diff --git a/cmd/zinject/zinject.c b/cmd/zinject/zinject.c index d6298f8f49a4..ee15ff8a4f11 100644 --- a/cmd/zinject/zinject.c +++ b/cmd/zinject/zinject.c @@ -1,1190 +1,1192 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2015 by Delphix. All rights reserved. * Copyright (c) 2017, Intel Corporation. */ /* * ZFS Fault Injector * * This userland component takes a set of options and uses libzpool to translate * from a user-visible object type and name to an internal representation. * There are two basic types of faults: device faults and data faults. * * * DEVICE FAULTS * * Errors can be injected into a particular vdev using the '-d' option. This * option takes a path or vdev GUID to uniquely identify the device within a * pool. There are four types of errors that can be injected, IO, ENXIO, * ECHILD, and EILSEQ. These can be controlled through the '-e' option and the * default is ENXIO. For EIO failures, any attempt to read data from the device * will return EIO, but a subsequent attempt to reopen the device will succeed. * For ENXIO failures, any attempt to read from the device will return EIO, but * any attempt to reopen the device will also return ENXIO. The EILSEQ failures * only apply to read operations (-T read) and will flip a bit after the device * has read the original data. * * For label faults, the -L option must be specified. This allows faults * to be injected into either the nvlist, uberblock, pad1, or pad2 region * of all the labels for the specified device. * * This form of the command looks like: * * zinject -d device [-e errno] [-L ] pool * * * DATA FAULTS * * We begin with a tuple of the form: * * * * type A string describing the type of data to target. Each type * implicitly describes how to interpret 'object'. Currently, * the following values are supported: * * data User data for a file * dnode Dnode for a file or directory * * The following MOS objects are special. Instead of injecting * errors on a particular object or blkid, we inject errors across * all objects of the given type. * * mos Any data in the MOS * mosdir object directory * config pool configuration * bpobj blkptr list * spacemap spacemap * metaslab metaslab * errlog persistent error log * * level Object level. Defaults to '0', not applicable to all types. If * a range is given, this corresponds to the indirect block * corresponding to the specific range. * * range A numerical range [start,end) within the object. Defaults to * the full size of the file. * * object A string describing the logical location of the object. For * files and directories (currently the only supported types), * this is the path of the object on disk. * * This is translated, via libzpool, into the following internal representation: * * * * These types should be self-explanatory. This tuple is then passed to the * kernel via a special ioctl() to initiate fault injection for the given * object. Note that 'type' is not strictly necessary for fault injection, but * is used when translating existing faults into a human-readable string. * * * The command itself takes one of the forms: * * zinject * zinject <-a | -u pool> * zinject -c * zinject [-q] <-t type> [-f freq] [-u] [-a] [-m] [-e errno] [-l level] * [-r range] * zinject [-f freq] [-a] [-m] [-u] -b objset:object:level:start:end pool * * With no arguments, the command prints all currently registered injection * handlers, with their numeric identifiers. * * The '-c' option will clear the given handler, or all handlers if 'all' is * specified. * * The '-e' option takes a string describing the errno to simulate. This must - * be one of 'io', 'checksum', or 'decrypt'. In most cases this will result - * in the same behavior, but RAID-Z will produce a different set of ereports - * for this situation. + * be one of 'io', 'checksum', 'decompress', or 'decrypt'. In most cases this + * will result in the same behavior, but RAID-Z will produce a different set of + * ereports for this situation. * * The '-a', '-u', and '-m' flags toggle internal flush behavior. If '-a' is * specified, then the ARC cache is flushed appropriately. If '-u' is * specified, then the underlying SPA is unloaded. Either of these flags can be * specified independently of any other handlers. The '-m' flag automatically * does an unmount and remount of the underlying dataset to aid in flushing the * cache. * * The '-f' flag controls the frequency of errors injected, expressed as a * real number percentage between 0.0001 and 100. The default is 100. * * The this form is responsible for actually injecting the handler into the * framework. It takes the arguments described above, translates them to the * internal tuple using libzpool, and then issues an ioctl() to register the * handler. * * The final form can target a specific bookmark, regardless of whether a * human-readable interface has been designed. It allows developers to specify * a particular block by number. */ #include #include #include #include #include #include #include #include #include #undef verify /* both libzfs.h and zfs_context.h want to define this */ #include "zinject.h" libzfs_handle_t *g_zfs; int zfs_fd; #define ECKSUM EBADE static const char *errtable[TYPE_INVAL] = { "data", "dnode", "mos", "mosdir", "metaslab", "config", "bpobj", "spacemap", "errlog", "uber", "nvlist", "pad1", "pad2" }; static err_type_t name_to_type(const char *arg) { int i; for (i = 0; i < TYPE_INVAL; i++) if (strcmp(errtable[i], arg) == 0) return (i); return (TYPE_INVAL); } static const char * type_to_name(uint64_t type) { switch (type) { case DMU_OT_OBJECT_DIRECTORY: return ("mosdir"); case DMU_OT_OBJECT_ARRAY: return ("metaslab"); case DMU_OT_PACKED_NVLIST: return ("config"); case DMU_OT_BPOBJ: return ("bpobj"); case DMU_OT_SPACE_MAP: return ("spacemap"); case DMU_OT_ERROR_LOG: return ("errlog"); default: return ("-"); } } /* * Print usage message. */ void usage(void) { (void) printf( "usage:\n" "\n" "\tzinject\n" "\n" "\t\tList all active injection records.\n" "\n" "\tzinject -c \n" "\n" "\t\tClear the particular record (if given a numeric ID), or\n" "\t\tall records if 'all' is specified.\n" "\n" "\tzinject -p pool\n" "\t\tInject a panic fault at the specified function. Only \n" "\t\tfunctions which call spa_vdev_config_exit(), or \n" "\t\tspa_vdev_exit() will trigger a panic.\n" "\n" "\tzinject -d device [-e errno] [-L ] [-F]\n" "\t\t[-T ] [-f frequency] pool\n\n" "\t\tInject a fault into a particular device or the device's\n" "\t\tlabel. Label injection can either be 'nvlist', 'uber',\n " "\t\t'pad1', or 'pad2'.\n" "\t\t'errno' can be 'nxio' (the default), 'io', 'dtl', or\n" "\t\t'corrupt' (bit flip).\n" "\t\t'frequency' is a value between 0.0001 and 100.0 that limits\n" "\t\tdevice error injection to a percentage of the IOs.\n" "\n" "\tzinject -d device -A -D pool\n" "\t\tPerform a specific action on a particular device.\n" "\n" "\tzinject -d device -D latency:lanes pool\n" "\n" "\t\tAdd an artificial delay to IO requests on a particular\n" "\t\tdevice, such that the requests take a minimum of 'latency'\n" "\t\tmilliseconds to complete. Each delay has an associated\n" "\t\tnumber of 'lanes' which defines the number of concurrent\n" "\t\tIO requests that can be processed.\n" "\n" "\t\tFor example, with a single lane delay of 10 ms (-D 10:1),\n" "\t\tthe device will only be able to service a single IO request\n" "\t\tat a time with each request taking 10 ms to complete. So,\n" "\t\tif only a single request is submitted every 10 ms, the\n" "\t\taverage latency will be 10 ms; but if more than one request\n" "\t\tis submitted every 10 ms, the average latency will be more\n" "\t\tthan 10 ms.\n" "\n" "\t\tSimilarly, if a delay of 10 ms is specified to have two\n" "\t\tlanes (-D 10:2), then the device will be able to service\n" "\t\ttwo requests at a time, each with a minimum latency of\n" "\t\t10 ms. So, if two requests are submitted every 10 ms, then\n" "\t\tthe average latency will be 10 ms; but if more than two\n" "\t\trequests are submitted every 10 ms, the average latency\n" "\t\twill be more than 10 ms.\n" "\n" "\t\tAlso note, these delays are additive. So two invocations\n" "\t\tof '-D 10:1', is roughly equivalent to a single invocation\n" "\t\tof '-D 10:2'. This also means, one can specify multiple\n" "\t\tlanes with differing target latencies. For example, an\n" "\t\tinvocation of '-D 10:1' followed by '-D 25:2' will\n" "\t\tcreate 3 lanes on the device; one lane with a latency\n" "\t\tof 10 ms and two lanes with a 25 ms latency.\n" "\n" "\tzinject -I [-s | -g ] pool\n" "\t\tCause the pool to stop writing blocks yet not\n" "\t\treport errors for a duration. Simulates buggy hardware\n" "\t\tthat fails to honor cache flush requests.\n" "\t\tDefault duration is 30 seconds. The machine is panicked\n" "\t\tat the end of the duration.\n" "\n" "\tzinject -b objset:object:level:blkid pool\n" "\n" "\t\tInject an error into pool 'pool' with the numeric bookmark\n" "\t\tspecified by the remaining tuple. Each number is in\n" "\t\thexadecimal, and only one block can be specified.\n" "\n" "\tzinject [-q] <-t type> [-e errno] [-l level] [-r range]\n" "\t [-a] [-m] [-u] [-f freq] \n" "\n" "\t\tInject an error into the object specified by the '-t' option\n" "\t\tand the object descriptor. The 'object' parameter is\n" "\t\tinterpreted depending on the '-t' option.\n" "\n" "\t\t-q\tQuiet mode. Only print out the handler number added.\n" "\t\t-e\tInject a specific error. Must be one of 'io',\n" - "\t\t\t'checksum', or decrypt. Default is 'io'.\n" + "\t\t\t'checksum', 'decompress', or decrypt. Default is 'io'.\n" "\t\t-l\tInject error at a particular block level. Default is " "0.\n" "\t\t-m\tAutomatically remount underlying filesystem.\n" "\t\t-r\tInject error over a particular logical range of an\n" "\t\t\tobject. Will be translated to the appropriate blkid\n" "\t\t\trange according to the object's properties.\n" "\t\t-a\tFlush the ARC cache. Can be specified without any\n" "\t\t\tassociated object.\n" "\t\t-u\tUnload the associated pool. Can be specified with only\n" "\t\t\ta pool object.\n" "\t\t-f\tOnly inject errors a fraction of the time. Expressed as\n" "\t\t\ta percentage between 0.0001 and 100.\n" "\n" "\t-t data\t\tInject an error into the plain file contents of a\n" "\t\t\tfile. The object must be specified as a complete path\n" "\t\t\tto a file on a ZFS filesystem.\n" "\n" "\t-t dnode\tInject an error into the metadnode in the block\n" "\t\t\tcorresponding to the dnode for a file or directory. The\n" "\t\t\t'-r' option is incompatible with this mode. The object\n" "\t\t\tis specified as a complete path to a file or directory\n" "\t\t\ton a ZFS filesystem.\n" "\n" "\t-t \tInject errors into the MOS for objects of the given\n" "\t\t\ttype. Valid types are: mos, mosdir, config, bpobj,\n" "\t\t\tspacemap, metaslab, errlog. The only valid is\n" "\t\t\tthe poolname.\n"); } static int iter_handlers(int (*func)(int, const char *, zinject_record_t *, void *), void *data) { zfs_cmd_t zc = {"\0"}; int ret; while (ioctl(zfs_fd, ZFS_IOC_INJECT_LIST_NEXT, &zc) == 0) if ((ret = func((int)zc.zc_guid, zc.zc_name, &zc.zc_inject_record, data)) != 0) return (ret); if (errno != ENOENT) { (void) fprintf(stderr, "Unable to list handlers: %s\n", strerror(errno)); return (-1); } return (0); } static int print_data_handler(int id, const char *pool, zinject_record_t *record, void *data) { int *count = data; if (record->zi_guid != 0 || record->zi_func[0] != '\0') return (0); if (*count == 0) { (void) printf("%3s %-15s %-6s %-6s %-8s %3s %-15s\n", "ID", "POOL", "OBJSET", "OBJECT", "TYPE", "LVL", "RANGE"); (void) printf("--- --------------- ------ " "------ -------- --- ---------------\n"); } *count += 1; (void) printf("%3d %-15s %-6llu %-6llu %-8s %3d ", id, pool, (u_longlong_t)record->zi_objset, (u_longlong_t)record->zi_object, type_to_name(record->zi_type), record->zi_level); if (record->zi_start == 0 && record->zi_end == -1ULL) (void) printf("all\n"); else (void) printf("[%llu, %llu]\n", (u_longlong_t)record->zi_start, (u_longlong_t)record->zi_end); return (0); } static int print_device_handler(int id, const char *pool, zinject_record_t *record, void *data) { int *count = data; if (record->zi_guid == 0 || record->zi_func[0] != '\0') return (0); if (record->zi_cmd == ZINJECT_DELAY_IO) return (0); if (*count == 0) { (void) printf("%3s %-15s %s\n", "ID", "POOL", "GUID"); (void) printf("--- --------------- ----------------\n"); } *count += 1; (void) printf("%3d %-15s %llx\n", id, pool, (u_longlong_t)record->zi_guid); return (0); } static int print_delay_handler(int id, const char *pool, zinject_record_t *record, void *data) { int *count = data; if (record->zi_guid == 0 || record->zi_func[0] != '\0') return (0); if (record->zi_cmd != ZINJECT_DELAY_IO) return (0); if (*count == 0) { (void) printf("%3s %-15s %-15s %-15s %s\n", "ID", "POOL", "DELAY (ms)", "LANES", "GUID"); (void) printf("--- --------------- --------------- " "--------------- ----------------\n"); } *count += 1; (void) printf("%3d %-15s %-15llu %-15llu %llx\n", id, pool, (u_longlong_t)NSEC2MSEC(record->zi_timer), (u_longlong_t)record->zi_nlanes, (u_longlong_t)record->zi_guid); return (0); } static int print_panic_handler(int id, const char *pool, zinject_record_t *record, void *data) { int *count = data; if (record->zi_func[0] == '\0') return (0); if (*count == 0) { (void) printf("%3s %-15s %s\n", "ID", "POOL", "FUNCTION"); (void) printf("--- --------------- ----------------\n"); } *count += 1; (void) printf("%3d %-15s %s\n", id, pool, record->zi_func); return (0); } /* * Print all registered error handlers. Returns the number of handlers * registered. */ static int print_all_handlers(void) { int count = 0, total = 0; (void) iter_handlers(print_device_handler, &count); if (count > 0) { total += count; (void) printf("\n"); count = 0; } (void) iter_handlers(print_delay_handler, &count); if (count > 0) { total += count; (void) printf("\n"); count = 0; } (void) iter_handlers(print_data_handler, &count); if (count > 0) { total += count; (void) printf("\n"); count = 0; } (void) iter_handlers(print_panic_handler, &count); return (count + total); } /* ARGSUSED */ static int cancel_one_handler(int id, const char *pool, zinject_record_t *record, void *data) { zfs_cmd_t zc = {"\0"}; zc.zc_guid = (uint64_t)id; if (ioctl(zfs_fd, ZFS_IOC_CLEAR_FAULT, &zc) != 0) { (void) fprintf(stderr, "failed to remove handler %d: %s\n", id, strerror(errno)); return (1); } return (0); } /* * Remove all fault injection handlers. */ static int cancel_all_handlers(void) { int ret = iter_handlers(cancel_one_handler, NULL); if (ret == 0) (void) printf("removed all registered handlers\n"); return (ret); } /* * Remove a specific fault injection handler. */ static int cancel_handler(int id) { zfs_cmd_t zc = {"\0"}; zc.zc_guid = (uint64_t)id; if (ioctl(zfs_fd, ZFS_IOC_CLEAR_FAULT, &zc) != 0) { (void) fprintf(stderr, "failed to remove handler %d: %s\n", id, strerror(errno)); return (1); } (void) printf("removed handler %d\n", id); return (0); } /* * Register a new fault injection handler. */ static int register_handler(const char *pool, int flags, zinject_record_t *record, int quiet) { zfs_cmd_t zc = {"\0"}; (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); zc.zc_inject_record = *record; zc.zc_guid = flags; if (ioctl(zfs_fd, ZFS_IOC_INJECT_FAULT, &zc) != 0) { (void) fprintf(stderr, "failed to add handler: %s\n", strerror(errno)); return (1); } if (flags & ZINJECT_NULL) return (0); if (quiet) { (void) printf("%llu\n", (u_longlong_t)zc.zc_guid); } else { (void) printf("Added handler %llu with the following " "properties:\n", (u_longlong_t)zc.zc_guid); (void) printf(" pool: %s\n", pool); if (record->zi_guid) { (void) printf(" vdev: %llx\n", (u_longlong_t)record->zi_guid); } else if (record->zi_func[0] != '\0') { (void) printf(" panic function: %s\n", record->zi_func); } else if (record->zi_duration > 0) { (void) printf(" time: %lld seconds\n", (u_longlong_t)record->zi_duration); } else if (record->zi_duration < 0) { (void) printf(" txgs: %lld \n", (u_longlong_t)-record->zi_duration); } else { (void) printf("objset: %llu\n", (u_longlong_t)record->zi_objset); (void) printf("object: %llu\n", (u_longlong_t)record->zi_object); (void) printf(" type: %llu\n", (u_longlong_t)record->zi_type); (void) printf(" level: %d\n", record->zi_level); if (record->zi_start == 0 && record->zi_end == -1ULL) (void) printf(" range: all\n"); else (void) printf(" range: [%llu, %llu)\n", (u_longlong_t)record->zi_start, (u_longlong_t)record->zi_end); } } return (0); } int perform_action(const char *pool, zinject_record_t *record, int cmd) { zfs_cmd_t zc = {"\0"}; ASSERT(cmd == VDEV_STATE_DEGRADED || cmd == VDEV_STATE_FAULTED); (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); zc.zc_guid = record->zi_guid; zc.zc_cookie = cmd; if (ioctl(zfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) return (0); return (1); } static int parse_delay(char *str, uint64_t *delay, uint64_t *nlanes) { unsigned long scan_delay; unsigned long scan_nlanes; if (sscanf(str, "%lu:%lu", &scan_delay, &scan_nlanes) != 2) return (1); /* * We explicitly disallow a delay of zero here, because we key * off this value being non-zero in translate_device(), to * determine if the fault is a ZINJECT_DELAY_IO fault or not. */ if (scan_delay == 0) return (1); /* * The units for the CLI delay parameter is milliseconds, but * the data passed to the kernel is interpreted as nanoseconds. * Thus we scale the milliseconds to nanoseconds here, and this * nanosecond value is used to pass the delay to the kernel. */ *delay = MSEC2NSEC(scan_delay); *nlanes = scan_nlanes; return (0); } static int parse_frequency(const char *str, uint32_t *percent) { double val; char *post; val = strtod(str, &post); if (post == NULL || *post != '\0') return (EINVAL); /* valid range is [0.0001, 100.0] */ val /= 100.0f; if (val < 0.000001f || val > 1.0f) return (ERANGE); /* convert to an integer for use by kernel */ *percent = ((uint32_t)(val * ZI_PERCENTAGE_MAX)); return (0); } int main(int argc, char **argv) { int c; char *range = NULL; char *cancel = NULL; char *end; char *raw = NULL; char *device = NULL; int level = 0; int quiet = 0; int error = 0; int domount = 0; int io_type = ZIO_TYPES; int action = VDEV_STATE_UNKNOWN; err_type_t type = TYPE_INVAL; err_type_t label = TYPE_INVAL; zinject_record_t record = { 0 }; char pool[MAXNAMELEN] = ""; char dataset[MAXNAMELEN] = ""; zfs_handle_t *zhp = NULL; int nowrites = 0; int dur_txg = 0; int dur_secs = 0; int ret; int flags = 0; if ((g_zfs = libzfs_init()) == NULL) { (void) fprintf(stderr, "%s", libzfs_error_init(errno)); return (1); } libzfs_print_on_error(g_zfs, B_TRUE); if ((zfs_fd = open(ZFS_DEV, O_RDWR)) < 0) { (void) fprintf(stderr, "failed to open ZFS device\n"); libzfs_fini(g_zfs); return (1); } if (argc == 1) { /* * No arguments. Print the available handlers. If there are no * available handlers, direct the user to '-h' for help * information. */ if (print_all_handlers() == 0) { (void) printf("No handlers registered.\n"); (void) printf("Run 'zinject -h' for usage " "information.\n"); } libzfs_fini(g_zfs); return (0); } while ((c = getopt(argc, argv, ":aA:b:d:D:f:Fg:qhIc:t:T:l:mr:s:e:uL:p:")) != -1) { switch (c) { case 'a': flags |= ZINJECT_FLUSH_ARC; break; case 'A': if (strcasecmp(optarg, "degrade") == 0) { action = VDEV_STATE_DEGRADED; } else if (strcasecmp(optarg, "fault") == 0) { action = VDEV_STATE_FAULTED; } else { (void) fprintf(stderr, "invalid action '%s': " "must be 'degrade' or 'fault'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'b': raw = optarg; break; case 'c': cancel = optarg; break; case 'd': device = optarg; break; case 'D': errno = 0; ret = parse_delay(optarg, &record.zi_timer, &record.zi_nlanes); if (ret != 0) { (void) fprintf(stderr, "invalid i/o delay " "value: '%s'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'e': if (strcasecmp(optarg, "io") == 0) { error = EIO; } else if (strcasecmp(optarg, "checksum") == 0) { error = ECKSUM; + } else if (strcasecmp(optarg, "decompress") == 0) { + error = EINVAL; } else if (strcasecmp(optarg, "decrypt") == 0) { error = EACCES; } else if (strcasecmp(optarg, "nxio") == 0) { error = ENXIO; } else if (strcasecmp(optarg, "dtl") == 0) { error = ECHILD; } else if (strcasecmp(optarg, "corrupt") == 0) { error = EILSEQ; } else { (void) fprintf(stderr, "invalid error type " "'%s': must be 'io', 'checksum' or " "'nxio'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'f': ret = parse_frequency(optarg, &record.zi_freq); if (ret != 0) { (void) fprintf(stderr, "%sfrequency value must " "be in the range [0.0001, 100.0]\n", ret == EINVAL ? "invalid value: " : ret == ERANGE ? "out of range: " : ""); libzfs_fini(g_zfs); return (1); } break; case 'F': record.zi_failfast = B_TRUE; break; case 'g': dur_txg = 1; record.zi_duration = (int)strtol(optarg, &end, 10); if (record.zi_duration <= 0 || *end != '\0') { (void) fprintf(stderr, "invalid duration '%s': " "must be a positive integer\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } /* store duration of txgs as its negative */ record.zi_duration *= -1; break; case 'h': usage(); libzfs_fini(g_zfs); return (0); case 'I': /* default duration, if one hasn't yet been defined */ nowrites = 1; if (dur_secs == 0 && dur_txg == 0) record.zi_duration = 30; break; case 'l': level = (int)strtol(optarg, &end, 10); if (*end != '\0') { (void) fprintf(stderr, "invalid level '%s': " "must be an integer\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'm': domount = 1; break; case 'p': (void) strlcpy(record.zi_func, optarg, sizeof (record.zi_func)); record.zi_cmd = ZINJECT_PANIC; break; case 'q': quiet = 1; break; case 'r': range = optarg; break; case 's': dur_secs = 1; record.zi_duration = (int)strtol(optarg, &end, 10); if (record.zi_duration <= 0 || *end != '\0') { (void) fprintf(stderr, "invalid duration '%s': " "must be a positive integer\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'T': if (strcasecmp(optarg, "read") == 0) { io_type = ZIO_TYPE_READ; } else if (strcasecmp(optarg, "write") == 0) { io_type = ZIO_TYPE_WRITE; } else if (strcasecmp(optarg, "free") == 0) { io_type = ZIO_TYPE_FREE; } else if (strcasecmp(optarg, "claim") == 0) { io_type = ZIO_TYPE_CLAIM; } else if (strcasecmp(optarg, "all") == 0) { io_type = ZIO_TYPES; } else { (void) fprintf(stderr, "invalid I/O type " "'%s': must be 'read', 'write', 'free', " "'claim' or 'all'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 't': if ((type = name_to_type(optarg)) == TYPE_INVAL && !MOS_TYPE(type)) { (void) fprintf(stderr, "invalid type '%s'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case 'u': flags |= ZINJECT_UNLOAD_SPA; break; case 'L': if ((label = name_to_type(optarg)) == TYPE_INVAL && !LABEL_TYPE(type)) { (void) fprintf(stderr, "invalid label type " "'%s'\n", optarg); usage(); libzfs_fini(g_zfs); return (1); } break; case ':': (void) fprintf(stderr, "option -%c requires an " "operand\n", optopt); usage(); libzfs_fini(g_zfs); return (1); case '?': (void) fprintf(stderr, "invalid option '%c'\n", optopt); usage(); libzfs_fini(g_zfs); return (2); } } argc -= optind; argv += optind; if (record.zi_duration != 0) record.zi_cmd = ZINJECT_IGNORED_WRITES; if (cancel != NULL) { /* * '-c' is invalid with any other options. */ if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED || record.zi_freq > 0) { (void) fprintf(stderr, "cancel (-c) incompatible with " "any other options\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc != 0) { (void) fprintf(stderr, "extraneous argument to '-c'\n"); usage(); libzfs_fini(g_zfs); return (2); } if (strcmp(cancel, "all") == 0) { return (cancel_all_handlers()); } else { int id = (int)strtol(cancel, &end, 10); if (*end != '\0') { (void) fprintf(stderr, "invalid handle id '%s':" " must be an integer or 'all'\n", cancel); usage(); libzfs_fini(g_zfs); return (1); } return (cancel_handler(id)); } } if (device != NULL) { /* * Device (-d) injection uses a completely different mechanism * for doing injection, so handle it separately here. */ if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED) { (void) fprintf(stderr, "device (-d) incompatible with " "data error injection\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc != 1) { (void) fprintf(stderr, "device (-d) injection requires " "a single pool name\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; if (error == ECKSUM) { (void) fprintf(stderr, "device error type must be " "'io', 'nxio' or 'corrupt'\n"); libzfs_fini(g_zfs); return (1); } if (error == EILSEQ && (record.zi_freq == 0 || io_type != ZIO_TYPE_READ)) { (void) fprintf(stderr, "device corrupt errors require " "io type read and a frequency value\n"); libzfs_fini(g_zfs); return (1); } record.zi_iotype = io_type; if (translate_device(pool, device, label, &record) != 0) { libzfs_fini(g_zfs); return (1); } if (!error) error = ENXIO; if (action != VDEV_STATE_UNKNOWN) return (perform_action(pool, &record, action)); } else if (raw != NULL) { if (range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED || record.zi_freq > 0) { (void) fprintf(stderr, "raw (-b) format with " "any other options\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc != 1) { (void) fprintf(stderr, "raw (-b) format expects a " "single pool name\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; if (error == ENXIO) { (void) fprintf(stderr, "data error type must be " "'checksum' or 'io'\n"); libzfs_fini(g_zfs); return (1); } record.zi_cmd = ZINJECT_DATA_FAULT; if (translate_raw(raw, &record) != 0) { libzfs_fini(g_zfs); return (1); } if (!error) error = EIO; } else if (record.zi_cmd == ZINJECT_PANIC) { if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || device != NULL || record.zi_freq > 0) { (void) fprintf(stderr, "panic (-p) incompatible with " "other options\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc < 1 || argc > 2) { (void) fprintf(stderr, "panic (-p) injection requires " "a single pool name and an optional id\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); if (argv[1] != NULL) record.zi_type = atoi(argv[1]); dataset[0] = '\0'; } else if (record.zi_cmd == ZINJECT_IGNORED_WRITES) { if (nowrites == 0) { (void) fprintf(stderr, "-s or -g meaningless " "without -I (ignore writes)\n"); usage(); libzfs_fini(g_zfs); return (2); } else if (dur_secs && dur_txg) { (void) fprintf(stderr, "choose a duration either " "in seconds (-s) or a number of txgs (-g) " "but not both\n"); usage(); libzfs_fini(g_zfs); return (2); } else if (argc != 1) { (void) fprintf(stderr, "ignore writes (-I) " "injection requires a single pool name\n"); usage(); libzfs_fini(g_zfs); return (2); } (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; } else if (type == TYPE_INVAL) { if (flags == 0) { (void) fprintf(stderr, "at least one of '-b', '-d', " "'-t', '-a', '-p', '-I' or '-u' " "must be specified\n"); usage(); libzfs_fini(g_zfs); return (2); } if (argc == 1 && (flags & ZINJECT_UNLOAD_SPA)) { (void) strlcpy(pool, argv[0], sizeof (pool)); dataset[0] = '\0'; } else if (argc != 0) { (void) fprintf(stderr, "extraneous argument for " "'-f'\n"); usage(); libzfs_fini(g_zfs); return (2); } flags |= ZINJECT_NULL; } else { if (argc != 1) { (void) fprintf(stderr, "missing object\n"); usage(); libzfs_fini(g_zfs); return (2); } if (error == ENXIO || error == EILSEQ) { (void) fprintf(stderr, "data error type must be " "'checksum' or 'io'\n"); libzfs_fini(g_zfs); return (1); } if (error == EACCES) { if (type != TYPE_DATA) { (void) fprintf(stderr, "decryption errors " "may only be injected for 'data' types\n"); libzfs_fini(g_zfs); return (1); } record.zi_cmd = ZINJECT_DECRYPT_FAULT; /* * Internally, ZFS actually uses ECKSUM for decryption * errors since EACCES is used to indicate the key was * not found. */ error = ECKSUM; } else { record.zi_cmd = ZINJECT_DATA_FAULT; } if (translate_record(type, argv[0], range, level, &record, pool, dataset) != 0) { libzfs_fini(g_zfs); return (1); } if (!error) error = EIO; } /* * If this is pool-wide metadata, unmount everything. The ioctl() will * unload the pool, so that we trigger spa-wide reopen of metadata next * time we access the pool. */ if (dataset[0] != '\0' && domount) { if ((zhp = zfs_open(g_zfs, dataset, ZFS_TYPE_DATASET)) == NULL) { libzfs_fini(g_zfs); return (1); } if (zfs_unmount(zhp, NULL, 0) != 0) { libzfs_fini(g_zfs); return (1); } } record.zi_error = error; ret = register_handler(pool, flags, &record, quiet); if (dataset[0] != '\0' && domount) ret = (zfs_mount(zhp, NULL, 0) != 0); libzfs_fini(g_zfs); return (ret); } diff --git a/cmd/ztest/ztest.c b/cmd/ztest/ztest.c index 8fb6200d55e6..71d5ed6469b2 100644 --- a/cmd/ztest/ztest.c +++ b/cmd/ztest/ztest.c @@ -1,7583 +1,7581 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2017 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2013 Steven Hartland. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright 2017 Joyent, Inc. */ /* * The objective of this program is to provide a DMU/ZAP/SPA stress test * that runs entirely in userland, is easy to use, and easy to extend. * * The overall design of the ztest program is as follows: * * (1) For each major functional area (e.g. adding vdevs to a pool, * creating and destroying datasets, reading and writing objects, etc) * we have a simple routine to test that functionality. These * individual routines do not have to do anything "stressful". * * (2) We turn these simple functionality tests into a stress test by * running them all in parallel, with as many threads as desired, * and spread across as many datasets, objects, and vdevs as desired. * * (3) While all this is happening, we inject faults into the pool to * verify that self-healing data really works. * * (4) Every time we open a dataset, we change its checksum and compression * functions. Thus even individual objects vary from block to block * in which checksum they use and whether they're compressed. * * (5) To verify that we never lose on-disk consistency after a crash, * we run the entire test in a child of the main process. * At random times, the child self-immolates with a SIGKILL. * This is the software equivalent of pulling the power cord. * The parent then runs the test again, using the existing * storage pool, as many times as desired. If backwards compatibility * testing is enabled ztest will sometimes run the "older" version * of ztest after a SIGKILL. * * (6) To verify that we don't have future leaks or temporal incursions, * many of the functional tests record the transaction group number * as part of their data. When reading old data, they verify that * the transaction group number is less than the current, open txg. * If you add a new test, please do this if applicable. * * (7) Threads are created with a reduced stack size, for sanity checking. * Therefore, it's important not to allocate huge buffers on the stack. * * When run with no arguments, ztest runs for about five minutes and * produces no output if successful. To get a little bit of information, * specify -V. To get more information, specify -VV, and so on. * * To turn this into an overnight stress test, use -T to specify run time. * * You can ask more more vdevs [-v], datasets [-d], or threads [-t] * to increase the pool capacity, fanout, and overall stress level. * * Use the -k option to set the desired frequency of kills. * * When ztest invokes itself it passes all relevant information through a * temporary file which is mmap-ed in the child process. This allows shared * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always * stored at offset 0 of this file and contains information on the size and * number of shared structures in the file. The information stored in this file * must remain backwards compatible with older versions of ztest so that * ztest can invoke them during backwards compatibility testing (-B). */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef __GLIBC__ #include /* for backtrace() */ #endif static int ztest_fd_data = -1; static int ztest_fd_rand = -1; typedef struct ztest_shared_hdr { uint64_t zh_hdr_size; uint64_t zh_opts_size; uint64_t zh_size; uint64_t zh_stats_size; uint64_t zh_stats_count; uint64_t zh_ds_size; uint64_t zh_ds_count; } ztest_shared_hdr_t; static ztest_shared_hdr_t *ztest_shared_hdr; typedef struct ztest_shared_opts { char zo_pool[ZFS_MAX_DATASET_NAME_LEN]; char zo_dir[ZFS_MAX_DATASET_NAME_LEN]; char zo_alt_ztest[MAXNAMELEN]; char zo_alt_libpath[MAXNAMELEN]; uint64_t zo_vdevs; uint64_t zo_vdevtime; size_t zo_vdev_size; int zo_ashift; int zo_mirrors; int zo_raidz; int zo_raidz_parity; int zo_datasets; int zo_threads; uint64_t zo_passtime; uint64_t zo_killrate; int zo_verbose; int zo_init; uint64_t zo_time; uint64_t zo_maxloops; uint64_t zo_metaslab_force_ganging; int zo_mmp_test; } ztest_shared_opts_t; static const ztest_shared_opts_t ztest_opts_defaults = { .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, .zo_dir = { '/', 't', 'm', 'p', '\0' }, .zo_alt_ztest = { '\0' }, .zo_alt_libpath = { '\0' }, .zo_vdevs = 5, .zo_ashift = SPA_MINBLOCKSHIFT, .zo_mirrors = 2, .zo_raidz = 4, .zo_raidz_parity = 1, .zo_vdev_size = SPA_MINDEVSIZE * 4, /* 256m default size */ .zo_datasets = 7, .zo_threads = 23, .zo_passtime = 60, /* 60 seconds */ .zo_killrate = 70, /* 70% kill rate */ .zo_verbose = 0, .zo_mmp_test = 0, .zo_init = 1, .zo_time = 300, /* 5 minutes */ .zo_maxloops = 50, /* max loops during spa_freeze() */ - .zo_metaslab_force_ganging = 32 << 10 + .zo_metaslab_force_ganging = 32 << 10, }; extern uint64_t metaslab_force_ganging; extern uint64_t metaslab_df_alloc_threshold; extern unsigned long zfs_deadman_synctime_ms; extern int metaslab_preload_limit; extern boolean_t zfs_compressed_arc_enabled; extern int zfs_abd_scatter_enabled; extern int dmu_object_alloc_chunk_shift; extern boolean_t zfs_force_some_double_word_sm_entries; +extern unsigned long zio_decompress_fail_fraction; static ztest_shared_opts_t *ztest_shared_opts; static ztest_shared_opts_t ztest_opts; static char *ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345"; typedef struct ztest_shared_ds { uint64_t zd_seq; } ztest_shared_ds_t; static ztest_shared_ds_t *ztest_shared_ds; #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) #define BT_MAGIC 0x123456789abcdefULL #define MAXFAULTS(zs) \ (MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) enum ztest_io_type { ZTEST_IO_WRITE_TAG, ZTEST_IO_WRITE_PATTERN, ZTEST_IO_WRITE_ZEROES, ZTEST_IO_TRUNCATE, ZTEST_IO_SETATTR, ZTEST_IO_REWRITE, ZTEST_IO_TYPES }; typedef struct ztest_block_tag { uint64_t bt_magic; uint64_t bt_objset; uint64_t bt_object; uint64_t bt_dnodesize; uint64_t bt_offset; uint64_t bt_gen; uint64_t bt_txg; uint64_t bt_crtxg; } ztest_block_tag_t; typedef struct bufwad { uint64_t bw_index; uint64_t bw_txg; uint64_t bw_data; } bufwad_t; typedef struct rll { void *rll_writer; int rll_readers; kmutex_t rll_lock; kcondvar_t rll_cv; } rll_t; typedef struct zll { list_t z_list; kmutex_t z_lock; } zll_t; #define ZTEST_RANGE_LOCKS 64 #define ZTEST_OBJECT_LOCKS 64 /* * Object descriptor. Used as a template for object lookup/create/remove. */ typedef struct ztest_od { uint64_t od_dir; uint64_t od_object; dmu_object_type_t od_type; dmu_object_type_t od_crtype; uint64_t od_blocksize; uint64_t od_crblocksize; uint64_t od_crdnodesize; uint64_t od_gen; uint64_t od_crgen; char od_name[ZFS_MAX_DATASET_NAME_LEN]; } ztest_od_t; /* * Per-dataset state. */ typedef struct ztest_ds { ztest_shared_ds_t *zd_shared; objset_t *zd_os; pthread_rwlock_t zd_zilog_lock; zilog_t *zd_zilog; ztest_od_t *zd_od; /* debugging aid */ char zd_name[ZFS_MAX_DATASET_NAME_LEN]; kmutex_t zd_dirobj_lock; rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; zll_t zd_range_lock[ZTEST_RANGE_LOCKS]; } ztest_ds_t; /* * Per-iteration state. */ typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); typedef struct ztest_info { ztest_func_t *zi_func; /* test function */ uint64_t zi_iters; /* iterations per execution */ uint64_t *zi_interval; /* execute every seconds */ const char *zi_funcname; /* name of test function */ } ztest_info_t; typedef struct ztest_shared_callstate { uint64_t zc_count; /* per-pass count */ uint64_t zc_time; /* per-pass time */ uint64_t zc_next; /* next time to call this function */ } ztest_shared_callstate_t; static ztest_shared_callstate_t *ztest_shared_callstate; #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) ztest_func_t ztest_dmu_read_write; ztest_func_t ztest_dmu_write_parallel; ztest_func_t ztest_dmu_object_alloc_free; ztest_func_t ztest_dmu_object_next_chunk; ztest_func_t ztest_dmu_commit_callbacks; ztest_func_t ztest_zap; ztest_func_t ztest_zap_parallel; ztest_func_t ztest_zil_commit; ztest_func_t ztest_zil_remount; ztest_func_t ztest_dmu_read_write_zcopy; ztest_func_t ztest_dmu_objset_create_destroy; ztest_func_t ztest_dmu_prealloc; ztest_func_t ztest_fzap; ztest_func_t ztest_dmu_snapshot_create_destroy; ztest_func_t ztest_dsl_prop_get_set; ztest_func_t ztest_spa_prop_get_set; ztest_func_t ztest_spa_create_destroy; ztest_func_t ztest_fault_inject; ztest_func_t ztest_ddt_repair; ztest_func_t ztest_dmu_snapshot_hold; ztest_func_t ztest_mmp_enable_disable; ztest_func_t ztest_spa_rename; ztest_func_t ztest_scrub; ztest_func_t ztest_dsl_dataset_promote_busy; ztest_func_t ztest_vdev_attach_detach; ztest_func_t ztest_vdev_LUN_growth; ztest_func_t ztest_vdev_add_remove; ztest_func_t ztest_vdev_aux_add_remove; ztest_func_t ztest_split_pool; ztest_func_t ztest_reguid; ztest_func_t ztest_spa_upgrade; ztest_func_t ztest_device_removal; ztest_func_t ztest_remap_blocks; ztest_func_t ztest_spa_checkpoint_create_discard; ztest_func_t ztest_fletcher; ztest_func_t ztest_fletcher_incr; ztest_func_t ztest_verify_dnode_bt; uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ #define ZTI_INIT(func, iters, interval) \ { .zi_func = (func), \ .zi_iters = (iters), \ .zi_interval = (interval), \ .zi_funcname = # func } ztest_info_t ztest_info[] = { ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes), ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), ZTI_INIT(ztest_zap, 30, &zopt_always), ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), ZTI_INIT(ztest_split_pool, 1, &zopt_always), ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant), ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes), ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often), ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often), ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often), ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes), #if 0 ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes), #endif ZTI_INIT(ztest_fzap, 1, &zopt_sometimes), ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes), ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes), ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes), ZTI_INIT(ztest_ddt_repair, 1, &zopt_sometimes), ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes), ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes), ZTI_INIT(ztest_reguid, 1, &zopt_rarely), ZTI_INIT(ztest_spa_rename, 1, &zopt_rarely), ZTI_INIT(ztest_scrub, 1, &zopt_rarely), ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely), ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely), ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes), ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely), ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime), ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime), ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes), ZTI_INIT(ztest_remap_blocks, 1, &zopt_sometimes), ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely), ZTI_INIT(ztest_fletcher, 1, &zopt_rarely), ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely), ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes), }; #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) /* * The following struct is used to hold a list of uncalled commit callbacks. * The callbacks are ordered by txg number. */ typedef struct ztest_cb_list { kmutex_t zcl_callbacks_lock; list_t zcl_callbacks; } ztest_cb_list_t; /* * Stuff we need to share writably between parent and child. */ typedef struct ztest_shared { boolean_t zs_do_init; hrtime_t zs_proc_start; hrtime_t zs_proc_stop; hrtime_t zs_thread_start; hrtime_t zs_thread_stop; hrtime_t zs_thread_kill; uint64_t zs_enospc_count; uint64_t zs_vdev_next_leaf; uint64_t zs_vdev_aux; uint64_t zs_alloc; uint64_t zs_space; uint64_t zs_splits; uint64_t zs_mirrors; uint64_t zs_metaslab_sz; uint64_t zs_metaslab_df_alloc_threshold; uint64_t zs_guid; } ztest_shared_t; #define ID_PARALLEL -1ULL static char ztest_dev_template[] = "%s/%s.%llua"; static char ztest_aux_template[] = "%s/%s.%s.%llu"; ztest_shared_t *ztest_shared; static spa_t *ztest_spa = NULL; static ztest_ds_t *ztest_ds; static kmutex_t ztest_vdev_lock; static boolean_t ztest_device_removal_active = B_FALSE; static kmutex_t ztest_checkpoint_lock; /* * The ztest_name_lock protects the pool and dataset namespace used by * the individual tests. To modify the namespace, consumers must grab * this lock as writer. Grabbing the lock as reader will ensure that the * namespace does not change while the lock is held. */ static pthread_rwlock_t ztest_name_lock; static boolean_t ztest_dump_core = B_TRUE; static boolean_t ztest_dump_debug_buffer = B_FALSE; static boolean_t ztest_exiting; /* Global commit callback list */ static ztest_cb_list_t zcl; /* Commit cb delay */ static uint64_t zc_min_txg_delay = UINT64_MAX; static int zc_cb_counter = 0; /* * Minimum number of commit callbacks that need to be registered for us to check * whether the minimum txg delay is acceptable. */ #define ZTEST_COMMIT_CB_MIN_REG 100 /* * If a number of txgs equal to this threshold have been created after a commit * callback has been registered but not called, then we assume there is an * implementation bug. */ #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) -extern uint64_t metaslab_gang_bang; -extern uint64_t metaslab_df_alloc_threshold; - enum ztest_object { ZTEST_META_DNODE = 0, ZTEST_DIROBJ, ZTEST_OBJECTS }; static void usage(boolean_t) __NORETURN; /* * These libumem hooks provide a reasonable set of defaults for the allocator's * debugging facilities. */ const char * _umem_debug_init(void) { return ("default,verbose"); /* $UMEM_DEBUG setting */ } const char * _umem_logging_init(void) { return ("fail,contents"); /* $UMEM_LOGGING setting */ } static void dump_debug_buffer(void) { if (!ztest_dump_debug_buffer) return; (void) printf("\n"); zfs_dbgmsg_print("ztest"); } #define BACKTRACE_SZ 100 static void sig_handler(int signo) { struct sigaction action; #ifdef __GLIBC__ /* backtrace() is a GNU extension */ int nptrs; void *buffer[BACKTRACE_SZ]; nptrs = backtrace(buffer, BACKTRACE_SZ); backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO); #endif dump_debug_buffer(); /* * Restore default action and re-raise signal so SIGSEGV and * SIGABRT can trigger a core dump. */ action.sa_handler = SIG_DFL; sigemptyset(&action.sa_mask); action.sa_flags = 0; (void) sigaction(signo, &action, NULL); raise(signo); } #define FATAL_MSG_SZ 1024 char *fatal_msg; static void fatal(int do_perror, char *message, ...) { va_list args; int save_errno = errno; char *buf; (void) fflush(stdout); buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); va_start(args, message); (void) sprintf(buf, "ztest: "); /* LINTED */ (void) vsprintf(buf + strlen(buf), message, args); va_end(args); if (do_perror) { (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), ": %s", strerror(save_errno)); } (void) fprintf(stderr, "%s\n", buf); fatal_msg = buf; /* to ease debugging */ dump_debug_buffer(); if (ztest_dump_core) abort(); exit(3); } static int str2shift(const char *buf) { const char *ends = "BKMGTPEZ"; int i; if (buf[0] == '\0') return (0); for (i = 0; i < strlen(ends); i++) { if (toupper(buf[0]) == ends[i]) break; } if (i == strlen(ends)) { (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); usage(B_FALSE); } if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { return (10*i); } (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); usage(B_FALSE); /* NOTREACHED */ } static uint64_t nicenumtoull(const char *buf) { char *end; uint64_t val; val = strtoull(buf, &end, 0); if (end == buf) { (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); usage(B_FALSE); } else if (end[0] == '.') { double fval = strtod(buf, &end); fval *= pow(2, str2shift(end)); if (fval > UINT64_MAX) { (void) fprintf(stderr, "ztest: value too large: %s\n", buf); usage(B_FALSE); } val = (uint64_t)fval; } else { int shift = str2shift(end); if (shift >= 64 || (val << shift) >> shift != val) { (void) fprintf(stderr, "ztest: value too large: %s\n", buf); usage(B_FALSE); } val <<= shift; } return (val); } static void usage(boolean_t requested) { const ztest_shared_opts_t *zo = &ztest_opts_defaults; char nice_vdev_size[NN_NUMBUF_SZ]; char nice_force_ganging[NN_NUMBUF_SZ]; FILE *fp = requested ? stdout : stderr; nicenum(zo->zo_vdev_size, nice_vdev_size, sizeof (nice_vdev_size)); nicenum(zo->zo_metaslab_force_ganging, nice_force_ganging, sizeof (nice_force_ganging)); (void) fprintf(fp, "Usage: %s\n" "\t[-v vdevs (default: %llu)]\n" "\t[-s size_of_each_vdev (default: %s)]\n" "\t[-a alignment_shift (default: %d)] use 0 for random\n" "\t[-m mirror_copies (default: %d)]\n" "\t[-r raidz_disks (default: %d)]\n" "\t[-R raidz_parity (default: %d)]\n" "\t[-d datasets (default: %d)]\n" "\t[-t threads (default: %d)]\n" "\t[-g gang_block_threshold (default: %s)]\n" "\t[-i init_count (default: %d)] initialize pool i times\n" "\t[-k kill_percentage (default: %llu%%)]\n" "\t[-p pool_name (default: %s)]\n" "\t[-f dir (default: %s)] file directory for vdev files\n" "\t[-M] Multi-host simulate pool imported on remote host\n" "\t[-V] verbose (use multiple times for ever more blather)\n" "\t[-E] use existing pool instead of creating new one\n" "\t[-T time (default: %llu sec)] total run time\n" "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" "\t[-P passtime (default: %llu sec)] time per pass\n" "\t[-B alt_ztest (default: )] alternate ztest path\n" "\t[-o variable=value] ... set global variable to an unsigned\n" "\t 32-bit integer value\n" "\t[-G dump zfs_dbgmsg buffer before exiting due to an error\n" "\t[-h] (print help)\n" "", zo->zo_pool, (u_longlong_t)zo->zo_vdevs, /* -v */ nice_vdev_size, /* -s */ zo->zo_ashift, /* -a */ zo->zo_mirrors, /* -m */ zo->zo_raidz, /* -r */ zo->zo_raidz_parity, /* -R */ zo->zo_datasets, /* -d */ zo->zo_threads, /* -t */ nice_force_ganging, /* -g */ zo->zo_init, /* -i */ (u_longlong_t)zo->zo_killrate, /* -k */ zo->zo_pool, /* -p */ zo->zo_dir, /* -f */ (u_longlong_t)zo->zo_time, /* -T */ (u_longlong_t)zo->zo_maxloops, /* -F */ (u_longlong_t)zo->zo_passtime); exit(requested ? 0 : 1); } static void process_options(int argc, char **argv) { char *path; ztest_shared_opts_t *zo = &ztest_opts; int opt; uint64_t value; char altdir[MAXNAMELEN] = { 0 }; bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); while ((opt = getopt(argc, argv, "v:s:a:m:r:R:d:t:g:i:k:p:f:MVET:P:hF:B:o:G")) != EOF) { value = 0; switch (opt) { case 'v': case 's': case 'a': case 'm': case 'r': case 'R': case 'd': case 't': case 'g': case 'i': case 'k': case 'T': case 'P': case 'F': value = nicenumtoull(optarg); } switch (opt) { case 'v': zo->zo_vdevs = value; break; case 's': zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); break; case 'a': zo->zo_ashift = value; break; case 'm': zo->zo_mirrors = value; break; case 'r': zo->zo_raidz = MAX(1, value); break; case 'R': zo->zo_raidz_parity = MIN(MAX(value, 1), 3); break; case 'd': zo->zo_datasets = MAX(1, value); break; case 't': zo->zo_threads = MAX(1, value); break; case 'g': zo->zo_metaslab_force_ganging = MAX(SPA_MINBLOCKSIZE << 1, value); break; case 'i': zo->zo_init = value; break; case 'k': zo->zo_killrate = value; break; case 'p': (void) strlcpy(zo->zo_pool, optarg, sizeof (zo->zo_pool)); break; case 'f': path = realpath(optarg, NULL); if (path == NULL) { (void) fprintf(stderr, "error: %s: %s\n", optarg, strerror(errno)); usage(B_FALSE); } else { (void) strlcpy(zo->zo_dir, path, sizeof (zo->zo_dir)); free(path); } break; case 'M': zo->zo_mmp_test = 1; break; case 'V': zo->zo_verbose++; break; case 'E': zo->zo_init = 0; break; case 'T': zo->zo_time = value; break; case 'P': zo->zo_passtime = MAX(1, value); break; case 'F': zo->zo_maxloops = MAX(1, value); break; case 'B': (void) strlcpy(altdir, optarg, sizeof (altdir)); break; case 'o': if (set_global_var(optarg) != 0) usage(B_FALSE); break; case 'G': ztest_dump_debug_buffer = B_TRUE; break; case 'h': usage(B_TRUE); break; case '?': default: usage(B_FALSE); break; } } zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); zo->zo_vdevtime = (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : UINT64_MAX >> 2); if (strlen(altdir) > 0) { char *cmd; char *realaltdir; char *bin; char *ztest; char *isa; int isalen; cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); VERIFY(NULL != realpath(getexecname(), cmd)); if (0 != access(altdir, F_OK)) { ztest_dump_core = B_FALSE; fatal(B_TRUE, "invalid alternate ztest path: %s", altdir); } VERIFY(NULL != realpath(altdir, realaltdir)); /* * 'cmd' should be of the form "/usr/bin//ztest". * We want to extract to determine if we should use * 32 or 64 bit binaries. */ bin = strstr(cmd, "/usr/bin/"); ztest = strstr(bin, "/ztest"); isa = bin + 9; isalen = ztest - isa; (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), "%s/usr/lib/%.*s", realaltdir, isalen, isa); if (0 != access(zo->zo_alt_ztest, X_OK)) { ztest_dump_core = B_FALSE; fatal(B_TRUE, "invalid alternate ztest: %s", zo->zo_alt_ztest); } else if (0 != access(zo->zo_alt_libpath, X_OK)) { ztest_dump_core = B_FALSE; fatal(B_TRUE, "invalid alternate lib directory %s", zo->zo_alt_libpath); } umem_free(cmd, MAXPATHLEN); umem_free(realaltdir, MAXPATHLEN); } } static void ztest_kill(ztest_shared_t *zs) { zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); /* * Before we kill off ztest, make sure that the config is updated. * See comment above spa_write_cachefile(). */ mutex_enter(&spa_namespace_lock); spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE); mutex_exit(&spa_namespace_lock); (void) kill(getpid(), SIGKILL); } static uint64_t ztest_random(uint64_t range) { uint64_t r; ASSERT3S(ztest_fd_rand, >=, 0); if (range == 0) return (0); if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) fatal(1, "short read from /dev/urandom"); return (r % range); } /* ARGSUSED */ static void ztest_record_enospc(const char *s) { ztest_shared->zs_enospc_count++; } static uint64_t ztest_get_ashift(void) { if (ztest_opts.zo_ashift == 0) return (SPA_MINBLOCKSHIFT + ztest_random(5)); return (ztest_opts.zo_ashift); } static nvlist_t * make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) { char *pathbuf; uint64_t vdev; nvlist_t *file; pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); if (ashift == 0) ashift = ztest_get_ashift(); if (path == NULL) { path = pathbuf; if (aux != NULL) { vdev = ztest_shared->zs_vdev_aux; (void) snprintf(path, MAXPATHLEN, ztest_aux_template, ztest_opts.zo_dir, pool == NULL ? ztest_opts.zo_pool : pool, aux, vdev); } else { vdev = ztest_shared->zs_vdev_next_leaf++; (void) snprintf(path, MAXPATHLEN, ztest_dev_template, ztest_opts.zo_dir, pool == NULL ? ztest_opts.zo_pool : pool, vdev); } } if (size != 0) { int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); if (fd == -1) fatal(1, "can't open %s", path); if (ftruncate(fd, size) != 0) fatal(1, "can't ftruncate %s", path); (void) close(fd); } VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); umem_free(pathbuf, MAXPATHLEN); return (file); } static nvlist_t * make_vdev_raidz(char *path, char *aux, char *pool, size_t size, uint64_t ashift, int r) { nvlist_t *raidz, **child; int c; if (r < 2) return (make_vdev_file(path, aux, pool, size, ashift)); child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); for (c = 0; c < r; c++) child[c] = make_vdev_file(path, aux, pool, size, ashift); VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, VDEV_TYPE_RAIDZ) == 0); VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, ztest_opts.zo_raidz_parity) == 0); VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, child, r) == 0); for (c = 0; c < r; c++) nvlist_free(child[c]); umem_free(child, r * sizeof (nvlist_t *)); return (raidz); } static nvlist_t * make_vdev_mirror(char *path, char *aux, char *pool, size_t size, uint64_t ashift, int r, int m) { nvlist_t *mirror, **child; int c; if (m < 1) return (make_vdev_raidz(path, aux, pool, size, ashift, r)); child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); for (c = 0; c < m; c++) child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r); VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, VDEV_TYPE_MIRROR) == 0); VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, child, m) == 0); for (c = 0; c < m; c++) nvlist_free(child[c]); umem_free(child, m * sizeof (nvlist_t *)); return (mirror); } static nvlist_t * make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, int log, int r, int m, int t) { nvlist_t *root, **child; int c; ASSERT(t > 0); child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); for (c = 0; c < t; c++) { child[c] = make_vdev_mirror(path, aux, pool, size, ashift, r, m); VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, log) == 0); } VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, child, t) == 0); for (c = 0; c < t; c++) nvlist_free(child[c]); umem_free(child, t * sizeof (nvlist_t *)); return (root); } /* * Find a random spa version. Returns back a random spa version in the * range [initial_version, SPA_VERSION_FEATURES]. */ static uint64_t ztest_random_spa_version(uint64_t initial_version) { uint64_t version = initial_version; if (version <= SPA_VERSION_BEFORE_FEATURES) { version = version + ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); } if (version > SPA_VERSION_BEFORE_FEATURES) version = SPA_VERSION_FEATURES; ASSERT(SPA_VERSION_IS_SUPPORTED(version)); return (version); } static int ztest_random_blocksize(void) { /* * Choose a block size >= the ashift. * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks. */ int maxbs = SPA_OLD_MAXBLOCKSHIFT; if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE) maxbs = 20; uint64_t block_shift = ztest_random(maxbs - ztest_spa->spa_max_ashift + 1); return (1 << (SPA_MINBLOCKSHIFT + block_shift)); } static int ztest_random_dnodesize(void) { int slots; int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT; if (max_slots == DNODE_MIN_SLOTS) return (DNODE_MIN_SIZE); /* * Weight the random distribution more heavily toward smaller * dnode sizes since that is more likely to reflect real-world * usage. */ ASSERT3U(max_slots, >, 4); switch (ztest_random(10)) { case 0: slots = 5 + ztest_random(max_slots - 4); break; case 1 ... 4: slots = 2 + ztest_random(3); break; default: slots = 1; break; } return (slots << DNODE_SHIFT); } static int ztest_random_ibshift(void) { return (DN_MIN_INDBLKSHIFT + ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); } static uint64_t ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) { uint64_t top; vdev_t *rvd = spa->spa_root_vdev; vdev_t *tvd; ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); do { top = ztest_random(rvd->vdev_children); tvd = rvd->vdev_child[top]; } while (!vdev_is_concrete(tvd) || (tvd->vdev_islog && !log_ok) || tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); return (top); } static uint64_t ztest_random_dsl_prop(zfs_prop_t prop) { uint64_t value; do { value = zfs_prop_random_value(prop, ztest_random(-1ULL)); } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); return (value); } static int ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, boolean_t inherit) { const char *propname = zfs_prop_to_name(prop); const char *valname; char *setpoint; uint64_t curval; int error; error = dsl_prop_set_int(osname, propname, (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); if (error == ENOSPC) { ztest_record_enospc(FTAG); return (error); } ASSERT0(error); setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); if (ztest_opts.zo_verbose >= 6) { int err; err = zfs_prop_index_to_string(prop, curval, &valname); if (err) (void) printf("%s %s = %llu at '%s'\n", osname, propname, (unsigned long long)curval, setpoint); else (void) printf("%s %s = %s at '%s'\n", osname, propname, valname, setpoint); } umem_free(setpoint, MAXPATHLEN); return (error); } static int ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) { spa_t *spa = ztest_spa; nvlist_t *props = NULL; int error; VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); error = spa_prop_set(spa, props); nvlist_free(props); if (error == ENOSPC) { ztest_record_enospc(FTAG); return (error); } ASSERT0(error); return (error); } static int ztest_dmu_objset_own(const char *name, dmu_objset_type_t type, boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) { int err; err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); if (decrypt && err == EACCES) { char ddname[ZFS_MAX_DATASET_NAME_LEN]; dsl_crypto_params_t *dcp; nvlist_t *crypto_args = fnvlist_alloc(); char *cp = NULL; /* spa_keystore_load_wkey() expects a dsl dir name */ strcpy(ddname, name); cp = strchr(ddname, '@'); if (cp != NULL) *cp = '\0'; fnvlist_add_uint8_array(crypto_args, "wkeydata", (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL, crypto_args, &dcp)); err = spa_keystore_load_wkey(ddname, dcp, B_FALSE); dsl_crypto_params_free(dcp, B_FALSE); fnvlist_free(crypto_args); if (err != 0) return (err); err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); } return (err); } /* * Object and range lock mechanics */ typedef struct { list_node_t z_lnode; refcount_t z_refcnt; uint64_t z_object; zfs_rlock_t z_range_lock; } ztest_znode_t; typedef struct { rl_t *z_rl; ztest_znode_t *z_ztznode; } ztest_zrl_t; static ztest_znode_t * ztest_znode_init(uint64_t object) { ztest_znode_t *zp = umem_alloc(sizeof (*zp), UMEM_NOFAIL); list_link_init(&zp->z_lnode); refcount_create(&zp->z_refcnt); zp->z_object = object; zfs_rlock_init(&zp->z_range_lock); return (zp); } static void ztest_znode_fini(ztest_znode_t *zp) { ASSERT(refcount_is_zero(&zp->z_refcnt)); zfs_rlock_destroy(&zp->z_range_lock); zp->z_object = 0; refcount_destroy(&zp->z_refcnt); list_link_init(&zp->z_lnode); umem_free(zp, sizeof (*zp)); } static void ztest_zll_init(zll_t *zll) { mutex_init(&zll->z_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zll->z_list, sizeof (ztest_znode_t), offsetof(ztest_znode_t, z_lnode)); } static void ztest_zll_destroy(zll_t *zll) { list_destroy(&zll->z_list); mutex_destroy(&zll->z_lock); } #define RL_TAG "range_lock" static ztest_znode_t * ztest_znode_get(ztest_ds_t *zd, uint64_t object) { zll_t *zll = &zd->zd_range_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; ztest_znode_t *zp = NULL; mutex_enter(&zll->z_lock); for (zp = list_head(&zll->z_list); (zp); zp = list_next(&zll->z_list, zp)) { if (zp->z_object == object) { refcount_add(&zp->z_refcnt, RL_TAG); break; } } if (zp == NULL) { zp = ztest_znode_init(object); refcount_add(&zp->z_refcnt, RL_TAG); list_insert_head(&zll->z_list, zp); } mutex_exit(&zll->z_lock); return (zp); } static void ztest_znode_put(ztest_ds_t *zd, ztest_znode_t *zp) { zll_t *zll = NULL; ASSERT3U(zp->z_object, !=, 0); zll = &zd->zd_range_lock[zp->z_object & (ZTEST_OBJECT_LOCKS - 1)]; mutex_enter(&zll->z_lock); refcount_remove(&zp->z_refcnt, RL_TAG); if (refcount_is_zero(&zp->z_refcnt)) { list_remove(&zll->z_list, zp); ztest_znode_fini(zp); } mutex_exit(&zll->z_lock); } static void ztest_rll_init(rll_t *rll) { rll->rll_writer = NULL; rll->rll_readers = 0; mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); } static void ztest_rll_destroy(rll_t *rll) { ASSERT(rll->rll_writer == NULL); ASSERT(rll->rll_readers == 0); mutex_destroy(&rll->rll_lock); cv_destroy(&rll->rll_cv); } static void ztest_rll_lock(rll_t *rll, rl_type_t type) { mutex_enter(&rll->rll_lock); if (type == RL_READER) { while (rll->rll_writer != NULL) (void) cv_wait(&rll->rll_cv, &rll->rll_lock); rll->rll_readers++; } else { while (rll->rll_writer != NULL || rll->rll_readers) (void) cv_wait(&rll->rll_cv, &rll->rll_lock); rll->rll_writer = curthread; } mutex_exit(&rll->rll_lock); } static void ztest_rll_unlock(rll_t *rll) { mutex_enter(&rll->rll_lock); if (rll->rll_writer) { ASSERT(rll->rll_readers == 0); rll->rll_writer = NULL; } else { ASSERT(rll->rll_readers != 0); ASSERT(rll->rll_writer == NULL); rll->rll_readers--; } if (rll->rll_writer == NULL && rll->rll_readers == 0) cv_broadcast(&rll->rll_cv); mutex_exit(&rll->rll_lock); } static void ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) { rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; ztest_rll_lock(rll, type); } static void ztest_object_unlock(ztest_ds_t *zd, uint64_t object) { rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; ztest_rll_unlock(rll); } static ztest_zrl_t * ztest_zrl_init(rl_t *rl, ztest_znode_t *zp) { ztest_zrl_t *zrl = umem_alloc(sizeof (*zrl), UMEM_NOFAIL); zrl->z_rl = rl; zrl->z_ztznode = zp; return (zrl); } static void ztest_zrl_fini(ztest_zrl_t *zrl) { umem_free(zrl, sizeof (*zrl)); } static ztest_zrl_t * ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, rl_type_t type) { ztest_znode_t *zp = ztest_znode_get(zd, object); rl_t *rl = zfs_range_lock(&zp->z_range_lock, offset, size, type); return (ztest_zrl_init(rl, zp)); } static void ztest_range_unlock(ztest_ds_t *zd, ztest_zrl_t *zrl) { zfs_range_unlock(zrl->z_rl); ztest_znode_put(zd, zrl->z_ztznode); ztest_zrl_fini(zrl); } static void ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) { zd->zd_os = os; zd->zd_zilog = dmu_objset_zil(os); zd->zd_shared = szd; dmu_objset_name(os, zd->zd_name); int l; if (zd->zd_shared != NULL) zd->zd_shared->zd_seq = 0; VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL)); mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) ztest_rll_init(&zd->zd_object_lock[l]); for (l = 0; l < ZTEST_RANGE_LOCKS; l++) ztest_zll_init(&zd->zd_range_lock[l]); } static void ztest_zd_fini(ztest_ds_t *zd) { int l; mutex_destroy(&zd->zd_dirobj_lock); (void) pthread_rwlock_destroy(&zd->zd_zilog_lock); for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) ztest_rll_destroy(&zd->zd_object_lock[l]); for (l = 0; l < ZTEST_RANGE_LOCKS; l++) ztest_zll_destroy(&zd->zd_range_lock[l]); } #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) static uint64_t ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) { uint64_t txg; int error; /* * Attempt to assign tx to some transaction group. */ error = dmu_tx_assign(tx, txg_how); if (error) { if (error == ERESTART) { ASSERT(txg_how == TXG_NOWAIT); dmu_tx_wait(tx); } else { ASSERT3U(error, ==, ENOSPC); ztest_record_enospc(tag); } dmu_tx_abort(tx); return (0); } txg = dmu_tx_get_txg(tx); ASSERT(txg != 0); return (txg); } static void ztest_pattern_set(void *buf, uint64_t size, uint64_t value) { uint64_t *ip = buf; uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); while (ip < ip_end) *ip++ = value; } #ifndef NDEBUG static boolean_t ztest_pattern_match(void *buf, uint64_t size, uint64_t value) { uint64_t *ip = buf; uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); uint64_t diff = 0; while (ip < ip_end) diff |= (value - *ip++); return (diff == 0); } #endif static void ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) { bt->bt_magic = BT_MAGIC; bt->bt_objset = dmu_objset_id(os); bt->bt_object = object; bt->bt_dnodesize = dnodesize; bt->bt_offset = offset; bt->bt_gen = gen; bt->bt_txg = txg; bt->bt_crtxg = crtxg; } static void ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) { ASSERT3U(bt->bt_magic, ==, BT_MAGIC); ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os)); ASSERT3U(bt->bt_object, ==, object); ASSERT3U(bt->bt_dnodesize, ==, dnodesize); ASSERT3U(bt->bt_offset, ==, offset); ASSERT3U(bt->bt_gen, <=, gen); ASSERT3U(bt->bt_txg, <=, txg); ASSERT3U(bt->bt_crtxg, ==, crtxg); } static ztest_block_tag_t * ztest_bt_bonus(dmu_buf_t *db) { dmu_object_info_t doi; ztest_block_tag_t *bt; dmu_object_info_from_db(db, &doi); ASSERT3U(doi.doi_bonus_size, <=, db->db_size); ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); return (bt); } /* * Generate a token to fill up unused bonus buffer space. Try to make * it unique to the object, generation, and offset to verify that data * is not getting overwritten by data from other dnodes. */ #define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \ (((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset)) /* * Fill up the unused bonus buffer region before the block tag with a * verifiable pattern. Filling the whole bonus area with non-zero data * helps ensure that all dnode traversal code properly skips the * interior regions of large dnodes. */ void ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, objset_t *os, uint64_t gen) { uint64_t *bonusp; ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8)); for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), gen, bonusp - (uint64_t *)db->db_data); *bonusp = token; } } /* * Verify that the unused area of a bonus buffer is filled with the * expected tokens. */ void ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, objset_t *os, uint64_t gen) { uint64_t *bonusp; for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), gen, bonusp - (uint64_t *)db->db_data); VERIFY3U(*bonusp, ==, token); } } /* * ZIL logging ops */ #define lrz_type lr_mode #define lrz_blocksize lr_uid #define lrz_ibshift lr_gid #define lrz_bonustype lr_rdev #define lrz_dnodesize lr_crtime[1] static void ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) { char *name = (void *)(lr + 1); /* name follows lr */ size_t namesize = strlen(name) + 1; itx_t *itx; if (zil_replaying(zd->zd_zilog, tx)) return; itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, sizeof (*lr) + namesize - sizeof (lr_t)); zil_itx_assign(zd->zd_zilog, itx, tx); } static void ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) { char *name = (void *)(lr + 1); /* name follows lr */ size_t namesize = strlen(name) + 1; itx_t *itx; if (zil_replaying(zd->zd_zilog, tx)) return; itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, sizeof (*lr) + namesize - sizeof (lr_t)); itx->itx_oid = object; zil_itx_assign(zd->zd_zilog, itx, tx); } static void ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) { itx_t *itx; itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); if (zil_replaying(zd->zd_zilog, tx)) return; if (lr->lr_length > ZIL_MAX_LOG_DATA) write_state = WR_INDIRECT; itx = zil_itx_create(TX_WRITE, sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); if (write_state == WR_COPIED && dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { zil_itx_destroy(itx); itx = zil_itx_create(TX_WRITE, sizeof (*lr)); write_state = WR_NEED_COPY; } itx->itx_private = zd; itx->itx_wr_state = write_state; itx->itx_sync = (ztest_random(8) == 0); bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, sizeof (*lr) - sizeof (lr_t)); zil_itx_assign(zd->zd_zilog, itx, tx); } static void ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) { itx_t *itx; if (zil_replaying(zd->zd_zilog, tx)) return; itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, sizeof (*lr) - sizeof (lr_t)); itx->itx_sync = B_FALSE; zil_itx_assign(zd->zd_zilog, itx, tx); } static void ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) { itx_t *itx; if (zil_replaying(zd->zd_zilog, tx)) return; itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, sizeof (*lr) - sizeof (lr_t)); itx->itx_sync = B_FALSE; zil_itx_assign(zd->zd_zilog, itx, tx); } /* * ZIL replay ops */ static int ztest_replay_create(void *arg1, void *arg2, boolean_t byteswap) { ztest_ds_t *zd = arg1; lr_create_t *lr = arg2; char *name = (void *)(lr + 1); /* name follows lr */ objset_t *os = zd->zd_os; ztest_block_tag_t *bbt; dmu_buf_t *db; dmu_tx_t *tx; uint64_t txg; int error = 0; int bonuslen; if (byteswap) byteswap_uint64_array(lr, sizeof (*lr)); ASSERT(lr->lr_doid == ZTEST_DIROBJ); ASSERT(name[0] != '\0'); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); if (lr->lrz_type == DMU_OT_ZAP_OTHER) { dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); } else { dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); } txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); if (txg == 0) return (ENOSPC); ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize); if (lr->lrz_type == DMU_OT_ZAP_OTHER) { if (lr->lr_foid == 0) { lr->lr_foid = zap_create_dnsize(os, lr->lrz_type, lr->lrz_bonustype, bonuslen, lr->lrz_dnodesize, tx); } else { error = zap_create_claim_dnsize(os, lr->lr_foid, lr->lrz_type, lr->lrz_bonustype, bonuslen, lr->lrz_dnodesize, tx); } } else { if (lr->lr_foid == 0) { lr->lr_foid = dmu_object_alloc_dnsize(os, lr->lrz_type, 0, lr->lrz_bonustype, bonuslen, lr->lrz_dnodesize, tx); } else { error = dmu_object_claim_dnsize(os, lr->lr_foid, lr->lrz_type, 0, lr->lrz_bonustype, bonuslen, lr->lrz_dnodesize, tx); } } if (error) { ASSERT3U(error, ==, EEXIST); ASSERT(zd->zd_zilog->zl_replay); dmu_tx_commit(tx); return (error); } ASSERT(lr->lr_foid != 0); if (lr->lrz_type != DMU_OT_ZAP_OTHER) VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, lr->lrz_blocksize, lr->lrz_ibshift, tx)); VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); bbt = ztest_bt_bonus(db); dmu_buf_will_dirty(db, tx); ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL, lr->lr_gen, txg, txg); ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen); dmu_buf_rele(db, FTAG); VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, &lr->lr_foid, tx)); (void) ztest_log_create(zd, tx, lr); dmu_tx_commit(tx); return (0); } static int ztest_replay_remove(void *arg1, void *arg2, boolean_t byteswap) { ztest_ds_t *zd = arg1; lr_remove_t *lr = arg2; char *name = (void *)(lr + 1); /* name follows lr */ objset_t *os = zd->zd_os; dmu_object_info_t doi; dmu_tx_t *tx; uint64_t object, txg; if (byteswap) byteswap_uint64_array(lr, sizeof (*lr)); ASSERT(lr->lr_doid == ZTEST_DIROBJ); ASSERT(name[0] != '\0'); VERIFY3U(0, ==, zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); ASSERT(object != 0); ztest_object_lock(zd, object, RL_WRITER); VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); if (txg == 0) { ztest_object_unlock(zd, object); return (ENOSPC); } if (doi.doi_type == DMU_OT_ZAP_OTHER) { VERIFY3U(0, ==, zap_destroy(os, object, tx)); } else { VERIFY3U(0, ==, dmu_object_free(os, object, tx)); } VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); (void) ztest_log_remove(zd, tx, lr, object); dmu_tx_commit(tx); ztest_object_unlock(zd, object); return (0); } static int ztest_replay_write(void *arg1, void *arg2, boolean_t byteswap) { ztest_ds_t *zd = arg1; lr_write_t *lr = arg2; objset_t *os = zd->zd_os; void *data = lr + 1; /* data follows lr */ uint64_t offset, length; ztest_block_tag_t *bt = data; ztest_block_tag_t *bbt; uint64_t gen, txg, lrtxg, crtxg; dmu_object_info_t doi; dmu_tx_t *tx; dmu_buf_t *db; arc_buf_t *abuf = NULL; ztest_zrl_t *rl; if (byteswap) byteswap_uint64_array(lr, sizeof (*lr)); offset = lr->lr_offset; length = lr->lr_length; /* If it's a dmu_sync() block, write the whole block */ if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); if (length < blocksize) { offset -= offset % blocksize; length = blocksize; } } if (bt->bt_magic == BSWAP_64(BT_MAGIC)) byteswap_uint64_array(bt, sizeof (*bt)); if (bt->bt_magic != BT_MAGIC) bt = NULL; ztest_object_lock(zd, lr->lr_foid, RL_READER); rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); dmu_object_info_from_db(db, &doi); bbt = ztest_bt_bonus(db); ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); gen = bbt->bt_gen; crtxg = bbt->bt_crtxg; lrtxg = lr->lr_common.lrc_txg; tx = dmu_tx_create(os); dmu_tx_hold_write(tx, lr->lr_foid, offset, length); if (ztest_random(8) == 0 && length == doi.doi_data_block_size && P2PHASE(offset, length) == 0) abuf = dmu_request_arcbuf(db, length); txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); if (txg == 0) { if (abuf != NULL) dmu_return_arcbuf(abuf); dmu_buf_rele(db, FTAG); ztest_range_unlock(zd, rl); ztest_object_unlock(zd, lr->lr_foid); return (ENOSPC); } if (bt != NULL) { /* * Usually, verify the old data before writing new data -- * but not always, because we also want to verify correct * behavior when the data was not recently read into cache. */ ASSERT(offset % doi.doi_data_block_size == 0); if (ztest_random(4) != 0) { int prefetch = ztest_random(2) ? DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; ztest_block_tag_t rbt; VERIFY(dmu_read(os, lr->lr_foid, offset, sizeof (rbt), &rbt, prefetch) == 0); if (rbt.bt_magic == BT_MAGIC) { ztest_bt_verify(&rbt, os, lr->lr_foid, 0, offset, gen, txg, crtxg); } } /* * Writes can appear to be newer than the bonus buffer because * the ztest_get_data() callback does a dmu_read() of the * open-context data, which may be different than the data * as it was when the write was generated. */ if (zd->zd_zilog->zl_replay) { ztest_bt_verify(bt, os, lr->lr_foid, 0, offset, MAX(gen, bt->bt_gen), MAX(txg, lrtxg), bt->bt_crtxg); } /* * Set the bt's gen/txg to the bonus buffer's gen/txg * so that all of the usual ASSERTs will work. */ ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg, crtxg); } if (abuf == NULL) { dmu_write(os, lr->lr_foid, offset, length, data, tx); } else { bcopy(data, abuf->b_data, length); dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx); } (void) ztest_log_write(zd, tx, lr); dmu_buf_rele(db, FTAG); dmu_tx_commit(tx); ztest_range_unlock(zd, rl); ztest_object_unlock(zd, lr->lr_foid); return (0); } static int ztest_replay_truncate(void *arg1, void *arg2, boolean_t byteswap) { ztest_ds_t *zd = arg1; lr_truncate_t *lr = arg2; objset_t *os = zd->zd_os; dmu_tx_t *tx; uint64_t txg; ztest_zrl_t *rl; if (byteswap) byteswap_uint64_array(lr, sizeof (*lr)); ztest_object_lock(zd, lr->lr_foid, RL_READER); rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, RL_WRITER); tx = dmu_tx_create(os); dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); if (txg == 0) { ztest_range_unlock(zd, rl); ztest_object_unlock(zd, lr->lr_foid); return (ENOSPC); } VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, lr->lr_length, tx) == 0); (void) ztest_log_truncate(zd, tx, lr); dmu_tx_commit(tx); ztest_range_unlock(zd, rl); ztest_object_unlock(zd, lr->lr_foid); return (0); } static int ztest_replay_setattr(void *arg1, void *arg2, boolean_t byteswap) { ztest_ds_t *zd = arg1; lr_setattr_t *lr = arg2; objset_t *os = zd->zd_os; dmu_tx_t *tx; dmu_buf_t *db; ztest_block_tag_t *bbt; uint64_t txg, lrtxg, crtxg, dnodesize; if (byteswap) byteswap_uint64_array(lr, sizeof (*lr)); ztest_object_lock(zd, lr->lr_foid, RL_WRITER); VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); tx = dmu_tx_create(os); dmu_tx_hold_bonus(tx, lr->lr_foid); txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); if (txg == 0) { dmu_buf_rele(db, FTAG); ztest_object_unlock(zd, lr->lr_foid); return (ENOSPC); } bbt = ztest_bt_bonus(db); ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); crtxg = bbt->bt_crtxg; lrtxg = lr->lr_common.lrc_txg; dnodesize = bbt->bt_dnodesize; if (zd->zd_zilog->zl_replay) { ASSERT(lr->lr_size != 0); ASSERT(lr->lr_mode != 0); ASSERT(lrtxg != 0); } else { /* * Randomly change the size and increment the generation. */ lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * sizeof (*bbt); lr->lr_mode = bbt->bt_gen + 1; ASSERT(lrtxg == 0); } /* * Verify that the current bonus buffer is not newer than our txg. */ ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, MAX(txg, lrtxg), crtxg); dmu_buf_will_dirty(db, tx); ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); ASSERT3U(lr->lr_size, <=, db->db_size); VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); bbt = ztest_bt_bonus(db); ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, txg, crtxg); ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen); dmu_buf_rele(db, FTAG); (void) ztest_log_setattr(zd, tx, lr); dmu_tx_commit(tx); ztest_object_unlock(zd, lr->lr_foid); return (0); } zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { NULL, /* 0 no such transaction type */ ztest_replay_create, /* TX_CREATE */ NULL, /* TX_MKDIR */ NULL, /* TX_MKXATTR */ NULL, /* TX_SYMLINK */ ztest_replay_remove, /* TX_REMOVE */ NULL, /* TX_RMDIR */ NULL, /* TX_LINK */ NULL, /* TX_RENAME */ ztest_replay_write, /* TX_WRITE */ ztest_replay_truncate, /* TX_TRUNCATE */ ztest_replay_setattr, /* TX_SETATTR */ NULL, /* TX_ACL */ NULL, /* TX_CREATE_ACL */ NULL, /* TX_CREATE_ATTR */ NULL, /* TX_CREATE_ACL_ATTR */ NULL, /* TX_MKDIR_ACL */ NULL, /* TX_MKDIR_ATTR */ NULL, /* TX_MKDIR_ACL_ATTR */ NULL, /* TX_WRITE2 */ }; /* * ZIL get_data callbacks */ typedef struct ztest_zgd_private { ztest_ds_t *z_zd; ztest_zrl_t *z_rl; uint64_t z_object; } ztest_zgd_private_t; static void ztest_get_done(zgd_t *zgd, int error) { ztest_zgd_private_t *zzp = zgd->zgd_private; ztest_ds_t *zd = zzp->z_zd; uint64_t object = zzp->z_object; if (zgd->zgd_db) dmu_buf_rele(zgd->zgd_db, zgd); ztest_range_unlock(zd, zzp->z_rl); ztest_object_unlock(zd, object); if (error == 0 && zgd->zgd_bp) zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp); umem_free(zgd, sizeof (*zgd)); umem_free(zzp, sizeof (*zzp)); } static int ztest_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio) { ztest_ds_t *zd = arg; objset_t *os = zd->zd_os; uint64_t object = lr->lr_foid; uint64_t offset = lr->lr_offset; uint64_t size = lr->lr_length; uint64_t txg = lr->lr_common.lrc_txg; uint64_t crtxg; dmu_object_info_t doi; dmu_buf_t *db; zgd_t *zgd; int error; ztest_zgd_private_t *zgd_private; ASSERT3P(lwb, !=, NULL); ASSERT3P(zio, !=, NULL); ASSERT3U(size, !=, 0); ztest_object_lock(zd, object, RL_READER); error = dmu_bonus_hold(os, object, FTAG, &db); if (error) { ztest_object_unlock(zd, object); return (error); } crtxg = ztest_bt_bonus(db)->bt_crtxg; if (crtxg == 0 || crtxg > txg) { dmu_buf_rele(db, FTAG); ztest_object_unlock(zd, object); return (ENOENT); } dmu_object_info_from_db(db, &doi); dmu_buf_rele(db, FTAG); db = NULL; zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); zgd->zgd_lwb = lwb; zgd_private = umem_zalloc(sizeof (ztest_zgd_private_t), UMEM_NOFAIL); zgd_private->z_zd = zd; zgd_private->z_object = object; zgd->zgd_private = zgd_private; if (buf != NULL) { /* immediate write */ zgd_private->z_rl = ztest_range_lock(zd, object, offset, size, RL_READER); zgd->zgd_rl = zgd_private->z_rl->z_rl; error = dmu_read(os, object, offset, size, buf, DMU_READ_NO_PREFETCH); ASSERT(error == 0); } else { size = doi.doi_data_block_size; if (ISP2(size)) { offset = P2ALIGN(offset, size); } else { ASSERT(offset < size); offset = 0; } zgd_private->z_rl = ztest_range_lock(zd, object, offset, size, RL_READER); zgd->zgd_rl = zgd_private->z_rl->z_rl; error = dmu_buf_hold(os, object, offset, zgd, &db, DMU_READ_NO_PREFETCH); if (error == 0) { blkptr_t *bp = &lr->lr_blkptr; zgd->zgd_db = db; zgd->zgd_bp = bp; ASSERT(db->db_offset == offset); ASSERT(db->db_size == size); error = dmu_sync(zio, lr->lr_common.lrc_txg, ztest_get_done, zgd); if (error == 0) return (0); } } ztest_get_done(zgd, error); return (error); } static void * ztest_lr_alloc(size_t lrsize, char *name) { char *lr; size_t namesize = name ? strlen(name) + 1 : 0; lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); if (name) bcopy(name, lr + lrsize, namesize); return (lr); } void ztest_lr_free(void *lr, size_t lrsize, char *name) { size_t namesize = name ? strlen(name) + 1 : 0; umem_free(lr, lrsize + namesize); } /* * Lookup a bunch of objects. Returns the number of objects not found. */ static int ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) { int missing = 0; int error; int i; ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); for (i = 0; i < count; i++, od++) { od->od_object = 0; error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, sizeof (uint64_t), 1, &od->od_object); if (error) { ASSERT(error == ENOENT); ASSERT(od->od_object == 0); missing++; } else { dmu_buf_t *db; ztest_block_tag_t *bbt; dmu_object_info_t doi; ASSERT(od->od_object != 0); ASSERT(missing == 0); /* there should be no gaps */ ztest_object_lock(zd, od->od_object, RL_READER); VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, od->od_object, FTAG, &db)); dmu_object_info_from_db(db, &doi); bbt = ztest_bt_bonus(db); ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); od->od_type = doi.doi_type; od->od_blocksize = doi.doi_data_block_size; od->od_gen = bbt->bt_gen; dmu_buf_rele(db, FTAG); ztest_object_unlock(zd, od->od_object); } } return (missing); } static int ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) { int missing = 0; int i; ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); for (i = 0; i < count; i++, od++) { if (missing) { od->od_object = 0; missing++; continue; } lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); lr->lr_doid = od->od_dir; lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ lr->lrz_type = od->od_crtype; lr->lrz_blocksize = od->od_crblocksize; lr->lrz_ibshift = ztest_random_ibshift(); lr->lrz_bonustype = DMU_OT_UINT64_OTHER; lr->lrz_dnodesize = od->od_crdnodesize; lr->lr_gen = od->od_crgen; lr->lr_crtime[0] = time(NULL); if (ztest_replay_create(zd, lr, B_FALSE) != 0) { ASSERT(missing == 0); od->od_object = 0; missing++; } else { od->od_object = lr->lr_foid; od->od_type = od->od_crtype; od->od_blocksize = od->od_crblocksize; od->od_gen = od->od_crgen; ASSERT(od->od_object != 0); } ztest_lr_free(lr, sizeof (*lr), od->od_name); } return (missing); } static int ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) { int missing = 0; int error; int i; ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); od += count - 1; for (i = count - 1; i >= 0; i--, od--) { if (missing) { missing++; continue; } /* * No object was found. */ if (od->od_object == 0) continue; lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); lr->lr_doid = od->od_dir; if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { ASSERT3U(error, ==, ENOSPC); missing++; } else { od->od_object = 0; } ztest_lr_free(lr, sizeof (*lr), od->od_name); } return (missing); } static int ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, void *data) { lr_write_t *lr; int error; lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); lr->lr_foid = object; lr->lr_offset = offset; lr->lr_length = size; lr->lr_blkoff = 0; BP_ZERO(&lr->lr_blkptr); bcopy(data, lr + 1, size); error = ztest_replay_write(zd, lr, B_FALSE); ztest_lr_free(lr, sizeof (*lr) + size, NULL); return (error); } static int ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) { lr_truncate_t *lr; int error; lr = ztest_lr_alloc(sizeof (*lr), NULL); lr->lr_foid = object; lr->lr_offset = offset; lr->lr_length = size; error = ztest_replay_truncate(zd, lr, B_FALSE); ztest_lr_free(lr, sizeof (*lr), NULL); return (error); } static int ztest_setattr(ztest_ds_t *zd, uint64_t object) { lr_setattr_t *lr; int error; lr = ztest_lr_alloc(sizeof (*lr), NULL); lr->lr_foid = object; lr->lr_size = 0; lr->lr_mode = 0; error = ztest_replay_setattr(zd, lr, B_FALSE); ztest_lr_free(lr, sizeof (*lr), NULL); return (error); } static void ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) { objset_t *os = zd->zd_os; dmu_tx_t *tx; uint64_t txg; ztest_zrl_t *rl; txg_wait_synced(dmu_objset_pool(os), 0); ztest_object_lock(zd, object, RL_READER); rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); tx = dmu_tx_create(os); dmu_tx_hold_write(tx, object, offset, size); txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); if (txg != 0) { dmu_prealloc(os, object, offset, size, tx); dmu_tx_commit(tx); txg_wait_synced(dmu_objset_pool(os), txg); } else { (void) dmu_free_long_range(os, object, offset, size); } ztest_range_unlock(zd, rl); ztest_object_unlock(zd, object); } static void ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) { int err; ztest_block_tag_t wbt; dmu_object_info_t doi; enum ztest_io_type io_type; uint64_t blocksize; void *data; VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); blocksize = doi.doi_data_block_size; data = umem_alloc(blocksize, UMEM_NOFAIL); /* * Pick an i/o type at random, biased toward writing block tags. */ io_type = ztest_random(ZTEST_IO_TYPES); if (ztest_random(2) == 0) io_type = ZTEST_IO_WRITE_TAG; (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); switch (io_type) { case ZTEST_IO_WRITE_TAG: ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize, offset, 0, 0, 0); (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); break; case ZTEST_IO_WRITE_PATTERN: (void) memset(data, 'a' + (object + offset) % 5, blocksize); if (ztest_random(2) == 0) { /* * Induce fletcher2 collisions to ensure that * zio_ddt_collision() detects and resolves them * when using fletcher2-verify for deduplication. */ ((uint64_t *)data)[0] ^= 1ULL << 63; ((uint64_t *)data)[4] ^= 1ULL << 63; } (void) ztest_write(zd, object, offset, blocksize, data); break; case ZTEST_IO_WRITE_ZEROES: bzero(data, blocksize); (void) ztest_write(zd, object, offset, blocksize, data); break; case ZTEST_IO_TRUNCATE: (void) ztest_truncate(zd, object, offset, blocksize); break; case ZTEST_IO_SETATTR: (void) ztest_setattr(zd, object); break; default: break; case ZTEST_IO_REWRITE: (void) pthread_rwlock_rdlock(&ztest_name_lock); err = ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa), B_FALSE); VERIFY(err == 0 || err == ENOSPC); err = ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COMPRESSION, ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), B_FALSE); VERIFY(err == 0 || err == ENOSPC); (void) pthread_rwlock_unlock(&ztest_name_lock); VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data, DMU_READ_NO_PREFETCH)); (void) ztest_write(zd, object, offset, blocksize, data); break; } (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); umem_free(data, blocksize); } /* * Initialize an object description template. */ static void ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize, uint64_t gen) { od->od_dir = ZTEST_DIROBJ; od->od_object = 0; od->od_crtype = type; od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize(); od->od_crgen = gen; od->od_type = DMU_OT_NONE; od->od_blocksize = 0; od->od_gen = 0; (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", tag, (longlong_t)id, (u_longlong_t)index); } /* * Lookup or create the objects for a test using the od template. * If the objects do not all exist, or if 'remove' is specified, * remove any existing objects and create new ones. Otherwise, * use the existing objects. */ static int ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) { int count = size / sizeof (*od); int rv = 0; mutex_enter(&zd->zd_dirobj_lock); if ((ztest_lookup(zd, od, count) != 0 || remove) && (ztest_remove(zd, od, count) != 0 || ztest_create(zd, od, count) != 0)) rv = -1; zd->zd_od = od; mutex_exit(&zd->zd_dirobj_lock); return (rv); } /* ARGSUSED */ void ztest_zil_commit(ztest_ds_t *zd, uint64_t id) { zilog_t *zilog = zd->zd_zilog; (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); /* * Remember the committed values in zd, which is in parent/child * shared memory. If we die, the next iteration of ztest_run() * will verify that the log really does contain this record. */ mutex_enter(&zilog->zl_lock); ASSERT(zd->zd_shared != NULL); ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; mutex_exit(&zilog->zl_lock); (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); } /* * This function is designed to simulate the operations that occur during a * mount/unmount operation. We hold the dataset across these operations in an * attempt to expose any implicit assumptions about ZIL management. */ /* ARGSUSED */ void ztest_zil_remount(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; /* * We grab the zd_dirobj_lock to ensure that no other thread is * updating the zil (i.e. adding in-memory log records) and the * zd_zilog_lock to block any I/O. */ mutex_enter(&zd->zd_dirobj_lock); (void) pthread_rwlock_wrlock(&zd->zd_zilog_lock); /* zfsvfs_teardown() */ zil_close(zd->zd_zilog); /* zfsvfs_setup() */ VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); zil_replay(os, zd, ztest_replay_vector); (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); mutex_exit(&zd->zd_dirobj_lock); } /* * Verify that we can't destroy an active pool, create an existing pool, * or create a pool with a bad vdev spec. */ /* ARGSUSED */ void ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) { ztest_shared_opts_t *zo = &ztest_opts; spa_t *spa; nvlist_t *nvroot; if (zo->zo_mmp_test) return; /* * Attempt to create using a bad file. */ nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); VERIFY3U(ENOENT, ==, spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); nvlist_free(nvroot); /* * Attempt to create using a bad mirror. */ nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 2, 1); VERIFY3U(ENOENT, ==, spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); nvlist_free(nvroot); /* * Attempt to create an existing pool. It shouldn't matter * what's in the nvroot; we should fail with EEXIST. */ (void) pthread_rwlock_rdlock(&ztest_name_lock); nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); VERIFY3U(EEXIST, ==, spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); nvlist_free(nvroot); VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool)); spa_close(spa, FTAG); (void) pthread_rwlock_unlock(&ztest_name_lock); } /* * Start and then stop the MMP threads to ensure the startup and shutdown code * works properly. Actual protection and property-related code tested via ZTS. */ /* ARGSUSED */ void ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id) { ztest_shared_opts_t *zo = &ztest_opts; spa_t *spa = ztest_spa; if (zo->zo_mmp_test) return; spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); mutex_enter(&spa->spa_props_lock); zfs_multihost_fail_intervals = 0; if (!spa_multihost(spa)) { spa->spa_multihost = B_TRUE; mmp_thread_start(spa); } mutex_exit(&spa->spa_props_lock); spa_config_exit(spa, SCL_CONFIG, FTAG); txg_wait_synced(spa_get_dsl(spa), 0); mmp_signal_all_threads(); txg_wait_synced(spa_get_dsl(spa), 0); spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); mutex_enter(&spa->spa_props_lock); if (spa_multihost(spa)) { mmp_thread_stop(spa); spa->spa_multihost = B_FALSE; } mutex_exit(&spa->spa_props_lock); spa_config_exit(spa, SCL_CONFIG, FTAG); } /* ARGSUSED */ void ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) { spa_t *spa; uint64_t initial_version = SPA_VERSION_INITIAL; uint64_t version, newversion; nvlist_t *nvroot, *props; char *name; if (ztest_opts.zo_mmp_test) return; mutex_enter(&ztest_vdev_lock); name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); /* * Clean up from previous runs. */ (void) spa_destroy(name); nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, 0, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1); /* * If we're configuring a RAIDZ device then make sure that the * the initial version is capable of supporting that feature. */ switch (ztest_opts.zo_raidz_parity) { case 0: case 1: initial_version = SPA_VERSION_INITIAL; break; case 2: initial_version = SPA_VERSION_RAIDZ2; break; case 3: initial_version = SPA_VERSION_RAIDZ3; break; } /* * Create a pool with a spa version that can be upgraded. Pick * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. */ do { version = ztest_random_spa_version(initial_version); } while (version > SPA_VERSION_BEFORE_FEATURES); props = fnvlist_alloc(); fnvlist_add_uint64(props, zpool_prop_to_name(ZPOOL_PROP_VERSION), version); VERIFY3S(spa_create(name, nvroot, props, NULL, NULL), ==, 0); fnvlist_free(nvroot); fnvlist_free(props); VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); VERIFY3U(spa_version(spa), ==, version); newversion = ztest_random_spa_version(version + 1); if (ztest_opts.zo_verbose >= 4) { (void) printf("upgrading spa version from %llu to %llu\n", (u_longlong_t)version, (u_longlong_t)newversion); } spa_upgrade(spa, newversion); VERIFY3U(spa_version(spa), >, version); VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, zpool_prop_to_name(ZPOOL_PROP_VERSION))); spa_close(spa, FTAG); strfree(name); mutex_exit(&ztest_vdev_lock); } static void ztest_spa_checkpoint(spa_t *spa) { ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); int error = spa_checkpoint(spa->spa_name); switch (error) { case 0: case ZFS_ERR_DEVRM_IN_PROGRESS: case ZFS_ERR_DISCARDING_CHECKPOINT: case ZFS_ERR_CHECKPOINT_EXISTS: break; case ENOSPC: ztest_record_enospc(FTAG); break; default: fatal(0, "spa_checkpoint(%s) = %d", spa->spa_name, error); } } static void ztest_spa_discard_checkpoint(spa_t *spa) { ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); int error = spa_checkpoint_discard(spa->spa_name); switch (error) { case 0: case ZFS_ERR_DISCARDING_CHECKPOINT: case ZFS_ERR_NO_CHECKPOINT: break; default: fatal(0, "spa_discard_checkpoint(%s) = %d", spa->spa_name, error); } } /* ARGSUSED */ void ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id) { spa_t *spa = ztest_spa; mutex_enter(&ztest_checkpoint_lock); if (ztest_random(2) == 0) { ztest_spa_checkpoint(spa); } else { ztest_spa_discard_checkpoint(spa); } mutex_exit(&ztest_checkpoint_lock); } static vdev_t * vdev_lookup_by_path(vdev_t *vd, const char *path) { vdev_t *mvd; int c; if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) return (vd); for (c = 0; c < vd->vdev_children; c++) if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != NULL) return (mvd); return (NULL); } /* * Find the first available hole which can be used as a top-level. */ int find_vdev_hole(spa_t *spa) { vdev_t *rvd = spa->spa_root_vdev; int c; ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); for (c = 0; c < rvd->vdev_children; c++) { vdev_t *cvd = rvd->vdev_child[c]; if (cvd->vdev_ishole) break; } return (c); } /* * Verify that vdev_add() works as expected. */ /* ARGSUSED */ void ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) { ztest_shared_t *zs = ztest_shared; spa_t *spa = ztest_spa; uint64_t leaves; uint64_t guid; nvlist_t *nvroot; int error; if (ztest_opts.zo_mmp_test) return; mutex_enter(&ztest_vdev_lock); leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; /* * If we have slogs then remove them 1/4 of the time. */ if (spa_has_slogs(spa) && ztest_random(4) == 0) { /* * Grab the guid from the head of the log class rotor. */ guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; spa_config_exit(spa, SCL_VDEV, FTAG); /* * We have to grab the zs_name_lock as writer to * prevent a race between removing a slog (dmu_objset_find) * and destroying a dataset. Removing the slog will * grab a reference on the dataset which may cause * dsl_destroy_head() to fail with EBUSY thus * leaving the dataset in an inconsistent state. */ pthread_rwlock_wrlock(&ztest_name_lock); error = spa_vdev_remove(spa, guid, B_FALSE); pthread_rwlock_unlock(&ztest_name_lock); switch (error) { case 0: case EEXIST: /* Generic zil_reset() error */ case EBUSY: /* Replay required */ case EACCES: /* Crypto key not loaded */ case ZFS_ERR_CHECKPOINT_EXISTS: case ZFS_ERR_DISCARDING_CHECKPOINT: break; default: fatal(0, "spa_vdev_remove() = %d", error); } } else { spa_config_exit(spa, SCL_VDEV, FTAG); /* * Make 1/4 of the devices be log devices. */ nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, ztest_random(4) == 0, ztest_opts.zo_raidz, zs->zs_mirrors, 1); error = spa_vdev_add(spa, nvroot); nvlist_free(nvroot); switch (error) { case 0: break; case ENOSPC: ztest_record_enospc("spa_vdev_add"); break; default: fatal(0, "spa_vdev_add() = %d", error); } } mutex_exit(&ztest_vdev_lock); } /* * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. */ /* ARGSUSED */ void ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) { ztest_shared_t *zs = ztest_shared; spa_t *spa = ztest_spa; vdev_t *rvd = spa->spa_root_vdev; spa_aux_vdev_t *sav; char *aux; char *path; uint64_t guid = 0; int error; if (ztest_opts.zo_mmp_test) return; path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); if (ztest_random(2) == 0) { sav = &spa->spa_spares; aux = ZPOOL_CONFIG_SPARES; } else { sav = &spa->spa_l2cache; aux = ZPOOL_CONFIG_L2CACHE; } mutex_enter(&ztest_vdev_lock); spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); if (sav->sav_count != 0 && ztest_random(4) == 0) { /* * Pick a random device to remove. */ guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; } else { /* * Find an unused device we can add. */ zs->zs_vdev_aux = 0; for (;;) { int c; (void) snprintf(path, MAXPATHLEN, ztest_aux_template, ztest_opts.zo_dir, ztest_opts.zo_pool, aux, zs->zs_vdev_aux); for (c = 0; c < sav->sav_count; c++) if (strcmp(sav->sav_vdevs[c]->vdev_path, path) == 0) break; if (c == sav->sav_count && vdev_lookup_by_path(rvd, path) == NULL) break; zs->zs_vdev_aux++; } } spa_config_exit(spa, SCL_VDEV, FTAG); if (guid == 0) { /* * Add a new device. */ nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1); error = spa_vdev_add(spa, nvroot); switch (error) { case 0: break; default: fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); } nvlist_free(nvroot); } else { /* * Remove an existing device. Sometimes, dirty its * vdev state first to make sure we handle removal * of devices that have pending state changes. */ if (ztest_random(2) == 0) (void) vdev_online(spa, guid, 0, NULL); error = spa_vdev_remove(spa, guid, B_FALSE); switch (error) { case 0: case EBUSY: case ZFS_ERR_CHECKPOINT_EXISTS: case ZFS_ERR_DISCARDING_CHECKPOINT: break; default: fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); } } mutex_exit(&ztest_vdev_lock); umem_free(path, MAXPATHLEN); } /* * split a pool if it has mirror tlvdevs */ /* ARGSUSED */ void ztest_split_pool(ztest_ds_t *zd, uint64_t id) { ztest_shared_t *zs = ztest_shared; spa_t *spa = ztest_spa; vdev_t *rvd = spa->spa_root_vdev; nvlist_t *tree, **child, *config, *split, **schild; uint_t c, children, schildren = 0, lastlogid = 0; int error = 0; if (ztest_opts.zo_mmp_test) return; mutex_enter(&ztest_vdev_lock); /* ensure we have a useable config; mirrors of raidz aren't supported */ if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { mutex_exit(&ztest_vdev_lock); return; } /* clean up the old pool, if any */ (void) spa_destroy("splitp"); spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); /* generate a config from the existing config */ mutex_enter(&spa->spa_props_lock); VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, &tree) == 0); mutex_exit(&spa->spa_props_lock); VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, &children) == 0); schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); for (c = 0; c < children; c++) { vdev_t *tvd = rvd->vdev_child[c]; nvlist_t **mchild; uint_t mchildren; if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_string(schild[schildren], ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); VERIFY(nvlist_add_uint64(schild[schildren], ZPOOL_CONFIG_IS_HOLE, 1) == 0); if (lastlogid == 0) lastlogid = schildren; ++schildren; continue; } lastlogid = 0; VERIFY(nvlist_lookup_nvlist_array(child[c], ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); } /* OK, create a config that can be used to split */ VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, lastlogid != 0 ? lastlogid : schildren) == 0); VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); for (c = 0; c < schildren; c++) nvlist_free(schild[c]); free(schild); nvlist_free(split); spa_config_exit(spa, SCL_VDEV, FTAG); (void) pthread_rwlock_wrlock(&ztest_name_lock); error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); (void) pthread_rwlock_unlock(&ztest_name_lock); nvlist_free(config); if (error == 0) { (void) printf("successful split - results:\n"); mutex_enter(&spa_namespace_lock); show_pool_stats(spa); show_pool_stats(spa_lookup("splitp")); mutex_exit(&spa_namespace_lock); ++zs->zs_splits; --zs->zs_mirrors; } mutex_exit(&ztest_vdev_lock); } /* * Verify that we can attach and detach devices. */ /* ARGSUSED */ void ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) { ztest_shared_t *zs = ztest_shared; spa_t *spa = ztest_spa; spa_aux_vdev_t *sav = &spa->spa_spares; vdev_t *rvd = spa->spa_root_vdev; vdev_t *oldvd, *newvd, *pvd; nvlist_t *root; uint64_t leaves; uint64_t leaf, top; uint64_t ashift = ztest_get_ashift(); uint64_t oldguid, pguid; uint64_t oldsize, newsize; char *oldpath, *newpath; int replacing; int oldvd_has_siblings = B_FALSE; int newvd_is_spare = B_FALSE; int oldvd_is_log; int error, expected_error; if (ztest_opts.zo_mmp_test) return; oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); mutex_enter(&ztest_vdev_lock); leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); /* * If a vdev is in the process of being removed, its removal may * finish while we are in progress, leading to an unexpected error * value. Don't bother trying to attach while we are in the middle * of removal. */ if (ztest_device_removal_active) { spa_config_exit(spa, SCL_ALL, FTAG); mutex_exit(&ztest_vdev_lock); return; } /* * Decide whether to do an attach or a replace. */ replacing = ztest_random(2); /* * Pick a random top-level vdev. */ top = ztest_random_vdev_top(spa, B_TRUE); /* * Pick a random leaf within it. */ leaf = ztest_random(leaves); /* * Locate this vdev. */ oldvd = rvd->vdev_child[top]; if (zs->zs_mirrors >= 1) { ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); ASSERT(oldvd->vdev_children >= zs->zs_mirrors); oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; } if (ztest_opts.zo_raidz > 1) { ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; } /* * If we're already doing an attach or replace, oldvd may be a * mirror vdev -- in which case, pick a random child. */ while (oldvd->vdev_children != 0) { oldvd_has_siblings = B_TRUE; ASSERT(oldvd->vdev_children >= 2); oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; } oldguid = oldvd->vdev_guid; oldsize = vdev_get_min_asize(oldvd); oldvd_is_log = oldvd->vdev_top->vdev_islog; (void) strcpy(oldpath, oldvd->vdev_path); pvd = oldvd->vdev_parent; pguid = pvd->vdev_guid; /* * If oldvd has siblings, then half of the time, detach it. */ if (oldvd_has_siblings && ztest_random(2) == 0) { spa_config_exit(spa, SCL_ALL, FTAG); error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); if (error != 0 && error != ENODEV && error != EBUSY && error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS && error != ZFS_ERR_DISCARDING_CHECKPOINT) fatal(0, "detach (%s) returned %d", oldpath, error); goto out; } /* * For the new vdev, choose with equal probability between the two * standard paths (ending in either 'a' or 'b') or a random hot spare. */ if (sav->sav_count != 0 && ztest_random(3) == 0) { newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; newvd_is_spare = B_TRUE; (void) strcpy(newpath, newvd->vdev_path); } else { (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, ztest_opts.zo_dir, ztest_opts.zo_pool, top * leaves + leaf); if (ztest_random(2) == 0) newpath[strlen(newpath) - 1] = 'b'; newvd = vdev_lookup_by_path(rvd, newpath); } if (newvd) { /* * Reopen to ensure the vdev's asize field isn't stale. */ vdev_reopen(newvd); newsize = vdev_get_min_asize(newvd); } else { /* * Make newsize a little bigger or smaller than oldsize. * If it's smaller, the attach should fail. * If it's larger, and we're doing a replace, * we should get dynamic LUN growth when we're done. */ newsize = 10 * oldsize / (9 + ztest_random(3)); } /* * If pvd is not a mirror or root, the attach should fail with ENOTSUP, * unless it's a replace; in that case any non-replacing parent is OK. * * If newvd is already part of the pool, it should fail with EBUSY. * * If newvd is too small, it should fail with EOVERFLOW. */ if (pvd->vdev_ops != &vdev_mirror_ops && pvd->vdev_ops != &vdev_root_ops && (!replacing || pvd->vdev_ops == &vdev_replacing_ops || pvd->vdev_ops == &vdev_spare_ops)) expected_error = ENOTSUP; else if (newvd_is_spare && (!replacing || oldvd_is_log)) expected_error = ENOTSUP; else if (newvd == oldvd) expected_error = replacing ? 0 : EBUSY; else if (vdev_lookup_by_path(rvd, newpath) != NULL) expected_error = EBUSY; else if (newsize < oldsize) expected_error = EOVERFLOW; else if (ashift > oldvd->vdev_top->vdev_ashift) expected_error = EDOM; else expected_error = 0; spa_config_exit(spa, SCL_ALL, FTAG); /* * Build the nvlist describing newpath. */ root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, ashift, 0, 0, 0, 1); error = spa_vdev_attach(spa, oldguid, root, replacing); nvlist_free(root); /* * If our parent was the replacing vdev, but the replace completed, * then instead of failing with ENOTSUP we may either succeed, * fail with ENODEV, or fail with EOVERFLOW. */ if (expected_error == ENOTSUP && (error == 0 || error == ENODEV || error == EOVERFLOW)) expected_error = error; /* * If someone grew the LUN, the replacement may be too small. */ if (error == EOVERFLOW || error == EBUSY) expected_error = error; if (error == ZFS_ERR_CHECKPOINT_EXISTS || error == ZFS_ERR_DISCARDING_CHECKPOINT) expected_error = error; /* XXX workaround 6690467 */ if (error != expected_error && expected_error != EBUSY) { fatal(0, "attach (%s %llu, %s %llu, %d) " "returned %d, expected %d", oldpath, oldsize, newpath, newsize, replacing, error, expected_error); } out: mutex_exit(&ztest_vdev_lock); umem_free(oldpath, MAXPATHLEN); umem_free(newpath, MAXPATHLEN); } /* ARGSUSED */ void ztest_device_removal(ztest_ds_t *zd, uint64_t id) { spa_t *spa = ztest_spa; vdev_t *vd; uint64_t guid; int error; mutex_enter(&ztest_vdev_lock); if (ztest_device_removal_active) { mutex_exit(&ztest_vdev_lock); return; } /* * Remove a random top-level vdev and wait for removal to finish. */ spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE)); guid = vd->vdev_guid; spa_config_exit(spa, SCL_VDEV, FTAG); error = spa_vdev_remove(spa, guid, B_FALSE); if (error == 0) { ztest_device_removal_active = B_TRUE; mutex_exit(&ztest_vdev_lock); while (spa->spa_vdev_removal != NULL) txg_wait_synced(spa_get_dsl(spa), 0); } else { mutex_exit(&ztest_vdev_lock); return; } /* * The pool needs to be scrubbed after completing device removal. * Failure to do so may result in checksum errors due to the * strategy employed by ztest_fault_inject() when selecting which * offset are redundant and can be damaged. */ error = spa_scan(spa, POOL_SCAN_SCRUB); if (error == 0) { while (dsl_scan_scrubbing(spa_get_dsl(spa))) txg_wait_synced(spa_get_dsl(spa), 0); } mutex_enter(&ztest_vdev_lock); ztest_device_removal_active = B_FALSE; mutex_exit(&ztest_vdev_lock); } /* * Callback function which expands the physical size of the vdev. */ vdev_t * grow_vdev(vdev_t *vd, void *arg) { ASSERTV(spa_t *spa = vd->vdev_spa); size_t *newsize = arg; size_t fsize; int fd; ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); ASSERT(vd->vdev_ops->vdev_op_leaf); if ((fd = open(vd->vdev_path, O_RDWR)) == -1) return (vd); fsize = lseek(fd, 0, SEEK_END); VERIFY(ftruncate(fd, *newsize) == 0); if (ztest_opts.zo_verbose >= 6) { (void) printf("%s grew from %lu to %lu bytes\n", vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); } (void) close(fd); return (NULL); } /* * Callback function which expands a given vdev by calling vdev_online(). */ /* ARGSUSED */ vdev_t * online_vdev(vdev_t *vd, void *arg) { spa_t *spa = vd->vdev_spa; vdev_t *tvd = vd->vdev_top; uint64_t guid = vd->vdev_guid; uint64_t generation = spa->spa_config_generation + 1; vdev_state_t newstate = VDEV_STATE_UNKNOWN; int error; ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); ASSERT(vd->vdev_ops->vdev_op_leaf); /* Calling vdev_online will initialize the new metaslabs */ spa_config_exit(spa, SCL_STATE, spa); error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); spa_config_enter(spa, SCL_STATE, spa, RW_READER); /* * If vdev_online returned an error or the underlying vdev_open * failed then we abort the expand. The only way to know that * vdev_open fails is by checking the returned newstate. */ if (error || newstate != VDEV_STATE_HEALTHY) { if (ztest_opts.zo_verbose >= 5) { (void) printf("Unable to expand vdev, state %llu, " "error %d\n", (u_longlong_t)newstate, error); } return (vd); } ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); /* * Since we dropped the lock we need to ensure that we're * still talking to the original vdev. It's possible this * vdev may have been detached/replaced while we were * trying to online it. */ if (generation != spa->spa_config_generation) { if (ztest_opts.zo_verbose >= 5) { (void) printf("vdev configuration has changed, " "guid %llu, state %llu, expected gen %llu, " "got gen %llu\n", (u_longlong_t)guid, (u_longlong_t)tvd->vdev_state, (u_longlong_t)generation, (u_longlong_t)spa->spa_config_generation); } return (vd); } return (NULL); } /* * Traverse the vdev tree calling the supplied function. * We continue to walk the tree until we either have walked all * children or we receive a non-NULL return from the callback. * If a NULL callback is passed, then we just return back the first * leaf vdev we encounter. */ vdev_t * vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) { uint_t c; if (vd->vdev_ops->vdev_op_leaf) { if (func == NULL) return (vd); else return (func(vd, arg)); } for (c = 0; c < vd->vdev_children; c++) { vdev_t *cvd = vd->vdev_child[c]; if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) return (cvd); } return (NULL); } /* * Verify that dynamic LUN growth works as expected. */ /* ARGSUSED */ void ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) { spa_t *spa = ztest_spa; vdev_t *vd, *tvd; metaslab_class_t *mc; metaslab_group_t *mg; size_t psize, newsize; uint64_t top; uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; mutex_enter(&ztest_checkpoint_lock); mutex_enter(&ztest_vdev_lock); spa_config_enter(spa, SCL_STATE, spa, RW_READER); /* * If there is a vdev removal in progress, it could complete while * we are running, in which case we would not be able to verify * that the metaslab_class space increased (because it decreases * when the device removal completes). */ if (ztest_device_removal_active) { spa_config_exit(spa, SCL_STATE, spa); mutex_exit(&ztest_vdev_lock); mutex_exit(&ztest_checkpoint_lock); return; } top = ztest_random_vdev_top(spa, B_TRUE); tvd = spa->spa_root_vdev->vdev_child[top]; mg = tvd->vdev_mg; mc = mg->mg_class; old_ms_count = tvd->vdev_ms_count; old_class_space = metaslab_class_get_space(mc); /* * Determine the size of the first leaf vdev associated with * our top-level device. */ vd = vdev_walk_tree(tvd, NULL, NULL); ASSERT3P(vd, !=, NULL); ASSERT(vd->vdev_ops->vdev_op_leaf); psize = vd->vdev_psize; /* * We only try to expand the vdev if it's healthy, less than 4x its * original size, and it has a valid psize. */ if (tvd->vdev_state != VDEV_STATE_HEALTHY || psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { spa_config_exit(spa, SCL_STATE, spa); mutex_exit(&ztest_vdev_lock); mutex_exit(&ztest_checkpoint_lock); return; } ASSERT(psize > 0); newsize = psize + psize / 8; ASSERT3U(newsize, >, psize); if (ztest_opts.zo_verbose >= 6) { (void) printf("Expanding LUN %s from %lu to %lu\n", vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); } /* * Growing the vdev is a two step process: * 1). expand the physical size (i.e. relabel) * 2). online the vdev to create the new metaslabs */ if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || vdev_walk_tree(tvd, online_vdev, NULL) != NULL || tvd->vdev_state != VDEV_STATE_HEALTHY) { if (ztest_opts.zo_verbose >= 5) { (void) printf("Could not expand LUN because " "the vdev configuration changed.\n"); } spa_config_exit(spa, SCL_STATE, spa); mutex_exit(&ztest_vdev_lock); mutex_exit(&ztest_checkpoint_lock); return; } spa_config_exit(spa, SCL_STATE, spa); /* * Expanding the LUN will update the config asynchronously, * thus we must wait for the async thread to complete any * pending tasks before proceeding. */ for (;;) { boolean_t done; mutex_enter(&spa->spa_async_lock); done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); mutex_exit(&spa->spa_async_lock); if (done) break; txg_wait_synced(spa_get_dsl(spa), 0); (void) poll(NULL, 0, 100); } spa_config_enter(spa, SCL_STATE, spa, RW_READER); tvd = spa->spa_root_vdev->vdev_child[top]; new_ms_count = tvd->vdev_ms_count; new_class_space = metaslab_class_get_space(mc); if (tvd->vdev_mg != mg || mg->mg_class != mc) { if (ztest_opts.zo_verbose >= 5) { (void) printf("Could not verify LUN expansion due to " "intervening vdev offline or remove.\n"); } spa_config_exit(spa, SCL_STATE, spa); mutex_exit(&ztest_vdev_lock); mutex_exit(&ztest_checkpoint_lock); return; } /* * Make sure we were able to grow the vdev. */ if (new_ms_count <= old_ms_count) { fatal(0, "LUN expansion failed: ms_count %llu < %llu\n", old_ms_count, new_ms_count); } /* * Make sure we were able to grow the pool. */ if (new_class_space <= old_class_space) { fatal(0, "LUN expansion failed: class_space %llu < %llu\n", old_class_space, new_class_space); } if (ztest_opts.zo_verbose >= 5) { char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ]; nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf)); nicenum(new_class_space, newnumbuf, sizeof (newnumbuf)); (void) printf("%s grew from %s to %s\n", spa->spa_name, oldnumbuf, newnumbuf); } spa_config_exit(spa, SCL_STATE, spa); mutex_exit(&ztest_vdev_lock); mutex_exit(&ztest_checkpoint_lock); } /* * Verify that dmu_objset_{create,destroy,open,close} work as expected. */ /* ARGSUSED */ static void ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) { /* * Create the objects common to all ztest datasets. */ VERIFY(zap_create_claim(os, ZTEST_DIROBJ, DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); } static int ztest_dataset_create(char *dsname) { int err; uint64_t rand; dsl_crypto_params_t *dcp = NULL; /* * 50% of the time, we create encrypted datasets * using a random cipher suite and a hard-coded * wrapping key. */ rand = ztest_random(2); if (rand != 0) { nvlist_t *crypto_args = fnvlist_alloc(); nvlist_t *props = fnvlist_alloc(); /* slight bias towards the default cipher suite */ rand = ztest_random(ZIO_CRYPT_FUNCTIONS); if (rand < ZIO_CRYPT_AES_128_CCM) rand = ZIO_CRYPT_ON; fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand); fnvlist_add_uint8_array(crypto_args, "wkeydata", (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); /* * These parameters aren't really used by the kernel. They * are simply stored so that userspace knows how to load * the wrapping key. */ fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW); fnvlist_add_string(props, zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt"); fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL); fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL); VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props, crypto_args, &dcp)); /* * Cycle through all available encryption implementations * to verify interoperability. */ VERIFY0(gcm_impl_set("cycle")); VERIFY0(aes_impl_set("cycle")); fnvlist_free(crypto_args); fnvlist_free(props); } err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp, ztest_objset_create_cb, NULL); dsl_crypto_params_free(dcp, !!err); rand = ztest_random(100); if (err || rand < 80) return (err); if (ztest_opts.zo_verbose >= 5) (void) printf("Setting dataset %s to sync always\n", dsname); return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, ZFS_SYNC_ALWAYS, B_FALSE)); } /* ARGSUSED */ static int ztest_objset_destroy_cb(const char *name, void *arg) { objset_t *os; dmu_object_info_t doi; int error; /* * Verify that the dataset contains a directory object. */ VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, B_TRUE, FTAG, &os)); error = dmu_object_info(os, ZTEST_DIROBJ, &doi); if (error != ENOENT) { /* We could have crashed in the middle of destroying it */ ASSERT0(error); ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); ASSERT3S(doi.doi_physical_blocks_512, >=, 0); } dmu_objset_disown(os, B_TRUE, FTAG); /* * Destroy the dataset. */ if (strchr(name, '@') != NULL) { VERIFY0(dsl_destroy_snapshot(name, B_TRUE)); } else { error = dsl_destroy_head(name); /* There could be a hold on this dataset */ if (error != EBUSY) ASSERT0(error); } return (0); } static boolean_t ztest_snapshot_create(char *osname, uint64_t id) { char snapname[ZFS_MAX_DATASET_NAME_LEN]; int error; (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); error = dmu_objset_snapshot_one(osname, snapname); if (error == ENOSPC) { ztest_record_enospc(FTAG); return (B_FALSE); } if (error != 0 && error != EEXIST) { fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, snapname, error); } return (B_TRUE); } static boolean_t ztest_snapshot_destroy(char *osname, uint64_t id) { char snapname[ZFS_MAX_DATASET_NAME_LEN]; int error; (void) snprintf(snapname, sizeof (snapname), "%s@%llu", osname, (u_longlong_t)id); error = dsl_destroy_snapshot(snapname, B_FALSE); if (error != 0 && error != ENOENT) fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); return (B_TRUE); } /* ARGSUSED */ void ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) { ztest_ds_t *zdtmp; int iters; int error; objset_t *os, *os2; char name[ZFS_MAX_DATASET_NAME_LEN]; zilog_t *zilog; int i; zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); (void) pthread_rwlock_rdlock(&ztest_name_lock); (void) snprintf(name, sizeof (name), "%s/temp_%llu", ztest_opts.zo_pool, (u_longlong_t)id); /* * If this dataset exists from a previous run, process its replay log * half of the time. If we don't replay it, then dsl_destroy_head() * (invoked from ztest_objset_destroy_cb()) should just throw it away. */ if (ztest_random(2) == 0 && ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, FTAG, &os) == 0) { ztest_zd_init(zdtmp, NULL, os); zil_replay(os, zdtmp, ztest_replay_vector); ztest_zd_fini(zdtmp); txg_wait_synced(dmu_objset_pool(os), 0); dmu_objset_disown(os, B_TRUE, FTAG); } /* * There may be an old instance of the dataset we're about to * create lying around from a previous run. If so, destroy it * and all of its snapshots. */ (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); /* * Verify that the destroyed dataset is no longer in the namespace. */ VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, B_TRUE, FTAG, &os)); /* * Verify that we can create a new dataset. */ error = ztest_dataset_create(name); if (error) { if (error == ENOSPC) { ztest_record_enospc(FTAG); goto out; } fatal(0, "dmu_objset_create(%s) = %d", name, error); } VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, FTAG, &os)); ztest_zd_init(zdtmp, NULL, os); /* * Open the intent log for it. */ zilog = zil_open(os, ztest_get_data); /* * Put some objects in there, do a little I/O to them, * and randomly take a couple of snapshots along the way. */ iters = ztest_random(5); for (i = 0; i < iters; i++) { ztest_dmu_object_alloc_free(zdtmp, id); if (ztest_random(iters) == 0) (void) ztest_snapshot_create(name, i); } /* * Verify that we cannot create an existing dataset. */ VERIFY3U(EEXIST, ==, dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL)); /* * Verify that we can hold an objset that is also owned. */ VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); dmu_objset_rele(os2, FTAG); /* * Verify that we cannot own an objset that is already owned. */ VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, FTAG, &os2)); zil_close(zilog); txg_wait_synced(spa_get_dsl(os->os_spa), 0); dmu_objset_disown(os, B_TRUE, FTAG); ztest_zd_fini(zdtmp); out: (void) pthread_rwlock_unlock(&ztest_name_lock); umem_free(zdtmp, sizeof (ztest_ds_t)); } /* * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. */ void ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) { (void) pthread_rwlock_rdlock(&ztest_name_lock); (void) ztest_snapshot_destroy(zd->zd_name, id); (void) ztest_snapshot_create(zd->zd_name, id); (void) pthread_rwlock_unlock(&ztest_name_lock); } /* * Cleanup non-standard snapshots and clones. */ void ztest_dsl_dataset_cleanup(char *osname, uint64_t id) { char *snap1name; char *clone1name; char *snap2name; char *clone2name; char *snap3name; int error; snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%llu", osname, (u_longlong_t)id); (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%llu", osname, (u_longlong_t)id); (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%llu", clone1name, (u_longlong_t)id); (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%llu", osname, (u_longlong_t)id); (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%llu", clone1name, (u_longlong_t)id); error = dsl_destroy_head(clone2name); if (error && error != ENOENT) fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); error = dsl_destroy_snapshot(snap3name, B_FALSE); if (error && error != ENOENT) fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); error = dsl_destroy_snapshot(snap2name, B_FALSE); if (error && error != ENOENT) fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); error = dsl_destroy_head(clone1name); if (error && error != ENOENT) fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); error = dsl_destroy_snapshot(snap1name, B_FALSE); if (error && error != ENOENT) fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); } /* * Verify dsl_dataset_promote handles EBUSY */ void ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) { objset_t *os; char *snap1name; char *clone1name; char *snap2name; char *clone2name; char *snap3name; char *osname = zd->zd_name; int error; snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); (void) pthread_rwlock_rdlock(&ztest_name_lock); ztest_dsl_dataset_cleanup(osname, id); (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, "%s@s1_%llu", osname, (u_longlong_t)id); (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, "%s/c1_%llu", osname, (u_longlong_t)id); (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, "%s@s2_%llu", clone1name, (u_longlong_t)id); (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, "%s/c2_%llu", osname, (u_longlong_t)id); (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, "%s@s3_%llu", clone1name, (u_longlong_t)id); error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); if (error && error != EEXIST) { if (error == ENOSPC) { ztest_record_enospc(FTAG); goto out; } fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); } error = dmu_objset_clone(clone1name, snap1name); if (error) { if (error == ENOSPC) { ztest_record_enospc(FTAG); goto out; } fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); } error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); if (error && error != EEXIST) { if (error == ENOSPC) { ztest_record_enospc(FTAG); goto out; } fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); } error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); if (error && error != EEXIST) { if (error == ENOSPC) { ztest_record_enospc(FTAG); goto out; } fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); } error = dmu_objset_clone(clone2name, snap3name); if (error) { if (error == ENOSPC) { ztest_record_enospc(FTAG); goto out; } fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); } error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os); if (error) fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); error = dsl_dataset_promote(clone2name, NULL); if (error == ENOSPC) { dmu_objset_disown(os, B_TRUE, FTAG); ztest_record_enospc(FTAG); goto out; } if (error != EBUSY) fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, error); dmu_objset_disown(os, B_TRUE, FTAG); out: ztest_dsl_dataset_cleanup(osname, id); (void) pthread_rwlock_unlock(&ztest_name_lock); umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); } #undef OD_ARRAY_SIZE #define OD_ARRAY_SIZE 4 /* * Verify that dmu_object_{alloc,free} work as expected. */ void ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) { ztest_od_t *od; int batchsize; int size; int b; size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; od = umem_alloc(size, UMEM_NOFAIL); batchsize = OD_ARRAY_SIZE; for (b = 0; b < batchsize; b++) ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, 0, 0, 0); /* * Destroy the previous batch of objects, create a new batch, * and do some I/O on the new objects. */ if (ztest_object_init(zd, od, size, B_TRUE) != 0) return; while (ztest_random(4 * batchsize) != 0) ztest_io(zd, od[ztest_random(batchsize)].od_object, ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); umem_free(od, size); } /* * Rewind the global allocator to verify object allocation backfilling. */ void ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; uint64_t object; /* * Rewind the global allocator randomly back to a lower object number * to force backfilling and reclamation of recently freed dnodes. */ mutex_enter(&os->os_obj_lock); object = ztest_random(os->os_obj_next_chunk); os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk); mutex_exit(&os->os_obj_lock); } #undef OD_ARRAY_SIZE #define OD_ARRAY_SIZE 2 /* * Verify that dmu_{read,write} work as expected. */ void ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) { int size; ztest_od_t *od; objset_t *os = zd->zd_os; size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; od = umem_alloc(size, UMEM_NOFAIL); dmu_tx_t *tx; int i, freeit, error; uint64_t n, s, txg; bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); uint64_t regions = 997; uint64_t stride = 123456789ULL; uint64_t width = 40; int free_percent = 5; /* * This test uses two objects, packobj and bigobj, that are always * updated together (i.e. in the same tx) so that their contents are * in sync and can be compared. Their contents relate to each other * in a simple way: packobj is a dense array of 'bufwad' structures, * while bigobj is a sparse array of the same bufwads. Specifically, * for any index n, there are three bufwads that should be identical: * * packobj, at offset n * sizeof (bufwad_t) * bigobj, at the head of the nth chunk * bigobj, at the tail of the nth chunk * * The chunk size is arbitrary. It doesn't have to be a power of two, * and it doesn't have any relation to the object blocksize. * The only requirement is that it can hold at least two bufwads. * * Normally, we write the bufwad to each of these locations. * However, free_percent of the time we instead write zeroes to * packobj and perform a dmu_free_range() on bigobj. By comparing * bigobj to packobj, we can verify that the DMU is correctly * tracking which parts of an object are allocated and free, * and that the contents of the allocated blocks are correct. */ /* * Read the directory info. If it's the first time, set things up. */ ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize); ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, chunksize); if (ztest_object_init(zd, od, size, B_FALSE) != 0) { umem_free(od, size); return; } bigobj = od[0].od_object; packobj = od[1].od_object; chunksize = od[0].od_gen; ASSERT(chunksize == od[1].od_gen); /* * Prefetch a random chunk of the big object. * Our aim here is to get some async reads in flight * for blocks that we may free below; the DMU should * handle this race correctly. */ n = ztest_random(regions) * stride + ztest_random(width); s = 1 + ztest_random(2 * width - 1); dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize, ZIO_PRIORITY_SYNC_READ); /* * Pick a random index and compute the offsets into packobj and bigobj. */ n = ztest_random(regions) * stride + ztest_random(width); s = 1 + ztest_random(width - 1); packoff = n * sizeof (bufwad_t); packsize = s * sizeof (bufwad_t); bigoff = n * chunksize; bigsize = s * chunksize; packbuf = umem_alloc(packsize, UMEM_NOFAIL); bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); /* * free_percent of the time, free a range of bigobj rather than * overwriting it. */ freeit = (ztest_random(100) < free_percent); /* * Read the current contents of our objects. */ error = dmu_read(os, packobj, packoff, packsize, packbuf, DMU_READ_PREFETCH); ASSERT0(error); error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, DMU_READ_PREFETCH); ASSERT0(error); /* * Get a tx for the mods to both packobj and bigobj. */ tx = dmu_tx_create(os); dmu_tx_hold_write(tx, packobj, packoff, packsize); if (freeit) dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); else dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); /* This accounts for setting the checksum/compression. */ dmu_tx_hold_bonus(tx, bigobj); txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); if (txg == 0) { umem_free(packbuf, packsize); umem_free(bigbuf, bigsize); umem_free(od, size); return; } enum zio_checksum cksum; do { cksum = (enum zio_checksum) ztest_random_dsl_prop(ZFS_PROP_CHECKSUM); } while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS); dmu_object_set_checksum(os, bigobj, cksum, tx); enum zio_compress comp; do { comp = (enum zio_compress) ztest_random_dsl_prop(ZFS_PROP_COMPRESSION); } while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS); dmu_object_set_compress(os, bigobj, comp, tx); /* * For each index from n to n + s, verify that the existing bufwad * in packobj matches the bufwads at the head and tail of the * corresponding chunk in bigobj. Then update all three bufwads * with the new values we want to write out. */ for (i = 0; i < s; i++) { /* LINTED */ pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); /* LINTED */ bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); /* LINTED */ bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); if (pack->bw_txg > txg) fatal(0, "future leak: got %llx, open txg is %llx", pack->bw_txg, txg); if (pack->bw_data != 0 && pack->bw_index != n + i) fatal(0, "wrong index: got %llx, wanted %llx+%llx", pack->bw_index, n, i); if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); if (freeit) { bzero(pack, sizeof (bufwad_t)); } else { pack->bw_index = n + i; pack->bw_txg = txg; pack->bw_data = 1 + ztest_random(-2ULL); } *bigH = *pack; *bigT = *pack; } /* * We've verified all the old bufwads, and made new ones. * Now write them out. */ dmu_write(os, packobj, packoff, packsize, packbuf, tx); if (freeit) { if (ztest_opts.zo_verbose >= 7) { (void) printf("freeing offset %llx size %llx" " txg %llx\n", (u_longlong_t)bigoff, (u_longlong_t)bigsize, (u_longlong_t)txg); } VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); } else { if (ztest_opts.zo_verbose >= 7) { (void) printf("writing offset %llx size %llx" " txg %llx\n", (u_longlong_t)bigoff, (u_longlong_t)bigsize, (u_longlong_t)txg); } dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); } dmu_tx_commit(tx); /* * Sanity check the stuff we just wrote. */ { void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); VERIFY(0 == dmu_read(os, packobj, packoff, packsize, packcheck, DMU_READ_PREFETCH)); VERIFY(0 == dmu_read(os, bigobj, bigoff, bigsize, bigcheck, DMU_READ_PREFETCH)); ASSERT(bcmp(packbuf, packcheck, packsize) == 0); ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); umem_free(packcheck, packsize); umem_free(bigcheck, bigsize); } umem_free(packbuf, packsize); umem_free(bigbuf, bigsize); umem_free(od, size); } void compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) { uint64_t i; bufwad_t *pack; bufwad_t *bigH; bufwad_t *bigT; /* * For each index from n to n + s, verify that the existing bufwad * in packobj matches the bufwads at the head and tail of the * corresponding chunk in bigobj. Then update all three bufwads * with the new values we want to write out. */ for (i = 0; i < s; i++) { /* LINTED */ pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); /* LINTED */ bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); /* LINTED */ bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); if (pack->bw_txg > txg) fatal(0, "future leak: got %llx, open txg is %llx", pack->bw_txg, txg); if (pack->bw_data != 0 && pack->bw_index != n + i) fatal(0, "wrong index: got %llx, wanted %llx+%llx", pack->bw_index, n, i); if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); pack->bw_index = n + i; pack->bw_txg = txg; pack->bw_data = 1 + ztest_random(-2ULL); *bigH = *pack; *bigT = *pack; } } #undef OD_ARRAY_SIZE #define OD_ARRAY_SIZE 2 void ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; ztest_od_t *od; dmu_tx_t *tx; uint64_t i; int error; int size; uint64_t n, s, txg; bufwad_t *packbuf, *bigbuf; uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; uint64_t blocksize = ztest_random_blocksize(); uint64_t chunksize = blocksize; uint64_t regions = 997; uint64_t stride = 123456789ULL; uint64_t width = 9; dmu_buf_t *bonus_db; arc_buf_t **bigbuf_arcbufs; dmu_object_info_t doi; size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; od = umem_alloc(size, UMEM_NOFAIL); /* * This test uses two objects, packobj and bigobj, that are always * updated together (i.e. in the same tx) so that their contents are * in sync and can be compared. Their contents relate to each other * in a simple way: packobj is a dense array of 'bufwad' structures, * while bigobj is a sparse array of the same bufwads. Specifically, * for any index n, there are three bufwads that should be identical: * * packobj, at offset n * sizeof (bufwad_t) * bigobj, at the head of the nth chunk * bigobj, at the tail of the nth chunk * * The chunk size is set equal to bigobj block size so that * dmu_assign_arcbuf_by_dbuf() can be tested for object updates. */ /* * Read the directory info. If it's the first time, set things up. */ ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, chunksize); if (ztest_object_init(zd, od, size, B_FALSE) != 0) { umem_free(od, size); return; } bigobj = od[0].od_object; packobj = od[1].od_object; blocksize = od[0].od_blocksize; chunksize = blocksize; ASSERT(chunksize == od[1].od_gen); VERIFY(dmu_object_info(os, bigobj, &doi) == 0); VERIFY(ISP2(doi.doi_data_block_size)); VERIFY(chunksize == doi.doi_data_block_size); VERIFY(chunksize >= 2 * sizeof (bufwad_t)); /* * Pick a random index and compute the offsets into packobj and bigobj. */ n = ztest_random(regions) * stride + ztest_random(width); s = 1 + ztest_random(width - 1); packoff = n * sizeof (bufwad_t); packsize = s * sizeof (bufwad_t); bigoff = n * chunksize; bigsize = s * chunksize; packbuf = umem_zalloc(packsize, UMEM_NOFAIL); bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); /* * Iteration 0 test zcopy for DB_UNCACHED dbufs. * Iteration 1 test zcopy to already referenced dbufs. * Iteration 2 test zcopy to dirty dbuf in the same txg. * Iteration 3 test zcopy to dbuf dirty in previous txg. * Iteration 4 test zcopy when dbuf is no longer dirty. * Iteration 5 test zcopy when it can't be done. * Iteration 6 one more zcopy write. */ for (i = 0; i < 7; i++) { uint64_t j; uint64_t off; /* * In iteration 5 (i == 5) use arcbufs * that don't match bigobj blksz to test * dmu_assign_arcbuf_by_dbuf() when it can't directly * assign an arcbuf to a dbuf. */ for (j = 0; j < s; j++) { if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { bigbuf_arcbufs[j] = dmu_request_arcbuf(bonus_db, chunksize); } else { bigbuf_arcbufs[2 * j] = dmu_request_arcbuf(bonus_db, chunksize / 2); bigbuf_arcbufs[2 * j + 1] = dmu_request_arcbuf(bonus_db, chunksize / 2); } } /* * Get a tx for the mods to both packobj and bigobj. */ tx = dmu_tx_create(os); dmu_tx_hold_write(tx, packobj, packoff, packsize); dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); if (txg == 0) { umem_free(packbuf, packsize); umem_free(bigbuf, bigsize); for (j = 0; j < s; j++) { if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { dmu_return_arcbuf(bigbuf_arcbufs[j]); } else { dmu_return_arcbuf( bigbuf_arcbufs[2 * j]); dmu_return_arcbuf( bigbuf_arcbufs[2 * j + 1]); } } umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); umem_free(od, size); dmu_buf_rele(bonus_db, FTAG); return; } /* * 50% of the time don't read objects in the 1st iteration to * test dmu_assign_arcbuf_by_dbuf() for the case when there are * no existing dbufs for the specified offsets. */ if (i != 0 || ztest_random(2) != 0) { error = dmu_read(os, packobj, packoff, packsize, packbuf, DMU_READ_PREFETCH); ASSERT0(error); error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, DMU_READ_PREFETCH); ASSERT0(error); } compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, n, chunksize, txg); /* * We've verified all the old bufwads, and made new ones. * Now write them out. */ dmu_write(os, packobj, packoff, packsize, packbuf, tx); if (ztest_opts.zo_verbose >= 7) { (void) printf("writing offset %llx size %llx" " txg %llx\n", (u_longlong_t)bigoff, (u_longlong_t)bigsize, (u_longlong_t)txg); } for (off = bigoff, j = 0; j < s; j++, off += chunksize) { dmu_buf_t *dbt; if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { bcopy((caddr_t)bigbuf + (off - bigoff), bigbuf_arcbufs[j]->b_data, chunksize); } else { bcopy((caddr_t)bigbuf + (off - bigoff), bigbuf_arcbufs[2 * j]->b_data, chunksize / 2); bcopy((caddr_t)bigbuf + (off - bigoff) + chunksize / 2, bigbuf_arcbufs[2 * j + 1]->b_data, chunksize / 2); } if (i == 1) { VERIFY(dmu_buf_hold(os, bigobj, off, FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); } if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { dmu_assign_arcbuf_by_dbuf(bonus_db, off, bigbuf_arcbufs[j], tx); } else { dmu_assign_arcbuf_by_dbuf(bonus_db, off, bigbuf_arcbufs[2 * j], tx); dmu_assign_arcbuf_by_dbuf(bonus_db, off + chunksize / 2, bigbuf_arcbufs[2 * j + 1], tx); } if (i == 1) { dmu_buf_rele(dbt, FTAG); } } dmu_tx_commit(tx); /* * Sanity check the stuff we just wrote. */ { void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); VERIFY(0 == dmu_read(os, packobj, packoff, packsize, packcheck, DMU_READ_PREFETCH)); VERIFY(0 == dmu_read(os, bigobj, bigoff, bigsize, bigcheck, DMU_READ_PREFETCH)); ASSERT(bcmp(packbuf, packcheck, packsize) == 0); ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); umem_free(packcheck, packsize); umem_free(bigcheck, bigsize); } if (i == 2) { txg_wait_open(dmu_objset_pool(os), 0); } else if (i == 3) { txg_wait_synced(dmu_objset_pool(os), 0); } } dmu_buf_rele(bonus_db, FTAG); umem_free(packbuf, packsize); umem_free(bigbuf, bigsize); umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); umem_free(od, size); } /* ARGSUSED */ void ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) { ztest_od_t *od; od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); uint64_t offset = (1ULL << (ztest_random(20) + 43)) + (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); /* * Have multiple threads write to large offsets in an object * to verify that parallel writes to an object -- even to the * same blocks within the object -- doesn't cause any trouble. */ ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) return; while (ztest_random(10) != 0) ztest_io(zd, od->od_object, offset); umem_free(od, sizeof (ztest_od_t)); } void ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) { ztest_od_t *od; uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); uint64_t count = ztest_random(20) + 1; uint64_t blocksize = ztest_random_blocksize(); void *data; od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) { umem_free(od, sizeof (ztest_od_t)); return; } if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { umem_free(od, sizeof (ztest_od_t)); return; } ztest_prealloc(zd, od->od_object, offset, count * blocksize); data = umem_zalloc(blocksize, UMEM_NOFAIL); while (ztest_random(count) != 0) { uint64_t randoff = offset + (ztest_random(count) * blocksize); if (ztest_write(zd, od->od_object, randoff, blocksize, data) != 0) break; while (ztest_random(4) != 0) ztest_io(zd, od->od_object, randoff); } umem_free(data, blocksize); umem_free(od, sizeof (ztest_od_t)); } /* * Verify that zap_{create,destroy,add,remove,update} work as expected. */ #define ZTEST_ZAP_MIN_INTS 1 #define ZTEST_ZAP_MAX_INTS 4 #define ZTEST_ZAP_MAX_PROPS 1000 void ztest_zap(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; ztest_od_t *od; uint64_t object; uint64_t txg, last_txg; uint64_t value[ZTEST_ZAP_MAX_INTS]; uint64_t zl_ints, zl_intsize, prop; int i, ints; dmu_tx_t *tx; char propname[100], txgname[100]; int error; char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) goto out; object = od->od_object; /* * Generate a known hash collision, and verify that * we can lookup and remove both entries. */ tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, object, B_TRUE, NULL); txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); if (txg == 0) goto out; for (i = 0; i < 2; i++) { value[i] = i; VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), 1, &value[i], tx)); } for (i = 0; i < 2; i++) { VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], sizeof (uint64_t), 1, &value[i], tx)); VERIFY3U(0, ==, zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); ASSERT3U(zl_ints, ==, 1); } for (i = 0; i < 2; i++) { VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); } dmu_tx_commit(tx); /* * Generate a buch of random entries. */ ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); prop = ztest_random(ZTEST_ZAP_MAX_PROPS); (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); bzero(value, sizeof (value)); last_txg = 0; /* * If these zap entries already exist, validate their contents. */ error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); if (error == 0) { ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); ASSERT3U(zl_ints, ==, 1); VERIFY(zap_lookup(os, object, txgname, zl_intsize, zl_ints, &last_txg) == 0); VERIFY(zap_length(os, object, propname, &zl_intsize, &zl_ints) == 0); ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); ASSERT3U(zl_ints, ==, ints); VERIFY(zap_lookup(os, object, propname, zl_intsize, zl_ints, value) == 0); for (i = 0; i < ints; i++) { ASSERT3U(value[i], ==, last_txg + object + i); } } else { ASSERT3U(error, ==, ENOENT); } /* * Atomically update two entries in our zap object. * The first is named txg_%llu, and contains the txg * in which the property was last updated. The second * is named prop_%llu, and the nth element of its value * should be txg + object + n. */ tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, object, B_TRUE, NULL); txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); if (txg == 0) goto out; if (last_txg > txg) fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); for (i = 0; i < ints; i++) value[i] = txg + object + i; VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), 1, &txg, tx)); VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), ints, value, tx)); dmu_tx_commit(tx); /* * Remove a random pair of entries. */ prop = ztest_random(ZTEST_ZAP_MAX_PROPS); (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); if (error == ENOENT) goto out; ASSERT0(error); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, object, B_TRUE, NULL); txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); if (txg == 0) goto out; VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); dmu_tx_commit(tx); out: umem_free(od, sizeof (ztest_od_t)); } /* * Testcase to test the upgrading of a microzap to fatzap. */ void ztest_fzap(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; ztest_od_t *od; uint64_t object, txg; int i; od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) goto out; object = od->od_object; /* * Add entries to this ZAP and make sure it spills over * and gets upgraded to a fatzap. Also, since we are adding * 2050 entries we should see ptrtbl growth and leaf-block split. */ for (i = 0; i < 2050; i++) { char name[ZFS_MAX_DATASET_NAME_LEN]; uint64_t value = i; dmu_tx_t *tx; int error; (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", (u_longlong_t)id, (u_longlong_t)value); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, object, B_TRUE, name); txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); if (txg == 0) goto out; error = zap_add(os, object, name, sizeof (uint64_t), 1, &value, tx); ASSERT(error == 0 || error == EEXIST); dmu_tx_commit(tx); } out: umem_free(od, sizeof (ztest_od_t)); } /* ARGSUSED */ void ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; ztest_od_t *od; uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; dmu_tx_t *tx; int i, namelen, error; int micro = ztest_random(2); char name[20], string_value[20]; void *data; od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0); if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { umem_free(od, sizeof (ztest_od_t)); return; } object = od->od_object; /* * Generate a random name of the form 'xxx.....' where each * x is a random printable character and the dots are dots. * There are 94 such characters, and the name length goes from * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. */ namelen = ztest_random(sizeof (name) - 5) + 5 + 1; for (i = 0; i < 3; i++) name[i] = '!' + ztest_random('~' - '!' + 1); for (; i < namelen - 1; i++) name[i] = '.'; name[i] = '\0'; if ((namelen & 1) || micro) { wsize = sizeof (txg); wc = 1; data = &txg; } else { wsize = 1; wc = namelen; data = string_value; } count = -1ULL; VERIFY0(zap_count(os, object, &count)); ASSERT(count != -1ULL); /* * Select an operation: length, lookup, add, update, remove. */ i = ztest_random(5); if (i >= 2) { tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, object, B_TRUE, NULL); txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); if (txg == 0) { umem_free(od, sizeof (ztest_od_t)); return; } bcopy(name, string_value, namelen); } else { tx = NULL; txg = 0; bzero(string_value, namelen); } switch (i) { case 0: error = zap_length(os, object, name, &zl_wsize, &zl_wc); if (error == 0) { ASSERT3U(wsize, ==, zl_wsize); ASSERT3U(wc, ==, zl_wc); } else { ASSERT3U(error, ==, ENOENT); } break; case 1: error = zap_lookup(os, object, name, wsize, wc, data); if (error == 0) { if (data == string_value && bcmp(name, data, namelen) != 0) fatal(0, "name '%s' != val '%s' len %d", name, data, namelen); } else { ASSERT3U(error, ==, ENOENT); } break; case 2: error = zap_add(os, object, name, wsize, wc, data, tx); ASSERT(error == 0 || error == EEXIST); break; case 3: VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); break; case 4: error = zap_remove(os, object, name, tx); ASSERT(error == 0 || error == ENOENT); break; } if (tx != NULL) dmu_tx_commit(tx); umem_free(od, sizeof (ztest_od_t)); } /* * Commit callback data. */ typedef struct ztest_cb_data { list_node_t zcd_node; uint64_t zcd_txg; int zcd_expected_err; boolean_t zcd_added; boolean_t zcd_called; spa_t *zcd_spa; } ztest_cb_data_t; /* This is the actual commit callback function */ static void ztest_commit_callback(void *arg, int error) { ztest_cb_data_t *data = arg; uint64_t synced_txg; VERIFY(data != NULL); VERIFY3S(data->zcd_expected_err, ==, error); VERIFY(!data->zcd_called); synced_txg = spa_last_synced_txg(data->zcd_spa); if (data->zcd_txg > synced_txg) fatal(0, "commit callback of txg %" PRIu64 " called prematurely" ", last synced txg = %" PRIu64 "\n", data->zcd_txg, synced_txg); data->zcd_called = B_TRUE; if (error == ECANCELED) { ASSERT0(data->zcd_txg); ASSERT(!data->zcd_added); /* * The private callback data should be destroyed here, but * since we are going to check the zcd_called field after * dmu_tx_abort(), we will destroy it there. */ return; } ASSERT(data->zcd_added); ASSERT3U(data->zcd_txg, !=, 0); (void) mutex_enter(&zcl.zcl_callbacks_lock); /* See if this cb was called more quickly */ if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) zc_min_txg_delay = synced_txg - data->zcd_txg; /* Remove our callback from the list */ list_remove(&zcl.zcl_callbacks, data); (void) mutex_exit(&zcl.zcl_callbacks_lock); umem_free(data, sizeof (ztest_cb_data_t)); } /* Allocate and initialize callback data structure */ static ztest_cb_data_t * ztest_create_cb_data(objset_t *os, uint64_t txg) { ztest_cb_data_t *cb_data; cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); cb_data->zcd_txg = txg; cb_data->zcd_spa = dmu_objset_spa(os); list_link_init(&cb_data->zcd_node); return (cb_data); } /* * Commit callback test. */ void ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; ztest_od_t *od; dmu_tx_t *tx; ztest_cb_data_t *cb_data[3], *tmp_cb; uint64_t old_txg, txg; int i, error = 0; od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { umem_free(od, sizeof (ztest_od_t)); return; } tx = dmu_tx_create(os); cb_data[0] = ztest_create_cb_data(os, 0); dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); /* Every once in a while, abort the transaction on purpose */ if (ztest_random(100) == 0) error = -1; if (!error) error = dmu_tx_assign(tx, TXG_NOWAIT); txg = error ? 0 : dmu_tx_get_txg(tx); cb_data[0]->zcd_txg = txg; cb_data[1] = ztest_create_cb_data(os, txg); dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); if (error) { /* * It's not a strict requirement to call the registered * callbacks from inside dmu_tx_abort(), but that's what * it's supposed to happen in the current implementation * so we will check for that. */ for (i = 0; i < 2; i++) { cb_data[i]->zcd_expected_err = ECANCELED; VERIFY(!cb_data[i]->zcd_called); } dmu_tx_abort(tx); for (i = 0; i < 2; i++) { VERIFY(cb_data[i]->zcd_called); umem_free(cb_data[i], sizeof (ztest_cb_data_t)); } umem_free(od, sizeof (ztest_od_t)); return; } cb_data[2] = ztest_create_cb_data(os, txg); dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); /* * Read existing data to make sure there isn't a future leak. */ VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), &old_txg, DMU_READ_PREFETCH)); if (old_txg > txg) fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, old_txg, txg); dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); (void) mutex_enter(&zcl.zcl_callbacks_lock); /* * Since commit callbacks don't have any ordering requirement and since * it is theoretically possible for a commit callback to be called * after an arbitrary amount of time has elapsed since its txg has been * synced, it is difficult to reliably determine whether a commit * callback hasn't been called due to high load or due to a flawed * implementation. * * In practice, we will assume that if after a certain number of txgs a * commit callback hasn't been called, then most likely there's an * implementation bug.. */ tmp_cb = list_head(&zcl.zcl_callbacks); if (tmp_cb != NULL && tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { fatal(0, "Commit callback threshold exceeded, oldest txg: %" PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); } /* * Let's find the place to insert our callbacks. * * Even though the list is ordered by txg, it is possible for the * insertion point to not be the end because our txg may already be * quiescing at this point and other callbacks in the open txg * (from other objsets) may have sneaked in. */ tmp_cb = list_tail(&zcl.zcl_callbacks); while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); /* Add the 3 callbacks to the list */ for (i = 0; i < 3; i++) { if (tmp_cb == NULL) list_insert_head(&zcl.zcl_callbacks, cb_data[i]); else list_insert_after(&zcl.zcl_callbacks, tmp_cb, cb_data[i]); cb_data[i]->zcd_added = B_TRUE; VERIFY(!cb_data[i]->zcd_called); tmp_cb = cb_data[i]; } zc_cb_counter += 3; (void) mutex_exit(&zcl.zcl_callbacks_lock); dmu_tx_commit(tx); umem_free(od, sizeof (ztest_od_t)); } /* * Visit each object in the dataset. Verify that its properties * are consistent what was stored in the block tag when it was created, * and that its unused bonus buffer space has not been overwritten. */ /* ARGSUSED */ void ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id) { objset_t *os = zd->zd_os; uint64_t obj; int err = 0; for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { ztest_block_tag_t *bt = NULL; dmu_object_info_t doi; dmu_buf_t *db; ztest_object_lock(zd, obj, RL_READER); if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) { ztest_object_unlock(zd, obj); continue; } dmu_object_info_from_db(db, &doi); if (doi.doi_bonus_size >= sizeof (*bt)) bt = ztest_bt_bonus(db); if (bt && bt->bt_magic == BT_MAGIC) { ztest_bt_verify(bt, os, obj, doi.doi_dnodesize, bt->bt_offset, bt->bt_gen, bt->bt_txg, bt->bt_crtxg); ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen); } dmu_buf_rele(db, FTAG); ztest_object_unlock(zd, obj); } } /* ARGSUSED */ void ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) { zfs_prop_t proplist[] = { ZFS_PROP_CHECKSUM, ZFS_PROP_COMPRESSION, ZFS_PROP_COPIES, ZFS_PROP_DEDUP }; int p; (void) pthread_rwlock_rdlock(&ztest_name_lock); for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE, ztest_random_blocksize(), (int)ztest_random(2))); (void) pthread_rwlock_unlock(&ztest_name_lock); } /* ARGSUSED */ void ztest_remap_blocks(ztest_ds_t *zd, uint64_t id) { (void) pthread_rwlock_rdlock(&ztest_name_lock); int error = dmu_objset_remap_indirects(zd->zd_name); if (error == ENOSPC) error = 0; ASSERT0(error); (void) pthread_rwlock_unlock(&ztest_name_lock); } /* ARGSUSED */ void ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) { nvlist_t *props = NULL; (void) pthread_rwlock_rdlock(&ztest_name_lock); (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO, ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); VERIFY0(spa_prop_get(ztest_spa, &props)); if (ztest_opts.zo_verbose >= 6) dump_nvlist(props, 4); nvlist_free(props); (void) pthread_rwlock_unlock(&ztest_name_lock); } static int user_release_one(const char *snapname, const char *holdname) { nvlist_t *snaps, *holds; int error; snaps = fnvlist_alloc(); holds = fnvlist_alloc(); fnvlist_add_boolean(holds, holdname); fnvlist_add_nvlist(snaps, snapname, holds); fnvlist_free(holds); error = dsl_dataset_user_release(snaps, NULL); fnvlist_free(snaps); return (error); } /* * Test snapshot hold/release and deferred destroy. */ void ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) { int error; objset_t *os = zd->zd_os; objset_t *origin; char snapname[100]; char fullname[100]; char clonename[100]; char tag[100]; char osname[ZFS_MAX_DATASET_NAME_LEN]; nvlist_t *holds; (void) pthread_rwlock_rdlock(&ztest_name_lock); dmu_objset_name(os, osname); (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", (u_longlong_t)id); (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); (void) snprintf(clonename, sizeof (clonename), "%s/ch1_%llu", osname, (u_longlong_t)id); (void) snprintf(tag, sizeof (tag), "tag_%llu", (u_longlong_t)id); /* * Clean up from any previous run. */ error = dsl_destroy_head(clonename); if (error != ENOENT) ASSERT0(error); error = user_release_one(fullname, tag); if (error != ESRCH && error != ENOENT) ASSERT0(error); error = dsl_destroy_snapshot(fullname, B_FALSE); if (error != ENOENT) ASSERT0(error); /* * Create snapshot, clone it, mark snap for deferred destroy, * destroy clone, verify snap was also destroyed. */ error = dmu_objset_snapshot_one(osname, snapname); if (error) { if (error == ENOSPC) { ztest_record_enospc("dmu_objset_snapshot"); goto out; } fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); } error = dmu_objset_clone(clonename, fullname); if (error) { if (error == ENOSPC) { ztest_record_enospc("dmu_objset_clone"); goto out; } fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); } error = dsl_destroy_snapshot(fullname, B_TRUE); if (error) { fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", fullname, error); } error = dsl_destroy_head(clonename); if (error) fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); error = dmu_objset_hold(fullname, FTAG, &origin); if (error != ENOENT) fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); /* * Create snapshot, add temporary hold, verify that we can't * destroy a held snapshot, mark for deferred destroy, * release hold, verify snapshot was destroyed. */ error = dmu_objset_snapshot_one(osname, snapname); if (error) { if (error == ENOSPC) { ztest_record_enospc("dmu_objset_snapshot"); goto out; } fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); } holds = fnvlist_alloc(); fnvlist_add_string(holds, fullname, tag); error = dsl_dataset_user_hold(holds, 0, NULL); fnvlist_free(holds); if (error == ENOSPC) { ztest_record_enospc("dsl_dataset_user_hold"); goto out; } else if (error) { fatal(0, "dsl_dataset_user_hold(%s, %s) = %u", fullname, tag, error); } error = dsl_destroy_snapshot(fullname, B_FALSE); if (error != EBUSY) { fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", fullname, error); } error = dsl_destroy_snapshot(fullname, B_TRUE); if (error) { fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", fullname, error); } error = user_release_one(fullname, tag); if (error) fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); out: (void) pthread_rwlock_unlock(&ztest_name_lock); } /* * Inject random faults into the on-disk data. */ /* ARGSUSED */ void ztest_fault_inject(ztest_ds_t *zd, uint64_t id) { ztest_shared_t *zs = ztest_shared; spa_t *spa = ztest_spa; int fd; uint64_t offset; uint64_t leaves; uint64_t bad = 0x1990c0ffeedecadeull; uint64_t top, leaf; char *path0; char *pathrand; size_t fsize; int bshift = SPA_MAXBLOCKSHIFT + 2; int iters = 1000; int maxfaults; int mirror_save; vdev_t *vd0 = NULL; uint64_t guid0 = 0; boolean_t islog = B_FALSE; path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); mutex_enter(&ztest_vdev_lock); /* * Device removal is in progress, fault injection must be disabled * until it completes and the pool is scrubbed. The fault injection * strategy for damaging blocks does not take in to account evacuated * blocks which may have already been damaged. */ if (ztest_device_removal_active) { mutex_exit(&ztest_vdev_lock); goto out; } maxfaults = MAXFAULTS(zs); leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; mirror_save = zs->zs_mirrors; mutex_exit(&ztest_vdev_lock); ASSERT(leaves >= 1); /* * Grab the name lock as reader. There are some operations * which don't like to have their vdevs changed while * they are in progress (i.e. spa_change_guid). Those * operations will have grabbed the name lock as writer. */ (void) pthread_rwlock_rdlock(&ztest_name_lock); /* * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. */ spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); if (ztest_random(2) == 0) { /* * Inject errors on a normal data device or slog device. */ top = ztest_random_vdev_top(spa, B_TRUE); leaf = ztest_random(leaves) + zs->zs_splits; /* * Generate paths to the first leaf in this top-level vdev, * and to the random leaf we selected. We'll induce transient * write failures and random online/offline activity on leaf 0, * and we'll write random garbage to the randomly chosen leaf. */ (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, ztest_opts.zo_dir, ztest_opts.zo_pool, top * leaves + zs->zs_splits); (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, ztest_opts.zo_dir, ztest_opts.zo_pool, top * leaves + leaf); vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); if (vd0 != NULL && vd0->vdev_top->vdev_islog) islog = B_TRUE; /* * If the top-level vdev needs to be resilvered * then we only allow faults on the device that is * resilvering. */ if (vd0 != NULL && maxfaults != 1 && (!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) || vd0->vdev_resilver_txg != 0)) { /* * Make vd0 explicitly claim to be unreadable, * or unwriteable, or reach behind its back * and close the underlying fd. We can do this if * maxfaults == 0 because we'll fail and reexecute, * and we can do it if maxfaults >= 2 because we'll * have enough redundancy. If maxfaults == 1, the * combination of this with injection of random data * corruption below exceeds the pool's fault tolerance. */ vdev_file_t *vf = vd0->vdev_tsd; zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d", (long long)vd0->vdev_id, (int)maxfaults); if (vf != NULL && ztest_random(3) == 0) { (void) close(vf->vf_vnode->v_fd); vf->vf_vnode->v_fd = -1; } else if (ztest_random(2) == 0) { vd0->vdev_cant_read = B_TRUE; } else { vd0->vdev_cant_write = B_TRUE; } guid0 = vd0->vdev_guid; } } else { /* * Inject errors on an l2cache device. */ spa_aux_vdev_t *sav = &spa->spa_l2cache; if (sav->sav_count == 0) { spa_config_exit(spa, SCL_STATE, FTAG); (void) pthread_rwlock_unlock(&ztest_name_lock); goto out; } vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; guid0 = vd0->vdev_guid; (void) strcpy(path0, vd0->vdev_path); (void) strcpy(pathrand, vd0->vdev_path); leaf = 0; leaves = 1; maxfaults = INT_MAX; /* no limit on cache devices */ } spa_config_exit(spa, SCL_STATE, FTAG); (void) pthread_rwlock_unlock(&ztest_name_lock); /* * If we can tolerate two or more faults, or we're dealing * with a slog, randomly online/offline vd0. */ if ((maxfaults >= 2 || islog) && guid0 != 0) { if (ztest_random(10) < 6) { int flags = (ztest_random(2) == 0 ? ZFS_OFFLINE_TEMPORARY : 0); /* * We have to grab the zs_name_lock as writer to * prevent a race between offlining a slog and * destroying a dataset. Offlining the slog will * grab a reference on the dataset which may cause * dsl_destroy_head() to fail with EBUSY thus * leaving the dataset in an inconsistent state. */ if (islog) (void) pthread_rwlock_wrlock(&ztest_name_lock); VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); if (islog) (void) pthread_rwlock_unlock(&ztest_name_lock); } else { /* * Ideally we would like to be able to randomly * call vdev_[on|off]line without holding locks * to force unpredictable failures but the side * effects of vdev_[on|off]line prevent us from * doing so. We grab the ztest_vdev_lock here to * prevent a race between injection testing and * aux_vdev removal. */ mutex_enter(&ztest_vdev_lock); (void) vdev_online(spa, guid0, 0, NULL); mutex_exit(&ztest_vdev_lock); } } if (maxfaults == 0) goto out; /* * We have at least single-fault tolerance, so inject data corruption. */ fd = open(pathrand, O_RDWR); if (fd == -1) /* we hit a gap in the device namespace */ goto out; fsize = lseek(fd, 0, SEEK_END); while (--iters != 0) { /* * The offset must be chosen carefully to ensure that * we do not inject a given logical block with errors * on two different leaf devices, because ZFS can not * tolerate that (if maxfaults==1). * * We divide each leaf into chunks of size * (# leaves * SPA_MAXBLOCKSIZE * 4). Within each chunk * there is a series of ranges to which we can inject errors. * Each range can accept errors on only a single leaf vdev. * The error injection ranges are separated by ranges * which we will not inject errors on any device (DMZs). * Each DMZ must be large enough such that a single block * can not straddle it, so that a single block can not be * a target in two different injection ranges (on different * leaf vdevs). * * For example, with 3 leaves, each chunk looks like: * 0 to 32M: injection range for leaf 0 * 32M to 64M: DMZ - no injection allowed * 64M to 96M: injection range for leaf 1 * 96M to 128M: DMZ - no injection allowed * 128M to 160M: injection range for leaf 2 * 160M to 192M: DMZ - no injection allowed */ offset = ztest_random(fsize / (leaves << bshift)) * (leaves << bshift) + (leaf << bshift) + (ztest_random(1ULL << (bshift - 1)) & -8ULL); /* * Only allow damage to the labels at one end of the vdev. * * If all labels are damaged, the device will be totally * inaccessible, which will result in loss of data, * because we also damage (parts of) the other side of * the mirror/raidz. * * Additionally, we will always have both an even and an * odd label, so that we can handle crashes in the * middle of vdev_config_sync(). */ if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE) continue; /* * The two end labels are stored at the "end" of the disk, but * the end of the disk (vdev_psize) is aligned to * sizeof (vdev_label_t). */ uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t)); if ((leaf & 1) == 1 && offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE) continue; mutex_enter(&ztest_vdev_lock); if (mirror_save != zs->zs_mirrors) { mutex_exit(&ztest_vdev_lock); (void) close(fd); goto out; } if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) fatal(1, "can't inject bad word at 0x%llx in %s", offset, pathrand); mutex_exit(&ztest_vdev_lock); if (ztest_opts.zo_verbose >= 7) (void) printf("injected bad word into %s," " offset 0x%llx\n", pathrand, (u_longlong_t)offset); } (void) close(fd); out: umem_free(path0, MAXPATHLEN); umem_free(pathrand, MAXPATHLEN); } /* * Verify that DDT repair works as expected. */ void ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) { ztest_shared_t *zs = ztest_shared; spa_t *spa = ztest_spa; objset_t *os = zd->zd_os; ztest_od_t *od; uint64_t object, blocksize, txg, pattern, psize; enum zio_checksum checksum = spa_dedup_checksum(spa); dmu_buf_t *db; dmu_tx_t *tx; abd_t *abd; blkptr_t blk; int copies = 2 * ZIO_DEDUPDITTO_MIN; int i; blocksize = ztest_random_blocksize(); blocksize = MIN(blocksize, 2048); /* because we write so many */ od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { umem_free(od, sizeof (ztest_od_t)); return; } /* * Take the name lock as writer to prevent anyone else from changing * the pool and dataset properies we need to maintain during this test. */ (void) pthread_rwlock_wrlock(&ztest_name_lock); if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, B_FALSE) != 0 || ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, B_FALSE) != 0) { (void) pthread_rwlock_unlock(&ztest_name_lock); umem_free(od, sizeof (ztest_od_t)); return; } dmu_objset_stats_t dds; dsl_pool_config_enter(dmu_objset_pool(os), FTAG); dmu_objset_fast_stat(os, &dds); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); object = od[0].od_object; blocksize = od[0].od_blocksize; pattern = zs->zs_guid ^ dds.dds_guid; ASSERT(object != 0); tx = dmu_tx_create(os); dmu_tx_hold_write(tx, object, 0, copies * blocksize); txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); if (txg == 0) { (void) pthread_rwlock_unlock(&ztest_name_lock); umem_free(od, sizeof (ztest_od_t)); return; } /* * Write all the copies of our block. */ for (i = 0; i < copies; i++) { uint64_t offset = i * blocksize; int error = dmu_buf_hold(os, object, offset, FTAG, &db, DMU_READ_NO_PREFETCH); if (error != 0) { fatal(B_FALSE, "dmu_buf_hold(%p, %llu, %llu) = %u", os, (long long)object, (long long) offset, error); } ASSERT(db->db_offset == offset); ASSERT(db->db_size == blocksize); ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || ztest_pattern_match(db->db_data, db->db_size, 0ULL)); dmu_buf_will_fill(db, tx); ztest_pattern_set(db->db_data, db->db_size, pattern); dmu_buf_rele(db, FTAG); } dmu_tx_commit(tx); txg_wait_synced(spa_get_dsl(spa), txg); /* * Find out what block we got. */ VERIFY0(dmu_buf_hold(os, object, 0, FTAG, &db, DMU_READ_NO_PREFETCH)); blk = *((dmu_buf_impl_t *)db)->db_blkptr; dmu_buf_rele(db, FTAG); /* * Damage the block. Dedup-ditto will save us when we read it later. */ psize = BP_GET_PSIZE(&blk); abd = abd_alloc_linear(psize, B_TRUE); ztest_pattern_set(abd_to_buf(abd), psize, ~pattern); (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, abd, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); abd_free(abd); (void) pthread_rwlock_unlock(&ztest_name_lock); umem_free(od, sizeof (ztest_od_t)); } /* * Scrub the pool. */ /* ARGSUSED */ void ztest_scrub(ztest_ds_t *zd, uint64_t id) { spa_t *spa = ztest_spa; /* * Scrub in progress by device removal. */ if (ztest_device_removal_active) return; (void) spa_scan(spa, POOL_SCAN_SCRUB); (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ (void) spa_scan(spa, POOL_SCAN_SCRUB); } /* * Change the guid for the pool. */ /* ARGSUSED */ void ztest_reguid(ztest_ds_t *zd, uint64_t id) { spa_t *spa = ztest_spa; uint64_t orig, load; int error; if (ztest_opts.zo_mmp_test) return; orig = spa_guid(spa); load = spa_load_guid(spa); (void) pthread_rwlock_wrlock(&ztest_name_lock); error = spa_change_guid(spa); (void) pthread_rwlock_unlock(&ztest_name_lock); if (error != 0) return; if (ztest_opts.zo_verbose >= 4) { (void) printf("Changed guid old %llu -> %llu\n", (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); } VERIFY3U(orig, !=, spa_guid(spa)); VERIFY3U(load, ==, spa_load_guid(spa)); } /* * Rename the pool to a different name and then rename it back. */ /* ARGSUSED */ void ztest_spa_rename(ztest_ds_t *zd, uint64_t id) { char *oldname, *newname; spa_t *spa; if (ztest_opts.zo_mmp_test) return; (void) pthread_rwlock_wrlock(&ztest_name_lock); oldname = ztest_opts.zo_pool; newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); (void) strcpy(newname, oldname); (void) strcat(newname, "_tmp"); /* * Do the rename */ VERIFY3U(0, ==, spa_rename(oldname, newname)); /* * Try to open it under the old name, which shouldn't exist */ VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); /* * Open it under the new name and make sure it's still the same spa_t. */ VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); ASSERT(spa == ztest_spa); spa_close(spa, FTAG); /* * Rename it back to the original */ VERIFY3U(0, ==, spa_rename(newname, oldname)); /* * Make sure it can still be opened */ VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); ASSERT(spa == ztest_spa); spa_close(spa, FTAG); umem_free(newname, strlen(newname) + 1); (void) pthread_rwlock_unlock(&ztest_name_lock); } void ztest_fletcher(ztest_ds_t *zd, uint64_t id) { hrtime_t end = gethrtime() + NANOSEC; while (gethrtime() <= end) { int run_count = 100; void *buf; struct abd *abd_data, *abd_meta; uint32_t size; int *ptr; int i; zio_cksum_t zc_ref; zio_cksum_t zc_ref_byteswap; size = ztest_random_blocksize(); buf = umem_alloc(size, UMEM_NOFAIL); abd_data = abd_alloc(size, B_FALSE); abd_meta = abd_alloc(size, B_TRUE); for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) *ptr = ztest_random(UINT_MAX); abd_copy_from_buf_off(abd_data, buf, 0, size); abd_copy_from_buf_off(abd_meta, buf, 0, size); VERIFY0(fletcher_4_impl_set("scalar")); fletcher_4_native(buf, size, NULL, &zc_ref); fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap); VERIFY0(fletcher_4_impl_set("cycle")); while (run_count-- > 0) { zio_cksum_t zc; zio_cksum_t zc_byteswap; fletcher_4_byteswap(buf, size, NULL, &zc_byteswap); fletcher_4_native(buf, size, NULL, &zc); VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, sizeof (zc_byteswap))); /* Test ABD - data */ abd_fletcher_4_byteswap(abd_data, size, NULL, &zc_byteswap); abd_fletcher_4_native(abd_data, size, NULL, &zc); VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, sizeof (zc_byteswap))); /* Test ABD - metadata */ abd_fletcher_4_byteswap(abd_meta, size, NULL, &zc_byteswap); abd_fletcher_4_native(abd_meta, size, NULL, &zc); VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, sizeof (zc_byteswap))); } umem_free(buf, size); abd_free(abd_data); abd_free(abd_meta); } } void ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id) { void *buf; size_t size; int *ptr; int i; zio_cksum_t zc_ref; zio_cksum_t zc_ref_bswap; hrtime_t end = gethrtime() + NANOSEC; while (gethrtime() <= end) { int run_count = 100; size = ztest_random_blocksize(); buf = umem_alloc(size, UMEM_NOFAIL); for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) *ptr = ztest_random(UINT_MAX); VERIFY0(fletcher_4_impl_set("scalar")); fletcher_4_native(buf, size, NULL, &zc_ref); fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap); VERIFY0(fletcher_4_impl_set("cycle")); while (run_count-- > 0) { zio_cksum_t zc; zio_cksum_t zc_bswap; size_t pos = 0; ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); while (pos < size) { size_t inc = 64 * ztest_random(size / 67); /* sometimes add few bytes to test non-simd */ if (ztest_random(100) < 10) inc += P2ALIGN(ztest_random(64), sizeof (uint32_t)); if (inc > (size - pos)) inc = size - pos; fletcher_4_incremental_native(buf + pos, inc, &zc); fletcher_4_incremental_byteswap(buf + pos, inc, &zc_bswap); pos += inc; } VERIFY3U(pos, ==, size); VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); /* * verify if incremental on the whole buffer is * equivalent to non-incremental version */ ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); fletcher_4_incremental_native(buf, size, &zc); fletcher_4_incremental_byteswap(buf, size, &zc_bswap); VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); } umem_free(buf, size); } } static int ztest_check_path(char *path) { struct stat s; /* return true on success */ return (!stat(path, &s)); } static void ztest_get_zdb_bin(char *bin, int len) { char *zdb_path; /* * Try to use ZDB_PATH and in-tree zdb path. If not successful, just * let popen to search through PATH. */ if ((zdb_path = getenv("ZDB_PATH"))) { strlcpy(bin, zdb_path, len); /* In env */ if (!ztest_check_path(bin)) { ztest_dump_core = 0; fatal(1, "invalid ZDB_PATH '%s'", bin); } return; } VERIFY(realpath(getexecname(), bin) != NULL); if (strstr(bin, "/ztest/")) { strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ strcat(bin, "/zdb/zdb"); if (ztest_check_path(bin)) return; } strcpy(bin, "zdb"); } /* * Verify pool integrity by running zdb. */ static void ztest_run_zdb(char *pool) { int status; char *bin; char *zdb; char *zbuf; const int len = MAXPATHLEN + MAXNAMELEN + 20; FILE *fp; bin = umem_alloc(len, UMEM_NOFAIL); zdb = umem_alloc(len, UMEM_NOFAIL); zbuf = umem_alloc(1024, UMEM_NOFAIL); ztest_get_zdb_bin(bin, len); (void) sprintf(zdb, "%s -bcc%s%s -G -d -U %s " "-o zfs_reconstruct_indirect_combinations_max=1000000 %s", bin, ztest_opts.zo_verbose >= 3 ? "s" : "", ztest_opts.zo_verbose >= 4 ? "v" : "", spa_config_path, pool); if (ztest_opts.zo_verbose >= 5) (void) printf("Executing %s\n", strstr(zdb, "zdb ")); fp = popen(zdb, "r"); while (fgets(zbuf, 1024, fp) != NULL) if (ztest_opts.zo_verbose >= 3) (void) printf("%s", zbuf); status = pclose(fp); if (status == 0) goto out; ztest_dump_core = 0; if (WIFEXITED(status)) fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); else fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); out: umem_free(bin, len); umem_free(zdb, len); umem_free(zbuf, 1024); } static void ztest_walk_pool_directory(char *header) { spa_t *spa = NULL; if (ztest_opts.zo_verbose >= 6) (void) printf("%s\n", header); mutex_enter(&spa_namespace_lock); while ((spa = spa_next(spa)) != NULL) if (ztest_opts.zo_verbose >= 6) (void) printf("\t%s\n", spa_name(spa)); mutex_exit(&spa_namespace_lock); } static void ztest_spa_import_export(char *oldname, char *newname) { nvlist_t *config, *newconfig; uint64_t pool_guid; spa_t *spa; int error; if (ztest_opts.zo_verbose >= 4) { (void) printf("import/export: old = %s, new = %s\n", oldname, newname); } /* * Clean up from previous runs. */ (void) spa_destroy(newname); /* * Get the pool's configuration and guid. */ VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); /* * Kick off a scrub to tickle scrub/export races. */ if (ztest_random(2) == 0) (void) spa_scan(spa, POOL_SCAN_SCRUB); pool_guid = spa_guid(spa); spa_close(spa, FTAG); ztest_walk_pool_directory("pools before export"); /* * Export it. */ VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); ztest_walk_pool_directory("pools after export"); /* * Try to import it. */ newconfig = spa_tryimport(config); ASSERT(newconfig != NULL); nvlist_free(newconfig); /* * Import it under the new name. */ error = spa_import(newname, config, NULL, 0); if (error != 0) { dump_nvlist(config, 0); fatal(B_FALSE, "couldn't import pool %s as %s: error %u", oldname, newname, error); } ztest_walk_pool_directory("pools after import"); /* * Try to import it again -- should fail with EEXIST. */ VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); /* * Try to import it under a different name -- should fail with EEXIST. */ VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); /* * Verify that the pool is no longer visible under the old name. */ VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); /* * Verify that we can open and close the pool using the new name. */ VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); ASSERT(pool_guid == spa_guid(spa)); spa_close(spa, FTAG); nvlist_free(config); } static void ztest_resume(spa_t *spa) { if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) (void) printf("resuming from suspended state\n"); spa_vdev_state_enter(spa, SCL_NONE); vdev_clear(spa, NULL); (void) spa_vdev_state_exit(spa, NULL, 0); (void) zio_resume(spa); } static void ztest_resume_thread(void *arg) { spa_t *spa = arg; while (!ztest_exiting) { if (spa_suspended(spa)) ztest_resume(spa); (void) poll(NULL, 0, 100); /* * Periodically change the zfs_compressed_arc_enabled setting. */ if (ztest_random(10) == 0) zfs_compressed_arc_enabled = ztest_random(2); /* * Periodically change the zfs_abd_scatter_enabled setting. */ if (ztest_random(10) == 0) zfs_abd_scatter_enabled = ztest_random(2); } thread_exit(); } static void ztest_deadman_thread(void *arg) { ztest_shared_t *zs = arg; spa_t *spa = ztest_spa; hrtime_t delta, overdue, total = 0; for (;;) { delta = zs->zs_thread_stop - zs->zs_thread_start + MSEC2NSEC(zfs_deadman_synctime_ms); (void) poll(NULL, 0, (int)NSEC2MSEC(delta)); /* * If the pool is suspended then fail immediately. Otherwise, * check to see if the pool is making any progress. If * vdev_deadman() discovers that there hasn't been any recent * I/Os then it will end up aborting the tests. */ if (spa_suspended(spa) || spa->spa_root_vdev == NULL) { fatal(0, "aborting test after %llu seconds because " "pool has transitioned to a suspended state.", zfs_deadman_synctime_ms / 1000); } vdev_deadman(spa->spa_root_vdev, FTAG); /* * If the process doesn't complete within a grace period of * zfs_deadman_synctime_ms over the expected finish time, * then it may be hung and is terminated. */ overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms); total += zfs_deadman_synctime_ms / 1000; if (gethrtime() > overdue) { fatal(0, "aborting test after %llu seconds because " "the process is overdue for termination.", total); } (void) printf("ztest has been running for %lld seconds\n", total); } } static void ztest_execute(int test, ztest_info_t *zi, uint64_t id) { ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); hrtime_t functime = gethrtime(); int i; for (i = 0; i < zi->zi_iters; i++) zi->zi_func(zd, id); functime = gethrtime() - functime; atomic_add_64(&zc->zc_count, 1); atomic_add_64(&zc->zc_time, functime); if (ztest_opts.zo_verbose >= 4) (void) printf("%6.2f sec in %s\n", (double)functime / NANOSEC, zi->zi_funcname); } static void ztest_thread(void *arg) { int rand; uint64_t id = (uintptr_t)arg; ztest_shared_t *zs = ztest_shared; uint64_t call_next; hrtime_t now; ztest_info_t *zi; ztest_shared_callstate_t *zc; while ((now = gethrtime()) < zs->zs_thread_stop) { /* * See if it's time to force a crash. */ if (now > zs->zs_thread_kill) ztest_kill(zs); /* * If we're getting ENOSPC with some regularity, stop. */ if (zs->zs_enospc_count > 10) break; /* * Pick a random function to execute. */ rand = ztest_random(ZTEST_FUNCS); zi = &ztest_info[rand]; zc = ZTEST_GET_SHARED_CALLSTATE(rand); call_next = zc->zc_next; if (now >= call_next && atomic_cas_64(&zc->zc_next, call_next, call_next + ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { ztest_execute(rand, zi, id); } } thread_exit(); } static void ztest_dataset_name(char *dsname, char *pool, int d) { (void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d); } static void ztest_dataset_destroy(int d) { char name[ZFS_MAX_DATASET_NAME_LEN]; int t; ztest_dataset_name(name, ztest_opts.zo_pool, d); if (ztest_opts.zo_verbose >= 3) (void) printf("Destroying %s to free up space\n", name); /* * Cleanup any non-standard clones and snapshots. In general, * ztest thread t operates on dataset (t % zopt_datasets), * so there may be more than one thing to clean up. */ for (t = d; t < ztest_opts.zo_threads; t += ztest_opts.zo_datasets) ztest_dsl_dataset_cleanup(name, t); (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); } static void ztest_dataset_dirobj_verify(ztest_ds_t *zd) { uint64_t usedobjs, dirobjs, scratch; /* * ZTEST_DIROBJ is the object directory for the entire dataset. * Therefore, the number of objects in use should equal the * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. * If not, we have an object leak. * * Note that we can only check this in ztest_dataset_open(), * when the open-context and syncing-context values agree. * That's because zap_count() returns the open-context value, * while dmu_objset_space() returns the rootbp fill count. */ VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); ASSERT3U(dirobjs + 1, ==, usedobjs); } static int ztest_dataset_open(int d) { ztest_ds_t *zd = &ztest_ds[d]; uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; objset_t *os; zilog_t *zilog; char name[ZFS_MAX_DATASET_NAME_LEN]; int error; ztest_dataset_name(name, ztest_opts.zo_pool, d); (void) pthread_rwlock_rdlock(&ztest_name_lock); error = ztest_dataset_create(name); if (error == ENOSPC) { (void) pthread_rwlock_unlock(&ztest_name_lock); ztest_record_enospc(FTAG); return (error); } ASSERT(error == 0 || error == EEXIST); VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, zd, &os)); (void) pthread_rwlock_unlock(&ztest_name_lock); ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); zilog = zd->zd_zilog; if (zilog->zl_header->zh_claim_lr_seq != 0 && zilog->zl_header->zh_claim_lr_seq < committed_seq) fatal(0, "missing log records: claimed %llu < committed %llu", zilog->zl_header->zh_claim_lr_seq, committed_seq); ztest_dataset_dirobj_verify(zd); zil_replay(os, zd, ztest_replay_vector); ztest_dataset_dirobj_verify(zd); if (ztest_opts.zo_verbose >= 6) (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", zd->zd_name, (u_longlong_t)zilog->zl_parse_blk_count, (u_longlong_t)zilog->zl_parse_lr_count, (u_longlong_t)zilog->zl_replaying_seq); zilog = zil_open(os, ztest_get_data); if (zilog->zl_replaying_seq != 0 && zilog->zl_replaying_seq < committed_seq) fatal(0, "missing log records: replayed %llu < committed %llu", zilog->zl_replaying_seq, committed_seq); return (0); } static void ztest_dataset_close(int d) { ztest_ds_t *zd = &ztest_ds[d]; zil_close(zd->zd_zilog); txg_wait_synced(spa_get_dsl(zd->zd_os->os_spa), 0); dmu_objset_disown(zd->zd_os, B_TRUE, zd); ztest_zd_fini(zd); } /* * Kick off threads to run tests on all datasets in parallel. */ static void ztest_run(ztest_shared_t *zs) { spa_t *spa; objset_t *os; kthread_t *resume_thread; kthread_t **run_threads; uint64_t object; int error; int t, d; ztest_exiting = B_FALSE; /* * Initialize parent/child shared state. */ mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); zs->zs_thread_start = gethrtime(); zs->zs_thread_stop = zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); zs->zs_thread_kill = zs->zs_thread_stop; if (ztest_random(100) < ztest_opts.zo_killrate) { zs->zs_thread_kill -= ztest_random(ztest_opts.zo_passtime * NANOSEC); } mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), offsetof(ztest_cb_data_t, zcd_node)); /* * Open our pool. */ kernel_init(FREAD | FWRITE); VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); metaslab_preload_limit = ztest_random(20) + 1; ztest_spa = spa; dmu_objset_stats_t dds; VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool, DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os)); dsl_pool_config_enter(dmu_objset_pool(os), FTAG); dmu_objset_fast_stat(os, &dds); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); zs->zs_guid = dds.dds_guid; dmu_objset_disown(os, B_TRUE, FTAG); spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; /* * Create a thread to periodically resume suspended I/O. */ resume_thread = thread_create(NULL, 0, ztest_resume_thread, spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); /* * Create a deadman thread and set to panic if we hang. */ (void) thread_create(NULL, 0, ztest_deadman_thread, zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; /* * Verify that we can safely inquire about any object, * whether it's allocated or not. To make it interesting, * we probe a 5-wide window around each power of two. * This hits all edge cases, including zero and the max. */ for (t = 0; t < 64; t++) { for (d = -5; d <= 5; d++) { error = dmu_object_info(spa->spa_meta_objset, (1ULL << t) + d, NULL); ASSERT(error == 0 || error == ENOENT || error == EINVAL); } } /* * If we got any ENOSPC errors on the previous run, destroy something. */ if (zs->zs_enospc_count != 0) { int d = ztest_random(ztest_opts.zo_datasets); ztest_dataset_destroy(d); } zs->zs_enospc_count = 0; run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *), UMEM_NOFAIL); if (ztest_opts.zo_verbose >= 4) (void) printf("starting main threads...\n"); /* * Kick off all the tests that run in parallel. */ for (t = 0; t < ztest_opts.zo_threads; t++) { if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) { umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); return; } run_threads[t] = thread_create(NULL, 0, ztest_thread, (void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); } /* * Wait for all of the tests to complete. We go in reverse order * so we don't close datasets while threads are still using them. */ for (t = ztest_opts.zo_threads - 1; t >= 0; t--) { VERIFY0(thread_join(run_threads[t])); if (t < ztest_opts.zo_datasets) ztest_dataset_close(t); } txg_wait_synced(spa_get_dsl(spa), 0); zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); /* Kill the resume thread */ ztest_exiting = B_TRUE; VERIFY0(thread_join(resume_thread)); ztest_resume(spa); /* * Right before closing the pool, kick off a bunch of async I/O; * spa_close() should wait for it to complete. */ for (object = 1; object < 50; object++) { dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20, ZIO_PRIORITY_SYNC_READ); } /* Verify that at least one commit cb was called in a timely fashion */ if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) VERIFY0(zc_min_txg_delay); spa_close(spa, FTAG); /* * Verify that we can loop over all pools. */ mutex_enter(&spa_namespace_lock); for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) if (ztest_opts.zo_verbose > 3) (void) printf("spa_next: found %s\n", spa_name(spa)); mutex_exit(&spa_namespace_lock); /* * Verify that we can export the pool and reimport it under a * different name. */ if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) { char name[ZFS_MAX_DATASET_NAME_LEN]; (void) snprintf(name, sizeof (name), "%s_import", ztest_opts.zo_pool); ztest_spa_import_export(ztest_opts.zo_pool, name); ztest_spa_import_export(name, ztest_opts.zo_pool); } kernel_fini(); list_destroy(&zcl.zcl_callbacks); mutex_destroy(&zcl.zcl_callbacks_lock); (void) pthread_rwlock_destroy(&ztest_name_lock); mutex_destroy(&ztest_vdev_lock); mutex_destroy(&ztest_checkpoint_lock); } static void ztest_freeze(void) { ztest_ds_t *zd = &ztest_ds[0]; spa_t *spa; int numloops = 0; if (ztest_opts.zo_verbose >= 3) (void) printf("testing spa_freeze()...\n"); kernel_init(FREAD | FWRITE); VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); VERIFY3U(0, ==, ztest_dataset_open(0)); ztest_spa = spa; /* * Force the first log block to be transactionally allocated. * We have to do this before we freeze the pool -- otherwise * the log chain won't be anchored. */ while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { ztest_dmu_object_alloc_free(zd, 0); zil_commit(zd->zd_zilog, 0); } txg_wait_synced(spa_get_dsl(spa), 0); /* * Freeze the pool. This stops spa_sync() from doing anything, * so that the only way to record changes from now on is the ZIL. */ spa_freeze(spa); /* * Because it is hard to predict how much space a write will actually * require beforehand, we leave ourselves some fudge space to write over * capacity. */ uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2; /* * Run tests that generate log records but don't alter the pool config * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). * We do a txg_wait_synced() after each iteration to force the txg * to increase well beyond the last synced value in the uberblock. * The ZIL should be OK with that. * * Run a random number of times less than zo_maxloops and ensure we do * not run out of space on the pool. */ while (ztest_random(10) != 0 && numloops++ < ztest_opts.zo_maxloops && metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) { ztest_od_t od; ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE)); ztest_io(zd, od.od_object, ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); txg_wait_synced(spa_get_dsl(spa), 0); } /* * Commit all of the changes we just generated. */ zil_commit(zd->zd_zilog, 0); txg_wait_synced(spa_get_dsl(spa), 0); /* * Close our dataset and close the pool. */ ztest_dataset_close(0); spa_close(spa, FTAG); kernel_fini(); /* * Open and close the pool and dataset to induce log replay. */ kernel_init(FREAD | FWRITE); VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); ASSERT(spa_freeze_txg(spa) == UINT64_MAX); VERIFY3U(0, ==, ztest_dataset_open(0)); ztest_spa = spa; txg_wait_synced(spa_get_dsl(spa), 0); ztest_dataset_close(0); ztest_reguid(NULL, 0); spa_close(spa, FTAG); kernel_fini(); } void print_time(hrtime_t t, char *timebuf) { hrtime_t s = t / NANOSEC; hrtime_t m = s / 60; hrtime_t h = m / 60; hrtime_t d = h / 24; s -= m * 60; m -= h * 60; h -= d * 24; timebuf[0] = '\0'; if (d) (void) sprintf(timebuf, "%llud%02lluh%02llum%02llus", d, h, m, s); else if (h) (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); else if (m) (void) sprintf(timebuf, "%llum%02llus", m, s); else (void) sprintf(timebuf, "%llus", s); } static nvlist_t * make_random_props(void) { nvlist_t *props; VERIFY0(nvlist_alloc(&props, NV_UNIQUE_NAME, 0)); if (ztest_random(2) == 0) return (props); VERIFY0(nvlist_add_uint64(props, zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1)); return (props); } /* * Import a storage pool with the given name. */ static void ztest_import(ztest_shared_t *zs) { libzfs_handle_t *hdl; importargs_t args = { 0 }; spa_t *spa; nvlist_t *cfg = NULL; int nsearch = 1; char *searchdirs[nsearch]; char *name = ztest_opts.zo_pool; int flags = ZFS_IMPORT_MISSING_LOG; int error; mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); kernel_init(FREAD | FWRITE); hdl = libzfs_init(); searchdirs[0] = ztest_opts.zo_dir; args.paths = nsearch; args.path = searchdirs; args.can_be_active = B_FALSE; error = zpool_tryimport(hdl, name, &cfg, &args); if (error) (void) fatal(0, "No pools found\n"); VERIFY0(spa_import(name, cfg, NULL, flags)); VERIFY0(spa_open(name, &spa, FTAG)); zs->zs_metaslab_sz = 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; spa_close(spa, FTAG); libzfs_fini(hdl); kernel_fini(); if (!ztest_opts.zo_mmp_test) { ztest_run_zdb(ztest_opts.zo_pool); ztest_freeze(); ztest_run_zdb(ztest_opts.zo_pool); } (void) pthread_rwlock_destroy(&ztest_name_lock); mutex_destroy(&ztest_vdev_lock); mutex_destroy(&ztest_checkpoint_lock); } /* * Create a storage pool with the given name and initial vdev size. * Then test spa_freeze() functionality. */ static void ztest_init(ztest_shared_t *zs) { spa_t *spa; nvlist_t *nvroot, *props; int i; mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); kernel_init(FREAD | FWRITE); /* * Create the storage pool. */ (void) spa_destroy(ztest_opts.zo_pool); ztest_shared->zs_vdev_next_leaf = 0; zs->zs_splits = 0; zs->zs_mirrors = ztest_opts.zo_mirrors; nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, 0, ztest_opts.zo_raidz, zs->zs_mirrors, 1); props = make_random_props(); /* * We don't expect the pool to suspend unless maxfaults == 0, * in which case ztest_fault_inject() temporarily takes away * the only valid replica. */ VERIFY0(nvlist_add_uint64(props, zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE), MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT)); for (i = 0; i < SPA_FEATURES; i++) { char *buf; VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", spa_feature_table[i].fi_uname)); VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); free(buf); } VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL)); nvlist_free(nvroot); nvlist_free(props); VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); zs->zs_metaslab_sz = 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; spa_close(spa, FTAG); kernel_fini(); if (!ztest_opts.zo_mmp_test) { ztest_run_zdb(ztest_opts.zo_pool); ztest_freeze(); ztest_run_zdb(ztest_opts.zo_pool); } (void) pthread_rwlock_destroy(&ztest_name_lock); mutex_destroy(&ztest_vdev_lock); mutex_destroy(&ztest_checkpoint_lock); } static void setup_data_fd(void) { static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; ztest_fd_data = mkstemp(ztest_name_data); ASSERT3S(ztest_fd_data, >=, 0); (void) unlink(ztest_name_data); } static int shared_data_size(ztest_shared_hdr_t *hdr) { int size; size = hdr->zh_hdr_size; size += hdr->zh_opts_size; size += hdr->zh_size; size += hdr->zh_stats_size * hdr->zh_stats_count; size += hdr->zh_ds_size * hdr->zh_ds_count; return (size); } static void setup_hdr(void) { int size; ztest_shared_hdr_t *hdr; hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); ASSERT(hdr != MAP_FAILED); VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); hdr->zh_opts_size = sizeof (ztest_shared_opts_t); hdr->zh_size = sizeof (ztest_shared_t); hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); hdr->zh_stats_count = ZTEST_FUNCS; hdr->zh_ds_size = sizeof (ztest_shared_ds_t); hdr->zh_ds_count = ztest_opts.zo_datasets; size = shared_data_size(hdr); VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); } static void setup_data(void) { int size, offset; ztest_shared_hdr_t *hdr; uint8_t *buf; hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), PROT_READ, MAP_SHARED, ztest_fd_data, 0); ASSERT(hdr != MAP_FAILED); size = shared_data_size(hdr); (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); ASSERT(hdr != MAP_FAILED); buf = (uint8_t *)hdr; offset = hdr->zh_hdr_size; ztest_shared_opts = (void *)&buf[offset]; offset += hdr->zh_opts_size; ztest_shared = (void *)&buf[offset]; offset += hdr->zh_size; ztest_shared_callstate = (void *)&buf[offset]; offset += hdr->zh_stats_size * hdr->zh_stats_count; ztest_shared_ds = (void *)&buf[offset]; } static boolean_t exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) { pid_t pid; int status; char *cmdbuf = NULL; pid = fork(); if (cmd == NULL) { cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); cmd = cmdbuf; } if (pid == -1) fatal(1, "fork failed"); if (pid == 0) { /* child */ char *emptyargv[2] = { cmd, NULL }; char fd_data_str[12]; struct rlimit rl = { 1024, 1024 }; (void) setrlimit(RLIMIT_NOFILE, &rl); (void) close(ztest_fd_rand); VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); (void) enable_extended_FILE_stdio(-1, -1); if (libpath != NULL) VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); (void) execv(cmd, emptyargv); ztest_dump_core = B_FALSE; fatal(B_TRUE, "exec failed: %s", cmd); } if (cmdbuf != NULL) { umem_free(cmdbuf, MAXPATHLEN); cmd = NULL; } while (waitpid(pid, &status, 0) != pid) continue; if (statusp != NULL) *statusp = status; if (WIFEXITED(status)) { if (WEXITSTATUS(status) != 0) { (void) fprintf(stderr, "child exited with code %d\n", WEXITSTATUS(status)); exit(2); } return (B_FALSE); } else if (WIFSIGNALED(status)) { if (!ignorekill || WTERMSIG(status) != SIGKILL) { (void) fprintf(stderr, "child died with signal %d\n", WTERMSIG(status)); exit(3); } return (B_TRUE); } else { (void) fprintf(stderr, "something strange happened to child\n"); exit(4); /* NOTREACHED */ } } static void ztest_run_init(void) { int i; ztest_shared_t *zs = ztest_shared; /* * Blow away any existing copy of zpool.cache */ (void) remove(spa_config_path); if (ztest_opts.zo_init == 0) { if (ztest_opts.zo_verbose >= 1) (void) printf("Importing pool %s\n", ztest_opts.zo_pool); ztest_import(zs); return; } /* * Create and initialize our storage pool. */ for (i = 1; i <= ztest_opts.zo_init; i++) { bzero(zs, sizeof (ztest_shared_t)); if (ztest_opts.zo_verbose >= 3 && ztest_opts.zo_init != 1) { (void) printf("ztest_init(), pass %d\n", i); } ztest_init(zs); } } int main(int argc, char **argv) { int kills = 0; int iters = 0; int older = 0; int newer = 0; ztest_shared_t *zs; ztest_info_t *zi; ztest_shared_callstate_t *zc; char timebuf[100]; char numbuf[NN_NUMBUF_SZ]; spa_t *spa; char *cmd; boolean_t hasalt; int f; char *fd_data_str = getenv("ZTEST_FD_DATA"); struct sigaction action; (void) setvbuf(stdout, NULL, _IOLBF, 0); dprintf_setup(&argc, argv); zfs_deadman_synctime_ms = 300000; /* * As two-word space map entries may not come up often (especially * if pool and vdev sizes are small) we want to force at least some * of them so the feature get tested. */ zfs_force_some_double_word_sm_entries = B_TRUE; action.sa_handler = sig_handler; sigemptyset(&action.sa_mask); action.sa_flags = 0; if (sigaction(SIGSEGV, &action, NULL) < 0) { (void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n", strerror(errno)); exit(EXIT_FAILURE); } if (sigaction(SIGABRT, &action, NULL) < 0) { (void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n", strerror(errno)); exit(EXIT_FAILURE); } /* * Force random_get_bytes() to use /dev/urandom in order to prevent * ztest from needlessly depleting the system entropy pool. */ random_path = "/dev/urandom"; ztest_fd_rand = open(random_path, O_RDONLY); ASSERT3S(ztest_fd_rand, >=, 0); if (!fd_data_str) { process_options(argc, argv); setup_data_fd(); setup_hdr(); setup_data(); bcopy(&ztest_opts, ztest_shared_opts, sizeof (*ztest_shared_opts)); } else { ztest_fd_data = atoi(fd_data_str); setup_data(); bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); } ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); /* Override location of zpool.cache */ VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", ztest_opts.zo_dir) != -1); ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), UMEM_NOFAIL); zs = ztest_shared; if (fd_data_str) { metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging; metaslab_df_alloc_threshold = zs->zs_metaslab_df_alloc_threshold; if (zs->zs_do_init) ztest_run_init(); else ztest_run(zs); exit(0); } hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); if (ztest_opts.zo_verbose >= 1) { (void) printf("%llu vdevs, %d datasets, %d threads," " %llu seconds...\n", (u_longlong_t)ztest_opts.zo_vdevs, ztest_opts.zo_datasets, ztest_opts.zo_threads, (u_longlong_t)ztest_opts.zo_time); } cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); (void) strlcpy(cmd, getexecname(), MAXNAMELEN); zs->zs_do_init = B_TRUE; if (strlen(ztest_opts.zo_alt_ztest) != 0) { if (ztest_opts.zo_verbose >= 1) { (void) printf("Executing older ztest for " "initialization: %s\n", ztest_opts.zo_alt_ztest); } VERIFY(!exec_child(ztest_opts.zo_alt_ztest, ztest_opts.zo_alt_libpath, B_FALSE, NULL)); } else { VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); } zs->zs_do_init = B_FALSE; zs->zs_proc_start = gethrtime(); zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; for (f = 0; f < ZTEST_FUNCS; f++) { zi = &ztest_info[f]; zc = ZTEST_GET_SHARED_CALLSTATE(f); if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) zc->zc_next = UINT64_MAX; else zc->zc_next = zs->zs_proc_start + ztest_random(2 * zi->zi_interval[0] + 1); } /* * Run the tests in a loop. These tests include fault injection * to verify that self-healing data works, and forced crashes * to verify that we never lose on-disk consistency. */ while (gethrtime() < zs->zs_proc_stop) { int status; boolean_t killed; /* * Initialize the workload counters for each function. */ for (f = 0; f < ZTEST_FUNCS; f++) { zc = ZTEST_GET_SHARED_CALLSTATE(f); zc->zc_count = 0; zc->zc_time = 0; } /* Set the allocation switch size */ zs->zs_metaslab_df_alloc_threshold = ztest_random(zs->zs_metaslab_sz / 4) + 1; if (!hasalt || ztest_random(2) == 0) { if (hasalt && ztest_opts.zo_verbose >= 1) { (void) printf("Executing newer ztest: %s\n", cmd); } newer++; killed = exec_child(cmd, NULL, B_TRUE, &status); } else { if (hasalt && ztest_opts.zo_verbose >= 1) { (void) printf("Executing older ztest: %s\n", ztest_opts.zo_alt_ztest); } older++; killed = exec_child(ztest_opts.zo_alt_ztest, ztest_opts.zo_alt_libpath, B_TRUE, &status); } if (killed) kills++; iters++; if (ztest_opts.zo_verbose >= 1) { hrtime_t now = gethrtime(); now = MIN(now, zs->zs_proc_stop); print_time(zs->zs_proc_stop - now, timebuf); nicenum(zs->zs_space, numbuf, sizeof (numbuf)); (void) printf("Pass %3d, %8s, %3llu ENOSPC, " "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", iters, WIFEXITED(status) ? "Complete" : "SIGKILL", (u_longlong_t)zs->zs_enospc_count, 100.0 * zs->zs_alloc / zs->zs_space, numbuf, 100.0 * (now - zs->zs_proc_start) / (ztest_opts.zo_time * NANOSEC), timebuf); } if (ztest_opts.zo_verbose >= 2) { (void) printf("\nWorkload summary:\n\n"); (void) printf("%7s %9s %s\n", "Calls", "Time", "Function"); (void) printf("%7s %9s %s\n", "-----", "----", "--------"); for (f = 0; f < ZTEST_FUNCS; f++) { zi = &ztest_info[f]; zc = ZTEST_GET_SHARED_CALLSTATE(f); print_time(zc->zc_time, timebuf); (void) printf("%7llu %9s %s\n", (u_longlong_t)zc->zc_count, timebuf, zi->zi_funcname); } (void) printf("\n"); } /* * It's possible that we killed a child during a rename test, * in which case we'll have a 'ztest_tmp' pool lying around * instead of 'ztest'. Do a blind rename in case this happened. */ kernel_init(FREAD); if (spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0) { spa_close(spa, FTAG); } else { char tmpname[ZFS_MAX_DATASET_NAME_LEN]; kernel_fini(); kernel_init(FREAD | FWRITE); (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", ztest_opts.zo_pool); (void) spa_rename(tmpname, ztest_opts.zo_pool); } kernel_fini(); if (!ztest_opts.zo_mmp_test) ztest_run_zdb(ztest_opts.zo_pool); } if (ztest_opts.zo_verbose >= 1) { if (hasalt) { (void) printf("%d runs of older ztest: %s\n", older, ztest_opts.zo_alt_ztest); (void) printf("%d runs of newer ztest: %s\n", newer, cmd); } (void) printf("%d killed, %d completed, %.0f%% kill rate\n", kills, iters - kills, (100.0 * kills) / MAX(1, iters)); } umem_free(cmd, MAXNAMELEN); return (0); } diff --git a/man/man5/zfs-module-parameters.5 b/man/man5/zfs-module-parameters.5 index eae8dc42870f..62a16f662907 100644 --- a/man/man5/zfs-module-parameters.5 +++ b/man/man5/zfs-module-parameters.5 @@ -1,2852 +1,2865 @@ '\" te .\" Copyright (c) 2013 by Turbo Fredriksson . All rights reserved. .\" Copyright (c) 2017 Datto Inc. .\" Copyright (c) 2018 by Delphix. All rights reserved. .\" The contents of this file are subject to the terms of the Common Development .\" and Distribution License (the "License"). You may not use this file except .\" in compliance with the License. You can obtain a copy of the license at .\" usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing. .\" .\" See the License for the specific language governing permissions and .\" limitations under the License. When distributing Covered Code, include this .\" CDDL HEADER in each file and include the License file at .\" usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this .\" CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your .\" own identifying information: .\" Portions Copyright [yyyy] [name of copyright owner] .TH ZFS-MODULE-PARAMETERS 5 "Oct 28, 2017" .SH NAME zfs\-module\-parameters \- ZFS module parameters .SH DESCRIPTION .sp .LP Description of the different parameters to the ZFS module. .SS "Module parameters" .sp .LP .sp .ne 2 .na \fBdbuf_cache_max_bytes\fR (ulong) .ad .RS 12n Maximum size in bytes of the dbuf cache. When \fB0\fR this value will default to \fB1/2^dbuf_cache_shift\fR (1/32) of the target ARC size, otherwise the provided value in bytes will be used. The behavior of the dbuf cache and its associated settings can be observed via the \fB/proc/spl/kstat/zfs/dbufstats\fR kstat. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBdbuf_metadata_cache_max_bytes\fR (ulong) .ad .RS 12n Maximum size in bytes of the metadata dbuf cache. When \fB0\fR this value will default to \fB1/2^dbuf_cache_shift\fR (1/16) of the target ARC size, otherwise the provided value in bytes will be used. The behavior of the metadata dbuf cache and its associated settings can be observed via the \fB/proc/spl/kstat/zfs/dbufstats\fR kstat. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBdbuf_cache_hiwater_pct\fR (uint) .ad .RS 12n The percentage over \fBdbuf_cache_max_bytes\fR when dbufs must be evicted directly. .sp Default value: \fB10\fR%. .RE .sp .ne 2 .na \fBdbuf_cache_lowater_pct\fR (uint) .ad .RS 12n The percentage below \fBdbuf_cache_max_bytes\fR when the evict thread stops evicting dbufs. .sp Default value: \fB10\fR%. .RE .sp .ne 2 .na \fBdbuf_cache_shift\fR (int) .ad .RS 12n Set the size of the dbuf cache, \fBdbuf_cache_max_bytes\fR, to a log2 fraction of the target arc size. .sp Default value: \fB5\fR. .RE .sp .ne 2 .na \fBdbuf_metadata_cache_shift\fR (int) .ad .RS 12n Set the size of the dbuf metadata cache, \fBdbuf_metadata_cache_max_bytes\fR, to a log2 fraction of the target arc size. .sp Default value: \fB6\fR. .RE .sp .ne 2 .na \fBignore_hole_birth\fR (int) .ad .RS 12n When set, the hole_birth optimization will not be used, and all holes will always be sent on zfs send. Useful if you suspect your datasets are affected by a bug in hole_birth. .sp Use \fB1\fR for on (default) and \fB0\fR for off. .RE .sp .ne 2 .na \fBl2arc_feed_again\fR (int) .ad .RS 12n Turbo L2ARC warm-up. When the L2ARC is cold the fill interval will be set as fast as possible. .sp Use \fB1\fR for yes (default) and \fB0\fR to disable. .RE .sp .ne 2 .na \fBl2arc_feed_min_ms\fR (ulong) .ad .RS 12n Min feed interval in milliseconds. Requires \fBl2arc_feed_again=1\fR and only applicable in related situations. .sp Default value: \fB200\fR. .RE .sp .ne 2 .na \fBl2arc_feed_secs\fR (ulong) .ad .RS 12n Seconds between L2ARC writing .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBl2arc_headroom\fR (ulong) .ad .RS 12n How far through the ARC lists to search for L2ARC cacheable content, expressed as a multiplier of \fBl2arc_write_max\fR .sp Default value: \fB2\fR. .RE .sp .ne 2 .na \fBl2arc_headroom_boost\fR (ulong) .ad .RS 12n Scales \fBl2arc_headroom\fR by this percentage when L2ARC contents are being successfully compressed before writing. A value of 100 disables this feature. .sp Default value: \fB200\fR%. .RE .sp .ne 2 .na \fBl2arc_noprefetch\fR (int) .ad .RS 12n Do not write buffers to L2ARC if they were prefetched but not used by applications .sp Use \fB1\fR for yes (default) and \fB0\fR to disable. .RE .sp .ne 2 .na \fBl2arc_norw\fR (int) .ad .RS 12n No reads during writes .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBl2arc_write_boost\fR (ulong) .ad .RS 12n Cold L2ARC devices will have \fBl2arc_write_max\fR increased by this amount while they remain cold. .sp Default value: \fB8,388,608\fR. .RE .sp .ne 2 .na \fBl2arc_write_max\fR (ulong) .ad .RS 12n Max write bytes per interval .sp Default value: \fB8,388,608\fR. .RE .sp .ne 2 .na \fBmetaslab_aliquot\fR (ulong) .ad .RS 12n Metaslab granularity, in bytes. This is roughly similar to what would be referred to as the "stripe size" in traditional RAID arrays. In normal operation, ZFS will try to write this amount of data to a top-level vdev before moving on to the next one. .sp Default value: \fB524,288\fR. .RE .sp .ne 2 .na \fBmetaslab_bias_enabled\fR (int) .ad .RS 12n Enable metaslab group biasing based on its vdev's over- or under-utilization relative to the pool. .sp Use \fB1\fR for yes (default) and \fB0\fR for no. .RE .sp .ne 2 .na \fBmetaslab_force_ganging\fR (ulong) .ad .RS 12n Make some blocks above a certain size be gang blocks. This option is used by the test suite to facilitate testing. .sp Default value: \fB16,777,217\fR. .RE .sp .ne 2 .na \fBzfs_metaslab_segment_weight_enabled\fR (int) .ad .RS 12n Enable/disable segment-based metaslab selection. .sp Use \fB1\fR for yes (default) and \fB0\fR for no. .RE .sp .ne 2 .na \fBzfs_metaslab_switch_threshold\fR (int) .ad .RS 12n When using segment-based metaslab selection, continue allocating from the active metaslab until \fBzfs_metaslab_switch_threshold\fR worth of buckets have been exhausted. .sp Default value: \fB2\fR. .RE .sp .ne 2 .na \fBmetaslab_debug_load\fR (int) .ad .RS 12n Load all metaslabs during pool import. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBmetaslab_debug_unload\fR (int) .ad .RS 12n Prevent metaslabs from being unloaded. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBmetaslab_fragmentation_factor_enabled\fR (int) .ad .RS 12n Enable use of the fragmentation metric in computing metaslab weights. .sp Use \fB1\fR for yes (default) and \fB0\fR for no. .RE .sp .ne 2 .na \fBvdev_max_ms_count\fR (int) .ad .RS 12n When a vdev is added target this number of metaslabs per top-level vdev. .sp Default value: \fB200\fR. .RE .sp .ne 2 .na \fBvdev_min_ms_count\fR (int) .ad .RS 12n Minimum number of metaslabs to create in a top-level vdev. .sp Default value: \fB16\fR. .RE .sp .ne 2 .na \fBvdev_ms_count_limit\fR (int) .ad .RS 12n Practical upper limit of total metaslabs per top-level vdev. .sp Default value: \fB131,072\fR. .RE .sp .ne 2 .na \fBmetaslab_preload_enabled\fR (int) .ad .RS 12n Enable metaslab group preloading. .sp Use \fB1\fR for yes (default) and \fB0\fR for no. .RE .sp .ne 2 .na \fBmetaslab_lba_weighting_enabled\fR (int) .ad .RS 12n Give more weight to metaslabs with lower LBAs, assuming they have greater bandwidth as is typically the case on a modern constant angular velocity disk drive. .sp Use \fB1\fR for yes (default) and \fB0\fR for no. .RE .sp .ne 2 .na \fBspa_config_path\fR (charp) .ad .RS 12n SPA config file .sp Default value: \fB/etc/zfs/zpool.cache\fR. .RE .sp .ne 2 .na \fBspa_asize_inflation\fR (int) .ad .RS 12n Multiplication factor used to estimate actual disk consumption from the size of data being written. The default value is a worst case estimate, but lower values may be valid for a given pool depending on its configuration. Pool administrators who understand the factors involved may wish to specify a more realistic inflation factor, particularly if they operate close to quota or capacity limits. .sp Default value: \fB24\fR. .RE .sp .ne 2 .na \fBspa_load_print_vdev_tree\fR (int) .ad .RS 12n Whether to print the vdev tree in the debugging message buffer during pool import. Use 0 to disable and 1 to enable. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBspa_load_verify_data\fR (int) .ad .RS 12n Whether to traverse data blocks during an "extreme rewind" (\fB-X\fR) import. Use 0 to disable and 1 to enable. An extreme rewind import normally performs a full traversal of all blocks in the pool for verification. If this parameter is set to 0, the traversal skips non-metadata blocks. It can be toggled once the import has started to stop or start the traversal of non-metadata blocks. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBspa_load_verify_metadata\fR (int) .ad .RS 12n Whether to traverse blocks during an "extreme rewind" (\fB-X\fR) pool import. Use 0 to disable and 1 to enable. An extreme rewind import normally performs a full traversal of all blocks in the pool for verification. If this parameter is set to 0, the traversal is not performed. It can be toggled once the import has started to stop or start the traversal. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBspa_load_verify_maxinflight\fR (int) .ad .RS 12n Maximum concurrent I/Os during the traversal performed during an "extreme rewind" (\fB-X\fR) pool import. .sp Default value: \fB10000\fR. .RE .sp .ne 2 .na \fBspa_slop_shift\fR (int) .ad .RS 12n Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in the pool to be consumed. This ensures that we don't run the pool completely out of space, due to unaccounted changes (e.g. to the MOS). It also limits the worst-case time to allocate space. If we have less than this amount of free space, most ZPL operations (e.g. write, create) will return ENOSPC. .sp Default value: \fB5\fR. .RE .sp .ne 2 .na \fBvdev_removal_max_span\fR (int) .ad .RS 12n During top-level vdev removal, chunks of data are copied from the vdev which may include free space in order to trade bandwidth for IOPS. This parameter determines the maximum span of free space (in bytes) which will be included as "unnecessary" data in a chunk of copied data. The default value here was chosen to align with \fBzfs_vdev_read_gap_limit\fR, which is a similar concept when doing regular reads (but there's no reason it has to be the same). .sp Default value: \fB32,768\fR. .RE .sp .ne 2 .na \fBzfetch_array_rd_sz\fR (ulong) .ad .RS 12n If prefetching is enabled, disable prefetching for reads larger than this size. .sp Default value: \fB1,048,576\fR. .RE .sp .ne 2 .na \fBzfetch_max_distance\fR (uint) .ad .RS 12n Max bytes to prefetch per stream (default 8MB). .sp Default value: \fB8,388,608\fR. .RE .sp .ne 2 .na \fBzfetch_max_streams\fR (uint) .ad .RS 12n Max number of streams per zfetch (prefetch streams per file). .sp Default value: \fB8\fR. .RE .sp .ne 2 .na \fBzfetch_min_sec_reap\fR (uint) .ad .RS 12n Min time before an active prefetch stream can be reclaimed .sp Default value: \fB2\fR. .RE .sp .ne 2 .na \fBzfs_arc_dnode_limit\fR (ulong) .ad .RS 12n When the number of bytes consumed by dnodes in the ARC exceeds this number of bytes, try to unpin some of it in response to demand for non-metadata. This value acts as a ceiling to the amount of dnode metadata, and defaults to 0 which indicates that a percent which is based on \fBzfs_arc_dnode_limit_percent\fR of the ARC meta buffers that may be used for dnodes. See also \fBzfs_arc_meta_prune\fR which serves a similar purpose but is used when the amount of metadata in the ARC exceeds \fBzfs_arc_meta_limit\fR rather than in response to overall demand for non-metadata. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_dnode_limit_percent\fR (ulong) .ad .RS 12n Percentage that can be consumed by dnodes of ARC meta buffers. .sp See also \fBzfs_arc_dnode_limit\fR which serves a similar purpose but has a higher priority if set to nonzero value. .sp Default value: \fB10\fR%. .RE .sp .ne 2 .na \fBzfs_arc_dnode_reduce_percent\fR (ulong) .ad .RS 12n Percentage of ARC dnodes to try to scan in response to demand for non-metadata when the number of bytes consumed by dnodes exceeds \fBzfs_arc_dnode_limit\fR. .sp Default value: \fB10\fR% of the number of dnodes in the ARC. .RE .sp .ne 2 .na \fBzfs_arc_average_blocksize\fR (int) .ad .RS 12n The ARC's buffer hash table is sized based on the assumption of an average block size of \fBzfs_arc_average_blocksize\fR (default 8K). This works out to roughly 1MB of hash table per 1GB of physical memory with 8-byte pointers. For configurations with a known larger average block size this value can be increased to reduce the memory footprint. .sp Default value: \fB8192\fR. .RE .sp .ne 2 .na \fBzfs_arc_evict_batch_limit\fR (int) .ad .RS 12n Number ARC headers to evict per sub-list before proceeding to another sub-list. This batch-style operation prevents entire sub-lists from being evicted at once but comes at a cost of additional unlocking and locking. .sp Default value: \fB10\fR. .RE .sp .ne 2 .na \fBzfs_arc_grow_retry\fR (int) .ad .RS 12n If set to a non zero value, it will replace the arc_grow_retry value with this value. The arc_grow_retry value (default 5) is the number of seconds the ARC will wait before trying to resume growth after a memory pressure event. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_lotsfree_percent\fR (int) .ad .RS 12n Throttle I/O when free system memory drops below this percentage of total system memory. Setting this value to 0 will disable the throttle. .sp Default value: \fB10\fR%. .RE .sp .ne 2 .na \fBzfs_arc_max\fR (ulong) .ad .RS 12n Max arc size of ARC in bytes. If set to 0 then it will consume 1/2 of system RAM. This value must be at least 67108864 (64 megabytes). .sp This value can be changed dynamically with some caveats. It cannot be set back to 0 while running and reducing it below the current ARC size will not cause the ARC to shrink without memory pressure to induce shrinking. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_meta_adjust_restarts\fR (ulong) .ad .RS 12n The number of restart passes to make while scanning the ARC attempting the free buffers in order to stay below the \fBzfs_arc_meta_limit\fR. This value should not need to be tuned but is available to facilitate performance analysis. .sp Default value: \fB4096\fR. .RE .sp .ne 2 .na \fBzfs_arc_meta_limit\fR (ulong) .ad .RS 12n The maximum allowed size in bytes that meta data buffers are allowed to consume in the ARC. When this limit is reached meta data buffers will be reclaimed even if the overall arc_c_max has not been reached. This value defaults to 0 which indicates that a percent which is based on \fBzfs_arc_meta_limit_percent\fR of the ARC may be used for meta data. .sp This value my be changed dynamically except that it cannot be set back to 0 for a specific percent of the ARC; it must be set to an explicit value. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_meta_limit_percent\fR (ulong) .ad .RS 12n Percentage of ARC buffers that can be used for meta data. See also \fBzfs_arc_meta_limit\fR which serves a similar purpose but has a higher priority if set to nonzero value. .sp Default value: \fB75\fR%. .RE .sp .ne 2 .na \fBzfs_arc_meta_min\fR (ulong) .ad .RS 12n The minimum allowed size in bytes that meta data buffers may consume in the ARC. This value defaults to 0 which disables a floor on the amount of the ARC devoted meta data. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_meta_prune\fR (int) .ad .RS 12n The number of dentries and inodes to be scanned looking for entries which can be dropped. This may be required when the ARC reaches the \fBzfs_arc_meta_limit\fR because dentries and inodes can pin buffers in the ARC. Increasing this value will cause to dentry and inode caches to be pruned more aggressively. Setting this value to 0 will disable pruning the inode and dentry caches. .sp Default value: \fB10,000\fR. .RE .sp .ne 2 .na \fBzfs_arc_meta_strategy\fR (int) .ad .RS 12n Define the strategy for ARC meta data buffer eviction (meta reclaim strategy). A value of 0 (META_ONLY) will evict only the ARC meta data buffers. A value of 1 (BALANCED) indicates that additional data buffers may be evicted if that is required to in order to evict the required number of meta data buffers. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_arc_min\fR (ulong) .ad .RS 12n Min arc size of ARC in bytes. If set to 0 then arc_c_min will default to consuming the larger of 32M or 1/32 of total system memory. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_min_prefetch_ms\fR (int) .ad .RS 12n Minimum time prefetched blocks are locked in the ARC, specified in ms. A value of \fB0\fR will default to 1000 ms. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_min_prescient_prefetch_ms\fR (int) .ad .RS 12n Minimum time "prescient prefetched" blocks are locked in the ARC, specified in ms. These blocks are meant to be prefetched fairly aggresively ahead of the code that may use them. A value of \fB0\fR will default to 6000 ms. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_max_missing_tvds\fR (int) .ad .RS 12n Number of missing top-level vdevs which will be allowed during pool import (only in read-only mode). .sp Default value: \fB0\fR .RE .sp .ne 2 .na \fBzfs_multilist_num_sublists\fR (int) .ad .RS 12n To allow more fine-grained locking, each ARC state contains a series of lists for both data and meta data objects. Locking is performed at the level of these "sub-lists". This parameters controls the number of sub-lists per ARC state, and also applies to other uses of the multilist data structure. .sp Default value: \fB4\fR or the number of online CPUs, whichever is greater .RE .sp .ne 2 .na \fBzfs_arc_overflow_shift\fR (int) .ad .RS 12n The ARC size is considered to be overflowing if it exceeds the current ARC target size (arc_c) by a threshold determined by this parameter. The threshold is calculated as a fraction of arc_c using the formula "arc_c >> \fBzfs_arc_overflow_shift\fR". The default value of 8 causes the ARC to be considered to be overflowing if it exceeds the target size by 1/256th (0.3%) of the target size. When the ARC is overflowing, new buffer allocations are stalled until the reclaim thread catches up and the overflow condition no longer exists. .sp Default value: \fB8\fR. .RE .sp .ne 2 .na \fBzfs_arc_p_min_shift\fR (int) .ad .RS 12n If set to a non zero value, this will update arc_p_min_shift (default 4) with the new value. arc_p_min_shift is used to shift of arc_c for calculating both min and max max arc_p .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_p_dampener_disable\fR (int) .ad .RS 12n Disable arc_p adapt dampener .sp Use \fB1\fR for yes (default) and \fB0\fR to disable. .RE .sp .ne 2 .na \fBzfs_arc_shrink_shift\fR (int) .ad .RS 12n If set to a non zero value, this will update arc_shrink_shift (default 7) with the new value. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_arc_pc_percent\fR (uint) .ad .RS 12n Percent of pagecache to reclaim arc to This tunable allows ZFS arc to play more nicely with the kernel's LRU pagecache. It can guarantee that the arc size won't collapse under scanning pressure on the pagecache, yet still allows arc to be reclaimed down to zfs_arc_min if necessary. This value is specified as percent of pagecache size (as measured by NR_FILE_PAGES) where that percent may exceed 100. This only operates during memory pressure/reclaim. .sp Default value: \fB0\fR% (disabled). .RE .sp .ne 2 .na \fBzfs_arc_sys_free\fR (ulong) .ad .RS 12n The target number of bytes the ARC should leave as free memory on the system. Defaults to the larger of 1/64 of physical memory or 512K. Setting this option to a non-zero value will override the default. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_autoimport_disable\fR (int) .ad .RS 12n Disable pool import at module load by ignoring the cache file (typically \fB/etc/zfs/zpool.cache\fR). .sp Use \fB1\fR for yes (default) and \fB0\fR for no. .RE .sp .ne 2 .na \fBzfs_checksums_per_second\fR (int) .ad .RS 12n Rate limit checksum events to this many per second. Note that this should not be set below the zed thresholds (currently 10 checksums over 10 sec) or else zed may not trigger any action. .sp Default value: 20 .RE .sp .ne 2 .na \fBzfs_commit_timeout_pct\fR (int) .ad .RS 12n This controls the amount of time that a ZIL block (lwb) will remain "open" when it isn't "full", and it has a thread waiting for it to be committed to stable storage. The timeout is scaled based on a percentage of the last lwb latency to avoid significantly impacting the latency of each individual transaction record (itx). .sp Default value: \fB5\fR%. .RE .sp .ne 2 .na \fBzfs_condense_indirect_vdevs_enable\fR (int) .ad .RS 12n Enable condensing indirect vdev mappings. When set to a non-zero value, attempt to condense indirect vdev mappings if the mapping uses more than \fBzfs_condense_min_mapping_bytes\fR bytes of memory and if the obsolete space map object uses more than \fBzfs_condense_max_obsolete_bytes\fR bytes on-disk. The condensing process is an attempt to save memory by removing obsolete mappings. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_condense_max_obsolete_bytes\fR (ulong) .ad .RS 12n Only attempt to condense indirect vdev mappings if the on-disk size of the obsolete space map object is greater than this number of bytes (see \fBfBzfs_condense_indirect_vdevs_enable\fR). .sp Default value: \fB1,073,741,824\fR. .RE .sp .ne 2 .na \fBzfs_condense_min_mapping_bytes\fR (ulong) .ad .RS 12n Minimum size vdev mapping to attempt to condense (see \fBzfs_condense_indirect_vdevs_enable\fR). .sp Default value: \fB131,072\fR. .RE .sp .ne 2 .na \fBzfs_dbgmsg_enable\fR (int) .ad .RS 12n Internally ZFS keeps a small log to facilitate debugging. By default the log is disabled, to enable it set this option to 1. The contents of the log can be accessed by reading the /proc/spl/kstat/zfs/dbgmsg file. Writing 0 to this proc file clears the log. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_dbgmsg_maxsize\fR (int) .ad .RS 12n The maximum size in bytes of the internal ZFS debug log. .sp Default value: \fB4M\fR. .RE .sp .ne 2 .na \fBzfs_dbuf_state_index\fR (int) .ad .RS 12n This feature is currently unused. It is normally used for controlling what reporting is available under /proc/spl/kstat/zfs. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_deadman_enabled\fR (int) .ad .RS 12n When a pool sync operation takes longer than \fBzfs_deadman_synctime_ms\fR milliseconds, or when an individual I/O takes longer than \fBzfs_deadman_ziotime_ms\fR milliseconds, then the operation is considered to be "hung". If \fBzfs_deadman_enabled\fR is set then the deadman behavior is invoked as described by the \fBzfs_deadman_failmode\fR module option. By default the deadman is enabled and configured to \fBwait\fR which results in "hung" I/Os only being logged. The deadman is automatically disabled when a pool gets suspended. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_deadman_failmode\fR (charp) .ad .RS 12n Controls the failure behavior when the deadman detects a "hung" I/O. Valid values are \fBwait\fR, \fBcontinue\fR, and \fBpanic\fR. .sp \fBwait\fR - Wait for a "hung" I/O to complete. For each "hung" I/O a "deadman" event will be posted describing that I/O. .sp \fBcontinue\fR - Attempt to recover from a "hung" I/O by re-dispatching it to the I/O pipeline if possible. .sp \fBpanic\fR - Panic the system. This can be used to facilitate an automatic fail-over to a properly configured fail-over partner. .sp Default value: \fBwait\fR. .RE .sp .ne 2 .na \fBzfs_deadman_checktime_ms\fR (int) .ad .RS 12n Check time in milliseconds. This defines the frequency at which we check for hung I/O and potentially invoke the \fBzfs_deadman_failmode\fR behavior. .sp Default value: \fB60,000\fR. .RE .sp .ne 2 .na \fBzfs_deadman_synctime_ms\fR (ulong) .ad .RS 12n Interval in milliseconds after which the deadman is triggered and also the interval after which a pool sync operation is considered to be "hung". Once this limit is exceeded the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR milliseconds until the pool sync completes. .sp Default value: \fB600,000\fR. .RE .sp .ne 2 .na \fBzfs_deadman_ziotime_ms\fR (ulong) .ad .RS 12n Interval in milliseconds after which the deadman is triggered and an individual IO operation is considered to be "hung". As long as the I/O remains "hung" the deadman will be invoked every \fBzfs_deadman_checktime_ms\fR milliseconds until the I/O completes. .sp Default value: \fB300,000\fR. .RE .sp .ne 2 .na \fBzfs_dedup_prefetch\fR (int) .ad .RS 12n Enable prefetching dedup-ed blks .sp Use \fB1\fR for yes and \fB0\fR to disable (default). .RE .sp .ne 2 .na \fBzfs_delay_min_dirty_percent\fR (int) .ad .RS 12n Start to delay each transaction once there is this amount of dirty data, expressed as a percentage of \fBzfs_dirty_data_max\fR. This value should be >= zfs_vdev_async_write_active_max_dirty_percent. See the section "ZFS TRANSACTION DELAY". .sp Default value: \fB60\fR%. .RE .sp .ne 2 .na \fBzfs_delay_scale\fR (int) .ad .RS 12n This controls how quickly the transaction delay approaches infinity. Larger values cause longer delays for a given amount of dirty data. .sp For the smoothest delay, this value should be about 1 billion divided by the maximum number of operations per second. This will smoothly handle between 10x and 1/10th this number. .sp See the section "ZFS TRANSACTION DELAY". .sp Note: \fBzfs_delay_scale\fR * \fBzfs_dirty_data_max\fR must be < 2^64. .sp Default value: \fB500,000\fR. .RE .sp .ne 2 .na \fBzfs_delays_per_second\fR (int) .ad .RS 12n Rate limit IO delay events to this many per second. .sp Default value: 20 .RE .sp .ne 2 .na \fBzfs_delete_blocks\fR (ulong) .ad .RS 12n This is the used to define a large file for the purposes of delete. Files containing more than \fBzfs_delete_blocks\fR will be deleted asynchronously while smaller files are deleted synchronously. Decreasing this value will reduce the time spent in an unlink(2) system call at the expense of a longer delay before the freed space is available. .sp Default value: \fB20,480\fR. .RE .sp .ne 2 .na \fBzfs_dirty_data_max\fR (int) .ad .RS 12n Determines the dirty space limit in bytes. Once this limit is exceeded, new writes are halted until space frees up. This parameter takes precedence over \fBzfs_dirty_data_max_percent\fR. See the section "ZFS TRANSACTION DELAY". .sp Default value: \fB10\fR% of physical RAM, capped at \fBzfs_dirty_data_max_max\fR. .RE .sp .ne 2 .na \fBzfs_dirty_data_max_max\fR (int) .ad .RS 12n Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed in bytes. This limit is only enforced at module load time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed. This parameter takes precedence over \fBzfs_dirty_data_max_max_percent\fR. See the section "ZFS TRANSACTION DELAY". .sp Default value: \fB25\fR% of physical RAM. .RE .sp .ne 2 .na \fBzfs_dirty_data_max_max_percent\fR (int) .ad .RS 12n Maximum allowable value of \fBzfs_dirty_data_max\fR, expressed as a percentage of physical RAM. This limit is only enforced at module load time, and will be ignored if \fBzfs_dirty_data_max\fR is later changed. The parameter \fBzfs_dirty_data_max_max\fR takes precedence over this one. See the section "ZFS TRANSACTION DELAY". .sp Default value: \fB25\fR%. .RE .sp .ne 2 .na \fBzfs_dirty_data_max_percent\fR (int) .ad .RS 12n Determines the dirty space limit, expressed as a percentage of all memory. Once this limit is exceeded, new writes are halted until space frees up. The parameter \fBzfs_dirty_data_max\fR takes precedence over this one. See the section "ZFS TRANSACTION DELAY". .sp Default value: \fB10\fR%, subject to \fBzfs_dirty_data_max_max\fR. .RE .sp .ne 2 .na \fBzfs_dirty_data_sync\fR (int) .ad .RS 12n Start syncing out a transaction group if there is at least this much dirty data. .sp Default value: \fB67,108,864\fR. .RE .sp .ne 2 .na \fBzfs_fletcher_4_impl\fR (string) .ad .RS 12n Select a fletcher 4 implementation. .sp Supported selectors are: \fBfastest\fR, \fBscalar\fR, \fBsse2\fR, \fBssse3\fR, \fBavx2\fR, \fBavx512f\fR, and \fBaarch64_neon\fR. All of the selectors except \fBfastest\fR and \fBscalar\fR require instruction set extensions to be available and will only appear if ZFS detects that they are present at runtime. If multiple implementations of fletcher 4 are available, the \fBfastest\fR will be chosen using a micro benchmark. Selecting \fBscalar\fR results in the original, CPU based calculation, being used. Selecting any option other than \fBfastest\fR and \fBscalar\fR results in vector instructions from the respective CPU instruction set being used. .sp Default value: \fBfastest\fR. .RE .sp .ne 2 .na \fBzfs_free_bpobj_enabled\fR (int) .ad .RS 12n Enable/disable the processing of the free_bpobj object. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_async_block_max_blocks\fR (ulong) .ad .RS 12n Maximum number of blocks freed in a single txg. .sp Default value: \fB100,000\fR. .RE .sp .ne 2 .na \fBzfs_override_estimate_recordsize\fR (ulong) .ad .RS 12n Record size calculation override for zfs send estimates. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_vdev_async_read_max_active\fR (int) .ad .RS 12n Maximum asynchronous read I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB3\fR. .RE .sp .ne 2 .na \fBzfs_vdev_async_read_min_active\fR (int) .ad .RS 12n Minimum asynchronous read I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_vdev_async_write_active_max_dirty_percent\fR (int) .ad .RS 12n When the pool has more than \fBzfs_vdev_async_write_active_max_dirty_percent\fR dirty data, use \fBzfs_vdev_async_write_max_active\fR to limit active async writes. If the dirty data is between min and max, the active I/O limit is linearly interpolated. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB60\fR%. .RE .sp .ne 2 .na \fBzfs_vdev_async_write_active_min_dirty_percent\fR (int) .ad .RS 12n When the pool has less than \fBzfs_vdev_async_write_active_min_dirty_percent\fR dirty data, use \fBzfs_vdev_async_write_min_active\fR to limit active async writes. If the dirty data is between min and max, the active I/O limit is linearly interpolated. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB30\fR%. .RE .sp .ne 2 .na \fBzfs_vdev_async_write_max_active\fR (int) .ad .RS 12n Maximum asynchronous write I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB10\fR. .RE .sp .ne 2 .na \fBzfs_vdev_async_write_min_active\fR (int) .ad .RS 12n Minimum asynchronous write I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Lower values are associated with better latency on rotational media but poorer resilver performance. The default value of 2 was chosen as a compromise. A value of 3 has been shown to improve resilver performance further at a cost of further increasing latency. .sp Default value: \fB2\fR. .RE .sp .ne 2 .na \fBzfs_vdev_max_active\fR (int) .ad .RS 12n The maximum number of I/Os active to each device. Ideally, this will be >= the sum of each queue's max_active. It must be at least the sum of each queue's min_active. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB1,000\fR. .RE .sp .ne 2 .na \fBzfs_vdev_scrub_max_active\fR (int) .ad .RS 12n Maximum scrub I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB2\fR. .RE .sp .ne 2 .na \fBzfs_vdev_scrub_min_active\fR (int) .ad .RS 12n Minimum scrub I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_vdev_sync_read_max_active\fR (int) .ad .RS 12n Maximum synchronous read I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB10\fR. .RE .sp .ne 2 .na \fBzfs_vdev_sync_read_min_active\fR (int) .ad .RS 12n Minimum synchronous read I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB10\fR. .RE .sp .ne 2 .na \fBzfs_vdev_sync_write_max_active\fR (int) .ad .RS 12n Maximum synchronous write I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB10\fR. .RE .sp .ne 2 .na \fBzfs_vdev_sync_write_min_active\fR (int) .ad .RS 12n Minimum synchronous write I/Os active to each device. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB10\fR. .RE .sp .ne 2 .na \fBzfs_vdev_queue_depth_pct\fR (int) .ad .RS 12n Maximum number of queued allocations per top-level vdev expressed as a percentage of \fBzfs_vdev_async_write_max_active\fR which allows the system to detect devices that are more capable of handling allocations and to allocate more blocks to those devices. It allows for dynamic allocation distribution when devices are imbalanced as fuller devices will tend to be slower than empty devices. See also \fBzio_dva_throttle_enabled\fR. .sp Default value: \fB1000\fR%. .RE .sp .ne 2 .na \fBzfs_expire_snapshot\fR (int) .ad .RS 12n Seconds to expire .zfs/snapshot .sp Default value: \fB300\fR. .RE .sp .ne 2 .na \fBzfs_admin_snapshot\fR (int) .ad .RS 12n Allow the creation, removal, or renaming of entries in the .zfs/snapshot directory to cause the creation, destruction, or renaming of snapshots. When enabled this functionality works both locally and over NFS exports which have the 'no_root_squash' option set. This functionality is disabled by default. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_flags\fR (int) .ad .RS 12n Set additional debugging flags. The following flags may be bitwise-or'd together. .sp .TS box; rB lB lB lB r l. Value Symbolic Name Description _ 1 ZFS_DEBUG_DPRINTF Enable dprintf entries in the debug log. _ 2 ZFS_DEBUG_DBUF_VERIFY * Enable extra dbuf verifications. _ 4 ZFS_DEBUG_DNODE_VERIFY * Enable extra dnode verifications. _ 8 ZFS_DEBUG_SNAPNAMES Enable snapshot name verification. _ 16 ZFS_DEBUG_MODIFY Check for illegally modified ARC buffers. _ 64 ZFS_DEBUG_ZIO_FREE Enable verification of block frees. _ 128 ZFS_DEBUG_HISTOGRAM_VERIFY Enable extra spacemap histogram verifications. _ 256 ZFS_DEBUG_METASLAB_VERIFY Verify space accounting on disk matches in-core range_trees. _ 512 ZFS_DEBUG_SET_ERROR Enable SET_ERROR and dprintf entries in the debug log. .TE .sp * Requires debug build. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_free_leak_on_eio\fR (int) .ad .RS 12n If destroy encounters an EIO while reading metadata (e.g. indirect blocks), space referenced by the missing metadata can not be freed. Normally this causes the background destroy to become "stalled", as it is unable to make forward progress. While in this stalled state, all remaining space to free from the error-encountering filesystem is "temporarily leaked". Set this flag to cause it to ignore the EIO, permanently leak the space from indirect blocks that can not be read, and continue to free everything else that it can. The default, "stalling" behavior is useful if the storage partially fails (i.e. some but not all i/os fail), and then later recovers. In this case, we will be able to continue pool operations while it is partially failed, and when it recovers, we can continue to free the space, with no leaks. However, note that this case is actually fairly rare. Typically pools either (a) fail completely (but perhaps temporarily, e.g. a top-level vdev going offline), or (b) have localized, permanent errors (e.g. disk returns the wrong data due to bit flip or firmware bug). In case (a), this setting does not matter because the pool will be suspended and the sync thread will not be able to make forward progress regardless. In case (b), because the error is permanent, the best we can do is leak the minimum amount of space, which is what setting this flag will do. Therefore, it is reasonable for this flag to normally be set, but we chose the more conservative approach of not setting it, so that there is no possibility of leaking space in the "partial temporary" failure case. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_free_min_time_ms\fR (int) .ad .RS 12n During a \fBzfs destroy\fR operation using \fBfeature@async_destroy\fR a minimum of this much time will be spent working on freeing blocks per txg. .sp Default value: \fB1,000\fR. .RE .sp .ne 2 .na \fBzfs_immediate_write_sz\fR (long) .ad .RS 12n Largest data block to write to zil. Larger blocks will be treated as if the dataset being written to had the property setting \fBlogbias=throughput\fR. .sp Default value: \fB32,768\fR. .RE .sp .ne 2 .na \fBzfs_lua_max_instrlimit\fR (ulong) .ad .RS 12n The maximum execution time limit that can be set for a ZFS channel program, specified as a number of Lua instructions. .sp Default value: \fB100,000,000\fR. .RE .sp .ne 2 .na \fBzfs_lua_max_memlimit\fR (ulong) .ad .RS 12n The maximum memory limit that can be set for a ZFS channel program, specified in bytes. .sp Default value: \fB104,857,600\fR. .RE .sp .ne 2 .na \fBzfs_max_dataset_nesting\fR (int) .ad .RS 12n The maximum depth of nested datasets. This value can be tuned temporarily to fix existing datasets that exceed the predefined limit. .sp Default value: \fB50\fR. .RE .sp .ne 2 .na \fBzfs_max_recordsize\fR (int) .ad .RS 12n We currently support block sizes from 512 bytes to 16MB. The benefits of larger blocks, and thus larger IO, need to be weighed against the cost of COWing a giant block to modify one byte. Additionally, very large blocks can have an impact on i/o latency, and also potentially on the memory allocator. Therefore, we do not allow the recordsize to be set larger than zfs_max_recordsize (default 1MB). Larger blocks can be created by changing this tunable, and pools with larger blocks can always be imported and used, regardless of this setting. .sp Default value: \fB1,048,576\fR. .RE .sp .ne 2 .na \fBzfs_metaslab_fragmentation_threshold\fR (int) .ad .RS 12n Allow metaslabs to keep their active state as long as their fragmentation percentage is less than or equal to this value. An active metaslab that exceeds this threshold will no longer keep its active status allowing better metaslabs to be selected. .sp Default value: \fB70\fR. .RE .sp .ne 2 .na \fBzfs_mg_fragmentation_threshold\fR (int) .ad .RS 12n Metaslab groups are considered eligible for allocations if their fragmentation metric (measured as a percentage) is less than or equal to this value. If a metaslab group exceeds this threshold then it will be skipped unless all metaslab groups within the metaslab class have also crossed this threshold. .sp Default value: \fB85\fR. .RE .sp .ne 2 .na \fBzfs_mg_noalloc_threshold\fR (int) .ad .RS 12n Defines a threshold at which metaslab groups should be eligible for allocations. The value is expressed as a percentage of free space beyond which a metaslab group is always eligible for allocations. If a metaslab group's free space is less than or equal to the threshold, the allocator will avoid allocating to that group unless all groups in the pool have reached the threshold. Once all groups have reached the threshold, all groups are allowed to accept allocations. The default value of 0 disables the feature and causes all metaslab groups to be eligible for allocations. This parameter allows one to deal with pools having heavily imbalanced vdevs such as would be the case when a new vdev has been added. Setting the threshold to a non-zero percentage will stop allocations from being made to vdevs that aren't filled to the specified percentage and allow lesser filled vdevs to acquire more allocations than they otherwise would under the old \fBzfs_mg_alloc_failures\fR facility. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_multihost_history\fR (int) .ad .RS 12n Historical statistics for the last N multihost updates will be available in \fB/proc/spl/kstat/zfs//multihost\fR .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_multihost_interval\fR (ulong) .ad .RS 12n Used to control the frequency of multihost writes which are performed when the \fBmultihost\fR pool property is on. This is one factor used to determine the length of the activity check during import. .sp The multihost write period is \fBzfs_multihost_interval / leaf-vdevs\fR milliseconds. This means that on average a multihost write will be issued for each leaf vdev every \fBzfs_multihost_interval\fR milliseconds. In practice, the observed period can vary with the I/O load and this observed value is the delay which is stored in the uberblock. .sp On import the activity check waits a minimum amount of time determined by \fBzfs_multihost_interval * zfs_multihost_import_intervals\fR. The activity check time may be further extended if the value of mmp delay found in the best uberblock indicates actual multihost updates happened at longer intervals than \fBzfs_multihost_interval\fR. A minimum value of \fB100ms\fR is enforced. .sp Default value: \fB1000\fR. .RE .sp .ne 2 .na \fBzfs_multihost_import_intervals\fR (uint) .ad .RS 12n Used to control the duration of the activity test on import. Smaller values of \fBzfs_multihost_import_intervals\fR will reduce the import time but increase the risk of failing to detect an active pool. The total activity check time is never allowed to drop below one second. A value of 0 is ignored and treated as if it was set to 1 .sp Default value: \fB10\fR. .RE .sp .ne 2 .na \fBzfs_multihost_fail_intervals\fR (uint) .ad .RS 12n Controls the behavior of the pool when multihost write failures are detected. .sp When \fBzfs_multihost_fail_intervals = 0\fR then multihost write failures are ignored. The failures will still be reported to the ZED which depending on its configuration may take action such as suspending the pool or offlining a device. .sp When \fBzfs_multihost_fail_intervals > 0\fR then sequential multihost write failures will cause the pool to be suspended. This occurs when \fBzfs_multihost_fail_intervals * zfs_multihost_interval\fR milliseconds have passed since the last successful multihost write. This guarantees the activity test will see multihost writes if the pool is imported. .sp Default value: \fB5\fR. .RE .sp .ne 2 .na \fBzfs_no_scrub_io\fR (int) .ad .RS 12n Set for no scrub I/O. This results in scrubs not actually scrubbing data and simply doing a metadata crawl of the pool instead. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_no_scrub_prefetch\fR (int) .ad .RS 12n Set to disable block prefetching for scrubs. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_nocacheflush\fR (int) .ad .RS 12n Disable cache flush operations on disks when writing. Beware, this may cause corruption if disks re-order writes. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_nopwrite_enabled\fR (int) .ad .RS 12n Enable NOP writes .sp Use \fB1\fR for yes (default) and \fB0\fR to disable. .RE .sp .ne 2 .na \fBzfs_dmu_offset_next_sync\fR (int) .ad .RS 12n Enable forcing txg sync to find holes. When enabled forces ZFS to act like prior versions when SEEK_HOLE or SEEK_DATA flags are used, which when a dnode is dirty causes txg's to be synced so that this data can be found. .sp Use \fB1\fR for yes and \fB0\fR to disable (default). .RE .sp .ne 2 .na \fBzfs_pd_bytes_max\fR (int) .ad .RS 12n The number of bytes which should be prefetched during a pool traversal (eg: \fBzfs send\fR or other data crawling operations) .sp Default value: \fB52,428,800\fR. .RE .sp .ne 2 .na \fBzfs_per_txg_dirty_frees_percent \fR (ulong) .ad .RS 12n Tunable to control percentage of dirtied blocks from frees in one TXG. After this threshold is crossed, additional dirty blocks from frees wait until the next TXG. A value of zero will disable this throttle. .sp Default value: \fB30\fR and \fB0\fR to disable. .RE .sp .ne 2 .na \fBzfs_prefetch_disable\fR (int) .ad .RS 12n This tunable disables predictive prefetch. Note that it leaves "prescient" prefetch (e.g. prefetch for zfs send) intact. Unlike predictive prefetch, prescient prefetch never issues i/os that end up not being needed, so it can't hurt performance. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_read_chunk_size\fR (long) .ad .RS 12n Bytes to read per chunk .sp Default value: \fB1,048,576\fR. .RE .sp .ne 2 .na \fBzfs_read_history\fR (int) .ad .RS 12n Historical statistics for the last N reads will be available in \fB/proc/spl/kstat/zfs//reads\fR .sp Default value: \fB0\fR (no data is kept). .RE .sp .ne 2 .na \fBzfs_read_history_hits\fR (int) .ad .RS 12n Include cache hits in read history .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_reconstruct_indirect_combinations_max\fR (int) .ad .RS 12na If an indirect split block contains more than this many possible unique combinations when being reconstructed, consider it too computationally expensive to check them all. Instead, try at most \fBzfs_reconstruct_indirect_combinations_max\fR randomly-selected combinations each time the block is accessed. This allows all segment copies to participate fairly in the reconstruction when all combinations cannot be checked and prevents repeated use of one bad copy. .sp Default value: \fB100\fR. .RE .sp .ne 2 .na \fBzfs_recover\fR (int) .ad .RS 12n Set to attempt to recover from fatal errors. This should only be used as a last resort, as it typically results in leaked space, or worse. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_resilver_min_time_ms\fR (int) .ad .RS 12n Resilvers are processed by the sync thread. While resilvering it will spend at least this much time working on a resilver between txg flushes. .sp Default value: \fB3,000\fR. .RE .sp .ne 2 .na \fBzfs_scan_ignore_errors\fR (int) .ad .RS 12n If set to a nonzero value, remove the DTL (dirty time list) upon completion of a pool scan (scrub) even if there were unrepairable errors. It is intended to be used during pool repair or recovery to stop resilvering when the pool is next imported. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_scrub_min_time_ms\fR (int) .ad .RS 12n Scrubs are processed by the sync thread. While scrubbing it will spend at least this much time working on a scrub between txg flushes. .sp Default value: \fB1,000\fR. .RE .sp .ne 2 .na \fBzfs_scan_checkpoint_intval\fR (int) .ad .RS 12n To preserve progress across reboots the sequential scan algorithm periodically needs to stop metadata scanning and issue all the verifications I/Os to disk. The frequency of this flushing is determined by the \fBzfs_scan_checkpoint_intval\fR tunable. .sp Default value: \fB7200\fR seconds (every 2 hours). .RE .sp .ne 2 .na \fBzfs_scan_fill_weight\fR (int) .ad .RS 12n This tunable affects how scrub and resilver I/O segments are ordered. A higher number indicates that we care more about how filled in a segment is, while a lower number indicates we care more about the size of the extent without considering the gaps within a segment. This value is only tunable upon module insertion. Changing the value afterwards will have no affect on scrub or resilver performance. .sp Default value: \fB3\fR. .RE .sp .ne 2 .na \fBzfs_scan_issue_strategy\fR (int) .ad .RS 12n Determines the order that data will be verified while scrubbing or resilvering. If set to \fB1\fR, data will be verified as sequentially as possible, given the amount of memory reserved for scrubbing (see \fBzfs_scan_mem_lim_fact\fR). This may improve scrub performance if the pool's data is very fragmented. If set to \fB2\fR, the largest mostly-contiguous chunk of found data will be verified first. By deferring scrubbing of small segments, we may later find adjacent data to coalesce and increase the segment size. If set to \fB0\fR, zfs will use strategy \fB1\fR during normal verification and strategy \fB2\fR while taking a checkpoint. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_scan_legacy\fR (int) .ad .RS 12n A value of 0 indicates that scrubs and resilvers will gather metadata in memory before issuing sequential I/O. A value of 1 indicates that the legacy algorithm will be used where I/O is initiated as soon as it is discovered. Changing this value to 0 will not affect scrubs or resilvers that are already in progress. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_scan_max_ext_gap\fR (int) .ad .RS 12n Indicates the largest gap in bytes between scrub / resilver I/Os that will still be considered sequential for sorting purposes. Changing this value will not affect scrubs or resilvers that are already in progress. .sp Default value: \fB2097152 (2 MB)\fR. .RE .sp .ne 2 .na \fBzfs_scan_mem_lim_fact\fR (int) .ad .RS 12n Maximum fraction of RAM used for I/O sorting by sequential scan algorithm. This tunable determines the hard limit for I/O sorting memory usage. When the hard limit is reached we stop scanning metadata and start issuing data verification I/O. This is done until we get below the soft limit. .sp Default value: \fB20\fR which is 5% of RAM (1/20). .RE .sp .ne 2 .na \fBzfs_scan_mem_lim_soft_fact\fR (int) .ad .RS 12n The fraction of the hard limit used to determined the soft limit for I/O sorting by the sequential scan algorithm. When we cross this limit from bellow no action is taken. When we cross this limit from above it is because we are issuing verification I/O. In this case (unless the metadata scan is done) we stop issuing verification I/O and start scanning metadata again until we get to the hard limit. .sp Default value: \fB20\fR which is 5% of the hard limit (1/20). .RE .sp .ne 2 .na \fBzfs_scan_vdev_limit\fR (int) .ad .RS 12n Maximum amount of data that can be concurrently issued at once for scrubs and resilvers per leaf device, given in bytes. .sp Default value: \fB41943040\fR. .RE .sp .ne 2 .na \fBzfs_send_corrupt_data\fR (int) .ad .RS 12n Allow sending of corrupt data (ignore read/checksum errors when sending data) .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_send_queue_length\fR (int) .ad .RS 12n The maximum number of bytes allowed in the \fBzfs send\fR queue. This value must be at least twice the maximum block size in use. .sp Default value: \fB16,777,216\fR. .RE .sp .ne 2 .na \fBzfs_recv_queue_length\fR (int) .ad .RS 12n .sp The maximum number of bytes allowed in the \fBzfs receive\fR queue. This value must be at least twice the maximum block size in use. .sp Default value: \fB16,777,216\fR. .RE .sp .ne 2 .na \fBzfs_sync_pass_deferred_free\fR (int) .ad .RS 12n Flushing of data to disk is done in passes. Defer frees starting in this pass .sp Default value: \fB2\fR. .RE .sp .ne 2 .na \fBzfs_spa_discard_memory_limit\fR (int) .ad .RS 12n Maximum memory used for prefetching a checkpoint's space map on each vdev while discarding the checkpoint. .sp Default value: \fB16,777,216\fR. .RE .sp .ne 2 .na \fBzfs_sync_pass_dont_compress\fR (int) .ad .RS 12n Don't compress starting in this pass .sp Default value: \fB5\fR. .RE .sp .ne 2 .na \fBzfs_sync_pass_rewrite\fR (int) .ad .RS 12n Rewrite new block pointers starting in this pass .sp Default value: \fB2\fR. .RE .sp .ne 2 .na \fBzfs_sync_taskq_batch_pct\fR (int) .ad .RS 12n This controls the number of threads used by the dp_sync_taskq. The default value of 75% will create a maximum of one thread per cpu. .sp Default value: \fB75\fR%. .RE .sp .ne 2 .na \fBzfs_txg_history\fR (int) .ad .RS 12n Historical statistics for the last N txgs will be available in \fB/proc/spl/kstat/zfs//txgs\fR .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_txg_timeout\fR (int) .ad .RS 12n Flush dirty data to disk at least every N seconds (maximum txg duration) .sp Default value: \fB5\fR. .RE .sp .ne 2 .na \fBzfs_vdev_aggregation_limit\fR (int) .ad .RS 12n Max vdev I/O aggregation size .sp Default value: \fB131,072\fR. .RE .sp .ne 2 .na \fBzfs_vdev_cache_bshift\fR (int) .ad .RS 12n Shift size to inflate reads too .sp Default value: \fB16\fR (effectively 65536). .RE .sp .ne 2 .na \fBzfs_vdev_cache_max\fR (int) .ad .RS 12n Inflate reads smaller than this value to meet the \fBzfs_vdev_cache_bshift\fR size (default 64k). .sp Default value: \fB16384\fR. .RE .sp .ne 2 .na \fBzfs_vdev_cache_size\fR (int) .ad .RS 12n Total size of the per-disk cache in bytes. .sp Currently this feature is disabled as it has been found to not be helpful for performance and in some cases harmful. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_vdev_mirror_rotating_inc\fR (int) .ad .RS 12n A number by which the balancing algorithm increments the load calculation for the purpose of selecting the least busy mirror member when an I/O immediately follows its predecessor on rotational vdevs for the purpose of making decisions based on load. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_vdev_mirror_rotating_seek_inc\fR (int) .ad .RS 12n A number by which the balancing algorithm increments the load calculation for the purpose of selecting the least busy mirror member when an I/O lacks locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within this that are not immediately following the previous I/O are incremented by half. .sp Default value: \fB5\fR. .RE .sp .ne 2 .na \fBzfs_vdev_mirror_rotating_seek_offset\fR (int) .ad .RS 12n The maximum distance for the last queued I/O in which the balancing algorithm considers an I/O to have locality. See the section "ZFS I/O SCHEDULER". .sp Default value: \fB1048576\fR. .RE .sp .ne 2 .na \fBzfs_vdev_mirror_non_rotating_inc\fR (int) .ad .RS 12n A number by which the balancing algorithm increments the load calculation for the purpose of selecting the least busy mirror member on non-rotational vdevs when I/Os do not immediately follow one another. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_vdev_mirror_non_rotating_seek_inc\fR (int) .ad .RS 12n A number by which the balancing algorithm increments the load calculation for the purpose of selecting the least busy mirror member when an I/O lacks locality as defined by the zfs_vdev_mirror_rotating_seek_offset. I/Os within this that are not immediately following the previous I/O are incremented by half. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_vdev_read_gap_limit\fR (int) .ad .RS 12n Aggregate read I/O operations if the gap on-disk between them is within this threshold. .sp Default value: \fB32,768\fR. .RE .sp .ne 2 .na \fBzfs_vdev_scheduler\fR (charp) .ad .RS 12n Set the Linux I/O scheduler on whole disk vdevs to this scheduler. Valid options are noop, cfq, bfq & deadline .sp Default value: \fBnoop\fR. .RE .sp .ne 2 .na \fBzfs_vdev_write_gap_limit\fR (int) .ad .RS 12n Aggregate write I/O over gap .sp Default value: \fB4,096\fR. .RE .sp .ne 2 .na \fBzfs_vdev_raidz_impl\fR (string) .ad .RS 12n Parameter for selecting raidz parity implementation to use. Options marked (always) below may be selected on module load as they are supported on all systems. The remaining options may only be set after the module is loaded, as they are available only if the implementations are compiled in and supported on the running system. Once the module is loaded, the content of /sys/module/zfs/parameters/zfs_vdev_raidz_impl will show available options with the currently selected one enclosed in []. Possible options are: fastest - (always) implementation selected using built-in benchmark original - (always) original raidz implementation scalar - (always) scalar raidz implementation sse2 - implementation using SSE2 instruction set (64bit x86 only) ssse3 - implementation using SSSE3 instruction set (64bit x86 only) avx2 - implementation using AVX2 instruction set (64bit x86 only) avx512f - implementation using AVX512F instruction set (64bit x86 only) avx512bw - implementation using AVX512F & AVX512BW instruction sets (64bit x86 only) aarch64_neon - implementation using NEON (Aarch64/64 bit ARMv8 only) aarch64_neonx2 - implementation using NEON with more unrolling (Aarch64/64 bit ARMv8 only) .sp Default value: \fBfastest\fR. .RE .sp .ne 2 .na \fBzfs_zevent_cols\fR (int) .ad .RS 12n When zevents are logged to the console use this as the word wrap width. .sp Default value: \fB80\fR. .RE .sp .ne 2 .na \fBzfs_zevent_console\fR (int) .ad .RS 12n Log events to the console .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzfs_zevent_len_max\fR (int) .ad .RS 12n Max event queue length. A value of 0 will result in a calculated value which increases with the number of CPUs in the system (minimum 64 events). Events in the queue can be viewed with the \fBzpool events\fR command. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzfs_zil_clean_taskq_maxalloc\fR (int) .ad .RS 12n The maximum number of taskq entries that are allowed to be cached. When this limit is exceeded transaction records (itxs) will be cleaned synchronously. .sp Default value: \fB1048576\fR. .RE .sp .ne 2 .na \fBzfs_zil_clean_taskq_minalloc\fR (int) .ad .RS 12n The number of taskq entries that are pre-populated when the taskq is first created and are immediately available for use. .sp Default value: \fB1024\fR. .RE .sp .ne 2 .na \fBzfs_zil_clean_taskq_nthr_pct\fR (int) .ad .RS 12n This controls the number of threads used by the dp_zil_clean_taskq. The default value of 100% will create a maximum of one thread per cpu. .sp Default value: \fB100\fR%. .RE .sp .ne 2 .na \fBzil_replay_disable\fR (int) .ad .RS 12n Disable intent logging replay. Can be disabled for recovery from corrupted ZIL .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzil_slog_bulk\fR (ulong) .ad .RS 12n Limit SLOG write size per commit executed with synchronous priority. Any writes above that will be executed with lower (asynchronous) priority to limit potential SLOG device abuse by single active ZIL writer. .sp Default value: \fB786,432\fR. .RE +.sp +.ne 2 +.na +\fBzio_decompress_fail_fraction\fR (int) +.ad +.RS 12n +If non-zero, this value represents the denominator of the probability that zfs +should induce a decompression failure. For instance, for a 5% decompression +failure rate, this value should be set to 20. +.sp +Default value: \fB0\fR. +.RE + .sp .ne 2 .na \fBzio_delay_max\fR (int) .ad .RS 12n A zevent will be logged if a ZIO operation takes more than N milliseconds to complete. Note that this is only a logging facility, not a timeout on operations. .sp Default value: \fB30,000\fR. .RE .sp .ne 2 .na \fBzio_dva_throttle_enabled\fR (int) .ad .RS 12n Throttle block allocations in the ZIO pipeline. This allows for dynamic allocation distribution when devices are imbalanced. When enabled, the maximum number of pending allocations per top-level vdev is limited by \fBzfs_vdev_queue_depth_pct\fR. .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzio_requeue_io_start_cut_in_line\fR (int) .ad .RS 12n Prioritize requeued I/O .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzio_taskq_batch_pct\fR (uint) .ad .RS 12n Percentage of online CPUs (or CPU cores, etc) which will run a worker thread for IO. These workers are responsible for IO work such as compression and checksum calculations. Fractional number of CPUs will be rounded down. .sp The default value of 75 was chosen to avoid using all CPUs which can result in latency issues and inconsistent application performance, especially when high compression is enabled. .sp Default value: \fB75\fR. .RE .sp .ne 2 .na \fBzvol_inhibit_dev\fR (uint) .ad .RS 12n Do not create zvol device nodes. This may slightly improve startup time on systems with a very large number of zvols. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .sp .ne 2 .na \fBzvol_major\fR (uint) .ad .RS 12n Major number for zvol block devices .sp Default value: \fB230\fR. .RE .sp .ne 2 .na \fBzvol_max_discard_blocks\fR (ulong) .ad .RS 12n Discard (aka TRIM) operations done on zvols will be done in batches of this many blocks, where block size is determined by the \fBvolblocksize\fR property of a zvol. .sp Default value: \fB16,384\fR. .RE .sp .ne 2 .na \fBzvol_prefetch_bytes\fR (uint) .ad .RS 12n When adding a zvol to the system prefetch \fBzvol_prefetch_bytes\fR from the start and end of the volume. Prefetching these regions of the volume is desirable because they are likely to be accessed immediately by \fBblkid(8)\fR or by the kernel scanning for a partition table. .sp Default value: \fB131,072\fR. .RE .sp .ne 2 .na \fBzvol_request_sync\fR (uint) .ad .RS 12n When processing I/O requests for a zvol submit them synchronously. This effectively limits the queue depth to 1 for each I/O submitter. When set to 0 requests are handled asynchronously by a thread pool. The number of requests which can be handled concurrently is controller by \fBzvol_threads\fR. .sp Default value: \fB0\fR. .RE .sp .ne 2 .na \fBzvol_threads\fR (uint) .ad .RS 12n Max number of threads which can handle zvol I/O requests concurrently. .sp Default value: \fB32\fR. .RE .sp .ne 2 .na \fBzvol_volmode\fR (uint) .ad .RS 12n Defines zvol block devices behaviour when \fBvolmode\fR is set to \fBdefault\fR. Valid values are \fB1\fR (full), \fB2\fR (dev) and \fB3\fR (none). .sp Default value: \fB1\fR. .RE .sp .ne 2 .na \fBzfs_qat_disable\fR (int) .ad .RS 12n This tunable disables qat hardware acceleration for gzip compression and. AES-GCM encryption. It is available only if qat acceleration is compiled in and the qat driver is present. .sp Use \fB1\fR for yes and \fB0\fR for no (default). .RE .SH ZFS I/O SCHEDULER ZFS issues I/O operations to leaf vdevs to satisfy and complete I/Os. The I/O scheduler determines when and in what order those operations are issued. The I/O scheduler divides operations into five I/O classes prioritized in the following order: sync read, sync write, async read, async write, and scrub/resilver. Each queue defines the minimum and maximum number of concurrent operations that may be issued to the device. In addition, the device has an aggregate maximum, \fBzfs_vdev_max_active\fR. Note that the sum of the per-queue minimums must not exceed the aggregate maximum. If the sum of the per-queue maximums exceeds the aggregate maximum, then the number of active I/Os may reach \fBzfs_vdev_max_active\fR, in which case no further I/Os will be issued regardless of whether all per-queue minimums have been met. .sp For many physical devices, throughput increases with the number of concurrent operations, but latency typically suffers. Further, physical devices typically have a limit at which more concurrent operations have no effect on throughput or can actually cause it to decrease. .sp The scheduler selects the next operation to issue by first looking for an I/O class whose minimum has not been satisfied. Once all are satisfied and the aggregate maximum has not been hit, the scheduler looks for classes whose maximum has not been satisfied. Iteration through the I/O classes is done in the order specified above. No further operations are issued if the aggregate maximum number of concurrent operations has been hit or if there are no operations queued for an I/O class that has not hit its maximum. Every time an I/O is queued or an operation completes, the I/O scheduler looks for new operations to issue. .sp In general, smaller max_active's will lead to lower latency of synchronous operations. Larger max_active's may lead to higher overall throughput, depending on underlying storage. .sp The ratio of the queues' max_actives determines the balance of performance between reads, writes, and scrubs. E.g., increasing \fBzfs_vdev_scrub_max_active\fR will cause the scrub or resilver to complete more quickly, but reads and writes to have higher latency and lower throughput. .sp All I/O classes have a fixed maximum number of outstanding operations except for the async write class. Asynchronous writes represent the data that is committed to stable storage during the syncing stage for transaction groups. Transaction groups enter the syncing state periodically so the number of queued async writes will quickly burst up and then bleed down to zero. Rather than servicing them as quickly as possible, the I/O scheduler changes the maximum number of active async write I/Os according to the amount of dirty data in the pool. Since both throughput and latency typically increase with the number of concurrent operations issued to physical devices, reducing the burstiness in the number of concurrent operations also stabilizes the response time of operations from other -- and in particular synchronous -- queues. In broad strokes, the I/O scheduler will issue more concurrent operations from the async write queue as there's more dirty data in the pool. .sp Async Writes .sp The number of concurrent operations issued for the async write I/O class follows a piece-wise linear function defined by a few adjustable points. .nf | o---------| <-- zfs_vdev_async_write_max_active ^ | /^ | | | / | | active | / | | I/O | / | | count | / | | | / | | |-------o | | <-- zfs_vdev_async_write_min_active 0|_______^______|_________| 0% | | 100% of zfs_dirty_data_max | | | `-- zfs_vdev_async_write_active_max_dirty_percent `--------- zfs_vdev_async_write_active_min_dirty_percent .fi Until the amount of dirty data exceeds a minimum percentage of the dirty data allowed in the pool, the I/O scheduler will limit the number of concurrent operations to the minimum. As that threshold is crossed, the number of concurrent operations issued increases linearly to the maximum at the specified maximum percentage of the dirty data allowed in the pool. .sp Ideally, the amount of dirty data on a busy pool will stay in the sloped part of the function between \fBzfs_vdev_async_write_active_min_dirty_percent\fR and \fBzfs_vdev_async_write_active_max_dirty_percent\fR. If it exceeds the maximum percentage, this indicates that the rate of incoming data is greater than the rate that the backend storage can handle. In this case, we must further throttle incoming writes, as described in the next section. .SH ZFS TRANSACTION DELAY We delay transactions when we've determined that the backend storage isn't able to accommodate the rate of incoming writes. .sp If there is already a transaction waiting, we delay relative to when that transaction will finish waiting. This way the calculated delay time is independent of the number of threads concurrently executing transactions. .sp If we are the only waiter, wait relative to when the transaction started, rather than the current time. This credits the transaction for "time already served", e.g. reading indirect blocks. .sp The minimum time for a transaction to take is calculated as: .nf min_time = zfs_delay_scale * (dirty - min) / (max - dirty) min_time is then capped at 100 milliseconds. .fi .sp The delay has two degrees of freedom that can be adjusted via tunables. The percentage of dirty data at which we start to delay is defined by \fBzfs_delay_min_dirty_percent\fR. This should typically be at or above \fBzfs_vdev_async_write_active_max_dirty_percent\fR so that we only start to delay after writing at full speed has failed to keep up with the incoming write rate. The scale of the curve is defined by \fBzfs_delay_scale\fR. Roughly speaking, this variable determines the amount of delay at the midpoint of the curve. .sp .nf delay 10ms +-------------------------------------------------------------*+ | *| 9ms + *+ | *| 8ms + *+ | * | 7ms + * + | * | 6ms + * + | * | 5ms + * + | * | 4ms + * + | * | 3ms + * + | * | 2ms + (midpoint) * + | | ** | 1ms + v *** + | zfs_delay_scale ----------> ******** | 0 +-------------------------------------*********----------------+ 0% <- zfs_dirty_data_max -> 100% .fi .sp Note that since the delay is added to the outstanding time remaining on the most recent transaction, the delay is effectively the inverse of IOPS. Here the midpoint of 500us translates to 2000 IOPS. The shape of the curve was chosen such that small changes in the amount of accumulated dirty data in the first 3/4 of the curve yield relatively small differences in the amount of delay. .sp The effects can be easier to understand when the amount of delay is represented on a log scale: .sp .nf delay 100ms +-------------------------------------------------------------++ + + | | + *+ 10ms + *+ + ** + | (midpoint) ** | + | ** + 1ms + v **** + + zfs_delay_scale ----------> ***** + | **** | + **** + 100us + ** + + * + | * | + * + 10us + * + + + | | + + +--------------------------------------------------------------+ 0% <- zfs_dirty_data_max -> 100% .fi .sp Note here that only as the amount of dirty data approaches its limit does the delay start to increase rapidly. The goal of a properly tuned system should be to keep the amount of dirty data out of that range by first ensuring that the appropriate limits are set for the I/O scheduler to reach optimal throughput on the backend storage, and then by changing the value of \fBzfs_delay_scale\fR to increase the steepness of the curve. diff --git a/man/man8/zinject.8 b/man/man8/zinject.8 index 7b664415f5ff..e97db839b808 100644 --- a/man/man8/zinject.8 +++ b/man/man8/zinject.8 @@ -1,189 +1,190 @@ '\" t .\" .\" CDDL HEADER START .\" .\" The contents of this file are subject to the terms of the .\" Common Development and Distribution License (the "License"). .\" You may not use this file except in compliance with the License. .\" .\" You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE .\" or http://www.opensolaris.org/os/licensing. .\" See the License for the specific language governing permissions .\" and limitations under the License. .\" .\" When distributing Covered Code, include this CDDL HEADER in each .\" file and include the License file at usr/src/OPENSOLARIS.LICENSE. .\" If applicable, add the following below this CDDL HEADER, with the .\" fields enclosed by brackets "[]" replaced with your own identifying .\" information: Portions Copyright [yyyy] [name of copyright owner] .\" .\" CDDL HEADER END .\" .\" .\" Copyright 2013 Darik Horn . All rights reserved. .\" .TH zinject 8 "2013 FEB 28" "ZFS on Linux" "System Administration Commands" .SH NAME zinject \- ZFS Fault Injector .SH DESCRIPTION .BR zinject creates artificial problems in a ZFS pool by simulating data corruption or device failures. This program is dangerous. .SH SYNOPSIS .TP .B "zinject" List injection records. .TP .B "zinject \-b \fIobjset:object:level:blkd\fB [\-f \fIfrequency\fB] [\-amu] \fIpool\fB" Force an error into the pool at a bookmark. .TP .B "zinject \-c <\fIid\fB | all> Cancel injection records. .TP .B "zinject \-d \fIvdev\fB \-A \fIpool\fB Force a vdev into the DEGRADED or FAULTED state. .TP .B "zinject -d \fIvdev\fB -D latency:lanes \fIpool\fB Add an artificial delay to IO requests on a particular device, such that the requests take a minimum of 'latency' milliseconds to complete. Each delay has an associated number of 'lanes' which defines the number of concurrent IO requests that can be processed. For example, with a single lane delay of 10 ms (-D 10:1), the device will only be able to service a single IO request at a time with each request taking 10 ms to complete. So, if only a single request is submitted every 10 ms, the average latency will be 10 ms; but if more than one request is submitted every 10 ms, the average latency will be more than 10 ms. Similarly, if a delay of 10 ms is specified to have two lanes (-D 10:2), then the device will be able to service two requests at a time, each with a minimum latency of 10 ms. So, if two requests are submitted every 10 ms, then the average latency will be 10 ms; but if more than two requests are submitted every 10 ms, the average latency will be more than 10 ms. Also note, these delays are additive. So two invocations of '-D 10:1', is roughly equivalent to a single invocation of '-D 10:2'. This also means, one can specify multiple lanes with differing target latencies. For example, an invocation of '-D 10:1' followed by '-D 25:2' will create 3 lanes on the device; one lane with a latency of 10 ms and two lanes with a 25 ms latency. .TP .B "zinject \-d \fIvdev\fB [\-e \fIdevice_error\fB] [\-L \fIlabel_error\fB] [\-T \fIfailure\fB] [\-f \fIfrequency\fB] [\-F] \fIpool\fB" Force a vdev error. .TP .B "zinject \-I [\-s \fIseconds\fB | \-g \fItxgs\fB] \fIpool\fB" Simulate a hardware failure that fails to honor a cache flush. .TP .B "zinject \-p \fIfunction\fB \fIpool\fB Panic inside the specified function. .TP .B "zinject \-t data [\-e \fIdevice_error\fB] [\-f \fIfrequency\fB] [\-l \fIlevel\fB] [\-r \fIrange\fB] [\-amq] \fIpath\fB" Force an error into the contents of a file. .TP .B "zinject \-t dnode [\-e \fIdevice_error\fB] [\-f \fIfrequency\fB] [\-l \fIlevel\fB] [\-amq] \fIpath\fB" Force an error into the metadnode for a file or directory. .TP .B "zinject \-t \fImos_type\fB [\-e \fIdevice_error\fB] [\-f \fIfrequency\fB] [\-l \fIlevel\fB] [\-r \fIrange\fB] [\-amqu] \fIpool\fB" Force an error into the MOS of a pool. .SH OPTIONS .TP .BI "\-a" Flush the ARC before injection. .TP .BI "\-b" " objset:object:level:start:end" Force an error into the pool at this bookmark tuple. Each number is in hexadecimal, and only one block can be specified. .TP .BI "\-d" " vdev" A vdev specified by path or GUID. .TP .BI "\-e" " device_error" Specify .BR "checksum" " for an ECKSUM error," +.BR "decompress" " for a data decompression error," .BR "decrypt" " for a data decryption error," .BR "corrupt" " to flip a bit in the data after a read," .BR "dtl" " for an ECHILD error," .BR "io" " for an EIO error where reopening the device will succeed, or" .BR "nxio" " for an ENXIO error where reopening the device will fail." For EIO and ENXIO, the "failed" reads or writes still occur. The probe simply sets the error value reported by the I/O pipeline so it appears the read or write failed. Decryption errors only currently work with file data. .TP .BI "\-f" " frequency" Only inject errors a fraction of the time. Expressed as a real number percentage between 0.0001 and 100. .TP .BI "\-F" Fail faster. Do fewer checks. .TP .BI "\-g" " txgs" Run for this many transaction groups before reporting failure. .TP .BI "\-h" Print the usage message. .TP .BI "\-l" " level" Inject an error at a particular block level. The default is 0. .TP .BI "\-L" " label_error" Set the label error region to one of .BR " nvlist" "," .BR " pad1" "," .BR " pad2" ", or" .BR " uber" "." .TP .BI "\-m" Automatically remount the underlying filesystem. .TP .BI "\-q" Quiet mode. Only print the handler number added. .TP .BI "\-r" " range" Inject an error over a particular logical range of an object, which will be translated to the appropriate blkid range according to the object's properties. .TP .BI "\-s" " seconds" Run for this many seconds before reporting failure. .TP .BI "\-T" " failure" Set the failure type to one of .BR " all" "," .BR " claim" "," .BR " free" "," .BR " read" ", or" .BR " write" "." .TP .BI "\-t" " mos_type" Set this to .BR "mos " "for any data in the MOS," .BR "mosdir " "for an object directory," .BR "config " "for the pool configuration," .BR "bpobj " "for the block pointer list," .BR "spacemap " "for the space map," .BR "metaslab " "for the metaslab, or" .BR "errlog " "for the persistent error log." .TP .BI "\-u" Unload the pool after injection. .SH "ENVIRONMENT VARIABLES" .TP .B "ZINJECT_DEBUG" Run \fBzinject\fR in debug mode. .SH "AUTHORS" This man page was written by Darik Horn excerpting the \fBzinject\fR usage message and source code. .SH "SEE ALSO" .BR zpool (8), .BR zfs (8) diff --git a/module/zfs/arc.c b/module/zfs/arc.c index f9436d7c471b..727b21f3fcb9 100644 --- a/module/zfs/arc.c +++ b/module/zfs/arc.c @@ -1,9323 +1,9354 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, Joyent, Inc. All rights reserved. - * Copyright (c) 2011, 2017 by Delphix. All rights reserved. + * Copyright (c) 2011, 2018 by Delphix. All rights reserved. * Copyright (c) 2014 by Saso Kiselkov. All rights reserved. * Copyright 2015 Nexenta Systems, Inc. All rights reserved. */ /* * DVA-based Adjustable Replacement Cache * * While much of the theory of operation used here is * based on the self-tuning, low overhead replacement cache * presented by Megiddo and Modha at FAST 2003, there are some * significant differences: * * 1. The Megiddo and Modha model assumes any page is evictable. * Pages in its cache cannot be "locked" into memory. This makes * the eviction algorithm simple: evict the last page in the list. * This also make the performance characteristics easy to reason * about. Our cache is not so simple. At any given moment, some * subset of the blocks in the cache are un-evictable because we * have handed out a reference to them. Blocks are only evictable * when there are no external references active. This makes * eviction far more problematic: we choose to evict the evictable * blocks that are the "lowest" in the list. * * There are times when it is not possible to evict the requested * space. In these circumstances we are unable to adjust the cache * size. To prevent the cache growing unbounded at these times we * implement a "cache throttle" that slows the flow of new data * into the cache until we can make space available. * * 2. The Megiddo and Modha model assumes a fixed cache size. * Pages are evicted when the cache is full and there is a cache * miss. Our model has a variable sized cache. It grows with * high use, but also tries to react to memory pressure from the * operating system: decreasing its size when system memory is * tight. * * 3. The Megiddo and Modha model assumes a fixed page size. All * elements of the cache are therefore exactly the same size. So * when adjusting the cache size following a cache miss, its simply * a matter of choosing a single page to evict. In our model, we * have variable sized cache blocks (rangeing from 512 bytes to * 128K bytes). We therefore choose a set of blocks to evict to make * space for a cache miss that approximates as closely as possible * the space used by the new block. * * See also: "ARC: A Self-Tuning, Low Overhead Replacement Cache" * by N. Megiddo & D. Modha, FAST 2003 */ /* * The locking model: * * A new reference to a cache buffer can be obtained in two * ways: 1) via a hash table lookup using the DVA as a key, * or 2) via one of the ARC lists. The arc_read() interface * uses method 1, while the internal ARC algorithms for * adjusting the cache use method 2. We therefore provide two * types of locks: 1) the hash table lock array, and 2) the * ARC list locks. * * Buffers do not have their own mutexes, rather they rely on the * hash table mutexes for the bulk of their protection (i.e. most * fields in the arc_buf_hdr_t are protected by these mutexes). * * buf_hash_find() returns the appropriate mutex (held) when it * locates the requested buffer in the hash table. It returns * NULL for the mutex if the buffer was not in the table. * * buf_hash_remove() expects the appropriate hash mutex to be * already held before it is invoked. * * Each ARC state also has a mutex which is used to protect the * buffer list associated with the state. When attempting to * obtain a hash table lock while holding an ARC list lock you * must use: mutex_tryenter() to avoid deadlock. Also note that * the active state mutex must be held before the ghost state mutex. * * It as also possible to register a callback which is run when the * arc_meta_limit is reached and no buffers can be safely evicted. In * this case the arc user should drop a reference on some arc buffers so * they can be reclaimed and the arc_meta_limit honored. For example, * when using the ZPL each dentry holds a references on a znode. These * dentries must be pruned before the arc buffer holding the znode can * be safely evicted. * * Note that the majority of the performance stats are manipulated * with atomic operations. * * The L2ARC uses the l2ad_mtx on each vdev for the following: * * - L2ARC buflist creation * - L2ARC buflist eviction * - L2ARC write completion, which walks L2ARC buflists * - ARC header destruction, as it removes from L2ARC buflists * - ARC header release, as it removes from L2ARC buflists */ /* * ARC operation: * * Every block that is in the ARC is tracked by an arc_buf_hdr_t structure. * This structure can point either to a block that is still in the cache or to * one that is only accessible in an L2 ARC device, or it can provide * information about a block that was recently evicted. If a block is * only accessible in the L2ARC, then the arc_buf_hdr_t only has enough * information to retrieve it from the L2ARC device. This information is * stored in the l2arc_buf_hdr_t sub-structure of the arc_buf_hdr_t. A block * that is in this state cannot access the data directly. * * Blocks that are actively being referenced or have not been evicted * are cached in the L1ARC. The L1ARC (l1arc_buf_hdr_t) is a structure within * the arc_buf_hdr_t that will point to the data block in memory. A block can * only be read by a consumer if it has an l1arc_buf_hdr_t. The L1ARC * caches data in two ways -- in a list of ARC buffers (arc_buf_t) and * also in the arc_buf_hdr_t's private physical data block pointer (b_pabd). * * The L1ARC's data pointer may or may not be uncompressed. The ARC has the * ability to store the physical data (b_pabd) associated with the DVA of the * arc_buf_hdr_t. Since the b_pabd is a copy of the on-disk physical block, * it will match its on-disk compression characteristics. This behavior can be * disabled by setting 'zfs_compressed_arc_enabled' to B_FALSE. When the * compressed ARC functionality is disabled, the b_pabd will point to an * uncompressed version of the on-disk data. * * Data in the L1ARC is not accessed by consumers of the ARC directly. Each * arc_buf_hdr_t can have multiple ARC buffers (arc_buf_t) which reference it. * Each ARC buffer (arc_buf_t) is being actively accessed by a specific ARC * consumer. The ARC will provide references to this data and will keep it * cached until it is no longer in use. The ARC caches only the L1ARC's physical * data block and will evict any arc_buf_t that is no longer referenced. The * amount of memory consumed by the arc_buf_ts' data buffers can be seen via the * "overhead_size" kstat. * * Depending on the consumer, an arc_buf_t can be requested in uncompressed or * compressed form. The typical case is that consumers will want uncompressed * data, and when that happens a new data buffer is allocated where the data is * decompressed for them to use. Currently the only consumer who wants * compressed arc_buf_t's is "zfs send", when it streams data exactly as it * exists on disk. When this happens, the arc_buf_t's data buffer is shared * with the arc_buf_hdr_t. * * Here is a diagram showing an arc_buf_hdr_t referenced by two arc_buf_t's. The * first one is owned by a compressed send consumer (and therefore references * the same compressed data buffer as the arc_buf_hdr_t) and the second could be * used by any other consumer (and has its own uncompressed copy of the data * buffer). * * arc_buf_hdr_t * +-----------+ * | fields | * | common to | * | L1- and | * | L2ARC | * +-----------+ * | l2arc_buf_hdr_t * | | * +-----------+ * | l1arc_buf_hdr_t * | | arc_buf_t * | b_buf +------------>+-----------+ arc_buf_t * | b_pabd +-+ |b_next +---->+-----------+ * +-----------+ | |-----------| |b_next +-->NULL * | |b_comp = T | +-----------+ * | |b_data +-+ |b_comp = F | * | +-----------+ | |b_data +-+ * +->+------+ | +-----------+ | * compressed | | | | * data | |<--------------+ | uncompressed * +------+ compressed, | data * shared +-->+------+ * data | | * | | * +------+ * * When a consumer reads a block, the ARC must first look to see if the * arc_buf_hdr_t is cached. If the hdr is cached then the ARC allocates a new * arc_buf_t and either copies uncompressed data into a new data buffer from an * existing uncompressed arc_buf_t, decompresses the hdr's b_pabd buffer into a * new data buffer, or shares the hdr's b_pabd buffer, depending on whether the * hdr is compressed and the desired compression characteristics of the * arc_buf_t consumer. If the arc_buf_t ends up sharing data with the * arc_buf_hdr_t and both of them are uncompressed then the arc_buf_t must be * the last buffer in the hdr's b_buf list, however a shared compressed buf can * be anywhere in the hdr's list. * * The diagram below shows an example of an uncompressed ARC hdr that is * sharing its data with an arc_buf_t (note that the shared uncompressed buf is * the last element in the buf list): * * arc_buf_hdr_t * +-----------+ * | | * | | * | | * +-----------+ * l2arc_buf_hdr_t| | * | | * +-----------+ * l1arc_buf_hdr_t| | * | | arc_buf_t (shared) * | b_buf +------------>+---------+ arc_buf_t * | | |b_next +---->+---------+ * | b_pabd +-+ |---------| |b_next +-->NULL * +-----------+ | | | +---------+ * | |b_data +-+ | | * | +---------+ | |b_data +-+ * +->+------+ | +---------+ | * | | | | * uncompressed | | | | * data +------+ | | * ^ +->+------+ | * | uncompressed | | | * | data | | | * | +------+ | * +---------------------------------+ * * Writing to the ARC requires that the ARC first discard the hdr's b_pabd * since the physical block is about to be rewritten. The new data contents * will be contained in the arc_buf_t. As the I/O pipeline performs the write, * it may compress the data before writing it to disk. The ARC will be called * with the transformed data and will bcopy the transformed on-disk block into * a newly allocated b_pabd. Writes are always done into buffers which have * either been loaned (and hence are new and don't have other readers) or * buffers which have been released (and hence have their own hdr, if there * were originally other readers of the buf's original hdr). This ensures that * the ARC only needs to update a single buf and its hdr after a write occurs. * * When the L2ARC is in use, it will also take advantage of the b_pabd. The * L2ARC will always write the contents of b_pabd to the L2ARC. This means * that when compressed ARC is enabled that the L2ARC blocks are identical * to the on-disk block in the main data pool. This provides a significant * advantage since the ARC can leverage the bp's checksum when reading from the * L2ARC to determine if the contents are valid. However, if the compressed * ARC is disabled, then the L2ARC's block must be transformed to look * like the physical block in the main data pool before comparing the * checksum and determining its validity. * * The L1ARC has a slightly different system for storing encrypted data. * Raw (encrypted + possibly compressed) data has a few subtle differences from * data that is just compressed. The biggest difference is that it is not * possible to decrypt encrypted data (or visa versa) if the keys aren't loaded. * The other difference is that encryption cannot be treated as a suggestion. * If a caller would prefer compressed data, but they actually wind up with * uncompressed data the worst thing that could happen is there might be a * performance hit. If the caller requests encrypted data, however, we must be * sure they actually get it or else secret information could be leaked. Raw * data is stored in hdr->b_crypt_hdr.b_rabd. An encrypted header, therefore, * may have both an encrypted version and a decrypted version of its data at * once. When a caller needs a raw arc_buf_t, it is allocated and the data is * copied out of this header. To avoid complications with b_pabd, raw buffers * cannot be shared. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef _KERNEL #include #include #include #include #endif #include #include #include #include #include #include #include #include #ifndef _KERNEL /* set with ZFS_DEBUG=watch, to enable watchpoints on frozen buffers */ boolean_t arc_watch = B_FALSE; #endif static kmutex_t arc_reclaim_lock; static kcondvar_t arc_reclaim_thread_cv; static boolean_t arc_reclaim_thread_exit; static kcondvar_t arc_reclaim_waiters_cv; /* * The number of headers to evict in arc_evict_state_impl() before * dropping the sublist lock and evicting from another sublist. A lower * value means we're more likely to evict the "correct" header (i.e. the * oldest header in the arc state), but comes with higher overhead * (i.e. more invocations of arc_evict_state_impl()). */ int zfs_arc_evict_batch_limit = 10; /* number of seconds before growing cache again */ static int arc_grow_retry = 5; /* shift of arc_c for calculating overflow limit in arc_get_data_impl */ int zfs_arc_overflow_shift = 8; /* shift of arc_c for calculating both min and max arc_p */ static int arc_p_min_shift = 4; /* log2(fraction of arc to reclaim) */ static int arc_shrink_shift = 7; /* percent of pagecache to reclaim arc to */ #ifdef _KERNEL static uint_t zfs_arc_pc_percent = 0; #endif /* * log2(fraction of ARC which must be free to allow growing). * I.e. If there is less than arc_c >> arc_no_grow_shift free memory, * when reading a new block into the ARC, we will evict an equal-sized block * from the ARC. * * This must be less than arc_shrink_shift, so that when we shrink the ARC, * we will still not allow it to grow. */ int arc_no_grow_shift = 5; /* * minimum lifespan of a prefetch block in clock ticks * (initialized in arc_init()) */ static int arc_min_prefetch_ms; static int arc_min_prescient_prefetch_ms; /* * If this percent of memory is free, don't throttle. */ int arc_lotsfree_percent = 10; static int arc_dead; /* * The arc has filled available memory and has now warmed up. */ static boolean_t arc_warm; /* * log2 fraction of the zio arena to keep free. */ int arc_zio_arena_free_shift = 2; /* * These tunables are for performance analysis. */ unsigned long zfs_arc_max = 0; unsigned long zfs_arc_min = 0; unsigned long zfs_arc_meta_limit = 0; unsigned long zfs_arc_meta_min = 0; unsigned long zfs_arc_dnode_limit = 0; unsigned long zfs_arc_dnode_reduce_percent = 10; int zfs_arc_grow_retry = 0; int zfs_arc_shrink_shift = 0; int zfs_arc_p_min_shift = 0; int zfs_arc_average_blocksize = 8 * 1024; /* 8KB */ /* * ARC dirty data constraints for arc_tempreserve_space() throttle. */ unsigned long zfs_arc_dirty_limit_percent = 50; /* total dirty data limit */ unsigned long zfs_arc_anon_limit_percent = 25; /* anon block dirty limit */ unsigned long zfs_arc_pool_dirty_percent = 20; /* each pool's anon allowance */ /* * Enable or disable compressed arc buffers. */ int zfs_compressed_arc_enabled = B_TRUE; /* * ARC will evict meta buffers that exceed arc_meta_limit. This * tunable make arc_meta_limit adjustable for different workloads. */ unsigned long zfs_arc_meta_limit_percent = 75; /* * Percentage that can be consumed by dnodes of ARC meta buffers. */ unsigned long zfs_arc_dnode_limit_percent = 10; /* * These tunables are Linux specific */ unsigned long zfs_arc_sys_free = 0; int zfs_arc_min_prefetch_ms = 0; int zfs_arc_min_prescient_prefetch_ms = 0; int zfs_arc_p_dampener_disable = 1; int zfs_arc_meta_prune = 10000; int zfs_arc_meta_strategy = ARC_STRATEGY_META_BALANCED; int zfs_arc_meta_adjust_restarts = 4096; int zfs_arc_lotsfree_percent = 10; /* The 6 states: */ static arc_state_t ARC_anon; static arc_state_t ARC_mru; static arc_state_t ARC_mru_ghost; static arc_state_t ARC_mfu; static arc_state_t ARC_mfu_ghost; static arc_state_t ARC_l2c_only; typedef struct arc_stats { kstat_named_t arcstat_hits; kstat_named_t arcstat_misses; kstat_named_t arcstat_demand_data_hits; kstat_named_t arcstat_demand_data_misses; kstat_named_t arcstat_demand_metadata_hits; kstat_named_t arcstat_demand_metadata_misses; kstat_named_t arcstat_prefetch_data_hits; kstat_named_t arcstat_prefetch_data_misses; kstat_named_t arcstat_prefetch_metadata_hits; kstat_named_t arcstat_prefetch_metadata_misses; kstat_named_t arcstat_mru_hits; kstat_named_t arcstat_mru_ghost_hits; kstat_named_t arcstat_mfu_hits; kstat_named_t arcstat_mfu_ghost_hits; kstat_named_t arcstat_deleted; /* * Number of buffers that could not be evicted because the hash lock * was held by another thread. The lock may not necessarily be held * by something using the same buffer, since hash locks are shared * by multiple buffers. */ kstat_named_t arcstat_mutex_miss; /* * Number of buffers skipped when updating the access state due to the * header having already been released after acquiring the hash lock. */ kstat_named_t arcstat_access_skip; /* * Number of buffers skipped because they have I/O in progress, are * indirect prefetch buffers that have not lived long enough, or are * not from the spa we're trying to evict from. */ kstat_named_t arcstat_evict_skip; /* * Number of times arc_evict_state() was unable to evict enough * buffers to reach its target amount. */ kstat_named_t arcstat_evict_not_enough; kstat_named_t arcstat_evict_l2_cached; kstat_named_t arcstat_evict_l2_eligible; kstat_named_t arcstat_evict_l2_ineligible; kstat_named_t arcstat_evict_l2_skip; kstat_named_t arcstat_hash_elements; kstat_named_t arcstat_hash_elements_max; kstat_named_t arcstat_hash_collisions; kstat_named_t arcstat_hash_chains; kstat_named_t arcstat_hash_chain_max; kstat_named_t arcstat_p; kstat_named_t arcstat_c; kstat_named_t arcstat_c_min; kstat_named_t arcstat_c_max; /* Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_size; /* * Number of compressed bytes stored in the arc_buf_hdr_t's b_pabd. * Note that the compressed bytes may match the uncompressed bytes * if the block is either not compressed or compressed arc is disabled. */ kstat_named_t arcstat_compressed_size; /* * Uncompressed size of the data stored in b_pabd. If compressed * arc is disabled then this value will be identical to the stat * above. */ kstat_named_t arcstat_uncompressed_size; /* * Number of bytes stored in all the arc_buf_t's. This is classified * as "overhead" since this data is typically short-lived and will * be evicted from the arc when it becomes unreferenced unless the * zfs_keep_uncompressed_metadata or zfs_keep_uncompressed_level * values have been set (see comment in dbuf.c for more information). */ kstat_named_t arcstat_overhead_size; /* * Number of bytes consumed by internal ARC structures necessary * for tracking purposes; these structures are not actually * backed by ARC buffers. This includes arc_buf_hdr_t structures * (allocated via arc_buf_hdr_t_full and arc_buf_hdr_t_l2only * caches), and arc_buf_t structures (allocated via arc_buf_t * cache). * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_hdr_size; /* * Number of bytes consumed by ARC buffers of type equal to * ARC_BUFC_DATA. This is generally consumed by buffers backing * on disk user data (e.g. plain file contents). * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_data_size; /* * Number of bytes consumed by ARC buffers of type equal to * ARC_BUFC_METADATA. This is generally consumed by buffers * backing on disk data that is used for internal ZFS * structures (e.g. ZAP, dnode, indirect blocks, etc). * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_metadata_size; /* * Number of bytes consumed by dmu_buf_impl_t objects. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_dbuf_size; /* * Number of bytes consumed by dnode_t objects. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_dnode_size; /* * Number of bytes consumed by bonus buffers. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_bonus_size; /* * Total number of bytes consumed by ARC buffers residing in the * arc_anon state. This includes *all* buffers in the arc_anon * state; e.g. data, metadata, evictable, and unevictable buffers * are all included in this value. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_anon_size; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_DATA, * residing in the arc_anon state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_anon_evictable_data; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_METADATA, * residing in the arc_anon state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_anon_evictable_metadata; /* * Total number of bytes consumed by ARC buffers residing in the * arc_mru state. This includes *all* buffers in the arc_mru * state; e.g. data, metadata, evictable, and unevictable buffers * are all included in this value. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mru_size; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_DATA, * residing in the arc_mru state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mru_evictable_data; /* * Number of bytes consumed by ARC buffers that meet the * following criteria: backing buffers of type ARC_BUFC_METADATA, * residing in the arc_mru state, and are eligible for eviction * (e.g. have no outstanding holds on the buffer). * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mru_evictable_metadata; /* * Total number of bytes that *would have been* consumed by ARC * buffers in the arc_mru_ghost state. The key thing to note * here, is the fact that this size doesn't actually indicate * RAM consumption. The ghost lists only consist of headers and * don't actually have ARC buffers linked off of these headers. * Thus, *if* the headers had associated ARC buffers, these * buffers *would have* consumed this number of bytes. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mru_ghost_size; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_DATA, and linked off the arc_mru_ghost state. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mru_ghost_evictable_data; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_METADATA, and linked off the arc_mru_ghost state. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mru_ghost_evictable_metadata; /* * Total number of bytes consumed by ARC buffers residing in the * arc_mfu state. This includes *all* buffers in the arc_mfu * state; e.g. data, metadata, evictable, and unevictable buffers * are all included in this value. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mfu_size; /* * Number of bytes consumed by ARC buffers that are eligible for * eviction, of type ARC_BUFC_DATA, and reside in the arc_mfu * state. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mfu_evictable_data; /* * Number of bytes consumed by ARC buffers that are eligible for * eviction, of type ARC_BUFC_METADATA, and reside in the * arc_mfu state. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mfu_evictable_metadata; /* * Total number of bytes that *would have been* consumed by ARC * buffers in the arc_mfu_ghost state. See the comment above * arcstat_mru_ghost_size for more details. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mfu_ghost_size; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_DATA, and linked off the arc_mfu_ghost state. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mfu_ghost_evictable_data; /* * Number of bytes that *would have been* consumed by ARC * buffers that are eligible for eviction, of type * ARC_BUFC_METADATA, and linked off the arc_mru_ghost state. * Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_mfu_ghost_evictable_metadata; kstat_named_t arcstat_l2_hits; kstat_named_t arcstat_l2_misses; kstat_named_t arcstat_l2_feeds; kstat_named_t arcstat_l2_rw_clash; kstat_named_t arcstat_l2_read_bytes; kstat_named_t arcstat_l2_write_bytes; kstat_named_t arcstat_l2_writes_sent; kstat_named_t arcstat_l2_writes_done; kstat_named_t arcstat_l2_writes_error; kstat_named_t arcstat_l2_writes_lock_retry; kstat_named_t arcstat_l2_evict_lock_retry; kstat_named_t arcstat_l2_evict_reading; kstat_named_t arcstat_l2_evict_l1cached; kstat_named_t arcstat_l2_free_on_write; kstat_named_t arcstat_l2_abort_lowmem; kstat_named_t arcstat_l2_cksum_bad; kstat_named_t arcstat_l2_io_error; kstat_named_t arcstat_l2_lsize; kstat_named_t arcstat_l2_psize; /* Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_l2_hdr_size; kstat_named_t arcstat_memory_throttle_count; kstat_named_t arcstat_memory_direct_count; kstat_named_t arcstat_memory_indirect_count; kstat_named_t arcstat_memory_all_bytes; kstat_named_t arcstat_memory_free_bytes; kstat_named_t arcstat_memory_available_bytes; kstat_named_t arcstat_no_grow; kstat_named_t arcstat_tempreserve; kstat_named_t arcstat_loaned_bytes; kstat_named_t arcstat_prune; /* Not updated directly; only synced in arc_kstat_update. */ kstat_named_t arcstat_meta_used; kstat_named_t arcstat_meta_limit; kstat_named_t arcstat_dnode_limit; kstat_named_t arcstat_meta_max; kstat_named_t arcstat_meta_min; kstat_named_t arcstat_async_upgrade_sync; kstat_named_t arcstat_demand_hit_predictive_prefetch; kstat_named_t arcstat_demand_hit_prescient_prefetch; kstat_named_t arcstat_need_free; kstat_named_t arcstat_sys_free; kstat_named_t arcstat_raw_size; } arc_stats_t; static arc_stats_t arc_stats = { { "hits", KSTAT_DATA_UINT64 }, { "misses", KSTAT_DATA_UINT64 }, { "demand_data_hits", KSTAT_DATA_UINT64 }, { "demand_data_misses", KSTAT_DATA_UINT64 }, { "demand_metadata_hits", KSTAT_DATA_UINT64 }, { "demand_metadata_misses", KSTAT_DATA_UINT64 }, { "prefetch_data_hits", KSTAT_DATA_UINT64 }, { "prefetch_data_misses", KSTAT_DATA_UINT64 }, { "prefetch_metadata_hits", KSTAT_DATA_UINT64 }, { "prefetch_metadata_misses", KSTAT_DATA_UINT64 }, { "mru_hits", KSTAT_DATA_UINT64 }, { "mru_ghost_hits", KSTAT_DATA_UINT64 }, { "mfu_hits", KSTAT_DATA_UINT64 }, { "mfu_ghost_hits", KSTAT_DATA_UINT64 }, { "deleted", KSTAT_DATA_UINT64 }, { "mutex_miss", KSTAT_DATA_UINT64 }, { "access_skip", KSTAT_DATA_UINT64 }, { "evict_skip", KSTAT_DATA_UINT64 }, { "evict_not_enough", KSTAT_DATA_UINT64 }, { "evict_l2_cached", KSTAT_DATA_UINT64 }, { "evict_l2_eligible", KSTAT_DATA_UINT64 }, { "evict_l2_ineligible", KSTAT_DATA_UINT64 }, { "evict_l2_skip", KSTAT_DATA_UINT64 }, { "hash_elements", KSTAT_DATA_UINT64 }, { "hash_elements_max", KSTAT_DATA_UINT64 }, { "hash_collisions", KSTAT_DATA_UINT64 }, { "hash_chains", KSTAT_DATA_UINT64 }, { "hash_chain_max", KSTAT_DATA_UINT64 }, { "p", KSTAT_DATA_UINT64 }, { "c", KSTAT_DATA_UINT64 }, { "c_min", KSTAT_DATA_UINT64 }, { "c_max", KSTAT_DATA_UINT64 }, { "size", KSTAT_DATA_UINT64 }, { "compressed_size", KSTAT_DATA_UINT64 }, { "uncompressed_size", KSTAT_DATA_UINT64 }, { "overhead_size", KSTAT_DATA_UINT64 }, { "hdr_size", KSTAT_DATA_UINT64 }, { "data_size", KSTAT_DATA_UINT64 }, { "metadata_size", KSTAT_DATA_UINT64 }, { "dbuf_size", KSTAT_DATA_UINT64 }, { "dnode_size", KSTAT_DATA_UINT64 }, { "bonus_size", KSTAT_DATA_UINT64 }, { "anon_size", KSTAT_DATA_UINT64 }, { "anon_evictable_data", KSTAT_DATA_UINT64 }, { "anon_evictable_metadata", KSTAT_DATA_UINT64 }, { "mru_size", KSTAT_DATA_UINT64 }, { "mru_evictable_data", KSTAT_DATA_UINT64 }, { "mru_evictable_metadata", KSTAT_DATA_UINT64 }, { "mru_ghost_size", KSTAT_DATA_UINT64 }, { "mru_ghost_evictable_data", KSTAT_DATA_UINT64 }, { "mru_ghost_evictable_metadata", KSTAT_DATA_UINT64 }, { "mfu_size", KSTAT_DATA_UINT64 }, { "mfu_evictable_data", KSTAT_DATA_UINT64 }, { "mfu_evictable_metadata", KSTAT_DATA_UINT64 }, { "mfu_ghost_size", KSTAT_DATA_UINT64 }, { "mfu_ghost_evictable_data", KSTAT_DATA_UINT64 }, { "mfu_ghost_evictable_metadata", KSTAT_DATA_UINT64 }, { "l2_hits", KSTAT_DATA_UINT64 }, { "l2_misses", KSTAT_DATA_UINT64 }, { "l2_feeds", KSTAT_DATA_UINT64 }, { "l2_rw_clash", KSTAT_DATA_UINT64 }, { "l2_read_bytes", KSTAT_DATA_UINT64 }, { "l2_write_bytes", KSTAT_DATA_UINT64 }, { "l2_writes_sent", KSTAT_DATA_UINT64 }, { "l2_writes_done", KSTAT_DATA_UINT64 }, { "l2_writes_error", KSTAT_DATA_UINT64 }, { "l2_writes_lock_retry", KSTAT_DATA_UINT64 }, { "l2_evict_lock_retry", KSTAT_DATA_UINT64 }, { "l2_evict_reading", KSTAT_DATA_UINT64 }, { "l2_evict_l1cached", KSTAT_DATA_UINT64 }, { "l2_free_on_write", KSTAT_DATA_UINT64 }, { "l2_abort_lowmem", KSTAT_DATA_UINT64 }, { "l2_cksum_bad", KSTAT_DATA_UINT64 }, { "l2_io_error", KSTAT_DATA_UINT64 }, { "l2_size", KSTAT_DATA_UINT64 }, { "l2_asize", KSTAT_DATA_UINT64 }, { "l2_hdr_size", KSTAT_DATA_UINT64 }, { "memory_throttle_count", KSTAT_DATA_UINT64 }, { "memory_direct_count", KSTAT_DATA_UINT64 }, { "memory_indirect_count", KSTAT_DATA_UINT64 }, { "memory_all_bytes", KSTAT_DATA_UINT64 }, { "memory_free_bytes", KSTAT_DATA_UINT64 }, { "memory_available_bytes", KSTAT_DATA_INT64 }, { "arc_no_grow", KSTAT_DATA_UINT64 }, { "arc_tempreserve", KSTAT_DATA_UINT64 }, { "arc_loaned_bytes", KSTAT_DATA_UINT64 }, { "arc_prune", KSTAT_DATA_UINT64 }, { "arc_meta_used", KSTAT_DATA_UINT64 }, { "arc_meta_limit", KSTAT_DATA_UINT64 }, { "arc_dnode_limit", KSTAT_DATA_UINT64 }, { "arc_meta_max", KSTAT_DATA_UINT64 }, { "arc_meta_min", KSTAT_DATA_UINT64 }, { "async_upgrade_sync", KSTAT_DATA_UINT64 }, { "demand_hit_predictive_prefetch", KSTAT_DATA_UINT64 }, { "demand_hit_prescient_prefetch", KSTAT_DATA_UINT64 }, { "arc_need_free", KSTAT_DATA_UINT64 }, { "arc_sys_free", KSTAT_DATA_UINT64 }, { "arc_raw_size", KSTAT_DATA_UINT64 } }; #define ARCSTAT(stat) (arc_stats.stat.value.ui64) #define ARCSTAT_INCR(stat, val) \ atomic_add_64(&arc_stats.stat.value.ui64, (val)) #define ARCSTAT_BUMP(stat) ARCSTAT_INCR(stat, 1) #define ARCSTAT_BUMPDOWN(stat) ARCSTAT_INCR(stat, -1) #define ARCSTAT_MAX(stat, val) { \ uint64_t m; \ while ((val) > (m = arc_stats.stat.value.ui64) && \ (m != atomic_cas_64(&arc_stats.stat.value.ui64, m, (val)))) \ continue; \ } #define ARCSTAT_MAXSTAT(stat) \ ARCSTAT_MAX(stat##_max, arc_stats.stat.value.ui64) /* * We define a macro to allow ARC hits/misses to be easily broken down by * two separate conditions, giving a total of four different subtypes for * each of hits and misses (so eight statistics total). */ #define ARCSTAT_CONDSTAT(cond1, stat1, notstat1, cond2, stat2, notstat2, stat) \ if (cond1) { \ if (cond2) { \ ARCSTAT_BUMP(arcstat_##stat1##_##stat2##_##stat); \ } else { \ ARCSTAT_BUMP(arcstat_##stat1##_##notstat2##_##stat); \ } \ } else { \ if (cond2) { \ ARCSTAT_BUMP(arcstat_##notstat1##_##stat2##_##stat); \ } else { \ ARCSTAT_BUMP(arcstat_##notstat1##_##notstat2##_##stat);\ } \ } kstat_t *arc_ksp; static arc_state_t *arc_anon; static arc_state_t *arc_mru; static arc_state_t *arc_mru_ghost; static arc_state_t *arc_mfu; static arc_state_t *arc_mfu_ghost; static arc_state_t *arc_l2c_only; /* * There are several ARC variables that are critical to export as kstats -- * but we don't want to have to grovel around in the kstat whenever we wish to * manipulate them. For these variables, we therefore define them to be in * terms of the statistic variable. This assures that we are not introducing * the possibility of inconsistency by having shadow copies of the variables, * while still allowing the code to be readable. */ #define arc_p ARCSTAT(arcstat_p) /* target size of MRU */ #define arc_c ARCSTAT(arcstat_c) /* target size of cache */ #define arc_c_min ARCSTAT(arcstat_c_min) /* min target cache size */ #define arc_c_max ARCSTAT(arcstat_c_max) /* max target cache size */ #define arc_no_grow ARCSTAT(arcstat_no_grow) /* do not grow cache size */ #define arc_tempreserve ARCSTAT(arcstat_tempreserve) #define arc_loaned_bytes ARCSTAT(arcstat_loaned_bytes) #define arc_meta_limit ARCSTAT(arcstat_meta_limit) /* max size for metadata */ #define arc_dnode_limit ARCSTAT(arcstat_dnode_limit) /* max size for dnodes */ #define arc_meta_min ARCSTAT(arcstat_meta_min) /* min size for metadata */ #define arc_meta_max ARCSTAT(arcstat_meta_max) /* max size of metadata */ #define arc_need_free ARCSTAT(arcstat_need_free) /* bytes to be freed */ #define arc_sys_free ARCSTAT(arcstat_sys_free) /* target system free bytes */ /* size of all b_rabd's in entire arc */ #define arc_raw_size ARCSTAT(arcstat_raw_size) /* compressed size of entire arc */ #define arc_compressed_size ARCSTAT(arcstat_compressed_size) /* uncompressed size of entire arc */ #define arc_uncompressed_size ARCSTAT(arcstat_uncompressed_size) /* number of bytes in the arc from arc_buf_t's */ #define arc_overhead_size ARCSTAT(arcstat_overhead_size) /* * There are also some ARC variables that we want to export, but that are * updated so often that having the canonical representation be the statistic * variable causes a performance bottleneck. We want to use aggsum_t's for these * instead, but still be able to export the kstat in the same way as before. * The solution is to always use the aggsum version, except in the kstat update * callback. */ aggsum_t arc_size; aggsum_t arc_meta_used; aggsum_t astat_data_size; aggsum_t astat_metadata_size; aggsum_t astat_dbuf_size; aggsum_t astat_dnode_size; aggsum_t astat_bonus_size; aggsum_t astat_hdr_size; aggsum_t astat_l2_hdr_size; static list_t arc_prune_list; static kmutex_t arc_prune_mtx; static taskq_t *arc_prune_taskq; #define GHOST_STATE(state) \ ((state) == arc_mru_ghost || (state) == arc_mfu_ghost || \ (state) == arc_l2c_only) #define HDR_IN_HASH_TABLE(hdr) ((hdr)->b_flags & ARC_FLAG_IN_HASH_TABLE) #define HDR_IO_IN_PROGRESS(hdr) ((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) #define HDR_IO_ERROR(hdr) ((hdr)->b_flags & ARC_FLAG_IO_ERROR) #define HDR_PREFETCH(hdr) ((hdr)->b_flags & ARC_FLAG_PREFETCH) #define HDR_PRESCIENT_PREFETCH(hdr) \ ((hdr)->b_flags & ARC_FLAG_PRESCIENT_PREFETCH) #define HDR_COMPRESSION_ENABLED(hdr) \ ((hdr)->b_flags & ARC_FLAG_COMPRESSED_ARC) #define HDR_L2CACHE(hdr) ((hdr)->b_flags & ARC_FLAG_L2CACHE) #define HDR_L2_READING(hdr) \ (((hdr)->b_flags & ARC_FLAG_IO_IN_PROGRESS) && \ ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR)) #define HDR_L2_WRITING(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITING) #define HDR_L2_EVICTED(hdr) ((hdr)->b_flags & ARC_FLAG_L2_EVICTED) #define HDR_L2_WRITE_HEAD(hdr) ((hdr)->b_flags & ARC_FLAG_L2_WRITE_HEAD) #define HDR_PROTECTED(hdr) ((hdr)->b_flags & ARC_FLAG_PROTECTED) #define HDR_NOAUTH(hdr) ((hdr)->b_flags & ARC_FLAG_NOAUTH) #define HDR_SHARED_DATA(hdr) ((hdr)->b_flags & ARC_FLAG_SHARED_DATA) #define HDR_ISTYPE_METADATA(hdr) \ ((hdr)->b_flags & ARC_FLAG_BUFC_METADATA) #define HDR_ISTYPE_DATA(hdr) (!HDR_ISTYPE_METADATA(hdr)) #define HDR_HAS_L1HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L1HDR) #define HDR_HAS_L2HDR(hdr) ((hdr)->b_flags & ARC_FLAG_HAS_L2HDR) #define HDR_HAS_RABD(hdr) \ (HDR_HAS_L1HDR(hdr) && HDR_PROTECTED(hdr) && \ (hdr)->b_crypt_hdr.b_rabd != NULL) #define HDR_ENCRYPTED(hdr) \ (HDR_PROTECTED(hdr) && DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot)) #define HDR_AUTHENTICATED(hdr) \ (HDR_PROTECTED(hdr) && !DMU_OT_IS_ENCRYPTED((hdr)->b_crypt_hdr.b_ot)) /* For storing compression mode in b_flags */ #define HDR_COMPRESS_OFFSET (highbit64(ARC_FLAG_COMPRESS_0) - 1) #define HDR_GET_COMPRESS(hdr) ((enum zio_compress)BF32_GET((hdr)->b_flags, \ HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS)) #define HDR_SET_COMPRESS(hdr, cmp) BF32_SET((hdr)->b_flags, \ HDR_COMPRESS_OFFSET, SPA_COMPRESSBITS, (cmp)); #define ARC_BUF_LAST(buf) ((buf)->b_next == NULL) #define ARC_BUF_SHARED(buf) ((buf)->b_flags & ARC_BUF_FLAG_SHARED) #define ARC_BUF_COMPRESSED(buf) ((buf)->b_flags & ARC_BUF_FLAG_COMPRESSED) #define ARC_BUF_ENCRYPTED(buf) ((buf)->b_flags & ARC_BUF_FLAG_ENCRYPTED) /* * Other sizes */ #define HDR_FULL_CRYPT_SIZE ((int64_t)sizeof (arc_buf_hdr_t)) #define HDR_FULL_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_crypt_hdr)) #define HDR_L2ONLY_SIZE ((int64_t)offsetof(arc_buf_hdr_t, b_l1hdr)) /* * Hash table routines */ #define HT_LOCK_ALIGN 64 #define HT_LOCK_PAD (P2NPHASE(sizeof (kmutex_t), (HT_LOCK_ALIGN))) struct ht_lock { kmutex_t ht_lock; #ifdef _KERNEL unsigned char pad[HT_LOCK_PAD]; #endif }; #define BUF_LOCKS 8192 typedef struct buf_hash_table { uint64_t ht_mask; arc_buf_hdr_t **ht_table; struct ht_lock ht_locks[BUF_LOCKS]; } buf_hash_table_t; static buf_hash_table_t buf_hash_table; #define BUF_HASH_INDEX(spa, dva, birth) \ (buf_hash(spa, dva, birth) & buf_hash_table.ht_mask) #define BUF_HASH_LOCK_NTRY(idx) (buf_hash_table.ht_locks[idx & (BUF_LOCKS-1)]) #define BUF_HASH_LOCK(idx) (&(BUF_HASH_LOCK_NTRY(idx).ht_lock)) #define HDR_LOCK(hdr) \ (BUF_HASH_LOCK(BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth))) uint64_t zfs_crc64_table[256]; /* * Level 2 ARC */ #define L2ARC_WRITE_SIZE (8 * 1024 * 1024) /* initial write max */ #define L2ARC_HEADROOM 2 /* num of writes */ /* * If we discover during ARC scan any buffers to be compressed, we boost * our headroom for the next scanning cycle by this percentage multiple. */ #define L2ARC_HEADROOM_BOOST 200 #define L2ARC_FEED_SECS 1 /* caching interval secs */ #define L2ARC_FEED_MIN_MS 200 /* min caching interval ms */ /* * We can feed L2ARC from two states of ARC buffers, mru and mfu, * and each of the state has two types: data and metadata. */ #define L2ARC_FEED_TYPES 4 #define l2arc_writes_sent ARCSTAT(arcstat_l2_writes_sent) #define l2arc_writes_done ARCSTAT(arcstat_l2_writes_done) /* L2ARC Performance Tunables */ unsigned long l2arc_write_max = L2ARC_WRITE_SIZE; /* def max write size */ unsigned long l2arc_write_boost = L2ARC_WRITE_SIZE; /* extra warmup write */ unsigned long l2arc_headroom = L2ARC_HEADROOM; /* # of dev writes */ unsigned long l2arc_headroom_boost = L2ARC_HEADROOM_BOOST; unsigned long l2arc_feed_secs = L2ARC_FEED_SECS; /* interval seconds */ unsigned long l2arc_feed_min_ms = L2ARC_FEED_MIN_MS; /* min interval msecs */ int l2arc_noprefetch = B_TRUE; /* don't cache prefetch bufs */ int l2arc_feed_again = B_TRUE; /* turbo warmup */ int l2arc_norw = B_FALSE; /* no reads during writes */ /* * L2ARC Internals */ static list_t L2ARC_dev_list; /* device list */ static list_t *l2arc_dev_list; /* device list pointer */ static kmutex_t l2arc_dev_mtx; /* device list mutex */ static l2arc_dev_t *l2arc_dev_last; /* last device used */ static list_t L2ARC_free_on_write; /* free after write buf list */ static list_t *l2arc_free_on_write; /* free after write list ptr */ static kmutex_t l2arc_free_on_write_mtx; /* mutex for list */ static uint64_t l2arc_ndev; /* number of devices */ typedef struct l2arc_read_callback { arc_buf_hdr_t *l2rcb_hdr; /* read header */ blkptr_t l2rcb_bp; /* original blkptr */ zbookmark_phys_t l2rcb_zb; /* original bookmark */ int l2rcb_flags; /* original flags */ abd_t *l2rcb_abd; /* temporary buffer */ } l2arc_read_callback_t; typedef struct l2arc_data_free { /* protected by l2arc_free_on_write_mtx */ abd_t *l2df_abd; size_t l2df_size; arc_buf_contents_t l2df_type; list_node_t l2df_list_node; } l2arc_data_free_t; typedef enum arc_fill_flags { ARC_FILL_LOCKED = 1 << 0, /* hdr lock is held */ ARC_FILL_COMPRESSED = 1 << 1, /* fill with compressed data */ ARC_FILL_ENCRYPTED = 1 << 2, /* fill with encrypted data */ ARC_FILL_NOAUTH = 1 << 3, /* don't attempt to authenticate */ ARC_FILL_IN_PLACE = 1 << 4 /* fill in place (special case) */ } arc_fill_flags_t; static kmutex_t l2arc_feed_thr_lock; static kcondvar_t l2arc_feed_thr_cv; static uint8_t l2arc_thread_exit; static abd_t *arc_get_data_abd(arc_buf_hdr_t *, uint64_t, void *); static void *arc_get_data_buf(arc_buf_hdr_t *, uint64_t, void *); static void arc_get_data_impl(arc_buf_hdr_t *, uint64_t, void *); static void arc_free_data_abd(arc_buf_hdr_t *, abd_t *, uint64_t, void *); static void arc_free_data_buf(arc_buf_hdr_t *, void *, uint64_t, void *); static void arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size, void *tag); static void arc_hdr_free_abd(arc_buf_hdr_t *, boolean_t); static void arc_hdr_alloc_abd(arc_buf_hdr_t *, boolean_t); static void arc_access(arc_buf_hdr_t *, kmutex_t *); static boolean_t arc_is_overflowing(void); static void arc_buf_watch(arc_buf_t *); static void arc_tuning_update(void); static void arc_prune_async(int64_t); static uint64_t arc_all_memory(void); static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *); static uint32_t arc_bufc_to_flags(arc_buf_contents_t); static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags); static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags); static boolean_t l2arc_write_eligible(uint64_t, arc_buf_hdr_t *); static void l2arc_read_done(zio_t *); /* * We use Cityhash for this. It's fast, and has good hash properties without * requiring any large static buffers. */ static uint64_t buf_hash(uint64_t spa, const dva_t *dva, uint64_t birth) { return (cityhash4(spa, dva->dva_word[0], dva->dva_word[1], birth)); } #define HDR_EMPTY(hdr) \ ((hdr)->b_dva.dva_word[0] == 0 && \ (hdr)->b_dva.dva_word[1] == 0) #define HDR_EQUAL(spa, dva, birth, hdr) \ ((hdr)->b_dva.dva_word[0] == (dva)->dva_word[0]) && \ ((hdr)->b_dva.dva_word[1] == (dva)->dva_word[1]) && \ ((hdr)->b_birth == birth) && ((hdr)->b_spa == spa) static void buf_discard_identity(arc_buf_hdr_t *hdr) { hdr->b_dva.dva_word[0] = 0; hdr->b_dva.dva_word[1] = 0; hdr->b_birth = 0; } static arc_buf_hdr_t * buf_hash_find(uint64_t spa, const blkptr_t *bp, kmutex_t **lockp) { const dva_t *dva = BP_IDENTITY(bp); uint64_t birth = BP_PHYSICAL_BIRTH(bp); uint64_t idx = BUF_HASH_INDEX(spa, dva, birth); kmutex_t *hash_lock = BUF_HASH_LOCK(idx); arc_buf_hdr_t *hdr; mutex_enter(hash_lock); for (hdr = buf_hash_table.ht_table[idx]; hdr != NULL; hdr = hdr->b_hash_next) { if (HDR_EQUAL(spa, dva, birth, hdr)) { *lockp = hash_lock; return (hdr); } } mutex_exit(hash_lock); *lockp = NULL; return (NULL); } /* * Insert an entry into the hash table. If there is already an element * equal to elem in the hash table, then the already existing element * will be returned and the new element will not be inserted. * Otherwise returns NULL. * If lockp == NULL, the caller is assumed to already hold the hash lock. */ static arc_buf_hdr_t * buf_hash_insert(arc_buf_hdr_t *hdr, kmutex_t **lockp) { uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth); kmutex_t *hash_lock = BUF_HASH_LOCK(idx); arc_buf_hdr_t *fhdr; uint32_t i; ASSERT(!DVA_IS_EMPTY(&hdr->b_dva)); ASSERT(hdr->b_birth != 0); ASSERT(!HDR_IN_HASH_TABLE(hdr)); if (lockp != NULL) { *lockp = hash_lock; mutex_enter(hash_lock); } else { ASSERT(MUTEX_HELD(hash_lock)); } for (fhdr = buf_hash_table.ht_table[idx], i = 0; fhdr != NULL; fhdr = fhdr->b_hash_next, i++) { if (HDR_EQUAL(hdr->b_spa, &hdr->b_dva, hdr->b_birth, fhdr)) return (fhdr); } hdr->b_hash_next = buf_hash_table.ht_table[idx]; buf_hash_table.ht_table[idx] = hdr; arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE); /* collect some hash table performance data */ if (i > 0) { ARCSTAT_BUMP(arcstat_hash_collisions); if (i == 1) ARCSTAT_BUMP(arcstat_hash_chains); ARCSTAT_MAX(arcstat_hash_chain_max, i); } ARCSTAT_BUMP(arcstat_hash_elements); ARCSTAT_MAXSTAT(arcstat_hash_elements); return (NULL); } static void buf_hash_remove(arc_buf_hdr_t *hdr) { arc_buf_hdr_t *fhdr, **hdrp; uint64_t idx = BUF_HASH_INDEX(hdr->b_spa, &hdr->b_dva, hdr->b_birth); ASSERT(MUTEX_HELD(BUF_HASH_LOCK(idx))); ASSERT(HDR_IN_HASH_TABLE(hdr)); hdrp = &buf_hash_table.ht_table[idx]; while ((fhdr = *hdrp) != hdr) { ASSERT3P(fhdr, !=, NULL); hdrp = &fhdr->b_hash_next; } *hdrp = hdr->b_hash_next; hdr->b_hash_next = NULL; arc_hdr_clear_flags(hdr, ARC_FLAG_IN_HASH_TABLE); /* collect some hash table performance data */ ARCSTAT_BUMPDOWN(arcstat_hash_elements); if (buf_hash_table.ht_table[idx] && buf_hash_table.ht_table[idx]->b_hash_next == NULL) ARCSTAT_BUMPDOWN(arcstat_hash_chains); } /* * Global data structures and functions for the buf kmem cache. */ static kmem_cache_t *hdr_full_cache; static kmem_cache_t *hdr_full_crypt_cache; static kmem_cache_t *hdr_l2only_cache; static kmem_cache_t *buf_cache; static void buf_fini(void) { int i; #if defined(_KERNEL) /* * Large allocations which do not require contiguous pages * should be using vmem_free() in the linux kernel\ */ vmem_free(buf_hash_table.ht_table, (buf_hash_table.ht_mask + 1) * sizeof (void *)); #else kmem_free(buf_hash_table.ht_table, (buf_hash_table.ht_mask + 1) * sizeof (void *)); #endif for (i = 0; i < BUF_LOCKS; i++) mutex_destroy(&buf_hash_table.ht_locks[i].ht_lock); kmem_cache_destroy(hdr_full_cache); kmem_cache_destroy(hdr_full_crypt_cache); kmem_cache_destroy(hdr_l2only_cache); kmem_cache_destroy(buf_cache); } /* * Constructor callback - called when the cache is empty * and a new buf is requested. */ /* ARGSUSED */ static int hdr_full_cons(void *vbuf, void *unused, int kmflag) { arc_buf_hdr_t *hdr = vbuf; bzero(hdr, HDR_FULL_SIZE); hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; cv_init(&hdr->b_l1hdr.b_cv, NULL, CV_DEFAULT, NULL); refcount_create(&hdr->b_l1hdr.b_refcnt); mutex_init(&hdr->b_l1hdr.b_freeze_lock, NULL, MUTEX_DEFAULT, NULL); list_link_init(&hdr->b_l1hdr.b_arc_node); list_link_init(&hdr->b_l2hdr.b_l2node); multilist_link_init(&hdr->b_l1hdr.b_arc_node); arc_space_consume(HDR_FULL_SIZE, ARC_SPACE_HDRS); return (0); } /* ARGSUSED */ static int hdr_full_crypt_cons(void *vbuf, void *unused, int kmflag) { arc_buf_hdr_t *hdr = vbuf; hdr_full_cons(vbuf, unused, kmflag); bzero(&hdr->b_crypt_hdr, sizeof (hdr->b_crypt_hdr)); arc_space_consume(sizeof (hdr->b_crypt_hdr), ARC_SPACE_HDRS); return (0); } /* ARGSUSED */ static int hdr_l2only_cons(void *vbuf, void *unused, int kmflag) { arc_buf_hdr_t *hdr = vbuf; bzero(hdr, HDR_L2ONLY_SIZE); arc_space_consume(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS); return (0); } /* ARGSUSED */ static int buf_cons(void *vbuf, void *unused, int kmflag) { arc_buf_t *buf = vbuf; bzero(buf, sizeof (arc_buf_t)); mutex_init(&buf->b_evict_lock, NULL, MUTEX_DEFAULT, NULL); arc_space_consume(sizeof (arc_buf_t), ARC_SPACE_HDRS); return (0); } /* * Destructor callback - called when a cached buf is * no longer required. */ /* ARGSUSED */ static void hdr_full_dest(void *vbuf, void *unused) { arc_buf_hdr_t *hdr = vbuf; ASSERT(HDR_EMPTY(hdr)); cv_destroy(&hdr->b_l1hdr.b_cv); refcount_destroy(&hdr->b_l1hdr.b_refcnt); mutex_destroy(&hdr->b_l1hdr.b_freeze_lock); ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); arc_space_return(HDR_FULL_SIZE, ARC_SPACE_HDRS); } /* ARGSUSED */ static void hdr_full_crypt_dest(void *vbuf, void *unused) { arc_buf_hdr_t *hdr = vbuf; hdr_full_dest(vbuf, unused); arc_space_return(sizeof (hdr->b_crypt_hdr), ARC_SPACE_HDRS); } /* ARGSUSED */ static void hdr_l2only_dest(void *vbuf, void *unused) { ASSERTV(arc_buf_hdr_t *hdr = vbuf); ASSERT(HDR_EMPTY(hdr)); arc_space_return(HDR_L2ONLY_SIZE, ARC_SPACE_L2HDRS); } /* ARGSUSED */ static void buf_dest(void *vbuf, void *unused) { arc_buf_t *buf = vbuf; mutex_destroy(&buf->b_evict_lock); arc_space_return(sizeof (arc_buf_t), ARC_SPACE_HDRS); } /* * Reclaim callback -- invoked when memory is low. */ /* ARGSUSED */ static void hdr_recl(void *unused) { dprintf("hdr_recl called\n"); /* * umem calls the reclaim func when we destroy the buf cache, * which is after we do arc_fini(). */ if (!arc_dead) cv_signal(&arc_reclaim_thread_cv); } static void buf_init(void) { uint64_t *ct = NULL; uint64_t hsize = 1ULL << 12; int i, j; /* * The hash table is big enough to fill all of physical memory * with an average block size of zfs_arc_average_blocksize (default 8K). * By default, the table will take up * totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers). */ while (hsize * zfs_arc_average_blocksize < arc_all_memory()) hsize <<= 1; retry: buf_hash_table.ht_mask = hsize - 1; #if defined(_KERNEL) /* * Large allocations which do not require contiguous pages * should be using vmem_alloc() in the linux kernel */ buf_hash_table.ht_table = vmem_zalloc(hsize * sizeof (void*), KM_SLEEP); #else buf_hash_table.ht_table = kmem_zalloc(hsize * sizeof (void*), KM_NOSLEEP); #endif if (buf_hash_table.ht_table == NULL) { ASSERT(hsize > (1ULL << 8)); hsize >>= 1; goto retry; } hdr_full_cache = kmem_cache_create("arc_buf_hdr_t_full", HDR_FULL_SIZE, 0, hdr_full_cons, hdr_full_dest, hdr_recl, NULL, NULL, 0); hdr_full_crypt_cache = kmem_cache_create("arc_buf_hdr_t_full_crypt", HDR_FULL_CRYPT_SIZE, 0, hdr_full_crypt_cons, hdr_full_crypt_dest, hdr_recl, NULL, NULL, 0); hdr_l2only_cache = kmem_cache_create("arc_buf_hdr_t_l2only", HDR_L2ONLY_SIZE, 0, hdr_l2only_cons, hdr_l2only_dest, hdr_recl, NULL, NULL, 0); buf_cache = kmem_cache_create("arc_buf_t", sizeof (arc_buf_t), 0, buf_cons, buf_dest, NULL, NULL, NULL, 0); for (i = 0; i < 256; i++) for (ct = zfs_crc64_table + i, *ct = i, j = 8; j > 0; j--) *ct = (*ct >> 1) ^ (-(*ct & 1) & ZFS_CRC64_POLY); for (i = 0; i < BUF_LOCKS; i++) { mutex_init(&buf_hash_table.ht_locks[i].ht_lock, NULL, MUTEX_DEFAULT, NULL); } } #define ARC_MINTIME (hz>>4) /* 62 ms */ /* * This is the size that the buf occupies in memory. If the buf is compressed, * it will correspond to the compressed size. You should use this method of * getting the buf size unless you explicitly need the logical size. */ uint64_t arc_buf_size(arc_buf_t *buf) { return (ARC_BUF_COMPRESSED(buf) ? HDR_GET_PSIZE(buf->b_hdr) : HDR_GET_LSIZE(buf->b_hdr)); } uint64_t arc_buf_lsize(arc_buf_t *buf) { return (HDR_GET_LSIZE(buf->b_hdr)); } /* * This function will return B_TRUE if the buffer is encrypted in memory. * This buffer can be decrypted by calling arc_untransform(). */ boolean_t arc_is_encrypted(arc_buf_t *buf) { return (ARC_BUF_ENCRYPTED(buf) != 0); } /* * Returns B_TRUE if the buffer represents data that has not had its MAC * verified yet. */ boolean_t arc_is_unauthenticated(arc_buf_t *buf) { return (HDR_NOAUTH(buf->b_hdr) != 0); } void arc_get_raw_params(arc_buf_t *buf, boolean_t *byteorder, uint8_t *salt, uint8_t *iv, uint8_t *mac) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(HDR_PROTECTED(hdr)); bcopy(hdr->b_crypt_hdr.b_salt, salt, ZIO_DATA_SALT_LEN); bcopy(hdr->b_crypt_hdr.b_iv, iv, ZIO_DATA_IV_LEN); bcopy(hdr->b_crypt_hdr.b_mac, mac, ZIO_DATA_MAC_LEN); *byteorder = (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ? ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER; } /* * Indicates how this buffer is compressed in memory. If it is not compressed * the value will be ZIO_COMPRESS_OFF. It can be made normally readable with * arc_untransform() as long as it is also unencrypted. */ enum zio_compress arc_get_compression(arc_buf_t *buf) { return (ARC_BUF_COMPRESSED(buf) ? HDR_GET_COMPRESS(buf->b_hdr) : ZIO_COMPRESS_OFF); } /* * Return the compression algorithm used to store this data in the ARC. If ARC * compression is enabled or this is an encrypted block, this will be the same * as what's used to store it on-disk. Otherwise, this will be ZIO_COMPRESS_OFF. */ static inline enum zio_compress arc_hdr_get_compress(arc_buf_hdr_t *hdr) { return (HDR_COMPRESSION_ENABLED(hdr) ? HDR_GET_COMPRESS(hdr) : ZIO_COMPRESS_OFF); } static inline boolean_t arc_buf_is_shared(arc_buf_t *buf) { boolean_t shared = (buf->b_data != NULL && buf->b_hdr->b_l1hdr.b_pabd != NULL && abd_is_linear(buf->b_hdr->b_l1hdr.b_pabd) && buf->b_data == abd_to_buf(buf->b_hdr->b_l1hdr.b_pabd)); IMPLY(shared, HDR_SHARED_DATA(buf->b_hdr)); IMPLY(shared, ARC_BUF_SHARED(buf)); IMPLY(shared, ARC_BUF_COMPRESSED(buf) || ARC_BUF_LAST(buf)); /* * It would be nice to assert arc_can_share() too, but the "hdr isn't * already being shared" requirement prevents us from doing that. */ return (shared); } /* * Free the checksum associated with this header. If there is no checksum, this * is a no-op. */ static inline void arc_cksum_free(arc_buf_hdr_t *hdr) { ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&hdr->b_l1hdr.b_freeze_lock); if (hdr->b_l1hdr.b_freeze_cksum != NULL) { kmem_free(hdr->b_l1hdr.b_freeze_cksum, sizeof (zio_cksum_t)); hdr->b_l1hdr.b_freeze_cksum = NULL; } mutex_exit(&hdr->b_l1hdr.b_freeze_lock); } /* * Return true iff at least one of the bufs on hdr is not compressed. * Encrypted buffers count as compressed. */ static boolean_t arc_hdr_has_uncompressed_buf(arc_buf_hdr_t *hdr) { ASSERT(hdr->b_l1hdr.b_state == arc_anon || MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); for (arc_buf_t *b = hdr->b_l1hdr.b_buf; b != NULL; b = b->b_next) { if (!ARC_BUF_COMPRESSED(b)) { return (B_TRUE); } } return (B_FALSE); } /* * If we've turned on the ZFS_DEBUG_MODIFY flag, verify that the buf's data * matches the checksum that is stored in the hdr. If there is no checksum, * or if the buf is compressed, this is a no-op. */ static void arc_cksum_verify(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; zio_cksum_t zc; if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; if (ARC_BUF_COMPRESSED(buf)) return; ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&hdr->b_l1hdr.b_freeze_lock); if (hdr->b_l1hdr.b_freeze_cksum == NULL || HDR_IO_ERROR(hdr)) { mutex_exit(&hdr->b_l1hdr.b_freeze_lock); return; } fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL, &zc); if (!ZIO_CHECKSUM_EQUAL(*hdr->b_l1hdr.b_freeze_cksum, zc)) panic("buffer modified while frozen!"); mutex_exit(&hdr->b_l1hdr.b_freeze_lock); } /* * This function makes the assumption that data stored in the L2ARC * will be transformed exactly as it is in the main pool. Because of * this we can verify the checksum against the reading process's bp. */ static boolean_t arc_cksum_is_equal(arc_buf_hdr_t *hdr, zio_t *zio) { ASSERT(!BP_IS_EMBEDDED(zio->io_bp)); VERIFY3U(BP_GET_PSIZE(zio->io_bp), ==, HDR_GET_PSIZE(hdr)); /* * Block pointers always store the checksum for the logical data. * If the block pointer has the gang bit set, then the checksum * it represents is for the reconstituted data and not for an * individual gang member. The zio pipeline, however, must be able to * determine the checksum of each of the gang constituents so it * treats the checksum comparison differently than what we need * for l2arc blocks. This prevents us from using the * zio_checksum_error() interface directly. Instead we must call the * zio_checksum_error_impl() so that we can ensure the checksum is * generated using the correct checksum algorithm and accounts for the * logical I/O size and not just a gang fragment. */ return (zio_checksum_error_impl(zio->io_spa, zio->io_bp, BP_GET_CHECKSUM(zio->io_bp), zio->io_abd, zio->io_size, zio->io_offset, NULL) == 0); } /* * Given a buf full of data, if ZFS_DEBUG_MODIFY is enabled this computes a * checksum and attaches it to the buf's hdr so that we can ensure that the buf * isn't modified later on. If buf is compressed or there is already a checksum * on the hdr, this is a no-op (we only checksum uncompressed bufs). */ static void arc_cksum_compute(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; ASSERT(HDR_HAS_L1HDR(hdr)); mutex_enter(&buf->b_hdr->b_l1hdr.b_freeze_lock); if (hdr->b_l1hdr.b_freeze_cksum != NULL || ARC_BUF_COMPRESSED(buf)) { mutex_exit(&hdr->b_l1hdr.b_freeze_lock); return; } ASSERT(!ARC_BUF_ENCRYPTED(buf)); ASSERT(!ARC_BUF_COMPRESSED(buf)); hdr->b_l1hdr.b_freeze_cksum = kmem_alloc(sizeof (zio_cksum_t), KM_SLEEP); fletcher_2_native(buf->b_data, arc_buf_size(buf), NULL, hdr->b_l1hdr.b_freeze_cksum); mutex_exit(&hdr->b_l1hdr.b_freeze_lock); arc_buf_watch(buf); } #ifndef _KERNEL void arc_buf_sigsegv(int sig, siginfo_t *si, void *unused) { panic("Got SIGSEGV at address: 0x%lx\n", (long)si->si_addr); } #endif /* ARGSUSED */ static void arc_buf_unwatch(arc_buf_t *buf) { #ifndef _KERNEL if (arc_watch) { ASSERT0(mprotect(buf->b_data, arc_buf_size(buf), PROT_READ | PROT_WRITE)); } #endif } /* ARGSUSED */ static void arc_buf_watch(arc_buf_t *buf) { #ifndef _KERNEL if (arc_watch) ASSERT0(mprotect(buf->b_data, arc_buf_size(buf), PROT_READ)); #endif } static arc_buf_contents_t arc_buf_type(arc_buf_hdr_t *hdr) { arc_buf_contents_t type; if (HDR_ISTYPE_METADATA(hdr)) { type = ARC_BUFC_METADATA; } else { type = ARC_BUFC_DATA; } VERIFY3U(hdr->b_type, ==, type); return (type); } boolean_t arc_is_metadata(arc_buf_t *buf) { return (HDR_ISTYPE_METADATA(buf->b_hdr) != 0); } static uint32_t arc_bufc_to_flags(arc_buf_contents_t type) { switch (type) { case ARC_BUFC_DATA: /* metadata field is 0 if buffer contains normal data */ return (0); case ARC_BUFC_METADATA: return (ARC_FLAG_BUFC_METADATA); default: break; } panic("undefined ARC buffer type!"); return ((uint32_t)-1); } void arc_buf_thaw(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); arc_cksum_verify(buf); /* * Compressed buffers do not manipulate the b_freeze_cksum. */ if (ARC_BUF_COMPRESSED(buf)) return; ASSERT(HDR_HAS_L1HDR(hdr)); arc_cksum_free(hdr); arc_buf_unwatch(buf); } void arc_buf_freeze(arc_buf_t *buf) { if (!(zfs_flags & ZFS_DEBUG_MODIFY)) return; if (ARC_BUF_COMPRESSED(buf)) return; ASSERT(HDR_HAS_L1HDR(buf->b_hdr)); arc_cksum_compute(buf); } /* * The arc_buf_hdr_t's b_flags should never be modified directly. Instead, * the following functions should be used to ensure that the flags are * updated in a thread-safe way. When manipulating the flags either * the hash_lock must be held or the hdr must be undiscoverable. This * ensures that we're not racing with any other threads when updating * the flags. */ static inline void arc_hdr_set_flags(arc_buf_hdr_t *hdr, arc_flags_t flags) { ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); hdr->b_flags |= flags; } static inline void arc_hdr_clear_flags(arc_buf_hdr_t *hdr, arc_flags_t flags) { ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); hdr->b_flags &= ~flags; } /* * Setting the compression bits in the arc_buf_hdr_t's b_flags is * done in a special way since we have to clear and set bits * at the same time. Consumers that wish to set the compression bits * must use this function to ensure that the flags are updated in * thread-safe manner. */ static void arc_hdr_set_compress(arc_buf_hdr_t *hdr, enum zio_compress cmp) { ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); /* * Holes and embedded blocks will always have a psize = 0 so * we ignore the compression of the blkptr and set the * want to uncompress them. Mark them as uncompressed. */ if (!zfs_compressed_arc_enabled || HDR_GET_PSIZE(hdr) == 0) { arc_hdr_clear_flags(hdr, ARC_FLAG_COMPRESSED_ARC); ASSERT(!HDR_COMPRESSION_ENABLED(hdr)); } else { arc_hdr_set_flags(hdr, ARC_FLAG_COMPRESSED_ARC); ASSERT(HDR_COMPRESSION_ENABLED(hdr)); } HDR_SET_COMPRESS(hdr, cmp); ASSERT3U(HDR_GET_COMPRESS(hdr), ==, cmp); } /* * Looks for another buf on the same hdr which has the data decompressed, copies * from it, and returns true. If no such buf exists, returns false. */ static boolean_t arc_buf_try_copy_decompressed_data(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; boolean_t copied = B_FALSE; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3P(buf->b_data, !=, NULL); ASSERT(!ARC_BUF_COMPRESSED(buf)); for (arc_buf_t *from = hdr->b_l1hdr.b_buf; from != NULL; from = from->b_next) { /* can't use our own data buffer */ if (from == buf) { continue; } if (!ARC_BUF_COMPRESSED(from)) { bcopy(from->b_data, buf->b_data, arc_buf_size(buf)); copied = B_TRUE; break; } } /* * There were no decompressed bufs, so there should not be a * checksum on the hdr either. */ EQUIV(!copied, hdr->b_l1hdr.b_freeze_cksum == NULL); return (copied); } /* * Return the size of the block, b_pabd, that is stored in the arc_buf_hdr_t. */ static uint64_t arc_hdr_size(arc_buf_hdr_t *hdr) { uint64_t size; if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF && HDR_GET_PSIZE(hdr) > 0) { size = HDR_GET_PSIZE(hdr); } else { ASSERT3U(HDR_GET_LSIZE(hdr), !=, 0); size = HDR_GET_LSIZE(hdr); } return (size); } static int arc_hdr_authenticate(arc_buf_hdr_t *hdr, spa_t *spa, uint64_t dsobj) { int ret; uint64_t csize; uint64_t lsize = HDR_GET_LSIZE(hdr); uint64_t psize = HDR_GET_PSIZE(hdr); void *tmpbuf = NULL; abd_t *abd = hdr->b_l1hdr.b_pabd; ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); ASSERT(HDR_AUTHENTICATED(hdr)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); /* * The MAC is calculated on the compressed data that is stored on disk. * However, if compressed arc is disabled we will only have the * decompressed data available to us now. Compress it into a temporary * abd so we can verify the MAC. The performance overhead of this will * be relatively low, since most objects in an encrypted objset will * be encrypted (instead of authenticated) anyway. */ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { tmpbuf = zio_buf_alloc(lsize); abd = abd_get_from_buf(tmpbuf, lsize); abd_take_ownership_of_buf(abd, B_TRUE); csize = zio_compress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, tmpbuf, lsize); ASSERT3U(csize, <=, psize); abd_zero_off(abd, csize, psize - csize); } /* * Authentication is best effort. We authenticate whenever the key is * available. If we succeed we clear ARC_FLAG_NOAUTH. */ if (hdr->b_crypt_hdr.b_ot == DMU_OT_OBJSET) { ASSERT3U(HDR_GET_COMPRESS(hdr), ==, ZIO_COMPRESS_OFF); ASSERT3U(lsize, ==, psize); ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa, dsobj, abd, psize, hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS); } else { ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj, abd, psize, hdr->b_crypt_hdr.b_mac); } if (ret == 0) arc_hdr_clear_flags(hdr, ARC_FLAG_NOAUTH); else if (ret != ENOENT) goto error; if (tmpbuf != NULL) abd_free(abd); return (0); error: if (tmpbuf != NULL) abd_free(abd); return (ret); } /* * This function will take a header that only has raw encrypted data in * b_crypt_hdr.b_rabd and decrypt it into a new buffer which is stored in * b_l1hdr.b_pabd. If designated in the header flags, this function will * also decompress the data. */ static int arc_hdr_decrypt(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb) { int ret; abd_t *cabd = NULL; void *tmp = NULL; boolean_t no_crypt = B_FALSE; boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS); ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); ASSERT(HDR_ENCRYPTED(hdr)); arc_hdr_alloc_abd(hdr, B_FALSE); ret = spa_do_crypt_abd(B_FALSE, spa, zb, hdr->b_crypt_hdr.b_ot, B_FALSE, bswap, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv, hdr->b_crypt_hdr.b_mac, HDR_GET_PSIZE(hdr), hdr->b_l1hdr.b_pabd, hdr->b_crypt_hdr.b_rabd, &no_crypt); if (ret != 0) goto error; if (no_crypt) { abd_copy(hdr->b_l1hdr.b_pabd, hdr->b_crypt_hdr.b_rabd, HDR_GET_PSIZE(hdr)); } /* * If this header has disabled arc compression but the b_pabd is * compressed after decrypting it, we need to decompress the newly * decrypted data. */ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { /* * We want to make sure that we are correctly honoring the * zfs_abd_scatter_enabled setting, so we allocate an abd here * and then loan a buffer from it, rather than allocating a * linear buffer and wrapping it in an abd later. */ cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr); tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr)); ret = zio_decompress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr)); if (ret != 0) { abd_return_buf(cabd, tmp, arc_hdr_size(hdr)); goto error; } abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr)); arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, arc_hdr_size(hdr), hdr); hdr->b_l1hdr.b_pabd = cabd; } return (0); error: arc_hdr_free_abd(hdr, B_FALSE); if (cabd != NULL) arc_free_data_buf(hdr, cabd, arc_hdr_size(hdr), hdr); return (ret); } /* * This function is called during arc_buf_fill() to prepare the header's * abd plaintext pointer for use. This involves authenticated protected * data and decrypting encrypted data into the plaintext abd. */ static int arc_fill_hdr_crypt(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, spa_t *spa, const zbookmark_phys_t *zb, boolean_t noauth) { int ret; ASSERT(HDR_PROTECTED(hdr)); if (hash_lock != NULL) mutex_enter(hash_lock); if (HDR_NOAUTH(hdr) && !noauth) { /* * The caller requested authenticated data but our data has * not been authenticated yet. Verify the MAC now if we can. */ ret = arc_hdr_authenticate(hdr, spa, zb->zb_objset); if (ret != 0) goto error; } else if (HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd == NULL) { /* * If we only have the encrypted version of the data, but the * unencrypted version was requested we take this opportunity * to store the decrypted version in the header for future use. */ ret = arc_hdr_decrypt(hdr, spa, zb); if (ret != 0) goto error; } ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); if (hash_lock != NULL) mutex_exit(hash_lock); return (0); error: if (hash_lock != NULL) mutex_exit(hash_lock); return (ret); } /* * This function is used by the dbuf code to decrypt bonus buffers in place. * The dbuf code itself doesn't have any locking for decrypting a shared dnode * block, so we use the hash lock here to protect against concurrent calls to * arc_buf_fill(). */ static void arc_buf_untransform_in_place(arc_buf_t *buf, kmutex_t *hash_lock) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(HDR_ENCRYPTED(hdr)); ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE); ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); zio_crypt_copy_dnode_bonus(hdr->b_l1hdr.b_pabd, buf->b_data, arc_buf_size(buf)); buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED; buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; hdr->b_crypt_hdr.b_ebufcnt -= 1; } /* * Given a buf that has a data buffer attached to it, this function will * efficiently fill the buf with data of the specified compression setting from * the hdr and update the hdr's b_freeze_cksum if necessary. If the buf and hdr * are already sharing a data buf, no copy is performed. * * If the buf is marked as compressed but uncompressed data was requested, this * will allocate a new data buffer for the buf, remove that flag, and fill the * buf with uncompressed data. You can't request a compressed buf on a hdr with * uncompressed data, and (since we haven't added support for it yet) if you * want compressed data your buf must already be marked as compressed and have * the correct-sized data buffer. */ static int arc_buf_fill(arc_buf_t *buf, spa_t *spa, const zbookmark_phys_t *zb, arc_fill_flags_t flags) { int error = 0; arc_buf_hdr_t *hdr = buf->b_hdr; boolean_t hdr_compressed = (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF); boolean_t compressed = (flags & ARC_FILL_COMPRESSED) != 0; boolean_t encrypted = (flags & ARC_FILL_ENCRYPTED) != 0; dmu_object_byteswap_t bswap = hdr->b_l1hdr.b_byteswap; kmutex_t *hash_lock = (flags & ARC_FILL_LOCKED) ? NULL : HDR_LOCK(hdr); ASSERT3P(buf->b_data, !=, NULL); IMPLY(compressed, hdr_compressed || ARC_BUF_ENCRYPTED(buf)); IMPLY(compressed, ARC_BUF_COMPRESSED(buf)); IMPLY(encrypted, HDR_ENCRYPTED(hdr)); IMPLY(encrypted, ARC_BUF_ENCRYPTED(buf)); IMPLY(encrypted, ARC_BUF_COMPRESSED(buf)); IMPLY(encrypted, !ARC_BUF_SHARED(buf)); /* * If the caller wanted encrypted data we just need to copy it from * b_rabd and potentially byteswap it. We won't be able to do any * further transforms on it. */ if (encrypted) { ASSERT(HDR_HAS_RABD(hdr)); abd_copy_to_buf(buf->b_data, hdr->b_crypt_hdr.b_rabd, HDR_GET_PSIZE(hdr)); goto byteswap; } /* * Adjust encrypted and authenticated headers to accomodate * the request if needed. Dnode blocks (ARC_FILL_IN_PLACE) are * allowed to fail decryption due to keys not being loaded * without being marked as an IO error. */ if (HDR_PROTECTED(hdr)) { error = arc_fill_hdr_crypt(hdr, hash_lock, spa, zb, !!(flags & ARC_FILL_NOAUTH)); if (error == EACCES && (flags & ARC_FILL_IN_PLACE) != 0) { return (error); } else if (error != 0) { if (hash_lock != NULL) mutex_enter(hash_lock); arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR); if (hash_lock != NULL) mutex_exit(hash_lock); return (error); } } /* * There is a special case here for dnode blocks which are * decrypting their bonus buffers. These blocks may request to * be decrypted in-place. This is necessary because there may * be many dnodes pointing into this buffer and there is * currently no method to synchronize replacing the backing * b_data buffer and updating all of the pointers. Here we use * the hash lock to ensure there are no races. If the need * arises for other types to be decrypted in-place, they must * add handling here as well. */ if ((flags & ARC_FILL_IN_PLACE) != 0) { ASSERT(!hdr_compressed); ASSERT(!compressed); ASSERT(!encrypted); if (HDR_ENCRYPTED(hdr) && ARC_BUF_ENCRYPTED(buf)) { ASSERT3U(hdr->b_crypt_hdr.b_ot, ==, DMU_OT_DNODE); if (hash_lock != NULL) mutex_enter(hash_lock); arc_buf_untransform_in_place(buf, hash_lock); if (hash_lock != NULL) mutex_exit(hash_lock); /* Compute the hdr's checksum if necessary */ arc_cksum_compute(buf); } return (0); } if (hdr_compressed == compressed) { if (!arc_buf_is_shared(buf)) { abd_copy_to_buf(buf->b_data, hdr->b_l1hdr.b_pabd, arc_buf_size(buf)); } } else { ASSERT(hdr_compressed); ASSERT(!compressed); ASSERT3U(HDR_GET_LSIZE(hdr), !=, HDR_GET_PSIZE(hdr)); /* * If the buf is sharing its data with the hdr, unlink it and * allocate a new data buffer for the buf. */ if (arc_buf_is_shared(buf)) { ASSERT(ARC_BUF_COMPRESSED(buf)); /* We need to give the buf it's own b_data */ buf->b_flags &= ~ARC_BUF_FLAG_SHARED; buf->b_data = arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf); arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA); /* Previously overhead was 0; just add new overhead */ ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr)); } else if (ARC_BUF_COMPRESSED(buf)) { /* We need to reallocate the buf's b_data */ arc_free_data_buf(hdr, buf->b_data, HDR_GET_PSIZE(hdr), buf); buf->b_data = arc_get_data_buf(hdr, HDR_GET_LSIZE(hdr), buf); /* We increased the size of b_data; update overhead */ ARCSTAT_INCR(arcstat_overhead_size, HDR_GET_LSIZE(hdr) - HDR_GET_PSIZE(hdr)); } /* * Regardless of the buf's previous compression settings, it * should not be compressed at the end of this function. */ buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; /* * Try copying the data from another buf which already has a * decompressed version. If that's not possible, it's time to * bite the bullet and decompress the data from the hdr. */ if (arc_buf_try_copy_decompressed_data(buf)) { /* Skip byteswapping and checksumming (already done) */ ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, !=, NULL); return (0); } else { error = zio_decompress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, buf->b_data, HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr)); /* * Absent hardware errors or software bugs, this should * be impossible, but log it anyway so we can debug it. */ if (error != 0) { zfs_dbgmsg( "hdr %p, compress %d, psize %d, lsize %d", hdr, arc_hdr_get_compress(hdr), HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr)); if (hash_lock != NULL) mutex_enter(hash_lock); arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR); if (hash_lock != NULL) mutex_exit(hash_lock); return (SET_ERROR(EIO)); } } } byteswap: /* Byteswap the buf's data if necessary */ if (bswap != DMU_BSWAP_NUMFUNCS) { ASSERT(!HDR_SHARED_DATA(hdr)); ASSERT3U(bswap, <, DMU_BSWAP_NUMFUNCS); dmu_ot_byteswap[bswap].ob_func(buf->b_data, HDR_GET_LSIZE(hdr)); } /* Compute the hdr's checksum if necessary */ arc_cksum_compute(buf); return (0); } /* * If this function is being called to decrypt an encrypted buffer or verify an * authenticated one, the key must be loaded and a mapping must be made * available in the keystore via spa_keystore_create_mapping() or one of its * callers. */ int arc_untransform(arc_buf_t *buf, spa_t *spa, const zbookmark_phys_t *zb, boolean_t in_place) { int ret; arc_fill_flags_t flags = 0; if (in_place) flags |= ARC_FILL_IN_PLACE; ret = arc_buf_fill(buf, spa, zb, flags); if (ret == ECKSUM) { /* * Convert authentication and decryption errors to EIO * (and generate an ereport) before leaving the ARC. */ ret = SET_ERROR(EIO); spa_log_error(spa, zb); zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION, spa, NULL, zb, NULL, 0, 0); } return (ret); } /* * Increment the amount of evictable space in the arc_state_t's refcount. * We account for the space used by the hdr and the arc buf individually * so that we can add and remove them from the refcount individually. */ static void arc_evictable_space_increment(arc_buf_hdr_t *hdr, arc_state_t *state) { arc_buf_contents_t type = arc_buf_type(hdr); ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(state)) { ASSERT0(hdr->b_l1hdr.b_bufcnt); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); (void) refcount_add_many(&state->arcs_esize[type], HDR_GET_LSIZE(hdr), hdr); return; } ASSERT(!GHOST_STATE(state)); if (hdr->b_l1hdr.b_pabd != NULL) { (void) refcount_add_many(&state->arcs_esize[type], arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) refcount_add_many(&state->arcs_esize[type], HDR_GET_PSIZE(hdr), hdr); } for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { if (arc_buf_is_shared(buf)) continue; (void) refcount_add_many(&state->arcs_esize[type], arc_buf_size(buf), buf); } } /* * Decrement the amount of evictable space in the arc_state_t's refcount. * We account for the space used by the hdr and the arc buf individually * so that we can add and remove them from the refcount individually. */ static void arc_evictable_space_decrement(arc_buf_hdr_t *hdr, arc_state_t *state) { arc_buf_contents_t type = arc_buf_type(hdr); ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(state)) { ASSERT0(hdr->b_l1hdr.b_bufcnt); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); (void) refcount_remove_many(&state->arcs_esize[type], HDR_GET_LSIZE(hdr), hdr); return; } ASSERT(!GHOST_STATE(state)); if (hdr->b_l1hdr.b_pabd != NULL) { (void) refcount_remove_many(&state->arcs_esize[type], arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) refcount_remove_many(&state->arcs_esize[type], HDR_GET_PSIZE(hdr), hdr); } for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { if (arc_buf_is_shared(buf)) continue; (void) refcount_remove_many(&state->arcs_esize[type], arc_buf_size(buf), buf); } } /* * Add a reference to this hdr indicating that someone is actively * referencing that memory. When the refcount transitions from 0 to 1, * we remove it from the respective arc_state_t list to indicate that * it is not evictable. */ static void add_reference(arc_buf_hdr_t *hdr, void *tag) { arc_state_t *state; ASSERT(HDR_HAS_L1HDR(hdr)); if (!MUTEX_HELD(HDR_LOCK(hdr))) { ASSERT(hdr->b_l1hdr.b_state == arc_anon); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); } state = hdr->b_l1hdr.b_state; if ((refcount_add(&hdr->b_l1hdr.b_refcnt, tag) == 1) && (state != arc_anon)) { /* We don't use the L2-only state list. */ if (state != arc_l2c_only) { multilist_remove(state->arcs_list[arc_buf_type(hdr)], hdr); arc_evictable_space_decrement(hdr, state); } /* remove the prefetch flag if we get a reference */ arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH); } } /* * Remove a reference from this hdr. When the reference transitions from * 1 to 0 and we're not anonymous, then we add this hdr to the arc_state_t's * list making it eligible for eviction. */ static int remove_reference(arc_buf_hdr_t *hdr, kmutex_t *hash_lock, void *tag) { int cnt; arc_state_t *state = hdr->b_l1hdr.b_state; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(state == arc_anon || MUTEX_HELD(hash_lock)); ASSERT(!GHOST_STATE(state)); /* * arc_l2c_only counts as a ghost state so we don't need to explicitly * check to prevent usage of the arc_l2c_only list. */ if (((cnt = refcount_remove(&hdr->b_l1hdr.b_refcnt, tag)) == 0) && (state != arc_anon)) { multilist_insert(state->arcs_list[arc_buf_type(hdr)], hdr); ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0); arc_evictable_space_increment(hdr, state); } return (cnt); } /* * Returns detailed information about a specific arc buffer. When the * state_index argument is set the function will calculate the arc header * list position for its arc state. Since this requires a linear traversal * callers are strongly encourage not to do this. However, it can be helpful * for targeted analysis so the functionality is provided. */ void arc_buf_info(arc_buf_t *ab, arc_buf_info_t *abi, int state_index) { arc_buf_hdr_t *hdr = ab->b_hdr; l1arc_buf_hdr_t *l1hdr = NULL; l2arc_buf_hdr_t *l2hdr = NULL; arc_state_t *state = NULL; memset(abi, 0, sizeof (arc_buf_info_t)); if (hdr == NULL) return; abi->abi_flags = hdr->b_flags; if (HDR_HAS_L1HDR(hdr)) { l1hdr = &hdr->b_l1hdr; state = l1hdr->b_state; } if (HDR_HAS_L2HDR(hdr)) l2hdr = &hdr->b_l2hdr; if (l1hdr) { abi->abi_bufcnt = l1hdr->b_bufcnt; abi->abi_access = l1hdr->b_arc_access; abi->abi_mru_hits = l1hdr->b_mru_hits; abi->abi_mru_ghost_hits = l1hdr->b_mru_ghost_hits; abi->abi_mfu_hits = l1hdr->b_mfu_hits; abi->abi_mfu_ghost_hits = l1hdr->b_mfu_ghost_hits; abi->abi_holds = refcount_count(&l1hdr->b_refcnt); } if (l2hdr) { abi->abi_l2arc_dattr = l2hdr->b_daddr; abi->abi_l2arc_hits = l2hdr->b_hits; } abi->abi_state_type = state ? state->arcs_state : ARC_STATE_ANON; abi->abi_state_contents = arc_buf_type(hdr); abi->abi_size = arc_hdr_size(hdr); } /* * Move the supplied buffer to the indicated state. The hash lock * for the buffer must be held by the caller. */ static void arc_change_state(arc_state_t *new_state, arc_buf_hdr_t *hdr, kmutex_t *hash_lock) { arc_state_t *old_state; int64_t refcnt; uint32_t bufcnt; boolean_t update_old, update_new; arc_buf_contents_t buftype = arc_buf_type(hdr); /* * We almost always have an L1 hdr here, since we call arc_hdr_realloc() * in arc_read() when bringing a buffer out of the L2ARC. However, the * L1 hdr doesn't always exist when we change state to arc_anon before * destroying a header, in which case reallocating to add the L1 hdr is * pointless. */ if (HDR_HAS_L1HDR(hdr)) { old_state = hdr->b_l1hdr.b_state; refcnt = refcount_count(&hdr->b_l1hdr.b_refcnt); bufcnt = hdr->b_l1hdr.b_bufcnt; update_old = (bufcnt > 0 || hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); } else { old_state = arc_l2c_only; refcnt = 0; bufcnt = 0; update_old = B_FALSE; } update_new = update_old; ASSERT(MUTEX_HELD(hash_lock)); ASSERT3P(new_state, !=, old_state); ASSERT(!GHOST_STATE(new_state) || bufcnt == 0); ASSERT(old_state != arc_anon || bufcnt <= 1); /* * If this buffer is evictable, transfer it from the * old state list to the new state list. */ if (refcnt == 0) { if (old_state != arc_anon && old_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); multilist_remove(old_state->arcs_list[buftype], hdr); if (GHOST_STATE(old_state)) { ASSERT0(bufcnt); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); update_old = B_TRUE; } arc_evictable_space_decrement(hdr, old_state); } if (new_state != arc_anon && new_state != arc_l2c_only) { /* * An L1 header always exists here, since if we're * moving to some L1-cached state (i.e. not l2c_only or * anonymous), we realloc the header to add an L1hdr * beforehand. */ ASSERT(HDR_HAS_L1HDR(hdr)); multilist_insert(new_state->arcs_list[buftype], hdr); if (GHOST_STATE(new_state)) { ASSERT0(bufcnt); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); update_new = B_TRUE; } arc_evictable_space_increment(hdr, new_state); } } ASSERT(!HDR_EMPTY(hdr)); if (new_state == arc_anon && HDR_IN_HASH_TABLE(hdr)) buf_hash_remove(hdr); /* adjust state sizes (ignore arc_l2c_only) */ if (update_new && new_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(new_state)) { ASSERT0(bufcnt); /* * When moving a header to a ghost state, we first * remove all arc buffers. Thus, we'll have a * bufcnt of zero, and no arc buffer to use for * the reference. As a result, we use the arc * header pointer for the reference. */ (void) refcount_add_many(&new_state->arcs_size, HDR_GET_LSIZE(hdr), hdr); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); } else { uint32_t buffers = 0; /* * Each individual buffer holds a unique reference, * thus we must remove each of these references one * at a time. */ for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { ASSERT3U(bufcnt, !=, 0); buffers++; /* * When the arc_buf_t is sharing the data * block with the hdr, the owner of the * reference belongs to the hdr. Only * add to the refcount if the arc_buf_t is * not shared. */ if (arc_buf_is_shared(buf)) continue; (void) refcount_add_many(&new_state->arcs_size, arc_buf_size(buf), buf); } ASSERT3U(bufcnt, ==, buffers); if (hdr->b_l1hdr.b_pabd != NULL) { (void) refcount_add_many(&new_state->arcs_size, arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) refcount_add_many(&new_state->arcs_size, HDR_GET_PSIZE(hdr), hdr); } } } if (update_old && old_state != arc_l2c_only) { ASSERT(HDR_HAS_L1HDR(hdr)); if (GHOST_STATE(old_state)) { ASSERT0(bufcnt); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); /* * When moving a header off of a ghost state, * the header will not contain any arc buffers. * We use the arc header pointer for the reference * which is exactly what we did when we put the * header on the ghost state. */ (void) refcount_remove_many(&old_state->arcs_size, HDR_GET_LSIZE(hdr), hdr); } else { uint32_t buffers = 0; /* * Each individual buffer holds a unique reference, * thus we must remove each of these references one * at a time. */ for (arc_buf_t *buf = hdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { ASSERT3U(bufcnt, !=, 0); buffers++; /* * When the arc_buf_t is sharing the data * block with the hdr, the owner of the * reference belongs to the hdr. Only * add to the refcount if the arc_buf_t is * not shared. */ if (arc_buf_is_shared(buf)) continue; (void) refcount_remove_many( &old_state->arcs_size, arc_buf_size(buf), buf); } ASSERT3U(bufcnt, ==, buffers); ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); if (hdr->b_l1hdr.b_pabd != NULL) { (void) refcount_remove_many( &old_state->arcs_size, arc_hdr_size(hdr), hdr); } if (HDR_HAS_RABD(hdr)) { (void) refcount_remove_many( &old_state->arcs_size, HDR_GET_PSIZE(hdr), hdr); } } } if (HDR_HAS_L1HDR(hdr)) hdr->b_l1hdr.b_state = new_state; /* * L2 headers should never be on the L2 state list since they don't * have L1 headers allocated. */ ASSERT(multilist_is_empty(arc_l2c_only->arcs_list[ARC_BUFC_DATA]) && multilist_is_empty(arc_l2c_only->arcs_list[ARC_BUFC_METADATA])); } void arc_space_consume(uint64_t space, arc_space_type_t type) { ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES); switch (type) { default: break; case ARC_SPACE_DATA: aggsum_add(&astat_data_size, space); break; case ARC_SPACE_META: aggsum_add(&astat_metadata_size, space); break; case ARC_SPACE_BONUS: aggsum_add(&astat_bonus_size, space); break; case ARC_SPACE_DNODE: aggsum_add(&astat_dnode_size, space); break; case ARC_SPACE_DBUF: aggsum_add(&astat_dbuf_size, space); break; case ARC_SPACE_HDRS: aggsum_add(&astat_hdr_size, space); break; case ARC_SPACE_L2HDRS: aggsum_add(&astat_l2_hdr_size, space); break; } if (type != ARC_SPACE_DATA) aggsum_add(&arc_meta_used, space); aggsum_add(&arc_size, space); } void arc_space_return(uint64_t space, arc_space_type_t type) { ASSERT(type >= 0 && type < ARC_SPACE_NUMTYPES); switch (type) { default: break; case ARC_SPACE_DATA: aggsum_add(&astat_data_size, -space); break; case ARC_SPACE_META: aggsum_add(&astat_metadata_size, -space); break; case ARC_SPACE_BONUS: aggsum_add(&astat_bonus_size, -space); break; case ARC_SPACE_DNODE: aggsum_add(&astat_dnode_size, -space); break; case ARC_SPACE_DBUF: aggsum_add(&astat_dbuf_size, -space); break; case ARC_SPACE_HDRS: aggsum_add(&astat_hdr_size, -space); break; case ARC_SPACE_L2HDRS: aggsum_add(&astat_l2_hdr_size, -space); break; } if (type != ARC_SPACE_DATA) { ASSERT(aggsum_compare(&arc_meta_used, space) >= 0); /* * We use the upper bound here rather than the precise value * because the arc_meta_max value doesn't need to be * precise. It's only consumed by humans via arcstats. */ if (arc_meta_max < aggsum_upper_bound(&arc_meta_used)) arc_meta_max = aggsum_upper_bound(&arc_meta_used); aggsum_add(&arc_meta_used, -space); } ASSERT(aggsum_compare(&arc_size, space) >= 0); aggsum_add(&arc_size, -space); } /* * Given a hdr and a buf, returns whether that buf can share its b_data buffer * with the hdr's b_pabd. */ static boolean_t arc_can_share(arc_buf_hdr_t *hdr, arc_buf_t *buf) { /* * The criteria for sharing a hdr's data are: * 1. the buffer is not encrypted * 2. the hdr's compression matches the buf's compression * 3. the hdr doesn't need to be byteswapped * 4. the hdr isn't already being shared * 5. the buf is either compressed or it is the last buf in the hdr list * * Criterion #5 maintains the invariant that shared uncompressed * bufs must be the final buf in the hdr's b_buf list. Reading this, you * might ask, "if a compressed buf is allocated first, won't that be the * last thing in the list?", but in that case it's impossible to create * a shared uncompressed buf anyway (because the hdr must be compressed * to have the compressed buf). You might also think that #3 is * sufficient to make this guarantee, however it's possible * (specifically in the rare L2ARC write race mentioned in * arc_buf_alloc_impl()) there will be an existing uncompressed buf that * is sharable, but wasn't at the time of its allocation. Rather than * allow a new shared uncompressed buf to be created and then shuffle * the list around to make it the last element, this simply disallows * sharing if the new buf isn't the first to be added. */ ASSERT3P(buf->b_hdr, ==, hdr); boolean_t hdr_compressed = arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF; boolean_t buf_compressed = ARC_BUF_COMPRESSED(buf) != 0; return (!ARC_BUF_ENCRYPTED(buf) && buf_compressed == hdr_compressed && hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS && !HDR_SHARED_DATA(hdr) && (ARC_BUF_LAST(buf) || ARC_BUF_COMPRESSED(buf))); } /* * Allocate a buf for this hdr. If you care about the data that's in the hdr, * or if you want a compressed buffer, pass those flags in. Returns 0 if the * copy was made successfully, or an error code otherwise. */ static int arc_buf_alloc_impl(arc_buf_hdr_t *hdr, spa_t *spa, const zbookmark_phys_t *zb, void *tag, boolean_t encrypted, boolean_t compressed, boolean_t noauth, boolean_t fill, arc_buf_t **ret) { arc_buf_t *buf; arc_fill_flags_t flags = ARC_FILL_LOCKED; ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3U(HDR_GET_LSIZE(hdr), >, 0); VERIFY(hdr->b_type == ARC_BUFC_DATA || hdr->b_type == ARC_BUFC_METADATA); ASSERT3P(ret, !=, NULL); ASSERT3P(*ret, ==, NULL); IMPLY(encrypted, compressed); hdr->b_l1hdr.b_mru_hits = 0; hdr->b_l1hdr.b_mru_ghost_hits = 0; hdr->b_l1hdr.b_mfu_hits = 0; hdr->b_l1hdr.b_mfu_ghost_hits = 0; hdr->b_l1hdr.b_l2_hits = 0; buf = *ret = kmem_cache_alloc(buf_cache, KM_PUSHPAGE); buf->b_hdr = hdr; buf->b_data = NULL; buf->b_next = hdr->b_l1hdr.b_buf; buf->b_flags = 0; add_reference(hdr, tag); /* * We're about to change the hdr's b_flags. We must either * hold the hash_lock or be undiscoverable. */ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); /* * Only honor requests for compressed bufs if the hdr is actually * compressed. This must be overriden if the buffer is encrypted since * encrypted buffers cannot be decompressed. */ if (encrypted) { buf->b_flags |= ARC_BUF_FLAG_COMPRESSED; buf->b_flags |= ARC_BUF_FLAG_ENCRYPTED; flags |= ARC_FILL_COMPRESSED | ARC_FILL_ENCRYPTED; } else if (compressed && arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) { buf->b_flags |= ARC_BUF_FLAG_COMPRESSED; flags |= ARC_FILL_COMPRESSED; } if (noauth) { ASSERT0(encrypted); flags |= ARC_FILL_NOAUTH; } /* * If the hdr's data can be shared then we share the data buffer and * set the appropriate bit in the hdr's b_flags to indicate the hdr is * allocate a new buffer to store the buf's data. * * There are two additional restrictions here because we're sharing * hdr -> buf instead of the usual buf -> hdr. First, the hdr can't be * actively involved in an L2ARC write, because if this buf is used by * an arc_write() then the hdr's data buffer will be released when the * write completes, even though the L2ARC write might still be using it. * Second, the hdr's ABD must be linear so that the buf's user doesn't * need to be ABD-aware. */ boolean_t can_share = arc_can_share(hdr, buf) && !HDR_L2_WRITING(hdr) && hdr->b_l1hdr.b_pabd != NULL && abd_is_linear(hdr->b_l1hdr.b_pabd); /* Set up b_data and sharing */ if (can_share) { buf->b_data = abd_to_buf(hdr->b_l1hdr.b_pabd); buf->b_flags |= ARC_BUF_FLAG_SHARED; arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA); } else { buf->b_data = arc_get_data_buf(hdr, arc_buf_size(buf), buf); ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf)); } VERIFY3P(buf->b_data, !=, NULL); hdr->b_l1hdr.b_buf = buf; hdr->b_l1hdr.b_bufcnt += 1; if (encrypted) hdr->b_crypt_hdr.b_ebufcnt += 1; /* * If the user wants the data from the hdr, we need to either copy or * decompress the data. */ if (fill) { ASSERT3P(zb, !=, NULL); return (arc_buf_fill(buf, spa, zb, flags)); } return (0); } static char *arc_onloan_tag = "onloan"; static inline void arc_loaned_bytes_update(int64_t delta) { atomic_add_64(&arc_loaned_bytes, delta); /* assert that it did not wrap around */ ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0); } /* * Loan out an anonymous arc buffer. Loaned buffers are not counted as in * flight data by arc_tempreserve_space() until they are "returned". Loaned * buffers must be returned to the arc before they can be used by the DMU or * freed. */ arc_buf_t * arc_loan_buf(spa_t *spa, boolean_t is_metadata, int size) { arc_buf_t *buf = arc_alloc_buf(spa, arc_onloan_tag, is_metadata ? ARC_BUFC_METADATA : ARC_BUFC_DATA, size); arc_loaned_bytes_update(arc_buf_size(buf)); return (buf); } arc_buf_t * arc_loan_compressed_buf(spa_t *spa, uint64_t psize, uint64_t lsize, enum zio_compress compression_type) { arc_buf_t *buf = arc_alloc_compressed_buf(spa, arc_onloan_tag, psize, lsize, compression_type); arc_loaned_bytes_update(arc_buf_size(buf)); return (buf); } arc_buf_t * arc_loan_raw_buf(spa_t *spa, uint64_t dsobj, boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_object_type_t ot, uint64_t psize, uint64_t lsize, enum zio_compress compression_type) { arc_buf_t *buf = arc_alloc_raw_buf(spa, arc_onloan_tag, dsobj, byteorder, salt, iv, mac, ot, psize, lsize, compression_type); atomic_add_64(&arc_loaned_bytes, psize); return (buf); } /* * Return a loaned arc buffer to the arc. */ void arc_return_buf(arc_buf_t *buf, void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(buf->b_data, !=, NULL); ASSERT(HDR_HAS_L1HDR(hdr)); (void) refcount_add(&hdr->b_l1hdr.b_refcnt, tag); (void) refcount_remove(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag); arc_loaned_bytes_update(-arc_buf_size(buf)); } /* Detach an arc_buf from a dbuf (tag) */ void arc_loan_inuse_buf(arc_buf_t *buf, void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(buf->b_data, !=, NULL); ASSERT(HDR_HAS_L1HDR(hdr)); (void) refcount_add(&hdr->b_l1hdr.b_refcnt, arc_onloan_tag); (void) refcount_remove(&hdr->b_l1hdr.b_refcnt, tag); arc_loaned_bytes_update(arc_buf_size(buf)); } static void l2arc_free_abd_on_write(abd_t *abd, size_t size, arc_buf_contents_t type) { l2arc_data_free_t *df = kmem_alloc(sizeof (*df), KM_SLEEP); df->l2df_abd = abd; df->l2df_size = size; df->l2df_type = type; mutex_enter(&l2arc_free_on_write_mtx); list_insert_head(l2arc_free_on_write, df); mutex_exit(&l2arc_free_on_write_mtx); } static void arc_hdr_free_on_write(arc_buf_hdr_t *hdr, boolean_t free_rdata) { arc_state_t *state = hdr->b_l1hdr.b_state; arc_buf_contents_t type = arc_buf_type(hdr); uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr); /* protected by hash lock, if in the hash table */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT(state != arc_anon && state != arc_l2c_only); (void) refcount_remove_many(&state->arcs_esize[type], size, hdr); } (void) refcount_remove_many(&state->arcs_size, size, hdr); if (type == ARC_BUFC_METADATA) { arc_space_return(size, ARC_SPACE_META); } else { ASSERT(type == ARC_BUFC_DATA); arc_space_return(size, ARC_SPACE_DATA); } if (free_rdata) { l2arc_free_abd_on_write(hdr->b_crypt_hdr.b_rabd, size, type); } else { l2arc_free_abd_on_write(hdr->b_l1hdr.b_pabd, size, type); } } /* * Share the arc_buf_t's data with the hdr. Whenever we are sharing the * data buffer, we transfer the refcount ownership to the hdr and update * the appropriate kstats. */ static void arc_share_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf) { ASSERT(arc_can_share(hdr, buf)); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!ARC_BUF_ENCRYPTED(buf)); ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); /* * Start sharing the data buffer. We transfer the * refcount ownership to the hdr since it always owns * the refcount whenever an arc_buf_t is shared. */ refcount_transfer_ownership(&hdr->b_l1hdr.b_state->arcs_size, buf, hdr); hdr->b_l1hdr.b_pabd = abd_get_from_buf(buf->b_data, arc_buf_size(buf)); abd_take_ownership_of_buf(hdr->b_l1hdr.b_pabd, HDR_ISTYPE_METADATA(hdr)); arc_hdr_set_flags(hdr, ARC_FLAG_SHARED_DATA); buf->b_flags |= ARC_BUF_FLAG_SHARED; /* * Since we've transferred ownership to the hdr we need * to increment its compressed and uncompressed kstats and * decrement the overhead size. */ ARCSTAT_INCR(arcstat_compressed_size, arc_hdr_size(hdr)); ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr)); ARCSTAT_INCR(arcstat_overhead_size, -arc_buf_size(buf)); } static void arc_unshare_buf(arc_buf_hdr_t *hdr, arc_buf_t *buf) { ASSERT(arc_buf_is_shared(buf)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); /* * We are no longer sharing this buffer so we need * to transfer its ownership to the rightful owner. */ refcount_transfer_ownership(&hdr->b_l1hdr.b_state->arcs_size, hdr, buf); arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA); abd_release_ownership_of_buf(hdr->b_l1hdr.b_pabd); abd_put(hdr->b_l1hdr.b_pabd); hdr->b_l1hdr.b_pabd = NULL; buf->b_flags &= ~ARC_BUF_FLAG_SHARED; /* * Since the buffer is no longer shared between * the arc buf and the hdr, count it as overhead. */ ARCSTAT_INCR(arcstat_compressed_size, -arc_hdr_size(hdr)); ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr)); ARCSTAT_INCR(arcstat_overhead_size, arc_buf_size(buf)); } /* * Remove an arc_buf_t from the hdr's buf list and return the last * arc_buf_t on the list. If no buffers remain on the list then return * NULL. */ static arc_buf_t * arc_buf_remove(arc_buf_hdr_t *hdr, arc_buf_t *buf) { ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); arc_buf_t **bufp = &hdr->b_l1hdr.b_buf; arc_buf_t *lastbuf = NULL; /* * Remove the buf from the hdr list and locate the last * remaining buffer on the list. */ while (*bufp != NULL) { if (*bufp == buf) *bufp = buf->b_next; /* * If we've removed a buffer in the middle of * the list then update the lastbuf and update * bufp. */ if (*bufp != NULL) { lastbuf = *bufp; bufp = &(*bufp)->b_next; } } buf->b_next = NULL; ASSERT3P(lastbuf, !=, buf); IMPLY(hdr->b_l1hdr.b_bufcnt > 0, lastbuf != NULL); IMPLY(hdr->b_l1hdr.b_bufcnt > 0, hdr->b_l1hdr.b_buf != NULL); IMPLY(lastbuf != NULL, ARC_BUF_LAST(lastbuf)); return (lastbuf); } /* * Free up buf->b_data and pull the arc_buf_t off of the the arc_buf_hdr_t's * list and free it. */ static void arc_buf_destroy_impl(arc_buf_t *buf) { arc_buf_hdr_t *hdr = buf->b_hdr; /* * Free up the data associated with the buf but only if we're not * sharing this with the hdr. If we are sharing it with the hdr, the * hdr is responsible for doing the free. */ if (buf->b_data != NULL) { /* * We're about to change the hdr's b_flags. We must either * hold the hash_lock or be undiscoverable. */ ASSERT(MUTEX_HELD(HDR_LOCK(hdr)) || HDR_EMPTY(hdr)); arc_cksum_verify(buf); arc_buf_unwatch(buf); if (arc_buf_is_shared(buf)) { arc_hdr_clear_flags(hdr, ARC_FLAG_SHARED_DATA); } else { uint64_t size = arc_buf_size(buf); arc_free_data_buf(hdr, buf->b_data, size, buf); ARCSTAT_INCR(arcstat_overhead_size, -size); } buf->b_data = NULL; ASSERT(hdr->b_l1hdr.b_bufcnt > 0); hdr->b_l1hdr.b_bufcnt -= 1; if (ARC_BUF_ENCRYPTED(buf)) { hdr->b_crypt_hdr.b_ebufcnt -= 1; /* * If we have no more encrypted buffers and we've * already gotten a copy of the decrypted data we can * free b_rabd to save some space. */ if (hdr->b_crypt_hdr.b_ebufcnt == 0 && HDR_HAS_RABD(hdr) && hdr->b_l1hdr.b_pabd != NULL && !HDR_IO_IN_PROGRESS(hdr)) { arc_hdr_free_abd(hdr, B_TRUE); } } } arc_buf_t *lastbuf = arc_buf_remove(hdr, buf); if (ARC_BUF_SHARED(buf) && !ARC_BUF_COMPRESSED(buf)) { /* * If the current arc_buf_t is sharing its data buffer with the * hdr, then reassign the hdr's b_pabd to share it with the new * buffer at the end of the list. The shared buffer is always * the last one on the hdr's buffer list. * * There is an equivalent case for compressed bufs, but since * they aren't guaranteed to be the last buf in the list and * that is an exceedingly rare case, we just allow that space be * wasted temporarily. We must also be careful not to share * encrypted buffers, since they cannot be shared. */ if (lastbuf != NULL && !ARC_BUF_ENCRYPTED(lastbuf)) { /* Only one buf can be shared at once */ VERIFY(!arc_buf_is_shared(lastbuf)); /* hdr is uncompressed so can't have compressed buf */ VERIFY(!ARC_BUF_COMPRESSED(lastbuf)); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); arc_hdr_free_abd(hdr, B_FALSE); /* * We must setup a new shared block between the * last buffer and the hdr. The data would have * been allocated by the arc buf so we need to transfer * ownership to the hdr since it's now being shared. */ arc_share_buf(hdr, lastbuf); } } else if (HDR_SHARED_DATA(hdr)) { /* * Uncompressed shared buffers are always at the end * of the list. Compressed buffers don't have the * same requirements. This makes it hard to * simply assert that the lastbuf is shared so * we rely on the hdr's compression flags to determine * if we have a compressed, shared buffer. */ ASSERT3P(lastbuf, !=, NULL); ASSERT(arc_buf_is_shared(lastbuf) || arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF); } /* * Free the checksum if we're removing the last uncompressed buf from * this hdr. */ if (!arc_hdr_has_uncompressed_buf(hdr)) { arc_cksum_free(hdr); } /* clean up the buf */ buf->b_hdr = NULL; kmem_cache_free(buf_cache, buf); } static void arc_hdr_alloc_abd(arc_buf_hdr_t *hdr, boolean_t alloc_rdata) { uint64_t size; ASSERT3U(HDR_GET_LSIZE(hdr), >, 0); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!HDR_SHARED_DATA(hdr) || alloc_rdata); IMPLY(alloc_rdata, HDR_PROTECTED(hdr)); if (alloc_rdata) { size = HDR_GET_PSIZE(hdr); ASSERT3P(hdr->b_crypt_hdr.b_rabd, ==, NULL); hdr->b_crypt_hdr.b_rabd = arc_get_data_abd(hdr, size, hdr); ASSERT3P(hdr->b_crypt_hdr.b_rabd, !=, NULL); ARCSTAT_INCR(arcstat_raw_size, size); } else { size = arc_hdr_size(hdr); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); hdr->b_l1hdr.b_pabd = arc_get_data_abd(hdr, size, hdr); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); } ARCSTAT_INCR(arcstat_compressed_size, size); ARCSTAT_INCR(arcstat_uncompressed_size, HDR_GET_LSIZE(hdr)); } static void arc_hdr_free_abd(arc_buf_hdr_t *hdr, boolean_t free_rdata) { uint64_t size = (free_rdata) ? HDR_GET_PSIZE(hdr) : arc_hdr_size(hdr); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); IMPLY(free_rdata, HDR_HAS_RABD(hdr)); /* * If the hdr is currently being written to the l2arc then * we defer freeing the data by adding it to the l2arc_free_on_write * list. The l2arc will free the data once it's finished * writing it to the l2arc device. */ if (HDR_L2_WRITING(hdr)) { arc_hdr_free_on_write(hdr, free_rdata); ARCSTAT_BUMP(arcstat_l2_free_on_write); } else if (free_rdata) { arc_free_data_abd(hdr, hdr->b_crypt_hdr.b_rabd, size, hdr); } else { arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, size, hdr); } if (free_rdata) { hdr->b_crypt_hdr.b_rabd = NULL; ARCSTAT_INCR(arcstat_raw_size, -size); } else { hdr->b_l1hdr.b_pabd = NULL; } if (hdr->b_l1hdr.b_pabd == NULL && !HDR_HAS_RABD(hdr)) hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; ARCSTAT_INCR(arcstat_compressed_size, -size); ARCSTAT_INCR(arcstat_uncompressed_size, -HDR_GET_LSIZE(hdr)); } static arc_buf_hdr_t * arc_hdr_alloc(uint64_t spa, int32_t psize, int32_t lsize, boolean_t protected, enum zio_compress compression_type, arc_buf_contents_t type, boolean_t alloc_rdata) { arc_buf_hdr_t *hdr; VERIFY(type == ARC_BUFC_DATA || type == ARC_BUFC_METADATA); if (protected) { hdr = kmem_cache_alloc(hdr_full_crypt_cache, KM_PUSHPAGE); } else { hdr = kmem_cache_alloc(hdr_full_cache, KM_PUSHPAGE); } ASSERT(HDR_EMPTY(hdr)); ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); HDR_SET_PSIZE(hdr, psize); HDR_SET_LSIZE(hdr, lsize); hdr->b_spa = spa; hdr->b_type = type; hdr->b_flags = 0; arc_hdr_set_flags(hdr, arc_bufc_to_flags(type) | ARC_FLAG_HAS_L1HDR); arc_hdr_set_compress(hdr, compression_type); if (protected) arc_hdr_set_flags(hdr, ARC_FLAG_PROTECTED); hdr->b_l1hdr.b_state = arc_anon; hdr->b_l1hdr.b_arc_access = 0; hdr->b_l1hdr.b_bufcnt = 0; hdr->b_l1hdr.b_buf = NULL; /* * Allocate the hdr's buffer. This will contain either * the compressed or uncompressed data depending on the block * it references and compressed arc enablement. */ arc_hdr_alloc_abd(hdr, alloc_rdata); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); return (hdr); } /* * Transition between the two allocation states for the arc_buf_hdr struct. * The arc_buf_hdr struct can be allocated with (hdr_full_cache) or without * (hdr_l2only_cache) the fields necessary for the L1 cache - the smaller * version is used when a cache buffer is only in the L2ARC in order to reduce * memory usage. */ static arc_buf_hdr_t * arc_hdr_realloc(arc_buf_hdr_t *hdr, kmem_cache_t *old, kmem_cache_t *new) { ASSERT(HDR_HAS_L2HDR(hdr)); arc_buf_hdr_t *nhdr; l2arc_dev_t *dev = hdr->b_l2hdr.b_dev; ASSERT((old == hdr_full_cache && new == hdr_l2only_cache) || (old == hdr_l2only_cache && new == hdr_full_cache)); /* * if the caller wanted a new full header and the header is to be * encrypted we will actually allocate the header from the full crypt * cache instead. The same applies to freeing from the old cache. */ if (HDR_PROTECTED(hdr) && new == hdr_full_cache) new = hdr_full_crypt_cache; if (HDR_PROTECTED(hdr) && old == hdr_full_cache) old = hdr_full_crypt_cache; nhdr = kmem_cache_alloc(new, KM_PUSHPAGE); ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); buf_hash_remove(hdr); bcopy(hdr, nhdr, HDR_L2ONLY_SIZE); if (new == hdr_full_cache || new == hdr_full_crypt_cache) { arc_hdr_set_flags(nhdr, ARC_FLAG_HAS_L1HDR); /* * arc_access and arc_change_state need to be aware that a * header has just come out of L2ARC, so we set its state to * l2c_only even though it's about to change. */ nhdr->b_l1hdr.b_state = arc_l2c_only; /* Verify previous threads set to NULL before freeing */ ASSERT3P(nhdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); } else { ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); ASSERT0(hdr->b_l1hdr.b_bufcnt); ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); /* * If we've reached here, We must have been called from * arc_evict_hdr(), as such we should have already been * removed from any ghost list we were previously on * (which protects us from racing with arc_evict_state), * thus no locking is needed during this check. */ ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); /* * A buffer must not be moved into the arc_l2c_only * state if it's not finished being written out to the * l2arc device. Otherwise, the b_l1hdr.b_pabd field * might try to be accessed, even though it was removed. */ VERIFY(!HDR_L2_WRITING(hdr)); VERIFY3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); arc_hdr_clear_flags(nhdr, ARC_FLAG_HAS_L1HDR); } /* * The header has been reallocated so we need to re-insert it into any * lists it was on. */ (void) buf_hash_insert(nhdr, NULL); ASSERT(list_link_active(&hdr->b_l2hdr.b_l2node)); mutex_enter(&dev->l2ad_mtx); /* * We must place the realloc'ed header back into the list at * the same spot. Otherwise, if it's placed earlier in the list, * l2arc_write_buffers() could find it during the function's * write phase, and try to write it out to the l2arc. */ list_insert_after(&dev->l2ad_buflist, hdr, nhdr); list_remove(&dev->l2ad_buflist, hdr); mutex_exit(&dev->l2ad_mtx); /* * Since we're using the pointer address as the tag when * incrementing and decrementing the l2ad_alloc refcount, we * must remove the old pointer (that we're about to destroy) and * add the new pointer to the refcount. Otherwise we'd remove * the wrong pointer address when calling arc_hdr_destroy() later. */ (void) refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); (void) refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(nhdr), nhdr); buf_discard_identity(hdr); kmem_cache_free(old, hdr); return (nhdr); } /* * This function allows an L1 header to be reallocated as a crypt * header and vice versa. If we are going to a crypt header, the * new fields will be zeroed out. */ static arc_buf_hdr_t * arc_hdr_realloc_crypt(arc_buf_hdr_t *hdr, boolean_t need_crypt) { arc_buf_hdr_t *nhdr; arc_buf_t *buf; kmem_cache_t *ncache, *ocache; unsigned nsize, osize; /* * This function requires that hdr is in the arc_anon state. * Therefore it won't have any L2ARC data for us to worry * about copying. */ ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!HDR_HAS_L2HDR(hdr)); ASSERT3U(!!HDR_PROTECTED(hdr), !=, need_crypt); ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT(!list_link_active(&hdr->b_l2hdr.b_l2node)); ASSERT3P(hdr->b_hash_next, ==, NULL); if (need_crypt) { ncache = hdr_full_crypt_cache; nsize = sizeof (hdr->b_crypt_hdr); ocache = hdr_full_cache; osize = HDR_FULL_SIZE; } else { ncache = hdr_full_cache; nsize = HDR_FULL_SIZE; ocache = hdr_full_crypt_cache; osize = sizeof (hdr->b_crypt_hdr); } nhdr = kmem_cache_alloc(ncache, KM_PUSHPAGE); /* * Copy all members that aren't locks or condvars to the new header. * No lists are pointing to us (as we asserted above), so we don't * need to worry about the list nodes. */ nhdr->b_dva = hdr->b_dva; nhdr->b_birth = hdr->b_birth; nhdr->b_type = hdr->b_type; nhdr->b_flags = hdr->b_flags; nhdr->b_psize = hdr->b_psize; nhdr->b_lsize = hdr->b_lsize; nhdr->b_spa = hdr->b_spa; nhdr->b_l1hdr.b_freeze_cksum = hdr->b_l1hdr.b_freeze_cksum; nhdr->b_l1hdr.b_bufcnt = hdr->b_l1hdr.b_bufcnt; nhdr->b_l1hdr.b_byteswap = hdr->b_l1hdr.b_byteswap; nhdr->b_l1hdr.b_state = hdr->b_l1hdr.b_state; nhdr->b_l1hdr.b_arc_access = hdr->b_l1hdr.b_arc_access; nhdr->b_l1hdr.b_mru_hits = hdr->b_l1hdr.b_mru_hits; nhdr->b_l1hdr.b_mru_ghost_hits = hdr->b_l1hdr.b_mru_ghost_hits; nhdr->b_l1hdr.b_mfu_hits = hdr->b_l1hdr.b_mfu_hits; nhdr->b_l1hdr.b_mfu_ghost_hits = hdr->b_l1hdr.b_mfu_ghost_hits; nhdr->b_l1hdr.b_l2_hits = hdr->b_l1hdr.b_l2_hits; nhdr->b_l1hdr.b_acb = hdr->b_l1hdr.b_acb; nhdr->b_l1hdr.b_pabd = hdr->b_l1hdr.b_pabd; /* * This refcount_add() exists only to ensure that the individual * arc buffers always point to a header that is referenced, avoiding * a small race condition that could trigger ASSERTs. */ (void) refcount_add(&nhdr->b_l1hdr.b_refcnt, FTAG); nhdr->b_l1hdr.b_buf = hdr->b_l1hdr.b_buf; for (buf = nhdr->b_l1hdr.b_buf; buf != NULL; buf = buf->b_next) { mutex_enter(&buf->b_evict_lock); buf->b_hdr = nhdr; mutex_exit(&buf->b_evict_lock); } refcount_transfer(&nhdr->b_l1hdr.b_refcnt, &hdr->b_l1hdr.b_refcnt); (void) refcount_remove(&nhdr->b_l1hdr.b_refcnt, FTAG); ASSERT0(refcount_count(&hdr->b_l1hdr.b_refcnt)); if (need_crypt) { arc_hdr_set_flags(nhdr, ARC_FLAG_PROTECTED); } else { arc_hdr_clear_flags(nhdr, ARC_FLAG_PROTECTED); } /* unset all members of the original hdr */ bzero(&hdr->b_dva, sizeof (dva_t)); hdr->b_birth = 0; hdr->b_type = ARC_BUFC_INVALID; hdr->b_flags = 0; hdr->b_psize = 0; hdr->b_lsize = 0; hdr->b_spa = 0; hdr->b_l1hdr.b_freeze_cksum = NULL; hdr->b_l1hdr.b_buf = NULL; hdr->b_l1hdr.b_bufcnt = 0; hdr->b_l1hdr.b_byteswap = 0; hdr->b_l1hdr.b_state = NULL; hdr->b_l1hdr.b_arc_access = 0; hdr->b_l1hdr.b_mru_hits = 0; hdr->b_l1hdr.b_mru_ghost_hits = 0; hdr->b_l1hdr.b_mfu_hits = 0; hdr->b_l1hdr.b_mfu_ghost_hits = 0; hdr->b_l1hdr.b_l2_hits = 0; hdr->b_l1hdr.b_acb = NULL; hdr->b_l1hdr.b_pabd = NULL; if (ocache == hdr_full_crypt_cache) { ASSERT(!HDR_HAS_RABD(hdr)); hdr->b_crypt_hdr.b_ot = DMU_OT_NONE; hdr->b_crypt_hdr.b_ebufcnt = 0; hdr->b_crypt_hdr.b_dsobj = 0; bzero(hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN); bzero(hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN); bzero(hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN); } buf_discard_identity(hdr); kmem_cache_free(ocache, hdr); return (nhdr); } /* * This function is used by the send / receive code to convert a newly * allocated arc_buf_t to one that is suitable for a raw encrypted write. It * is also used to allow the root objset block to be uupdated without altering * its embedded MACs. Both block types will always be uncompressed so we do not * have to worry about compression type or psize. */ void arc_convert_to_raw(arc_buf_t *buf, uint64_t dsobj, boolean_t byteorder, dmu_object_type_t ot, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac) { arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT(ot == DMU_OT_DNODE || ot == DMU_OT_OBJSET); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); buf->b_flags |= (ARC_BUF_FLAG_COMPRESSED | ARC_BUF_FLAG_ENCRYPTED); if (!HDR_PROTECTED(hdr)) hdr = arc_hdr_realloc_crypt(hdr, B_TRUE); hdr->b_crypt_hdr.b_dsobj = dsobj; hdr->b_crypt_hdr.b_ot = ot; hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ? DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot); if (!arc_hdr_has_uncompressed_buf(hdr)) arc_cksum_free(hdr); if (salt != NULL) bcopy(salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN); if (iv != NULL) bcopy(iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN); if (mac != NULL) bcopy(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN); } /* * Allocate a new arc_buf_hdr_t and arc_buf_t and return the buf to the caller. * The buf is returned thawed since we expect the consumer to modify it. */ arc_buf_t * arc_alloc_buf(spa_t *spa, void *tag, arc_buf_contents_t type, int32_t size) { arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), size, size, B_FALSE, ZIO_COMPRESS_OFF, type, B_FALSE); ASSERT(!MUTEX_HELD(HDR_LOCK(hdr))); arc_buf_t *buf = NULL; VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_FALSE, B_FALSE, B_FALSE, &buf)); arc_buf_thaw(buf); return (buf); } /* * Allocate a compressed buf in the same manner as arc_alloc_buf. Don't use this * for bufs containing metadata. */ arc_buf_t * arc_alloc_compressed_buf(spa_t *spa, void *tag, uint64_t psize, uint64_t lsize, enum zio_compress compression_type) { ASSERT3U(lsize, >, 0); ASSERT3U(lsize, >=, psize); ASSERT3U(compression_type, >, ZIO_COMPRESS_OFF); ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS); arc_buf_hdr_t *hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_FALSE, compression_type, ARC_BUFC_DATA, B_FALSE); ASSERT(!MUTEX_HELD(HDR_LOCK(hdr))); arc_buf_t *buf = NULL; VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_FALSE, B_TRUE, B_FALSE, B_FALSE, &buf)); arc_buf_thaw(buf); ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); if (!arc_buf_is_shared(buf)) { /* * To ensure that the hdr has the correct data in it if we call * arc_untransform() on this buf before it's been written to * disk, it's easiest if we just set up sharing between the * buf and the hdr. */ ASSERT(!abd_is_linear(hdr->b_l1hdr.b_pabd)); arc_hdr_free_abd(hdr, B_FALSE); arc_share_buf(hdr, buf); } return (buf); } arc_buf_t * arc_alloc_raw_buf(spa_t *spa, void *tag, uint64_t dsobj, boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_object_type_t ot, uint64_t psize, uint64_t lsize, enum zio_compress compression_type) { arc_buf_hdr_t *hdr; arc_buf_t *buf; arc_buf_contents_t type = DMU_OT_IS_METADATA(ot) ? ARC_BUFC_METADATA : ARC_BUFC_DATA; ASSERT3U(lsize, >, 0); ASSERT3U(lsize, >=, psize); ASSERT3U(compression_type, >=, ZIO_COMPRESS_OFF); ASSERT3U(compression_type, <, ZIO_COMPRESS_FUNCTIONS); hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, B_TRUE, compression_type, type, B_TRUE); ASSERT(!MUTEX_HELD(HDR_LOCK(hdr))); hdr->b_crypt_hdr.b_dsobj = dsobj; hdr->b_crypt_hdr.b_ot = ot; hdr->b_l1hdr.b_byteswap = (byteorder == ZFS_HOST_BYTEORDER) ? DMU_BSWAP_NUMFUNCS : DMU_OT_BYTESWAP(ot); bcopy(salt, hdr->b_crypt_hdr.b_salt, ZIO_DATA_SALT_LEN); bcopy(iv, hdr->b_crypt_hdr.b_iv, ZIO_DATA_IV_LEN); bcopy(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN); /* * This buffer will be considered encrypted even if the ot is not an * encrypted type. It will become authenticated instead in * arc_write_ready(). */ buf = NULL; VERIFY0(arc_buf_alloc_impl(hdr, spa, NULL, tag, B_TRUE, B_TRUE, B_FALSE, B_FALSE, &buf)); arc_buf_thaw(buf); ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); return (buf); } static void arc_hdr_l2hdr_destroy(arc_buf_hdr_t *hdr) { l2arc_buf_hdr_t *l2hdr = &hdr->b_l2hdr; l2arc_dev_t *dev = l2hdr->b_dev; uint64_t psize = arc_hdr_size(hdr); ASSERT(MUTEX_HELD(&dev->l2ad_mtx)); ASSERT(HDR_HAS_L2HDR(hdr)); list_remove(&dev->l2ad_buflist, hdr); ARCSTAT_INCR(arcstat_l2_psize, -psize); ARCSTAT_INCR(arcstat_l2_lsize, -HDR_GET_LSIZE(hdr)); vdev_space_update(dev->l2ad_vdev, -psize, 0, 0); (void) refcount_remove_many(&dev->l2ad_alloc, psize, hdr); arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR); } static void arc_hdr_destroy(arc_buf_hdr_t *hdr) { if (HDR_HAS_L1HDR(hdr)) { ASSERT(hdr->b_l1hdr.b_buf == NULL || hdr->b_l1hdr.b_bufcnt > 0); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); } ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(!HDR_IN_HASH_TABLE(hdr)); if (!HDR_EMPTY(hdr)) buf_discard_identity(hdr); if (HDR_HAS_L2HDR(hdr)) { l2arc_dev_t *dev = hdr->b_l2hdr.b_dev; boolean_t buflist_held = MUTEX_HELD(&dev->l2ad_mtx); if (!buflist_held) mutex_enter(&dev->l2ad_mtx); /* * Even though we checked this conditional above, we * need to check this again now that we have the * l2ad_mtx. This is because we could be racing with * another thread calling l2arc_evict() which might have * destroyed this header's L2 portion as we were waiting * to acquire the l2ad_mtx. If that happens, we don't * want to re-destroy the header's L2 portion. */ if (HDR_HAS_L2HDR(hdr)) arc_hdr_l2hdr_destroy(hdr); if (!buflist_held) mutex_exit(&dev->l2ad_mtx); } if (HDR_HAS_L1HDR(hdr)) { arc_cksum_free(hdr); while (hdr->b_l1hdr.b_buf != NULL) arc_buf_destroy_impl(hdr->b_l1hdr.b_buf); if (hdr->b_l1hdr.b_pabd != NULL) { arc_hdr_free_abd(hdr, B_FALSE); } if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); } ASSERT3P(hdr->b_hash_next, ==, NULL); if (HDR_HAS_L1HDR(hdr)) { ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); if (!HDR_PROTECTED(hdr)) { kmem_cache_free(hdr_full_cache, hdr); } else { kmem_cache_free(hdr_full_crypt_cache, hdr); } } else { kmem_cache_free(hdr_l2only_cache, hdr); } } void arc_buf_destroy(arc_buf_t *buf, void* tag) { arc_buf_hdr_t *hdr = buf->b_hdr; kmutex_t *hash_lock = HDR_LOCK(hdr); if (hdr->b_l1hdr.b_state == arc_anon) { ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); VERIFY0(remove_reference(hdr, NULL, tag)); arc_hdr_destroy(hdr); return; } mutex_enter(hash_lock); ASSERT3P(hdr, ==, buf->b_hdr); ASSERT(hdr->b_l1hdr.b_bufcnt > 0); ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); ASSERT3P(hdr->b_l1hdr.b_state, !=, arc_anon); ASSERT3P(buf->b_data, !=, NULL); (void) remove_reference(hdr, hash_lock, tag); arc_buf_destroy_impl(buf); mutex_exit(hash_lock); } /* * Evict the arc_buf_hdr that is provided as a parameter. The resultant * state of the header is dependent on its state prior to entering this * function. The following transitions are possible: * * - arc_mru -> arc_mru_ghost * - arc_mfu -> arc_mfu_ghost * - arc_mru_ghost -> arc_l2c_only * - arc_mru_ghost -> deleted * - arc_mfu_ghost -> arc_l2c_only * - arc_mfu_ghost -> deleted */ static int64_t arc_evict_hdr(arc_buf_hdr_t *hdr, kmutex_t *hash_lock) { arc_state_t *evicted_state, *state; int64_t bytes_evicted = 0; int min_lifetime = HDR_PRESCIENT_PREFETCH(hdr) ? arc_min_prescient_prefetch_ms : arc_min_prefetch_ms; ASSERT(MUTEX_HELD(hash_lock)); ASSERT(HDR_HAS_L1HDR(hdr)); state = hdr->b_l1hdr.b_state; if (GHOST_STATE(state)) { ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); /* * l2arc_write_buffers() relies on a header's L1 portion * (i.e. its b_pabd field) during it's write phase. * Thus, we cannot push a header onto the arc_l2c_only * state (removing its L1 piece) until the header is * done being written to the l2arc. */ if (HDR_HAS_L2HDR(hdr) && HDR_L2_WRITING(hdr)) { ARCSTAT_BUMP(arcstat_evict_l2_skip); return (bytes_evicted); } ARCSTAT_BUMP(arcstat_deleted); bytes_evicted += HDR_GET_LSIZE(hdr); DTRACE_PROBE1(arc__delete, arc_buf_hdr_t *, hdr); if (HDR_HAS_L2HDR(hdr)) { ASSERT(hdr->b_l1hdr.b_pabd == NULL); ASSERT(!HDR_HAS_RABD(hdr)); /* * This buffer is cached on the 2nd Level ARC; * don't destroy the header. */ arc_change_state(arc_l2c_only, hdr, hash_lock); /* * dropping from L1+L2 cached to L2-only, * realloc to remove the L1 header. */ hdr = arc_hdr_realloc(hdr, hdr_full_cache, hdr_l2only_cache); } else { arc_change_state(arc_anon, hdr, hash_lock); arc_hdr_destroy(hdr); } return (bytes_evicted); } ASSERT(state == arc_mru || state == arc_mfu); evicted_state = (state == arc_mru) ? arc_mru_ghost : arc_mfu_ghost; /* prefetch buffers have a minimum lifespan */ if (HDR_IO_IN_PROGRESS(hdr) || ((hdr->b_flags & (ARC_FLAG_PREFETCH | ARC_FLAG_INDIRECT)) && ddi_get_lbolt() - hdr->b_l1hdr.b_arc_access < MSEC_TO_TICK(min_lifetime))) { ARCSTAT_BUMP(arcstat_evict_skip); return (bytes_evicted); } ASSERT0(refcount_count(&hdr->b_l1hdr.b_refcnt)); while (hdr->b_l1hdr.b_buf) { arc_buf_t *buf = hdr->b_l1hdr.b_buf; if (!mutex_tryenter(&buf->b_evict_lock)) { ARCSTAT_BUMP(arcstat_mutex_miss); break; } if (buf->b_data != NULL) bytes_evicted += HDR_GET_LSIZE(hdr); mutex_exit(&buf->b_evict_lock); arc_buf_destroy_impl(buf); } if (HDR_HAS_L2HDR(hdr)) { ARCSTAT_INCR(arcstat_evict_l2_cached, HDR_GET_LSIZE(hdr)); } else { if (l2arc_write_eligible(hdr->b_spa, hdr)) { ARCSTAT_INCR(arcstat_evict_l2_eligible, HDR_GET_LSIZE(hdr)); } else { ARCSTAT_INCR(arcstat_evict_l2_ineligible, HDR_GET_LSIZE(hdr)); } } if (hdr->b_l1hdr.b_bufcnt == 0) { arc_cksum_free(hdr); bytes_evicted += arc_hdr_size(hdr); /* * If this hdr is being evicted and has a compressed * buffer then we discard it here before we change states. * This ensures that the accounting is updated correctly * in arc_free_data_impl(). */ if (hdr->b_l1hdr.b_pabd != NULL) arc_hdr_free_abd(hdr, B_FALSE); if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); arc_change_state(evicted_state, hdr, hash_lock); ASSERT(HDR_IN_HASH_TABLE(hdr)); arc_hdr_set_flags(hdr, ARC_FLAG_IN_HASH_TABLE); DTRACE_PROBE1(arc__evict, arc_buf_hdr_t *, hdr); } return (bytes_evicted); } static uint64_t arc_evict_state_impl(multilist_t *ml, int idx, arc_buf_hdr_t *marker, uint64_t spa, int64_t bytes) { multilist_sublist_t *mls; uint64_t bytes_evicted = 0; arc_buf_hdr_t *hdr; kmutex_t *hash_lock; int evict_count = 0; ASSERT3P(marker, !=, NULL); IMPLY(bytes < 0, bytes == ARC_EVICT_ALL); mls = multilist_sublist_lock(ml, idx); for (hdr = multilist_sublist_prev(mls, marker); hdr != NULL; hdr = multilist_sublist_prev(mls, marker)) { if ((bytes != ARC_EVICT_ALL && bytes_evicted >= bytes) || (evict_count >= zfs_arc_evict_batch_limit)) break; /* * To keep our iteration location, move the marker * forward. Since we're not holding hdr's hash lock, we * must be very careful and not remove 'hdr' from the * sublist. Otherwise, other consumers might mistake the * 'hdr' as not being on a sublist when they call the * multilist_link_active() function (they all rely on * the hash lock protecting concurrent insertions and * removals). multilist_sublist_move_forward() was * specifically implemented to ensure this is the case * (only 'marker' will be removed and re-inserted). */ multilist_sublist_move_forward(mls, marker); /* * The only case where the b_spa field should ever be * zero, is the marker headers inserted by * arc_evict_state(). It's possible for multiple threads * to be calling arc_evict_state() concurrently (e.g. * dsl_pool_close() and zio_inject_fault()), so we must * skip any markers we see from these other threads. */ if (hdr->b_spa == 0) continue; /* we're only interested in evicting buffers of a certain spa */ if (spa != 0 && hdr->b_spa != spa) { ARCSTAT_BUMP(arcstat_evict_skip); continue; } hash_lock = HDR_LOCK(hdr); /* * We aren't calling this function from any code path * that would already be holding a hash lock, so we're * asserting on this assumption to be defensive in case * this ever changes. Without this check, it would be * possible to incorrectly increment arcstat_mutex_miss * below (e.g. if the code changed such that we called * this function with a hash lock held). */ ASSERT(!MUTEX_HELD(hash_lock)); if (mutex_tryenter(hash_lock)) { uint64_t evicted = arc_evict_hdr(hdr, hash_lock); mutex_exit(hash_lock); bytes_evicted += evicted; /* * If evicted is zero, arc_evict_hdr() must have * decided to skip this header, don't increment * evict_count in this case. */ if (evicted != 0) evict_count++; /* * If arc_size isn't overflowing, signal any * threads that might happen to be waiting. * * For each header evicted, we wake up a single * thread. If we used cv_broadcast, we could * wake up "too many" threads causing arc_size * to significantly overflow arc_c; since * arc_get_data_impl() doesn't check for overflow * when it's woken up (it doesn't because it's * possible for the ARC to be overflowing while * full of un-evictable buffers, and the * function should proceed in this case). * * If threads are left sleeping, due to not * using cv_broadcast, they will be woken up * just before arc_reclaim_thread() sleeps. */ mutex_enter(&arc_reclaim_lock); if (!arc_is_overflowing()) cv_signal(&arc_reclaim_waiters_cv); mutex_exit(&arc_reclaim_lock); } else { ARCSTAT_BUMP(arcstat_mutex_miss); } } multilist_sublist_unlock(mls); return (bytes_evicted); } /* * Evict buffers from the given arc state, until we've removed the * specified number of bytes. Move the removed buffers to the * appropriate evict state. * * This function makes a "best effort". It skips over any buffers * it can't get a hash_lock on, and so, may not catch all candidates. * It may also return without evicting as much space as requested. * * If bytes is specified using the special value ARC_EVICT_ALL, this * will evict all available (i.e. unlocked and evictable) buffers from * the given arc state; which is used by arc_flush(). */ static uint64_t arc_evict_state(arc_state_t *state, uint64_t spa, int64_t bytes, arc_buf_contents_t type) { uint64_t total_evicted = 0; multilist_t *ml = state->arcs_list[type]; int num_sublists; arc_buf_hdr_t **markers; IMPLY(bytes < 0, bytes == ARC_EVICT_ALL); num_sublists = multilist_get_num_sublists(ml); /* * If we've tried to evict from each sublist, made some * progress, but still have not hit the target number of bytes * to evict, we want to keep trying. The markers allow us to * pick up where we left off for each individual sublist, rather * than starting from the tail each time. */ markers = kmem_zalloc(sizeof (*markers) * num_sublists, KM_SLEEP); for (int i = 0; i < num_sublists; i++) { multilist_sublist_t *mls; markers[i] = kmem_cache_alloc(hdr_full_cache, KM_SLEEP); /* * A b_spa of 0 is used to indicate that this header is * a marker. This fact is used in arc_adjust_type() and * arc_evict_state_impl(). */ markers[i]->b_spa = 0; mls = multilist_sublist_lock(ml, i); multilist_sublist_insert_tail(mls, markers[i]); multilist_sublist_unlock(mls); } /* * While we haven't hit our target number of bytes to evict, or * we're evicting all available buffers. */ while (total_evicted < bytes || bytes == ARC_EVICT_ALL) { int sublist_idx = multilist_get_random_index(ml); uint64_t scan_evicted = 0; /* * Try to reduce pinned dnodes with a floor of arc_dnode_limit. * Request that 10% of the LRUs be scanned by the superblock * shrinker. */ if (type == ARC_BUFC_DATA && aggsum_compare(&astat_dnode_size, arc_dnode_limit) > 0) { arc_prune_async((aggsum_upper_bound(&astat_dnode_size) - arc_dnode_limit) / sizeof (dnode_t) / zfs_arc_dnode_reduce_percent); } /* * Start eviction using a randomly selected sublist, * this is to try and evenly balance eviction across all * sublists. Always starting at the same sublist * (e.g. index 0) would cause evictions to favor certain * sublists over others. */ for (int i = 0; i < num_sublists; i++) { uint64_t bytes_remaining; uint64_t bytes_evicted; if (bytes == ARC_EVICT_ALL) bytes_remaining = ARC_EVICT_ALL; else if (total_evicted < bytes) bytes_remaining = bytes - total_evicted; else break; bytes_evicted = arc_evict_state_impl(ml, sublist_idx, markers[sublist_idx], spa, bytes_remaining); scan_evicted += bytes_evicted; total_evicted += bytes_evicted; /* we've reached the end, wrap to the beginning */ if (++sublist_idx >= num_sublists) sublist_idx = 0; } /* * If we didn't evict anything during this scan, we have * no reason to believe we'll evict more during another * scan, so break the loop. */ if (scan_evicted == 0) { /* This isn't possible, let's make that obvious */ ASSERT3S(bytes, !=, 0); /* * When bytes is ARC_EVICT_ALL, the only way to * break the loop is when scan_evicted is zero. * In that case, we actually have evicted enough, * so we don't want to increment the kstat. */ if (bytes != ARC_EVICT_ALL) { ASSERT3S(total_evicted, <, bytes); ARCSTAT_BUMP(arcstat_evict_not_enough); } break; } } for (int i = 0; i < num_sublists; i++) { multilist_sublist_t *mls = multilist_sublist_lock(ml, i); multilist_sublist_remove(mls, markers[i]); multilist_sublist_unlock(mls); kmem_cache_free(hdr_full_cache, markers[i]); } kmem_free(markers, sizeof (*markers) * num_sublists); return (total_evicted); } /* * Flush all "evictable" data of the given type from the arc state * specified. This will not evict any "active" buffers (i.e. referenced). * * When 'retry' is set to B_FALSE, the function will make a single pass * over the state and evict any buffers that it can. Since it doesn't * continually retry the eviction, it might end up leaving some buffers * in the ARC due to lock misses. * * When 'retry' is set to B_TRUE, the function will continually retry the * eviction until *all* evictable buffers have been removed from the * state. As a result, if concurrent insertions into the state are * allowed (e.g. if the ARC isn't shutting down), this function might * wind up in an infinite loop, continually trying to evict buffers. */ static uint64_t arc_flush_state(arc_state_t *state, uint64_t spa, arc_buf_contents_t type, boolean_t retry) { uint64_t evicted = 0; while (refcount_count(&state->arcs_esize[type]) != 0) { evicted += arc_evict_state(state, spa, ARC_EVICT_ALL, type); if (!retry) break; } return (evicted); } /* * Helper function for arc_prune_async() it is responsible for safely * handling the execution of a registered arc_prune_func_t. */ static void arc_prune_task(void *ptr) { arc_prune_t *ap = (arc_prune_t *)ptr; arc_prune_func_t *func = ap->p_pfunc; if (func != NULL) func(ap->p_adjust, ap->p_private); refcount_remove(&ap->p_refcnt, func); } /* * Notify registered consumers they must drop holds on a portion of the ARC * buffered they reference. This provides a mechanism to ensure the ARC can * honor the arc_meta_limit and reclaim otherwise pinned ARC buffers. This * is analogous to dnlc_reduce_cache() but more generic. * * This operation is performed asynchronously so it may be safely called * in the context of the arc_reclaim_thread(). A reference is taken here * for each registered arc_prune_t and the arc_prune_task() is responsible * for releasing it once the registered arc_prune_func_t has completed. */ static void arc_prune_async(int64_t adjust) { arc_prune_t *ap; mutex_enter(&arc_prune_mtx); for (ap = list_head(&arc_prune_list); ap != NULL; ap = list_next(&arc_prune_list, ap)) { if (refcount_count(&ap->p_refcnt) >= 2) continue; refcount_add(&ap->p_refcnt, ap->p_pfunc); ap->p_adjust = adjust; if (taskq_dispatch(arc_prune_taskq, arc_prune_task, ap, TQ_SLEEP) == TASKQID_INVALID) { refcount_remove(&ap->p_refcnt, ap->p_pfunc); continue; } ARCSTAT_BUMP(arcstat_prune); } mutex_exit(&arc_prune_mtx); } /* * Evict the specified number of bytes from the state specified, * restricting eviction to the spa and type given. This function * prevents us from trying to evict more from a state's list than * is "evictable", and to skip evicting altogether when passed a * negative value for "bytes". In contrast, arc_evict_state() will * evict everything it can, when passed a negative value for "bytes". */ static uint64_t arc_adjust_impl(arc_state_t *state, uint64_t spa, int64_t bytes, arc_buf_contents_t type) { int64_t delta; if (bytes > 0 && refcount_count(&state->arcs_esize[type]) > 0) { delta = MIN(refcount_count(&state->arcs_esize[type]), bytes); return (arc_evict_state(state, spa, delta, type)); } return (0); } /* * The goal of this function is to evict enough meta data buffers from the * ARC in order to enforce the arc_meta_limit. Achieving this is slightly * more complicated than it appears because it is common for data buffers * to have holds on meta data buffers. In addition, dnode meta data buffers * will be held by the dnodes in the block preventing them from being freed. * This means we can't simply traverse the ARC and expect to always find * enough unheld meta data buffer to release. * * Therefore, this function has been updated to make alternating passes * over the ARC releasing data buffers and then newly unheld meta data * buffers. This ensures forward progress is maintained and meta_used * will decrease. Normally this is sufficient, but if required the ARC * will call the registered prune callbacks causing dentry and inodes to * be dropped from the VFS cache. This will make dnode meta data buffers * available for reclaim. */ static uint64_t arc_adjust_meta_balanced(uint64_t meta_used) { int64_t delta, prune = 0, adjustmnt; uint64_t total_evicted = 0; arc_buf_contents_t type = ARC_BUFC_DATA; int restarts = MAX(zfs_arc_meta_adjust_restarts, 0); restart: /* * This slightly differs than the way we evict from the mru in * arc_adjust because we don't have a "target" value (i.e. no * "meta" arc_p). As a result, I think we can completely * cannibalize the metadata in the MRU before we evict the * metadata from the MFU. I think we probably need to implement a * "metadata arc_p" value to do this properly. */ adjustmnt = meta_used - arc_meta_limit; if (adjustmnt > 0 && refcount_count(&arc_mru->arcs_esize[type]) > 0) { delta = MIN(refcount_count(&arc_mru->arcs_esize[type]), adjustmnt); total_evicted += arc_adjust_impl(arc_mru, 0, delta, type); adjustmnt -= delta; } /* * We can't afford to recalculate adjustmnt here. If we do, * new metadata buffers can sneak into the MRU or ANON lists, * thus penalize the MFU metadata. Although the fudge factor is * small, it has been empirically shown to be significant for * certain workloads (e.g. creating many empty directories). As * such, we use the original calculation for adjustmnt, and * simply decrement the amount of data evicted from the MRU. */ if (adjustmnt > 0 && refcount_count(&arc_mfu->arcs_esize[type]) > 0) { delta = MIN(refcount_count(&arc_mfu->arcs_esize[type]), adjustmnt); total_evicted += arc_adjust_impl(arc_mfu, 0, delta, type); } adjustmnt = meta_used - arc_meta_limit; if (adjustmnt > 0 && refcount_count(&arc_mru_ghost->arcs_esize[type]) > 0) { delta = MIN(adjustmnt, refcount_count(&arc_mru_ghost->arcs_esize[type])); total_evicted += arc_adjust_impl(arc_mru_ghost, 0, delta, type); adjustmnt -= delta; } if (adjustmnt > 0 && refcount_count(&arc_mfu_ghost->arcs_esize[type]) > 0) { delta = MIN(adjustmnt, refcount_count(&arc_mfu_ghost->arcs_esize[type])); total_evicted += arc_adjust_impl(arc_mfu_ghost, 0, delta, type); } /* * If after attempting to make the requested adjustment to the ARC * the meta limit is still being exceeded then request that the * higher layers drop some cached objects which have holds on ARC * meta buffers. Requests to the upper layers will be made with * increasingly large scan sizes until the ARC is below the limit. */ if (meta_used > arc_meta_limit) { if (type == ARC_BUFC_DATA) { type = ARC_BUFC_METADATA; } else { type = ARC_BUFC_DATA; if (zfs_arc_meta_prune) { prune += zfs_arc_meta_prune; arc_prune_async(prune); } } if (restarts > 0) { restarts--; goto restart; } } return (total_evicted); } /* * Evict metadata buffers from the cache, such that arc_meta_used is * capped by the arc_meta_limit tunable. */ static uint64_t arc_adjust_meta_only(uint64_t meta_used) { uint64_t total_evicted = 0; int64_t target; /* * If we're over the meta limit, we want to evict enough * metadata to get back under the meta limit. We don't want to * evict so much that we drop the MRU below arc_p, though. If * we're over the meta limit more than we're over arc_p, we * evict some from the MRU here, and some from the MFU below. */ target = MIN((int64_t)(meta_used - arc_meta_limit), (int64_t)(refcount_count(&arc_anon->arcs_size) + refcount_count(&arc_mru->arcs_size) - arc_p)); total_evicted += arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_METADATA); /* * Similar to the above, we want to evict enough bytes to get us * below the meta limit, but not so much as to drop us below the * space allotted to the MFU (which is defined as arc_c - arc_p). */ target = MIN((int64_t)(meta_used - arc_meta_limit), (int64_t)(refcount_count(&arc_mfu->arcs_size) - (arc_c - arc_p))); total_evicted += arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_METADATA); return (total_evicted); } static uint64_t arc_adjust_meta(uint64_t meta_used) { if (zfs_arc_meta_strategy == ARC_STRATEGY_META_ONLY) return (arc_adjust_meta_only(meta_used)); else return (arc_adjust_meta_balanced(meta_used)); } /* * Return the type of the oldest buffer in the given arc state * * This function will select a random sublist of type ARC_BUFC_DATA and * a random sublist of type ARC_BUFC_METADATA. The tail of each sublist * is compared, and the type which contains the "older" buffer will be * returned. */ static arc_buf_contents_t arc_adjust_type(arc_state_t *state) { multilist_t *data_ml = state->arcs_list[ARC_BUFC_DATA]; multilist_t *meta_ml = state->arcs_list[ARC_BUFC_METADATA]; int data_idx = multilist_get_random_index(data_ml); int meta_idx = multilist_get_random_index(meta_ml); multilist_sublist_t *data_mls; multilist_sublist_t *meta_mls; arc_buf_contents_t type; arc_buf_hdr_t *data_hdr; arc_buf_hdr_t *meta_hdr; /* * We keep the sublist lock until we're finished, to prevent * the headers from being destroyed via arc_evict_state(). */ data_mls = multilist_sublist_lock(data_ml, data_idx); meta_mls = multilist_sublist_lock(meta_ml, meta_idx); /* * These two loops are to ensure we skip any markers that * might be at the tail of the lists due to arc_evict_state(). */ for (data_hdr = multilist_sublist_tail(data_mls); data_hdr != NULL; data_hdr = multilist_sublist_prev(data_mls, data_hdr)) { if (data_hdr->b_spa != 0) break; } for (meta_hdr = multilist_sublist_tail(meta_mls); meta_hdr != NULL; meta_hdr = multilist_sublist_prev(meta_mls, meta_hdr)) { if (meta_hdr->b_spa != 0) break; } if (data_hdr == NULL && meta_hdr == NULL) { type = ARC_BUFC_DATA; } else if (data_hdr == NULL) { ASSERT3P(meta_hdr, !=, NULL); type = ARC_BUFC_METADATA; } else if (meta_hdr == NULL) { ASSERT3P(data_hdr, !=, NULL); type = ARC_BUFC_DATA; } else { ASSERT3P(data_hdr, !=, NULL); ASSERT3P(meta_hdr, !=, NULL); /* The headers can't be on the sublist without an L1 header */ ASSERT(HDR_HAS_L1HDR(data_hdr)); ASSERT(HDR_HAS_L1HDR(meta_hdr)); if (data_hdr->b_l1hdr.b_arc_access < meta_hdr->b_l1hdr.b_arc_access) { type = ARC_BUFC_DATA; } else { type = ARC_BUFC_METADATA; } } multilist_sublist_unlock(meta_mls); multilist_sublist_unlock(data_mls); return (type); } /* * Evict buffers from the cache, such that arc_size is capped by arc_c. */ static uint64_t arc_adjust(void) { uint64_t total_evicted = 0; uint64_t bytes; int64_t target; uint64_t asize = aggsum_value(&arc_size); uint64_t ameta = aggsum_value(&arc_meta_used); /* * If we're over arc_meta_limit, we want to correct that before * potentially evicting data buffers below. */ total_evicted += arc_adjust_meta(ameta); /* * Adjust MRU size * * If we're over the target cache size, we want to evict enough * from the list to get back to our target size. We don't want * to evict too much from the MRU, such that it drops below * arc_p. So, if we're over our target cache size more than * the MRU is over arc_p, we'll evict enough to get back to * arc_p here, and then evict more from the MFU below. */ target = MIN((int64_t)(asize - arc_c), (int64_t)(refcount_count(&arc_anon->arcs_size) + refcount_count(&arc_mru->arcs_size) + ameta - arc_p)); /* * If we're below arc_meta_min, always prefer to evict data. * Otherwise, try to satisfy the requested number of bytes to * evict from the type which contains older buffers; in an * effort to keep newer buffers in the cache regardless of their * type. If we cannot satisfy the number of bytes from this * type, spill over into the next type. */ if (arc_adjust_type(arc_mru) == ARC_BUFC_METADATA && ameta > arc_meta_min) { bytes = arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_METADATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * metadata, we try to get the rest from data. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_DATA); } else { bytes = arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_DATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * data, we try to get the rest from metadata. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mru, 0, target, ARC_BUFC_METADATA); } /* * Adjust MFU size * * Now that we've tried to evict enough from the MRU to get its * size back to arc_p, if we're still above the target cache * size, we evict the rest from the MFU. */ target = asize - arc_c; if (arc_adjust_type(arc_mfu) == ARC_BUFC_METADATA && ameta > arc_meta_min) { bytes = arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_METADATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * metadata, we try to get the rest from data. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_DATA); } else { bytes = arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_DATA); total_evicted += bytes; /* * If we couldn't evict our target number of bytes from * data, we try to get the rest from data. */ target -= bytes; total_evicted += arc_adjust_impl(arc_mfu, 0, target, ARC_BUFC_METADATA); } /* * Adjust ghost lists * * In addition to the above, the ARC also defines target values * for the ghost lists. The sum of the mru list and mru ghost * list should never exceed the target size of the cache, and * the sum of the mru list, mfu list, mru ghost list, and mfu * ghost list should never exceed twice the target size of the * cache. The following logic enforces these limits on the ghost * caches, and evicts from them as needed. */ target = refcount_count(&arc_mru->arcs_size) + refcount_count(&arc_mru_ghost->arcs_size) - arc_c; bytes = arc_adjust_impl(arc_mru_ghost, 0, target, ARC_BUFC_DATA); total_evicted += bytes; target -= bytes; total_evicted += arc_adjust_impl(arc_mru_ghost, 0, target, ARC_BUFC_METADATA); /* * We assume the sum of the mru list and mfu list is less than * or equal to arc_c (we enforced this above), which means we * can use the simpler of the two equations below: * * mru + mfu + mru ghost + mfu ghost <= 2 * arc_c * mru ghost + mfu ghost <= arc_c */ target = refcount_count(&arc_mru_ghost->arcs_size) + refcount_count(&arc_mfu_ghost->arcs_size) - arc_c; bytes = arc_adjust_impl(arc_mfu_ghost, 0, target, ARC_BUFC_DATA); total_evicted += bytes; target -= bytes; total_evicted += arc_adjust_impl(arc_mfu_ghost, 0, target, ARC_BUFC_METADATA); return (total_evicted); } void arc_flush(spa_t *spa, boolean_t retry) { uint64_t guid = 0; /* * If retry is B_TRUE, a spa must not be specified since we have * no good way to determine if all of a spa's buffers have been * evicted from an arc state. */ ASSERT(!retry || spa == 0); if (spa != NULL) guid = spa_load_guid(spa); (void) arc_flush_state(arc_mru, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mru, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mfu, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mfu, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mru_ghost, guid, ARC_BUFC_METADATA, retry); (void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_DATA, retry); (void) arc_flush_state(arc_mfu_ghost, guid, ARC_BUFC_METADATA, retry); } void arc_shrink(int64_t to_free) { uint64_t asize = aggsum_value(&arc_size); uint64_t c = arc_c; if (c > to_free && c - to_free > arc_c_min) { arc_c = c - to_free; atomic_add_64(&arc_p, -(arc_p >> arc_shrink_shift)); if (asize < arc_c) arc_c = MAX(asize, arc_c_min); if (arc_p > arc_c) arc_p = (arc_c >> 1); ASSERT(arc_c >= arc_c_min); ASSERT((int64_t)arc_p >= 0); } else { arc_c = arc_c_min; } if (asize > arc_c) (void) arc_adjust(); } /* * Return maximum amount of memory that we could possibly use. Reduced * to half of all memory in user space which is primarily used for testing. */ static uint64_t arc_all_memory(void) { #ifdef _KERNEL #ifdef CONFIG_HIGHMEM return (ptob(totalram_pages - totalhigh_pages)); #else return (ptob(totalram_pages)); #endif /* CONFIG_HIGHMEM */ #else return (ptob(physmem) / 2); #endif /* _KERNEL */ } /* * Return the amount of memory that is considered free. In user space * which is primarily used for testing we pretend that free memory ranges * from 0-20% of all memory. */ static uint64_t arc_free_memory(void) { #ifdef _KERNEL #ifdef CONFIG_HIGHMEM struct sysinfo si; si_meminfo(&si); return (ptob(si.freeram - si.freehigh)); #else return (ptob(nr_free_pages() + nr_inactive_file_pages() + nr_inactive_anon_pages() + nr_slab_reclaimable_pages())); #endif /* CONFIG_HIGHMEM */ #else return (spa_get_random(arc_all_memory() * 20 / 100)); #endif /* _KERNEL */ } typedef enum free_memory_reason_t { FMR_UNKNOWN, FMR_NEEDFREE, FMR_LOTSFREE, FMR_SWAPFS_MINFREE, FMR_PAGES_PP_MAXIMUM, FMR_HEAP_ARENA, FMR_ZIO_ARENA, } free_memory_reason_t; int64_t last_free_memory; free_memory_reason_t last_free_reason; #ifdef _KERNEL /* * Additional reserve of pages for pp_reserve. */ int64_t arc_pages_pp_reserve = 64; /* * Additional reserve of pages for swapfs. */ int64_t arc_swapfs_reserve = 64; #endif /* _KERNEL */ /* * Return the amount of memory that can be consumed before reclaim will be * needed. Positive if there is sufficient free memory, negative indicates * the amount of memory that needs to be freed up. */ static int64_t arc_available_memory(void) { int64_t lowest = INT64_MAX; free_memory_reason_t r = FMR_UNKNOWN; #ifdef _KERNEL int64_t n; #ifdef __linux__ #ifdef freemem #undef freemem #endif pgcnt_t needfree = btop(arc_need_free); pgcnt_t lotsfree = btop(arc_sys_free); pgcnt_t desfree = 0; pgcnt_t freemem = btop(arc_free_memory()); #endif if (needfree > 0) { n = PAGESIZE * (-needfree); if (n < lowest) { lowest = n; r = FMR_NEEDFREE; } } /* * check that we're out of range of the pageout scanner. It starts to * schedule paging if freemem is less than lotsfree and needfree. * lotsfree is the high-water mark for pageout, and needfree is the * number of needed free pages. We add extra pages here to make sure * the scanner doesn't start up while we're freeing memory. */ n = PAGESIZE * (freemem - lotsfree - needfree - desfree); if (n < lowest) { lowest = n; r = FMR_LOTSFREE; } #ifndef __linux__ /* * check to make sure that swapfs has enough space so that anon * reservations can still succeed. anon_resvmem() checks that the * availrmem is greater than swapfs_minfree, and the number of reserved * swap pages. We also add a bit of extra here just to prevent * circumstances from getting really dire. */ n = PAGESIZE * (availrmem - swapfs_minfree - swapfs_reserve - desfree - arc_swapfs_reserve); if (n < lowest) { lowest = n; r = FMR_SWAPFS_MINFREE; } /* * Check that we have enough availrmem that memory locking (e.g., via * mlock(3C) or memcntl(2)) can still succeed. (pages_pp_maximum * stores the number of pages that cannot be locked; when availrmem * drops below pages_pp_maximum, page locking mechanisms such as * page_pp_lock() will fail.) */ n = PAGESIZE * (availrmem - pages_pp_maximum - arc_pages_pp_reserve); if (n < lowest) { lowest = n; r = FMR_PAGES_PP_MAXIMUM; } #endif #if defined(_ILP32) /* * If we're on a 32-bit platform, it's possible that we'll exhaust the * kernel heap space before we ever run out of available physical * memory. Most checks of the size of the heap_area compare against * tune.t_minarmem, which is the minimum available real memory that we * can have in the system. However, this is generally fixed at 25 pages * which is so low that it's useless. In this comparison, we seek to * calculate the total heap-size, and reclaim if more than 3/4ths of the * heap is allocated. (Or, in the calculation, if less than 1/4th is * free) */ n = vmem_size(heap_arena, VMEM_FREE) - (vmem_size(heap_arena, VMEM_FREE | VMEM_ALLOC) >> 2); if (n < lowest) { lowest = n; r = FMR_HEAP_ARENA; } #endif /* * If zio data pages are being allocated out of a separate heap segment, * then enforce that the size of available vmem for this arena remains * above about 1/4th (1/(2^arc_zio_arena_free_shift)) free. * * Note that reducing the arc_zio_arena_free_shift keeps more virtual * memory (in the zio_arena) free, which can avoid memory * fragmentation issues. */ if (zio_arena != NULL) { n = (int64_t)vmem_size(zio_arena, VMEM_FREE) - (vmem_size(zio_arena, VMEM_ALLOC) >> arc_zio_arena_free_shift); if (n < lowest) { lowest = n; r = FMR_ZIO_ARENA; } } #else /* _KERNEL */ /* Every 100 calls, free a small amount */ if (spa_get_random(100) == 0) lowest = -1024; #endif /* _KERNEL */ last_free_memory = lowest; last_free_reason = r; return (lowest); } /* * Determine if the system is under memory pressure and is asking * to reclaim memory. A return value of B_TRUE indicates that the system * is under memory pressure and that the arc should adjust accordingly. */ static boolean_t arc_reclaim_needed(void) { return (arc_available_memory() < 0); } static void arc_kmem_reap_now(void) { size_t i; kmem_cache_t *prev_cache = NULL; kmem_cache_t *prev_data_cache = NULL; extern kmem_cache_t *zio_buf_cache[]; extern kmem_cache_t *zio_data_buf_cache[]; extern kmem_cache_t *range_seg_cache; #ifdef _KERNEL if ((aggsum_compare(&arc_meta_used, arc_meta_limit) >= 0) && zfs_arc_meta_prune) { /* * We are exceeding our meta-data cache limit. * Prune some entries to release holds on meta-data. */ arc_prune_async(zfs_arc_meta_prune); } #if defined(_ILP32) /* * Reclaim unused memory from all kmem caches. */ kmem_reap(); #endif #endif for (i = 0; i < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; i++) { #if defined(_ILP32) /* reach upper limit of cache size on 32-bit */ if (zio_buf_cache[i] == NULL) break; #endif if (zio_buf_cache[i] != prev_cache) { prev_cache = zio_buf_cache[i]; kmem_cache_reap_now(zio_buf_cache[i]); } if (zio_data_buf_cache[i] != prev_data_cache) { prev_data_cache = zio_data_buf_cache[i]; kmem_cache_reap_now(zio_data_buf_cache[i]); } } kmem_cache_reap_now(buf_cache); kmem_cache_reap_now(hdr_full_cache); kmem_cache_reap_now(hdr_l2only_cache); kmem_cache_reap_now(range_seg_cache); if (zio_arena != NULL) { /* * Ask the vmem arena to reclaim unused memory from its * quantum caches. */ vmem_qcache_reap(zio_arena); } } /* * Threads can block in arc_get_data_impl() waiting for this thread to evict * enough data and signal them to proceed. When this happens, the threads in * arc_get_data_impl() are sleeping while holding the hash lock for their * particular arc header. Thus, we must be careful to never sleep on a * hash lock in this thread. This is to prevent the following deadlock: * * - Thread A sleeps on CV in arc_get_data_impl() holding hash lock "L", * waiting for the reclaim thread to signal it. * * - arc_reclaim_thread() tries to acquire hash lock "L" using mutex_enter, * fails, and goes to sleep forever. * * This possible deadlock is avoided by always acquiring a hash lock * using mutex_tryenter() from arc_reclaim_thread(). */ /* ARGSUSED */ static void arc_reclaim_thread(void *unused) { fstrans_cookie_t cookie = spl_fstrans_mark(); hrtime_t growtime = 0; callb_cpr_t cpr; CALLB_CPR_INIT(&cpr, &arc_reclaim_lock, callb_generic_cpr, FTAG); mutex_enter(&arc_reclaim_lock); while (!arc_reclaim_thread_exit) { uint64_t evicted = 0; uint64_t need_free = arc_need_free; arc_tuning_update(); /* * This is necessary in order for the mdb ::arc dcmd to * show up to date information. Since the ::arc command * does not call the kstat's update function, without * this call, the command may show stale stats for the * anon, mru, mru_ghost, mfu, and mfu_ghost lists. Even * with this change, the data might be up to 1 second * out of date; but that should suffice. The arc_state_t * structures can be queried directly if more accurate * information is needed. */ #ifndef __linux__ if (arc_ksp != NULL) arc_ksp->ks_update(arc_ksp, KSTAT_READ); #endif mutex_exit(&arc_reclaim_lock); /* * We call arc_adjust() before (possibly) calling * arc_kmem_reap_now(), so that we can wake up * arc_get_data_buf() sooner. */ evicted = arc_adjust(); int64_t free_memory = arc_available_memory(); if (free_memory < 0) { arc_no_grow = B_TRUE; arc_warm = B_TRUE; /* * Wait at least zfs_grow_retry (default 5) seconds * before considering growing. */ growtime = gethrtime() + SEC2NSEC(arc_grow_retry); arc_kmem_reap_now(); /* * If we are still low on memory, shrink the ARC * so that we have arc_shrink_min free space. */ free_memory = arc_available_memory(); int64_t to_free = (arc_c >> arc_shrink_shift) - free_memory; if (to_free > 0) { #ifdef _KERNEL to_free = MAX(to_free, need_free); #endif arc_shrink(to_free); } } else if (free_memory < arc_c >> arc_no_grow_shift) { arc_no_grow = B_TRUE; } else if (gethrtime() >= growtime) { arc_no_grow = B_FALSE; } mutex_enter(&arc_reclaim_lock); /* * If evicted is zero, we couldn't evict anything via * arc_adjust(). This could be due to hash lock * collisions, but more likely due to the majority of * arc buffers being unevictable. Therefore, even if * arc_size is above arc_c, another pass is unlikely to * be helpful and could potentially cause us to enter an * infinite loop. */ if (aggsum_compare(&arc_size, arc_c) <= 0|| evicted == 0) { /* * We're either no longer overflowing, or we * can't evict anything more, so we should wake * up any threads before we go to sleep and remove * the bytes we were working on from arc_need_free * since nothing more will be done here. */ cv_broadcast(&arc_reclaim_waiters_cv); ARCSTAT_INCR(arcstat_need_free, -need_free); /* * Block until signaled, or after one second (we * might need to perform arc_kmem_reap_now() * even if we aren't being signalled) */ CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait_sig_hires(&arc_reclaim_thread_cv, &arc_reclaim_lock, SEC2NSEC(1), MSEC2NSEC(1), 0); CALLB_CPR_SAFE_END(&cpr, &arc_reclaim_lock); } } arc_reclaim_thread_exit = B_FALSE; cv_broadcast(&arc_reclaim_thread_cv); CALLB_CPR_EXIT(&cpr); /* drops arc_reclaim_lock */ spl_fstrans_unmark(cookie); thread_exit(); } #ifdef _KERNEL /* * Determine the amount of memory eligible for eviction contained in the * ARC. All clean data reported by the ghost lists can always be safely * evicted. Due to arc_c_min, the same does not hold for all clean data * contained by the regular mru and mfu lists. * * In the case of the regular mru and mfu lists, we need to report as * much clean data as possible, such that evicting that same reported * data will not bring arc_size below arc_c_min. Thus, in certain * circumstances, the total amount of clean data in the mru and mfu * lists might not actually be evictable. * * The following two distinct cases are accounted for: * * 1. The sum of the amount of dirty data contained by both the mru and * mfu lists, plus the ARC's other accounting (e.g. the anon list), * is greater than or equal to arc_c_min. * (i.e. amount of dirty data >= arc_c_min) * * This is the easy case; all clean data contained by the mru and mfu * lists is evictable. Evicting all clean data can only drop arc_size * to the amount of dirty data, which is greater than arc_c_min. * * 2. The sum of the amount of dirty data contained by both the mru and * mfu lists, plus the ARC's other accounting (e.g. the anon list), * is less than arc_c_min. * (i.e. arc_c_min > amount of dirty data) * * 2.1. arc_size is greater than or equal arc_c_min. * (i.e. arc_size >= arc_c_min > amount of dirty data) * * In this case, not all clean data from the regular mru and mfu * lists is actually evictable; we must leave enough clean data * to keep arc_size above arc_c_min. Thus, the maximum amount of * evictable data from the two lists combined, is exactly the * difference between arc_size and arc_c_min. * * 2.2. arc_size is less than arc_c_min * (i.e. arc_c_min > arc_size > amount of dirty data) * * In this case, none of the data contained in the mru and mfu * lists is evictable, even if it's clean. Since arc_size is * already below arc_c_min, evicting any more would only * increase this negative difference. */ static uint64_t arc_evictable_memory(void) { int64_t asize = aggsum_value(&arc_size); uint64_t arc_clean = refcount_count(&arc_mru->arcs_esize[ARC_BUFC_DATA]) + refcount_count(&arc_mru->arcs_esize[ARC_BUFC_METADATA]) + refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_DATA]) + refcount_count(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]); uint64_t arc_dirty = MAX((int64_t)asize - (int64_t)arc_clean, 0); /* * Scale reported evictable memory in proportion to page cache, cap * at specified min/max. */ uint64_t min = (ptob(nr_file_pages()) / 100) * zfs_arc_pc_percent; min = MAX(arc_c_min, MIN(arc_c_max, min)); if (arc_dirty >= min) return (arc_clean); return (MAX((int64_t)asize - (int64_t)min, 0)); } /* * If sc->nr_to_scan is zero, the caller is requesting a query of the * number of objects which can potentially be freed. If it is nonzero, * the request is to free that many objects. * * Linux kernels >= 3.12 have the count_objects and scan_objects callbacks * in struct shrinker and also require the shrinker to return the number * of objects freed. * * Older kernels require the shrinker to return the number of freeable * objects following the freeing of nr_to_free. */ static spl_shrinker_t __arc_shrinker_func(struct shrinker *shrink, struct shrink_control *sc) { int64_t pages; /* The arc is considered warm once reclaim has occurred */ if (unlikely(arc_warm == B_FALSE)) arc_warm = B_TRUE; /* Return the potential number of reclaimable pages */ pages = btop((int64_t)arc_evictable_memory()); if (sc->nr_to_scan == 0) return (pages); /* Not allowed to perform filesystem reclaim */ if (!(sc->gfp_mask & __GFP_FS)) return (SHRINK_STOP); /* Reclaim in progress */ if (mutex_tryenter(&arc_reclaim_lock) == 0) { ARCSTAT_INCR(arcstat_need_free, ptob(sc->nr_to_scan)); return (0); } mutex_exit(&arc_reclaim_lock); /* * Evict the requested number of pages by shrinking arc_c the * requested amount. */ if (pages > 0) { arc_shrink(ptob(sc->nr_to_scan)); if (current_is_kswapd()) arc_kmem_reap_now(); #ifdef HAVE_SPLIT_SHRINKER_CALLBACK pages = MAX((int64_t)pages - (int64_t)btop(arc_evictable_memory()), 0); #else pages = btop(arc_evictable_memory()); #endif /* * We've shrunk what we can, wake up threads. */ cv_broadcast(&arc_reclaim_waiters_cv); } else pages = SHRINK_STOP; /* * When direct reclaim is observed it usually indicates a rapid * increase in memory pressure. This occurs because the kswapd * threads were unable to asynchronously keep enough free memory * available. In this case set arc_no_grow to briefly pause arc * growth to avoid compounding the memory pressure. */ if (current_is_kswapd()) { ARCSTAT_BUMP(arcstat_memory_indirect_count); } else { arc_no_grow = B_TRUE; arc_kmem_reap_now(); ARCSTAT_BUMP(arcstat_memory_direct_count); } return (pages); } SPL_SHRINKER_CALLBACK_WRAPPER(arc_shrinker_func); SPL_SHRINKER_DECLARE(arc_shrinker, arc_shrinker_func, DEFAULT_SEEKS); #endif /* _KERNEL */ /* * Adapt arc info given the number of bytes we are trying to add and * the state that we are coming from. This function is only called * when we are adding new content to the cache. */ static void arc_adapt(int bytes, arc_state_t *state) { int mult; uint64_t arc_p_min = (arc_c >> arc_p_min_shift); int64_t mrug_size = refcount_count(&arc_mru_ghost->arcs_size); int64_t mfug_size = refcount_count(&arc_mfu_ghost->arcs_size); if (state == arc_l2c_only) return; ASSERT(bytes > 0); /* * Adapt the target size of the MRU list: * - if we just hit in the MRU ghost list, then increase * the target size of the MRU list. * - if we just hit in the MFU ghost list, then increase * the target size of the MFU list by decreasing the * target size of the MRU list. */ if (state == arc_mru_ghost) { mult = (mrug_size >= mfug_size) ? 1 : (mfug_size / mrug_size); if (!zfs_arc_p_dampener_disable) mult = MIN(mult, 10); /* avoid wild arc_p adjustment */ arc_p = MIN(arc_c - arc_p_min, arc_p + bytes * mult); } else if (state == arc_mfu_ghost) { uint64_t delta; mult = (mfug_size >= mrug_size) ? 1 : (mrug_size / mfug_size); if (!zfs_arc_p_dampener_disable) mult = MIN(mult, 10); delta = MIN(bytes * mult, arc_p); arc_p = MAX(arc_p_min, arc_p - delta); } ASSERT((int64_t)arc_p >= 0); if (arc_reclaim_needed()) { cv_signal(&arc_reclaim_thread_cv); return; } if (arc_no_grow) return; if (arc_c >= arc_c_max) return; /* * If we're within (2 * maxblocksize) bytes of the target * cache size, increment the target cache size */ ASSERT3U(arc_c, >=, 2ULL << SPA_MAXBLOCKSHIFT); if (aggsum_compare(&arc_size, arc_c - (2ULL << SPA_MAXBLOCKSHIFT)) >= 0) { atomic_add_64(&arc_c, (int64_t)bytes); if (arc_c > arc_c_max) arc_c = arc_c_max; else if (state == arc_anon) atomic_add_64(&arc_p, (int64_t)bytes); if (arc_p > arc_c) arc_p = arc_c; } ASSERT((int64_t)arc_p >= 0); } /* * Check if arc_size has grown past our upper threshold, determined by * zfs_arc_overflow_shift. */ static boolean_t arc_is_overflowing(void) { /* Always allow at least one block of overflow */ uint64_t overflow = MAX(SPA_MAXBLOCKSIZE, arc_c >> zfs_arc_overflow_shift); /* * We just compare the lower bound here for performance reasons. Our * primary goals are to make sure that the arc never grows without * bound, and that it can reach its maximum size. This check * accomplishes both goals. The maximum amount we could run over by is * 2 * aggsum_borrow_multiplier * NUM_CPUS * the average size of a block * in the ARC. In practice, that's in the tens of MB, which is low * enough to be safe. */ return (aggsum_lower_bound(&arc_size) >= arc_c + overflow); } static abd_t * arc_get_data_abd(arc_buf_hdr_t *hdr, uint64_t size, void *tag) { arc_buf_contents_t type = arc_buf_type(hdr); arc_get_data_impl(hdr, size, tag); if (type == ARC_BUFC_METADATA) { return (abd_alloc(size, B_TRUE)); } else { ASSERT(type == ARC_BUFC_DATA); return (abd_alloc(size, B_FALSE)); } } static void * arc_get_data_buf(arc_buf_hdr_t *hdr, uint64_t size, void *tag) { arc_buf_contents_t type = arc_buf_type(hdr); arc_get_data_impl(hdr, size, tag); if (type == ARC_BUFC_METADATA) { return (zio_buf_alloc(size)); } else { ASSERT(type == ARC_BUFC_DATA); return (zio_data_buf_alloc(size)); } } /* * Allocate a block and return it to the caller. If we are hitting the * hard limit for the cache size, we must sleep, waiting for the eviction * thread to catch up. If we're past the target size but below the hard * limit, we'll only signal the reclaim thread and continue on. */ static void arc_get_data_impl(arc_buf_hdr_t *hdr, uint64_t size, void *tag) { arc_state_t *state = hdr->b_l1hdr.b_state; arc_buf_contents_t type = arc_buf_type(hdr); arc_adapt(size, state); /* * If arc_size is currently overflowing, and has grown past our * upper limit, we must be adding data faster than the evict * thread can evict. Thus, to ensure we don't compound the * problem by adding more data and forcing arc_size to grow even * further past it's target size, we halt and wait for the * eviction thread to catch up. * * It's also possible that the reclaim thread is unable to evict * enough buffers to get arc_size below the overflow limit (e.g. * due to buffers being un-evictable, or hash lock collisions). * In this case, we want to proceed regardless if we're * overflowing; thus we don't use a while loop here. */ if (arc_is_overflowing()) { mutex_enter(&arc_reclaim_lock); /* * Now that we've acquired the lock, we may no longer be * over the overflow limit, lets check. * * We're ignoring the case of spurious wake ups. If that * were to happen, it'd let this thread consume an ARC * buffer before it should have (i.e. before we're under * the overflow limit and were signalled by the reclaim * thread). As long as that is a rare occurrence, it * shouldn't cause any harm. */ if (arc_is_overflowing()) { cv_signal(&arc_reclaim_thread_cv); cv_wait(&arc_reclaim_waiters_cv, &arc_reclaim_lock); } mutex_exit(&arc_reclaim_lock); } VERIFY3U(hdr->b_type, ==, type); if (type == ARC_BUFC_METADATA) { arc_space_consume(size, ARC_SPACE_META); } else { arc_space_consume(size, ARC_SPACE_DATA); } /* * Update the state size. Note that ghost states have a * "ghost size" and so don't need to be updated. */ if (!GHOST_STATE(state)) { (void) refcount_add_many(&state->arcs_size, size, tag); /* * If this is reached via arc_read, the link is * protected by the hash lock. If reached via * arc_buf_alloc, the header should not be accessed by * any other thread. And, if reached via arc_read_done, * the hash lock will protect it if it's found in the * hash table; otherwise no other thread should be * trying to [add|remove]_reference it. */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); (void) refcount_add_many(&state->arcs_esize[type], size, tag); } /* * If we are growing the cache, and we are adding anonymous * data, and we have outgrown arc_p, update arc_p */ if (aggsum_compare(&arc_size, arc_c) < 0 && hdr->b_l1hdr.b_state == arc_anon && (refcount_count(&arc_anon->arcs_size) + refcount_count(&arc_mru->arcs_size) > arc_p)) arc_p = MIN(arc_c, arc_p + size); } } static void arc_free_data_abd(arc_buf_hdr_t *hdr, abd_t *abd, uint64_t size, void *tag) { arc_free_data_impl(hdr, size, tag); abd_free(abd); } static void arc_free_data_buf(arc_buf_hdr_t *hdr, void *buf, uint64_t size, void *tag) { arc_buf_contents_t type = arc_buf_type(hdr); arc_free_data_impl(hdr, size, tag); if (type == ARC_BUFC_METADATA) { zio_buf_free(buf, size); } else { ASSERT(type == ARC_BUFC_DATA); zio_data_buf_free(buf, size); } } /* * Free the arc data buffer. */ static void arc_free_data_impl(arc_buf_hdr_t *hdr, uint64_t size, void *tag) { arc_state_t *state = hdr->b_l1hdr.b_state; arc_buf_contents_t type = arc_buf_type(hdr); /* protected by hash lock, if in the hash table */ if (multilist_link_active(&hdr->b_l1hdr.b_arc_node)) { ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); ASSERT(state != arc_anon && state != arc_l2c_only); (void) refcount_remove_many(&state->arcs_esize[type], size, tag); } (void) refcount_remove_many(&state->arcs_size, size, tag); VERIFY3U(hdr->b_type, ==, type); if (type == ARC_BUFC_METADATA) { arc_space_return(size, ARC_SPACE_META); } else { ASSERT(type == ARC_BUFC_DATA); arc_space_return(size, ARC_SPACE_DATA); } } /* * This routine is called whenever a buffer is accessed. * NOTE: the hash lock is dropped in this function. */ static void arc_access(arc_buf_hdr_t *hdr, kmutex_t *hash_lock) { clock_t now; ASSERT(MUTEX_HELD(hash_lock)); ASSERT(HDR_HAS_L1HDR(hdr)); if (hdr->b_l1hdr.b_state == arc_anon) { /* * This buffer is not in the cache, and does not * appear in our "ghost" list. Add the new buffer * to the MRU state. */ ASSERT0(hdr->b_l1hdr.b_arc_access); hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr); arc_change_state(arc_mru, hdr, hash_lock); } else if (hdr->b_l1hdr.b_state == arc_mru) { now = ddi_get_lbolt(); /* * If this buffer is here because of a prefetch, then either: * - clear the flag if this is a "referencing" read * (any subsequent access will bump this into the MFU state). * or * - move the buffer to the head of the list if this is * another prefetch (to make it less likely to be evicted). */ if (HDR_PREFETCH(hdr) || HDR_PRESCIENT_PREFETCH(hdr)) { if (refcount_count(&hdr->b_l1hdr.b_refcnt) == 0) { /* link protected by hash lock */ ASSERT(multilist_link_active( &hdr->b_l1hdr.b_arc_node)); } else { arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH | ARC_FLAG_PRESCIENT_PREFETCH); atomic_inc_32(&hdr->b_l1hdr.b_mru_hits); ARCSTAT_BUMP(arcstat_mru_hits); } hdr->b_l1hdr.b_arc_access = now; return; } /* * This buffer has been "accessed" only once so far, * but it is still in the cache. Move it to the MFU * state. */ if (ddi_time_after(now, hdr->b_l1hdr.b_arc_access + ARC_MINTIME)) { /* * More than 125ms have passed since we * instantiated this buffer. Move it to the * most frequently used state. */ hdr->b_l1hdr.b_arc_access = now; DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(arc_mfu, hdr, hash_lock); } atomic_inc_32(&hdr->b_l1hdr.b_mru_hits); ARCSTAT_BUMP(arcstat_mru_hits); } else if (hdr->b_l1hdr.b_state == arc_mru_ghost) { arc_state_t *new_state; /* * This buffer has been "accessed" recently, but * was evicted from the cache. Move it to the * MFU state. */ if (HDR_PREFETCH(hdr) || HDR_PRESCIENT_PREFETCH(hdr)) { new_state = arc_mru; if (refcount_count(&hdr->b_l1hdr.b_refcnt) > 0) { arc_hdr_clear_flags(hdr, ARC_FLAG_PREFETCH | ARC_FLAG_PRESCIENT_PREFETCH); } DTRACE_PROBE1(new_state__mru, arc_buf_hdr_t *, hdr); } else { new_state = arc_mfu; DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); } hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); arc_change_state(new_state, hdr, hash_lock); atomic_inc_32(&hdr->b_l1hdr.b_mru_ghost_hits); ARCSTAT_BUMP(arcstat_mru_ghost_hits); } else if (hdr->b_l1hdr.b_state == arc_mfu) { /* * This buffer has been accessed more than once and is * still in the cache. Keep it in the MFU state. * * NOTE: an add_reference() that occurred when we did * the arc_read() will have kicked this off the list. * If it was a prefetch, we will explicitly move it to * the head of the list now. */ atomic_inc_32(&hdr->b_l1hdr.b_mfu_hits); ARCSTAT_BUMP(arcstat_mfu_hits); hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); } else if (hdr->b_l1hdr.b_state == arc_mfu_ghost) { arc_state_t *new_state = arc_mfu; /* * This buffer has been accessed more than once but has * been evicted from the cache. Move it back to the * MFU state. */ if (HDR_PREFETCH(hdr) || HDR_PRESCIENT_PREFETCH(hdr)) { /* * This is a prefetch access... * move this block back to the MRU state. */ new_state = arc_mru; } hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(new_state, hdr, hash_lock); atomic_inc_32(&hdr->b_l1hdr.b_mfu_ghost_hits); ARCSTAT_BUMP(arcstat_mfu_ghost_hits); } else if (hdr->b_l1hdr.b_state == arc_l2c_only) { /* * This buffer is on the 2nd Level ARC. */ hdr->b_l1hdr.b_arc_access = ddi_get_lbolt(); DTRACE_PROBE1(new_state__mfu, arc_buf_hdr_t *, hdr); arc_change_state(arc_mfu, hdr, hash_lock); } else { cmn_err(CE_PANIC, "invalid arc state 0x%p", hdr->b_l1hdr.b_state); } } /* * This routine is called by dbuf_hold() to update the arc_access() state * which otherwise would be skipped for entries in the dbuf cache. */ void arc_buf_access(arc_buf_t *buf) { mutex_enter(&buf->b_evict_lock); arc_buf_hdr_t *hdr = buf->b_hdr; /* * Avoid taking the hash_lock when possible as an optimization. * The header must be checked again under the hash_lock in order * to handle the case where it is concurrently being released. */ if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) { mutex_exit(&buf->b_evict_lock); return; } kmutex_t *hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); if (hdr->b_l1hdr.b_state == arc_anon || HDR_EMPTY(hdr)) { mutex_exit(hash_lock); mutex_exit(&buf->b_evict_lock); ARCSTAT_BUMP(arcstat_access_skip); return; } mutex_exit(&buf->b_evict_lock); ASSERT(hdr->b_l1hdr.b_state == arc_mru || hdr->b_l1hdr.b_state == arc_mfu); DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr); arc_access(hdr, hash_lock); mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_hits); ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr) && !HDR_PRESCIENT_PREFETCH(hdr), demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, hits); } /* a generic arc_read_done_func_t which you can use */ /* ARGSUSED */ void arc_bcopy_func(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, arc_buf_t *buf, void *arg) { if (buf == NULL) return; bcopy(buf->b_data, arg, arc_buf_size(buf)); arc_buf_destroy(buf, arg); } /* a generic arc_read_done_func_t */ /* ARGSUSED */ void arc_getbuf_func(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, arc_buf_t *buf, void *arg) { arc_buf_t **bufp = arg; if (buf == NULL) { + ASSERT(zio == NULL || zio->io_error != 0); *bufp = NULL; } else { + ASSERT(zio == NULL || zio->io_error == 0); *bufp = buf; - ASSERT(buf->b_data); + ASSERT(buf->b_data != NULL); } } static void arc_hdr_verify(arc_buf_hdr_t *hdr, blkptr_t *bp) { if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) { ASSERT3U(HDR_GET_PSIZE(hdr), ==, 0); ASSERT3U(arc_hdr_get_compress(hdr), ==, ZIO_COMPRESS_OFF); } else { if (HDR_COMPRESSION_ENABLED(hdr)) { ASSERT3U(arc_hdr_get_compress(hdr), ==, BP_GET_COMPRESS(bp)); } ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp)); ASSERT3U(HDR_GET_PSIZE(hdr), ==, BP_GET_PSIZE(bp)); ASSERT3U(!!HDR_PROTECTED(hdr), ==, BP_IS_PROTECTED(bp)); } } static void arc_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; arc_buf_hdr_t *hdr = zio->io_private; kmutex_t *hash_lock = NULL; arc_callback_t *callback_list; arc_callback_t *acb; boolean_t freeable = B_FALSE; /* * The hdr was inserted into hash-table and removed from lists * prior to starting I/O. We should find this header, since * it's in the hash table, and it should be legit since it's * not possible to evict it during the I/O. The only possible * reason for it not to be found is if we were freed during the * read. */ if (HDR_IN_HASH_TABLE(hdr)) { arc_buf_hdr_t *found; ASSERT3U(hdr->b_birth, ==, BP_PHYSICAL_BIRTH(zio->io_bp)); ASSERT3U(hdr->b_dva.dva_word[0], ==, BP_IDENTITY(zio->io_bp)->dva_word[0]); ASSERT3U(hdr->b_dva.dva_word[1], ==, BP_IDENTITY(zio->io_bp)->dva_word[1]); found = buf_hash_find(hdr->b_spa, zio->io_bp, &hash_lock); ASSERT((found == hdr && DVA_EQUAL(&hdr->b_dva, BP_IDENTITY(zio->io_bp))) || (found == hdr && HDR_L2_READING(hdr))); ASSERT3P(hash_lock, !=, NULL); } if (BP_IS_PROTECTED(bp)) { hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp); hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset; zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv); if (BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG) { void *tmpbuf; tmpbuf = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t)); zio_crypt_decode_mac_zil(tmpbuf, hdr->b_crypt_hdr.b_mac); abd_return_buf(zio->io_abd, tmpbuf, sizeof (zil_chain_t)); } else { zio_crypt_decode_mac_bp(bp, hdr->b_crypt_hdr.b_mac); } } if (zio->io_error == 0) { /* byteswap if necessary */ if (BP_SHOULD_BYTESWAP(zio->io_bp)) { if (BP_GET_LEVEL(zio->io_bp) > 0) { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64; } else { hdr->b_l1hdr.b_byteswap = DMU_OT_BYTESWAP(BP_GET_TYPE(zio->io_bp)); } } else { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; } } arc_hdr_clear_flags(hdr, ARC_FLAG_L2_EVICTED); if (l2arc_noprefetch && HDR_PREFETCH(hdr)) arc_hdr_clear_flags(hdr, ARC_FLAG_L2CACHE); callback_list = hdr->b_l1hdr.b_acb; ASSERT3P(callback_list, !=, NULL); if (hash_lock && zio->io_error == 0 && hdr->b_l1hdr.b_state == arc_anon) { /* * Only call arc_access on anonymous buffers. This is because * if we've issued an I/O for an evicted buffer, we've already * called arc_access (to prevent any simultaneous readers from * getting confused). */ arc_access(hdr, hash_lock); } /* * If a read request has a callback (i.e. acb_done is not NULL), then we * make a buf containing the data according to the parameters which were * passed in. The implementation of arc_buf_alloc_impl() ensures that we * aren't needlessly decompressing the data multiple times. */ int callback_cnt = 0; for (acb = callback_list; acb != NULL; acb = acb->acb_next) { if (!acb->acb_done) continue; callback_cnt++; if (zio->io_error != 0) continue; int error = arc_buf_alloc_impl(hdr, zio->io_spa, &acb->acb_zb, acb->acb_private, acb->acb_encrypted, acb->acb_compressed, acb->acb_noauth, B_TRUE, &acb->acb_buf); - if (error != 0) { - (void) remove_reference(hdr, hash_lock, - acb->acb_private); - arc_buf_destroy_impl(acb->acb_buf); - acb->acb_buf = NULL; - } /* * Assert non-speculative zios didn't fail because an * encryption key wasn't loaded */ ASSERT((zio->io_flags & ZIO_FLAG_SPECULATIVE) || error != EACCES); /* * If we failed to decrypt, report an error now (as the zio * layer would have done if it had done the transforms). */ if (error == ECKSUM) { ASSERT(BP_IS_PROTECTED(bp)); error = SET_ERROR(EIO); if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(zio->io_spa, &acb->acb_zb); zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION, zio->io_spa, NULL, &acb->acb_zb, zio, 0, 0); } } - if (zio->io_error == 0) + if (error != 0) { + /* + * Decompression or decryption failed. Set + * io_error so that when we call acb_done + * (below), we will indicate that the read + * failed. Note that in the unusual case + * where one callback is compressed and another + * uncompressed, we will mark all of them + * as failed, even though the uncompressed + * one can't actually fail. In this case, + * the hdr will not be anonymous, because + * if there are multiple callbacks, it's + * because multiple threads found the same + * arc buf in the hash table. + */ zio->io_error = error; + } } + + /* + * If there are multiple callbacks, we must have the hash lock, + * because the only way for multiple threads to find this hdr is + * in the hash table. This ensures that if there are multiple + * callbacks, the hdr is not anonymous. If it were anonymous, + * we couldn't use arc_buf_destroy() in the error case below. + */ + ASSERT(callback_cnt < 2 || hash_lock != NULL); + hdr->b_l1hdr.b_acb = NULL; arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); if (callback_cnt == 0) ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); ASSERT(refcount_is_zero(&hdr->b_l1hdr.b_refcnt) || callback_list != NULL); if (zio->io_error == 0) { arc_hdr_verify(hdr, zio->io_bp); } else { arc_hdr_set_flags(hdr, ARC_FLAG_IO_ERROR); if (hdr->b_l1hdr.b_state != arc_anon) arc_change_state(arc_anon, hdr, hash_lock); if (HDR_IN_HASH_TABLE(hdr)) buf_hash_remove(hdr); freeable = refcount_is_zero(&hdr->b_l1hdr.b_refcnt); } /* * Broadcast before we drop the hash_lock to avoid the possibility * that the hdr (and hence the cv) might be freed before we get to * the cv_broadcast(). */ cv_broadcast(&hdr->b_l1hdr.b_cv); if (hash_lock != NULL) { mutex_exit(hash_lock); } else { /* * This block was freed while we waited for the read to * complete. It has been removed from the hash table and * moved to the anonymous state (so that it won't show up * in the cache). */ ASSERT3P(hdr->b_l1hdr.b_state, ==, arc_anon); freeable = refcount_is_zero(&hdr->b_l1hdr.b_refcnt); } /* execute each callback and free its structure */ while ((acb = callback_list) != NULL) { - if (acb->acb_done) { + if (acb->acb_done != NULL) { + if (zio->io_error != 0 && acb->acb_buf != NULL) { + /* + * If arc_buf_alloc_impl() fails during + * decompression, the buf will still be + * allocated, and needs to be freed here. + */ + arc_buf_destroy(acb->acb_buf, + acb->acb_private); + acb->acb_buf = NULL; + } acb->acb_done(zio, &zio->io_bookmark, zio->io_bp, acb->acb_buf, acb->acb_private); } if (acb->acb_zio_dummy != NULL) { acb->acb_zio_dummy->io_error = zio->io_error; zio_nowait(acb->acb_zio_dummy); } callback_list = acb->acb_next; kmem_free(acb, sizeof (arc_callback_t)); } if (freeable) arc_hdr_destroy(hdr); } /* * "Read" the block at the specified DVA (in bp) via the * cache. If the block is found in the cache, invoke the provided * callback immediately and return. Note that the `zio' parameter * in the callback will be NULL in this case, since no IO was * required. If the block is not in the cache pass the read request * on to the spa with a substitute callback function, so that the * requested block will be added to the cache. * * If a read request arrives for a block that has a read in-progress, * either wait for the in-progress read to complete (and return the * results); or, if this is a read with a "done" func, add a record * to the read to invoke the "done" func when the read completes, * and return; or just return. * * arc_read_done() will invoke all the requested "done" functions * for readers of this block. */ int arc_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, arc_read_done_func_t *done, void *private, zio_priority_t priority, int zio_flags, arc_flags_t *arc_flags, const zbookmark_phys_t *zb) { arc_buf_hdr_t *hdr = NULL; kmutex_t *hash_lock = NULL; zio_t *rzio; uint64_t guid = spa_load_guid(spa); boolean_t compressed_read = (zio_flags & ZIO_FLAG_RAW_COMPRESS) != 0; boolean_t encrypted_read = BP_IS_ENCRYPTED(bp) && (zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0; boolean_t noauth_read = BP_IS_AUTHENTICATED(bp) && (zio_flags & ZIO_FLAG_RAW_ENCRYPT) != 0; int rc = 0; ASSERT(!BP_IS_EMBEDDED(bp) || BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA); top: if (!BP_IS_EMBEDDED(bp)) { /* * Embedded BP's have no DVA and require no I/O to "read". * Create an anonymous arc buf to back it. */ hdr = buf_hash_find(guid, bp, &hash_lock); } /* * Determine if we have an L1 cache hit or a cache miss. For simplicity * we maintain encrypted data seperately from compressed / uncompressed * data. If the user is requesting raw encrypted data and we don't have * that in the header we will read from disk to guarantee that we can * get it even if the encryption keys aren't loaded. */ if (hdr != NULL && HDR_HAS_L1HDR(hdr) && (HDR_HAS_RABD(hdr) || (hdr->b_l1hdr.b_pabd != NULL && !encrypted_read))) { arc_buf_t *buf = NULL; *arc_flags |= ARC_FLAG_CACHED; if (HDR_IO_IN_PROGRESS(hdr)) { zio_t *head_zio = hdr->b_l1hdr.b_acb->acb_zio_head; ASSERT3P(head_zio, !=, NULL); if ((hdr->b_flags & ARC_FLAG_PRIO_ASYNC_READ) && priority == ZIO_PRIORITY_SYNC_READ) { /* * This is a sync read that needs to wait for * an in-flight async read. Request that the * zio have its priority upgraded. */ zio_change_priority(head_zio, priority); DTRACE_PROBE1(arc__async__upgrade__sync, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_async_upgrade_sync); } if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) { arc_hdr_clear_flags(hdr, ARC_FLAG_PREDICTIVE_PREFETCH); } if (*arc_flags & ARC_FLAG_WAIT) { cv_wait(&hdr->b_l1hdr.b_cv, hash_lock); mutex_exit(hash_lock); goto top; } ASSERT(*arc_flags & ARC_FLAG_NOWAIT); if (done) { arc_callback_t *acb = NULL; acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP); acb->acb_done = done; acb->acb_private = private; acb->acb_compressed = compressed_read; acb->acb_encrypted = encrypted_read; acb->acb_noauth = noauth_read; acb->acb_zb = *zb; if (pio != NULL) acb->acb_zio_dummy = zio_null(pio, spa, NULL, NULL, NULL, zio_flags); ASSERT3P(acb->acb_done, !=, NULL); acb->acb_zio_head = head_zio; acb->acb_next = hdr->b_l1hdr.b_acb; hdr->b_l1hdr.b_acb = acb; mutex_exit(hash_lock); goto out; } mutex_exit(hash_lock); goto out; } ASSERT(hdr->b_l1hdr.b_state == arc_mru || hdr->b_l1hdr.b_state == arc_mfu); if (done) { if (hdr->b_flags & ARC_FLAG_PREDICTIVE_PREFETCH) { /* * This is a demand read which does not have to * wait for i/o because we did a predictive * prefetch i/o for it, which has completed. */ DTRACE_PROBE1( arc__demand__hit__predictive__prefetch, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP( arcstat_demand_hit_predictive_prefetch); arc_hdr_clear_flags(hdr, ARC_FLAG_PREDICTIVE_PREFETCH); } if (hdr->b_flags & ARC_FLAG_PRESCIENT_PREFETCH) { ARCSTAT_BUMP( arcstat_demand_hit_prescient_prefetch); arc_hdr_clear_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH); } ASSERT(!BP_IS_EMBEDDED(bp) || !BP_IS_HOLE(bp)); /* Get a buf with the desired data in it. */ rc = arc_buf_alloc_impl(hdr, spa, zb, private, encrypted_read, compressed_read, noauth_read, B_TRUE, &buf); if (rc == ECKSUM) { /* * Convert authentication and decryption errors * to EIO (and generate an ereport if needed) * before leaving the ARC. */ rc = SET_ERROR(EIO); if ((zio_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(spa, zb); zfs_ereport_post( FM_EREPORT_ZFS_AUTHENTICATION, spa, NULL, zb, NULL, 0, 0); } } if (rc != 0) { (void) remove_reference(hdr, hash_lock, private); arc_buf_destroy_impl(buf); buf = NULL; } /* assert any errors weren't due to unloaded keys */ ASSERT((zio_flags & ZIO_FLAG_SPECULATIVE) || rc != EACCES); } else if (*arc_flags & ARC_FLAG_PREFETCH && refcount_count(&hdr->b_l1hdr.b_refcnt) == 0) { arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH); } DTRACE_PROBE1(arc__hit, arc_buf_hdr_t *, hdr); arc_access(hdr, hash_lock); if (*arc_flags & ARC_FLAG_PRESCIENT_PREFETCH) arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH); if (*arc_flags & ARC_FLAG_L2CACHE) arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE); mutex_exit(hash_lock); ARCSTAT_BUMP(arcstat_hits); ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr), demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, hits); if (done) done(NULL, zb, bp, buf, private); } else { uint64_t lsize = BP_GET_LSIZE(bp); uint64_t psize = BP_GET_PSIZE(bp); arc_callback_t *acb; vdev_t *vd = NULL; uint64_t addr = 0; boolean_t devw = B_FALSE; uint64_t size; abd_t *hdr_abd; /* * Gracefully handle a damaged logical block size as a * checksum error. */ if (lsize > spa_maxblocksize(spa)) { rc = SET_ERROR(ECKSUM); goto out; } if (hdr == NULL) { /* this block is not in the cache */ arc_buf_hdr_t *exists = NULL; arc_buf_contents_t type = BP_GET_BUFC_TYPE(bp); hdr = arc_hdr_alloc(spa_load_guid(spa), psize, lsize, BP_IS_PROTECTED(bp), BP_GET_COMPRESS(bp), type, encrypted_read); if (!BP_IS_EMBEDDED(bp)) { hdr->b_dva = *BP_IDENTITY(bp); hdr->b_birth = BP_PHYSICAL_BIRTH(bp); exists = buf_hash_insert(hdr, &hash_lock); } if (exists != NULL) { /* somebody beat us to the hash insert */ mutex_exit(hash_lock); buf_discard_identity(hdr); arc_hdr_destroy(hdr); goto top; /* restart the IO request */ } } else { /* * This block is in the ghost cache or encrypted data * was requested and we didn't have it. If it was * L2-only (and thus didn't have an L1 hdr), * we realloc the header to add an L1 hdr. */ if (!HDR_HAS_L1HDR(hdr)) { hdr = arc_hdr_realloc(hdr, hdr_l2only_cache, hdr_full_cache); } if (GHOST_STATE(hdr->b_l1hdr.b_state)) { ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT0(refcount_count(&hdr->b_l1hdr.b_refcnt)); ASSERT3P(hdr->b_l1hdr.b_buf, ==, NULL); ASSERT3P(hdr->b_l1hdr.b_freeze_cksum, ==, NULL); } else if (HDR_IO_IN_PROGRESS(hdr)) { /* * If this header already had an IO in progress * and we are performing another IO to fetch * encrypted data we must wait until the first * IO completes so as not to confuse * arc_read_done(). This should be very rare * and so the performance impact shouldn't * matter. */ cv_wait(&hdr->b_l1hdr.b_cv, hash_lock); mutex_exit(hash_lock); goto top; } /* * This is a delicate dance that we play here. * This hdr might be in the ghost list so we access * it to move it out of the ghost list before we * initiate the read. If it's a prefetch then * it won't have a callback so we'll remove the * reference that arc_buf_alloc_impl() created. We * do this after we've called arc_access() to * avoid hitting an assert in remove_reference(). */ arc_access(hdr, hash_lock); arc_hdr_alloc_abd(hdr, encrypted_read); } if (encrypted_read) { ASSERT(HDR_HAS_RABD(hdr)); size = HDR_GET_PSIZE(hdr); hdr_abd = hdr->b_crypt_hdr.b_rabd; zio_flags |= ZIO_FLAG_RAW; } else { ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); size = arc_hdr_size(hdr); hdr_abd = hdr->b_l1hdr.b_pabd; if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF) { zio_flags |= ZIO_FLAG_RAW_COMPRESS; } /* * For authenticated bp's, we do not ask the ZIO layer * to authenticate them since this will cause the entire * IO to fail if the key isn't loaded. Instead, we * defer authentication until arc_buf_fill(), which will * verify the data when the key is available. */ if (BP_IS_AUTHENTICATED(bp)) zio_flags |= ZIO_FLAG_RAW_ENCRYPT; } if (*arc_flags & ARC_FLAG_PREFETCH && refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) arc_hdr_set_flags(hdr, ARC_FLAG_PREFETCH); if (*arc_flags & ARC_FLAG_PRESCIENT_PREFETCH) arc_hdr_set_flags(hdr, ARC_FLAG_PRESCIENT_PREFETCH); if (*arc_flags & ARC_FLAG_L2CACHE) arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE); if (BP_IS_AUTHENTICATED(bp)) arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH); if (BP_GET_LEVEL(bp) > 0) arc_hdr_set_flags(hdr, ARC_FLAG_INDIRECT); if (*arc_flags & ARC_FLAG_PREDICTIVE_PREFETCH) arc_hdr_set_flags(hdr, ARC_FLAG_PREDICTIVE_PREFETCH); ASSERT(!GHOST_STATE(hdr->b_l1hdr.b_state)); acb = kmem_zalloc(sizeof (arc_callback_t), KM_SLEEP); acb->acb_done = done; acb->acb_private = private; acb->acb_compressed = compressed_read; acb->acb_encrypted = encrypted_read; acb->acb_noauth = noauth_read; acb->acb_zb = *zb; ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); hdr->b_l1hdr.b_acb = acb; arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); if (HDR_HAS_L2HDR(hdr) && (vd = hdr->b_l2hdr.b_dev->l2ad_vdev) != NULL) { devw = hdr->b_l2hdr.b_dev->l2ad_writing; addr = hdr->b_l2hdr.b_daddr; /* * Lock out L2ARC device removal. */ if (vdev_is_dead(vd) || !spa_config_tryenter(spa, SCL_L2ARC, vd, RW_READER)) vd = NULL; } /* * We count both async reads and scrub IOs as asynchronous so * that both can be upgraded in the event of a cache hit while * the read IO is still in-flight. */ if (priority == ZIO_PRIORITY_ASYNC_READ || priority == ZIO_PRIORITY_SCRUB) arc_hdr_set_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ); else arc_hdr_clear_flags(hdr, ARC_FLAG_PRIO_ASYNC_READ); /* * At this point, we have a level 1 cache miss. Try again in * L2ARC if possible. */ ASSERT3U(HDR_GET_LSIZE(hdr), ==, lsize); DTRACE_PROBE4(arc__miss, arc_buf_hdr_t *, hdr, blkptr_t *, bp, uint64_t, lsize, zbookmark_phys_t *, zb); ARCSTAT_BUMP(arcstat_misses); ARCSTAT_CONDSTAT(!HDR_PREFETCH(hdr), demand, prefetch, !HDR_ISTYPE_METADATA(hdr), data, metadata, misses); if (vd != NULL && l2arc_ndev != 0 && !(l2arc_norw && devw)) { /* * Read from the L2ARC if the following are true: * 1. The L2ARC vdev was previously cached. * 2. This buffer still has L2ARC metadata. * 3. This buffer isn't currently writing to the L2ARC. * 4. The L2ARC entry wasn't evicted, which may * also have invalidated the vdev. * 5. This isn't prefetch and l2arc_noprefetch is set. */ if (HDR_HAS_L2HDR(hdr) && !HDR_L2_WRITING(hdr) && !HDR_L2_EVICTED(hdr) && !(l2arc_noprefetch && HDR_PREFETCH(hdr))) { l2arc_read_callback_t *cb; abd_t *abd; uint64_t asize; DTRACE_PROBE1(l2arc__hit, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_hits); atomic_inc_32(&hdr->b_l2hdr.b_hits); cb = kmem_zalloc(sizeof (l2arc_read_callback_t), KM_SLEEP); cb->l2rcb_hdr = hdr; cb->l2rcb_bp = *bp; cb->l2rcb_zb = *zb; cb->l2rcb_flags = zio_flags; asize = vdev_psize_to_asize(vd, size); if (asize != size) { abd = abd_alloc_for_io(asize, HDR_ISTYPE_METADATA(hdr)); cb->l2rcb_abd = abd; } else { abd = hdr_abd; } ASSERT(addr >= VDEV_LABEL_START_SIZE && addr + asize <= vd->vdev_psize - VDEV_LABEL_END_SIZE); /* * l2arc read. The SCL_L2ARC lock will be * released by l2arc_read_done(). * Issue a null zio if the underlying buffer * was squashed to zero size by compression. */ ASSERT3U(arc_hdr_get_compress(hdr), !=, ZIO_COMPRESS_EMPTY); rzio = zio_read_phys(pio, vd, addr, asize, abd, ZIO_CHECKSUM_OFF, l2arc_read_done, cb, priority, zio_flags | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY, B_FALSE); acb->acb_zio_head = rzio; if (hash_lock != NULL) mutex_exit(hash_lock); DTRACE_PROBE2(l2arc__read, vdev_t *, vd, zio_t *, rzio); ARCSTAT_INCR(arcstat_l2_read_bytes, HDR_GET_PSIZE(hdr)); if (*arc_flags & ARC_FLAG_NOWAIT) { zio_nowait(rzio); goto out; } ASSERT(*arc_flags & ARC_FLAG_WAIT); if (zio_wait(rzio) == 0) goto out; /* l2arc read error; goto zio_read() */ if (hash_lock != NULL) mutex_enter(hash_lock); } else { DTRACE_PROBE1(l2arc__miss, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_misses); if (HDR_L2_WRITING(hdr)) ARCSTAT_BUMP(arcstat_l2_rw_clash); spa_config_exit(spa, SCL_L2ARC, vd); } } else { if (vd != NULL) spa_config_exit(spa, SCL_L2ARC, vd); if (l2arc_ndev != 0) { DTRACE_PROBE1(l2arc__miss, arc_buf_hdr_t *, hdr); ARCSTAT_BUMP(arcstat_l2_misses); } } rzio = zio_read(pio, spa, bp, hdr_abd, size, arc_read_done, hdr, priority, zio_flags, zb); acb->acb_zio_head = rzio; if (hash_lock != NULL) mutex_exit(hash_lock); if (*arc_flags & ARC_FLAG_WAIT) { rc = zio_wait(rzio); goto out; } ASSERT(*arc_flags & ARC_FLAG_NOWAIT); zio_nowait(rzio); } out: /* embedded bps don't actually go to disk */ if (!BP_IS_EMBEDDED(bp)) spa_read_history_add(spa, zb, *arc_flags); return (rc); } arc_prune_t * arc_add_prune_callback(arc_prune_func_t *func, void *private) { arc_prune_t *p; p = kmem_alloc(sizeof (*p), KM_SLEEP); p->p_pfunc = func; p->p_private = private; list_link_init(&p->p_node); refcount_create(&p->p_refcnt); mutex_enter(&arc_prune_mtx); refcount_add(&p->p_refcnt, &arc_prune_list); list_insert_head(&arc_prune_list, p); mutex_exit(&arc_prune_mtx); return (p); } void arc_remove_prune_callback(arc_prune_t *p) { boolean_t wait = B_FALSE; mutex_enter(&arc_prune_mtx); list_remove(&arc_prune_list, p); if (refcount_remove(&p->p_refcnt, &arc_prune_list) > 0) wait = B_TRUE; mutex_exit(&arc_prune_mtx); /* wait for arc_prune_task to finish */ if (wait) taskq_wait_outstanding(arc_prune_taskq, 0); ASSERT0(refcount_count(&p->p_refcnt)); refcount_destroy(&p->p_refcnt); kmem_free(p, sizeof (*p)); } /* * Notify the arc that a block was freed, and thus will never be used again. */ void arc_freed(spa_t *spa, const blkptr_t *bp) { arc_buf_hdr_t *hdr; kmutex_t *hash_lock; uint64_t guid = spa_load_guid(spa); ASSERT(!BP_IS_EMBEDDED(bp)); hdr = buf_hash_find(guid, bp, &hash_lock); if (hdr == NULL) return; /* * We might be trying to free a block that is still doing I/O * (i.e. prefetch) or has a reference (i.e. a dedup-ed, * dmu_sync-ed block). If this block is being prefetched, then it * would still have the ARC_FLAG_IO_IN_PROGRESS flag set on the hdr * until the I/O completes. A block may also have a reference if it is * part of a dedup-ed, dmu_synced write. The dmu_sync() function would * have written the new block to its final resting place on disk but * without the dedup flag set. This would have left the hdr in the MRU * state and discoverable. When the txg finally syncs it detects that * the block was overridden in open context and issues an override I/O. * Since this is a dedup block, the override I/O will determine if the * block is already in the DDT. If so, then it will replace the io_bp * with the bp from the DDT and allow the I/O to finish. When the I/O * reaches the done callback, dbuf_write_override_done, it will * check to see if the io_bp and io_bp_override are identical. * If they are not, then it indicates that the bp was replaced with * the bp in the DDT and the override bp is freed. This allows * us to arrive here with a reference on a block that is being * freed. So if we have an I/O in progress, or a reference to * this hdr, then we don't destroy the hdr. */ if (!HDR_HAS_L1HDR(hdr) || (!HDR_IO_IN_PROGRESS(hdr) && refcount_is_zero(&hdr->b_l1hdr.b_refcnt))) { arc_change_state(arc_anon, hdr, hash_lock); arc_hdr_destroy(hdr); mutex_exit(hash_lock); } else { mutex_exit(hash_lock); } } /* * Release this buffer from the cache, making it an anonymous buffer. This * must be done after a read and prior to modifying the buffer contents. * If the buffer has more than one reference, we must make * a new hdr for the buffer. */ void arc_release(arc_buf_t *buf, void *tag) { arc_buf_hdr_t *hdr = buf->b_hdr; /* * It would be nice to assert that if its DMU metadata (level > * 0 || it's the dnode file), then it must be syncing context. * But we don't know that information at this level. */ mutex_enter(&buf->b_evict_lock); ASSERT(HDR_HAS_L1HDR(hdr)); /* * We don't grab the hash lock prior to this check, because if * the buffer's header is in the arc_anon state, it won't be * linked into the hash table. */ if (hdr->b_l1hdr.b_state == arc_anon) { mutex_exit(&buf->b_evict_lock); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT(!HDR_IN_HASH_TABLE(hdr)); ASSERT(!HDR_HAS_L2HDR(hdr)); ASSERT(HDR_EMPTY(hdr)); ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1); ASSERT3S(refcount_count(&hdr->b_l1hdr.b_refcnt), ==, 1); ASSERT(!list_link_active(&hdr->b_l1hdr.b_arc_node)); hdr->b_l1hdr.b_arc_access = 0; /* * If the buf is being overridden then it may already * have a hdr that is not empty. */ buf_discard_identity(hdr); arc_buf_thaw(buf); return; } kmutex_t *hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); /* * This assignment is only valid as long as the hash_lock is * held, we must be careful not to reference state or the * b_state field after dropping the lock. */ arc_state_t *state = hdr->b_l1hdr.b_state; ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); ASSERT3P(state, !=, arc_anon); /* this buffer is not on any list */ ASSERT3S(refcount_count(&hdr->b_l1hdr.b_refcnt), >, 0); if (HDR_HAS_L2HDR(hdr)) { mutex_enter(&hdr->b_l2hdr.b_dev->l2ad_mtx); /* * We have to recheck this conditional again now that * we're holding the l2ad_mtx to prevent a race with * another thread which might be concurrently calling * l2arc_evict(). In that case, l2arc_evict() might have * destroyed the header's L2 portion as we were waiting * to acquire the l2ad_mtx. */ if (HDR_HAS_L2HDR(hdr)) arc_hdr_l2hdr_destroy(hdr); mutex_exit(&hdr->b_l2hdr.b_dev->l2ad_mtx); } /* * Do we have more than one buf? */ if (hdr->b_l1hdr.b_bufcnt > 1) { arc_buf_hdr_t *nhdr; uint64_t spa = hdr->b_spa; uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t lsize = HDR_GET_LSIZE(hdr); boolean_t protected = HDR_PROTECTED(hdr); enum zio_compress compress = arc_hdr_get_compress(hdr); arc_buf_contents_t type = arc_buf_type(hdr); VERIFY3U(hdr->b_type, ==, type); ASSERT(hdr->b_l1hdr.b_buf != buf || buf->b_next != NULL); (void) remove_reference(hdr, hash_lock, tag); if (arc_buf_is_shared(buf) && !ARC_BUF_COMPRESSED(buf)) { ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf); ASSERT(ARC_BUF_LAST(buf)); } /* * Pull the data off of this hdr and attach it to * a new anonymous hdr. Also find the last buffer * in the hdr's buffer list. */ arc_buf_t *lastbuf = arc_buf_remove(hdr, buf); ASSERT3P(lastbuf, !=, NULL); /* * If the current arc_buf_t and the hdr are sharing their data * buffer, then we must stop sharing that block. */ if (arc_buf_is_shared(buf)) { ASSERT3P(hdr->b_l1hdr.b_buf, !=, buf); VERIFY(!arc_buf_is_shared(lastbuf)); /* * First, sever the block sharing relationship between * buf and the arc_buf_hdr_t. */ arc_unshare_buf(hdr, buf); /* * Now we need to recreate the hdr's b_pabd. Since we * have lastbuf handy, we try to share with it, but if * we can't then we allocate a new b_pabd and copy the * data from buf into it. */ if (arc_can_share(hdr, lastbuf)) { arc_share_buf(hdr, lastbuf); } else { arc_hdr_alloc_abd(hdr, B_FALSE); abd_copy_from_buf(hdr->b_l1hdr.b_pabd, buf->b_data, psize); } VERIFY3P(lastbuf->b_data, !=, NULL); } else if (HDR_SHARED_DATA(hdr)) { /* * Uncompressed shared buffers are always at the end * of the list. Compressed buffers don't have the * same requirements. This makes it hard to * simply assert that the lastbuf is shared so * we rely on the hdr's compression flags to determine * if we have a compressed, shared buffer. */ ASSERT(arc_buf_is_shared(lastbuf) || arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF); ASSERT(!ARC_BUF_SHARED(buf)); } ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); ASSERT3P(state, !=, arc_l2c_only); (void) refcount_remove_many(&state->arcs_size, arc_buf_size(buf), buf); if (refcount_is_zero(&hdr->b_l1hdr.b_refcnt)) { ASSERT3P(state, !=, arc_l2c_only); (void) refcount_remove_many(&state->arcs_esize[type], arc_buf_size(buf), buf); } hdr->b_l1hdr.b_bufcnt -= 1; if (ARC_BUF_ENCRYPTED(buf)) hdr->b_crypt_hdr.b_ebufcnt -= 1; arc_cksum_verify(buf); arc_buf_unwatch(buf); /* if this is the last uncompressed buf free the checksum */ if (!arc_hdr_has_uncompressed_buf(hdr)) arc_cksum_free(hdr); mutex_exit(hash_lock); /* * Allocate a new hdr. The new hdr will contain a b_pabd * buffer which will be freed in arc_write(). */ nhdr = arc_hdr_alloc(spa, psize, lsize, protected, compress, type, HDR_HAS_RABD(hdr)); ASSERT3P(nhdr->b_l1hdr.b_buf, ==, NULL); ASSERT0(nhdr->b_l1hdr.b_bufcnt); ASSERT0(refcount_count(&nhdr->b_l1hdr.b_refcnt)); VERIFY3U(nhdr->b_type, ==, type); ASSERT(!HDR_SHARED_DATA(nhdr)); nhdr->b_l1hdr.b_buf = buf; nhdr->b_l1hdr.b_bufcnt = 1; if (ARC_BUF_ENCRYPTED(buf)) nhdr->b_crypt_hdr.b_ebufcnt = 1; nhdr->b_l1hdr.b_mru_hits = 0; nhdr->b_l1hdr.b_mru_ghost_hits = 0; nhdr->b_l1hdr.b_mfu_hits = 0; nhdr->b_l1hdr.b_mfu_ghost_hits = 0; nhdr->b_l1hdr.b_l2_hits = 0; (void) refcount_add(&nhdr->b_l1hdr.b_refcnt, tag); buf->b_hdr = nhdr; mutex_exit(&buf->b_evict_lock); (void) refcount_add_many(&arc_anon->arcs_size, HDR_GET_LSIZE(nhdr), buf); } else { mutex_exit(&buf->b_evict_lock); ASSERT(refcount_count(&hdr->b_l1hdr.b_refcnt) == 1); /* protected by hash lock, or hdr is on arc_anon */ ASSERT(!multilist_link_active(&hdr->b_l1hdr.b_arc_node)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); hdr->b_l1hdr.b_mru_hits = 0; hdr->b_l1hdr.b_mru_ghost_hits = 0; hdr->b_l1hdr.b_mfu_hits = 0; hdr->b_l1hdr.b_mfu_ghost_hits = 0; hdr->b_l1hdr.b_l2_hits = 0; arc_change_state(arc_anon, hdr, hash_lock); hdr->b_l1hdr.b_arc_access = 0; mutex_exit(hash_lock); buf_discard_identity(hdr); arc_buf_thaw(buf); } } int arc_released(arc_buf_t *buf) { int released; mutex_enter(&buf->b_evict_lock); released = (buf->b_data != NULL && buf->b_hdr->b_l1hdr.b_state == arc_anon); mutex_exit(&buf->b_evict_lock); return (released); } #ifdef ZFS_DEBUG int arc_referenced(arc_buf_t *buf) { int referenced; mutex_enter(&buf->b_evict_lock); referenced = (refcount_count(&buf->b_hdr->b_l1hdr.b_refcnt)); mutex_exit(&buf->b_evict_lock); return (referenced); } #endif static void arc_write_ready(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; arc_buf_hdr_t *hdr = buf->b_hdr; blkptr_t *bp = zio->io_bp; uint64_t psize = BP_IS_HOLE(bp) ? 0 : BP_GET_PSIZE(bp); fstrans_cookie_t cookie = spl_fstrans_mark(); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT(!refcount_is_zero(&buf->b_hdr->b_l1hdr.b_refcnt)); ASSERT(hdr->b_l1hdr.b_bufcnt > 0); /* * If we're reexecuting this zio because the pool suspended, then * cleanup any state that was previously set the first time the * callback was invoked. */ if (zio->io_flags & ZIO_FLAG_REEXECUTED) { arc_cksum_free(hdr); arc_buf_unwatch(buf); if (hdr->b_l1hdr.b_pabd != NULL) { if (arc_buf_is_shared(buf)) { arc_unshare_buf(hdr, buf); } else { arc_hdr_free_abd(hdr, B_FALSE); } } if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); } ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); ASSERT(!HDR_HAS_RABD(hdr)); ASSERT(!HDR_SHARED_DATA(hdr)); ASSERT(!arc_buf_is_shared(buf)); callback->awcb_ready(zio, buf, callback->awcb_private); if (HDR_IO_IN_PROGRESS(hdr)) ASSERT(zio->io_flags & ZIO_FLAG_REEXECUTED); arc_hdr_set_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); if (BP_IS_PROTECTED(bp) != !!HDR_PROTECTED(hdr)) hdr = arc_hdr_realloc_crypt(hdr, BP_IS_PROTECTED(bp)); if (BP_IS_PROTECTED(bp)) { /* ZIL blocks are written through zio_rewrite */ ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG); ASSERT(HDR_PROTECTED(hdr)); if (BP_SHOULD_BYTESWAP(bp)) { if (BP_GET_LEVEL(bp) > 0) { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_UINT64; } else { hdr->b_l1hdr.b_byteswap = DMU_OT_BYTESWAP(BP_GET_TYPE(bp)); } } else { hdr->b_l1hdr.b_byteswap = DMU_BSWAP_NUMFUNCS; } hdr->b_crypt_hdr.b_ot = BP_GET_TYPE(bp); hdr->b_crypt_hdr.b_dsobj = zio->io_bookmark.zb_objset; zio_crypt_decode_params_bp(bp, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv); zio_crypt_decode_mac_bp(bp, hdr->b_crypt_hdr.b_mac); } /* * If this block was written for raw encryption but the zio layer * ended up only authenticating it, adjust the buffer flags now. */ if (BP_IS_AUTHENTICATED(bp) && ARC_BUF_ENCRYPTED(buf)) { arc_hdr_set_flags(hdr, ARC_FLAG_NOAUTH); buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED; if (BP_GET_COMPRESS(bp) == ZIO_COMPRESS_OFF) buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; } else if (BP_IS_HOLE(bp) && ARC_BUF_ENCRYPTED(buf)) { buf->b_flags &= ~ARC_BUF_FLAG_ENCRYPTED; buf->b_flags &= ~ARC_BUF_FLAG_COMPRESSED; } /* this must be done after the buffer flags are adjusted */ arc_cksum_compute(buf); enum zio_compress compress; if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) { compress = ZIO_COMPRESS_OFF; } else { ASSERT3U(HDR_GET_LSIZE(hdr), ==, BP_GET_LSIZE(bp)); compress = BP_GET_COMPRESS(bp); } HDR_SET_PSIZE(hdr, psize); arc_hdr_set_compress(hdr, compress); if (zio->io_error != 0 || psize == 0) goto out; /* * Fill the hdr with data. If the buffer is encrypted we have no choice * but to copy the data into b_radb. If the hdr is compressed, the data * we want is available from the zio, otherwise we can take it from * the buf. * * We might be able to share the buf's data with the hdr here. However, * doing so would cause the ARC to be full of linear ABDs if we write a * lot of shareable data. As a compromise, we check whether scattered * ABDs are allowed, and assume that if they are then the user wants * the ARC to be primarily filled with them regardless of the data being * written. Therefore, if they're allowed then we allocate one and copy * the data into it; otherwise, we share the data directly if we can. */ if (ARC_BUF_ENCRYPTED(buf)) { ASSERT3U(psize, >, 0); ASSERT(ARC_BUF_COMPRESSED(buf)); arc_hdr_alloc_abd(hdr, B_TRUE); abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize); } else if (zfs_abd_scatter_enabled || !arc_can_share(hdr, buf)) { /* * Ideally, we would always copy the io_abd into b_pabd, but the * user may have disabled compressed ARC, thus we must check the * hdr's compression setting rather than the io_bp's. */ if (BP_IS_ENCRYPTED(bp)) { ASSERT3U(psize, >, 0); arc_hdr_alloc_abd(hdr, B_TRUE); abd_copy(hdr->b_crypt_hdr.b_rabd, zio->io_abd, psize); } else if (arc_hdr_get_compress(hdr) != ZIO_COMPRESS_OFF && !ARC_BUF_COMPRESSED(buf)) { ASSERT3U(psize, >, 0); arc_hdr_alloc_abd(hdr, B_FALSE); abd_copy(hdr->b_l1hdr.b_pabd, zio->io_abd, psize); } else { ASSERT3U(zio->io_orig_size, ==, arc_hdr_size(hdr)); arc_hdr_alloc_abd(hdr, B_FALSE); abd_copy_from_buf(hdr->b_l1hdr.b_pabd, buf->b_data, arc_buf_size(buf)); } } else { ASSERT3P(buf->b_data, ==, abd_to_buf(zio->io_orig_abd)); ASSERT3U(zio->io_orig_size, ==, arc_buf_size(buf)); ASSERT3U(hdr->b_l1hdr.b_bufcnt, ==, 1); arc_share_buf(hdr, buf); } out: arc_hdr_verify(hdr, bp); spl_fstrans_unmark(cookie); } static void arc_write_children_ready(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; callback->awcb_children_ready(zio, buf, callback->awcb_private); } /* * The SPA calls this callback for each physical write that happens on behalf * of a logical write. See the comment in dbuf_write_physdone() for details. */ static void arc_write_physdone(zio_t *zio) { arc_write_callback_t *cb = zio->io_private; if (cb->awcb_physdone != NULL) cb->awcb_physdone(zio, cb->awcb_buf, cb->awcb_private); } static void arc_write_done(zio_t *zio) { arc_write_callback_t *callback = zio->io_private; arc_buf_t *buf = callback->awcb_buf; arc_buf_hdr_t *hdr = buf->b_hdr; ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); if (zio->io_error == 0) { arc_hdr_verify(hdr, zio->io_bp); if (BP_IS_HOLE(zio->io_bp) || BP_IS_EMBEDDED(zio->io_bp)) { buf_discard_identity(hdr); } else { hdr->b_dva = *BP_IDENTITY(zio->io_bp); hdr->b_birth = BP_PHYSICAL_BIRTH(zio->io_bp); } } else { ASSERT(HDR_EMPTY(hdr)); } /* * If the block to be written was all-zero or compressed enough to be * embedded in the BP, no write was performed so there will be no * dva/birth/checksum. The buffer must therefore remain anonymous * (and uncached). */ if (!HDR_EMPTY(hdr)) { arc_buf_hdr_t *exists; kmutex_t *hash_lock; ASSERT3U(zio->io_error, ==, 0); arc_cksum_verify(buf); exists = buf_hash_insert(hdr, &hash_lock); if (exists != NULL) { /* * This can only happen if we overwrite for * sync-to-convergence, because we remove * buffers from the hash table when we arc_free(). */ if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp)) panic("bad overwrite, hdr=%p exists=%p", (void *)hdr, (void *)exists); ASSERT(refcount_is_zero( &exists->b_l1hdr.b_refcnt)); arc_change_state(arc_anon, exists, hash_lock); mutex_exit(hash_lock); arc_hdr_destroy(exists); exists = buf_hash_insert(hdr, &hash_lock); ASSERT3P(exists, ==, NULL); } else if (zio->io_flags & ZIO_FLAG_NOPWRITE) { /* nopwrite */ ASSERT(zio->io_prop.zp_nopwrite); if (!BP_EQUAL(&zio->io_bp_orig, zio->io_bp)) panic("bad nopwrite, hdr=%p exists=%p", (void *)hdr, (void *)exists); } else { /* Dedup */ ASSERT(hdr->b_l1hdr.b_bufcnt == 1); ASSERT(hdr->b_l1hdr.b_state == arc_anon); ASSERT(BP_GET_DEDUP(zio->io_bp)); ASSERT(BP_GET_LEVEL(zio->io_bp) == 0); } } arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); /* if it's not anon, we are doing a scrub */ if (exists == NULL && hdr->b_l1hdr.b_state == arc_anon) arc_access(hdr, hash_lock); mutex_exit(hash_lock); } else { arc_hdr_clear_flags(hdr, ARC_FLAG_IO_IN_PROGRESS); } ASSERT(!refcount_is_zero(&hdr->b_l1hdr.b_refcnt)); callback->awcb_done(zio, buf, callback->awcb_private); abd_put(zio->io_abd); kmem_free(callback, sizeof (arc_write_callback_t)); } zio_t * arc_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, arc_buf_t *buf, boolean_t l2arc, const zio_prop_t *zp, arc_write_done_func_t *ready, arc_write_done_func_t *children_ready, arc_write_done_func_t *physdone, arc_write_done_func_t *done, void *private, zio_priority_t priority, int zio_flags, const zbookmark_phys_t *zb) { arc_buf_hdr_t *hdr = buf->b_hdr; arc_write_callback_t *callback; zio_t *zio; zio_prop_t localprop = *zp; ASSERT3P(ready, !=, NULL); ASSERT3P(done, !=, NULL); ASSERT(!HDR_IO_ERROR(hdr)); ASSERT(!HDR_IO_IN_PROGRESS(hdr)); ASSERT3P(hdr->b_l1hdr.b_acb, ==, NULL); ASSERT3U(hdr->b_l1hdr.b_bufcnt, >, 0); if (l2arc) arc_hdr_set_flags(hdr, ARC_FLAG_L2CACHE); if (ARC_BUF_ENCRYPTED(buf)) { ASSERT(ARC_BUF_COMPRESSED(buf)); localprop.zp_encrypt = B_TRUE; localprop.zp_compress = HDR_GET_COMPRESS(hdr); localprop.zp_byteorder = (hdr->b_l1hdr.b_byteswap == DMU_BSWAP_NUMFUNCS) ? ZFS_HOST_BYTEORDER : !ZFS_HOST_BYTEORDER; bcopy(hdr->b_crypt_hdr.b_salt, localprop.zp_salt, ZIO_DATA_SALT_LEN); bcopy(hdr->b_crypt_hdr.b_iv, localprop.zp_iv, ZIO_DATA_IV_LEN); bcopy(hdr->b_crypt_hdr.b_mac, localprop.zp_mac, ZIO_DATA_MAC_LEN); if (DMU_OT_IS_ENCRYPTED(localprop.zp_type)) { localprop.zp_nopwrite = B_FALSE; localprop.zp_copies = MIN(localprop.zp_copies, SPA_DVAS_PER_BP - 1); } zio_flags |= ZIO_FLAG_RAW; } else if (ARC_BUF_COMPRESSED(buf)) { ASSERT3U(HDR_GET_LSIZE(hdr), !=, arc_buf_size(buf)); localprop.zp_compress = HDR_GET_COMPRESS(hdr); zio_flags |= ZIO_FLAG_RAW_COMPRESS; } callback = kmem_zalloc(sizeof (arc_write_callback_t), KM_SLEEP); callback->awcb_ready = ready; callback->awcb_children_ready = children_ready; callback->awcb_physdone = physdone; callback->awcb_done = done; callback->awcb_private = private; callback->awcb_buf = buf; /* * The hdr's b_pabd is now stale, free it now. A new data block * will be allocated when the zio pipeline calls arc_write_ready(). */ if (hdr->b_l1hdr.b_pabd != NULL) { /* * If the buf is currently sharing the data block with * the hdr then we need to break that relationship here. * The hdr will remain with a NULL data pointer and the * buf will take sole ownership of the block. */ if (arc_buf_is_shared(buf)) { arc_unshare_buf(hdr, buf); } else { arc_hdr_free_abd(hdr, B_FALSE); } VERIFY3P(buf->b_data, !=, NULL); } if (HDR_HAS_RABD(hdr)) arc_hdr_free_abd(hdr, B_TRUE); if (!(zio_flags & ZIO_FLAG_RAW)) arc_hdr_set_compress(hdr, ZIO_COMPRESS_OFF); ASSERT(!arc_buf_is_shared(buf)); ASSERT3P(hdr->b_l1hdr.b_pabd, ==, NULL); zio = zio_write(pio, spa, txg, bp, abd_get_from_buf(buf->b_data, HDR_GET_LSIZE(hdr)), HDR_GET_LSIZE(hdr), arc_buf_size(buf), &localprop, arc_write_ready, (children_ready != NULL) ? arc_write_children_ready : NULL, arc_write_physdone, arc_write_done, callback, priority, zio_flags, zb); return (zio); } static int arc_memory_throttle(spa_t *spa, uint64_t reserve, uint64_t txg) { #ifdef _KERNEL uint64_t available_memory = arc_free_memory(); #if defined(_ILP32) available_memory = MIN(available_memory, vmem_size(heap_arena, VMEM_FREE)); #endif if (available_memory > arc_all_memory() * arc_lotsfree_percent / 100) return (0); if (txg > spa->spa_lowmem_last_txg) { spa->spa_lowmem_last_txg = txg; spa->spa_lowmem_page_load = 0; } /* * If we are in pageout, we know that memory is already tight, * the arc is already going to be evicting, so we just want to * continue to let page writes occur as quickly as possible. */ if (current_is_kswapd()) { if (spa->spa_lowmem_page_load > MAX(arc_sys_free / 4, available_memory) / 4) { DMU_TX_STAT_BUMP(dmu_tx_memory_reclaim); return (SET_ERROR(ERESTART)); } /* Note: reserve is inflated, so we deflate */ atomic_add_64(&spa->spa_lowmem_page_load, reserve / 8); return (0); } else if (spa->spa_lowmem_page_load > 0 && arc_reclaim_needed()) { /* memory is low, delay before restarting */ ARCSTAT_INCR(arcstat_memory_throttle_count, 1); DMU_TX_STAT_BUMP(dmu_tx_memory_reclaim); return (SET_ERROR(EAGAIN)); } spa->spa_lowmem_page_load = 0; #endif /* _KERNEL */ return (0); } void arc_tempreserve_clear(uint64_t reserve) { atomic_add_64(&arc_tempreserve, -reserve); ASSERT((int64_t)arc_tempreserve >= 0); } int arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg) { int error; uint64_t anon_size; if (!arc_no_grow && reserve > arc_c/4 && reserve * 4 > (2ULL << SPA_MAXBLOCKSHIFT)) arc_c = MIN(arc_c_max, reserve * 4); /* * Throttle when the calculated memory footprint for the TXG * exceeds the target ARC size. */ if (reserve > arc_c) { DMU_TX_STAT_BUMP(dmu_tx_memory_reserve); return (SET_ERROR(ERESTART)); } /* * Don't count loaned bufs as in flight dirty data to prevent long * network delays from blocking transactions that are ready to be * assigned to a txg. */ /* assert that it has not wrapped around */ ASSERT3S(atomic_add_64_nv(&arc_loaned_bytes, 0), >=, 0); anon_size = MAX((int64_t)(refcount_count(&arc_anon->arcs_size) - arc_loaned_bytes), 0); /* * Writes will, almost always, require additional memory allocations * in order to compress/encrypt/etc the data. We therefore need to * make sure that there is sufficient available memory for this. */ error = arc_memory_throttle(spa, reserve, txg); if (error != 0) return (error); /* * Throttle writes when the amount of dirty data in the cache * gets too large. We try to keep the cache less than half full * of dirty blocks so that our sync times don't grow too large. * * In the case of one pool being built on another pool, we want * to make sure we don't end up throttling the lower (backing) * pool when the upper pool is the majority contributor to dirty * data. To insure we make forward progress during throttling, we * also check the current pool's net dirty data and only throttle * if it exceeds zfs_arc_pool_dirty_percent of the anonymous dirty * data in the cache. * * Note: if two requests come in concurrently, we might let them * both succeed, when one of them should fail. Not a huge deal. */ uint64_t total_dirty = reserve + arc_tempreserve + anon_size; uint64_t spa_dirty_anon = spa_dirty_data(spa); if (total_dirty > arc_c * zfs_arc_dirty_limit_percent / 100 && anon_size > arc_c * zfs_arc_anon_limit_percent / 100 && spa_dirty_anon > anon_size * zfs_arc_pool_dirty_percent / 100) { #ifdef ZFS_DEBUG uint64_t meta_esize = refcount_count(&arc_anon->arcs_esize[ARC_BUFC_METADATA]); uint64_t data_esize = refcount_count(&arc_anon->arcs_esize[ARC_BUFC_DATA]); dprintf("failing, arc_tempreserve=%lluK anon_meta=%lluK " "anon_data=%lluK tempreserve=%lluK arc_c=%lluK\n", arc_tempreserve >> 10, meta_esize >> 10, data_esize >> 10, reserve >> 10, arc_c >> 10); #endif DMU_TX_STAT_BUMP(dmu_tx_dirty_throttle); return (SET_ERROR(ERESTART)); } atomic_add_64(&arc_tempreserve, reserve); return (0); } static void arc_kstat_update_state(arc_state_t *state, kstat_named_t *size, kstat_named_t *evict_data, kstat_named_t *evict_metadata) { size->value.ui64 = refcount_count(&state->arcs_size); evict_data->value.ui64 = refcount_count(&state->arcs_esize[ARC_BUFC_DATA]); evict_metadata->value.ui64 = refcount_count(&state->arcs_esize[ARC_BUFC_METADATA]); } static int arc_kstat_update(kstat_t *ksp, int rw) { arc_stats_t *as = ksp->ks_data; if (rw == KSTAT_WRITE) { return (SET_ERROR(EACCES)); } else { arc_kstat_update_state(arc_anon, &as->arcstat_anon_size, &as->arcstat_anon_evictable_data, &as->arcstat_anon_evictable_metadata); arc_kstat_update_state(arc_mru, &as->arcstat_mru_size, &as->arcstat_mru_evictable_data, &as->arcstat_mru_evictable_metadata); arc_kstat_update_state(arc_mru_ghost, &as->arcstat_mru_ghost_size, &as->arcstat_mru_ghost_evictable_data, &as->arcstat_mru_ghost_evictable_metadata); arc_kstat_update_state(arc_mfu, &as->arcstat_mfu_size, &as->arcstat_mfu_evictable_data, &as->arcstat_mfu_evictable_metadata); arc_kstat_update_state(arc_mfu_ghost, &as->arcstat_mfu_ghost_size, &as->arcstat_mfu_ghost_evictable_data, &as->arcstat_mfu_ghost_evictable_metadata); ARCSTAT(arcstat_size) = aggsum_value(&arc_size); ARCSTAT(arcstat_meta_used) = aggsum_value(&arc_meta_used); ARCSTAT(arcstat_data_size) = aggsum_value(&astat_data_size); ARCSTAT(arcstat_metadata_size) = aggsum_value(&astat_metadata_size); ARCSTAT(arcstat_hdr_size) = aggsum_value(&astat_hdr_size); ARCSTAT(arcstat_l2_hdr_size) = aggsum_value(&astat_l2_hdr_size); ARCSTAT(arcstat_dbuf_size) = aggsum_value(&astat_dbuf_size); ARCSTAT(arcstat_dnode_size) = aggsum_value(&astat_dnode_size); ARCSTAT(arcstat_bonus_size) = aggsum_value(&astat_bonus_size); as->arcstat_memory_all_bytes.value.ui64 = arc_all_memory(); as->arcstat_memory_free_bytes.value.ui64 = arc_free_memory(); as->arcstat_memory_available_bytes.value.i64 = arc_available_memory(); } return (0); } /* * This function *must* return indices evenly distributed between all * sublists of the multilist. This is needed due to how the ARC eviction * code is laid out; arc_evict_state() assumes ARC buffers are evenly * distributed between all sublists and uses this assumption when * deciding which sublist to evict from and how much to evict from it. */ unsigned int arc_state_multilist_index_func(multilist_t *ml, void *obj) { arc_buf_hdr_t *hdr = obj; /* * We rely on b_dva to generate evenly distributed index * numbers using buf_hash below. So, as an added precaution, * let's make sure we never add empty buffers to the arc lists. */ ASSERT(!HDR_EMPTY(hdr)); /* * The assumption here, is the hash value for a given * arc_buf_hdr_t will remain constant throughout its lifetime * (i.e. its b_spa, b_dva, and b_birth fields don't change). * Thus, we don't need to store the header's sublist index * on insertion, as this index can be recalculated on removal. * * Also, the low order bits of the hash value are thought to be * distributed evenly. Otherwise, in the case that the multilist * has a power of two number of sublists, each sublists' usage * would not be evenly distributed. */ return (buf_hash(hdr->b_spa, &hdr->b_dva, hdr->b_birth) % multilist_get_num_sublists(ml)); } /* * Called during module initialization and periodically thereafter to * apply reasonable changes to the exposed performance tunings. Non-zero * zfs_* values which differ from the currently set values will be applied. */ static void arc_tuning_update(void) { uint64_t allmem = arc_all_memory(); unsigned long limit; /* Valid range: 64M - */ if ((zfs_arc_max) && (zfs_arc_max != arc_c_max) && (zfs_arc_max >= 64 << 20) && (zfs_arc_max < allmem) && (zfs_arc_max > arc_c_min)) { arc_c_max = zfs_arc_max; arc_c = arc_c_max; arc_p = (arc_c >> 1); if (arc_meta_limit > arc_c_max) arc_meta_limit = arc_c_max; if (arc_dnode_limit > arc_meta_limit) arc_dnode_limit = arc_meta_limit; } /* Valid range: 32M - */ if ((zfs_arc_min) && (zfs_arc_min != arc_c_min) && (zfs_arc_min >= 2ULL << SPA_MAXBLOCKSHIFT) && (zfs_arc_min <= arc_c_max)) { arc_c_min = zfs_arc_min; arc_c = MAX(arc_c, arc_c_min); } /* Valid range: 16M - */ if ((zfs_arc_meta_min) && (zfs_arc_meta_min != arc_meta_min) && (zfs_arc_meta_min >= 1ULL << SPA_MAXBLOCKSHIFT) && (zfs_arc_meta_min <= arc_c_max)) { arc_meta_min = zfs_arc_meta_min; if (arc_meta_limit < arc_meta_min) arc_meta_limit = arc_meta_min; if (arc_dnode_limit < arc_meta_min) arc_dnode_limit = arc_meta_min; } /* Valid range: - */ limit = zfs_arc_meta_limit ? zfs_arc_meta_limit : MIN(zfs_arc_meta_limit_percent, 100) * arc_c_max / 100; if ((limit != arc_meta_limit) && (limit >= arc_meta_min) && (limit <= arc_c_max)) arc_meta_limit = limit; /* Valid range: - */ limit = zfs_arc_dnode_limit ? zfs_arc_dnode_limit : MIN(zfs_arc_dnode_limit_percent, 100) * arc_meta_limit / 100; if ((limit != arc_dnode_limit) && (limit >= arc_meta_min) && (limit <= arc_meta_limit)) arc_dnode_limit = limit; /* Valid range: 1 - N */ if (zfs_arc_grow_retry) arc_grow_retry = zfs_arc_grow_retry; /* Valid range: 1 - N */ if (zfs_arc_shrink_shift) { arc_shrink_shift = zfs_arc_shrink_shift; arc_no_grow_shift = MIN(arc_no_grow_shift, arc_shrink_shift -1); } /* Valid range: 1 - N */ if (zfs_arc_p_min_shift) arc_p_min_shift = zfs_arc_p_min_shift; /* Valid range: 1 - N ms */ if (zfs_arc_min_prefetch_ms) arc_min_prefetch_ms = zfs_arc_min_prefetch_ms; /* Valid range: 1 - N ms */ if (zfs_arc_min_prescient_prefetch_ms) { arc_min_prescient_prefetch_ms = zfs_arc_min_prescient_prefetch_ms; } /* Valid range: 0 - 100 */ if ((zfs_arc_lotsfree_percent >= 0) && (zfs_arc_lotsfree_percent <= 100)) arc_lotsfree_percent = zfs_arc_lotsfree_percent; /* Valid range: 0 - */ if ((zfs_arc_sys_free) && (zfs_arc_sys_free != arc_sys_free)) arc_sys_free = MIN(MAX(zfs_arc_sys_free, 0), allmem); } static void arc_state_init(void) { arc_anon = &ARC_anon; arc_mru = &ARC_mru; arc_mru_ghost = &ARC_mru_ghost; arc_mfu = &ARC_mfu; arc_mfu_ghost = &ARC_mfu_ghost; arc_l2c_only = &ARC_l2c_only; arc_mru->arcs_list[ARC_BUFC_METADATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_mru->arcs_list[ARC_BUFC_DATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_mru_ghost->arcs_list[ARC_BUFC_METADATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_mru_ghost->arcs_list[ARC_BUFC_DATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_mfu->arcs_list[ARC_BUFC_METADATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_mfu->arcs_list[ARC_BUFC_DATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_mfu_ghost->arcs_list[ARC_BUFC_DATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_l2c_only->arcs_list[ARC_BUFC_METADATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); arc_l2c_only->arcs_list[ARC_BUFC_DATA] = multilist_create(sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l1hdr.b_arc_node), arc_state_multilist_index_func); refcount_create(&arc_anon->arcs_esize[ARC_BUFC_METADATA]); refcount_create(&arc_anon->arcs_esize[ARC_BUFC_DATA]); refcount_create(&arc_mru->arcs_esize[ARC_BUFC_METADATA]); refcount_create(&arc_mru->arcs_esize[ARC_BUFC_DATA]); refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]); refcount_create(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]); refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]); refcount_create(&arc_mfu->arcs_esize[ARC_BUFC_DATA]); refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]); refcount_create(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]); refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]); refcount_create(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]); refcount_create(&arc_anon->arcs_size); refcount_create(&arc_mru->arcs_size); refcount_create(&arc_mru_ghost->arcs_size); refcount_create(&arc_mfu->arcs_size); refcount_create(&arc_mfu_ghost->arcs_size); refcount_create(&arc_l2c_only->arcs_size); aggsum_init(&arc_meta_used, 0); aggsum_init(&arc_size, 0); aggsum_init(&astat_data_size, 0); aggsum_init(&astat_metadata_size, 0); aggsum_init(&astat_hdr_size, 0); aggsum_init(&astat_l2_hdr_size, 0); aggsum_init(&astat_bonus_size, 0); aggsum_init(&astat_dnode_size, 0); aggsum_init(&astat_dbuf_size, 0); arc_anon->arcs_state = ARC_STATE_ANON; arc_mru->arcs_state = ARC_STATE_MRU; arc_mru_ghost->arcs_state = ARC_STATE_MRU_GHOST; arc_mfu->arcs_state = ARC_STATE_MFU; arc_mfu_ghost->arcs_state = ARC_STATE_MFU_GHOST; arc_l2c_only->arcs_state = ARC_STATE_L2C_ONLY; } static void arc_state_fini(void) { refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_METADATA]); refcount_destroy(&arc_anon->arcs_esize[ARC_BUFC_DATA]); refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_METADATA]); refcount_destroy(&arc_mru->arcs_esize[ARC_BUFC_DATA]); refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_METADATA]); refcount_destroy(&arc_mru_ghost->arcs_esize[ARC_BUFC_DATA]); refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_METADATA]); refcount_destroy(&arc_mfu->arcs_esize[ARC_BUFC_DATA]); refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_METADATA]); refcount_destroy(&arc_mfu_ghost->arcs_esize[ARC_BUFC_DATA]); refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_METADATA]); refcount_destroy(&arc_l2c_only->arcs_esize[ARC_BUFC_DATA]); refcount_destroy(&arc_anon->arcs_size); refcount_destroy(&arc_mru->arcs_size); refcount_destroy(&arc_mru_ghost->arcs_size); refcount_destroy(&arc_mfu->arcs_size); refcount_destroy(&arc_mfu_ghost->arcs_size); refcount_destroy(&arc_l2c_only->arcs_size); multilist_destroy(arc_mru->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(arc_mru_ghost->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(arc_mfu->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(arc_mfu_ghost->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(arc_mru->arcs_list[ARC_BUFC_DATA]); multilist_destroy(arc_mru_ghost->arcs_list[ARC_BUFC_DATA]); multilist_destroy(arc_mfu->arcs_list[ARC_BUFC_DATA]); multilist_destroy(arc_mfu_ghost->arcs_list[ARC_BUFC_DATA]); multilist_destroy(arc_l2c_only->arcs_list[ARC_BUFC_METADATA]); multilist_destroy(arc_l2c_only->arcs_list[ARC_BUFC_DATA]); aggsum_fini(&arc_meta_used); aggsum_fini(&arc_size); aggsum_fini(&astat_data_size); aggsum_fini(&astat_metadata_size); aggsum_fini(&astat_hdr_size); aggsum_fini(&astat_l2_hdr_size); aggsum_fini(&astat_bonus_size); aggsum_fini(&astat_dnode_size); aggsum_fini(&astat_dbuf_size); } uint64_t arc_target_bytes(void) { return (arc_c); } void arc_init(void) { uint64_t percent, allmem = arc_all_memory(); mutex_init(&arc_reclaim_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&arc_reclaim_thread_cv, NULL, CV_DEFAULT, NULL); cv_init(&arc_reclaim_waiters_cv, NULL, CV_DEFAULT, NULL); arc_min_prefetch_ms = 1000; arc_min_prescient_prefetch_ms = 6000; #ifdef _KERNEL /* * Register a shrinker to support synchronous (direct) memory * reclaim from the arc. This is done to prevent kswapd from * swapping out pages when it is preferable to shrink the arc. */ spl_register_shrinker(&arc_shrinker); /* Set to 1/64 of all memory or a minimum of 512K */ arc_sys_free = MAX(allmem / 64, (512 * 1024)); arc_need_free = 0; #endif /* Set max to 1/2 of all memory */ arc_c_max = allmem / 2; #ifdef _KERNEL /* Set min cache to 1/32 of all memory, or 32MB, whichever is more */ arc_c_min = MAX(allmem / 32, 2ULL << SPA_MAXBLOCKSHIFT); #else /* * In userland, there's only the memory pressure that we artificially * create (see arc_available_memory()). Don't let arc_c get too * small, because it can cause transactions to be larger than * arc_c, causing arc_tempreserve_space() to fail. */ arc_c_min = MAX(arc_c_max / 2, 2ULL << SPA_MAXBLOCKSHIFT); #endif arc_c = arc_c_max; arc_p = (arc_c >> 1); /* Set min to 1/2 of arc_c_min */ arc_meta_min = 1ULL << SPA_MAXBLOCKSHIFT; /* Initialize maximum observed usage to zero */ arc_meta_max = 0; /* * Set arc_meta_limit to a percent of arc_c_max with a floor of * arc_meta_min, and a ceiling of arc_c_max. */ percent = MIN(zfs_arc_meta_limit_percent, 100); arc_meta_limit = MAX(arc_meta_min, (percent * arc_c_max) / 100); percent = MIN(zfs_arc_dnode_limit_percent, 100); arc_dnode_limit = (percent * arc_meta_limit) / 100; /* Apply user specified tunings */ arc_tuning_update(); /* if kmem_flags are set, lets try to use less memory */ if (kmem_debugging()) arc_c = arc_c / 2; if (arc_c < arc_c_min) arc_c = arc_c_min; arc_state_init(); buf_init(); list_create(&arc_prune_list, sizeof (arc_prune_t), offsetof(arc_prune_t, p_node)); mutex_init(&arc_prune_mtx, NULL, MUTEX_DEFAULT, NULL); arc_prune_taskq = taskq_create("arc_prune", max_ncpus, defclsyspri, max_ncpus, INT_MAX, TASKQ_PREPOPULATE | TASKQ_DYNAMIC); arc_reclaim_thread_exit = B_FALSE; arc_ksp = kstat_create("zfs", 0, "arcstats", "misc", KSTAT_TYPE_NAMED, sizeof (arc_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (arc_ksp != NULL) { arc_ksp->ks_data = &arc_stats; arc_ksp->ks_update = arc_kstat_update; kstat_install(arc_ksp); } (void) thread_create(NULL, 0, arc_reclaim_thread, NULL, 0, &p0, TS_RUN, defclsyspri); arc_dead = B_FALSE; arc_warm = B_FALSE; /* * Calculate maximum amount of dirty data per pool. * * If it has been set by a module parameter, take that. * Otherwise, use a percentage of physical memory defined by * zfs_dirty_data_max_percent (default 10%) with a cap at * zfs_dirty_data_max_max (default 4G or 25% of physical memory). */ if (zfs_dirty_data_max_max == 0) zfs_dirty_data_max_max = MIN(4ULL * 1024 * 1024 * 1024, allmem * zfs_dirty_data_max_max_percent / 100); if (zfs_dirty_data_max == 0) { zfs_dirty_data_max = allmem * zfs_dirty_data_max_percent / 100; zfs_dirty_data_max = MIN(zfs_dirty_data_max, zfs_dirty_data_max_max); } } void arc_fini(void) { arc_prune_t *p; #ifdef _KERNEL spl_unregister_shrinker(&arc_shrinker); #endif /* _KERNEL */ mutex_enter(&arc_reclaim_lock); arc_reclaim_thread_exit = B_TRUE; /* * The reclaim thread will set arc_reclaim_thread_exit back to * B_FALSE when it is finished exiting; we're waiting for that. */ while (arc_reclaim_thread_exit) { cv_signal(&arc_reclaim_thread_cv); cv_wait(&arc_reclaim_thread_cv, &arc_reclaim_lock); } mutex_exit(&arc_reclaim_lock); /* Use B_TRUE to ensure *all* buffers are evicted */ arc_flush(NULL, B_TRUE); arc_dead = B_TRUE; if (arc_ksp != NULL) { kstat_delete(arc_ksp); arc_ksp = NULL; } taskq_wait(arc_prune_taskq); taskq_destroy(arc_prune_taskq); mutex_enter(&arc_prune_mtx); while ((p = list_head(&arc_prune_list)) != NULL) { list_remove(&arc_prune_list, p); refcount_remove(&p->p_refcnt, &arc_prune_list); refcount_destroy(&p->p_refcnt); kmem_free(p, sizeof (*p)); } mutex_exit(&arc_prune_mtx); list_destroy(&arc_prune_list); mutex_destroy(&arc_prune_mtx); mutex_destroy(&arc_reclaim_lock); cv_destroy(&arc_reclaim_thread_cv); cv_destroy(&arc_reclaim_waiters_cv); arc_state_fini(); buf_fini(); ASSERT0(arc_loaned_bytes); } /* * Level 2 ARC * * The level 2 ARC (L2ARC) is a cache layer in-between main memory and disk. * It uses dedicated storage devices to hold cached data, which are populated * using large infrequent writes. The main role of this cache is to boost * the performance of random read workloads. The intended L2ARC devices * include short-stroked disks, solid state disks, and other media with * substantially faster read latency than disk. * * +-----------------------+ * | ARC | * +-----------------------+ * | ^ ^ * | | | * l2arc_feed_thread() arc_read() * | | | * | l2arc read | * V | | * +---------------+ | * | L2ARC | | * +---------------+ | * | ^ | * l2arc_write() | | * | | | * V | | * +-------+ +-------+ * | vdev | | vdev | * | cache | | cache | * +-------+ +-------+ * +=========+ .-----. * : L2ARC : |-_____-| * : devices : | Disks | * +=========+ `-_____-' * * Read requests are satisfied from the following sources, in order: * * 1) ARC * 2) vdev cache of L2ARC devices * 3) L2ARC devices * 4) vdev cache of disks * 5) disks * * Some L2ARC device types exhibit extremely slow write performance. * To accommodate for this there are some significant differences between * the L2ARC and traditional cache design: * * 1. There is no eviction path from the ARC to the L2ARC. Evictions from * the ARC behave as usual, freeing buffers and placing headers on ghost * lists. The ARC does not send buffers to the L2ARC during eviction as * this would add inflated write latencies for all ARC memory pressure. * * 2. The L2ARC attempts to cache data from the ARC before it is evicted. * It does this by periodically scanning buffers from the eviction-end of * the MFU and MRU ARC lists, copying them to the L2ARC devices if they are * not already there. It scans until a headroom of buffers is satisfied, * which itself is a buffer for ARC eviction. If a compressible buffer is * found during scanning and selected for writing to an L2ARC device, we * temporarily boost scanning headroom during the next scan cycle to make * sure we adapt to compression effects (which might significantly reduce * the data volume we write to L2ARC). The thread that does this is * l2arc_feed_thread(), illustrated below; example sizes are included to * provide a better sense of ratio than this diagram: * * head --> tail * +---------------------+----------+ * ARC_mfu |:::::#:::::::::::::::|o#o###o###|-->. # already on L2ARC * +---------------------+----------+ | o L2ARC eligible * ARC_mru |:#:::::::::::::::::::|#o#ooo####|-->| : ARC buffer * +---------------------+----------+ | * 15.9 Gbytes ^ 32 Mbytes | * headroom | * l2arc_feed_thread() * | * l2arc write hand <--[oooo]--' * | 8 Mbyte * | write max * V * +==============================+ * L2ARC dev |####|#|###|###| |####| ... | * +==============================+ * 32 Gbytes * * 3. If an ARC buffer is copied to the L2ARC but then hit instead of * evicted, then the L2ARC has cached a buffer much sooner than it probably * needed to, potentially wasting L2ARC device bandwidth and storage. It is * safe to say that this is an uncommon case, since buffers at the end of * the ARC lists have moved there due to inactivity. * * 4. If the ARC evicts faster than the L2ARC can maintain a headroom, * then the L2ARC simply misses copying some buffers. This serves as a * pressure valve to prevent heavy read workloads from both stalling the ARC * with waits and clogging the L2ARC with writes. This also helps prevent * the potential for the L2ARC to churn if it attempts to cache content too * quickly, such as during backups of the entire pool. * * 5. After system boot and before the ARC has filled main memory, there are * no evictions from the ARC and so the tails of the ARC_mfu and ARC_mru * lists can remain mostly static. Instead of searching from tail of these * lists as pictured, the l2arc_feed_thread() will search from the list heads * for eligible buffers, greatly increasing its chance of finding them. * * The L2ARC device write speed is also boosted during this time so that * the L2ARC warms up faster. Since there have been no ARC evictions yet, * there are no L2ARC reads, and no fear of degrading read performance * through increased writes. * * 6. Writes to the L2ARC devices are grouped and sent in-sequence, so that * the vdev queue can aggregate them into larger and fewer writes. Each * device is written to in a rotor fashion, sweeping writes through * available space then repeating. * * 7. The L2ARC does not store dirty content. It never needs to flush * write buffers back to disk based storage. * * 8. If an ARC buffer is written (and dirtied) which also exists in the * L2ARC, the now stale L2ARC buffer is immediately dropped. * * The performance of the L2ARC can be tweaked by a number of tunables, which * may be necessary for different workloads: * * l2arc_write_max max write bytes per interval * l2arc_write_boost extra write bytes during device warmup * l2arc_noprefetch skip caching prefetched buffers * l2arc_headroom number of max device writes to precache * l2arc_headroom_boost when we find compressed buffers during ARC * scanning, we multiply headroom by this * percentage factor for the next scan cycle, * since more compressed buffers are likely to * be present * l2arc_feed_secs seconds between L2ARC writing * * Tunables may be removed or added as future performance improvements are * integrated, and also may become zpool properties. * * There are three key functions that control how the L2ARC warms up: * * l2arc_write_eligible() check if a buffer is eligible to cache * l2arc_write_size() calculate how much to write * l2arc_write_interval() calculate sleep delay between writes * * These three functions determine what to write, how much, and how quickly * to send writes. */ static boolean_t l2arc_write_eligible(uint64_t spa_guid, arc_buf_hdr_t *hdr) { /* * A buffer is *not* eligible for the L2ARC if it: * 1. belongs to a different spa. * 2. is already cached on the L2ARC. * 3. has an I/O in progress (it may be an incomplete read). * 4. is flagged not eligible (zfs property). */ if (hdr->b_spa != spa_guid || HDR_HAS_L2HDR(hdr) || HDR_IO_IN_PROGRESS(hdr) || !HDR_L2CACHE(hdr)) return (B_FALSE); return (B_TRUE); } static uint64_t l2arc_write_size(void) { uint64_t size; /* * Make sure our globals have meaningful values in case the user * altered them. */ size = l2arc_write_max; if (size == 0) { cmn_err(CE_NOTE, "Bad value for l2arc_write_max, value must " "be greater than zero, resetting it to the default (%d)", L2ARC_WRITE_SIZE); size = l2arc_write_max = L2ARC_WRITE_SIZE; } if (arc_warm == B_FALSE) size += l2arc_write_boost; return (size); } static clock_t l2arc_write_interval(clock_t began, uint64_t wanted, uint64_t wrote) { clock_t interval, next, now; /* * If the ARC lists are busy, increase our write rate; if the * lists are stale, idle back. This is achieved by checking * how much we previously wrote - if it was more than half of * what we wanted, schedule the next write much sooner. */ if (l2arc_feed_again && wrote > (wanted / 2)) interval = (hz * l2arc_feed_min_ms) / 1000; else interval = hz * l2arc_feed_secs; now = ddi_get_lbolt(); next = MAX(now, MIN(now + interval, began + interval)); return (next); } /* * Cycle through L2ARC devices. This is how L2ARC load balances. * If a device is returned, this also returns holding the spa config lock. */ static l2arc_dev_t * l2arc_dev_get_next(void) { l2arc_dev_t *first, *next = NULL; /* * Lock out the removal of spas (spa_namespace_lock), then removal * of cache devices (l2arc_dev_mtx). Once a device has been selected, * both locks will be dropped and a spa config lock held instead. */ mutex_enter(&spa_namespace_lock); mutex_enter(&l2arc_dev_mtx); /* if there are no vdevs, there is nothing to do */ if (l2arc_ndev == 0) goto out; first = NULL; next = l2arc_dev_last; do { /* loop around the list looking for a non-faulted vdev */ if (next == NULL) { next = list_head(l2arc_dev_list); } else { next = list_next(l2arc_dev_list, next); if (next == NULL) next = list_head(l2arc_dev_list); } /* if we have come back to the start, bail out */ if (first == NULL) first = next; else if (next == first) break; } while (vdev_is_dead(next->l2ad_vdev)); /* if we were unable to find any usable vdevs, return NULL */ if (vdev_is_dead(next->l2ad_vdev)) next = NULL; l2arc_dev_last = next; out: mutex_exit(&l2arc_dev_mtx); /* * Grab the config lock to prevent the 'next' device from being * removed while we are writing to it. */ if (next != NULL) spa_config_enter(next->l2ad_spa, SCL_L2ARC, next, RW_READER); mutex_exit(&spa_namespace_lock); return (next); } /* * Free buffers that were tagged for destruction. */ static void l2arc_do_free_on_write(void) { list_t *buflist; l2arc_data_free_t *df, *df_prev; mutex_enter(&l2arc_free_on_write_mtx); buflist = l2arc_free_on_write; for (df = list_tail(buflist); df; df = df_prev) { df_prev = list_prev(buflist, df); ASSERT3P(df->l2df_abd, !=, NULL); abd_free(df->l2df_abd); list_remove(buflist, df); kmem_free(df, sizeof (l2arc_data_free_t)); } mutex_exit(&l2arc_free_on_write_mtx); } /* * A write to a cache device has completed. Update all headers to allow * reads from these buffers to begin. */ static void l2arc_write_done(zio_t *zio) { l2arc_write_callback_t *cb; l2arc_dev_t *dev; list_t *buflist; arc_buf_hdr_t *head, *hdr, *hdr_prev; kmutex_t *hash_lock; int64_t bytes_dropped = 0; cb = zio->io_private; ASSERT3P(cb, !=, NULL); dev = cb->l2wcb_dev; ASSERT3P(dev, !=, NULL); head = cb->l2wcb_head; ASSERT3P(head, !=, NULL); buflist = &dev->l2ad_buflist; ASSERT3P(buflist, !=, NULL); DTRACE_PROBE2(l2arc__iodone, zio_t *, zio, l2arc_write_callback_t *, cb); if (zio->io_error != 0) ARCSTAT_BUMP(arcstat_l2_writes_error); /* * All writes completed, or an error was hit. */ top: mutex_enter(&dev->l2ad_mtx); for (hdr = list_prev(buflist, head); hdr; hdr = hdr_prev) { hdr_prev = list_prev(buflist, hdr); hash_lock = HDR_LOCK(hdr); /* * We cannot use mutex_enter or else we can deadlock * with l2arc_write_buffers (due to swapping the order * the hash lock and l2ad_mtx are taken). */ if (!mutex_tryenter(hash_lock)) { /* * Missed the hash lock. We must retry so we * don't leave the ARC_FLAG_L2_WRITING bit set. */ ARCSTAT_BUMP(arcstat_l2_writes_lock_retry); /* * We don't want to rescan the headers we've * already marked as having been written out, so * we reinsert the head node so we can pick up * where we left off. */ list_remove(buflist, head); list_insert_after(buflist, hdr, head); mutex_exit(&dev->l2ad_mtx); /* * We wait for the hash lock to become available * to try and prevent busy waiting, and increase * the chance we'll be able to acquire the lock * the next time around. */ mutex_enter(hash_lock); mutex_exit(hash_lock); goto top; } /* * We could not have been moved into the arc_l2c_only * state while in-flight due to our ARC_FLAG_L2_WRITING * bit being set. Let's just ensure that's being enforced. */ ASSERT(HDR_HAS_L1HDR(hdr)); /* * Skipped - drop L2ARC entry and mark the header as no * longer L2 eligibile. */ if (zio->io_error != 0) { /* * Error - drop L2ARC entry. */ list_remove(buflist, hdr); arc_hdr_clear_flags(hdr, ARC_FLAG_HAS_L2HDR); ARCSTAT_INCR(arcstat_l2_psize, -arc_hdr_size(hdr)); ARCSTAT_INCR(arcstat_l2_lsize, -HDR_GET_LSIZE(hdr)); bytes_dropped += arc_hdr_size(hdr); (void) refcount_remove_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); } /* * Allow ARC to begin reads and ghost list evictions to * this L2ARC entry. */ arc_hdr_clear_flags(hdr, ARC_FLAG_L2_WRITING); mutex_exit(hash_lock); } atomic_inc_64(&l2arc_writes_done); list_remove(buflist, head); ASSERT(!HDR_HAS_L1HDR(head)); kmem_cache_free(hdr_l2only_cache, head); mutex_exit(&dev->l2ad_mtx); vdev_space_update(dev->l2ad_vdev, -bytes_dropped, 0, 0); l2arc_do_free_on_write(); kmem_free(cb, sizeof (l2arc_write_callback_t)); } static int l2arc_untransform(zio_t *zio, l2arc_read_callback_t *cb) { int ret; spa_t *spa = zio->io_spa; arc_buf_hdr_t *hdr = cb->l2rcb_hdr; blkptr_t *bp = zio->io_bp; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; boolean_t no_crypt = B_FALSE; /* * ZIL data is never be written to the L2ARC, so we don't need * special handling for its unique MAC storage. */ ASSERT3U(BP_GET_TYPE(bp), !=, DMU_OT_INTENT_LOG); ASSERT(MUTEX_HELD(HDR_LOCK(hdr))); ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); /* * If the data was encrypted, decrypt it now. Note that * we must check the bp here and not the hdr, since the * hdr does not have its encryption parameters updated * until arc_read_done(). */ if (BP_IS_ENCRYPTED(bp)) { abd_t *eabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr); zio_crypt_decode_params_bp(bp, salt, iv); zio_crypt_decode_mac_bp(bp, mac); ret = spa_do_crypt_abd(B_FALSE, spa, &cb->l2rcb_zb, BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, HDR_GET_PSIZE(hdr), eabd, hdr->b_l1hdr.b_pabd, &no_crypt); if (ret != 0) { arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr); goto error; } /* * If we actually performed decryption, replace b_pabd * with the decrypted data. Otherwise we can just throw * our decryption buffer away. */ if (!no_crypt) { arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, arc_hdr_size(hdr), hdr); hdr->b_l1hdr.b_pabd = eabd; zio->io_abd = eabd; } else { arc_free_data_abd(hdr, eabd, arc_hdr_size(hdr), hdr); } } /* * If the L2ARC block was compressed, but ARC compression * is disabled we decompress the data into a new buffer and * replace the existing data. */ if (HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { abd_t *cabd = arc_get_data_abd(hdr, arc_hdr_size(hdr), hdr); void *tmp = abd_borrow_buf(cabd, arc_hdr_size(hdr)); ret = zio_decompress_data(HDR_GET_COMPRESS(hdr), hdr->b_l1hdr.b_pabd, tmp, HDR_GET_PSIZE(hdr), HDR_GET_LSIZE(hdr)); if (ret != 0) { abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr)); arc_free_data_abd(hdr, cabd, arc_hdr_size(hdr), hdr); goto error; } abd_return_buf_copy(cabd, tmp, arc_hdr_size(hdr)); arc_free_data_abd(hdr, hdr->b_l1hdr.b_pabd, arc_hdr_size(hdr), hdr); hdr->b_l1hdr.b_pabd = cabd; zio->io_abd = cabd; zio->io_size = HDR_GET_LSIZE(hdr); } return (0); error: return (ret); } /* * A read to a cache device completed. Validate buffer contents before * handing over to the regular ARC routines. */ static void l2arc_read_done(zio_t *zio) { int tfm_error = 0; l2arc_read_callback_t *cb = zio->io_private; arc_buf_hdr_t *hdr; kmutex_t *hash_lock; boolean_t valid_cksum; boolean_t using_rdata = (BP_IS_ENCRYPTED(&cb->l2rcb_bp) && (cb->l2rcb_flags & ZIO_FLAG_RAW_ENCRYPT)); ASSERT3P(zio->io_vd, !=, NULL); ASSERT(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE); spa_config_exit(zio->io_spa, SCL_L2ARC, zio->io_vd); ASSERT3P(cb, !=, NULL); hdr = cb->l2rcb_hdr; ASSERT3P(hdr, !=, NULL); hash_lock = HDR_LOCK(hdr); mutex_enter(hash_lock); ASSERT3P(hash_lock, ==, HDR_LOCK(hdr)); /* * If the data was read into a temporary buffer, * move it and free the buffer. */ if (cb->l2rcb_abd != NULL) { ASSERT3U(arc_hdr_size(hdr), <, zio->io_size); if (zio->io_error == 0) { if (using_rdata) { abd_copy(hdr->b_crypt_hdr.b_rabd, cb->l2rcb_abd, arc_hdr_size(hdr)); } else { abd_copy(hdr->b_l1hdr.b_pabd, cb->l2rcb_abd, arc_hdr_size(hdr)); } } /* * The following must be done regardless of whether * there was an error: * - free the temporary buffer * - point zio to the real ARC buffer * - set zio size accordingly * These are required because zio is either re-used for * an I/O of the block in the case of the error * or the zio is passed to arc_read_done() and it * needs real data. */ abd_free(cb->l2rcb_abd); zio->io_size = zio->io_orig_size = arc_hdr_size(hdr); if (using_rdata) { ASSERT(HDR_HAS_RABD(hdr)); zio->io_abd = zio->io_orig_abd = hdr->b_crypt_hdr.b_rabd; } else { ASSERT3P(hdr->b_l1hdr.b_pabd, !=, NULL); zio->io_abd = zio->io_orig_abd = hdr->b_l1hdr.b_pabd; } } ASSERT3P(zio->io_abd, !=, NULL); /* * Check this survived the L2ARC journey. */ ASSERT(zio->io_abd == hdr->b_l1hdr.b_pabd || (HDR_HAS_RABD(hdr) && zio->io_abd == hdr->b_crypt_hdr.b_rabd)); zio->io_bp_copy = cb->l2rcb_bp; /* XXX fix in L2ARC 2.0 */ zio->io_bp = &zio->io_bp_copy; /* XXX fix in L2ARC 2.0 */ valid_cksum = arc_cksum_is_equal(hdr, zio); /* * b_rabd will always match the data as it exists on disk if it is * being used. Therefore if we are reading into b_rabd we do not * attempt to untransform the data. */ if (valid_cksum && !using_rdata) tfm_error = l2arc_untransform(zio, cb); if (valid_cksum && tfm_error == 0 && zio->io_error == 0 && !HDR_L2_EVICTED(hdr)) { mutex_exit(hash_lock); zio->io_private = hdr; arc_read_done(zio); } else { mutex_exit(hash_lock); /* * Buffer didn't survive caching. Increment stats and * reissue to the original storage device. */ if (zio->io_error != 0) { ARCSTAT_BUMP(arcstat_l2_io_error); } else { zio->io_error = SET_ERROR(EIO); } if (!valid_cksum || tfm_error != 0) ARCSTAT_BUMP(arcstat_l2_cksum_bad); /* * If there's no waiter, issue an async i/o to the primary * storage now. If there *is* a waiter, the caller must * issue the i/o in a context where it's OK to block. */ if (zio->io_waiter == NULL) { zio_t *pio = zio_unique_parent(zio); void *abd = (using_rdata) ? hdr->b_crypt_hdr.b_rabd : hdr->b_l1hdr.b_pabd; ASSERT(!pio || pio->io_child_type == ZIO_CHILD_LOGICAL); zio_nowait(zio_read(pio, zio->io_spa, zio->io_bp, abd, zio->io_size, arc_read_done, hdr, zio->io_priority, cb->l2rcb_flags, &cb->l2rcb_zb)); } } kmem_free(cb, sizeof (l2arc_read_callback_t)); } /* * This is the list priority from which the L2ARC will search for pages to * cache. This is used within loops (0..3) to cycle through lists in the * desired order. This order can have a significant effect on cache * performance. * * Currently the metadata lists are hit first, MFU then MRU, followed by * the data lists. This function returns a locked list, and also returns * the lock pointer. */ static multilist_sublist_t * l2arc_sublist_lock(int list_num) { multilist_t *ml = NULL; unsigned int idx; ASSERT(list_num >= 0 && list_num < L2ARC_FEED_TYPES); switch (list_num) { case 0: ml = arc_mfu->arcs_list[ARC_BUFC_METADATA]; break; case 1: ml = arc_mru->arcs_list[ARC_BUFC_METADATA]; break; case 2: ml = arc_mfu->arcs_list[ARC_BUFC_DATA]; break; case 3: ml = arc_mru->arcs_list[ARC_BUFC_DATA]; break; default: return (NULL); } /* * Return a randomly-selected sublist. This is acceptable * because the caller feeds only a little bit of data for each * call (8MB). Subsequent calls will result in different * sublists being selected. */ idx = multilist_get_random_index(ml); return (multilist_sublist_lock(ml, idx)); } /* * Evict buffers from the device write hand to the distance specified in * bytes. This distance may span populated buffers, it may span nothing. * This is clearing a region on the L2ARC device ready for writing. * If the 'all' boolean is set, every buffer is evicted. */ static void l2arc_evict(l2arc_dev_t *dev, uint64_t distance, boolean_t all) { list_t *buflist; arc_buf_hdr_t *hdr, *hdr_prev; kmutex_t *hash_lock; uint64_t taddr; buflist = &dev->l2ad_buflist; if (!all && dev->l2ad_first) { /* * This is the first sweep through the device. There is * nothing to evict. */ return; } if (dev->l2ad_hand >= (dev->l2ad_end - (2 * distance))) { /* * When nearing the end of the device, evict to the end * before the device write hand jumps to the start. */ taddr = dev->l2ad_end; } else { taddr = dev->l2ad_hand + distance; } DTRACE_PROBE4(l2arc__evict, l2arc_dev_t *, dev, list_t *, buflist, uint64_t, taddr, boolean_t, all); top: mutex_enter(&dev->l2ad_mtx); for (hdr = list_tail(buflist); hdr; hdr = hdr_prev) { hdr_prev = list_prev(buflist, hdr); hash_lock = HDR_LOCK(hdr); /* * We cannot use mutex_enter or else we can deadlock * with l2arc_write_buffers (due to swapping the order * the hash lock and l2ad_mtx are taken). */ if (!mutex_tryenter(hash_lock)) { /* * Missed the hash lock. Retry. */ ARCSTAT_BUMP(arcstat_l2_evict_lock_retry); mutex_exit(&dev->l2ad_mtx); mutex_enter(hash_lock); mutex_exit(hash_lock); goto top; } /* * A header can't be on this list if it doesn't have L2 header. */ ASSERT(HDR_HAS_L2HDR(hdr)); /* Ensure this header has finished being written. */ ASSERT(!HDR_L2_WRITING(hdr)); ASSERT(!HDR_L2_WRITE_HEAD(hdr)); if (!all && (hdr->b_l2hdr.b_daddr >= taddr || hdr->b_l2hdr.b_daddr < dev->l2ad_hand)) { /* * We've evicted to the target address, * or the end of the device. */ mutex_exit(hash_lock); break; } if (!HDR_HAS_L1HDR(hdr)) { ASSERT(!HDR_L2_READING(hdr)); /* * This doesn't exist in the ARC. Destroy. * arc_hdr_destroy() will call list_remove() * and decrement arcstat_l2_lsize. */ arc_change_state(arc_anon, hdr, hash_lock); arc_hdr_destroy(hdr); } else { ASSERT(hdr->b_l1hdr.b_state != arc_l2c_only); ARCSTAT_BUMP(arcstat_l2_evict_l1cached); /* * Invalidate issued or about to be issued * reads, since we may be about to write * over this location. */ if (HDR_L2_READING(hdr)) { ARCSTAT_BUMP(arcstat_l2_evict_reading); arc_hdr_set_flags(hdr, ARC_FLAG_L2_EVICTED); } arc_hdr_l2hdr_destroy(hdr); } mutex_exit(hash_lock); } mutex_exit(&dev->l2ad_mtx); } /* * Handle any abd transforms that might be required for writing to the L2ARC. * If successful, this function will always return an abd with the data * transformed as it is on disk in a new abd of asize bytes. */ static int l2arc_apply_transforms(spa_t *spa, arc_buf_hdr_t *hdr, uint64_t asize, abd_t **abd_out) { int ret; void *tmp = NULL; abd_t *cabd = NULL, *eabd = NULL, *to_write = hdr->b_l1hdr.b_pabd; enum zio_compress compress = HDR_GET_COMPRESS(hdr); uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t size = arc_hdr_size(hdr); boolean_t ismd = HDR_ISTYPE_METADATA(hdr); boolean_t bswap = (hdr->b_l1hdr.b_byteswap != DMU_BSWAP_NUMFUNCS); dsl_crypto_key_t *dck = NULL; uint8_t mac[ZIO_DATA_MAC_LEN] = { 0 }; boolean_t no_crypt = B_FALSE; ASSERT((HDR_GET_COMPRESS(hdr) != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) || HDR_ENCRYPTED(hdr) || HDR_SHARED_DATA(hdr) || psize != asize); ASSERT3U(psize, <=, asize); /* * If this data simply needs its own buffer, we simply allocate it * and copy the data. This may be done to elimiate a depedency on a * shared buffer or to reallocate the buffer to match asize. */ if (HDR_HAS_RABD(hdr) && asize != psize) { ASSERT3U(asize, >=, psize); to_write = abd_alloc_for_io(asize, ismd); abd_copy(to_write, hdr->b_crypt_hdr.b_rabd, psize); if (psize != asize) abd_zero_off(to_write, psize, asize - psize); goto out; } if ((compress == ZIO_COMPRESS_OFF || HDR_COMPRESSION_ENABLED(hdr)) && !HDR_ENCRYPTED(hdr)) { ASSERT3U(size, ==, psize); to_write = abd_alloc_for_io(asize, ismd); abd_copy(to_write, hdr->b_l1hdr.b_pabd, size); if (size != asize) abd_zero_off(to_write, size, asize - size); goto out; } if (compress != ZIO_COMPRESS_OFF && !HDR_COMPRESSION_ENABLED(hdr)) { cabd = abd_alloc_for_io(asize, ismd); tmp = abd_borrow_buf(cabd, asize); psize = zio_compress_data(compress, to_write, tmp, size); ASSERT3U(psize, <=, HDR_GET_PSIZE(hdr)); if (psize < asize) bzero((char *)tmp + psize, asize - psize); psize = HDR_GET_PSIZE(hdr); abd_return_buf_copy(cabd, tmp, asize); to_write = cabd; } if (HDR_ENCRYPTED(hdr)) { eabd = abd_alloc_for_io(asize, ismd); /* * If the dataset was disowned before the buffer * made it to this point, the key to re-encrypt * it won't be available. In this case we simply * won't write the buffer to the L2ARC. */ ret = spa_keystore_lookup_key(spa, hdr->b_crypt_hdr.b_dsobj, FTAG, &dck); if (ret != 0) goto error; ret = zio_do_crypt_abd(B_TRUE, &dck->dck_key, hdr->b_crypt_hdr.b_ot, bswap, hdr->b_crypt_hdr.b_salt, hdr->b_crypt_hdr.b_iv, mac, psize, to_write, eabd, &no_crypt); if (ret != 0) goto error; if (no_crypt) abd_copy(eabd, to_write, psize); if (psize != asize) abd_zero_off(eabd, psize, asize - psize); /* assert that the MAC we got here matches the one we saved */ ASSERT0(bcmp(mac, hdr->b_crypt_hdr.b_mac, ZIO_DATA_MAC_LEN)); spa_keystore_dsl_key_rele(spa, dck, FTAG); if (to_write == cabd) abd_free(cabd); to_write = eabd; } out: ASSERT3P(to_write, !=, hdr->b_l1hdr.b_pabd); *abd_out = to_write; return (0); error: if (dck != NULL) spa_keystore_dsl_key_rele(spa, dck, FTAG); if (cabd != NULL) abd_free(cabd); if (eabd != NULL) abd_free(eabd); *abd_out = NULL; return (ret); } /* * Find and write ARC buffers to the L2ARC device. * * An ARC_FLAG_L2_WRITING flag is set so that the L2ARC buffers are not valid * for reading until they have completed writing. * The headroom_boost is an in-out parameter used to maintain headroom boost * state between calls to this function. * * Returns the number of bytes actually written (which may be smaller than * the delta by which the device hand has changed due to alignment). */ static uint64_t l2arc_write_buffers(spa_t *spa, l2arc_dev_t *dev, uint64_t target_sz) { arc_buf_hdr_t *hdr, *hdr_prev, *head; uint64_t write_asize, write_psize, write_lsize, headroom; boolean_t full; l2arc_write_callback_t *cb; zio_t *pio, *wzio; uint64_t guid = spa_load_guid(spa); ASSERT3P(dev->l2ad_vdev, !=, NULL); pio = NULL; write_lsize = write_asize = write_psize = 0; full = B_FALSE; head = kmem_cache_alloc(hdr_l2only_cache, KM_PUSHPAGE); arc_hdr_set_flags(head, ARC_FLAG_L2_WRITE_HEAD | ARC_FLAG_HAS_L2HDR); /* * Copy buffers for L2ARC writing. */ for (int try = 0; try < L2ARC_FEED_TYPES; try++) { multilist_sublist_t *mls = l2arc_sublist_lock(try); uint64_t passed_sz = 0; VERIFY3P(mls, !=, NULL); /* * L2ARC fast warmup. * * Until the ARC is warm and starts to evict, read from the * head of the ARC lists rather than the tail. */ if (arc_warm == B_FALSE) hdr = multilist_sublist_head(mls); else hdr = multilist_sublist_tail(mls); headroom = target_sz * l2arc_headroom; if (zfs_compressed_arc_enabled) headroom = (headroom * l2arc_headroom_boost) / 100; for (; hdr; hdr = hdr_prev) { kmutex_t *hash_lock; abd_t *to_write = NULL; if (arc_warm == B_FALSE) hdr_prev = multilist_sublist_next(mls, hdr); else hdr_prev = multilist_sublist_prev(mls, hdr); hash_lock = HDR_LOCK(hdr); if (!mutex_tryenter(hash_lock)) { /* * Skip this buffer rather than waiting. */ continue; } passed_sz += HDR_GET_LSIZE(hdr); if (passed_sz > headroom) { /* * Searched too far. */ mutex_exit(hash_lock); break; } if (!l2arc_write_eligible(guid, hdr)) { mutex_exit(hash_lock); continue; } /* * We rely on the L1 portion of the header below, so * it's invalid for this header to have been evicted out * of the ghost cache, prior to being written out. The * ARC_FLAG_L2_WRITING bit ensures this won't happen. */ ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3U(HDR_GET_PSIZE(hdr), >, 0); ASSERT3U(arc_hdr_size(hdr), >, 0); ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); uint64_t psize = HDR_GET_PSIZE(hdr); uint64_t asize = vdev_psize_to_asize(dev->l2ad_vdev, psize); if ((write_asize + asize) > target_sz) { full = B_TRUE; mutex_exit(hash_lock); break; } /* * We rely on the L1 portion of the header below, so * it's invalid for this header to have been evicted out * of the ghost cache, prior to being written out. The * ARC_FLAG_L2_WRITING bit ensures this won't happen. */ arc_hdr_set_flags(hdr, ARC_FLAG_L2_WRITING); ASSERT(HDR_HAS_L1HDR(hdr)); ASSERT3U(HDR_GET_PSIZE(hdr), >, 0); ASSERT(hdr->b_l1hdr.b_pabd != NULL || HDR_HAS_RABD(hdr)); ASSERT3U(arc_hdr_size(hdr), >, 0); /* * If this header has b_rabd, we can use this since it * must always match the data exactly as it exists on * disk. Otherwise, the L2ARC can normally use the * hdr's data, but if we're sharing data between the * hdr and one of its bufs, L2ARC needs its own copy of * the data so that the ZIO below can't race with the * buf consumer. To ensure that this copy will be * available for the lifetime of the ZIO and be cleaned * up afterwards, we add it to the l2arc_free_on_write * queue. If we need to apply any transforms to the * data (compression, encryption) we will also need the * extra buffer. */ if (HDR_HAS_RABD(hdr) && psize == asize) { to_write = hdr->b_crypt_hdr.b_rabd; } else if ((HDR_COMPRESSION_ENABLED(hdr) || HDR_GET_COMPRESS(hdr) == ZIO_COMPRESS_OFF) && !HDR_ENCRYPTED(hdr) && !HDR_SHARED_DATA(hdr) && psize == asize) { to_write = hdr->b_l1hdr.b_pabd; } else { int ret; arc_buf_contents_t type = arc_buf_type(hdr); ret = l2arc_apply_transforms(spa, hdr, asize, &to_write); if (ret != 0) { arc_hdr_clear_flags(hdr, ARC_FLAG_L2_WRITING); mutex_exit(hash_lock); continue; } l2arc_free_abd_on_write(to_write, asize, type); } if (pio == NULL) { /* * Insert a dummy header on the buflist so * l2arc_write_done() can find where the * write buffers begin without searching. */ mutex_enter(&dev->l2ad_mtx); list_insert_head(&dev->l2ad_buflist, head); mutex_exit(&dev->l2ad_mtx); cb = kmem_alloc( sizeof (l2arc_write_callback_t), KM_SLEEP); cb->l2wcb_dev = dev; cb->l2wcb_head = head; pio = zio_root(spa, l2arc_write_done, cb, ZIO_FLAG_CANFAIL); } hdr->b_l2hdr.b_dev = dev; hdr->b_l2hdr.b_hits = 0; hdr->b_l2hdr.b_daddr = dev->l2ad_hand; arc_hdr_set_flags(hdr, ARC_FLAG_HAS_L2HDR); mutex_enter(&dev->l2ad_mtx); list_insert_head(&dev->l2ad_buflist, hdr); mutex_exit(&dev->l2ad_mtx); (void) refcount_add_many(&dev->l2ad_alloc, arc_hdr_size(hdr), hdr); wzio = zio_write_phys(pio, dev->l2ad_vdev, hdr->b_l2hdr.b_daddr, asize, to_write, ZIO_CHECKSUM_OFF, NULL, hdr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_CANFAIL, B_FALSE); write_lsize += HDR_GET_LSIZE(hdr); DTRACE_PROBE2(l2arc__write, vdev_t *, dev->l2ad_vdev, zio_t *, wzio); write_psize += psize; write_asize += asize; dev->l2ad_hand += asize; mutex_exit(hash_lock); (void) zio_nowait(wzio); } multilist_sublist_unlock(mls); if (full == B_TRUE) break; } /* No buffers selected for writing? */ if (pio == NULL) { ASSERT0(write_lsize); ASSERT(!HDR_HAS_L1HDR(head)); kmem_cache_free(hdr_l2only_cache, head); return (0); } ASSERT3U(write_asize, <=, target_sz); ARCSTAT_BUMP(arcstat_l2_writes_sent); ARCSTAT_INCR(arcstat_l2_write_bytes, write_psize); ARCSTAT_INCR(arcstat_l2_lsize, write_lsize); ARCSTAT_INCR(arcstat_l2_psize, write_psize); vdev_space_update(dev->l2ad_vdev, write_psize, 0, 0); /* * Bump device hand to the device start if it is approaching the end. * l2arc_evict() will already have evicted ahead for this case. */ if (dev->l2ad_hand >= (dev->l2ad_end - target_sz)) { dev->l2ad_hand = dev->l2ad_start; dev->l2ad_first = B_FALSE; } dev->l2ad_writing = B_TRUE; (void) zio_wait(pio); dev->l2ad_writing = B_FALSE; return (write_asize); } /* * This thread feeds the L2ARC at regular intervals. This is the beating * heart of the L2ARC. */ /* ARGSUSED */ static void l2arc_feed_thread(void *unused) { callb_cpr_t cpr; l2arc_dev_t *dev; spa_t *spa; uint64_t size, wrote; clock_t begin, next = ddi_get_lbolt(); fstrans_cookie_t cookie; CALLB_CPR_INIT(&cpr, &l2arc_feed_thr_lock, callb_generic_cpr, FTAG); mutex_enter(&l2arc_feed_thr_lock); cookie = spl_fstrans_mark(); while (l2arc_thread_exit == 0) { CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait_sig(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock, next); CALLB_CPR_SAFE_END(&cpr, &l2arc_feed_thr_lock); next = ddi_get_lbolt() + hz; /* * Quick check for L2ARC devices. */ mutex_enter(&l2arc_dev_mtx); if (l2arc_ndev == 0) { mutex_exit(&l2arc_dev_mtx); continue; } mutex_exit(&l2arc_dev_mtx); begin = ddi_get_lbolt(); /* * This selects the next l2arc device to write to, and in * doing so the next spa to feed from: dev->l2ad_spa. This * will return NULL if there are now no l2arc devices or if * they are all faulted. * * If a device is returned, its spa's config lock is also * held to prevent device removal. l2arc_dev_get_next() * will grab and release l2arc_dev_mtx. */ if ((dev = l2arc_dev_get_next()) == NULL) continue; spa = dev->l2ad_spa; ASSERT3P(spa, !=, NULL); /* * If the pool is read-only then force the feed thread to * sleep a little longer. */ if (!spa_writeable(spa)) { next = ddi_get_lbolt() + 5 * l2arc_feed_secs * hz; spa_config_exit(spa, SCL_L2ARC, dev); continue; } /* * Avoid contributing to memory pressure. */ if (arc_reclaim_needed()) { ARCSTAT_BUMP(arcstat_l2_abort_lowmem); spa_config_exit(spa, SCL_L2ARC, dev); continue; } ARCSTAT_BUMP(arcstat_l2_feeds); size = l2arc_write_size(); /* * Evict L2ARC buffers that will be overwritten. */ l2arc_evict(dev, size, B_FALSE); /* * Write ARC buffers. */ wrote = l2arc_write_buffers(spa, dev, size); /* * Calculate interval between writes. */ next = l2arc_write_interval(begin, size, wrote); spa_config_exit(spa, SCL_L2ARC, dev); } spl_fstrans_unmark(cookie); l2arc_thread_exit = 0; cv_broadcast(&l2arc_feed_thr_cv); CALLB_CPR_EXIT(&cpr); /* drops l2arc_feed_thr_lock */ thread_exit(); } boolean_t l2arc_vdev_present(vdev_t *vd) { l2arc_dev_t *dev; mutex_enter(&l2arc_dev_mtx); for (dev = list_head(l2arc_dev_list); dev != NULL; dev = list_next(l2arc_dev_list, dev)) { if (dev->l2ad_vdev == vd) break; } mutex_exit(&l2arc_dev_mtx); return (dev != NULL); } /* * Add a vdev for use by the L2ARC. By this point the spa has already * validated the vdev and opened it. */ void l2arc_add_vdev(spa_t *spa, vdev_t *vd) { l2arc_dev_t *adddev; ASSERT(!l2arc_vdev_present(vd)); /* * Create a new l2arc device entry. */ adddev = kmem_zalloc(sizeof (l2arc_dev_t), KM_SLEEP); adddev->l2ad_spa = spa; adddev->l2ad_vdev = vd; adddev->l2ad_start = VDEV_LABEL_START_SIZE; adddev->l2ad_end = VDEV_LABEL_START_SIZE + vdev_get_min_asize(vd); adddev->l2ad_hand = adddev->l2ad_start; adddev->l2ad_first = B_TRUE; adddev->l2ad_writing = B_FALSE; list_link_init(&adddev->l2ad_node); mutex_init(&adddev->l2ad_mtx, NULL, MUTEX_DEFAULT, NULL); /* * This is a list of all ARC buffers that are still valid on the * device. */ list_create(&adddev->l2ad_buflist, sizeof (arc_buf_hdr_t), offsetof(arc_buf_hdr_t, b_l2hdr.b_l2node)); vdev_space_update(vd, 0, 0, adddev->l2ad_end - adddev->l2ad_hand); refcount_create(&adddev->l2ad_alloc); /* * Add device to global list */ mutex_enter(&l2arc_dev_mtx); list_insert_head(l2arc_dev_list, adddev); atomic_inc_64(&l2arc_ndev); mutex_exit(&l2arc_dev_mtx); } /* * Remove a vdev from the L2ARC. */ void l2arc_remove_vdev(vdev_t *vd) { l2arc_dev_t *dev, *nextdev, *remdev = NULL; /* * Find the device by vdev */ mutex_enter(&l2arc_dev_mtx); for (dev = list_head(l2arc_dev_list); dev; dev = nextdev) { nextdev = list_next(l2arc_dev_list, dev); if (vd == dev->l2ad_vdev) { remdev = dev; break; } } ASSERT3P(remdev, !=, NULL); /* * Remove device from global list */ list_remove(l2arc_dev_list, remdev); l2arc_dev_last = NULL; /* may have been invalidated */ atomic_dec_64(&l2arc_ndev); mutex_exit(&l2arc_dev_mtx); /* * Clear all buflists and ARC references. L2ARC device flush. */ l2arc_evict(remdev, 0, B_TRUE); list_destroy(&remdev->l2ad_buflist); mutex_destroy(&remdev->l2ad_mtx); refcount_destroy(&remdev->l2ad_alloc); kmem_free(remdev, sizeof (l2arc_dev_t)); } void l2arc_init(void) { l2arc_thread_exit = 0; l2arc_ndev = 0; l2arc_writes_sent = 0; l2arc_writes_done = 0; mutex_init(&l2arc_feed_thr_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&l2arc_feed_thr_cv, NULL, CV_DEFAULT, NULL); mutex_init(&l2arc_dev_mtx, NULL, MUTEX_DEFAULT, NULL); mutex_init(&l2arc_free_on_write_mtx, NULL, MUTEX_DEFAULT, NULL); l2arc_dev_list = &L2ARC_dev_list; l2arc_free_on_write = &L2ARC_free_on_write; list_create(l2arc_dev_list, sizeof (l2arc_dev_t), offsetof(l2arc_dev_t, l2ad_node)); list_create(l2arc_free_on_write, sizeof (l2arc_data_free_t), offsetof(l2arc_data_free_t, l2df_list_node)); } void l2arc_fini(void) { /* * This is called from dmu_fini(), which is called from spa_fini(); * Because of this, we can assume that all l2arc devices have * already been removed when the pools themselves were removed. */ l2arc_do_free_on_write(); mutex_destroy(&l2arc_feed_thr_lock); cv_destroy(&l2arc_feed_thr_cv); mutex_destroy(&l2arc_dev_mtx); mutex_destroy(&l2arc_free_on_write_mtx); list_destroy(l2arc_dev_list); list_destroy(l2arc_free_on_write); } void l2arc_start(void) { if (!(spa_mode_global & FWRITE)) return; (void) thread_create(NULL, 0, l2arc_feed_thread, NULL, 0, &p0, TS_RUN, defclsyspri); } void l2arc_stop(void) { if (!(spa_mode_global & FWRITE)) return; mutex_enter(&l2arc_feed_thr_lock); cv_signal(&l2arc_feed_thr_cv); /* kick thread out of startup */ l2arc_thread_exit = 1; while (l2arc_thread_exit != 0) cv_wait(&l2arc_feed_thr_cv, &l2arc_feed_thr_lock); mutex_exit(&l2arc_feed_thr_lock); } #if defined(_KERNEL) EXPORT_SYMBOL(arc_buf_size); EXPORT_SYMBOL(arc_write); EXPORT_SYMBOL(arc_read); EXPORT_SYMBOL(arc_buf_info); EXPORT_SYMBOL(arc_getbuf_func); EXPORT_SYMBOL(arc_add_prune_callback); EXPORT_SYMBOL(arc_remove_prune_callback); /* BEGIN CSTYLED */ module_param(zfs_arc_min, ulong, 0644); MODULE_PARM_DESC(zfs_arc_min, "Min arc size"); module_param(zfs_arc_max, ulong, 0644); MODULE_PARM_DESC(zfs_arc_max, "Max arc size"); module_param(zfs_arc_meta_limit, ulong, 0644); MODULE_PARM_DESC(zfs_arc_meta_limit, "Meta limit for arc size"); module_param(zfs_arc_meta_limit_percent, ulong, 0644); MODULE_PARM_DESC(zfs_arc_meta_limit_percent, "Percent of arc size for arc meta limit"); module_param(zfs_arc_meta_min, ulong, 0644); MODULE_PARM_DESC(zfs_arc_meta_min, "Min arc metadata"); module_param(zfs_arc_meta_prune, int, 0644); MODULE_PARM_DESC(zfs_arc_meta_prune, "Meta objects to scan for prune"); module_param(zfs_arc_meta_adjust_restarts, int, 0644); MODULE_PARM_DESC(zfs_arc_meta_adjust_restarts, "Limit number of restarts in arc_adjust_meta"); module_param(zfs_arc_meta_strategy, int, 0644); MODULE_PARM_DESC(zfs_arc_meta_strategy, "Meta reclaim strategy"); module_param(zfs_arc_grow_retry, int, 0644); MODULE_PARM_DESC(zfs_arc_grow_retry, "Seconds before growing arc size"); module_param(zfs_arc_p_dampener_disable, int, 0644); MODULE_PARM_DESC(zfs_arc_p_dampener_disable, "disable arc_p adapt dampener"); module_param(zfs_arc_shrink_shift, int, 0644); MODULE_PARM_DESC(zfs_arc_shrink_shift, "log2(fraction of arc to reclaim)"); module_param(zfs_arc_pc_percent, uint, 0644); MODULE_PARM_DESC(zfs_arc_pc_percent, "Percent of pagecache to reclaim arc to"); module_param(zfs_arc_p_min_shift, int, 0644); MODULE_PARM_DESC(zfs_arc_p_min_shift, "arc_c shift to calc min/max arc_p"); module_param(zfs_arc_average_blocksize, int, 0444); MODULE_PARM_DESC(zfs_arc_average_blocksize, "Target average block size"); module_param(zfs_compressed_arc_enabled, int, 0644); MODULE_PARM_DESC(zfs_compressed_arc_enabled, "Disable compressed arc buffers"); module_param(zfs_arc_min_prefetch_ms, int, 0644); MODULE_PARM_DESC(zfs_arc_min_prefetch_ms, "Min life of prefetch block in ms"); module_param(zfs_arc_min_prescient_prefetch_ms, int, 0644); MODULE_PARM_DESC(zfs_arc_min_prescient_prefetch_ms, "Min life of prescient prefetched block in ms"); module_param(l2arc_write_max, ulong, 0644); MODULE_PARM_DESC(l2arc_write_max, "Max write bytes per interval"); module_param(l2arc_write_boost, ulong, 0644); MODULE_PARM_DESC(l2arc_write_boost, "Extra write bytes during device warmup"); module_param(l2arc_headroom, ulong, 0644); MODULE_PARM_DESC(l2arc_headroom, "Number of max device writes to precache"); module_param(l2arc_headroom_boost, ulong, 0644); MODULE_PARM_DESC(l2arc_headroom_boost, "Compressed l2arc_headroom multiplier"); module_param(l2arc_feed_secs, ulong, 0644); MODULE_PARM_DESC(l2arc_feed_secs, "Seconds between L2ARC writing"); module_param(l2arc_feed_min_ms, ulong, 0644); MODULE_PARM_DESC(l2arc_feed_min_ms, "Min feed interval in milliseconds"); module_param(l2arc_noprefetch, int, 0644); MODULE_PARM_DESC(l2arc_noprefetch, "Skip caching prefetched buffers"); module_param(l2arc_feed_again, int, 0644); MODULE_PARM_DESC(l2arc_feed_again, "Turbo L2ARC warmup"); module_param(l2arc_norw, int, 0644); MODULE_PARM_DESC(l2arc_norw, "No reads during writes"); module_param(zfs_arc_lotsfree_percent, int, 0644); MODULE_PARM_DESC(zfs_arc_lotsfree_percent, "System free memory I/O throttle in bytes"); module_param(zfs_arc_sys_free, ulong, 0644); MODULE_PARM_DESC(zfs_arc_sys_free, "System free memory target size in bytes"); module_param(zfs_arc_dnode_limit, ulong, 0644); MODULE_PARM_DESC(zfs_arc_dnode_limit, "Minimum bytes of dnodes in arc"); module_param(zfs_arc_dnode_limit_percent, ulong, 0644); MODULE_PARM_DESC(zfs_arc_dnode_limit_percent, "Percent of ARC meta buffers for dnodes"); module_param(zfs_arc_dnode_reduce_percent, ulong, 0644); MODULE_PARM_DESC(zfs_arc_dnode_reduce_percent, "Percentage of excess dnodes to try to unpin"); /* END CSTYLED */ #endif diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c index 1af0d8010807..43909cd59f3a 100644 --- a/module/zfs/dbuf.c +++ b/module/zfs/dbuf.c @@ -1,4603 +1,4606 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. - * Copyright (c) 2012, 2017 by Delphix. All rights reserved. + * Copyright (c) 2012, 2018 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include kstat_t *dbuf_ksp; typedef struct dbuf_stats { /* * Various statistics about the size of the dbuf cache. */ kstat_named_t cache_count; kstat_named_t cache_size_bytes; kstat_named_t cache_size_bytes_max; /* * Statistics regarding the bounds on the dbuf cache size. */ kstat_named_t cache_target_bytes; kstat_named_t cache_lowater_bytes; kstat_named_t cache_hiwater_bytes; /* * Total number of dbuf cache evictions that have occurred. */ kstat_named_t cache_total_evicts; /* * The distribution of dbuf levels in the dbuf cache and * the total size of all dbufs at each level. */ kstat_named_t cache_levels[DN_MAX_LEVELS]; kstat_named_t cache_levels_bytes[DN_MAX_LEVELS]; /* * Statistics about the dbuf hash table. */ kstat_named_t hash_hits; kstat_named_t hash_misses; kstat_named_t hash_collisions; kstat_named_t hash_elements; kstat_named_t hash_elements_max; /* * Number of sublists containing more than one dbuf in the dbuf * hash table. Keep track of the longest hash chain. */ kstat_named_t hash_chains; kstat_named_t hash_chain_max; /* * Number of times a dbuf_create() discovers that a dbuf was * already created and in the dbuf hash table. */ kstat_named_t hash_insert_race; /* * Statistics about the size of the metadata dbuf cache. */ kstat_named_t metadata_cache_count; kstat_named_t metadata_cache_size_bytes; kstat_named_t metadata_cache_size_bytes_max; /* * For diagnostic purposes, this is incremented whenever we can't add * something to the metadata cache because it's full, and instead put * the data in the regular dbuf cache. */ kstat_named_t metadata_cache_overflow; } dbuf_stats_t; dbuf_stats_t dbuf_stats = { { "cache_count", KSTAT_DATA_UINT64 }, { "cache_size_bytes", KSTAT_DATA_UINT64 }, { "cache_size_bytes_max", KSTAT_DATA_UINT64 }, { "cache_target_bytes", KSTAT_DATA_UINT64 }, { "cache_lowater_bytes", KSTAT_DATA_UINT64 }, { "cache_hiwater_bytes", KSTAT_DATA_UINT64 }, { "cache_total_evicts", KSTAT_DATA_UINT64 }, { { "cache_levels_N", KSTAT_DATA_UINT64 } }, { { "cache_levels_bytes_N", KSTAT_DATA_UINT64 } }, { "hash_hits", KSTAT_DATA_UINT64 }, { "hash_misses", KSTAT_DATA_UINT64 }, { "hash_collisions", KSTAT_DATA_UINT64 }, { "hash_elements", KSTAT_DATA_UINT64 }, { "hash_elements_max", KSTAT_DATA_UINT64 }, { "hash_chains", KSTAT_DATA_UINT64 }, { "hash_chain_max", KSTAT_DATA_UINT64 }, { "hash_insert_race", KSTAT_DATA_UINT64 }, { "metadata_cache_count", KSTAT_DATA_UINT64 }, { "metadata_cache_size_bytes", KSTAT_DATA_UINT64 }, { "metadata_cache_size_bytes_max", KSTAT_DATA_UINT64 }, { "metadata_cache_overflow", KSTAT_DATA_UINT64 } }; #define DBUF_STAT_INCR(stat, val) \ atomic_add_64(&dbuf_stats.stat.value.ui64, (val)); #define DBUF_STAT_DECR(stat, val) \ DBUF_STAT_INCR(stat, -(val)); #define DBUF_STAT_BUMP(stat) \ DBUF_STAT_INCR(stat, 1); #define DBUF_STAT_BUMPDOWN(stat) \ DBUF_STAT_INCR(stat, -1); #define DBUF_STAT_MAX(stat, v) { \ uint64_t _m; \ while ((v) > (_m = dbuf_stats.stat.value.ui64) && \ (_m != atomic_cas_64(&dbuf_stats.stat.value.ui64, _m, (v))))\ continue; \ } struct dbuf_hold_impl_data { /* Function arguments */ dnode_t *dh_dn; uint8_t dh_level; uint64_t dh_blkid; boolean_t dh_fail_sparse; boolean_t dh_fail_uncached; void *dh_tag; dmu_buf_impl_t **dh_dbp; /* Local variables */ dmu_buf_impl_t *dh_db; dmu_buf_impl_t *dh_parent; blkptr_t *dh_bp; int dh_err; dbuf_dirty_record_t *dh_dr; int dh_depth; }; static void __dbuf_hold_impl_init(struct dbuf_hold_impl_data *dh, dnode_t *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, void *tag, dmu_buf_impl_t **dbp, int depth); static int __dbuf_hold_impl(struct dbuf_hold_impl_data *dh); static boolean_t dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx); static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx); extern inline void dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp); /* * Global data structures and functions for the dbuf cache. */ static kmem_cache_t *dbuf_kmem_cache; static taskq_t *dbu_evict_taskq; static kthread_t *dbuf_cache_evict_thread; static kmutex_t dbuf_evict_lock; static kcondvar_t dbuf_evict_cv; static boolean_t dbuf_evict_thread_exit; /* * There are two dbuf caches; each dbuf can only be in one of them at a time. * * 1. Cache of metadata dbufs, to help make read-heavy administrative commands * from /sbin/zfs run faster. The "metadata cache" specifically stores dbufs * that represent the metadata that describes filesystems/snapshots/ * bookmarks/properties/etc. We only evict from this cache when we export a * pool, to short-circuit as much I/O as possible for all administrative * commands that need the metadata. There is no eviction policy for this * cache, because we try to only include types in it which would occupy a * very small amount of space per object but create a large impact on the * performance of these commands. Instead, after it reaches a maximum size * (which should only happen on very small memory systems with a very large * number of filesystem objects), we stop taking new dbufs into the * metadata cache, instead putting them in the normal dbuf cache. * * 2. LRU cache of dbufs. The dbuf cache maintains a list of dbufs that * are not currently held but have been recently released. These dbufs * are not eligible for arc eviction until they are aged out of the cache. * Dbufs that are aged out of the cache will be immediately destroyed and * become eligible for arc eviction. * * Dbufs are added to these caches once the last hold is released. If a dbuf is * later accessed and still exists in the dbuf cache, then it will be removed * from the cache and later re-added to the head of the cache. * * If a given dbuf meets the requirements for the metadata cache, it will go * there, otherwise it will be considered for the generic LRU dbuf cache. The * caches and the refcounts tracking their sizes are stored in an array indexed * by those caches' matching enum values (from dbuf_cached_state_t). */ typedef struct dbuf_cache { multilist_t *cache; refcount_t size; } dbuf_cache_t; dbuf_cache_t dbuf_caches[DB_CACHE_MAX]; /* Size limits for the caches */ unsigned long dbuf_cache_max_bytes = 0; unsigned long dbuf_metadata_cache_max_bytes = 0; /* Set the default sizes of the caches to log2 fraction of arc size */ int dbuf_cache_shift = 5; int dbuf_metadata_cache_shift = 6; /* * The LRU dbuf cache uses a three-stage eviction policy: * - A low water marker designates when the dbuf eviction thread * should stop evicting from the dbuf cache. * - When we reach the maximum size (aka mid water mark), we * signal the eviction thread to run. * - The high water mark indicates when the eviction thread * is unable to keep up with the incoming load and eviction must * happen in the context of the calling thread. * * The dbuf cache: * (max size) * low water mid water hi water * +----------------------------------------+----------+----------+ * | | | | * | | | | * | | | | * | | | | * +----------------------------------------+----------+----------+ * stop signal evict * evicting eviction directly * thread * * The high and low water marks indicate the operating range for the eviction * thread. The low water mark is, by default, 90% of the total size of the * cache and the high water mark is at 110% (both of these percentages can be * changed by setting dbuf_cache_lowater_pct and dbuf_cache_hiwater_pct, * respectively). The eviction thread will try to ensure that the cache remains * within this range by waking up every second and checking if the cache is * above the low water mark. The thread can also be woken up by callers adding * elements into the cache if the cache is larger than the mid water (i.e max * cache size). Once the eviction thread is woken up and eviction is required, * it will continue evicting buffers until it's able to reduce the cache size * to the low water mark. If the cache size continues to grow and hits the high * water mark, then callers adding elements to the cache will begin to evict * directly from the cache until the cache is no longer above the high water * mark. */ /* * The percentage above and below the maximum cache size. */ uint_t dbuf_cache_hiwater_pct = 10; uint_t dbuf_cache_lowater_pct = 10; /* ARGSUSED */ static int dbuf_cons(void *vdb, void *unused, int kmflag) { dmu_buf_impl_t *db = vdb; bzero(db, sizeof (dmu_buf_impl_t)); mutex_init(&db->db_mtx, NULL, MUTEX_DEFAULT, NULL); cv_init(&db->db_changed, NULL, CV_DEFAULT, NULL); multilist_link_init(&db->db_cache_link); refcount_create(&db->db_holds); return (0); } /* ARGSUSED */ static void dbuf_dest(void *vdb, void *unused) { dmu_buf_impl_t *db = vdb; mutex_destroy(&db->db_mtx); cv_destroy(&db->db_changed); ASSERT(!multilist_link_active(&db->db_cache_link)); refcount_destroy(&db->db_holds); } /* * dbuf hash table routines */ static dbuf_hash_table_t dbuf_hash_table; static uint64_t dbuf_hash_count; /* * We use Cityhash for this. It's fast, and has good hash properties without * requiring any large static buffers. */ static uint64_t dbuf_hash(void *os, uint64_t obj, uint8_t lvl, uint64_t blkid) { return (cityhash4((uintptr_t)os, obj, (uint64_t)lvl, blkid)); } #define DBUF_EQUAL(dbuf, os, obj, level, blkid) \ ((dbuf)->db.db_object == (obj) && \ (dbuf)->db_objset == (os) && \ (dbuf)->db_level == (level) && \ (dbuf)->db_blkid == (blkid)) dmu_buf_impl_t * dbuf_find(objset_t *os, uint64_t obj, uint8_t level, uint64_t blkid) { dbuf_hash_table_t *h = &dbuf_hash_table; uint64_t hv; uint64_t idx; dmu_buf_impl_t *db; hv = dbuf_hash(os, obj, level, blkid); idx = hv & h->hash_table_mask; mutex_enter(DBUF_HASH_MUTEX(h, idx)); for (db = h->hash_table[idx]; db != NULL; db = db->db_hash_next) { if (DBUF_EQUAL(db, os, obj, level, blkid)) { mutex_enter(&db->db_mtx); if (db->db_state != DB_EVICTING) { mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (db); } mutex_exit(&db->db_mtx); } } mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (NULL); } static dmu_buf_impl_t * dbuf_find_bonus(objset_t *os, uint64_t object) { dnode_t *dn; dmu_buf_impl_t *db = NULL; if (dnode_hold(os, object, FTAG, &dn) == 0) { rw_enter(&dn->dn_struct_rwlock, RW_READER); if (dn->dn_bonus != NULL) { db = dn->dn_bonus; mutex_enter(&db->db_mtx); } rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); } return (db); } /* * Insert an entry into the hash table. If there is already an element * equal to elem in the hash table, then the already existing element * will be returned and the new element will not be inserted. * Otherwise returns NULL. */ static dmu_buf_impl_t * dbuf_hash_insert(dmu_buf_impl_t *db) { dbuf_hash_table_t *h = &dbuf_hash_table; objset_t *os = db->db_objset; uint64_t obj = db->db.db_object; int level = db->db_level; uint64_t blkid, hv, idx; dmu_buf_impl_t *dbf; uint32_t i; blkid = db->db_blkid; hv = dbuf_hash(os, obj, level, blkid); idx = hv & h->hash_table_mask; mutex_enter(DBUF_HASH_MUTEX(h, idx)); for (dbf = h->hash_table[idx], i = 0; dbf != NULL; dbf = dbf->db_hash_next, i++) { if (DBUF_EQUAL(dbf, os, obj, level, blkid)) { mutex_enter(&dbf->db_mtx); if (dbf->db_state != DB_EVICTING) { mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (dbf); } mutex_exit(&dbf->db_mtx); } } if (i > 0) { DBUF_STAT_BUMP(hash_collisions); if (i == 1) DBUF_STAT_BUMP(hash_chains); DBUF_STAT_MAX(hash_chain_max, i); } mutex_enter(&db->db_mtx); db->db_hash_next = h->hash_table[idx]; h->hash_table[idx] = db; mutex_exit(DBUF_HASH_MUTEX(h, idx)); atomic_inc_64(&dbuf_hash_count); DBUF_STAT_MAX(hash_elements_max, dbuf_hash_count); return (NULL); } /* * This returns whether this dbuf should be stored in the metadata cache, which * is based on whether it's from one of the dnode types that store data related * to traversing dataset hierarchies. */ static boolean_t dbuf_include_in_metadata_cache(dmu_buf_impl_t *db) { DB_DNODE_ENTER(db); dmu_object_type_t type = DB_DNODE(db)->dn_type; DB_DNODE_EXIT(db); /* Check if this dbuf is one of the types we care about */ if (DMU_OT_IS_METADATA_CACHED(type)) { /* If we hit this, then we set something up wrong in dmu_ot */ ASSERT(DMU_OT_IS_METADATA(type)); /* * Sanity check for small-memory systems: don't allocate too * much memory for this purpose. */ if (refcount_count(&dbuf_caches[DB_DBUF_METADATA_CACHE].size) > dbuf_metadata_cache_max_bytes) { DBUF_STAT_BUMP(metadata_cache_overflow); return (B_FALSE); } return (B_TRUE); } return (B_FALSE); } /* * Remove an entry from the hash table. It must be in the EVICTING state. */ static void dbuf_hash_remove(dmu_buf_impl_t *db) { dbuf_hash_table_t *h = &dbuf_hash_table; uint64_t hv, idx; dmu_buf_impl_t *dbf, **dbp; hv = dbuf_hash(db->db_objset, db->db.db_object, db->db_level, db->db_blkid); idx = hv & h->hash_table_mask; /* * We mustn't hold db_mtx to maintain lock ordering: * DBUF_HASH_MUTEX > db_mtx. */ ASSERT(refcount_is_zero(&db->db_holds)); ASSERT(db->db_state == DB_EVICTING); ASSERT(!MUTEX_HELD(&db->db_mtx)); mutex_enter(DBUF_HASH_MUTEX(h, idx)); dbp = &h->hash_table[idx]; while ((dbf = *dbp) != db) { dbp = &dbf->db_hash_next; ASSERT(dbf != NULL); } *dbp = db->db_hash_next; db->db_hash_next = NULL; if (h->hash_table[idx] && h->hash_table[idx]->db_hash_next == NULL) DBUF_STAT_BUMPDOWN(hash_chains); mutex_exit(DBUF_HASH_MUTEX(h, idx)); atomic_dec_64(&dbuf_hash_count); } typedef enum { DBVU_EVICTING, DBVU_NOT_EVICTING } dbvu_verify_type_t; static void dbuf_verify_user(dmu_buf_impl_t *db, dbvu_verify_type_t verify_type) { #ifdef ZFS_DEBUG int64_t holds; if (db->db_user == NULL) return; /* Only data blocks support the attachment of user data. */ ASSERT(db->db_level == 0); /* Clients must resolve a dbuf before attaching user data. */ ASSERT(db->db.db_data != NULL); ASSERT3U(db->db_state, ==, DB_CACHED); holds = refcount_count(&db->db_holds); if (verify_type == DBVU_EVICTING) { /* * Immediate eviction occurs when holds == dirtycnt. * For normal eviction buffers, holds is zero on * eviction, except when dbuf_fix_old_data() calls * dbuf_clear_data(). However, the hold count can grow * during eviction even though db_mtx is held (see * dmu_bonus_hold() for an example), so we can only * test the generic invariant that holds >= dirtycnt. */ ASSERT3U(holds, >=, db->db_dirtycnt); } else { if (db->db_user_immediate_evict == TRUE) ASSERT3U(holds, >=, db->db_dirtycnt); else ASSERT3U(holds, >, 0); } #endif } static void dbuf_evict_user(dmu_buf_impl_t *db) { dmu_buf_user_t *dbu = db->db_user; ASSERT(MUTEX_HELD(&db->db_mtx)); if (dbu == NULL) return; dbuf_verify_user(db, DBVU_EVICTING); db->db_user = NULL; #ifdef ZFS_DEBUG if (dbu->dbu_clear_on_evict_dbufp != NULL) *dbu->dbu_clear_on_evict_dbufp = NULL; #endif /* * There are two eviction callbacks - one that we call synchronously * and one that we invoke via a taskq. The async one is useful for * avoiding lock order reversals and limiting stack depth. * * Note that if we have a sync callback but no async callback, * it's likely that the sync callback will free the structure * containing the dbu. In that case we need to take care to not * dereference dbu after calling the sync evict func. */ boolean_t has_async = (dbu->dbu_evict_func_async != NULL); if (dbu->dbu_evict_func_sync != NULL) dbu->dbu_evict_func_sync(dbu); if (has_async) { taskq_dispatch_ent(dbu_evict_taskq, dbu->dbu_evict_func_async, dbu, 0, &dbu->dbu_tqent); } } boolean_t dbuf_is_metadata(dmu_buf_impl_t *db) { /* * Consider indirect blocks and spill blocks to be meta data. */ if (db->db_level > 0 || db->db_blkid == DMU_SPILL_BLKID) { return (B_TRUE); } else { boolean_t is_metadata; DB_DNODE_ENTER(db); is_metadata = DMU_OT_IS_METADATA(DB_DNODE(db)->dn_type); DB_DNODE_EXIT(db); return (is_metadata); } } /* * This function *must* return indices evenly distributed between all * sublists of the multilist. This is needed due to how the dbuf eviction * code is laid out; dbuf_evict_thread() assumes dbufs are evenly * distributed between all sublists and uses this assumption when * deciding which sublist to evict from and how much to evict from it. */ unsigned int dbuf_cache_multilist_index_func(multilist_t *ml, void *obj) { dmu_buf_impl_t *db = obj; /* * The assumption here, is the hash value for a given * dmu_buf_impl_t will remain constant throughout it's lifetime * (i.e. it's objset, object, level and blkid fields don't change). * Thus, we don't need to store the dbuf's sublist index * on insertion, as this index can be recalculated on removal. * * Also, the low order bits of the hash value are thought to be * distributed evenly. Otherwise, in the case that the multilist * has a power of two number of sublists, each sublists' usage * would not be evenly distributed. */ return (dbuf_hash(db->db_objset, db->db.db_object, db->db_level, db->db_blkid) % multilist_get_num_sublists(ml)); } static inline unsigned long dbuf_cache_target_bytes(void) { return MIN(dbuf_cache_max_bytes, arc_target_bytes() >> dbuf_cache_shift); } static inline uint64_t dbuf_cache_hiwater_bytes(void) { uint64_t dbuf_cache_target = dbuf_cache_target_bytes(); return (dbuf_cache_target + (dbuf_cache_target * dbuf_cache_hiwater_pct) / 100); } static inline uint64_t dbuf_cache_lowater_bytes(void) { uint64_t dbuf_cache_target = dbuf_cache_target_bytes(); return (dbuf_cache_target - (dbuf_cache_target * dbuf_cache_lowater_pct) / 100); } static inline boolean_t dbuf_cache_above_hiwater(void) { return (refcount_count(&dbuf_caches[DB_DBUF_CACHE].size) > dbuf_cache_hiwater_bytes()); } static inline boolean_t dbuf_cache_above_lowater(void) { return (refcount_count(&dbuf_caches[DB_DBUF_CACHE].size) > dbuf_cache_lowater_bytes()); } /* * Evict the oldest eligible dbuf from the dbuf cache. */ static void dbuf_evict_one(void) { int idx = multilist_get_random_index(dbuf_caches[DB_DBUF_CACHE].cache); multilist_sublist_t *mls = multilist_sublist_lock( dbuf_caches[DB_DBUF_CACHE].cache, idx); ASSERT(!MUTEX_HELD(&dbuf_evict_lock)); dmu_buf_impl_t *db = multilist_sublist_tail(mls); while (db != NULL && mutex_tryenter(&db->db_mtx) == 0) { db = multilist_sublist_prev(mls, db); } DTRACE_PROBE2(dbuf__evict__one, dmu_buf_impl_t *, db, multilist_sublist_t *, mls); if (db != NULL) { multilist_sublist_remove(mls, db); multilist_sublist_unlock(mls); (void) refcount_remove_many(&dbuf_caches[DB_DBUF_CACHE].size, db->db.db_size, db); DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[db->db_level], db->db.db_size); ASSERT3U(db->db_caching_status, ==, DB_DBUF_CACHE); db->db_caching_status = DB_NO_CACHE; dbuf_destroy(db); DBUF_STAT_MAX(cache_size_bytes_max, refcount_count(&dbuf_caches[DB_DBUF_CACHE].size)); DBUF_STAT_BUMP(cache_total_evicts); } else { multilist_sublist_unlock(mls); } } /* * The dbuf evict thread is responsible for aging out dbufs from the * cache. Once the cache has reached it's maximum size, dbufs are removed * and destroyed. The eviction thread will continue running until the size * of the dbuf cache is at or below the maximum size. Once the dbuf is aged * out of the cache it is destroyed and becomes eligible for arc eviction. */ /* ARGSUSED */ static void dbuf_evict_thread(void *unused) { callb_cpr_t cpr; CALLB_CPR_INIT(&cpr, &dbuf_evict_lock, callb_generic_cpr, FTAG); mutex_enter(&dbuf_evict_lock); while (!dbuf_evict_thread_exit) { while (!dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) { CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait_sig_hires(&dbuf_evict_cv, &dbuf_evict_lock, SEC2NSEC(1), MSEC2NSEC(1), 0); CALLB_CPR_SAFE_END(&cpr, &dbuf_evict_lock); } mutex_exit(&dbuf_evict_lock); /* * Keep evicting as long as we're above the low water mark * for the cache. We do this without holding the locks to * minimize lock contention. */ while (dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) { dbuf_evict_one(); } mutex_enter(&dbuf_evict_lock); } dbuf_evict_thread_exit = B_FALSE; cv_broadcast(&dbuf_evict_cv); CALLB_CPR_EXIT(&cpr); /* drops dbuf_evict_lock */ thread_exit(); } /* * Wake up the dbuf eviction thread if the dbuf cache is at its max size. * If the dbuf cache is at its high water mark, then evict a dbuf from the * dbuf cache using the callers context. */ static void dbuf_evict_notify(void) { /* * We check if we should evict without holding the dbuf_evict_lock, * because it's OK to occasionally make the wrong decision here, * and grabbing the lock results in massive lock contention. */ if (refcount_count(&dbuf_caches[DB_DBUF_CACHE].size) > dbuf_cache_target_bytes()) { if (dbuf_cache_above_hiwater()) dbuf_evict_one(); cv_signal(&dbuf_evict_cv); } } static int dbuf_kstat_update(kstat_t *ksp, int rw) { dbuf_stats_t *ds = ksp->ks_data; if (rw == KSTAT_WRITE) { return (SET_ERROR(EACCES)); } else { ds->metadata_cache_size_bytes.value.ui64 = refcount_count(&dbuf_caches[DB_DBUF_METADATA_CACHE].size); ds->cache_size_bytes.value.ui64 = refcount_count(&dbuf_caches[DB_DBUF_CACHE].size); ds->cache_target_bytes.value.ui64 = dbuf_cache_target_bytes(); ds->cache_hiwater_bytes.value.ui64 = dbuf_cache_hiwater_bytes(); ds->cache_lowater_bytes.value.ui64 = dbuf_cache_lowater_bytes(); ds->hash_elements.value.ui64 = dbuf_hash_count; } return (0); } void dbuf_init(void) { uint64_t hsize = 1ULL << 16; dbuf_hash_table_t *h = &dbuf_hash_table; int i; /* * The hash table is big enough to fill all of physical memory * with an average block size of zfs_arc_average_blocksize (default 8K). * By default, the table will take up * totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers). */ while (hsize * zfs_arc_average_blocksize < physmem * PAGESIZE) hsize <<= 1; retry: h->hash_table_mask = hsize - 1; #if defined(_KERNEL) /* * Large allocations which do not require contiguous pages * should be using vmem_alloc() in the linux kernel */ h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP); #else h->hash_table = kmem_zalloc(hsize * sizeof (void *), KM_NOSLEEP); #endif if (h->hash_table == NULL) { /* XXX - we should really return an error instead of assert */ ASSERT(hsize > (1ULL << 10)); hsize >>= 1; goto retry; } dbuf_kmem_cache = kmem_cache_create("dmu_buf_impl_t", sizeof (dmu_buf_impl_t), 0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0); for (i = 0; i < DBUF_MUTEXES; i++) mutex_init(&h->hash_mutexes[i], NULL, MUTEX_DEFAULT, NULL); dbuf_stats_init(h); /* * Setup the parameters for the dbuf caches. We set the sizes of the * dbuf cache and the metadata cache to 1/32nd and 1/16th (default) * of the target size of the ARC. If the values has been specified as * a module option and they're not greater than the target size of the * ARC, then we honor that value. */ if (dbuf_cache_max_bytes == 0 || dbuf_cache_max_bytes >= arc_target_bytes()) { dbuf_cache_max_bytes = arc_target_bytes() >> dbuf_cache_shift; } if (dbuf_metadata_cache_max_bytes == 0 || dbuf_metadata_cache_max_bytes >= arc_target_bytes()) { dbuf_metadata_cache_max_bytes = arc_target_bytes() >> dbuf_metadata_cache_shift; } /* * All entries are queued via taskq_dispatch_ent(), so min/maxalloc * configuration is not required. */ dbu_evict_taskq = taskq_create("dbu_evict", 1, defclsyspri, 0, 0, 0); for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) { dbuf_caches[dcs].cache = multilist_create(sizeof (dmu_buf_impl_t), offsetof(dmu_buf_impl_t, db_cache_link), dbuf_cache_multilist_index_func); refcount_create(&dbuf_caches[dcs].size); } dbuf_evict_thread_exit = B_FALSE; mutex_init(&dbuf_evict_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&dbuf_evict_cv, NULL, CV_DEFAULT, NULL); dbuf_cache_evict_thread = thread_create(NULL, 0, dbuf_evict_thread, NULL, 0, &p0, TS_RUN, minclsyspri); dbuf_ksp = kstat_create("zfs", 0, "dbufstats", "misc", KSTAT_TYPE_NAMED, sizeof (dbuf_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (dbuf_ksp != NULL) { dbuf_ksp->ks_data = &dbuf_stats; dbuf_ksp->ks_update = dbuf_kstat_update; kstat_install(dbuf_ksp); for (i = 0; i < DN_MAX_LEVELS; i++) { snprintf(dbuf_stats.cache_levels[i].name, KSTAT_STRLEN, "cache_level_%d", i); dbuf_stats.cache_levels[i].data_type = KSTAT_DATA_UINT64; snprintf(dbuf_stats.cache_levels_bytes[i].name, KSTAT_STRLEN, "cache_level_%d_bytes", i); dbuf_stats.cache_levels_bytes[i].data_type = KSTAT_DATA_UINT64; } } } void dbuf_fini(void) { dbuf_hash_table_t *h = &dbuf_hash_table; int i; dbuf_stats_destroy(); for (i = 0; i < DBUF_MUTEXES; i++) mutex_destroy(&h->hash_mutexes[i]); #if defined(_KERNEL) /* * Large allocations which do not require contiguous pages * should be using vmem_free() in the linux kernel */ vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *)); #else kmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *)); #endif kmem_cache_destroy(dbuf_kmem_cache); taskq_destroy(dbu_evict_taskq); mutex_enter(&dbuf_evict_lock); dbuf_evict_thread_exit = B_TRUE; while (dbuf_evict_thread_exit) { cv_signal(&dbuf_evict_cv); cv_wait(&dbuf_evict_cv, &dbuf_evict_lock); } mutex_exit(&dbuf_evict_lock); mutex_destroy(&dbuf_evict_lock); cv_destroy(&dbuf_evict_cv); for (dbuf_cached_state_t dcs = 0; dcs < DB_CACHE_MAX; dcs++) { refcount_destroy(&dbuf_caches[dcs].size); multilist_destroy(dbuf_caches[dcs].cache); } if (dbuf_ksp != NULL) { kstat_delete(dbuf_ksp); dbuf_ksp = NULL; } } /* * Other stuff. */ #ifdef ZFS_DEBUG static void dbuf_verify(dmu_buf_impl_t *db) { dnode_t *dn; dbuf_dirty_record_t *dr; ASSERT(MUTEX_HELD(&db->db_mtx)); if (!(zfs_flags & ZFS_DEBUG_DBUF_VERIFY)) return; ASSERT(db->db_objset != NULL); DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dn == NULL) { ASSERT(db->db_parent == NULL); ASSERT(db->db_blkptr == NULL); } else { ASSERT3U(db->db.db_object, ==, dn->dn_object); ASSERT3P(db->db_objset, ==, dn->dn_objset); ASSERT3U(db->db_level, <, dn->dn_nlevels); ASSERT(db->db_blkid == DMU_BONUS_BLKID || db->db_blkid == DMU_SPILL_BLKID || !avl_is_empty(&dn->dn_dbufs)); } if (db->db_blkid == DMU_BONUS_BLKID) { ASSERT(dn != NULL); ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen); ASSERT3U(db->db.db_offset, ==, DMU_BONUS_BLKID); } else if (db->db_blkid == DMU_SPILL_BLKID) { ASSERT(dn != NULL); ASSERT0(db->db.db_offset); } else { ASSERT3U(db->db.db_offset, ==, db->db_blkid * db->db.db_size); } for (dr = db->db_data_pending; dr != NULL; dr = dr->dr_next) ASSERT(dr->dr_dbuf == db); for (dr = db->db_last_dirty; dr != NULL; dr = dr->dr_next) ASSERT(dr->dr_dbuf == db); /* * We can't assert that db_size matches dn_datablksz because it * can be momentarily different when another thread is doing * dnode_set_blksz(). */ if (db->db_level == 0 && db->db.db_object == DMU_META_DNODE_OBJECT) { dr = db->db_data_pending; /* * It should only be modified in syncing context, so * make sure we only have one copy of the data. */ ASSERT(dr == NULL || dr->dt.dl.dr_data == db->db_buf); } /* verify db->db_blkptr */ if (db->db_blkptr) { if (db->db_parent == dn->dn_dbuf) { /* db is pointed to by the dnode */ /* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */ if (DMU_OBJECT_IS_SPECIAL(db->db.db_object)) ASSERT(db->db_parent == NULL); else ASSERT(db->db_parent != NULL); if (db->db_blkid != DMU_SPILL_BLKID) ASSERT3P(db->db_blkptr, ==, &dn->dn_phys->dn_blkptr[db->db_blkid]); } else { /* db is pointed to by an indirect block */ ASSERTV(int epb = db->db_parent->db.db_size >> SPA_BLKPTRSHIFT); ASSERT3U(db->db_parent->db_level, ==, db->db_level+1); ASSERT3U(db->db_parent->db.db_object, ==, db->db.db_object); /* * dnode_grow_indblksz() can make this fail if we don't * have the struct_rwlock. XXX indblksz no longer * grows. safe to do this now? */ if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) { ASSERT3P(db->db_blkptr, ==, ((blkptr_t *)db->db_parent->db.db_data + db->db_blkid % epb)); } } } if ((db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr)) && (db->db_buf == NULL || db->db_buf->b_data) && db->db.db_data && db->db_blkid != DMU_BONUS_BLKID && db->db_state != DB_FILL && !dn->dn_free_txg) { /* * If the blkptr isn't set but they have nonzero data, * it had better be dirty, otherwise we'll lose that * data when we evict this buffer. * * There is an exception to this rule for indirect blocks; in * this case, if the indirect block is a hole, we fill in a few * fields on each of the child blocks (importantly, birth time) * to prevent hole birth times from being lost when you * partially fill in a hole. */ if (db->db_dirtycnt == 0) { if (db->db_level == 0) { uint64_t *buf = db->db.db_data; int i; for (i = 0; i < db->db.db_size >> 3; i++) { ASSERT(buf[i] == 0); } } else { blkptr_t *bps = db->db.db_data; ASSERT3U(1 << DB_DNODE(db)->dn_indblkshift, ==, db->db.db_size); /* * We want to verify that all the blkptrs in the * indirect block are holes, but we may have * automatically set up a few fields for them. * We iterate through each blkptr and verify * they only have those fields set. */ for (int i = 0; i < db->db.db_size / sizeof (blkptr_t); i++) { blkptr_t *bp = &bps[i]; ASSERT(ZIO_CHECKSUM_IS_ZERO( &bp->blk_cksum)); ASSERT( DVA_IS_EMPTY(&bp->blk_dva[0]) && DVA_IS_EMPTY(&bp->blk_dva[1]) && DVA_IS_EMPTY(&bp->blk_dva[2])); ASSERT0(bp->blk_fill); ASSERT0(bp->blk_pad[0]); ASSERT0(bp->blk_pad[1]); ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT(BP_IS_HOLE(bp)); ASSERT0(bp->blk_phys_birth); } } } } DB_DNODE_EXIT(db); } #endif static void dbuf_clear_data(dmu_buf_impl_t *db) { ASSERT(MUTEX_HELD(&db->db_mtx)); dbuf_evict_user(db); ASSERT3P(db->db_buf, ==, NULL); db->db.db_data = NULL; if (db->db_state != DB_NOFILL) db->db_state = DB_UNCACHED; } static void dbuf_set_data(dmu_buf_impl_t *db, arc_buf_t *buf) { ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(buf != NULL); db->db_buf = buf; ASSERT(buf->b_data != NULL); db->db.db_data = buf->b_data; } /* * Loan out an arc_buf for read. Return the loaned arc_buf. */ arc_buf_t * dbuf_loan_arcbuf(dmu_buf_impl_t *db) { arc_buf_t *abuf; ASSERT(db->db_blkid != DMU_BONUS_BLKID); mutex_enter(&db->db_mtx); if (arc_released(db->db_buf) || refcount_count(&db->db_holds) > 1) { int blksz = db->db.db_size; spa_t *spa = db->db_objset->os_spa; mutex_exit(&db->db_mtx); abuf = arc_loan_buf(spa, B_FALSE, blksz); bcopy(db->db.db_data, abuf->b_data, blksz); } else { abuf = db->db_buf; arc_loan_inuse_buf(abuf, db); db->db_buf = NULL; dbuf_clear_data(db); mutex_exit(&db->db_mtx); } return (abuf); } /* * Calculate which level n block references the data at the level 0 offset * provided. */ uint64_t dbuf_whichblock(const dnode_t *dn, const int64_t level, const uint64_t offset) { if (dn->dn_datablkshift != 0 && dn->dn_indblkshift != 0) { /* * The level n blkid is equal to the level 0 blkid divided by * the number of level 0s in a level n block. * * The level 0 blkid is offset >> datablkshift = * offset / 2^datablkshift. * * The number of level 0s in a level n is the number of block * pointers in an indirect block, raised to the power of level. * This is 2^(indblkshift - SPA_BLKPTRSHIFT)^level = * 2^(level*(indblkshift - SPA_BLKPTRSHIFT)). * * Thus, the level n blkid is: offset / * ((2^datablkshift)*(2^(level*(indblkshift-SPA_BLKPTRSHIFT)))) * = offset / 2^(datablkshift + level * * (indblkshift - SPA_BLKPTRSHIFT)) * = offset >> (datablkshift + level * * (indblkshift - SPA_BLKPTRSHIFT)) */ const unsigned exp = dn->dn_datablkshift + level * (dn->dn_indblkshift - SPA_BLKPTRSHIFT); if (exp >= 8 * sizeof (offset)) { /* This only happens on the highest indirection level */ ASSERT3U(level, ==, dn->dn_nlevels - 1); return (0); } ASSERT3U(exp, <, 8 * sizeof (offset)); return (offset >> exp); } else { ASSERT3U(offset, <, dn->dn_datablksz); return (0); } } static void dbuf_read_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; mutex_enter(&db->db_mtx); ASSERT3U(db->db_state, ==, DB_READ); /* * All reads are synchronous, so we must have a hold on the dbuf */ ASSERT(refcount_count(&db->db_holds) > 0); ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); - if (db->db_level == 0 && db->db_freed_in_flight) { - /* we were freed in flight; disregard any error */ - if (buf == NULL) { - buf = arc_alloc_buf(db->db_objset->os_spa, - db, DBUF_GET_BUFC_TYPE(db), db->db.db_size); - } + if (buf == NULL) { + /* i/o error */ + ASSERT(zio == NULL || zio->io_error != 0); + ASSERT(db->db_blkid != DMU_BONUS_BLKID); + ASSERT3P(db->db_buf, ==, NULL); + db->db_state = DB_UNCACHED; + } else if (db->db_level == 0 && db->db_freed_in_flight) { + /* freed in flight */ + ASSERT(zio == NULL || zio->io_error == 0); arc_release(buf, db); bzero(buf->b_data, db->db.db_size); arc_buf_freeze(buf); db->db_freed_in_flight = FALSE; dbuf_set_data(db, buf); db->db_state = DB_CACHED; - } else if (buf != NULL) { + } else { + /* success */ + ASSERT(zio == NULL || zio->io_error == 0); dbuf_set_data(db, buf); db->db_state = DB_CACHED; - } else { - ASSERT(db->db_blkid != DMU_BONUS_BLKID); - ASSERT3P(db->db_buf, ==, NULL); - db->db_state = DB_UNCACHED; } cv_broadcast(&db->db_changed); dbuf_rele_and_unlock(db, NULL, B_FALSE); } /* * This function ensures that, when doing a decrypting read of a block, * we make sure we have decrypted the dnode associated with it. We must do * this so that we ensure we are fully authenticating the checksum-of-MACs * tree from the root of the objset down to this block. Indirect blocks are * always verified against their secure checksum-of-MACs assuming that the * dnode containing them is correct. Now that we are doing a decrypting read, * we can be sure that the key is loaded and verify that assumption. This is * especially important considering that we always read encrypted dnode * blocks as raw data (without verifying their MACs) to start, and * decrypt / authenticate them when we need to read an encrypted bonus buffer. */ static int dbuf_read_verify_dnode_crypt(dmu_buf_impl_t *db, uint32_t flags) { int err = 0; objset_t *os = db->db_objset; arc_buf_t *dnode_abuf; dnode_t *dn; zbookmark_phys_t zb; ASSERT(MUTEX_HELD(&db->db_mtx)); if (!os->os_encrypted || os->os_raw_receive || (flags & DB_RF_NO_DECRYPT) != 0) return (0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dnode_abuf = (dn->dn_dbuf != NULL) ? dn->dn_dbuf->db_buf : NULL; if (dnode_abuf == NULL || !arc_is_encrypted(dnode_abuf)) { DB_DNODE_EXIT(db); return (0); } SET_BOOKMARK(&zb, dmu_objset_id(os), DMU_META_DNODE_OBJECT, 0, dn->dn_dbuf->db_blkid); err = arc_untransform(dnode_abuf, os->os_spa, &zb, B_TRUE); /* * An error code of EACCES tells us that the key is still not * available. This is ok if we are only reading authenticated * (and therefore non-encrypted) blocks. */ if (err == EACCES && ((db->db_blkid != DMU_BONUS_BLKID && !DMU_OT_IS_ENCRYPTED(dn->dn_type)) || (db->db_blkid == DMU_BONUS_BLKID && !DMU_OT_IS_ENCRYPTED(dn->dn_bonustype)))) err = 0; DB_DNODE_EXIT(db); return (err); } static int dbuf_read_impl(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags) { dnode_t *dn; zbookmark_phys_t zb; uint32_t aflags = ARC_FLAG_NOWAIT; int err, zio_flags = 0; DB_DNODE_ENTER(db); dn = DB_DNODE(db); ASSERT(!refcount_is_zero(&db->db_holds)); /* We need the struct_rwlock to prevent db_blkptr from changing. */ ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db_state == DB_UNCACHED); ASSERT(db->db_buf == NULL); if (db->db_blkid == DMU_BONUS_BLKID) { /* * The bonus length stored in the dnode may be less than * the maximum available space in the bonus buffer. */ int bonuslen = MIN(dn->dn_bonuslen, dn->dn_phys->dn_bonuslen); int max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); /* if the underlying dnode block is encrypted, decrypt it */ err = dbuf_read_verify_dnode_crypt(db, flags); if (err != 0) { DB_DNODE_EXIT(db); mutex_exit(&db->db_mtx); return (err); } ASSERT3U(bonuslen, <=, db->db.db_size); db->db.db_data = kmem_alloc(max_bonuslen, KM_SLEEP); arc_space_consume(max_bonuslen, ARC_SPACE_BONUS); if (bonuslen < max_bonuslen) bzero(db->db.db_data, max_bonuslen); if (bonuslen) bcopy(DN_BONUS(dn->dn_phys), db->db.db_data, bonuslen); DB_DNODE_EXIT(db); db->db_state = DB_CACHED; mutex_exit(&db->db_mtx); return (0); } /* * Recheck BP_IS_HOLE() after dnode_block_freed() in case dnode_sync() * processes the delete record and clears the bp while we are waiting * for the dn_mtx (resulting in a "no" from block_freed). */ if (db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr) || (db->db_level == 0 && (dnode_block_freed(dn, db->db_blkid) || BP_IS_HOLE(db->db_blkptr)))) { arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); dbuf_set_data(db, arc_alloc_buf(db->db_objset->os_spa, db, type, db->db.db_size)); bzero(db->db.db_data, db->db.db_size); if (db->db_blkptr != NULL && db->db_level > 0 && BP_IS_HOLE(db->db_blkptr) && db->db_blkptr->blk_birth != 0) { blkptr_t *bps = db->db.db_data; for (int i = 0; i < ((1 << DB_DNODE(db)->dn_indblkshift) / sizeof (blkptr_t)); i++) { blkptr_t *bp = &bps[i]; ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==, 1 << dn->dn_indblkshift); BP_SET_LSIZE(bp, BP_GET_LEVEL(db->db_blkptr) == 1 ? dn->dn_datablksz : BP_GET_LSIZE(db->db_blkptr)); BP_SET_TYPE(bp, BP_GET_TYPE(db->db_blkptr)); BP_SET_LEVEL(bp, BP_GET_LEVEL(db->db_blkptr) - 1); BP_SET_BIRTH(bp, db->db_blkptr->blk_birth, 0); } } DB_DNODE_EXIT(db); db->db_state = DB_CACHED; mutex_exit(&db->db_mtx); return (0); } SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); /* * All bps of an encrypted os should have the encryption bit set. * If this is not true it indicates tampering and we report an error. */ if (db->db_objset->os_encrypted && !BP_USES_CRYPT(db->db_blkptr)) { spa_log_error(db->db_objset->os_spa, &zb); zfs_panic_recover("unencrypted block in encrypted " "object set %llu", dmu_objset_id(db->db_objset)); DB_DNODE_EXIT(db); mutex_exit(&db->db_mtx); return (SET_ERROR(EIO)); } err = dbuf_read_verify_dnode_crypt(db, flags); if (err != 0) { DB_DNODE_EXIT(db); mutex_exit(&db->db_mtx); return (err); } DB_DNODE_EXIT(db); db->db_state = DB_READ; mutex_exit(&db->db_mtx); if (DBUF_IS_L2CACHEABLE(db)) aflags |= ARC_FLAG_L2CACHE; dbuf_add_ref(db, NULL); zio_flags = (flags & DB_RF_CANFAIL) ? ZIO_FLAG_CANFAIL : ZIO_FLAG_MUSTSUCCEED; if ((flags & DB_RF_NO_DECRYPT) && BP_IS_PROTECTED(db->db_blkptr)) zio_flags |= ZIO_FLAG_RAW; err = arc_read(zio, db->db_objset->os_spa, db->db_blkptr, dbuf_read_done, db, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); return (err); } /* * This is our just-in-time copy function. It makes a copy of buffers that * have been modified in a previous transaction group before we access them in * the current active group. * * This function is used in three places: when we are dirtying a buffer for the * first time in a txg, when we are freeing a range in a dnode that includes * this buffer, and when we are accessing a buffer which was received compressed * and later referenced in a WRITE_BYREF record. * * Note that when we are called from dbuf_free_range() we do not put a hold on * the buffer, we just traverse the active dbuf list for the dnode. */ static void dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg) { dbuf_dirty_record_t *dr = db->db_last_dirty; ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db.db_data != NULL); ASSERT(db->db_level == 0); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT); if (dr == NULL || (dr->dt.dl.dr_data != ((db->db_blkid == DMU_BONUS_BLKID) ? db->db.db_data : db->db_buf))) return; /* * If the last dirty record for this dbuf has not yet synced * and its referencing the dbuf data, either: * reset the reference to point to a new copy, * or (if there a no active holders) * just null out the current db_data pointer. */ ASSERT3U(dr->dr_txg, >=, txg - 2); if (db->db_blkid == DMU_BONUS_BLKID) { dnode_t *dn = DB_DNODE(db); int bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); dr->dt.dl.dr_data = kmem_alloc(bonuslen, KM_SLEEP); arc_space_consume(bonuslen, ARC_SPACE_BONUS); bcopy(db->db.db_data, dr->dt.dl.dr_data, bonuslen); } else if (refcount_count(&db->db_holds) > db->db_dirtycnt) { dnode_t *dn = DB_DNODE(db); int size = arc_buf_size(db->db_buf); arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); spa_t *spa = db->db_objset->os_spa; enum zio_compress compress_type = arc_get_compression(db->db_buf); if (arc_is_encrypted(db->db_buf)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(db->db_buf, &byteorder, salt, iv, mac); dr->dt.dl.dr_data = arc_alloc_raw_buf(spa, db, dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac, dn->dn_type, size, arc_buf_lsize(db->db_buf), compress_type); } else if (compress_type != ZIO_COMPRESS_OFF) { ASSERT3U(type, ==, ARC_BUFC_DATA); dr->dt.dl.dr_data = arc_alloc_compressed_buf(spa, db, size, arc_buf_lsize(db->db_buf), compress_type); } else { dr->dt.dl.dr_data = arc_alloc_buf(spa, db, type, size); } bcopy(db->db.db_data, dr->dt.dl.dr_data->b_data, size); } else { db->db_buf = NULL; dbuf_clear_data(db); } } int dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags) { int err = 0; boolean_t prefetch; dnode_t *dn; /* * We don't have to hold the mutex to check db_state because it * can't be freed while we have a hold on the buffer. */ ASSERT(!refcount_is_zero(&db->db_holds)); if (db->db_state == DB_NOFILL) return (SET_ERROR(EIO)); DB_DNODE_ENTER(db); dn = DB_DNODE(db); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_enter(&dn->dn_struct_rwlock, RW_READER); prefetch = db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID && (flags & DB_RF_NOPREFETCH) == 0 && dn != NULL && DBUF_IS_CACHEABLE(db); mutex_enter(&db->db_mtx); if (db->db_state == DB_CACHED) { spa_t *spa = dn->dn_objset->os_spa; /* * Ensure that this block's dnode has been decrypted if * the caller has requested decrypted data. */ err = dbuf_read_verify_dnode_crypt(db, flags); /* * If the arc buf is compressed or encrypted and the caller * requested uncompressed data, we need to untransform it * before returning. We also call arc_untransform() on any * unauthenticated blocks, which will verify their MAC if * the key is now available. */ if (err == 0 && db->db_buf != NULL && (flags & DB_RF_NO_DECRYPT) == 0 && (arc_is_encrypted(db->db_buf) || arc_is_unauthenticated(db->db_buf) || arc_get_compression(db->db_buf) != ZIO_COMPRESS_OFF)) { zbookmark_phys_t zb; SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); dbuf_fix_old_data(db, spa_syncing_txg(spa)); err = arc_untransform(db->db_buf, spa, &zb, B_FALSE); dbuf_set_data(db, db->db_buf); } mutex_exit(&db->db_mtx); if (err == 0 && prefetch) dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); DBUF_STAT_BUMP(hash_hits); } else if (db->db_state == DB_UNCACHED) { spa_t *spa = dn->dn_objset->os_spa; boolean_t need_wait = B_FALSE; if (zio == NULL && db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr)) { zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); need_wait = B_TRUE; } err = dbuf_read_impl(db, zio, flags); /* dbuf_read_impl has dropped db_mtx for us */ if (!err && prefetch) dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); DBUF_STAT_BUMP(hash_misses); if (!err && need_wait) err = zio_wait(zio); } else { /* * Another reader came in while the dbuf was in flight * between UNCACHED and CACHED. Either a writer will finish * writing the buffer (sending the dbuf to CACHED) or the * first reader's request will reach the read_done callback * and send the dbuf to CACHED. Otherwise, a failure * occurred and the dbuf went to UNCACHED. */ mutex_exit(&db->db_mtx); if (prefetch) dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); DBUF_STAT_BUMP(hash_misses); /* Skip the wait per the caller's request. */ mutex_enter(&db->db_mtx); if ((flags & DB_RF_NEVERWAIT) == 0) { while (db->db_state == DB_READ || db->db_state == DB_FILL) { ASSERT(db->db_state == DB_READ || (flags & DB_RF_HAVESTRUCT) == 0); DTRACE_PROBE2(blocked__read, dmu_buf_impl_t *, db, zio_t *, zio); cv_wait(&db->db_changed, &db->db_mtx); } if (db->db_state == DB_UNCACHED) err = SET_ERROR(EIO); } mutex_exit(&db->db_mtx); } return (err); } static void dbuf_noread(dmu_buf_impl_t *db) { ASSERT(!refcount_is_zero(&db->db_holds)); ASSERT(db->db_blkid != DMU_BONUS_BLKID); mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); if (db->db_state == DB_UNCACHED) { arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); spa_t *spa = db->db_objset->os_spa; ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); dbuf_set_data(db, arc_alloc_buf(spa, db, type, db->db.db_size)); db->db_state = DB_FILL; } else if (db->db_state == DB_NOFILL) { dbuf_clear_data(db); } else { ASSERT3U(db->db_state, ==, DB_CACHED); } mutex_exit(&db->db_mtx); } void dbuf_unoverride(dbuf_dirty_record_t *dr) { dmu_buf_impl_t *db = dr->dr_dbuf; blkptr_t *bp = &dr->dt.dl.dr_overridden_by; uint64_t txg = dr->dr_txg; ASSERT(MUTEX_HELD(&db->db_mtx)); /* * This assert is valid because dmu_sync() expects to be called by * a zilog's get_data while holding a range lock. This call only * comes from dbuf_dirty() callers who must also hold a range lock. */ ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC); ASSERT(db->db_level == 0); if (db->db_blkid == DMU_BONUS_BLKID || dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN) return; ASSERT(db->db_data_pending != dr); /* free this block */ if (!BP_IS_HOLE(bp) && !dr->dt.dl.dr_nopwrite) zio_free(db->db_objset->os_spa, txg, bp); dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; dr->dt.dl.dr_nopwrite = B_FALSE; dr->dt.dl.dr_has_raw_params = B_FALSE; /* * Release the already-written buffer, so we leave it in * a consistent dirty state. Note that all callers are * modifying the buffer, so they will immediately do * another (redundant) arc_release(). Therefore, leave * the buf thawed to save the effort of freezing & * immediately re-thawing it. */ arc_release(dr->dt.dl.dr_data, db); } /* * Evict (if its unreferenced) or clear (if its referenced) any level-0 * data blocks in the free range, so that any future readers will find * empty blocks. */ void dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid, dmu_tx_t *tx) { dmu_buf_impl_t *db_search; dmu_buf_impl_t *db, *db_next; uint64_t txg = tx->tx_txg; avl_index_t where; if (end_blkid > dn->dn_maxblkid && !(start_blkid == DMU_SPILL_BLKID || end_blkid == DMU_SPILL_BLKID)) end_blkid = dn->dn_maxblkid; dprintf_dnode(dn, "start=%llu end=%llu\n", start_blkid, end_blkid); db_search = kmem_alloc(sizeof (dmu_buf_impl_t), KM_SLEEP); db_search->db_level = 0; db_search->db_blkid = start_blkid; db_search->db_state = DB_SEARCH; mutex_enter(&dn->dn_dbufs_mtx); db = avl_find(&dn->dn_dbufs, db_search, &where); ASSERT3P(db, ==, NULL); db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER); for (; db != NULL; db = db_next) { db_next = AVL_NEXT(&dn->dn_dbufs, db); ASSERT(db->db_blkid != DMU_BONUS_BLKID); if (db->db_level != 0 || db->db_blkid > end_blkid) { break; } ASSERT3U(db->db_blkid, >=, start_blkid); /* found a level 0 buffer in the range */ mutex_enter(&db->db_mtx); if (dbuf_undirty(db, tx)) { /* mutex has been dropped and dbuf destroyed */ continue; } if (db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL || db->db_state == DB_EVICTING) { ASSERT(db->db.db_data == NULL); mutex_exit(&db->db_mtx); continue; } if (db->db_state == DB_READ || db->db_state == DB_FILL) { /* will be handled in dbuf_read_done or dbuf_rele */ db->db_freed_in_flight = TRUE; mutex_exit(&db->db_mtx); continue; } if (refcount_count(&db->db_holds) == 0) { ASSERT(db->db_buf); dbuf_destroy(db); continue; } /* The dbuf is referenced */ if (db->db_last_dirty != NULL) { dbuf_dirty_record_t *dr = db->db_last_dirty; if (dr->dr_txg == txg) { /* * This buffer is "in-use", re-adjust the file * size to reflect that this buffer may * contain new data when we sync. */ if (db->db_blkid != DMU_SPILL_BLKID && db->db_blkid > dn->dn_maxblkid) dn->dn_maxblkid = db->db_blkid; dbuf_unoverride(dr); } else { /* * This dbuf is not dirty in the open context. * Either uncache it (if its not referenced in * the open context) or reset its contents to * empty. */ dbuf_fix_old_data(db, txg); } } /* clear the contents if its cached */ if (db->db_state == DB_CACHED) { ASSERT(db->db.db_data != NULL); arc_release(db->db_buf, db); bzero(db->db.db_data, db->db.db_size); arc_buf_freeze(db->db_buf); } mutex_exit(&db->db_mtx); } kmem_free(db_search, sizeof (dmu_buf_impl_t)); mutex_exit(&dn->dn_dbufs_mtx); } void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx) { arc_buf_t *buf, *obuf; int osize = db->db.db_size; arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); dnode_t *dn; ASSERT(db->db_blkid != DMU_BONUS_BLKID); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* XXX does *this* func really need the lock? */ ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock)); /* * This call to dmu_buf_will_dirty() with the dn_struct_rwlock held * is OK, because there can be no other references to the db * when we are changing its size, so no concurrent DB_FILL can * be happening. */ /* * XXX we should be doing a dbuf_read, checking the return * value and returning that up to our callers */ dmu_buf_will_dirty(&db->db, tx); /* create the data buffer for the new block */ buf = arc_alloc_buf(dn->dn_objset->os_spa, db, type, size); /* copy old block data to the new block */ obuf = db->db_buf; bcopy(obuf->b_data, buf->b_data, MIN(osize, size)); /* zero the remainder */ if (size > osize) bzero((uint8_t *)buf->b_data + osize, size - osize); mutex_enter(&db->db_mtx); dbuf_set_data(db, buf); arc_buf_destroy(obuf, db); db->db.db_size = size; if (db->db_level == 0) { ASSERT3U(db->db_last_dirty->dr_txg, ==, tx->tx_txg); db->db_last_dirty->dt.dl.dr_data = buf; } mutex_exit(&db->db_mtx); dmu_objset_willuse_space(dn->dn_objset, size - osize, tx); DB_DNODE_EXIT(db); } void dbuf_release_bp(dmu_buf_impl_t *db) { ASSERTV(objset_t *os = db->db_objset); ASSERT(dsl_pool_sync_context(dmu_objset_pool(os))); ASSERT(arc_released(os->os_phys_buf) || list_link_active(&os->os_dsl_dataset->ds_synced_link)); ASSERT(db->db_parent == NULL || arc_released(db->db_parent->db_buf)); (void) arc_release(db->db_buf, db); } /* * We already have a dirty record for this TXG, and we are being * dirtied again. */ static void dbuf_redirty(dbuf_dirty_record_t *dr) { dmu_buf_impl_t *db = dr->dr_dbuf; ASSERT(MUTEX_HELD(&db->db_mtx)); if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID) { /* * If this buffer has already been written out, * we now need to reset its state. */ dbuf_unoverride(dr); if (db->db.db_object != DMU_META_DNODE_OBJECT && db->db_state != DB_NOFILL) { /* Already released on initial dirty, so just thaw. */ ASSERT(arc_released(db->db_buf)); arc_buf_thaw(db->db_buf); } } } dbuf_dirty_record_t * dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx) { dnode_t *dn; objset_t *os; dbuf_dirty_record_t **drp, *dr; int drop_struct_lock = FALSE; int txgoff = tx->tx_txg & TXG_MASK; ASSERT(tx->tx_txg != 0); ASSERT(!refcount_is_zero(&db->db_holds)); DMU_TX_DIRTY_BUF(tx, db); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* * Shouldn't dirty a regular buffer in syncing context. Private * objects may be dirtied in syncing context, but only if they * were already pre-dirtied in open context. */ #ifdef DEBUG if (dn->dn_objset->os_dsl_dataset != NULL) { rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); } ASSERT(!dmu_tx_is_syncing(tx) || BP_IS_HOLE(dn->dn_objset->os_rootbp) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) || dn->dn_objset->os_dsl_dataset == NULL); if (dn->dn_objset->os_dsl_dataset != NULL) rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG); #endif /* * We make this assert for private objects as well, but after we * check if we're already dirty. They are allowed to re-dirty * in syncing context. */ ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx == (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN)); mutex_enter(&db->db_mtx); /* * XXX make this true for indirects too? The problem is that * transactions created with dmu_tx_create_assigned() from * syncing context don't bother holding ahead. */ ASSERT(db->db_level != 0 || db->db_state == DB_CACHED || db->db_state == DB_FILL || db->db_state == DB_NOFILL); mutex_enter(&dn->dn_mtx); /* * Don't set dirtyctx to SYNC if we're just modifying this as we * initialize the objset. */ if (dn->dn_dirtyctx == DN_UNDIRTIED) { if (dn->dn_objset->os_dsl_dataset != NULL) { rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); } if (!BP_IS_HOLE(dn->dn_objset->os_rootbp)) { dn->dn_dirtyctx = (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN); ASSERT(dn->dn_dirtyctx_firstset == NULL); dn->dn_dirtyctx_firstset = kmem_alloc(1, KM_SLEEP); } if (dn->dn_objset->os_dsl_dataset != NULL) { rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG); } } if (tx->tx_txg > dn->dn_dirty_txg) dn->dn_dirty_txg = tx->tx_txg; mutex_exit(&dn->dn_mtx); if (db->db_blkid == DMU_SPILL_BLKID) dn->dn_have_spill = B_TRUE; /* * If this buffer is already dirty, we're done. */ drp = &db->db_last_dirty; ASSERT(*drp == NULL || (*drp)->dr_txg <= tx->tx_txg || db->db.db_object == DMU_META_DNODE_OBJECT); while ((dr = *drp) != NULL && dr->dr_txg > tx->tx_txg) drp = &dr->dr_next; if (dr && dr->dr_txg == tx->tx_txg) { DB_DNODE_EXIT(db); dbuf_redirty(dr); mutex_exit(&db->db_mtx); return (dr); } /* * Only valid if not already dirty. */ ASSERT(dn->dn_object == 0 || dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx == (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN)); ASSERT3U(dn->dn_nlevels, >, db->db_level); /* * We should only be dirtying in syncing context if it's the * mos or we're initializing the os or it's a special object. * However, we are allowed to dirty in syncing context provided * we already dirtied it in open context. Hence we must make * this assertion only if we're not already dirty. */ os = dn->dn_objset; VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(os->os_spa)); #ifdef DEBUG if (dn->dn_objset->os_dsl_dataset != NULL) rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); ASSERT(!dmu_tx_is_syncing(tx) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) || os->os_dsl_dataset == NULL || BP_IS_HOLE(os->os_rootbp)); if (dn->dn_objset->os_dsl_dataset != NULL) rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG); #endif ASSERT(db->db.db_size != 0); dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size); if (db->db_blkid != DMU_BONUS_BLKID) { dmu_objset_willuse_space(os, db->db.db_size, tx); } /* * If this buffer is dirty in an old transaction group we need * to make a copy of it so that the changes we make in this * transaction group won't leak out when we sync the older txg. */ dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP); list_link_init(&dr->dr_dirty_node); if (db->db_level == 0) { void *data_old = db->db_buf; if (db->db_state != DB_NOFILL) { if (db->db_blkid == DMU_BONUS_BLKID) { dbuf_fix_old_data(db, tx->tx_txg); data_old = db->db.db_data; } else if (db->db.db_object != DMU_META_DNODE_OBJECT) { /* * Release the data buffer from the cache so * that we can modify it without impacting * possible other users of this cached data * block. Note that indirect blocks and * private objects are not released until the * syncing state (since they are only modified * then). */ arc_release(db->db_buf, db); dbuf_fix_old_data(db, tx->tx_txg); data_old = db->db_buf; } ASSERT(data_old != NULL); } dr->dt.dl.dr_data = data_old; } else { mutex_init(&dr->dt.di.dr_mtx, NULL, MUTEX_NOLOCKDEP, NULL); list_create(&dr->dt.di.dr_children, sizeof (dbuf_dirty_record_t), offsetof(dbuf_dirty_record_t, dr_dirty_node)); } if (db->db_blkid != DMU_BONUS_BLKID && os->os_dsl_dataset != NULL) dr->dr_accounted = db->db.db_size; dr->dr_dbuf = db; dr->dr_txg = tx->tx_txg; dr->dr_next = *drp; *drp = dr; /* * We could have been freed_in_flight between the dbuf_noread * and dbuf_dirty. We win, as though the dbuf_noread() had * happened after the free. */ if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID && db->db_blkid != DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (dn->dn_free_ranges[txgoff] != NULL) { range_tree_clear(dn->dn_free_ranges[txgoff], db->db_blkid, 1); } mutex_exit(&dn->dn_mtx); db->db_freed_in_flight = FALSE; } /* * This buffer is now part of this txg */ dbuf_add_ref(db, (void *)(uintptr_t)tx->tx_txg); db->db_dirtycnt += 1; ASSERT3U(db->db_dirtycnt, <=, 3); mutex_exit(&db->db_mtx); if (db->db_blkid == DMU_BONUS_BLKID || db->db_blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&dn->dn_dirty_records[txgoff], dr); mutex_exit(&dn->dn_mtx); dnode_setdirty(dn, tx); DB_DNODE_EXIT(db); return (dr); } /* * The dn_struct_rwlock prevents db_blkptr from changing * due to a write from syncing context completing * while we are running, so we want to acquire it before * looking at db_blkptr. */ if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) { rw_enter(&dn->dn_struct_rwlock, RW_READER); drop_struct_lock = TRUE; } /* * We need to hold the dn_struct_rwlock to make this assertion, * because it protects dn_phys / dn_next_nlevels from changing. */ ASSERT((dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) || dn->dn_phys->dn_nlevels > db->db_level || dn->dn_next_nlevels[txgoff] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg-2) & TXG_MASK] > db->db_level); /* * If we are overwriting a dedup BP, then unless it is snapshotted, * when we get to syncing context we will need to decrement its * refcount in the DDT. Prefetch the relevant DDT block so that * syncing context won't have to wait for the i/o. */ ddt_prefetch(os->os_spa, db->db_blkptr); if (db->db_level == 0) { ASSERT(!db->db_objset->os_raw_receive || dn->dn_maxblkid >= db->db_blkid); dnode_new_blkid(dn, db->db_blkid, tx, drop_struct_lock); ASSERT(dn->dn_maxblkid >= db->db_blkid); } if (db->db_level+1 < dn->dn_nlevels) { dmu_buf_impl_t *parent = db->db_parent; dbuf_dirty_record_t *di; int parent_held = FALSE; if (db->db_parent == NULL || db->db_parent == dn->dn_dbuf) { int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; parent = dbuf_hold_level(dn, db->db_level+1, db->db_blkid >> epbs, FTAG); ASSERT(parent != NULL); parent_held = TRUE; } if (drop_struct_lock) rw_exit(&dn->dn_struct_rwlock); ASSERT3U(db->db_level+1, ==, parent->db_level); di = dbuf_dirty(parent, tx); if (parent_held) dbuf_rele(parent, FTAG); mutex_enter(&db->db_mtx); /* * Since we've dropped the mutex, it's possible that * dbuf_undirty() might have changed this out from under us. */ if (db->db_last_dirty == dr || dn->dn_object == DMU_META_DNODE_OBJECT) { mutex_enter(&di->dt.di.dr_mtx); ASSERT3U(di->dr_txg, ==, tx->tx_txg); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&di->dt.di.dr_children, dr); mutex_exit(&di->dt.di.dr_mtx); dr->dr_parent = di; } mutex_exit(&db->db_mtx); } else { ASSERT(db->db_level+1 == dn->dn_nlevels); ASSERT(db->db_blkid < dn->dn_nblkptr); ASSERT(db->db_parent == NULL || db->db_parent == dn->dn_dbuf); mutex_enter(&dn->dn_mtx); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&dn->dn_dirty_records[txgoff], dr); mutex_exit(&dn->dn_mtx); if (drop_struct_lock) rw_exit(&dn->dn_struct_rwlock); } dnode_setdirty(dn, tx); DB_DNODE_EXIT(db); return (dr); } /* * Undirty a buffer in the transaction group referenced by the given * transaction. Return whether this evicted the dbuf. */ static boolean_t dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx) { dnode_t *dn; uint64_t txg = tx->tx_txg; dbuf_dirty_record_t *dr, **drp; ASSERT(txg != 0); /* * Due to our use of dn_nlevels below, this can only be called * in open context, unless we are operating on the MOS. * From syncing context, dn_nlevels may be different from the * dn_nlevels used when dbuf was dirtied. */ ASSERT(db->db_objset == dmu_objset_pool(db->db_objset)->dp_meta_objset || txg != spa_syncing_txg(dmu_objset_spa(db->db_objset))); ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT0(db->db_level); ASSERT(MUTEX_HELD(&db->db_mtx)); /* * If this buffer is not dirty, we're done. */ for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) if (dr->dr_txg <= txg) break; if (dr == NULL || dr->dr_txg < txg) return (B_FALSE); ASSERT(dr->dr_txg == txg); ASSERT(dr->dr_dbuf == db); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size); ASSERT(db->db.db_size != 0); dsl_pool_undirty_space(dmu_objset_pool(dn->dn_objset), dr->dr_accounted, txg); *drp = dr->dr_next; /* * Note that there are three places in dbuf_dirty() * where this dirty record may be put on a list. * Make sure to do a list_remove corresponding to * every one of those list_insert calls. */ if (dr->dr_parent) { mutex_enter(&dr->dr_parent->dt.di.dr_mtx); list_remove(&dr->dr_parent->dt.di.dr_children, dr); mutex_exit(&dr->dr_parent->dt.di.dr_mtx); } else if (db->db_blkid == DMU_SPILL_BLKID || db->db_level + 1 == dn->dn_nlevels) { ASSERT(db->db_blkptr == NULL || db->db_parent == dn->dn_dbuf); mutex_enter(&dn->dn_mtx); list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr); mutex_exit(&dn->dn_mtx); } DB_DNODE_EXIT(db); if (db->db_state != DB_NOFILL) { dbuf_unoverride(dr); ASSERT(db->db_buf != NULL); ASSERT(dr->dt.dl.dr_data != NULL); if (dr->dt.dl.dr_data != db->db_buf) arc_buf_destroy(dr->dt.dl.dr_data, db); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; if (refcount_remove(&db->db_holds, (void *)(uintptr_t)txg) == 0) { ASSERT(db->db_state == DB_NOFILL || arc_released(db->db_buf)); dbuf_destroy(db); return (B_TRUE); } return (B_FALSE); } static void dmu_buf_will_dirty_impl(dmu_buf_t *db_fake, int flags, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; ASSERT(tx->tx_txg != 0); ASSERT(!refcount_is_zero(&db->db_holds)); /* * Quick check for dirtyness. For already dirty blocks, this * reduces runtime of this function by >90%, and overall performance * by 50% for some workloads (e.g. file deletion with indirect blocks * cached). */ mutex_enter(&db->db_mtx); dbuf_dirty_record_t *dr; for (dr = db->db_last_dirty; dr != NULL && dr->dr_txg >= tx->tx_txg; dr = dr->dr_next) { /* * It's possible that it is already dirty but not cached, * because there are some calls to dbuf_dirty() that don't * go through dmu_buf_will_dirty(). */ if (dr->dr_txg == tx->tx_txg && db->db_state == DB_CACHED) { /* This dbuf is already dirty and cached. */ dbuf_redirty(dr); mutex_exit(&db->db_mtx); return; } } mutex_exit(&db->db_mtx); DB_DNODE_ENTER(db); if (RW_WRITE_HELD(&DB_DNODE(db)->dn_struct_rwlock)) flags |= DB_RF_HAVESTRUCT; DB_DNODE_EXIT(db); (void) dbuf_read(db, NULL, flags); (void) dbuf_dirty(db, tx); } void dmu_buf_will_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_will_dirty_impl(db_fake, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH, tx); } void dmu_buf_will_not_fill(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; db->db_state = DB_NOFILL; dmu_buf_will_fill(db_fake, tx); } void dmu_buf_will_fill(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(tx->tx_txg != 0); ASSERT(db->db_level == 0); ASSERT(!refcount_is_zero(&db->db_holds)); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx)); dbuf_noread(db); (void) dbuf_dirty(db, tx); } /* * This function is effectively the same as dmu_buf_will_dirty(), but * indicates the caller expects raw encrypted data in the db, and provides * the crypt params (byteorder, salt, iv, mac) which should be stored in the * blkptr_t when this dbuf is written. This is only used for blocks of * dnodes, during raw receive. */ void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dbuf_dirty_record_t *dr; /* * dr_has_raw_params is only processed for blocks of dnodes * (see dbuf_sync_dnode_leaf_crypt()). */ ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT); ASSERT3U(db->db_level, ==, 0); ASSERT(db->db_objset->os_raw_receive); dmu_buf_will_dirty_impl(db_fake, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_NO_DECRYPT, tx); dr = db->db_last_dirty; while (dr != NULL && dr->dr_txg > tx->tx_txg) dr = dr->dr_next; ASSERT3P(dr, !=, NULL); ASSERT3U(dr->dr_txg, ==, tx->tx_txg); dr->dt.dl.dr_has_raw_params = B_TRUE; dr->dt.dl.dr_byteorder = byteorder; bcopy(salt, dr->dt.dl.dr_salt, ZIO_DATA_SALT_LEN); bcopy(iv, dr->dt.dl.dr_iv, ZIO_DATA_IV_LEN); bcopy(mac, dr->dt.dl.dr_mac, ZIO_DATA_MAC_LEN); } #pragma weak dmu_buf_fill_done = dbuf_fill_done /* ARGSUSED */ void dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx) { mutex_enter(&db->db_mtx); DBUF_VERIFY(db); if (db->db_state == DB_FILL) { if (db->db_level == 0 && db->db_freed_in_flight) { ASSERT(db->db_blkid != DMU_BONUS_BLKID); /* we were freed while filling */ /* XXX dbuf_undirty? */ bzero(db->db.db_data, db->db.db_size); db->db_freed_in_flight = FALSE; } db->db_state = DB_CACHED; cv_broadcast(&db->db_changed); } mutex_exit(&db->db_mtx); } void dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data, bp_embedded_type_t etype, enum zio_compress comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf; struct dirty_leaf *dl; dmu_object_type_t type; if (etype == BP_EMBEDDED_TYPE_DATA) { ASSERT(spa_feature_is_active(dmu_objset_spa(db->db_objset), SPA_FEATURE_EMBEDDED_DATA)); } DB_DNODE_ENTER(db); type = DB_DNODE(db)->dn_type; DB_DNODE_EXIT(db); ASSERT0(db->db_level); ASSERT(db->db_blkid != DMU_BONUS_BLKID); dmu_buf_will_not_fill(dbuf, tx); ASSERT3U(db->db_last_dirty->dr_txg, ==, tx->tx_txg); dl = &db->db_last_dirty->dt.dl; encode_embedded_bp_compressed(&dl->dr_overridden_by, data, comp, uncompressed_size, compressed_size); BPE_SET_ETYPE(&dl->dr_overridden_by, etype); BP_SET_TYPE(&dl->dr_overridden_by, type); BP_SET_LEVEL(&dl->dr_overridden_by, 0); BP_SET_BYTEORDER(&dl->dr_overridden_by, byteorder); dl->dr_override_state = DR_OVERRIDDEN; dl->dr_overridden_by.blk_birth = db->db_last_dirty->dr_txg; } /* * Directly assign a provided arc buf to a given dbuf if it's not referenced * by anybody except our caller. Otherwise copy arcbuf's contents to dbuf. */ void dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx) { ASSERT(!refcount_is_zero(&db->db_holds)); ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(db->db_level == 0); ASSERT3U(dbuf_is_metadata(db), ==, arc_is_metadata(buf)); ASSERT(buf != NULL); ASSERT(arc_buf_lsize(buf) == db->db.db_size); ASSERT(tx->tx_txg != 0); arc_return_buf(buf, db); ASSERT(arc_released(buf)); mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); ASSERT(db->db_state == DB_CACHED || db->db_state == DB_UNCACHED); if (db->db_state == DB_CACHED && refcount_count(&db->db_holds) - 1 > db->db_dirtycnt) { /* * In practice, we will never have a case where we have an * encrypted arc buffer while additional holds exist on the * dbuf. We don't handle this here so we simply assert that * fact instead. */ ASSERT(!arc_is_encrypted(buf)); mutex_exit(&db->db_mtx); (void) dbuf_dirty(db, tx); bcopy(buf->b_data, db->db.db_data, db->db.db_size); arc_buf_destroy(buf, db); xuio_stat_wbuf_copied(); return; } xuio_stat_wbuf_nocopy(); if (db->db_state == DB_CACHED) { dbuf_dirty_record_t *dr = db->db_last_dirty; ASSERT(db->db_buf != NULL); if (dr != NULL && dr->dr_txg == tx->tx_txg) { ASSERT(dr->dt.dl.dr_data == db->db_buf); if (!arc_released(db->db_buf)) { ASSERT(dr->dt.dl.dr_override_state == DR_OVERRIDDEN); arc_release(db->db_buf, db); } dr->dt.dl.dr_data = buf; arc_buf_destroy(db->db_buf, db); } else if (dr == NULL || dr->dt.dl.dr_data != db->db_buf) { arc_release(db->db_buf, db); arc_buf_destroy(db->db_buf, db); } db->db_buf = NULL; } ASSERT(db->db_buf == NULL); dbuf_set_data(db, buf); db->db_state = DB_FILL; mutex_exit(&db->db_mtx); (void) dbuf_dirty(db, tx); dmu_buf_fill_done(&db->db, tx); } void dbuf_destroy(dmu_buf_impl_t *db) { dnode_t *dn; dmu_buf_impl_t *parent = db->db_parent; dmu_buf_impl_t *dndb; ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(refcount_is_zero(&db->db_holds)); if (db->db_buf != NULL) { arc_buf_destroy(db->db_buf, db); db->db_buf = NULL; } if (db->db_blkid == DMU_BONUS_BLKID) { int slots = DB_DNODE(db)->dn_num_slots; int bonuslen = DN_SLOTS_TO_BONUSLEN(slots); if (db->db.db_data != NULL) { kmem_free(db->db.db_data, bonuslen); arc_space_return(bonuslen, ARC_SPACE_BONUS); db->db_state = DB_UNCACHED; } } dbuf_clear_data(db); if (multilist_link_active(&db->db_cache_link)) { ASSERT(db->db_caching_status == DB_DBUF_CACHE || db->db_caching_status == DB_DBUF_METADATA_CACHE); multilist_remove(dbuf_caches[db->db_caching_status].cache, db); (void) refcount_remove_many( &dbuf_caches[db->db_caching_status].size, db->db.db_size, db); if (db->db_caching_status == DB_DBUF_METADATA_CACHE) { DBUF_STAT_BUMPDOWN(metadata_cache_count); } else { DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[db->db_level], db->db.db_size); } db->db_caching_status = DB_NO_CACHE; } ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); ASSERT(db->db_data_pending == NULL); db->db_state = DB_EVICTING; db->db_blkptr = NULL; /* * Now that db_state is DB_EVICTING, nobody else can find this via * the hash table. We can now drop db_mtx, which allows us to * acquire the dn_dbufs_mtx. */ mutex_exit(&db->db_mtx); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dndb = dn->dn_dbuf; if (db->db_blkid != DMU_BONUS_BLKID) { boolean_t needlock = !MUTEX_HELD(&dn->dn_dbufs_mtx); if (needlock) mutex_enter(&dn->dn_dbufs_mtx); avl_remove(&dn->dn_dbufs, db); atomic_dec_32(&dn->dn_dbufs_count); membar_producer(); DB_DNODE_EXIT(db); if (needlock) mutex_exit(&dn->dn_dbufs_mtx); /* * Decrementing the dbuf count means that the hold corresponding * to the removed dbuf is no longer discounted in dnode_move(), * so the dnode cannot be moved until after we release the hold. * The membar_producer() ensures visibility of the decremented * value in dnode_move(), since DB_DNODE_EXIT doesn't actually * release any lock. */ mutex_enter(&dn->dn_mtx); dnode_rele_and_unlock(dn, db, B_TRUE); db->db_dnode_handle = NULL; dbuf_hash_remove(db); } else { DB_DNODE_EXIT(db); } ASSERT(refcount_is_zero(&db->db_holds)); db->db_parent = NULL; ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); ASSERT(db->db_hash_next == NULL); ASSERT(db->db_blkptr == NULL); ASSERT(db->db_data_pending == NULL); ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE); ASSERT(!multilist_link_active(&db->db_cache_link)); kmem_cache_free(dbuf_kmem_cache, db); arc_space_return(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); /* * If this dbuf is referenced from an indirect dbuf, * decrement the ref count on the indirect dbuf. */ if (parent && parent != dndb) { mutex_enter(&parent->db_mtx); dbuf_rele_and_unlock(parent, db, B_TRUE); } } /* * Note: While bpp will always be updated if the function returns success, * parentp will not be updated if the dnode does not have dn_dbuf filled in; * this happens when the dnode is the meta-dnode, or {user|group|project}used * object. */ __attribute__((always_inline)) static inline int dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse, dmu_buf_impl_t **parentp, blkptr_t **bpp, struct dbuf_hold_impl_data *dh) { *parentp = NULL; *bpp = NULL; ASSERT(blkid != DMU_BONUS_BLKID); if (blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (dn->dn_have_spill && (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) *bpp = DN_SPILL_BLKPTR(dn->dn_phys); else *bpp = NULL; dbuf_add_ref(dn->dn_dbuf, NULL); *parentp = dn->dn_dbuf; mutex_exit(&dn->dn_mtx); return (0); } int nlevels = (dn->dn_phys->dn_nlevels == 0) ? 1 : dn->dn_phys->dn_nlevels; int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT3U(level * epbs, <, 64); ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); /* * This assertion shouldn't trip as long as the max indirect block size * is less than 1M. The reason for this is that up to that point, * the number of levels required to address an entire object with blocks * of size SPA_MINBLOCKSIZE satisfies nlevels * epbs + 1 <= 64. In * other words, if N * epbs + 1 > 64, then if (N-1) * epbs + 1 > 55 * (i.e. we can address the entire object), objects will all use at most * N-1 levels and the assertion won't overflow. However, once epbs is * 13, 4 * 13 + 1 = 53, but 5 * 13 + 1 = 66. Then, 4 levels will not be * enough to address an entire object, so objects will have 5 levels, * but then this assertion will overflow. * * All this is to say that if we ever increase DN_MAX_INDBLKSHIFT, we * need to redo this logic to handle overflows. */ ASSERT(level >= nlevels || ((nlevels - level - 1) * epbs) + highbit64(dn->dn_phys->dn_nblkptr) <= 64); if (level >= nlevels || blkid >= ((uint64_t)dn->dn_phys->dn_nblkptr << ((nlevels - level - 1) * epbs)) || (fail_sparse && blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))) { /* the buffer has no parent yet */ return (SET_ERROR(ENOENT)); } else if (level < nlevels-1) { /* this block is referenced from an indirect block */ int err; if (dh == NULL) { err = dbuf_hold_impl(dn, level+1, blkid >> epbs, fail_sparse, FALSE, NULL, parentp); } else { __dbuf_hold_impl_init(dh + 1, dn, dh->dh_level + 1, blkid >> epbs, fail_sparse, FALSE, NULL, parentp, dh->dh_depth + 1); err = __dbuf_hold_impl(dh + 1); } if (err) return (err); err = dbuf_read(*parentp, NULL, (DB_RF_HAVESTRUCT | DB_RF_NOPREFETCH | DB_RF_CANFAIL)); if (err) { dbuf_rele(*parentp, NULL); *parentp = NULL; return (err); } *bpp = ((blkptr_t *)(*parentp)->db.db_data) + (blkid & ((1ULL << epbs) - 1)); if (blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs))) ASSERT(BP_IS_HOLE(*bpp)); return (0); } else { /* the block is referenced from the dnode */ ASSERT3U(level, ==, nlevels-1); ASSERT(dn->dn_phys->dn_nblkptr == 0 || blkid < dn->dn_phys->dn_nblkptr); if (dn->dn_dbuf) { dbuf_add_ref(dn->dn_dbuf, NULL); *parentp = dn->dn_dbuf; } *bpp = &dn->dn_phys->dn_blkptr[blkid]; return (0); } } static dmu_buf_impl_t * dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid, dmu_buf_impl_t *parent, blkptr_t *blkptr) { objset_t *os = dn->dn_objset; dmu_buf_impl_t *db, *odb; ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); ASSERT(dn->dn_type != DMU_OT_NONE); db = kmem_cache_alloc(dbuf_kmem_cache, KM_SLEEP); db->db_objset = os; db->db.db_object = dn->dn_object; db->db_level = level; db->db_blkid = blkid; db->db_last_dirty = NULL; db->db_dirtycnt = 0; db->db_dnode_handle = dn->dn_handle; db->db_parent = parent; db->db_blkptr = blkptr; db->db_user = NULL; db->db_user_immediate_evict = FALSE; db->db_freed_in_flight = FALSE; db->db_pending_evict = FALSE; if (blkid == DMU_BONUS_BLKID) { ASSERT3P(parent, ==, dn->dn_dbuf); db->db.db_size = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots) - (dn->dn_nblkptr-1) * sizeof (blkptr_t); ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen); db->db.db_offset = DMU_BONUS_BLKID; db->db_state = DB_UNCACHED; db->db_caching_status = DB_NO_CACHE; /* the bonus dbuf is not placed in the hash table */ arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); return (db); } else if (blkid == DMU_SPILL_BLKID) { db->db.db_size = (blkptr != NULL) ? BP_GET_LSIZE(blkptr) : SPA_MINBLOCKSIZE; db->db.db_offset = 0; } else { int blocksize = db->db_level ? 1 << dn->dn_indblkshift : dn->dn_datablksz; db->db.db_size = blocksize; db->db.db_offset = db->db_blkid * blocksize; } /* * Hold the dn_dbufs_mtx while we get the new dbuf * in the hash table *and* added to the dbufs list. * This prevents a possible deadlock with someone * trying to look up this dbuf before its added to the * dn_dbufs list. */ mutex_enter(&dn->dn_dbufs_mtx); db->db_state = DB_EVICTING; if ((odb = dbuf_hash_insert(db)) != NULL) { /* someone else inserted it first */ kmem_cache_free(dbuf_kmem_cache, db); mutex_exit(&dn->dn_dbufs_mtx); DBUF_STAT_BUMP(hash_insert_race); return (odb); } avl_add(&dn->dn_dbufs, db); db->db_state = DB_UNCACHED; db->db_caching_status = DB_NO_CACHE; mutex_exit(&dn->dn_dbufs_mtx); arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); if (parent && parent != dn->dn_dbuf) dbuf_add_ref(parent, db); ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || refcount_count(&dn->dn_holds) > 0); (void) refcount_add(&dn->dn_holds, db); atomic_inc_32(&dn->dn_dbufs_count); dprintf_dbuf(db, "db=%p\n", db); return (db); } typedef struct dbuf_prefetch_arg { spa_t *dpa_spa; /* The spa to issue the prefetch in. */ zbookmark_phys_t dpa_zb; /* The target block to prefetch. */ int dpa_epbs; /* Entries (blkptr_t's) Per Block Shift. */ int dpa_curlevel; /* The current level that we're reading */ dnode_t *dpa_dnode; /* The dnode associated with the prefetch */ zio_priority_t dpa_prio; /* The priority I/Os should be issued at. */ zio_t *dpa_zio; /* The parent zio_t for all prefetches. */ arc_flags_t dpa_aflags; /* Flags to pass to the final prefetch. */ } dbuf_prefetch_arg_t; /* * Actually issue the prefetch read for the block given. */ static void dbuf_issue_final_prefetch(dbuf_prefetch_arg_t *dpa, blkptr_t *bp) { if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) return; int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE; arc_flags_t aflags = dpa->dpa_aflags | ARC_FLAG_NOWAIT | ARC_FLAG_PREFETCH; /* dnodes are always read as raw and then converted later */ if (BP_GET_TYPE(bp) == DMU_OT_DNODE && BP_IS_PROTECTED(bp) && dpa->dpa_curlevel == 0) zio_flags |= ZIO_FLAG_RAW; ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp)); ASSERT3U(dpa->dpa_curlevel, ==, dpa->dpa_zb.zb_level); ASSERT(dpa->dpa_zio != NULL); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp, NULL, NULL, dpa->dpa_prio, zio_flags, &aflags, &dpa->dpa_zb); } /* * Called when an indirect block above our prefetch target is read in. This * will either read in the next indirect block down the tree or issue the actual * prefetch if the next block down is our target. */ static void dbuf_prefetch_indirect_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *iobp, arc_buf_t *abuf, void *private) { dbuf_prefetch_arg_t *dpa = private; ASSERT3S(dpa->dpa_zb.zb_level, <, dpa->dpa_curlevel); ASSERT3S(dpa->dpa_curlevel, >, 0); + if (abuf == NULL) { + ASSERT(zio == NULL || zio->io_error != 0); + kmem_free(dpa, sizeof (*dpa)); + return; + } + ASSERT(zio == NULL || zio->io_error == 0); + /* * The dpa_dnode is only valid if we are called with a NULL * zio. This indicates that the arc_read() returned without * first calling zio_read() to issue a physical read. Once * a physical read is made the dpa_dnode must be invalidated * as the locks guarding it may have been dropped. If the * dpa_dnode is still valid, then we want to add it to the dbuf * cache. To do so, we must hold the dbuf associated with the block * we just prefetched, read its contents so that we associate it * with an arc_buf_t, and then release it. */ if (zio != NULL) { ASSERT3S(BP_GET_LEVEL(zio->io_bp), ==, dpa->dpa_curlevel); if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS) { ASSERT3U(BP_GET_PSIZE(zio->io_bp), ==, zio->io_size); } else { ASSERT3U(BP_GET_LSIZE(zio->io_bp), ==, zio->io_size); } ASSERT3P(zio->io_spa, ==, dpa->dpa_spa); dpa->dpa_dnode = NULL; } else if (dpa->dpa_dnode != NULL) { uint64_t curblkid = dpa->dpa_zb.zb_blkid >> (dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level)); dmu_buf_impl_t *db = dbuf_hold_level(dpa->dpa_dnode, dpa->dpa_curlevel, curblkid, FTAG); (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_HAVESTRUCT); dbuf_rele(db, FTAG); } - if (abuf == NULL) { - kmem_free(dpa, sizeof (*dpa)); - return; - } - dpa->dpa_curlevel--; uint64_t nextblkid = dpa->dpa_zb.zb_blkid >> (dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level)); blkptr_t *bp = ((blkptr_t *)abuf->b_data) + P2PHASE(nextblkid, 1ULL << dpa->dpa_epbs); if (BP_IS_HOLE(bp)) { kmem_free(dpa, sizeof (*dpa)); } else if (dpa->dpa_curlevel == dpa->dpa_zb.zb_level) { ASSERT3U(nextblkid, ==, dpa->dpa_zb.zb_blkid); dbuf_issue_final_prefetch(dpa, bp); kmem_free(dpa, sizeof (*dpa)); } else { arc_flags_t iter_aflags = ARC_FLAG_NOWAIT; zbookmark_phys_t zb; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (dpa->dpa_aflags & ARC_FLAG_L2CACHE) iter_aflags |= ARC_FLAG_L2CACHE; ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp)); SET_BOOKMARK(&zb, dpa->dpa_zb.zb_objset, dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp, dbuf_prefetch_indirect_done, dpa, dpa->dpa_prio, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &iter_aflags, &zb); } arc_buf_destroy(abuf, private); } /* * Issue prefetch reads for the given block on the given level. If the indirect * blocks above that block are not in memory, we will read them in * asynchronously. As a result, this call never blocks waiting for a read to * complete. Note that the prefetch might fail if the dataset is encrypted and * the encryption key is unmapped before the IO completes. */ void dbuf_prefetch(dnode_t *dn, int64_t level, uint64_t blkid, zio_priority_t prio, arc_flags_t aflags) { blkptr_t bp; int epbs, nlevels, curlevel; uint64_t curblkid; ASSERT(blkid != DMU_BONUS_BLKID); ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); if (blkid > dn->dn_maxblkid) return; if (dnode_block_freed(dn, blkid)) return; /* * This dnode hasn't been written to disk yet, so there's nothing to * prefetch. */ nlevels = dn->dn_phys->dn_nlevels; if (level >= nlevels || dn->dn_phys->dn_nblkptr == 0) return; epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; if (dn->dn_phys->dn_maxblkid < blkid << (epbs * level)) return; dmu_buf_impl_t *db = dbuf_find(dn->dn_objset, dn->dn_object, level, blkid); if (db != NULL) { mutex_exit(&db->db_mtx); /* * This dbuf already exists. It is either CACHED, or * (we assume) about to be read or filled. */ return; } /* * Find the closest ancestor (indirect block) of the target block * that is present in the cache. In this indirect block, we will * find the bp that is at curlevel, curblkid. */ curlevel = level; curblkid = blkid; while (curlevel < nlevels - 1) { int parent_level = curlevel + 1; uint64_t parent_blkid = curblkid >> epbs; dmu_buf_impl_t *db; if (dbuf_hold_impl(dn, parent_level, parent_blkid, FALSE, TRUE, FTAG, &db) == 0) { blkptr_t *bpp = db->db_buf->b_data; bp = bpp[P2PHASE(curblkid, 1 << epbs)]; dbuf_rele(db, FTAG); break; } curlevel = parent_level; curblkid = parent_blkid; } if (curlevel == nlevels - 1) { /* No cached indirect blocks found. */ ASSERT3U(curblkid, <, dn->dn_phys->dn_nblkptr); bp = dn->dn_phys->dn_blkptr[curblkid]; } if (BP_IS_HOLE(&bp)) return; ASSERT3U(curlevel, ==, BP_GET_LEVEL(&bp)); zio_t *pio = zio_root(dmu_objset_spa(dn->dn_objset), NULL, NULL, ZIO_FLAG_CANFAIL); dbuf_prefetch_arg_t *dpa = kmem_zalloc(sizeof (*dpa), KM_SLEEP); dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset; SET_BOOKMARK(&dpa->dpa_zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET, dn->dn_object, level, blkid); dpa->dpa_curlevel = curlevel; dpa->dpa_prio = prio; dpa->dpa_aflags = aflags; dpa->dpa_spa = dn->dn_objset->os_spa; dpa->dpa_dnode = dn; dpa->dpa_epbs = epbs; dpa->dpa_zio = pio; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (DNODE_LEVEL_IS_L2CACHEABLE(dn, level)) dpa->dpa_aflags |= ARC_FLAG_L2CACHE; /* * If we have the indirect just above us, no need to do the asynchronous * prefetch chain; we'll just run the last step ourselves. If we're at * a higher level, though, we want to issue the prefetches for all the * indirect blocks asynchronously, so we can go on with whatever we were * doing. */ if (curlevel == level) { ASSERT3U(curblkid, ==, blkid); dbuf_issue_final_prefetch(dpa, &bp); kmem_free(dpa, sizeof (*dpa)); } else { arc_flags_t iter_aflags = ARC_FLAG_NOWAIT; zbookmark_phys_t zb; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (DNODE_LEVEL_IS_L2CACHEABLE(dn, level)) iter_aflags |= ARC_FLAG_L2CACHE; SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET, dn->dn_object, curlevel, curblkid); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, &bp, dbuf_prefetch_indirect_done, dpa, prio, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &iter_aflags, &zb); } /* * We use pio here instead of dpa_zio since it's possible that * dpa may have already been freed. */ zio_nowait(pio); } #define DBUF_HOLD_IMPL_MAX_DEPTH 20 /* * Helper function for __dbuf_hold_impl() to copy a buffer. Handles * the case of encrypted, compressed and uncompressed buffers by * allocating the new buffer, respectively, with arc_alloc_raw_buf(), * arc_alloc_compressed_buf() or arc_alloc_buf().* * * NOTE: Declared noinline to avoid stack bloat in __dbuf_hold_impl(). */ noinline static void dbuf_hold_copy(struct dbuf_hold_impl_data *dh) { dnode_t *dn = dh->dh_dn; dmu_buf_impl_t *db = dh->dh_db; dbuf_dirty_record_t *dr = dh->dh_dr; arc_buf_t *data = dr->dt.dl.dr_data; enum zio_compress compress_type = arc_get_compression(data); if (arc_is_encrypted(data)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(data, &byteorder, salt, iv, mac); dbuf_set_data(db, arc_alloc_raw_buf(dn->dn_objset->os_spa, db, dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac, dn->dn_type, arc_buf_size(data), arc_buf_lsize(data), compress_type)); } else if (compress_type != ZIO_COMPRESS_OFF) { dbuf_set_data(db, arc_alloc_compressed_buf( dn->dn_objset->os_spa, db, arc_buf_size(data), arc_buf_lsize(data), compress_type)); } else { dbuf_set_data(db, arc_alloc_buf(dn->dn_objset->os_spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size)); } bcopy(data->b_data, db->db.db_data, arc_buf_size(data)); } /* * Returns with db_holds incremented, and db_mtx not held. * Note: dn_struct_rwlock must be held. */ static int __dbuf_hold_impl(struct dbuf_hold_impl_data *dh) { ASSERT3S(dh->dh_depth, <, DBUF_HOLD_IMPL_MAX_DEPTH); dh->dh_parent = NULL; ASSERT(dh->dh_blkid != DMU_BONUS_BLKID); ASSERT(RW_LOCK_HELD(&dh->dh_dn->dn_struct_rwlock)); ASSERT3U(dh->dh_dn->dn_nlevels, >, dh->dh_level); *(dh->dh_dbp) = NULL; /* dbuf_find() returns with db_mtx held */ dh->dh_db = dbuf_find(dh->dh_dn->dn_objset, dh->dh_dn->dn_object, dh->dh_level, dh->dh_blkid); if (dh->dh_db == NULL) { dh->dh_bp = NULL; if (dh->dh_fail_uncached) return (SET_ERROR(ENOENT)); ASSERT3P(dh->dh_parent, ==, NULL); dh->dh_err = dbuf_findbp(dh->dh_dn, dh->dh_level, dh->dh_blkid, dh->dh_fail_sparse, &dh->dh_parent, &dh->dh_bp, dh); if (dh->dh_fail_sparse) { if (dh->dh_err == 0 && dh->dh_bp && BP_IS_HOLE(dh->dh_bp)) dh->dh_err = SET_ERROR(ENOENT); if (dh->dh_err) { if (dh->dh_parent) dbuf_rele(dh->dh_parent, NULL); return (dh->dh_err); } } if (dh->dh_err && dh->dh_err != ENOENT) return (dh->dh_err); dh->dh_db = dbuf_create(dh->dh_dn, dh->dh_level, dh->dh_blkid, dh->dh_parent, dh->dh_bp); } if (dh->dh_fail_uncached && dh->dh_db->db_state != DB_CACHED) { mutex_exit(&dh->dh_db->db_mtx); return (SET_ERROR(ENOENT)); } if (dh->dh_db->db_buf != NULL) { arc_buf_access(dh->dh_db->db_buf); ASSERT3P(dh->dh_db->db.db_data, ==, dh->dh_db->db_buf->b_data); } ASSERT(dh->dh_db->db_buf == NULL || arc_referenced(dh->dh_db->db_buf)); /* * If this buffer is currently syncing out, and we are are * still referencing it from db_data, we need to make a copy * of it in case we decide we want to dirty it again in this txg. */ if (dh->dh_db->db_level == 0 && dh->dh_db->db_blkid != DMU_BONUS_BLKID && dh->dh_dn->dn_object != DMU_META_DNODE_OBJECT && dh->dh_db->db_state == DB_CACHED && dh->dh_db->db_data_pending) { dh->dh_dr = dh->dh_db->db_data_pending; if (dh->dh_dr->dt.dl.dr_data == dh->dh_db->db_buf) dbuf_hold_copy(dh); } if (multilist_link_active(&dh->dh_db->db_cache_link)) { ASSERT(refcount_is_zero(&dh->dh_db->db_holds)); ASSERT(dh->dh_db->db_caching_status == DB_DBUF_CACHE || dh->dh_db->db_caching_status == DB_DBUF_METADATA_CACHE); multilist_remove( dbuf_caches[dh->dh_db->db_caching_status].cache, dh->dh_db); (void) refcount_remove_many( &dbuf_caches[dh->dh_db->db_caching_status].size, dh->dh_db->db.db_size, dh->dh_db); if (dh->dh_db->db_caching_status == DB_DBUF_METADATA_CACHE) { DBUF_STAT_BUMPDOWN(metadata_cache_count); } else { DBUF_STAT_BUMPDOWN(cache_levels[dh->dh_db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[dh->dh_db->db_level], dh->dh_db->db.db_size); } dh->dh_db->db_caching_status = DB_NO_CACHE; } (void) refcount_add(&dh->dh_db->db_holds, dh->dh_tag); DBUF_VERIFY(dh->dh_db); mutex_exit(&dh->dh_db->db_mtx); /* NOTE: we can't rele the parent until after we drop the db_mtx */ if (dh->dh_parent) dbuf_rele(dh->dh_parent, NULL); ASSERT3P(DB_DNODE(dh->dh_db), ==, dh->dh_dn); ASSERT3U(dh->dh_db->db_blkid, ==, dh->dh_blkid); ASSERT3U(dh->dh_db->db_level, ==, dh->dh_level); *(dh->dh_dbp) = dh->dh_db; return (0); } /* * The following code preserves the recursive function dbuf_hold_impl() * but moves the local variables AND function arguments to the heap to * minimize the stack frame size. Enough space is initially allocated * on the stack for 20 levels of recursion. */ int dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, void *tag, dmu_buf_impl_t **dbp) { struct dbuf_hold_impl_data *dh; int error; dh = kmem_alloc(sizeof (struct dbuf_hold_impl_data) * DBUF_HOLD_IMPL_MAX_DEPTH, KM_SLEEP); __dbuf_hold_impl_init(dh, dn, level, blkid, fail_sparse, fail_uncached, tag, dbp, 0); error = __dbuf_hold_impl(dh); kmem_free(dh, sizeof (struct dbuf_hold_impl_data) * DBUF_HOLD_IMPL_MAX_DEPTH); return (error); } static void __dbuf_hold_impl_init(struct dbuf_hold_impl_data *dh, dnode_t *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, void *tag, dmu_buf_impl_t **dbp, int depth) { dh->dh_dn = dn; dh->dh_level = level; dh->dh_blkid = blkid; dh->dh_fail_sparse = fail_sparse; dh->dh_fail_uncached = fail_uncached; dh->dh_tag = tag; dh->dh_dbp = dbp; dh->dh_db = NULL; dh->dh_parent = NULL; dh->dh_bp = NULL; dh->dh_err = 0; dh->dh_dr = NULL; dh->dh_depth = depth; } dmu_buf_impl_t * dbuf_hold(dnode_t *dn, uint64_t blkid, void *tag) { return (dbuf_hold_level(dn, 0, blkid, tag)); } dmu_buf_impl_t * dbuf_hold_level(dnode_t *dn, int level, uint64_t blkid, void *tag) { dmu_buf_impl_t *db; int err = dbuf_hold_impl(dn, level, blkid, FALSE, FALSE, tag, &db); return (err ? NULL : db); } void dbuf_create_bonus(dnode_t *dn) { ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock)); ASSERT(dn->dn_bonus == NULL); dn->dn_bonus = dbuf_create(dn, 0, DMU_BONUS_BLKID, dn->dn_dbuf, NULL); } int dbuf_spill_set_blksz(dmu_buf_t *db_fake, uint64_t blksz, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; if (db->db_blkid != DMU_SPILL_BLKID) return (SET_ERROR(ENOTSUP)); if (blksz == 0) blksz = SPA_MINBLOCKSIZE; ASSERT3U(blksz, <=, spa_maxblocksize(dmu_objset_spa(db->db_objset))); blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE); DB_DNODE_ENTER(db); dn = DB_DNODE(db); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); dbuf_new_size(db, blksz, tx); rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); return (0); } void dbuf_rm_spill(dnode_t *dn, dmu_tx_t *tx) { dbuf_free_range(dn, DMU_SPILL_BLKID, DMU_SPILL_BLKID, tx); } #pragma weak dmu_buf_add_ref = dbuf_add_ref void dbuf_add_ref(dmu_buf_impl_t *db, void *tag) { int64_t holds = refcount_add(&db->db_holds, tag); VERIFY3S(holds, >, 1); } #pragma weak dmu_buf_try_add_ref = dbuf_try_add_ref boolean_t dbuf_try_add_ref(dmu_buf_t *db_fake, objset_t *os, uint64_t obj, uint64_t blkid, void *tag) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dmu_buf_impl_t *found_db; boolean_t result = B_FALSE; if (blkid == DMU_BONUS_BLKID) found_db = dbuf_find_bonus(os, obj); else found_db = dbuf_find(os, obj, 0, blkid); if (found_db != NULL) { if (db == found_db && dbuf_refcount(db) > db->db_dirtycnt) { (void) refcount_add(&db->db_holds, tag); result = B_TRUE; } mutex_exit(&found_db->db_mtx); } return (result); } /* * If you call dbuf_rele() you had better not be referencing the dnode handle * unless you have some other direct or indirect hold on the dnode. (An indirect * hold is a hold on one of the dnode's dbufs, including the bonus buffer.) * Without that, the dbuf_rele() could lead to a dnode_rele() followed by the * dnode's parent dbuf evicting its dnode handles. */ void dbuf_rele(dmu_buf_impl_t *db, void *tag) { mutex_enter(&db->db_mtx); dbuf_rele_and_unlock(db, tag, B_FALSE); } void dmu_buf_rele(dmu_buf_t *db, void *tag) { dbuf_rele((dmu_buf_impl_t *)db, tag); } /* * dbuf_rele() for an already-locked dbuf. This is necessary to allow * db_dirtycnt and db_holds to be updated atomically. The 'evicting' * argument should be set if we are already in the dbuf-evicting code * path, in which case we don't want to recursively evict. This allows us to * avoid deeply nested stacks that would have a call flow similar to this: * * dbuf_rele()-->dbuf_rele_and_unlock()-->dbuf_evict_notify() * ^ | * | | * +-----dbuf_destroy()<--dbuf_evict_one()<--------+ * */ void dbuf_rele_and_unlock(dmu_buf_impl_t *db, void *tag, boolean_t evicting) { int64_t holds; ASSERT(MUTEX_HELD(&db->db_mtx)); DBUF_VERIFY(db); /* * Remove the reference to the dbuf before removing its hold on the * dnode so we can guarantee in dnode_move() that a referenced bonus * buffer has a corresponding dnode hold. */ holds = refcount_remove(&db->db_holds, tag); ASSERT(holds >= 0); /* * We can't freeze indirects if there is a possibility that they * may be modified in the current syncing context. */ if (db->db_buf != NULL && holds == (db->db_level == 0 ? db->db_dirtycnt : 0)) { arc_buf_freeze(db->db_buf); } if (holds == db->db_dirtycnt && db->db_level == 0 && db->db_user_immediate_evict) dbuf_evict_user(db); if (holds == 0) { if (db->db_blkid == DMU_BONUS_BLKID) { dnode_t *dn; boolean_t evict_dbuf = db->db_pending_evict; /* * If the dnode moves here, we cannot cross this * barrier until the move completes. */ DB_DNODE_ENTER(db); dn = DB_DNODE(db); atomic_dec_32(&dn->dn_dbufs_count); /* * Decrementing the dbuf count means that the bonus * buffer's dnode hold is no longer discounted in * dnode_move(). The dnode cannot move until after * the dnode_rele() below. */ DB_DNODE_EXIT(db); /* * Do not reference db after its lock is dropped. * Another thread may evict it. */ mutex_exit(&db->db_mtx); if (evict_dbuf) dnode_evict_bonus(dn); dnode_rele(dn, db); } else if (db->db_buf == NULL) { /* * This is a special case: we never associated this * dbuf with any data allocated from the ARC. */ ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); dbuf_destroy(db); } else if (arc_released(db->db_buf)) { /* * This dbuf has anonymous data associated with it. */ dbuf_destroy(db); } else { boolean_t do_arc_evict = B_FALSE; blkptr_t bp; spa_t *spa = dmu_objset_spa(db->db_objset); if (!DBUF_IS_CACHEABLE(db) && db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr) && !BP_IS_EMBEDDED(db->db_blkptr)) { do_arc_evict = B_TRUE; bp = *db->db_blkptr; } if (!DBUF_IS_CACHEABLE(db) || db->db_pending_evict) { dbuf_destroy(db); } else if (!multilist_link_active(&db->db_cache_link)) { ASSERT3U(db->db_caching_status, ==, DB_NO_CACHE); dbuf_cached_state_t dcs = dbuf_include_in_metadata_cache(db) ? DB_DBUF_METADATA_CACHE : DB_DBUF_CACHE; db->db_caching_status = dcs; multilist_insert(dbuf_caches[dcs].cache, db); (void) refcount_add_many(&dbuf_caches[dcs].size, db->db.db_size, db); if (dcs == DB_DBUF_METADATA_CACHE) { DBUF_STAT_BUMP(metadata_cache_count); DBUF_STAT_MAX( metadata_cache_size_bytes_max, refcount_count( &dbuf_caches[dcs].size)); } else { DBUF_STAT_BUMP( cache_levels[db->db_level]); DBUF_STAT_BUMP(cache_count); DBUF_STAT_INCR( cache_levels_bytes[db->db_level], db->db.db_size); DBUF_STAT_MAX(cache_size_bytes_max, refcount_count( &dbuf_caches[dcs].size)); } mutex_exit(&db->db_mtx); if (db->db_caching_status == DB_DBUF_CACHE && !evicting) { dbuf_evict_notify(); } } if (do_arc_evict) arc_freed(spa, &bp); } } else { mutex_exit(&db->db_mtx); } } #pragma weak dmu_buf_refcount = dbuf_refcount uint64_t dbuf_refcount(dmu_buf_impl_t *db) { return (refcount_count(&db->db_holds)); } uint64_t dmu_buf_user_refcount(dmu_buf_t *db_fake) { uint64_t holds; dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; mutex_enter(&db->db_mtx); ASSERT3U(refcount_count(&db->db_holds), >=, db->db_dirtycnt); holds = refcount_count(&db->db_holds) - db->db_dirtycnt; mutex_exit(&db->db_mtx); return (holds); } void * dmu_buf_replace_user(dmu_buf_t *db_fake, dmu_buf_user_t *old_user, dmu_buf_user_t *new_user) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; mutex_enter(&db->db_mtx); dbuf_verify_user(db, DBVU_NOT_EVICTING); if (db->db_user == old_user) db->db_user = new_user; else old_user = db->db_user; dbuf_verify_user(db, DBVU_NOT_EVICTING); mutex_exit(&db->db_mtx); return (old_user); } void * dmu_buf_set_user(dmu_buf_t *db_fake, dmu_buf_user_t *user) { return (dmu_buf_replace_user(db_fake, NULL, user)); } void * dmu_buf_set_user_ie(dmu_buf_t *db_fake, dmu_buf_user_t *user) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; db->db_user_immediate_evict = TRUE; return (dmu_buf_set_user(db_fake, user)); } void * dmu_buf_remove_user(dmu_buf_t *db_fake, dmu_buf_user_t *user) { return (dmu_buf_replace_user(db_fake, user, NULL)); } void * dmu_buf_get_user(dmu_buf_t *db_fake) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dbuf_verify_user(db, DBVU_NOT_EVICTING); return (db->db_user); } void dmu_buf_user_evict_wait() { taskq_wait(dbu_evict_taskq); } blkptr_t * dmu_buf_get_blkptr(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; return (dbi->db_blkptr); } objset_t * dmu_buf_get_objset(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; return (dbi->db_objset); } dnode_t * dmu_buf_dnode_enter(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; DB_DNODE_ENTER(dbi); return (DB_DNODE(dbi)); } void dmu_buf_dnode_exit(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; DB_DNODE_EXIT(dbi); } static void dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db) { /* ASSERT(dmu_tx_is_syncing(tx) */ ASSERT(MUTEX_HELD(&db->db_mtx)); if (db->db_blkptr != NULL) return; if (db->db_blkid == DMU_SPILL_BLKID) { db->db_blkptr = DN_SPILL_BLKPTR(dn->dn_phys); BP_ZERO(db->db_blkptr); return; } if (db->db_level == dn->dn_phys->dn_nlevels-1) { /* * This buffer was allocated at a time when there was * no available blkptrs from the dnode, or it was * inappropriate to hook it in (i.e., nlevels mis-match). */ ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr); ASSERT(db->db_parent == NULL); db->db_parent = dn->dn_dbuf; db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid]; DBUF_VERIFY(db); } else { dmu_buf_impl_t *parent = db->db_parent; int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT(dn->dn_phys->dn_nlevels > 1); if (parent == NULL) { mutex_exit(&db->db_mtx); rw_enter(&dn->dn_struct_rwlock, RW_READER); parent = dbuf_hold_level(dn, db->db_level + 1, db->db_blkid >> epbs, db); rw_exit(&dn->dn_struct_rwlock); mutex_enter(&db->db_mtx); db->db_parent = parent; } db->db_blkptr = (blkptr_t *)parent->db.db_data + (db->db_blkid & ((1ULL << epbs) - 1)); DBUF_VERIFY(db); } } /* * When syncing out a blocks of dnodes, adjust the block to deal with * encryption. Normally, we make sure the block is decrypted before writing * it. If we have crypt params, then we are writing a raw (encrypted) block, * from a raw receive. In this case, set the ARC buf's crypt params so * that the BP will be filled with the correct byteorder, salt, iv, and mac. */ static void dbuf_prepare_encrypted_dnode_leaf(dbuf_dirty_record_t *dr) { int err; dmu_buf_impl_t *db = dr->dr_dbuf; ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT); ASSERT3U(db->db_level, ==, 0); if (!db->db_objset->os_raw_receive && arc_is_encrypted(db->db_buf)) { zbookmark_phys_t zb; /* * Unfortunately, there is currently no mechanism for * syncing context to handle decryption errors. An error * here is only possible if an attacker maliciously * changed a dnode block and updated the associated * checksums going up the block tree. */ SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); err = arc_untransform(db->db_buf, db->db_objset->os_spa, &zb, B_TRUE); if (err) panic("Invalid dnode block MAC"); } else if (dr->dt.dl.dr_has_raw_params) { (void) arc_release(dr->dt.dl.dr_data, db); arc_convert_to_raw(dr->dt.dl.dr_data, dmu_objset_id(db->db_objset), dr->dt.dl.dr_byteorder, DMU_OT_DNODE, dr->dt.dl.dr_salt, dr->dt.dl.dr_iv, dr->dt.dl.dr_mac); } } /* * dbuf_sync_indirect() is called recursively from dbuf_sync_list() so it * is critical the we not allow the compiler to inline this function in to * dbuf_sync_list() thereby drastically bloating the stack usage. */ noinline static void dbuf_sync_indirect(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn; zio_t *zio; ASSERT(dmu_tx_is_syncing(tx)); dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr); mutex_enter(&db->db_mtx); ASSERT(db->db_level > 0); DBUF_VERIFY(db); /* Read the block if it hasn't been read yet. */ if (db->db_buf == NULL) { mutex_exit(&db->db_mtx); (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED); mutex_enter(&db->db_mtx); } ASSERT3U(db->db_state, ==, DB_CACHED); ASSERT(db->db_buf != NULL); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* Indirect block size must match what the dnode thinks it is. */ ASSERT3U(db->db.db_size, ==, 1<dn_phys->dn_indblkshift); dbuf_check_blkptr(dn, db); DB_DNODE_EXIT(db); /* Provide the pending dirty record to child dbufs */ db->db_data_pending = dr; mutex_exit(&db->db_mtx); dbuf_write(dr, db->db_buf, tx); zio = dr->dr_zio; mutex_enter(&dr->dt.di.dr_mtx); dbuf_sync_list(&dr->dt.di.dr_children, db->db_level - 1, tx); ASSERT(list_head(&dr->dt.di.dr_children) == NULL); mutex_exit(&dr->dt.di.dr_mtx); zio_nowait(zio); } /* * dbuf_sync_leaf() is called recursively from dbuf_sync_list() so it is * critical the we not allow the compiler to inline this function in to * dbuf_sync_list() thereby drastically bloating the stack usage. */ noinline static void dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { arc_buf_t **datap = &dr->dt.dl.dr_data; dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn; objset_t *os; uint64_t txg = tx->tx_txg; ASSERT(dmu_tx_is_syncing(tx)); dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr); mutex_enter(&db->db_mtx); /* * To be synced, we must be dirtied. But we * might have been freed after the dirty. */ if (db->db_state == DB_UNCACHED) { /* This buffer has been freed since it was dirtied */ ASSERT(db->db.db_data == NULL); } else if (db->db_state == DB_FILL) { /* This buffer was freed and is now being re-filled */ ASSERT(db->db.db_data != dr->dt.dl.dr_data); } else { ASSERT(db->db_state == DB_CACHED || db->db_state == DB_NOFILL); } DBUF_VERIFY(db); DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (db->db_blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) { /* * In the previous transaction group, the bonus buffer * was entirely used to store the attributes for the * dnode which overrode the dn_spill field. However, * when adding more attributes to the file a spill * block was required to hold the extra attributes. * * Make sure to clear the garbage left in the dn_spill * field from the previous attributes in the bonus * buffer. Otherwise, after writing out the spill * block to the new allocated dva, it will free * the old block pointed to by the invalid dn_spill. */ db->db_blkptr = NULL; } dn->dn_phys->dn_flags |= DNODE_FLAG_SPILL_BLKPTR; mutex_exit(&dn->dn_mtx); } /* * If this is a bonus buffer, simply copy the bonus data into the * dnode. It will be written out when the dnode is synced (and it * will be synced, since it must have been dirty for dbuf_sync to * be called). */ if (db->db_blkid == DMU_BONUS_BLKID) { dbuf_dirty_record_t **drp; ASSERT(*datap != NULL); ASSERT0(db->db_level); ASSERT3U(DN_MAX_BONUS_LEN(dn->dn_phys), <=, DN_SLOTS_TO_BONUSLEN(dn->dn_phys->dn_extra_slots + 1)); bcopy(*datap, DN_BONUS(dn->dn_phys), DN_MAX_BONUS_LEN(dn->dn_phys)); DB_DNODE_EXIT(db); if (*datap != db->db.db_data) { int slots = DB_DNODE(db)->dn_num_slots; int bonuslen = DN_SLOTS_TO_BONUSLEN(slots); kmem_free(*datap, bonuslen); arc_space_return(bonuslen, ARC_SPACE_BONUS); } db->db_data_pending = NULL; drp = &db->db_last_dirty; while (*drp != dr) drp = &(*drp)->dr_next; ASSERT(dr->dr_next == NULL); ASSERT(dr->dr_dbuf == db); *drp = dr->dr_next; if (dr->dr_dbuf->db_level != 0) { mutex_destroy(&dr->dt.di.dr_mtx); list_destroy(&dr->dt.di.dr_children); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; dbuf_rele_and_unlock(db, (void *)(uintptr_t)txg, B_FALSE); return; } os = dn->dn_objset; /* * This function may have dropped the db_mtx lock allowing a dmu_sync * operation to sneak in. As a result, we need to ensure that we * don't check the dr_override_state until we have returned from * dbuf_check_blkptr. */ dbuf_check_blkptr(dn, db); /* * If this buffer is in the middle of an immediate write, * wait for the synchronous IO to complete. */ while (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) { ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); cv_wait(&db->db_changed, &db->db_mtx); ASSERT(dr->dt.dl.dr_override_state != DR_NOT_OVERRIDDEN); } /* * If this is a dnode block, ensure it is appropriately encrypted * or decrypted, depending on what we are writing to it this txg. */ if (os->os_encrypted && dn->dn_object == DMU_META_DNODE_OBJECT) dbuf_prepare_encrypted_dnode_leaf(dr); if (db->db_state != DB_NOFILL && dn->dn_object != DMU_META_DNODE_OBJECT && refcount_count(&db->db_holds) > 1 && dr->dt.dl.dr_override_state != DR_OVERRIDDEN && *datap == db->db_buf) { /* * If this buffer is currently "in use" (i.e., there * are active holds and db_data still references it), * then make a copy before we start the write so that * any modifications from the open txg will not leak * into this write. * * NOTE: this copy does not need to be made for * objects only modified in the syncing context (e.g. * DNONE_DNODE blocks). */ int psize = arc_buf_size(*datap); int lsize = arc_buf_lsize(*datap); arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); enum zio_compress compress_type = arc_get_compression(*datap); if (arc_is_encrypted(*datap)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(*datap, &byteorder, salt, iv, mac); *datap = arc_alloc_raw_buf(os->os_spa, db, dmu_objset_id(os), byteorder, salt, iv, mac, dn->dn_type, psize, lsize, compress_type); } else if (compress_type != ZIO_COMPRESS_OFF) { ASSERT3U(type, ==, ARC_BUFC_DATA); *datap = arc_alloc_compressed_buf(os->os_spa, db, psize, lsize, compress_type); } else { *datap = arc_alloc_buf(os->os_spa, db, type, psize); } bcopy(db->db.db_data, (*datap)->b_data, psize); } db->db_data_pending = dr; mutex_exit(&db->db_mtx); dbuf_write(dr, *datap, tx); ASSERT(!list_link_active(&dr->dr_dirty_node)); if (dn->dn_object == DMU_META_DNODE_OBJECT) { list_insert_tail(&dn->dn_dirty_records[txg&TXG_MASK], dr); DB_DNODE_EXIT(db); } else { /* * Although zio_nowait() does not "wait for an IO", it does * initiate the IO. If this is an empty write it seems plausible * that the IO could actually be completed before the nowait * returns. We need to DB_DNODE_EXIT() first in case * zio_nowait() invalidates the dbuf. */ DB_DNODE_EXIT(db); zio_nowait(dr->dr_zio); } } void dbuf_sync_list(list_t *list, int level, dmu_tx_t *tx) { dbuf_dirty_record_t *dr; while ((dr = list_head(list))) { if (dr->dr_zio != NULL) { /* * If we find an already initialized zio then we * are processing the meta-dnode, and we have finished. * The dbufs for all dnodes are put back on the list * during processing, so that we can zio_wait() * these IOs after initiating all child IOs. */ ASSERT3U(dr->dr_dbuf->db.db_object, ==, DMU_META_DNODE_OBJECT); break; } if (dr->dr_dbuf->db_blkid != DMU_BONUS_BLKID && dr->dr_dbuf->db_blkid != DMU_SPILL_BLKID) { VERIFY3U(dr->dr_dbuf->db_level, ==, level); } list_remove(list, dr); if (dr->dr_dbuf->db_level > 0) dbuf_sync_indirect(dr, tx); else dbuf_sync_leaf(dr, tx); } } /* ARGSUSED */ static void dbuf_write_ready(zio_t *zio, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; dnode_t *dn; blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; spa_t *spa = zio->io_spa; int64_t delta; uint64_t fill = 0; int i; ASSERT3P(db->db_blkptr, !=, NULL); ASSERT3P(&db->db_data_pending->dr_bp_copy, ==, bp); DB_DNODE_ENTER(db); dn = DB_DNODE(db); delta = bp_get_dsize_sync(spa, bp) - bp_get_dsize_sync(spa, bp_orig); dnode_diduse_space(dn, delta - zio->io_prev_space_delta); zio->io_prev_space_delta = delta; if (bp->blk_birth != 0) { ASSERT((db->db_blkid != DMU_SPILL_BLKID && BP_GET_TYPE(bp) == dn->dn_type) || (db->db_blkid == DMU_SPILL_BLKID && BP_GET_TYPE(bp) == dn->dn_bonustype) || BP_IS_EMBEDDED(bp)); ASSERT(BP_GET_LEVEL(bp) == db->db_level); } mutex_enter(&db->db_mtx); #ifdef ZFS_DEBUG if (db->db_blkid == DMU_SPILL_BLKID) { ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR); ASSERT(!(BP_IS_HOLE(bp)) && db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys)); } #endif if (db->db_level == 0) { mutex_enter(&dn->dn_mtx); if (db->db_blkid > dn->dn_phys->dn_maxblkid && db->db_blkid != DMU_SPILL_BLKID) { ASSERT0(db->db_objset->os_raw_receive); dn->dn_phys->dn_maxblkid = db->db_blkid; } mutex_exit(&dn->dn_mtx); if (dn->dn_type == DMU_OT_DNODE) { i = 0; while (i < db->db.db_size) { dnode_phys_t *dnp = (void *)(((char *)db->db.db_data) + i); i += DNODE_MIN_SIZE; if (dnp->dn_type != DMU_OT_NONE) { fill++; i += dnp->dn_extra_slots * DNODE_MIN_SIZE; } } } else { if (BP_IS_HOLE(bp)) { fill = 0; } else { fill = 1; } } } else { blkptr_t *ibp = db->db.db_data; ASSERT3U(db->db.db_size, ==, 1<dn_phys->dn_indblkshift); for (i = db->db.db_size >> SPA_BLKPTRSHIFT; i > 0; i--, ibp++) { if (BP_IS_HOLE(ibp)) continue; fill += BP_GET_FILL(ibp); } } DB_DNODE_EXIT(db); if (!BP_IS_EMBEDDED(bp)) BP_SET_FILL(bp, fill); mutex_exit(&db->db_mtx); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); *db->db_blkptr = *bp; rw_exit(&dn->dn_struct_rwlock); } /* ARGSUSED */ /* * This function gets called just prior to running through the compression * stage of the zio pipeline. If we're an indirect block comprised of only * holes, then we want this indirect to be compressed away to a hole. In * order to do that we must zero out any information about the holes that * this indirect points to prior to before we try to compress it. */ static void dbuf_write_children_ready(zio_t *zio, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; dnode_t *dn; blkptr_t *bp; unsigned int epbs, i; ASSERT3U(db->db_level, >, 0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT3U(epbs, <, 31); /* Determine if all our children are holes */ for (i = 0, bp = db->db.db_data; i < 1ULL << epbs; i++, bp++) { if (!BP_IS_HOLE(bp)) break; } /* * If all the children are holes, then zero them all out so that * we may get compressed away. */ if (i == 1ULL << epbs) { /* * We only found holes. Grab the rwlock to prevent * anybody from reading the blocks we're about to * zero out. */ rw_enter(&dn->dn_struct_rwlock, RW_WRITER); bzero(db->db.db_data, db->db.db_size); rw_exit(&dn->dn_struct_rwlock); } DB_DNODE_EXIT(db); } /* * The SPA will call this callback several times for each zio - once * for every physical child i/o (zio->io_phys_children times). This * allows the DMU to monitor the progress of each logical i/o. For example, * there may be 2 copies of an indirect block, or many fragments of a RAID-Z * block. There may be a long delay before all copies/fragments are completed, * so this callback allows us to retire dirty space gradually, as the physical * i/os complete. */ /* ARGSUSED */ static void dbuf_write_physdone(zio_t *zio, arc_buf_t *buf, void *arg) { dmu_buf_impl_t *db = arg; objset_t *os = db->db_objset; dsl_pool_t *dp = dmu_objset_pool(os); dbuf_dirty_record_t *dr; int delta = 0; dr = db->db_data_pending; ASSERT3U(dr->dr_txg, ==, zio->io_txg); /* * The callback will be called io_phys_children times. Retire one * portion of our dirty space each time we are called. Any rounding * error will be cleaned up by dsl_pool_sync()'s call to * dsl_pool_undirty_space(). */ delta = dr->dr_accounted / zio->io_phys_children; dsl_pool_undirty_space(dp, delta, zio->io_txg); } /* ARGSUSED */ static void dbuf_write_done(zio_t *zio, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; blkptr_t *bp_orig = &zio->io_bp_orig; blkptr_t *bp = db->db_blkptr; objset_t *os = db->db_objset; dmu_tx_t *tx = os->os_synctx; dbuf_dirty_record_t **drp, *dr; ASSERT0(zio->io_error); ASSERT(db->db_blkptr == bp); /* * For nopwrites and rewrites we ensure that the bp matches our * original and bypass all the accounting. */ if (zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)) { ASSERT(BP_EQUAL(bp, bp_orig)); } else { dsl_dataset_t *ds = os->os_dsl_dataset; (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); dsl_dataset_block_born(ds, bp, tx); } mutex_enter(&db->db_mtx); DBUF_VERIFY(db); drp = &db->db_last_dirty; while ((dr = *drp) != db->db_data_pending) drp = &dr->dr_next; ASSERT(!list_link_active(&dr->dr_dirty_node)); ASSERT(dr->dr_dbuf == db); ASSERT(dr->dr_next == NULL); *drp = dr->dr_next; #ifdef ZFS_DEBUG if (db->db_blkid == DMU_SPILL_BLKID) { dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR); ASSERT(!(BP_IS_HOLE(db->db_blkptr)) && db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys)); DB_DNODE_EXIT(db); } #endif if (db->db_level == 0) { ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN); if (db->db_state != DB_NOFILL) { if (dr->dt.dl.dr_data != db->db_buf) arc_buf_destroy(dr->dt.dl.dr_data, db); } } else { dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); ASSERT(list_head(&dr->dt.di.dr_children) == NULL); ASSERT3U(db->db.db_size, ==, 1 << dn->dn_phys->dn_indblkshift); if (!BP_IS_HOLE(db->db_blkptr)) { ASSERTV(int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT); ASSERT3U(db->db_blkid, <=, dn->dn_phys->dn_maxblkid >> (db->db_level * epbs)); ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==, db->db.db_size); } DB_DNODE_EXIT(db); mutex_destroy(&dr->dt.di.dr_mtx); list_destroy(&dr->dt.di.dr_children); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); cv_broadcast(&db->db_changed); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; db->db_data_pending = NULL; dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg, B_FALSE); } static void dbuf_write_nofill_ready(zio_t *zio) { dbuf_write_ready(zio, NULL, zio->io_private); } static void dbuf_write_nofill_done(zio_t *zio) { dbuf_write_done(zio, NULL, zio->io_private); } static void dbuf_write_override_ready(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; dmu_buf_impl_t *db = dr->dr_dbuf; dbuf_write_ready(zio, NULL, db); } static void dbuf_write_override_done(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; dmu_buf_impl_t *db = dr->dr_dbuf; blkptr_t *obp = &dr->dt.dl.dr_overridden_by; mutex_enter(&db->db_mtx); if (!BP_EQUAL(zio->io_bp, obp)) { if (!BP_IS_HOLE(obp)) dsl_free(spa_get_dsl(zio->io_spa), zio->io_txg, obp); arc_release(dr->dt.dl.dr_data, db); } mutex_exit(&db->db_mtx); dbuf_write_done(zio, NULL, db); if (zio->io_abd != NULL) abd_put(zio->io_abd); } typedef struct dbuf_remap_impl_callback_arg { objset_t *drica_os; uint64_t drica_blk_birth; dmu_tx_t *drica_tx; } dbuf_remap_impl_callback_arg_t; static void dbuf_remap_impl_callback(uint64_t vdev, uint64_t offset, uint64_t size, void *arg) { dbuf_remap_impl_callback_arg_t *drica = arg; objset_t *os = drica->drica_os; spa_t *spa = dmu_objset_spa(os); dmu_tx_t *tx = drica->drica_tx; ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); if (os == spa_meta_objset(spa)) { spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx); } else { dsl_dataset_block_remapped(dmu_objset_ds(os), vdev, offset, size, drica->drica_blk_birth, tx); } } static void dbuf_remap_impl(dnode_t *dn, blkptr_t *bp, dmu_tx_t *tx) { blkptr_t bp_copy = *bp; spa_t *spa = dmu_objset_spa(dn->dn_objset); dbuf_remap_impl_callback_arg_t drica; ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); drica.drica_os = dn->dn_objset; drica.drica_blk_birth = bp->blk_birth; drica.drica_tx = tx; if (spa_remap_blkptr(spa, &bp_copy, dbuf_remap_impl_callback, &drica)) { /* * The struct_rwlock prevents dbuf_read_impl() from * dereferencing the BP while we are changing it. To * avoid lock contention, only grab it when we are actually * changing the BP. */ rw_enter(&dn->dn_struct_rwlock, RW_WRITER); *bp = bp_copy; rw_exit(&dn->dn_struct_rwlock); } } /* * Returns true if a dbuf_remap would modify the dbuf. We do this by attempting * to remap a copy of every bp in the dbuf. */ boolean_t dbuf_can_remap(const dmu_buf_impl_t *db) { spa_t *spa = dmu_objset_spa(db->db_objset); blkptr_t *bp = db->db.db_data; boolean_t ret = B_FALSE; ASSERT3U(db->db_level, >, 0); ASSERT3S(db->db_state, ==, DB_CACHED); ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) { blkptr_t bp_copy = bp[i]; if (spa_remap_blkptr(spa, &bp_copy, NULL, NULL)) { ret = B_TRUE; break; } } spa_config_exit(spa, SCL_VDEV, FTAG); return (ret); } boolean_t dnode_needs_remap(const dnode_t *dn) { spa_t *spa = dmu_objset_spa(dn->dn_objset); boolean_t ret = B_FALSE; if (dn->dn_phys->dn_nlevels == 0) { return (B_FALSE); } ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); for (int j = 0; j < dn->dn_phys->dn_nblkptr; j++) { blkptr_t bp_copy = dn->dn_phys->dn_blkptr[j]; if (spa_remap_blkptr(spa, &bp_copy, NULL, NULL)) { ret = B_TRUE; break; } } spa_config_exit(spa, SCL_VDEV, FTAG); return (ret); } /* * Remap any existing BP's to concrete vdevs, if possible. */ static void dbuf_remap(dnode_t *dn, dmu_buf_impl_t *db, dmu_tx_t *tx) { spa_t *spa = dmu_objset_spa(db->db_objset); ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); if (!spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)) return; if (db->db_level > 0) { blkptr_t *bp = db->db.db_data; for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) { dbuf_remap_impl(dn, &bp[i], tx); } } else if (db->db.db_object == DMU_META_DNODE_OBJECT) { dnode_phys_t *dnp = db->db.db_data; ASSERT3U(db->db_dnode_handle->dnh_dnode->dn_type, ==, DMU_OT_DNODE); for (int i = 0; i < db->db.db_size >> DNODE_SHIFT; i += dnp[i].dn_extra_slots + 1) { for (int j = 0; j < dnp[i].dn_nblkptr; j++) { dbuf_remap_impl(dn, &dnp[i].dn_blkptr[j], tx); } } } } /* Issue I/O to commit a dirty buffer to disk. */ static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn; objset_t *os; dmu_buf_impl_t *parent = db->db_parent; uint64_t txg = tx->tx_txg; zbookmark_phys_t zb; zio_prop_t zp; zio_t *zio; int wp_flag = 0; ASSERT(dmu_tx_is_syncing(tx)); DB_DNODE_ENTER(db); dn = DB_DNODE(db); os = dn->dn_objset; if (db->db_state != DB_NOFILL) { if (db->db_level > 0 || dn->dn_type == DMU_OT_DNODE) { /* * Private object buffers are released here rather * than in dbuf_dirty() since they are only modified * in the syncing context and we don't want the * overhead of making multiple copies of the data. */ if (BP_IS_HOLE(db->db_blkptr)) { arc_buf_thaw(data); } else { dbuf_release_bp(db); } dbuf_remap(dn, db, tx); } } if (parent != dn->dn_dbuf) { /* Our parent is an indirect block. */ /* We have a dirty parent that has been scheduled for write. */ ASSERT(parent && parent->db_data_pending); /* Our parent's buffer is one level closer to the dnode. */ ASSERT(db->db_level == parent->db_level-1); /* * We're about to modify our parent's db_data by modifying * our block pointer, so the parent must be released. */ ASSERT(arc_released(parent->db_buf)); zio = parent->db_data_pending->dr_zio; } else { /* Our parent is the dnode itself. */ ASSERT((db->db_level == dn->dn_phys->dn_nlevels-1 && db->db_blkid != DMU_SPILL_BLKID) || (db->db_blkid == DMU_SPILL_BLKID && db->db_level == 0)); if (db->db_blkid != DMU_SPILL_BLKID) ASSERT3P(db->db_blkptr, ==, &dn->dn_phys->dn_blkptr[db->db_blkid]); zio = dn->dn_zio; } ASSERT(db->db_level == 0 || data == db->db_buf); ASSERT3U(db->db_blkptr->blk_birth, <=, txg); ASSERT(zio); SET_BOOKMARK(&zb, os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : DMU_META_OBJSET, db->db.db_object, db->db_level, db->db_blkid); if (db->db_blkid == DMU_SPILL_BLKID) wp_flag = WP_SPILL; wp_flag |= (db->db_state == DB_NOFILL) ? WP_NOFILL : 0; dmu_write_policy(os, dn, db->db_level, wp_flag, &zp); DB_DNODE_EXIT(db); /* * We copy the blkptr now (rather than when we instantiate the dirty * record), because its value can change between open context and * syncing context. We do not need to hold dn_struct_rwlock to read * db_blkptr because we are in syncing context. */ dr->dr_bp_copy = *db->db_blkptr; if (db->db_level == 0 && dr->dt.dl.dr_override_state == DR_OVERRIDDEN) { /* * The BP for this block has been provided by open context * (by dmu_sync() or dmu_buf_write_embedded()). */ abd_t *contents = (data != NULL) ? abd_get_from_buf(data->b_data, arc_buf_size(data)) : NULL; dr->dr_zio = zio_write(zio, os->os_spa, txg, &dr->dr_bp_copy, contents, db->db.db_size, db->db.db_size, &zp, dbuf_write_override_ready, NULL, NULL, dbuf_write_override_done, dr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); mutex_enter(&db->db_mtx); dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; zio_write_override(dr->dr_zio, &dr->dt.dl.dr_overridden_by, dr->dt.dl.dr_copies, dr->dt.dl.dr_nopwrite); mutex_exit(&db->db_mtx); } else if (db->db_state == DB_NOFILL) { ASSERT(zp.zp_checksum == ZIO_CHECKSUM_OFF || zp.zp_checksum == ZIO_CHECKSUM_NOPARITY); dr->dr_zio = zio_write(zio, os->os_spa, txg, &dr->dr_bp_copy, NULL, db->db.db_size, db->db.db_size, &zp, dbuf_write_nofill_ready, NULL, NULL, dbuf_write_nofill_done, db, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED | ZIO_FLAG_NODATA, &zb); } else { ASSERT(arc_released(data)); /* * For indirect blocks, we want to setup the children * ready callback so that we can properly handle an indirect * block that only contains holes. */ arc_write_done_func_t *children_ready_cb = NULL; if (db->db_level != 0) children_ready_cb = dbuf_write_children_ready; dr->dr_zio = arc_write(zio, os->os_spa, txg, &dr->dr_bp_copy, data, DBUF_IS_L2CACHEABLE(db), &zp, dbuf_write_ready, children_ready_cb, dbuf_write_physdone, dbuf_write_done, db, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); } } #if defined(_KERNEL) EXPORT_SYMBOL(dbuf_find); EXPORT_SYMBOL(dbuf_is_metadata); EXPORT_SYMBOL(dbuf_destroy); EXPORT_SYMBOL(dbuf_loan_arcbuf); EXPORT_SYMBOL(dbuf_whichblock); EXPORT_SYMBOL(dbuf_read); EXPORT_SYMBOL(dbuf_unoverride); EXPORT_SYMBOL(dbuf_free_range); EXPORT_SYMBOL(dbuf_new_size); EXPORT_SYMBOL(dbuf_release_bp); EXPORT_SYMBOL(dbuf_dirty); EXPORT_SYMBOL(dmu_buf_set_crypt_params); EXPORT_SYMBOL(dmu_buf_will_dirty); EXPORT_SYMBOL(dmu_buf_will_not_fill); EXPORT_SYMBOL(dmu_buf_will_fill); EXPORT_SYMBOL(dmu_buf_fill_done); EXPORT_SYMBOL(dmu_buf_rele); EXPORT_SYMBOL(dbuf_assign_arcbuf); EXPORT_SYMBOL(dbuf_prefetch); EXPORT_SYMBOL(dbuf_hold_impl); EXPORT_SYMBOL(dbuf_hold); EXPORT_SYMBOL(dbuf_hold_level); EXPORT_SYMBOL(dbuf_create_bonus); EXPORT_SYMBOL(dbuf_spill_set_blksz); EXPORT_SYMBOL(dbuf_rm_spill); EXPORT_SYMBOL(dbuf_add_ref); EXPORT_SYMBOL(dbuf_rele); EXPORT_SYMBOL(dbuf_rele_and_unlock); EXPORT_SYMBOL(dbuf_refcount); EXPORT_SYMBOL(dbuf_sync_list); EXPORT_SYMBOL(dmu_buf_set_user); EXPORT_SYMBOL(dmu_buf_set_user_ie); EXPORT_SYMBOL(dmu_buf_get_user); EXPORT_SYMBOL(dmu_buf_get_blkptr); /* BEGIN CSTYLED */ module_param(dbuf_cache_max_bytes, ulong, 0644); MODULE_PARM_DESC(dbuf_cache_max_bytes, "Maximum size in bytes of the dbuf cache."); module_param(dbuf_cache_hiwater_pct, uint, 0644); MODULE_PARM_DESC(dbuf_cache_hiwater_pct, "Percentage over dbuf_cache_max_bytes when dbufs must be evicted " "directly."); module_param(dbuf_cache_lowater_pct, uint, 0644); MODULE_PARM_DESC(dbuf_cache_lowater_pct, "Percentage below dbuf_cache_max_bytes when the evict thread stops " "evicting dbufs."); module_param(dbuf_metadata_cache_max_bytes, ulong, 0644); MODULE_PARM_DESC(dbuf_metadata_cache_max_bytes, "Maximum size in bytes of the dbuf metadata cache."); module_param(dbuf_cache_shift, int, 0644); MODULE_PARM_DESC(dbuf_cache_shift, "Set the size of the dbuf cache to a log2 fraction of arc size."); module_param(dbuf_metadata_cache_shift, int, 0644); MODULE_PARM_DESC(dbuf_cache_shift, "Set the size of the dbuf metadata cache to a log2 fraction of " "arc size."); /* END CSTYLED */ #endif diff --git a/module/zfs/zio.c b/module/zfs/zio.c index a8f6a928fdc2..654c81ef9b71 100644 --- a/module/zfs/zio.c +++ b/module/zfs/zio.c @@ -1,4776 +1,4779 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2018 by Delphix. All rights reserved. * Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * ========================================================================== * I/O type descriptions * ========================================================================== */ const char *zio_type_name[ZIO_TYPES] = { /* * Note: Linux kernel thread name length is limited * so these names will differ from upstream open zfs. */ "z_null", "z_rd", "z_wr", "z_fr", "z_cl", "z_ioctl" }; int zio_dva_throttle_enabled = B_TRUE; /* * ========================================================================== * I/O kmem caches * ========================================================================== */ kmem_cache_t *zio_cache; kmem_cache_t *zio_link_cache; kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; #if defined(ZFS_DEBUG) && !defined(_KERNEL) uint64_t zio_buf_cache_allocs[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; uint64_t zio_buf_cache_frees[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; #endif int zio_delay_max = ZIO_DELAY_MAX; #define BP_SPANB(indblkshift, level) \ (((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT))) #define COMPARE_META_LEVEL 0x80000000ul /* * The following actions directly effect the spa's sync-to-convergence logic. * The values below define the sync pass when we start performing the action. * Care should be taken when changing these values as they directly impact * spa_sync() performance. Tuning these values may introduce subtle performance * pathologies and should only be done in the context of performance analysis. * These tunables will eventually be removed and replaced with #defines once * enough analysis has been done to determine optimal values. * * The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that * regular blocks are not deferred. */ int zfs_sync_pass_deferred_free = 2; /* defer frees starting in this pass */ int zfs_sync_pass_dont_compress = 5; /* don't compress starting in this pass */ int zfs_sync_pass_rewrite = 2; /* rewrite new bps starting in this pass */ /* * An allocating zio is one that either currently has the DVA allocate * stage set or will have it later in its lifetime. */ #define IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE) int zio_requeue_io_start_cut_in_line = 1; #ifdef ZFS_DEBUG int zio_buf_debug_limit = 16384; #else int zio_buf_debug_limit = 0; #endif static inline void __zio_execute(zio_t *zio); static void zio_taskq_dispatch(zio_t *, zio_taskq_type_t, boolean_t); void zio_init(void) { size_t c; vmem_t *data_alloc_arena = NULL; zio_cache = kmem_cache_create("zio_cache", sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0); zio_link_cache = kmem_cache_create("zio_link_cache", sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0); /* * For small buffers, we want a cache for each multiple of * SPA_MINBLOCKSIZE. For larger buffers, we want a cache * for each quarter-power of 2. */ for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) { size_t size = (c + 1) << SPA_MINBLOCKSHIFT; size_t p2 = size; size_t align = 0; size_t cflags = (size > zio_buf_debug_limit) ? KMC_NODEBUG : 0; #if defined(_ILP32) && defined(_KERNEL) /* * Cache size limited to 1M on 32-bit platforms until ARC * buffers no longer require virtual address space. */ if (size > zfs_max_recordsize) break; #endif while (!ISP2(p2)) p2 &= p2 - 1; #ifndef _KERNEL /* * If we are using watchpoints, put each buffer on its own page, * to eliminate the performance overhead of trapping to the * kernel when modifying a non-watched buffer that shares the * page with a watched buffer. */ if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE)) continue; /* * Here's the problem - on 4K native devices in userland on * Linux using O_DIRECT, buffers must be 4K aligned or I/O * will fail with EINVAL, causing zdb (and others) to coredump. * Since userland probably doesn't need optimized buffer caches, * we just force 4K alignment on everything. */ align = 8 * SPA_MINBLOCKSIZE; #else if (size < PAGESIZE) { align = SPA_MINBLOCKSIZE; } else if (IS_P2ALIGNED(size, p2 >> 2)) { align = PAGESIZE; } #endif if (align != 0) { char name[36]; (void) sprintf(name, "zio_buf_%lu", (ulong_t)size); zio_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, cflags); (void) sprintf(name, "zio_data_buf_%lu", (ulong_t)size); zio_data_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, data_alloc_arena, cflags); } } while (--c != 0) { ASSERT(zio_buf_cache[c] != NULL); if (zio_buf_cache[c - 1] == NULL) zio_buf_cache[c - 1] = zio_buf_cache[c]; ASSERT(zio_data_buf_cache[c] != NULL); if (zio_data_buf_cache[c - 1] == NULL) zio_data_buf_cache[c - 1] = zio_data_buf_cache[c]; } zio_inject_init(); lz4_init(); } void zio_fini(void) { size_t c; kmem_cache_t *last_cache = NULL; kmem_cache_t *last_data_cache = NULL; for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) { #ifdef _ILP32 /* * Cache size limited to 1M on 32-bit platforms until ARC * buffers no longer require virtual address space. */ if (((c + 1) << SPA_MINBLOCKSHIFT) > zfs_max_recordsize) break; #endif #if defined(ZFS_DEBUG) && !defined(_KERNEL) if (zio_buf_cache_allocs[c] != zio_buf_cache_frees[c]) (void) printf("zio_fini: [%d] %llu != %llu\n", (int)((c + 1) << SPA_MINBLOCKSHIFT), (long long unsigned)zio_buf_cache_allocs[c], (long long unsigned)zio_buf_cache_frees[c]); #endif if (zio_buf_cache[c] != last_cache) { last_cache = zio_buf_cache[c]; kmem_cache_destroy(zio_buf_cache[c]); } zio_buf_cache[c] = NULL; if (zio_data_buf_cache[c] != last_data_cache) { last_data_cache = zio_data_buf_cache[c]; kmem_cache_destroy(zio_data_buf_cache[c]); } zio_data_buf_cache[c] = NULL; } kmem_cache_destroy(zio_link_cache); kmem_cache_destroy(zio_cache); zio_inject_fini(); lz4_fini(); } /* * ========================================================================== * Allocate and free I/O buffers * ========================================================================== */ /* * Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a * crashdump if the kernel panics, so use it judiciously. Obviously, it's * useful to inspect ZFS metadata, but if possible, we should avoid keeping * excess / transient data in-core during a crashdump. */ void * zio_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); #if defined(ZFS_DEBUG) && !defined(_KERNEL) atomic_add_64(&zio_buf_cache_allocs[c], 1); #endif return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE)); } /* * Use zio_data_buf_alloc to allocate data. The data will not appear in a * crashdump if the kernel panics. This exists so that we will limit the amount * of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount * of kernel heap dumped to disk when the kernel panics) */ void * zio_data_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE)); } void zio_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); #if defined(ZFS_DEBUG) && !defined(_KERNEL) atomic_add_64(&zio_buf_cache_frees[c], 1); #endif kmem_cache_free(zio_buf_cache[c], buf); } void zio_data_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); kmem_cache_free(zio_data_buf_cache[c], buf); } static void zio_abd_free(void *abd, size_t size) { abd_free((abd_t *)abd); } /* * ========================================================================== * Push and pop I/O transform buffers * ========================================================================== */ void zio_push_transform(zio_t *zio, abd_t *data, uint64_t size, uint64_t bufsize, zio_transform_func_t *transform) { zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP); /* * Ensure that anyone expecting this zio to contain a linear ABD isn't * going to get a nasty surprise when they try to access the data. */ IMPLY(abd_is_linear(zio->io_abd), abd_is_linear(data)); zt->zt_orig_abd = zio->io_abd; zt->zt_orig_size = zio->io_size; zt->zt_bufsize = bufsize; zt->zt_transform = transform; zt->zt_next = zio->io_transform_stack; zio->io_transform_stack = zt; zio->io_abd = data; zio->io_size = size; } void zio_pop_transforms(zio_t *zio) { zio_transform_t *zt; while ((zt = zio->io_transform_stack) != NULL) { if (zt->zt_transform != NULL) zt->zt_transform(zio, zt->zt_orig_abd, zt->zt_orig_size); if (zt->zt_bufsize != 0) abd_free(zio->io_abd); zio->io_abd = zt->zt_orig_abd; zio->io_size = zt->zt_orig_size; zio->io_transform_stack = zt->zt_next; kmem_free(zt, sizeof (zio_transform_t)); } } /* * ========================================================================== * I/O transform callbacks for subblocks, decompression, and decryption * ========================================================================== */ static void zio_subblock(zio_t *zio, abd_t *data, uint64_t size) { ASSERT(zio->io_size > size); if (zio->io_type == ZIO_TYPE_READ) abd_copy(data, zio->io_abd, size); } static void zio_decompress(zio_t *zio, abd_t *data, uint64_t size) { if (zio->io_error == 0) { void *tmp = abd_borrow_buf(data, size); int ret = zio_decompress_data(BP_GET_COMPRESS(zio->io_bp), zio->io_abd, tmp, zio->io_size, size); abd_return_buf_copy(data, tmp, size); + if (zio_injection_enabled && ret == 0) + ret = zio_handle_fault_injection(zio, EINVAL); + if (ret != 0) zio->io_error = SET_ERROR(EIO); } } static void zio_decrypt(zio_t *zio, abd_t *data, uint64_t size) { int ret; void *tmp; blkptr_t *bp = zio->io_bp; spa_t *spa = zio->io_spa; uint64_t dsobj = zio->io_bookmark.zb_objset; uint64_t lsize = BP_GET_LSIZE(bp); dmu_object_type_t ot = BP_GET_TYPE(bp); uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; boolean_t no_crypt = B_FALSE; ASSERT(BP_USES_CRYPT(bp)); ASSERT3U(size, !=, 0); if (zio->io_error != 0) return; /* * Verify the cksum of MACs stored in an indirect bp. It will always * be possible to verify this since it does not require an encryption * key. */ if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) { zio_crypt_decode_mac_bp(bp, mac); if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) { /* * We haven't decompressed the data yet, but * zio_crypt_do_indirect_mac_checksum() requires * decompressed data to be able to parse out the MACs * from the indirect block. We decompress it now and * throw away the result after we are finished. */ tmp = zio_buf_alloc(lsize); ret = zio_decompress_data(BP_GET_COMPRESS(bp), zio->io_abd, tmp, zio->io_size, lsize); if (ret != 0) { ret = SET_ERROR(EIO); goto error; } ret = zio_crypt_do_indirect_mac_checksum(B_FALSE, tmp, lsize, BP_SHOULD_BYTESWAP(bp), mac); zio_buf_free(tmp, lsize); } else { ret = zio_crypt_do_indirect_mac_checksum_abd(B_FALSE, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp), mac); } abd_copy(data, zio->io_abd, size); if (zio_injection_enabled && ot != DMU_OT_DNODE && ret == 0) { ret = zio_handle_decrypt_injection(spa, &zio->io_bookmark, ot, ECKSUM); } if (ret != 0) goto error; return; } /* * If this is an authenticated block, just check the MAC. It would be * nice to separate this out into its own flag, but for the moment * enum zio_flag is out of bits. */ if (BP_IS_AUTHENTICATED(bp)) { if (ot == DMU_OT_OBJSET) { ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa, dsobj, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp)); } else { zio_crypt_decode_mac_bp(bp, mac); ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj, zio->io_abd, size, mac); if (zio_injection_enabled && ret == 0) { ret = zio_handle_decrypt_injection(spa, &zio->io_bookmark, ot, ECKSUM); } } abd_copy(data, zio->io_abd, size); if (ret != 0) goto error; return; } zio_crypt_decode_params_bp(bp, salt, iv); if (ot == DMU_OT_INTENT_LOG) { tmp = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t)); zio_crypt_decode_mac_zil(tmp, mac); abd_return_buf(zio->io_abd, tmp, sizeof (zil_chain_t)); } else { zio_crypt_decode_mac_bp(bp, mac); } ret = spa_do_crypt_abd(B_FALSE, spa, &zio->io_bookmark, BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, size, data, zio->io_abd, &no_crypt); if (no_crypt) abd_copy(data, zio->io_abd, size); if (ret != 0) goto error; return; error: /* assert that the key was found unless this was speculative */ ASSERT(ret != EACCES || (zio->io_flags & ZIO_FLAG_SPECULATIVE)); /* * If there was a decryption / authentication error return EIO as * the io_error. If this was not a speculative zio, create an ereport. */ if (ret == ECKSUM) { zio->io_error = SET_ERROR(EIO); if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) { spa_log_error(spa, &zio->io_bookmark); zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION, spa, NULL, &zio->io_bookmark, zio, 0, 0); } } else { zio->io_error = ret; } } /* * ========================================================================== * I/O parent/child relationships and pipeline interlocks * ========================================================================== */ zio_t * zio_walk_parents(zio_t *cio, zio_link_t **zl) { list_t *pl = &cio->io_parent_list; *zl = (*zl == NULL) ? list_head(pl) : list_next(pl, *zl); if (*zl == NULL) return (NULL); ASSERT((*zl)->zl_child == cio); return ((*zl)->zl_parent); } zio_t * zio_walk_children(zio_t *pio, zio_link_t **zl) { list_t *cl = &pio->io_child_list; ASSERT(MUTEX_HELD(&pio->io_lock)); *zl = (*zl == NULL) ? list_head(cl) : list_next(cl, *zl); if (*zl == NULL) return (NULL); ASSERT((*zl)->zl_parent == pio); return ((*zl)->zl_child); } zio_t * zio_unique_parent(zio_t *cio) { zio_link_t *zl = NULL; zio_t *pio = zio_walk_parents(cio, &zl); VERIFY3P(zio_walk_parents(cio, &zl), ==, NULL); return (pio); } void zio_add_child(zio_t *pio, zio_t *cio) { zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP); /* * Logical I/Os can have logical, gang, or vdev children. * Gang I/Os can have gang or vdev children. * Vdev I/Os can only have vdev children. * The following ASSERT captures all of these constraints. */ ASSERT3S(cio->io_child_type, <=, pio->io_child_type); zl->zl_parent = pio; zl->zl_child = cio; mutex_enter(&pio->io_lock); mutex_enter(&cio->io_lock); ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0); for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_children[cio->io_child_type][w] += !cio->io_state[w]; list_insert_head(&pio->io_child_list, zl); list_insert_head(&cio->io_parent_list, zl); pio->io_child_count++; cio->io_parent_count++; mutex_exit(&cio->io_lock); mutex_exit(&pio->io_lock); } static void zio_remove_child(zio_t *pio, zio_t *cio, zio_link_t *zl) { ASSERT(zl->zl_parent == pio); ASSERT(zl->zl_child == cio); mutex_enter(&pio->io_lock); mutex_enter(&cio->io_lock); list_remove(&pio->io_child_list, zl); list_remove(&cio->io_parent_list, zl); pio->io_child_count--; cio->io_parent_count--; mutex_exit(&cio->io_lock); mutex_exit(&pio->io_lock); kmem_cache_free(zio_link_cache, zl); } static boolean_t zio_wait_for_children(zio_t *zio, uint8_t childbits, enum zio_wait_type wait) { boolean_t waiting = B_FALSE; mutex_enter(&zio->io_lock); ASSERT(zio->io_stall == NULL); for (int c = 0; c < ZIO_CHILD_TYPES; c++) { if (!(ZIO_CHILD_BIT_IS_SET(childbits, c))) continue; uint64_t *countp = &zio->io_children[c][wait]; if (*countp != 0) { zio->io_stage >>= 1; ASSERT3U(zio->io_stage, !=, ZIO_STAGE_OPEN); zio->io_stall = countp; waiting = B_TRUE; break; } } mutex_exit(&zio->io_lock); return (waiting); } __attribute__((always_inline)) static inline void zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait, zio_t **next_to_executep) { uint64_t *countp = &pio->io_children[zio->io_child_type][wait]; int *errorp = &pio->io_child_error[zio->io_child_type]; mutex_enter(&pio->io_lock); if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE)) *errorp = zio_worst_error(*errorp, zio->io_error); pio->io_reexecute |= zio->io_reexecute; ASSERT3U(*countp, >, 0); (*countp)--; if (*countp == 0 && pio->io_stall == countp) { zio_taskq_type_t type = pio->io_stage < ZIO_STAGE_VDEV_IO_START ? ZIO_TASKQ_ISSUE : ZIO_TASKQ_INTERRUPT; pio->io_stall = NULL; mutex_exit(&pio->io_lock); /* * If we can tell the caller to execute this parent next, do * so. Otherwise dispatch the parent zio as its own task. * * Having the caller execute the parent when possible reduces * locking on the zio taskq's, reduces context switch * overhead, and has no recursion penalty. Note that one * read from disk typically causes at least 3 zio's: a * zio_null(), the logical zio_read(), and then a physical * zio. When the physical ZIO completes, we are able to call * zio_done() on all 3 of these zio's from one invocation of * zio_execute() by returning the parent back to * zio_execute(). Since the parent isn't executed until this * thread returns back to zio_execute(), the caller should do * so promptly. * * In other cases, dispatching the parent prevents * overflowing the stack when we have deeply nested * parent-child relationships, as we do with the "mega zio" * of writes for spa_sync(), and the chain of ZIL blocks. */ if (next_to_executep != NULL && *next_to_executep == NULL) { *next_to_executep = pio; } else { zio_taskq_dispatch(pio, type, B_FALSE); } } else { mutex_exit(&pio->io_lock); } } static void zio_inherit_child_errors(zio_t *zio, enum zio_child c) { if (zio->io_child_error[c] != 0 && zio->io_error == 0) zio->io_error = zio->io_child_error[c]; } int zio_bookmark_compare(const void *x1, const void *x2) { const zio_t *z1 = x1; const zio_t *z2 = x2; if (z1->io_bookmark.zb_objset < z2->io_bookmark.zb_objset) return (-1); if (z1->io_bookmark.zb_objset > z2->io_bookmark.zb_objset) return (1); if (z1->io_bookmark.zb_object < z2->io_bookmark.zb_object) return (-1); if (z1->io_bookmark.zb_object > z2->io_bookmark.zb_object) return (1); if (z1->io_bookmark.zb_level < z2->io_bookmark.zb_level) return (-1); if (z1->io_bookmark.zb_level > z2->io_bookmark.zb_level) return (1); if (z1->io_bookmark.zb_blkid < z2->io_bookmark.zb_blkid) return (-1); if (z1->io_bookmark.zb_blkid > z2->io_bookmark.zb_blkid) return (1); if (z1 < z2) return (-1); if (z1 > z2) return (1); return (0); } /* * ========================================================================== * Create the various types of I/O (read, write, free, etc) * ========================================================================== */ static zio_t * zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, abd_t *data, uint64_t lsize, uint64_t psize, zio_done_func_t *done, void *private, zio_type_t type, zio_priority_t priority, enum zio_flag flags, vdev_t *vd, uint64_t offset, const zbookmark_phys_t *zb, enum zio_stage stage, enum zio_stage pipeline) { zio_t *zio; ASSERT3U(psize, <=, SPA_MAXBLOCKSIZE); ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0); ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0); ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER)); ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER)); ASSERT(vd || stage == ZIO_STAGE_OPEN); IMPLY(lsize != psize, (flags & ZIO_FLAG_RAW_COMPRESS) != 0); zio = kmem_cache_alloc(zio_cache, KM_SLEEP); bzero(zio, sizeof (zio_t)); mutex_init(&zio->io_lock, NULL, MUTEX_NOLOCKDEP, NULL); cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL); list_create(&zio->io_parent_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_parent_node)); list_create(&zio->io_child_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_child_node)); metaslab_trace_init(&zio->io_alloc_list); if (vd != NULL) zio->io_child_type = ZIO_CHILD_VDEV; else if (flags & ZIO_FLAG_GANG_CHILD) zio->io_child_type = ZIO_CHILD_GANG; else if (flags & ZIO_FLAG_DDT_CHILD) zio->io_child_type = ZIO_CHILD_DDT; else zio->io_child_type = ZIO_CHILD_LOGICAL; if (bp != NULL) { zio->io_bp = (blkptr_t *)bp; zio->io_bp_copy = *bp; zio->io_bp_orig = *bp; if (type != ZIO_TYPE_WRITE || zio->io_child_type == ZIO_CHILD_DDT) zio->io_bp = &zio->io_bp_copy; /* so caller can free */ if (zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_logical = zio; if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp)) pipeline |= ZIO_GANG_STAGES; } zio->io_spa = spa; zio->io_txg = txg; zio->io_done = done; zio->io_private = private; zio->io_type = type; zio->io_priority = priority; zio->io_vd = vd; zio->io_offset = offset; zio->io_orig_abd = zio->io_abd = data; zio->io_orig_size = zio->io_size = psize; zio->io_lsize = lsize; zio->io_orig_flags = zio->io_flags = flags; zio->io_orig_stage = zio->io_stage = stage; zio->io_orig_pipeline = zio->io_pipeline = pipeline; zio->io_pipeline_trace = ZIO_STAGE_OPEN; zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY); zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE); if (zb != NULL) zio->io_bookmark = *zb; if (pio != NULL) { if (zio->io_logical == NULL) zio->io_logical = pio->io_logical; if (zio->io_child_type == ZIO_CHILD_GANG) zio->io_gang_leader = pio->io_gang_leader; zio_add_child(pio, zio); } taskq_init_ent(&zio->io_tqent); return (zio); } static void zio_destroy(zio_t *zio) { metaslab_trace_fini(&zio->io_alloc_list); list_destroy(&zio->io_parent_list); list_destroy(&zio->io_child_list); mutex_destroy(&zio->io_lock); cv_destroy(&zio->io_cv); kmem_cache_free(zio_cache, zio); } zio_t * zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done, void *private, enum zio_flag flags) { zio_t *zio; zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private, ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL, ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE); return (zio); } zio_t * zio_root(spa_t *spa, zio_done_func_t *done, void *private, enum zio_flag flags) { return (zio_null(NULL, spa, NULL, done, private, flags)); } void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp) { if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) { zfs_panic_recover("blkptr at %p has invalid TYPE %llu", bp, (longlong_t)BP_GET_TYPE(bp)); } if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS || BP_GET_CHECKSUM(bp) <= ZIO_CHECKSUM_ON) { zfs_panic_recover("blkptr at %p has invalid CHECKSUM %llu", bp, (longlong_t)BP_GET_CHECKSUM(bp)); } if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS || BP_GET_COMPRESS(bp) <= ZIO_COMPRESS_ON) { zfs_panic_recover("blkptr at %p has invalid COMPRESS %llu", bp, (longlong_t)BP_GET_COMPRESS(bp)); } if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) { zfs_panic_recover("blkptr at %p has invalid LSIZE %llu", bp, (longlong_t)BP_GET_LSIZE(bp)); } if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) { zfs_panic_recover("blkptr at %p has invalid PSIZE %llu", bp, (longlong_t)BP_GET_PSIZE(bp)); } if (BP_IS_EMBEDDED(bp)) { if (BPE_GET_ETYPE(bp) > NUM_BP_EMBEDDED_TYPES) { zfs_panic_recover("blkptr at %p has invalid ETYPE %llu", bp, (longlong_t)BPE_GET_ETYPE(bp)); } } /* * Do not verify individual DVAs if the config is not trusted. This * will be done once the zio is executed in vdev_mirror_map_alloc. */ if (!spa->spa_trust_config) return; /* * Pool-specific checks. * * Note: it would be nice to verify that the blk_birth and * BP_PHYSICAL_BIRTH() are not too large. However, spa_freeze() * allows the birth time of log blocks (and dmu_sync()-ed blocks * that are in the log) to be arbitrarily large. */ for (int i = 0; i < BP_GET_NDVAS(bp); i++) { uint64_t vdevid = DVA_GET_VDEV(&bp->blk_dva[i]); if (vdevid >= spa->spa_root_vdev->vdev_children) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid]; if (vd == NULL) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_hole_ops) { zfs_panic_recover("blkptr at %p DVA %u has hole " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_missing_ops) { /* * "missing" vdevs are valid during import, but we * don't have their detailed info (e.g. asize), so * we can't perform any more checks on them. */ continue; } uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]); uint64_t asize = DVA_GET_ASIZE(&bp->blk_dva[i]); if (BP_IS_GANG(bp)) asize = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE); if (offset + asize > vd->vdev_asize) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "OFFSET %llu", bp, i, (longlong_t)offset); } } } boolean_t zfs_dva_valid(spa_t *spa, const dva_t *dva, const blkptr_t *bp) { uint64_t vdevid = DVA_GET_VDEV(dva); if (vdevid >= spa->spa_root_vdev->vdev_children) return (B_FALSE); vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid]; if (vd == NULL) return (B_FALSE); if (vd->vdev_ops == &vdev_hole_ops) return (B_FALSE); if (vd->vdev_ops == &vdev_missing_ops) { return (B_FALSE); } uint64_t offset = DVA_GET_OFFSET(dva); uint64_t asize = DVA_GET_ASIZE(dva); if (BP_IS_GANG(bp)) asize = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE); if (offset + asize > vd->vdev_asize) return (B_FALSE); return (B_TRUE); } zio_t * zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, abd_t *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb) { zio_t *zio; zfs_blkptr_verify(spa, bp); zio = zio_create(pio, spa, BP_PHYSICAL_BIRTH(bp), bp, data, size, size, done, private, ZIO_TYPE_READ, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE); return (zio); } zio_t * zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data, uint64_t lsize, uint64_t psize, const zio_prop_t *zp, zio_done_func_t *ready, zio_done_func_t *children_ready, zio_done_func_t *physdone, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb) { zio_t *zio; ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF && zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS && zp->zp_compress >= ZIO_COMPRESS_OFF && zp->zp_compress < ZIO_COMPRESS_FUNCTIONS && DMU_OT_IS_VALID(zp->zp_type) && zp->zp_level < 32 && zp->zp_copies > 0 && zp->zp_copies <= spa_max_replication(spa)); zio = zio_create(pio, spa, txg, bp, data, lsize, psize, done, private, ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE); zio->io_ready = ready; zio->io_children_ready = children_ready; zio->io_physdone = physdone; zio->io_prop = *zp; /* * Data can be NULL if we are going to call zio_write_override() to * provide the already-allocated BP. But we may need the data to * verify a dedup hit (if requested). In this case, don't try to * dedup (just take the already-allocated BP verbatim). Encrypted * dedup blocks need data as well so we also disable dedup in this * case. */ if (data == NULL && (zio->io_prop.zp_dedup_verify || zio->io_prop.zp_encrypt)) { zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE; } return (zio); } zio_t * zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb) { zio_t *zio; zio = zio_create(pio, spa, txg, bp, data, size, size, done, private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_IO_REWRITE, NULL, 0, zb, ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE); return (zio); } void zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio->io_stage == ZIO_STAGE_OPEN); ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa)); /* * We must reset the io_prop to match the values that existed * when the bp was first written by dmu_sync() keeping in mind * that nopwrite and dedup are mutually exclusive. */ zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup; zio->io_prop.zp_nopwrite = nopwrite; zio->io_prop.zp_copies = copies; zio->io_bp_override = bp; } void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp) { zfs_blkptr_verify(spa, bp); /* * The check for EMBEDDED is a performance optimization. We * process the free here (by ignoring it) rather than * putting it on the list and then processing it in zio_free_sync(). */ if (BP_IS_EMBEDDED(bp)) return; metaslab_check_free(spa, bp); /* * Frees that are for the currently-syncing txg, are not going to be * deferred, and which will not need to do a read (i.e. not GANG or * DEDUP), can be processed immediately. Otherwise, put them on the * in-memory list for later processing. */ if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp) || txg != spa->spa_syncing_txg || spa_sync_pass(spa) >= zfs_sync_pass_deferred_free) { bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp); } else { VERIFY0(zio_wait(zio_free_sync(NULL, spa, txg, bp, 0))); } } zio_t * zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, enum zio_flag flags) { zio_t *zio; enum zio_stage stage = ZIO_FREE_PIPELINE; ASSERT(!BP_IS_HOLE(bp)); ASSERT(spa_syncing_txg(spa) == txg); ASSERT(spa_sync_pass(spa) < zfs_sync_pass_deferred_free); if (BP_IS_EMBEDDED(bp)) return (zio_null(pio, spa, NULL, NULL, NULL, 0)); metaslab_check_free(spa, bp); arc_freed(spa, bp); dsl_scan_freed(spa, bp); /* * GANG and DEDUP blocks can induce a read (for the gang block header, * or the DDT), so issue them asynchronously so that this thread is * not tied up. */ if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp)) stage |= ZIO_STAGE_ISSUE_ASYNC; zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp), BP_GET_PSIZE(bp), NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage); return (zio); } zio_t * zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_done_func_t *done, void *private, enum zio_flag flags) { zio_t *zio; zfs_blkptr_verify(spa, bp); if (BP_IS_EMBEDDED(bp)) return (zio_null(pio, spa, NULL, NULL, NULL, 0)); /* * A claim is an allocation of a specific block. Claims are needed * to support immediate writes in the intent log. The issue is that * immediate writes contain committed data, but in a txg that was * *not* committed. Upon opening the pool after an unclean shutdown, * the intent log claims all blocks that contain immediate write data * so that the SPA knows they're in use. * * All claims *must* be resolved in the first txg -- before the SPA * starts allocating blocks -- so that nothing is allocated twice. * If txg == 0 we just verify that the block is claimable. */ ASSERT3U(spa->spa_uberblock.ub_rootbp.blk_birth, <, spa_min_claim_txg(spa)); ASSERT(txg == spa_min_claim_txg(spa) || txg == 0); ASSERT(!BP_GET_DEDUP(bp) || !spa_writeable(spa)); /* zdb(1M) */ zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp), BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE); ASSERT0(zio->io_queued_timestamp); return (zio); } zio_t * zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd, zio_done_func_t *done, void *private, enum zio_flag flags) { zio_t *zio; int c; if (vd->vdev_children == 0) { zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private, ZIO_TYPE_IOCTL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL, ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE); zio->io_cmd = cmd; } else { zio = zio_null(pio, spa, NULL, NULL, NULL, flags); for (c = 0; c < vd->vdev_children; c++) zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd, done, private, flags)); } return (zio); } zio_t * zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, abd_t *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done, private, ZIO_TYPE_READ, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; return (zio); } zio_t * zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, abd_t *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done, private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { /* * zec checksums are necessarily destructive -- they modify * the end of the write buffer to hold the verifier/checksum. * Therefore, we must make a local copy in case the data is * being written to multiple places in parallel. */ abd_t *wbuf = abd_alloc_sametype(data, size); abd_copy(wbuf, data, size); zio_push_transform(zio, wbuf, size, size, NULL); } return (zio); } /* * Create a child I/O to do some work for us. */ zio_t * zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *private) { enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE; zio_t *zio; /* * vdev child I/Os do not propagate their error to the parent. * Therefore, for correct operation the caller *must* check for * and handle the error in the child i/o's done callback. * The only exceptions are i/os that we don't care about * (OPTIONAL or REPAIR). */ ASSERT((flags & ZIO_FLAG_OPTIONAL) || (flags & ZIO_FLAG_IO_REPAIR) || done != NULL); if (type == ZIO_TYPE_READ && bp != NULL) { /* * If we have the bp, then the child should perform the * checksum and the parent need not. This pushes error * detection as close to the leaves as possible and * eliminates redundant checksums in the interior nodes. */ pipeline |= ZIO_STAGE_CHECKSUM_VERIFY; pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } if (vd->vdev_ops->vdev_op_leaf) { ASSERT0(vd->vdev_children); offset += VDEV_LABEL_START_SIZE; } flags |= ZIO_VDEV_CHILD_FLAGS(pio); /* * If we've decided to do a repair, the write is not speculative -- * even if the original read was. */ if (flags & ZIO_FLAG_IO_REPAIR) flags &= ~ZIO_FLAG_SPECULATIVE; /* * If we're creating a child I/O that is not associated with a * top-level vdev, then the child zio is not an allocating I/O. * If this is a retried I/O then we ignore it since we will * have already processed the original allocating I/O. */ if (flags & ZIO_FLAG_IO_ALLOCATING && (vd != vd->vdev_top || (flags & ZIO_FLAG_IO_RETRY))) { ASSERTV(metaslab_class_t *mc = spa_normal_class(pio->io_spa)); ASSERT(mc->mc_alloc_throttle_enabled); ASSERT(type == ZIO_TYPE_WRITE); ASSERT(priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(flags & ZIO_FLAG_IO_REPAIR)); ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) || pio->io_child_type == ZIO_CHILD_GANG); flags &= ~ZIO_FLAG_IO_ALLOCATING; } zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size, done, private, type, priority, flags, vd, offset, &pio->io_bookmark, ZIO_STAGE_VDEV_IO_START >> 1, pipeline); ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV); zio->io_physdone = pio->io_physdone; if (vd->vdev_ops->vdev_op_leaf && zio->io_logical != NULL) zio->io_logical->io_phys_children++; return (zio); } zio_t * zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size, zio_type_t type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *private) { zio_t *zio; ASSERT(vd->vdev_ops->vdev_op_leaf); zio = zio_create(NULL, vd->vdev_spa, 0, NULL, data, size, size, done, private, type, priority, flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_DELEGATED, vd, offset, NULL, ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE); return (zio); } void zio_flush(zio_t *zio, vdev_t *vd) { zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY)); } void zio_shrink(zio_t *zio, uint64_t size) { ASSERT3P(zio->io_executor, ==, NULL); ASSERT3U(zio->io_orig_size, ==, zio->io_size); ASSERT3U(size, <=, zio->io_size); /* * We don't shrink for raidz because of problems with the * reconstruction when reading back less than the block size. * Note, BP_IS_RAIDZ() assumes no compression. */ ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF); if (!BP_IS_RAIDZ(zio->io_bp)) { /* we are not doing a raw write */ ASSERT3U(zio->io_size, ==, zio->io_lsize); zio->io_orig_size = zio->io_size = zio->io_lsize = size; } } /* * ========================================================================== * Prepare to read and write logical blocks * ========================================================================== */ static zio_t * zio_read_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; uint64_t psize = BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp); ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy); if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF && zio->io_child_type == ZIO_CHILD_LOGICAL && !(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) { zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize), psize, psize, zio_decompress); } if (((BP_IS_PROTECTED(bp) && !(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) || BP_HAS_INDIRECT_MAC_CKSUM(bp)) && zio->io_child_type == ZIO_CHILD_LOGICAL) { zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize), psize, psize, zio_decrypt); } if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) { int psize = BPE_GET_PSIZE(bp); void *data = abd_borrow_buf(zio->io_abd, psize); zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; decode_embedded_bp_compressed(bp, data); abd_return_buf_copy(zio->io_abd, data, psize); } else { ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy); } if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0) zio->io_flags |= ZIO_FLAG_DONT_CACHE; if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP) zio->io_flags |= ZIO_FLAG_DONT_CACHE; if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_pipeline = ZIO_DDT_READ_PIPELINE; return (zio); } static zio_t * zio_write_bp_init(zio_t *zio) { if (!IO_IS_ALLOCATING(zio)) return (zio); ASSERT(zio->io_child_type != ZIO_CHILD_DDT); if (zio->io_bp_override) { blkptr_t *bp = zio->io_bp; zio_prop_t *zp = &zio->io_prop; ASSERT(bp->blk_birth != zio->io_txg); ASSERT(BP_GET_DEDUP(zio->io_bp_override) == 0); *bp = *zio->io_bp_override; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (BP_IS_EMBEDDED(bp)) return (zio); /* * If we've been overridden and nopwrite is set then * set the flag accordingly to indicate that a nopwrite * has already occurred. */ if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) { ASSERT(!zp->zp_dedup); ASSERT3U(BP_GET_CHECKSUM(bp), ==, zp->zp_checksum); zio->io_flags |= ZIO_FLAG_NOPWRITE; return (zio); } ASSERT(!zp->zp_nopwrite); if (BP_IS_HOLE(bp) || !zp->zp_dedup) return (zio); ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_DEDUP) || zp->zp_dedup_verify); if (BP_GET_CHECKSUM(bp) == zp->zp_checksum && !zp->zp_encrypt) { BP_SET_DEDUP(bp, 1); zio->io_pipeline |= ZIO_STAGE_DDT_WRITE; return (zio); } /* * We were unable to handle this as an override bp, treat * it as a regular write I/O. */ zio->io_bp_override = NULL; *bp = zio->io_bp_orig; zio->io_pipeline = zio->io_orig_pipeline; } return (zio); } static zio_t * zio_write_compress(zio_t *zio) { spa_t *spa = zio->io_spa; zio_prop_t *zp = &zio->io_prop; enum zio_compress compress = zp->zp_compress; blkptr_t *bp = zio->io_bp; uint64_t lsize = zio->io_lsize; uint64_t psize = zio->io_size; int pass = 1; /* * If our children haven't all reached the ready stage, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT | ZIO_CHILD_GANG_BIT, ZIO_WAIT_READY)) { return (NULL); } if (!IO_IS_ALLOCATING(zio)) return (zio); if (zio->io_children_ready != NULL) { /* * Now that all our children are ready, run the callback * associated with this zio in case it wants to modify the * data to be written. */ ASSERT3U(zp->zp_level, >, 0); zio->io_children_ready(zio); } ASSERT(zio->io_child_type != ZIO_CHILD_DDT); ASSERT(zio->io_bp_override == NULL); if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg) { /* * We're rewriting an existing block, which means we're * working on behalf of spa_sync(). For spa_sync() to * converge, it must eventually be the case that we don't * have to allocate new blocks. But compression changes * the blocksize, which forces a reallocate, and makes * convergence take longer. Therefore, after the first * few passes, stop compressing to ensure convergence. */ pass = spa_sync_pass(spa); ASSERT(zio->io_txg == spa_syncing_txg(spa)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!BP_GET_DEDUP(bp)); if (pass >= zfs_sync_pass_dont_compress) compress = ZIO_COMPRESS_OFF; /* Make sure someone doesn't change their mind on overwrites */ ASSERT(BP_IS_EMBEDDED(bp) || MIN(zp->zp_copies + BP_IS_GANG(bp), spa_max_replication(spa)) == BP_GET_NDVAS(bp)); } /* If it's a compressed write that is not raw, compress the buffer. */ if (compress != ZIO_COMPRESS_OFF && !(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) { void *cbuf = zio_buf_alloc(lsize); psize = zio_compress_data(compress, zio->io_abd, cbuf, lsize); if (psize == 0 || psize == lsize) { compress = ZIO_COMPRESS_OFF; zio_buf_free(cbuf, lsize); } else if (!zp->zp_dedup && !zp->zp_encrypt && psize <= BPE_PAYLOAD_SIZE && zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) && spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) { encode_embedded_bp_compressed(bp, cbuf, compress, lsize, psize); BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA); BP_SET_TYPE(bp, zio->io_prop.zp_type); BP_SET_LEVEL(bp, zio->io_prop.zp_level); zio_buf_free(cbuf, lsize); bp->blk_birth = zio->io_txg; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; ASSERT(spa_feature_is_active(spa, SPA_FEATURE_EMBEDDED_DATA)); return (zio); } else { /* * Round up compressed size up to the ashift * of the smallest-ashift device, and zero the tail. * This ensures that the compressed size of the BP * (and thus compressratio property) are correct, * in that we charge for the padding used to fill out * the last sector. */ ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT); size_t rounded = (size_t)P2ROUNDUP(psize, 1ULL << spa->spa_min_ashift); if (rounded >= lsize) { compress = ZIO_COMPRESS_OFF; zio_buf_free(cbuf, lsize); psize = lsize; } else { abd_t *cdata = abd_get_from_buf(cbuf, lsize); abd_take_ownership_of_buf(cdata, B_TRUE); abd_zero_off(cdata, psize, rounded - psize); psize = rounded; zio_push_transform(zio, cdata, psize, lsize, NULL); } } /* * We were unable to handle this as an override bp, treat * it as a regular write I/O. */ zio->io_bp_override = NULL; *bp = zio->io_bp_orig; zio->io_pipeline = zio->io_orig_pipeline; } else if ((zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) != 0 && zp->zp_type == DMU_OT_DNODE) { /* * The DMU actually relies on the zio layer's compression * to free metadnode blocks that have had all contained * dnodes freed. As a result, even when doing a raw * receive, we must check whether the block can be compressed * to a hole. */ psize = zio_compress_data(ZIO_COMPRESS_EMPTY, zio->io_abd, NULL, lsize); if (psize == 0) compress = ZIO_COMPRESS_OFF; } else { ASSERT3U(psize, !=, 0); } /* * The final pass of spa_sync() must be all rewrites, but the first * few passes offer a trade-off: allocating blocks defers convergence, * but newly allocated blocks are sequential, so they can be written * to disk faster. Therefore, we allow the first few passes of * spa_sync() to allocate new blocks, but force rewrites after that. * There should only be a handful of blocks after pass 1 in any case. */ if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg && BP_GET_PSIZE(bp) == psize && pass >= zfs_sync_pass_rewrite) { ASSERT(psize != 0); enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES; zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages; zio->io_flags |= ZIO_FLAG_IO_REWRITE; } else { BP_ZERO(bp); zio->io_pipeline = ZIO_WRITE_PIPELINE; } if (psize == 0) { if (zio->io_bp_orig.blk_birth != 0 && spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) { BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_BIRTH(bp, zio->io_txg, 0); } zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; } else { ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER); BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_PSIZE(bp, psize); BP_SET_COMPRESS(bp, compress); BP_SET_CHECKSUM(bp, zp->zp_checksum); BP_SET_DEDUP(bp, zp->zp_dedup); BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER); if (zp->zp_dedup) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(!zp->zp_encrypt || DMU_OT_IS_ENCRYPTED(zp->zp_type)); zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE; } if (zp->zp_nopwrite) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); zio->io_pipeline |= ZIO_STAGE_NOP_WRITE; } } return (zio); } static zio_t * zio_free_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio->io_child_type == ZIO_CHILD_LOGICAL) { if (BP_GET_DEDUP(bp)) zio->io_pipeline = ZIO_DDT_FREE_PIPELINE; } ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy); return (zio); } /* * ========================================================================== * Execute the I/O pipeline * ========================================================================== */ static void zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline) { spa_t *spa = zio->io_spa; zio_type_t t = zio->io_type; int flags = (cutinline ? TQ_FRONT : 0); /* * If we're a config writer or a probe, the normal issue and * interrupt threads may all be blocked waiting for the config lock. * In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL. */ if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_PROBE)) t = ZIO_TYPE_NULL; /* * A similar issue exists for the L2ARC write thread until L2ARC 2.0. */ if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux) t = ZIO_TYPE_NULL; /* * If this is a high priority I/O, then use the high priority taskq if * available. */ if (zio->io_priority == ZIO_PRIORITY_NOW && spa->spa_zio_taskq[t][q + 1].stqs_count != 0) q++; ASSERT3U(q, <, ZIO_TASKQ_TYPES); /* * NB: We are assuming that the zio can only be dispatched * to a single taskq at a time. It would be a grievous error * to dispatch the zio to another taskq at the same time. */ ASSERT(taskq_empty_ent(&zio->io_tqent)); spa_taskq_dispatch_ent(spa, t, q, (task_func_t *)zio_execute, zio, flags, &zio->io_tqent); } static boolean_t zio_taskq_member(zio_t *zio, zio_taskq_type_t q) { kthread_t *executor = zio->io_executor; spa_t *spa = zio->io_spa; for (zio_type_t t = 0; t < ZIO_TYPES; t++) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; uint_t i; for (i = 0; i < tqs->stqs_count; i++) { if (taskq_member(tqs->stqs_taskq[i], executor)) return (B_TRUE); } } return (B_FALSE); } static zio_t * zio_issue_async(zio_t *zio) { zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (NULL); } void zio_interrupt(zio_t *zio) { zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE); } void zio_delay_interrupt(zio_t *zio) { /* * The timeout_generic() function isn't defined in userspace, so * rather than trying to implement the function, the zio delay * functionality has been disabled for userspace builds. */ #ifdef _KERNEL /* * If io_target_timestamp is zero, then no delay has been registered * for this IO, thus jump to the end of this function and "skip" the * delay; issuing it directly to the zio layer. */ if (zio->io_target_timestamp != 0) { hrtime_t now = gethrtime(); if (now >= zio->io_target_timestamp) { /* * This IO has already taken longer than the target * delay to complete, so we don't want to delay it * any longer; we "miss" the delay and issue it * directly to the zio layer. This is likely due to * the target latency being set to a value less than * the underlying hardware can satisfy (e.g. delay * set to 1ms, but the disks take 10ms to complete an * IO request). */ DTRACE_PROBE2(zio__delay__miss, zio_t *, zio, hrtime_t, now); zio_interrupt(zio); } else { taskqid_t tid; hrtime_t diff = zio->io_target_timestamp - now; clock_t expire_at_tick = ddi_get_lbolt() + NSEC_TO_TICK(diff); DTRACE_PROBE3(zio__delay__hit, zio_t *, zio, hrtime_t, now, hrtime_t, diff); if (NSEC_TO_TICK(diff) == 0) { /* Our delay is less than a jiffy - just spin */ zfs_sleep_until(zio->io_target_timestamp); } else { /* * Use taskq_dispatch_delay() in the place of * OpenZFS's timeout_generic(). */ tid = taskq_dispatch_delay(system_taskq, (task_func_t *)zio_interrupt, zio, TQ_NOSLEEP, expire_at_tick); if (tid == TASKQID_INVALID) { /* * Couldn't allocate a task. Just * finish the zio without a delay. */ zio_interrupt(zio); } } } return; } #endif DTRACE_PROBE1(zio__delay__skip, zio_t *, zio); zio_interrupt(zio); } static void zio_deadman_impl(zio_t *pio) { zio_t *cio, *cio_next; zio_link_t *zl = NULL; vdev_t *vd = pio->io_vd; if (vd != NULL && vd->vdev_ops->vdev_op_leaf) { vdev_queue_t *vq = &vd->vdev_queue; zbookmark_phys_t *zb = &pio->io_bookmark; uint64_t delta = gethrtime() - pio->io_timestamp; uint64_t failmode = spa_get_deadman_failmode(pio->io_spa); zfs_dbgmsg("slow zio: zio=%p timestamp=%llu " "delta=%llu queued=%llu io=%llu " "path=%s last=%llu " "type=%d priority=%d flags=0x%x " "stage=0x%x pipeline=0x%x pipeline-trace=0x%x " "objset=%llu object=%llu level=%llu blkid=%llu " "offset=%llu size=%llu error=%d", pio, pio->io_timestamp, delta, pio->io_delta, pio->io_delay, vd->vdev_path, vq->vq_io_complete_ts, pio->io_type, pio->io_priority, pio->io_flags, pio->io_state, pio->io_pipeline, pio->io_pipeline_trace, zb->zb_objset, zb->zb_object, zb->zb_level, zb->zb_blkid, pio->io_offset, pio->io_size, pio->io_error); zfs_ereport_post(FM_EREPORT_ZFS_DEADMAN, pio->io_spa, vd, zb, pio, 0, 0); if (failmode == ZIO_FAILURE_MODE_CONTINUE && taskq_empty_ent(&pio->io_tqent)) { zio_interrupt(pio); } } mutex_enter(&pio->io_lock); for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio, &zl); zio_deadman_impl(cio); } mutex_exit(&pio->io_lock); } /* * Log the critical information describing this zio and all of its children * using the zfs_dbgmsg() interface then post deadman event for the ZED. */ void zio_deadman(zio_t *pio, char *tag) { spa_t *spa = pio->io_spa; char *name = spa_name(spa); if (!zfs_deadman_enabled || spa_suspended(spa)) return; zio_deadman_impl(pio); switch (spa_get_deadman_failmode(spa)) { case ZIO_FAILURE_MODE_WAIT: zfs_dbgmsg("%s waiting for hung I/O to pool '%s'", tag, name); break; case ZIO_FAILURE_MODE_CONTINUE: zfs_dbgmsg("%s restarting hung I/O for pool '%s'", tag, name); break; case ZIO_FAILURE_MODE_PANIC: fm_panic("%s determined I/O to pool '%s' is hung.", tag, name); break; } } /* * Execute the I/O pipeline until one of the following occurs: * (1) the I/O completes; (2) the pipeline stalls waiting for * dependent child I/Os; (3) the I/O issues, so we're waiting * for an I/O completion interrupt; (4) the I/O is delegated by * vdev-level caching or aggregation; (5) the I/O is deferred * due to vdev-level queueing; (6) the I/O is handed off to * another thread. In all cases, the pipeline stops whenever * there's no CPU work; it never burns a thread in cv_wait_io(). * * There's no locking on io_stage because there's no legitimate way * for multiple threads to be attempting to process the same I/O. */ static zio_pipe_stage_t *zio_pipeline[]; /* * zio_execute() is a wrapper around the static function * __zio_execute() so that we can force __zio_execute() to be * inlined. This reduces stack overhead which is important * because __zio_execute() is called recursively in several zio * code paths. zio_execute() itself cannot be inlined because * it is externally visible. */ void zio_execute(zio_t *zio) { fstrans_cookie_t cookie; cookie = spl_fstrans_mark(); __zio_execute(zio); spl_fstrans_unmark(cookie); } /* * Used to determine if in the current context the stack is sized large * enough to allow zio_execute() to be called recursively. A minimum * stack size of 16K is required to avoid needing to re-dispatch the zio. */ boolean_t zio_execute_stack_check(zio_t *zio) { #if !defined(HAVE_LARGE_STACKS) dsl_pool_t *dp = spa_get_dsl(zio->io_spa); /* Executing in txg_sync_thread() context. */ if (dp && curthread == dp->dp_tx.tx_sync_thread) return (B_TRUE); /* Pool initialization outside of zio_taskq context. */ if (dp && spa_is_initializing(dp->dp_spa) && !zio_taskq_member(zio, ZIO_TASKQ_ISSUE) && !zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH)) return (B_TRUE); #endif /* HAVE_LARGE_STACKS */ return (B_FALSE); } __attribute__((always_inline)) static inline void __zio_execute(zio_t *zio) { ASSERT3U(zio->io_queued_timestamp, >, 0); while (zio->io_stage < ZIO_STAGE_DONE) { enum zio_stage pipeline = zio->io_pipeline; enum zio_stage stage = zio->io_stage; zio->io_executor = curthread; ASSERT(!MUTEX_HELD(&zio->io_lock)); ASSERT(ISP2(stage)); ASSERT(zio->io_stall == NULL); do { stage <<= 1; } while ((stage & pipeline) == 0); ASSERT(stage <= ZIO_STAGE_DONE); /* * If we are in interrupt context and this pipeline stage * will grab a config lock that is held across I/O, * or may wait for an I/O that needs an interrupt thread * to complete, issue async to avoid deadlock. * * For VDEV_IO_START, we cut in line so that the io will * be sent to disk promptly. */ if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL && zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) { boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ? zio_requeue_io_start_cut_in_line : B_FALSE; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut); return; } /* * If the current context doesn't have large enough stacks * the zio must be issued asynchronously to prevent overflow. */ if (zio_execute_stack_check(zio)) { boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ? zio_requeue_io_start_cut_in_line : B_FALSE; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut); return; } zio->io_stage = stage; zio->io_pipeline_trace |= zio->io_stage; /* * The zio pipeline stage returns the next zio to execute * (typically the same as this one), or NULL if we should * stop. */ zio = zio_pipeline[highbit64(stage) - 1](zio); if (zio == NULL) return; } } /* * ========================================================================== * Initiate I/O, either sync or async * ========================================================================== */ int zio_wait(zio_t *zio) { long timeout = MSEC_TO_TICK(zfs_deadman_ziotime_ms); int error; ASSERT3S(zio->io_stage, ==, ZIO_STAGE_OPEN); ASSERT3P(zio->io_executor, ==, NULL); zio->io_waiter = curthread; ASSERT0(zio->io_queued_timestamp); zio->io_queued_timestamp = gethrtime(); __zio_execute(zio); mutex_enter(&zio->io_lock); while (zio->io_executor != NULL) { error = cv_timedwait_io(&zio->io_cv, &zio->io_lock, ddi_get_lbolt() + timeout); if (zfs_deadman_enabled && error == -1 && gethrtime() - zio->io_queued_timestamp > spa_deadman_ziotime(zio->io_spa)) { mutex_exit(&zio->io_lock); timeout = MSEC_TO_TICK(zfs_deadman_checktime_ms); zio_deadman(zio, FTAG); mutex_enter(&zio->io_lock); } } mutex_exit(&zio->io_lock); error = zio->io_error; zio_destroy(zio); return (error); } void zio_nowait(zio_t *zio) { ASSERT3P(zio->io_executor, ==, NULL); if (zio->io_child_type == ZIO_CHILD_LOGICAL && zio_unique_parent(zio) == NULL) { zio_t *pio; /* * This is a logical async I/O with no parent to wait for it. * We add it to the spa_async_root_zio "Godfather" I/O which * will ensure they complete prior to unloading the pool. */ spa_t *spa = zio->io_spa; kpreempt_disable(); pio = spa->spa_async_zio_root[CPU_SEQID]; kpreempt_enable(); zio_add_child(pio, zio); } ASSERT0(zio->io_queued_timestamp); zio->io_queued_timestamp = gethrtime(); __zio_execute(zio); } /* * ========================================================================== * Reexecute, cancel, or suspend/resume failed I/O * ========================================================================== */ static void zio_reexecute(zio_t *pio) { zio_t *cio, *cio_next; ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN); ASSERT(pio->io_gang_leader == NULL); ASSERT(pio->io_gang_tree == NULL); pio->io_flags = pio->io_orig_flags; pio->io_stage = pio->io_orig_stage; pio->io_pipeline = pio->io_orig_pipeline; pio->io_reexecute = 0; pio->io_flags |= ZIO_FLAG_REEXECUTED; pio->io_pipeline_trace = 0; pio->io_error = 0; for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_state[w] = 0; for (int c = 0; c < ZIO_CHILD_TYPES; c++) pio->io_child_error[c] = 0; if (IO_IS_ALLOCATING(pio)) BP_ZERO(pio->io_bp); /* * As we reexecute pio's children, new children could be created. * New children go to the head of pio's io_child_list, however, * so we will (correctly) not reexecute them. The key is that * the remainder of pio's io_child_list, from 'cio_next' onward, * cannot be affected by any side effects of reexecuting 'cio'. */ zio_link_t *zl = NULL; mutex_enter(&pio->io_lock); for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio, &zl); for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_children[cio->io_child_type][w]++; mutex_exit(&pio->io_lock); zio_reexecute(cio); mutex_enter(&pio->io_lock); } mutex_exit(&pio->io_lock); /* * Now that all children have been reexecuted, execute the parent. * We don't reexecute "The Godfather" I/O here as it's the * responsibility of the caller to wait on it. */ if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) { pio->io_queued_timestamp = gethrtime(); __zio_execute(pio); } } void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t reason) { if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC) fm_panic("Pool '%s' has encountered an uncorrectable I/O " "failure and the failure mode property for this pool " "is set to panic.", spa_name(spa)); cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable I/O " "failure and has been suspended.\n", spa_name(spa)); zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL, NULL, NULL, 0, 0); mutex_enter(&spa->spa_suspend_lock); if (spa->spa_suspend_zio_root == NULL) spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); spa->spa_suspended = reason; if (zio != NULL) { ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); ASSERT(zio != spa->spa_suspend_zio_root); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio_unique_parent(zio) == NULL); ASSERT(zio->io_stage == ZIO_STAGE_DONE); zio_add_child(spa->spa_suspend_zio_root, zio); } mutex_exit(&spa->spa_suspend_lock); } int zio_resume(spa_t *spa) { zio_t *pio; /* * Reexecute all previously suspended i/o. */ mutex_enter(&spa->spa_suspend_lock); spa->spa_suspended = ZIO_SUSPEND_NONE; cv_broadcast(&spa->spa_suspend_cv); pio = spa->spa_suspend_zio_root; spa->spa_suspend_zio_root = NULL; mutex_exit(&spa->spa_suspend_lock); if (pio == NULL) return (0); zio_reexecute(pio); return (zio_wait(pio)); } void zio_resume_wait(spa_t *spa) { mutex_enter(&spa->spa_suspend_lock); while (spa_suspended(spa)) cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock); mutex_exit(&spa->spa_suspend_lock); } /* * ========================================================================== * Gang blocks. * * A gang block is a collection of small blocks that looks to the DMU * like one large block. When zio_dva_allocate() cannot find a block * of the requested size, due to either severe fragmentation or the pool * being nearly full, it calls zio_write_gang_block() to construct the * block from smaller fragments. * * A gang block consists of a gang header (zio_gbh_phys_t) and up to * three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like * an indirect block: it's an array of block pointers. It consumes * only one sector and hence is allocatable regardless of fragmentation. * The gang header's bps point to its gang members, which hold the data. * * Gang blocks are self-checksumming, using the bp's * as the verifier to ensure uniqueness of the SHA256 checksum. * Critically, the gang block bp's blk_cksum is the checksum of the data, * not the gang header. This ensures that data block signatures (needed for * deduplication) are independent of how the block is physically stored. * * Gang blocks can be nested: a gang member may itself be a gang block. * Thus every gang block is a tree in which root and all interior nodes are * gang headers, and the leaves are normal blocks that contain user data. * The root of the gang tree is called the gang leader. * * To perform any operation (read, rewrite, free, claim) on a gang block, * zio_gang_assemble() first assembles the gang tree (minus data leaves) * in the io_gang_tree field of the original logical i/o by recursively * reading the gang leader and all gang headers below it. This yields * an in-core tree containing the contents of every gang header and the * bps for every constituent of the gang block. * * With the gang tree now assembled, zio_gang_issue() just walks the gang tree * and invokes a callback on each bp. To free a gang block, zio_gang_issue() * calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp. * zio_claim_gang() provides a similarly trivial wrapper for zio_claim(). * zio_read_gang() is a wrapper around zio_read() that omits reading gang * headers, since we already have those in io_gang_tree. zio_rewrite_gang() * performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite() * of the gang header plus zio_checksum_compute() of the data to update the * gang header's blk_cksum as described above. * * The two-phase assemble/issue model solves the problem of partial failure -- * what if you'd freed part of a gang block but then couldn't read the * gang header for another part? Assembling the entire gang tree first * ensures that all the necessary gang header I/O has succeeded before * starting the actual work of free, claim, or write. Once the gang tree * is assembled, free and claim are in-memory operations that cannot fail. * * In the event that a gang write fails, zio_dva_unallocate() walks the * gang tree to immediately free (i.e. insert back into the space map) * everything we've allocated. This ensures that we don't get ENOSPC * errors during repeated suspend/resume cycles due to a flaky device. * * Gang rewrites only happen during sync-to-convergence. If we can't assemble * the gang tree, we won't modify the block, so we can safely defer the free * (knowing that the block is still intact). If we *can* assemble the gang * tree, then even if some of the rewrites fail, zio_dva_unallocate() will free * each constituent bp and we can allocate a new block on the next sync pass. * * In all cases, the gang tree allows complete recovery from partial failure. * ========================================================================== */ static void zio_gang_issue_func_done(zio_t *zio) { abd_put(zio->io_abd); } static zio_t * zio_read_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { if (gn != NULL) return (pio); return (zio_read(pio, pio->io_spa, bp, abd_get_offset(data, offset), BP_GET_PSIZE(bp), zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark)); } static zio_t * zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { zio_t *zio; if (gn != NULL) { abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * As we rewrite each gang header, the pipeline will compute * a new gang block header checksum for it; but no one will * compute a new data checksum, so we do that here. The one * exception is the gang leader: the pipeline already computed * its data checksum because that stage precedes gang assembly. * (Presently, nothing actually uses interior data checksums; * this is just good hygiene.) */ if (gn != pio->io_gang_leader->io_gang_tree) { abd_t *buf = abd_get_offset(data, offset); zio_checksum_compute(zio, BP_GET_CHECKSUM(bp), buf, BP_GET_PSIZE(bp)); abd_put(buf); } /* * If we are here to damage data for testing purposes, * leave the GBH alone so that we can detect the damage. */ if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE) zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } else { zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, abd_get_offset(data, offset), BP_GET_PSIZE(bp), zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); } return (zio); } /* ARGSUSED */ static zio_t * zio_free_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { return (zio_free_sync(pio, pio->io_spa, pio->io_txg, bp, ZIO_GANG_CHILD_FLAGS(pio))); } /* ARGSUSED */ static zio_t * zio_claim_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { return (zio_claim(pio, pio->io_spa, pio->io_txg, bp, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio))); } static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = { NULL, zio_read_gang, zio_rewrite_gang, zio_free_gang, zio_claim_gang, NULL }; static void zio_gang_tree_assemble_done(zio_t *zio); static zio_gang_node_t * zio_gang_node_alloc(zio_gang_node_t **gnpp) { zio_gang_node_t *gn; ASSERT(*gnpp == NULL); gn = kmem_zalloc(sizeof (*gn), KM_SLEEP); gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE); *gnpp = gn; return (gn); } static void zio_gang_node_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) ASSERT(gn->gn_child[g] == NULL); zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE); kmem_free(gn, sizeof (*gn)); *gnpp = NULL; } static void zio_gang_tree_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; if (gn == NULL) return; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) zio_gang_tree_free(&gn->gn_child[g]); zio_gang_node_free(gnpp); } static void zio_gang_tree_assemble(zio_t *gio, blkptr_t *bp, zio_gang_node_t **gnpp) { zio_gang_node_t *gn = zio_gang_node_alloc(gnpp); abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); ASSERT(gio->io_gang_leader == gio); ASSERT(BP_IS_GANG(bp)); zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_tree_assemble_done, gn, gio->io_priority, ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark)); } static void zio_gang_tree_assemble_done(zio_t *zio) { zio_t *gio = zio->io_gang_leader; zio_gang_node_t *gn = zio->io_private; blkptr_t *bp = zio->io_bp; ASSERT(gio == zio_unique_parent(zio)); ASSERT(zio->io_child_count == 0); if (zio->io_error) return; /* this ABD was created from a linear buf in zio_gang_tree_assemble */ if (BP_SHOULD_BYTESWAP(bp)) byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size); ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh); ASSERT(zio->io_size == SPA_GANGBLOCKSIZE); ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); abd_put(zio->io_abd); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (!BP_IS_GANG(gbp)) continue; zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]); } } static void zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, abd_t *data, uint64_t offset) { zio_t *gio = pio->io_gang_leader; zio_t *zio; ASSERT(BP_IS_GANG(bp) == !!gn); ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp)); ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) || gn == gio->io_gang_tree); /* * If you're a gang header, your data is in gn->gn_gbh. * If you're a gang member, your data is in 'data' and gn == NULL. */ zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset); if (gn != NULL) { ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (BP_IS_HOLE(gbp)) continue; zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data, offset); offset += BP_GET_PSIZE(gbp); } } if (gn == gio->io_gang_tree) ASSERT3U(gio->io_size, ==, offset); if (zio != pio) zio_nowait(zio); } static zio_t * zio_gang_assemble(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree); return (zio); } static zio_t * zio_gang_issue(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT, ZIO_WAIT_DONE)) { return (NULL); } ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); if (zio->io_child_error[ZIO_CHILD_GANG] == 0) zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_abd, 0); else zio_gang_tree_free(&zio->io_gang_tree); zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; return (zio); } static void zio_write_gang_member_ready(zio_t *zio) { zio_t *pio = zio_unique_parent(zio); dva_t *cdva = zio->io_bp->blk_dva; dva_t *pdva = pio->io_bp->blk_dva; uint64_t asize; ASSERTV(zio_t *gio = zio->io_gang_leader); if (BP_IS_HOLE(zio->io_bp)) return; ASSERT(BP_IS_HOLE(&zio->io_bp_orig)); ASSERT(zio->io_child_type == ZIO_CHILD_GANG); ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies); ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp)); ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp)); ASSERT3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp)); mutex_enter(&pio->io_lock); for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) { ASSERT(DVA_GET_GANG(&pdva[d])); asize = DVA_GET_ASIZE(&pdva[d]); asize += DVA_GET_ASIZE(&cdva[d]); DVA_SET_ASIZE(&pdva[d], asize); } mutex_exit(&pio->io_lock); } static void zio_write_gang_done(zio_t *zio) { abd_put(zio->io_abd); } static zio_t * zio_write_gang_block(zio_t *pio) { spa_t *spa = pio->io_spa; metaslab_class_t *mc = spa_normal_class(spa); blkptr_t *bp = pio->io_bp; zio_t *gio = pio->io_gang_leader; zio_t *zio; zio_gang_node_t *gn, **gnpp; zio_gbh_phys_t *gbh; abd_t *gbh_abd; uint64_t txg = pio->io_txg; uint64_t resid = pio->io_size; uint64_t lsize; int copies = gio->io_prop.zp_copies; int gbh_copies; zio_prop_t zp; int error; /* * encrypted blocks need DVA[2] free so encrypted gang headers can't * have a third copy. */ gbh_copies = MIN(copies + 1, spa_max_replication(spa)); if (gio->io_prop.zp_encrypt && gbh_copies >= SPA_DVAS_PER_BP) gbh_copies = SPA_DVAS_PER_BP - 1; int flags = METASLAB_HINTBP_FAVOR | METASLAB_GANG_HEADER; if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(pio->io_flags & ZIO_FLAG_NODATA)); flags |= METASLAB_ASYNC_ALLOC; VERIFY(refcount_held(&mc->mc_alloc_slots[pio->io_allocator], pio)); /* * The logical zio has already placed a reservation for * 'copies' allocation slots but gang blocks may require * additional copies. These additional copies * (i.e. gbh_copies - copies) are guaranteed to succeed * since metaslab_class_throttle_reserve() always allows * additional reservations for gang blocks. */ VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies - copies, pio->io_allocator, pio, flags)); } error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE, bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags, &pio->io_alloc_list, pio, pio->io_allocator); if (error) { if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(pio->io_flags & ZIO_FLAG_NODATA)); /* * If we failed to allocate the gang block header then * we remove any additional allocation reservations that * we placed here. The original reservation will * be removed when the logical I/O goes to the ready * stage. */ metaslab_class_throttle_unreserve(mc, gbh_copies - copies, pio->io_allocator, pio); } pio->io_error = error; return (pio); } if (pio == gio) { gnpp = &gio->io_gang_tree; } else { gnpp = pio->io_private; ASSERT(pio->io_ready == zio_write_gang_member_ready); } gn = zio_gang_node_alloc(gnpp); gbh = gn->gn_gbh; bzero(gbh, SPA_GANGBLOCKSIZE); gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE); /* * Create the gang header. */ zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_write_gang_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * Create and nowait the gang children. */ for (int g = 0; resid != 0; resid -= lsize, g++) { lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g), SPA_MINBLOCKSIZE); ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid); zp.zp_checksum = gio->io_prop.zp_checksum; zp.zp_compress = ZIO_COMPRESS_OFF; zp.zp_type = DMU_OT_NONE; zp.zp_level = 0; zp.zp_copies = gio->io_prop.zp_copies; zp.zp_dedup = B_FALSE; zp.zp_dedup_verify = B_FALSE; zp.zp_nopwrite = B_FALSE; zp.zp_encrypt = gio->io_prop.zp_encrypt; zp.zp_byteorder = gio->io_prop.zp_byteorder; bzero(zp.zp_salt, ZIO_DATA_SALT_LEN); bzero(zp.zp_iv, ZIO_DATA_IV_LEN); bzero(zp.zp_mac, ZIO_DATA_MAC_LEN); zio_t *cio = zio_write(zio, spa, txg, &gbh->zg_blkptr[g], abd_get_offset(pio->io_abd, pio->io_size - resid), lsize, lsize, &zp, zio_write_gang_member_ready, NULL, NULL, zio_write_gang_done, &gn->gn_child[g], pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(pio->io_flags & ZIO_FLAG_NODATA)); /* * Gang children won't throttle but we should * account for their work, so reserve an allocation * slot for them here. */ VERIFY(metaslab_class_throttle_reserve(mc, zp.zp_copies, cio->io_allocator, cio, flags)); } zio_nowait(cio); } /* * Set pio's pipeline to just wait for zio to finish. */ pio->io_pipeline = ZIO_INTERLOCK_PIPELINE; /* * We didn't allocate this bp, so make sure it doesn't get unmarked. */ pio->io_flags &= ~ZIO_FLAG_FASTWRITE; zio_nowait(zio); return (pio); } /* * The zio_nop_write stage in the pipeline determines if allocating a * new bp is necessary. The nopwrite feature can handle writes in * either syncing or open context (i.e. zil writes) and as a result is * mutually exclusive with dedup. * * By leveraging a cryptographically secure checksum, such as SHA256, we * can compare the checksums of the new data and the old to determine if * allocating a new block is required. Note that our requirements for * cryptographic strength are fairly weak: there can't be any accidental * hash collisions, but we don't need to be secure against intentional * (malicious) collisions. To trigger a nopwrite, you have to be able * to write the file to begin with, and triggering an incorrect (hash * collision) nopwrite is no worse than simply writing to the file. * That said, there are no known attacks against the checksum algorithms * used for nopwrite, assuming that the salt and the checksums * themselves remain secret. */ static zio_t * zio_nop_write(zio_t *zio) { blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; zio_prop_t *zp = &zio->io_prop; ASSERT(BP_GET_LEVEL(bp) == 0); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(zp->zp_nopwrite); ASSERT(!zp->zp_dedup); ASSERT(zio->io_bp_override == NULL); ASSERT(IO_IS_ALLOCATING(zio)); /* * Check to see if the original bp and the new bp have matching * characteristics (i.e. same checksum, compression algorithms, etc). * If they don't then just continue with the pipeline which will * allocate a new bp. */ if (BP_IS_HOLE(bp_orig) || !(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags & ZCHECKSUM_FLAG_NOPWRITE) || BP_IS_ENCRYPTED(bp) || BP_IS_ENCRYPTED(bp_orig) || BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) || BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) || BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) || zp->zp_copies != BP_GET_NDVAS(bp_orig)) return (zio); /* * If the checksums match then reset the pipeline so that we * avoid allocating a new bp and issuing any I/O. */ if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) { ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_NOPWRITE); ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig)); ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig)); ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF); ASSERT(bcmp(&bp->blk_prop, &bp_orig->blk_prop, sizeof (uint64_t)) == 0); *bp = *bp_orig; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; zio->io_flags |= ZIO_FLAG_NOPWRITE; } return (zio); } /* * ========================================================================== * Dedup * ========================================================================== */ static void zio_ddt_child_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp; zio_t *pio = zio_unique_parent(zio); mutex_enter(&pio->io_lock); ddp = ddt_phys_select(dde, bp); if (zio->io_error == 0) ddt_phys_clear(ddp); /* this ddp doesn't need repair */ if (zio->io_error == 0 && dde->dde_repair_abd == NULL) dde->dde_repair_abd = zio->io_abd; else abd_free(zio->io_abd); mutex_exit(&pio->io_lock); } static zio_t * zio_ddt_read_start(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = ddt_repair_start(ddt, bp); ddt_phys_t *ddp = dde->dde_phys; ddt_phys_t *ddp_self = ddt_phys_select(dde, bp); blkptr_t blk; ASSERT(zio->io_vsd == NULL); zio->io_vsd = dde; if (ddp_self == NULL) return (zio); for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) { if (ddp->ddp_phys_birth == 0 || ddp == ddp_self) continue; ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp, &blk); zio_nowait(zio_read(zio, zio->io_spa, &blk, abd_alloc_for_io(zio->io_size, B_TRUE), zio->io_size, zio_ddt_child_read_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) | ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark)); } return (zio); } zio_nowait(zio_read(zio, zio->io_spa, bp, zio->io_abd, zio->io_size, NULL, NULL, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark)); return (zio); } static zio_t * zio_ddt_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_DDT_BIT, ZIO_WAIT_DONE)) { return (NULL); } ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = zio->io_vsd; if (ddt == NULL) { ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE); return (zio); } if (dde == NULL) { zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (NULL); } if (dde->dde_repair_abd != NULL) { abd_copy(zio->io_abd, dde->dde_repair_abd, zio->io_size); zio->io_child_error[ZIO_CHILD_DDT] = 0; } ddt_repair_done(ddt, dde); zio->io_vsd = NULL; } ASSERT(zio->io_vsd == NULL); return (zio); } static boolean_t zio_ddt_collision(zio_t *zio, ddt_t *ddt, ddt_entry_t *dde) { spa_t *spa = zio->io_spa; boolean_t do_raw = !!(zio->io_flags & ZIO_FLAG_RAW); ASSERT(!(zio->io_bp_override && do_raw)); /* * Note: we compare the original data, not the transformed data, * because when zio->io_bp is an override bp, we will not have * pushed the I/O transforms. That's an important optimization * because otherwise we'd compress/encrypt all dmu_sync() data twice. * However, we should never get a raw, override zio so in these * cases we can compare the io_abd directly. This is useful because * it allows us to do dedup verification even if we don't have access * to the original data (for instance, if the encryption keys aren't * loaded). */ for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { zio_t *lio = dde->dde_lead_zio[p]; if (lio != NULL && do_raw) { return (lio->io_size != zio->io_size || abd_cmp(zio->io_abd, lio->io_abd) != 0); } else if (lio != NULL) { return (lio->io_orig_size != zio->io_orig_size || abd_cmp(zio->io_orig_abd, lio->io_orig_abd) != 0); } } for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { ddt_phys_t *ddp = &dde->dde_phys[p]; if (ddp->ddp_phys_birth != 0 && do_raw) { blkptr_t blk = *zio->io_bp; uint64_t psize; abd_t *tmpabd; int error; ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth); psize = BP_GET_PSIZE(&blk); if (psize != zio->io_size) return (B_TRUE); ddt_exit(ddt); tmpabd = abd_alloc_for_io(psize, B_TRUE); error = zio_wait(zio_read(NULL, spa, &blk, tmpabd, psize, NULL, NULL, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_RAW, &zio->io_bookmark)); if (error == 0) { if (abd_cmp(tmpabd, zio->io_abd) != 0) error = SET_ERROR(ENOENT); } abd_free(tmpabd); ddt_enter(ddt); return (error != 0); } else if (ddp->ddp_phys_birth != 0) { arc_buf_t *abuf = NULL; arc_flags_t aflags = ARC_FLAG_WAIT; blkptr_t blk = *zio->io_bp; int error; ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth); if (BP_GET_LSIZE(&blk) != zio->io_orig_size) return (B_TRUE); ddt_exit(ddt); error = arc_read(NULL, spa, &blk, arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &aflags, &zio->io_bookmark); if (error == 0) { if (abd_cmp_buf(zio->io_orig_abd, abuf->b_data, zio->io_orig_size) != 0) error = SET_ERROR(ENOENT); arc_buf_destroy(abuf, &abuf); } ddt_enter(ddt); return (error != 0); } } return (B_FALSE); } static void zio_ddt_child_write_ready(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; zio_t *pio; if (zio->io_error) return; ddt_enter(ddt); ASSERT(dde->dde_lead_zio[p] == zio); ddt_phys_fill(ddp, zio->io_bp); zio_link_t *zl = NULL; while ((pio = zio_walk_parents(zio, &zl)) != NULL) ddt_bp_fill(ddp, pio->io_bp, zio->io_txg); ddt_exit(ddt); } static void zio_ddt_child_write_done(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; ddt_enter(ddt); ASSERT(ddp->ddp_refcnt == 0); ASSERT(dde->dde_lead_zio[p] == zio); dde->dde_lead_zio[p] = NULL; if (zio->io_error == 0) { zio_link_t *zl = NULL; while (zio_walk_parents(zio, &zl) != NULL) ddt_phys_addref(ddp); } else { ddt_phys_clear(ddp); } ddt_exit(ddt); } static void zio_ddt_ditto_write_done(zio_t *zio) { int p = DDT_PHYS_DITTO; ASSERTV(zio_prop_t *zp = &zio->io_prop); blkptr_t *bp = zio->io_bp; ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; ddt_key_t *ddk = &dde->dde_key; ddt_enter(ddt); ASSERT(ddp->ddp_refcnt == 0); ASSERT(dde->dde_lead_zio[p] == zio); dde->dde_lead_zio[p] = NULL; if (zio->io_error == 0) { ASSERT(ZIO_CHECKSUM_EQUAL(bp->blk_cksum, ddk->ddk_cksum)); ASSERT(zp->zp_copies < SPA_DVAS_PER_BP); ASSERT(zp->zp_copies == BP_GET_NDVAS(bp) - BP_IS_GANG(bp)); if (ddp->ddp_phys_birth != 0) ddt_phys_free(ddt, ddk, ddp, zio->io_txg); ddt_phys_fill(ddp, bp); } ddt_exit(ddt); } static zio_t * zio_ddt_write(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; uint64_t txg = zio->io_txg; zio_prop_t *zp = &zio->io_prop; int p = zp->zp_copies; int ditto_copies; zio_t *cio = NULL; zio_t *dio = NULL; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum); ASSERT(BP_IS_HOLE(bp) || zio->io_bp_override); ASSERT(!(zio->io_bp_override && (zio->io_flags & ZIO_FLAG_RAW))); ddt_enter(ddt); dde = ddt_lookup(ddt, bp, B_TRUE); ddp = &dde->dde_phys[p]; if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) { /* * If we're using a weak checksum, upgrade to a strong checksum * and try again. If we're already using a strong checksum, * we can't resolve it, so just convert to an ordinary write. * (And automatically e-mail a paper to Nature?) */ if (!(zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_DEDUP)) { zp->zp_checksum = spa_dedup_checksum(spa); zio_pop_transforms(zio); zio->io_stage = ZIO_STAGE_OPEN; BP_ZERO(bp); } else { zp->zp_dedup = B_FALSE; } zio->io_pipeline = ZIO_WRITE_PIPELINE; ddt_exit(ddt); return (zio); } ditto_copies = ddt_ditto_copies_needed(ddt, dde, ddp); ASSERT(ditto_copies < SPA_DVAS_PER_BP); if (ditto_copies > ddt_ditto_copies_present(dde) && dde->dde_lead_zio[DDT_PHYS_DITTO] == NULL) { zio_prop_t czp = *zp; czp.zp_copies = ditto_copies; /* * If we arrived here with an override bp, we won't have run * the transform stack, so we won't have the data we need to * generate a child i/o. So, toss the override bp and restart. * This is safe, because using the override bp is just an * optimization; and it's rare, so the cost doesn't matter. */ if (zio->io_bp_override) { zio_pop_transforms(zio); zio->io_stage = ZIO_STAGE_OPEN; zio->io_pipeline = ZIO_WRITE_PIPELINE; zio->io_bp_override = NULL; BP_ZERO(bp); ddt_exit(ddt); return (zio); } dio = zio_write(zio, spa, txg, bp, zio->io_orig_abd, zio->io_orig_size, zio->io_orig_size, &czp, NULL, NULL, NULL, zio_ddt_ditto_write_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark); zio_push_transform(dio, zio->io_abd, zio->io_size, 0, NULL); dde->dde_lead_zio[DDT_PHYS_DITTO] = dio; } if (ddp->ddp_phys_birth != 0 || dde->dde_lead_zio[p] != NULL) { if (ddp->ddp_phys_birth != 0) ddt_bp_fill(ddp, bp, txg); if (dde->dde_lead_zio[p] != NULL) zio_add_child(zio, dde->dde_lead_zio[p]); else ddt_phys_addref(ddp); } else if (zio->io_bp_override) { ASSERT(bp->blk_birth == txg); ASSERT(BP_EQUAL(bp, zio->io_bp_override)); ddt_phys_fill(ddp, bp); ddt_phys_addref(ddp); } else { cio = zio_write(zio, spa, txg, bp, zio->io_orig_abd, zio->io_orig_size, zio->io_orig_size, zp, zio_ddt_child_write_ready, NULL, NULL, zio_ddt_child_write_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark); zio_push_transform(cio, zio->io_abd, zio->io_size, 0, NULL); dde->dde_lead_zio[p] = cio; } ddt_exit(ddt); if (cio) zio_nowait(cio); if (dio) zio_nowait(dio); return (zio); } ddt_entry_t *freedde; /* for debugging */ static zio_t * zio_ddt_free(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ddt_enter(ddt); freedde = dde = ddt_lookup(ddt, bp, B_TRUE); if (dde) { ddp = ddt_phys_select(dde, bp); if (ddp) ddt_phys_decref(ddp); } ddt_exit(ddt); return (zio); } /* * ========================================================================== * Allocate and free blocks * ========================================================================== */ static zio_t * zio_io_to_allocate(spa_t *spa, int allocator) { zio_t *zio; ASSERT(MUTEX_HELD(&spa->spa_alloc_locks[allocator])); zio = avl_first(&spa->spa_alloc_trees[allocator]); if (zio == NULL) return (NULL); ASSERT(IO_IS_ALLOCATING(zio)); /* * Try to place a reservation for this zio. If we're unable to * reserve then we throttle. */ ASSERT3U(zio->io_allocator, ==, allocator); if (!metaslab_class_throttle_reserve(spa_normal_class(spa), zio->io_prop.zp_copies, zio->io_allocator, zio, 0)) { return (NULL); } avl_remove(&spa->spa_alloc_trees[allocator], zio); ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE); return (zio); } static zio_t * zio_dva_throttle(zio_t *zio) { spa_t *spa = zio->io_spa; zio_t *nio; if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE || !spa_normal_class(zio->io_spa)->mc_alloc_throttle_enabled || zio->io_child_type == ZIO_CHILD_GANG || zio->io_flags & ZIO_FLAG_NODATA) { return (zio); } ASSERT(zio->io_child_type > ZIO_CHILD_GANG); ASSERT3U(zio->io_queued_timestamp, >, 0); ASSERT(zio->io_stage == ZIO_STAGE_DVA_THROTTLE); zbookmark_phys_t *bm = &zio->io_bookmark; /* * We want to try to use as many allocators as possible to help improve * performance, but we also want logically adjacent IOs to be physically * adjacent to improve sequential read performance. We chunk each object * into 2^20 block regions, and then hash based on the objset, object, * level, and region to accomplish both of these goals. */ zio->io_allocator = cityhash4(bm->zb_objset, bm->zb_object, bm->zb_level, bm->zb_blkid >> 20) % spa->spa_alloc_count; mutex_enter(&spa->spa_alloc_locks[zio->io_allocator]); ASSERT(zio->io_type == ZIO_TYPE_WRITE); avl_add(&spa->spa_alloc_trees[zio->io_allocator], zio); nio = zio_io_to_allocate(zio->io_spa, zio->io_allocator); mutex_exit(&spa->spa_alloc_locks[zio->io_allocator]); return (nio); } void zio_allocate_dispatch(spa_t *spa, int allocator) { zio_t *zio; mutex_enter(&spa->spa_alloc_locks[allocator]); zio = zio_io_to_allocate(spa, allocator); mutex_exit(&spa->spa_alloc_locks[allocator]); if (zio == NULL) return; ASSERT3U(zio->io_stage, ==, ZIO_STAGE_DVA_THROTTLE); ASSERT0(zio->io_error); zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_TRUE); } static zio_t * zio_dva_allocate(zio_t *zio) { spa_t *spa = zio->io_spa; metaslab_class_t *mc = spa_normal_class(spa); blkptr_t *bp = zio->io_bp; int error; int flags = 0; if (zio->io_gang_leader == NULL) { ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; } ASSERT(BP_IS_HOLE(bp)); ASSERT0(BP_GET_NDVAS(bp)); ASSERT3U(zio->io_prop.zp_copies, >, 0); ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa)); ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp)); flags |= (zio->io_flags & ZIO_FLAG_FASTWRITE) ? METASLAB_FASTWRITE : 0; if (zio->io_flags & ZIO_FLAG_NODATA) flags |= METASLAB_DONT_THROTTLE; if (zio->io_flags & ZIO_FLAG_GANG_CHILD) flags |= METASLAB_GANG_CHILD; if (zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE) flags |= METASLAB_ASYNC_ALLOC; error = metaslab_alloc(spa, mc, zio->io_size, bp, zio->io_prop.zp_copies, zio->io_txg, NULL, flags, &zio->io_alloc_list, zio, zio->io_allocator); if (error != 0) { zfs_dbgmsg("%s: metaslab allocation failure: zio %p, " "size %llu, error %d", spa_name(spa), zio, zio->io_size, error); if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE) return (zio_write_gang_block(zio)); zio->io_error = error; } return (zio); } static zio_t * zio_dva_free(zio_t *zio) { metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE); return (zio); } static zio_t * zio_dva_claim(zio_t *zio) { int error; error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg); if (error) zio->io_error = error; return (zio); } /* * Undo an allocation. This is used by zio_done() when an I/O fails * and we want to give back the block we just allocated. * This handles both normal blocks and gang blocks. */ static void zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp) { ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp)); ASSERT(zio->io_bp_override == NULL); if (!BP_IS_HOLE(bp)) metaslab_free(zio->io_spa, bp, bp->blk_birth, B_TRUE); if (gn != NULL) { for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { zio_dva_unallocate(zio, gn->gn_child[g], &gn->gn_gbh->zg_blkptr[g]); } } } /* * Try to allocate an intent log block. Return 0 on success, errno on failure. */ int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp, uint64_t size, boolean_t *slog) { int error = 1; zio_alloc_list_t io_alloc_list; ASSERT(txg > spa_syncing_txg(spa)); metaslab_trace_init(&io_alloc_list); /* * When allocating a zil block, we don't have information about * the final destination of the block except the objset it's part * of, so we just hash the objset ID to pick the allocator to get * some parallelism. */ error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1, txg, NULL, METASLAB_FASTWRITE, &io_alloc_list, NULL, cityhash4(0, 0, 0, os->os_dsl_dataset->ds_object) % spa->spa_alloc_count); if (error == 0) { *slog = TRUE; } else { error = metaslab_alloc(spa, spa_normal_class(spa), size, new_bp, 1, txg, NULL, METASLAB_FASTWRITE, &io_alloc_list, NULL, cityhash4(0, 0, 0, os->os_dsl_dataset->ds_object) % spa->spa_alloc_count); if (error == 0) *slog = FALSE; } metaslab_trace_fini(&io_alloc_list); if (error == 0) { BP_SET_LSIZE(new_bp, size); BP_SET_PSIZE(new_bp, size); BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF); BP_SET_CHECKSUM(new_bp, spa_version(spa) >= SPA_VERSION_SLIM_ZIL ? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG); BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG); BP_SET_LEVEL(new_bp, 0); BP_SET_DEDUP(new_bp, 0); BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER); /* * encrypted blocks will require an IV and salt. We generate * these now since we will not be rewriting the bp at * rewrite time. */ if (os->os_encrypted) { uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t salt[ZIO_DATA_SALT_LEN]; BP_SET_CRYPT(new_bp, B_TRUE); VERIFY0(spa_crypt_get_salt(spa, dmu_objset_id(os), salt)); VERIFY0(zio_crypt_generate_iv(iv)); zio_crypt_encode_params_bp(new_bp, salt, iv); } } else { zfs_dbgmsg("%s: zil block allocation failure: " "size %llu, error %d", spa_name(spa), size, error); } return (error); } /* * ========================================================================== * Read and write to physical devices * ========================================================================== */ /* * Issue an I/O to the underlying vdev. Typically the issue pipeline * stops after this stage and will resume upon I/O completion. * However, there are instances where the vdev layer may need to * continue the pipeline when an I/O was not issued. Since the I/O * that was sent to the vdev layer might be different than the one * currently active in the pipeline (see vdev_queue_io()), we explicitly * force the underlying vdev layers to call either zio_execute() or * zio_interrupt() to ensure that the pipeline continues with the correct I/O. */ static zio_t * zio_vdev_io_start(zio_t *zio) { vdev_t *vd = zio->io_vd; uint64_t align; spa_t *spa = zio->io_spa; zio->io_delay = 0; ASSERT(zio->io_error == 0); ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0); if (vd == NULL) { if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_enter(spa, SCL_ZIO, zio, RW_READER); /* * The mirror_ops handle multiple DVAs in a single BP. */ vdev_mirror_ops.vdev_op_io_start(zio); return (NULL); } ASSERT3P(zio->io_logical, !=, zio); if (zio->io_type == ZIO_TYPE_WRITE) { ASSERT(spa->spa_trust_config); /* * Note: the code can handle other kinds of writes, * but we don't expect them. */ if (zio->io_vd->vdev_removing) { ASSERT(zio->io_flags & (ZIO_FLAG_PHYSICAL | ZIO_FLAG_SELF_HEAL | ZIO_FLAG_RESILVER | ZIO_FLAG_INDUCE_DAMAGE)); } } align = 1ULL << vd->vdev_top->vdev_ashift; if (!(zio->io_flags & ZIO_FLAG_PHYSICAL) && P2PHASE(zio->io_size, align) != 0) { /* Transform logical writes to be a full physical block size. */ uint64_t asize = P2ROUNDUP(zio->io_size, align); abd_t *abuf = abd_alloc_sametype(zio->io_abd, asize); ASSERT(vd == vd->vdev_top); if (zio->io_type == ZIO_TYPE_WRITE) { abd_copy(abuf, zio->io_abd, zio->io_size); abd_zero_off(abuf, zio->io_size, asize - zio->io_size); } zio_push_transform(zio, abuf, asize, asize, zio_subblock); } /* * If this is not a physical io, make sure that it is properly aligned * before proceeding. */ if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) { ASSERT0(P2PHASE(zio->io_offset, align)); ASSERT0(P2PHASE(zio->io_size, align)); } else { /* * For physical writes, we allow 512b aligned writes and assume * the device will perform a read-modify-write as necessary. */ ASSERT0(P2PHASE(zio->io_offset, SPA_MINBLOCKSIZE)); ASSERT0(P2PHASE(zio->io_size, SPA_MINBLOCKSIZE)); } VERIFY(zio->io_type != ZIO_TYPE_WRITE || spa_writeable(spa)); /* * If this is a repair I/O, and there's no self-healing involved -- * that is, we're just resilvering what we expect to resilver -- * then don't do the I/O unless zio's txg is actually in vd's DTL. * This prevents spurious resilvering. * * There are a few ways that we can end up creating these spurious * resilver i/os: * * 1. A resilver i/o will be issued if any DVA in the BP has a * dirty DTL. The mirror code will issue resilver writes to * each DVA, including the one(s) that are not on vdevs with dirty * DTLs. * * 2. With nested replication, which happens when we have a * "replacing" or "spare" vdev that's a child of a mirror or raidz. * For example, given mirror(replacing(A+B), C), it's likely that * only A is out of date (it's the new device). In this case, we'll * read from C, then use the data to resilver A+B -- but we don't * actually want to resilver B, just A. The top-level mirror has no * way to know this, so instead we just discard unnecessary repairs * as we work our way down the vdev tree. * * 3. ZTEST also creates mirrors of mirrors, mirrors of raidz, etc. * The same logic applies to any form of nested replication: ditto * + mirror, RAID-Z + replacing, etc. * * However, indirect vdevs point off to other vdevs which may have * DTL's, so we never bypass them. The child i/os on concrete vdevs * will be properly bypassed instead. */ if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) && !(zio->io_flags & ZIO_FLAG_SELF_HEAL) && zio->io_txg != 0 && /* not a delegated i/o */ vd->vdev_ops != &vdev_indirect_ops && !vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); zio_vdev_io_bypass(zio); return (zio); } if (vd->vdev_ops->vdev_op_leaf && (zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE)) { if (zio->io_type == ZIO_TYPE_READ && vdev_cache_read(zio)) return (zio); if ((zio = vdev_queue_io(zio)) == NULL) return (NULL); if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); zio_interrupt(zio); return (NULL); } zio->io_delay = gethrtime(); } vd->vdev_ops->vdev_op_io_start(zio); return (NULL); } static zio_t * zio_vdev_io_done(zio_t *zio) { vdev_t *vd = zio->io_vd; vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops; boolean_t unexpected_error = B_FALSE; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) { return (NULL); } ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE); if (zio->io_delay) zio->io_delay = gethrtime() - zio->io_delay; if (vd != NULL && vd->vdev_ops->vdev_op_leaf) { vdev_queue_io_done(zio); if (zio->io_type == ZIO_TYPE_WRITE) vdev_cache_write(zio); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_device_injections(vd, zio, EIO, EILSEQ); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_label_injection(zio, EIO); if (zio->io_error) { if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); } else { unexpected_error = B_TRUE; } } } ops->vdev_op_io_done(zio); if (unexpected_error) VERIFY(vdev_probe(vd, zio) == NULL); return (zio); } /* * This function is used to change the priority of an existing zio that is * currently in-flight. This is used by the arc to upgrade priority in the * event that a demand read is made for a block that is currently queued * as a scrub or async read IO. Otherwise, the high priority read request * would end up having to wait for the lower priority IO. */ void zio_change_priority(zio_t *pio, zio_priority_t priority) { zio_t *cio, *cio_next; zio_link_t *zl = NULL; ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE); if (pio->io_vd != NULL && pio->io_vd->vdev_ops->vdev_op_leaf) { vdev_queue_change_io_priority(pio, priority); } else { pio->io_priority = priority; } mutex_enter(&pio->io_lock); for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio, &zl); zio_change_priority(cio, priority); } mutex_exit(&pio->io_lock); } /* * For non-raidz ZIOs, we can just copy aside the bad data read from the * disk, and use that to finish the checksum ereport later. */ static void zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr, const abd_t *good_buf) { /* no processing needed */ zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE); } /*ARGSUSED*/ void zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr, void *ignored) { void *abd = abd_alloc_sametype(zio->io_abd, zio->io_size); abd_copy(abd, zio->io_abd, zio->io_size); zcr->zcr_cbinfo = zio->io_size; zcr->zcr_cbdata = abd; zcr->zcr_finish = zio_vsd_default_cksum_finish; zcr->zcr_free = zio_abd_free; } static zio_t * zio_vdev_io_assess(zio_t *zio) { vdev_t *vd = zio->io_vd; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) { return (NULL); } if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_exit(zio->io_spa, SCL_ZIO, zio); if (zio->io_vsd != NULL) { zio->io_vsd_ops->vsd_free(zio); zio->io_vsd = NULL; } if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_fault_injection(zio, EIO); /* * If the I/O failed, determine whether we should attempt to retry it. * * On retry, we cut in line in the issue queue, since we don't want * compression/checksumming/etc. work to prevent our (cheap) IO reissue. */ if (zio->io_error && vd == NULL && !(zio->io_flags & (ZIO_FLAG_DONT_RETRY | ZIO_FLAG_IO_RETRY))) { ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */ ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */ zio->io_error = 0; zio->io_flags |= ZIO_FLAG_IO_RETRY | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE; zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, zio_requeue_io_start_cut_in_line); return (NULL); } /* * If we got an error on a leaf device, convert it to ENXIO * if the device is not accessible at all. */ if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf && !vdev_accessible(vd, zio)) zio->io_error = SET_ERROR(ENXIO); /* * If we can't write to an interior vdev (mirror or RAID-Z), * set vdev_cant_write so that we stop trying to allocate from it. */ if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE && vd != NULL && !vd->vdev_ops->vdev_op_leaf) { vd->vdev_cant_write = B_TRUE; } /* * If a cache flush returns ENOTSUP or ENOTTY, we know that no future * attempts will ever succeed. In this case we set a persistent bit so * that we don't bother with it in the future. */ if ((zio->io_error == ENOTSUP || zio->io_error == ENOTTY) && zio->io_type == ZIO_TYPE_IOCTL && zio->io_cmd == DKIOCFLUSHWRITECACHE && vd != NULL) vd->vdev_nowritecache = B_TRUE; if (zio->io_error) zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (vd != NULL && vd->vdev_ops->vdev_op_leaf && zio->io_physdone != NULL) { ASSERT(!(zio->io_flags & ZIO_FLAG_DELEGATED)); ASSERT(zio->io_child_type == ZIO_CHILD_VDEV); zio->io_physdone(zio->io_logical); } return (zio); } void zio_vdev_io_reissue(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_stage >>= 1; } void zio_vdev_io_redone(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE); zio->io_stage >>= 1; } void zio_vdev_io_bypass(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_flags |= ZIO_FLAG_IO_BYPASS; zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1; } /* * ========================================================================== * Encrypt and store encryption parameters * ========================================================================== */ /* * This function is used for ZIO_STAGE_ENCRYPT. It is responsible for * managing the storage of encryption parameters and passing them to the * lower-level encryption functions. */ static zio_t * zio_encrypt(zio_t *zio) { zio_prop_t *zp = &zio->io_prop; spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; uint64_t psize = BP_GET_PSIZE(bp); uint64_t dsobj = zio->io_bookmark.zb_objset; dmu_object_type_t ot = BP_GET_TYPE(bp); void *enc_buf = NULL; abd_t *eabd = NULL; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; boolean_t no_crypt = B_FALSE; /* the root zio already encrypted the data */ if (zio->io_child_type == ZIO_CHILD_GANG) return (zio); /* only ZIL blocks are re-encrypted on rewrite */ if (!IO_IS_ALLOCATING(zio) && ot != DMU_OT_INTENT_LOG) return (zio); if (!(zp->zp_encrypt || BP_IS_ENCRYPTED(bp))) { BP_SET_CRYPT(bp, B_FALSE); return (zio); } /* if we are doing raw encryption set the provided encryption params */ if (zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) { ASSERT0(BP_GET_LEVEL(bp)); BP_SET_CRYPT(bp, B_TRUE); BP_SET_BYTEORDER(bp, zp->zp_byteorder); if (ot != DMU_OT_OBJSET) zio_crypt_encode_mac_bp(bp, zp->zp_mac); /* dnode blocks must be written out in the provided byteorder */ if (zp->zp_byteorder != ZFS_HOST_BYTEORDER && ot == DMU_OT_DNODE) { void *bswap_buf = zio_buf_alloc(psize); abd_t *babd = abd_get_from_buf(bswap_buf, psize); ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); abd_copy_to_buf(bswap_buf, zio->io_abd, psize); dmu_ot_byteswap[DMU_OT_BYTESWAP(ot)].ob_func(bswap_buf, psize); abd_take_ownership_of_buf(babd, B_TRUE); zio_push_transform(zio, babd, psize, psize, NULL); } if (DMU_OT_IS_ENCRYPTED(ot)) zio_crypt_encode_params_bp(bp, zp->zp_salt, zp->zp_iv); return (zio); } /* indirect blocks only maintain a cksum of the lower level MACs */ if (BP_GET_LEVEL(bp) > 0) { BP_SET_CRYPT(bp, B_TRUE); VERIFY0(zio_crypt_do_indirect_mac_checksum_abd(B_TRUE, zio->io_orig_abd, BP_GET_LSIZE(bp), BP_SHOULD_BYTESWAP(bp), mac)); zio_crypt_encode_mac_bp(bp, mac); return (zio); } /* * Objset blocks are a special case since they have 2 256-bit MACs * embedded within them. */ if (ot == DMU_OT_OBJSET) { ASSERT0(DMU_OT_IS_ENCRYPTED(ot)); ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); BP_SET_CRYPT(bp, B_TRUE); VERIFY0(spa_do_crypt_objset_mac_abd(B_TRUE, spa, dsobj, zio->io_abd, psize, BP_SHOULD_BYTESWAP(bp))); return (zio); } /* unencrypted object types are only authenticated with a MAC */ if (!DMU_OT_IS_ENCRYPTED(ot)) { BP_SET_CRYPT(bp, B_TRUE); VERIFY0(spa_do_crypt_mac_abd(B_TRUE, spa, dsobj, zio->io_abd, psize, mac)); zio_crypt_encode_mac_bp(bp, mac); return (zio); } /* * Later passes of sync-to-convergence may decide to rewrite data * in place to avoid more disk reallocations. This presents a problem * for encryption because this consitutes rewriting the new data with * the same encryption key and IV. However, this only applies to blocks * in the MOS (particularly the spacemaps) and we do not encrypt the * MOS. We assert that the zio is allocating or an intent log write * to enforce this. */ ASSERT(IO_IS_ALLOCATING(zio) || ot == DMU_OT_INTENT_LOG); ASSERT(BP_GET_LEVEL(bp) == 0 || ot == DMU_OT_INTENT_LOG); ASSERT(spa_feature_is_active(spa, SPA_FEATURE_ENCRYPTION)); ASSERT3U(psize, !=, 0); enc_buf = zio_buf_alloc(psize); eabd = abd_get_from_buf(enc_buf, psize); abd_take_ownership_of_buf(eabd, B_TRUE); /* * For an explanation of what encryption parameters are stored * where, see the block comment in zio_crypt.c. */ if (ot == DMU_OT_INTENT_LOG) { zio_crypt_decode_params_bp(bp, salt, iv); } else { BP_SET_CRYPT(bp, B_TRUE); } /* Perform the encryption. This should not fail */ VERIFY0(spa_do_crypt_abd(B_TRUE, spa, &zio->io_bookmark, BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, psize, zio->io_abd, eabd, &no_crypt)); /* encode encryption metadata into the bp */ if (ot == DMU_OT_INTENT_LOG) { /* * ZIL blocks store the MAC in the embedded checksum, so the * transform must always be applied. */ zio_crypt_encode_mac_zil(enc_buf, mac); zio_push_transform(zio, eabd, psize, psize, NULL); } else { BP_SET_CRYPT(bp, B_TRUE); zio_crypt_encode_params_bp(bp, salt, iv); zio_crypt_encode_mac_bp(bp, mac); if (no_crypt) { ASSERT3U(ot, ==, DMU_OT_DNODE); abd_free(eabd); } else { zio_push_transform(zio, eabd, psize, psize, NULL); } } return (zio); } /* * ========================================================================== * Generate and verify checksums * ========================================================================== */ static zio_t * zio_checksum_generate(zio_t *zio) { blkptr_t *bp = zio->io_bp; enum zio_checksum checksum; if (bp == NULL) { /* * This is zio_write_phys(). * We're either generating a label checksum, or none at all. */ checksum = zio->io_prop.zp_checksum; if (checksum == ZIO_CHECKSUM_OFF) return (zio); ASSERT(checksum == ZIO_CHECKSUM_LABEL); } else { if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) { ASSERT(!IO_IS_ALLOCATING(zio)); checksum = ZIO_CHECKSUM_GANG_HEADER; } else { checksum = BP_GET_CHECKSUM(bp); } } zio_checksum_compute(zio, checksum, zio->io_abd, zio->io_size); return (zio); } static zio_t * zio_checksum_verify(zio_t *zio) { zio_bad_cksum_t info; blkptr_t *bp = zio->io_bp; int error; ASSERT(zio->io_vd != NULL); if (bp == NULL) { /* * This is zio_read_phys(). * We're either verifying a label checksum, or nothing at all. */ if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF) return (zio); ASSERT(zio->io_prop.zp_checksum == ZIO_CHECKSUM_LABEL); } if ((error = zio_checksum_error(zio, &info)) != 0) { zio->io_error = error; if (error == ECKSUM && !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) { zfs_ereport_start_checksum(zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, zio->io_offset, zio->io_size, NULL, &info); } } return (zio); } /* * Called by RAID-Z to ensure we don't compute the checksum twice. */ void zio_checksum_verified(zio_t *zio) { zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } /* * ========================================================================== * Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other. * An error of 0 indicates success. ENXIO indicates whole-device failure, * which may be transient (e.g. unplugged) or permament. ECKSUM and EIO * indicate errors that are specific to one I/O, and most likely permanent. * Any other error is presumed to be worse because we weren't expecting it. * ========================================================================== */ int zio_worst_error(int e1, int e2) { static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO }; int r1, r2; for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++) if (e1 == zio_error_rank[r1]) break; for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++) if (e2 == zio_error_rank[r2]) break; return (r1 > r2 ? e1 : e2); } /* * ========================================================================== * I/O completion * ========================================================================== */ static zio_t * zio_ready(zio_t *zio) { blkptr_t *bp = zio->io_bp; zio_t *pio, *pio_next; zio_link_t *zl = NULL; if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT | ZIO_CHILD_DDT_BIT, ZIO_WAIT_READY)) { return (NULL); } if (zio->io_ready) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp) || (zio->io_flags & ZIO_FLAG_NOPWRITE)); ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0); zio->io_ready(zio); } if (bp != NULL && bp != &zio->io_bp_copy) zio->io_bp_copy = *bp; if (zio->io_error != 0) { zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); /* * We were unable to allocate anything, unreserve and * issue the next I/O to allocate. */ metaslab_class_throttle_unreserve( spa_normal_class(zio->io_spa), zio->io_prop.zp_copies, zio->io_allocator, zio); zio_allocate_dispatch(zio->io_spa, zio->io_allocator); } } mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_READY] = 1; pio = zio_walk_parents(zio, &zl); mutex_exit(&zio->io_lock); /* * As we notify zio's parents, new parents could be added. * New parents go to the head of zio's io_parent_list, however, * so we will (correctly) not notify them. The remainder of zio's * io_parent_list, from 'pio_next' onward, cannot change because * all parents must wait for us to be done before they can be done. */ for (; pio != NULL; pio = pio_next) { pio_next = zio_walk_parents(zio, &zl); zio_notify_parent(pio, zio, ZIO_WAIT_READY, NULL); } if (zio->io_flags & ZIO_FLAG_NODATA) { if (BP_IS_GANG(bp)) { zio->io_flags &= ~ZIO_FLAG_NODATA; } else { ASSERT((uintptr_t)zio->io_abd < SPA_MAXBLOCKSIZE); zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } } if (zio_injection_enabled && zio->io_spa->spa_syncing_txg == zio->io_txg) zio_handle_ignored_writes(zio); return (zio); } /* * Update the allocation throttle accounting. */ static void zio_dva_throttle_done(zio_t *zio) { ASSERTV(zio_t *lio = zio->io_logical); zio_t *pio = zio_unique_parent(zio); vdev_t *vd = zio->io_vd; int flags = METASLAB_ASYNC_ALLOC; ASSERT3P(zio->io_bp, !=, NULL); ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE); ASSERT3U(zio->io_priority, ==, ZIO_PRIORITY_ASYNC_WRITE); ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV); ASSERT(vd != NULL); ASSERT3P(vd, ==, vd->vdev_top); ASSERT(zio_injection_enabled || !(zio->io_flags & ZIO_FLAG_IO_RETRY)); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR)); ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING); ASSERT(!(lio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(!(lio->io_orig_flags & ZIO_FLAG_NODATA)); /* * Parents of gang children can have two flavors -- ones that * allocated the gang header (will have ZIO_FLAG_IO_REWRITE set) * and ones that allocated the constituent blocks. The allocation * throttle needs to know the allocating parent zio so we must find * it here. */ if (pio->io_child_type == ZIO_CHILD_GANG) { /* * If our parent is a rewrite gang child then our grandparent * would have been the one that performed the allocation. */ if (pio->io_flags & ZIO_FLAG_IO_REWRITE) pio = zio_unique_parent(pio); flags |= METASLAB_GANG_CHILD; } ASSERT(IO_IS_ALLOCATING(pio)); ASSERT3P(zio, !=, zio->io_logical); ASSERT(zio->io_logical != NULL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR)); ASSERT0(zio->io_flags & ZIO_FLAG_NOPWRITE); mutex_enter(&pio->io_lock); metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, pio, flags, pio->io_allocator, B_TRUE); mutex_exit(&pio->io_lock); metaslab_class_throttle_unreserve(spa_normal_class(zio->io_spa), 1, pio->io_allocator, pio); /* * Call into the pipeline to see if there is more work that * needs to be done. If there is work to be done it will be * dispatched to another taskq thread. */ zio_allocate_dispatch(zio->io_spa, pio->io_allocator); } static zio_t * zio_done(zio_t *zio) { /* * Always attempt to keep stack usage minimal here since * we can be called recurisvely up to 19 levels deep. */ const uint64_t psize = zio->io_size; zio_t *pio, *pio_next; ASSERTV(metaslab_class_t *mc = spa_normal_class(zio->io_spa)); zio_link_t *zl = NULL; /* * If our children haven't all completed, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_ALL_BITS, ZIO_WAIT_DONE)) { return (NULL); } /* * If the allocation throttle is enabled, then update the accounting. * We only track child I/Os that are part of an allocating async * write. We must do this since the allocation is performed * by the logical I/O but the actual write is done by child I/Os. */ if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING && zio->io_child_type == ZIO_CHILD_VDEV) { ASSERT(mc->mc_alloc_throttle_enabled); zio_dva_throttle_done(zio); } /* * If the allocation throttle is enabled, verify that * we have decremented the refcounts for every I/O that was throttled. */ if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(zio->io_bp != NULL); metaslab_group_alloc_verify(zio->io_spa, zio->io_bp, zio, zio->io_allocator); VERIFY(refcount_not_held(&mc->mc_alloc_slots[zio->io_allocator], zio)); } for (int c = 0; c < ZIO_CHILD_TYPES; c++) for (int w = 0; w < ZIO_WAIT_TYPES; w++) ASSERT(zio->io_children[c][w] == 0); if (zio->io_bp != NULL && !BP_IS_EMBEDDED(zio->io_bp)) { ASSERT(zio->io_bp->blk_pad[0] == 0); ASSERT(zio->io_bp->blk_pad[1] == 0); ASSERT(bcmp(zio->io_bp, &zio->io_bp_copy, sizeof (blkptr_t)) == 0 || (zio->io_bp == zio_unique_parent(zio)->io_bp)); if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(zio->io_bp) && zio->io_bp_override == NULL && !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) { ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp)); ASSERT(BP_COUNT_GANG(zio->io_bp) == 0 || (BP_COUNT_GANG(zio->io_bp) == BP_GET_NDVAS(zio->io_bp))); } if (zio->io_flags & ZIO_FLAG_NOPWRITE) VERIFY(BP_EQUAL(zio->io_bp, &zio->io_bp_orig)); } /* * If there were child vdev/gang/ddt errors, they apply to us now. */ zio_inherit_child_errors(zio, ZIO_CHILD_VDEV); zio_inherit_child_errors(zio, ZIO_CHILD_GANG); zio_inherit_child_errors(zio, ZIO_CHILD_DDT); /* * If the I/O on the transformed data was successful, generate any * checksum reports now while we still have the transformed data. */ if (zio->io_error == 0) { while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; uint64_t align = zcr->zcr_align; uint64_t asize = P2ROUNDUP(psize, align); abd_t *adata = zio->io_abd; if (asize != psize) { adata = abd_alloc(asize, B_TRUE); abd_copy(adata, zio->io_abd, psize); abd_zero_off(adata, psize, asize - psize); } zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, adata); zfs_ereport_free_checksum(zcr); if (asize != psize) abd_free(adata); } } zio_pop_transforms(zio); /* note: may set zio->io_error */ vdev_stat_update(zio, psize); /* * If this I/O is attached to a particular vdev is slow, exceeding * 30 seconds to complete, post an error described the I/O delay. * We ignore these errors if the device is currently unavailable. */ if (zio->io_delay >= MSEC2NSEC(zio_delay_max)) { if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) zfs_ereport_post(FM_EREPORT_ZFS_DELAY, zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0, 0); } if (zio->io_error) { /* * If this I/O is attached to a particular vdev, * generate an error message describing the I/O failure * at the block level. We ignore these errors if the * device is currently unavailable. */ if (zio->io_error != ECKSUM && zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) zfs_ereport_post(FM_EREPORT_ZFS_IO, zio->io_spa, zio->io_vd, &zio->io_bookmark, zio, 0, 0); if ((zio->io_error == EIO || !(zio->io_flags & (ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_PROPAGATE))) && zio == zio->io_logical) { /* * For logical I/O requests, tell the SPA to log the * error and generate a logical data ereport. */ spa_log_error(zio->io_spa, &zio->io_bookmark); zfs_ereport_post(FM_EREPORT_ZFS_DATA, zio->io_spa, NULL, &zio->io_bookmark, zio, 0, 0); } } if (zio->io_error && zio == zio->io_logical) { /* * Determine whether zio should be reexecuted. This will * propagate all the way to the root via zio_notify_parent(). */ ASSERT(zio->io_vd == NULL && zio->io_bp != NULL); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (IO_IS_ALLOCATING(zio) && !(zio->io_flags & ZIO_FLAG_CANFAIL)) { if (zio->io_error != ENOSPC) zio->io_reexecute |= ZIO_REEXECUTE_NOW; else zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; } if ((zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_FREE) && !(zio->io_flags & ZIO_FLAG_SCAN_THREAD) && zio->io_error == ENXIO && spa_load_state(zio->io_spa) == SPA_LOAD_NONE && spa_get_failmode(zio->io_spa) != ZIO_FAILURE_MODE_CONTINUE) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; /* * Here is a possibly good place to attempt to do * either combinatorial reconstruction or error correction * based on checksums. It also might be a good place * to send out preliminary ereports before we suspend * processing. */ } /* * If there were logical child errors, they apply to us now. * We defer this until now to avoid conflating logical child * errors with errors that happened to the zio itself when * updating vdev stats and reporting FMA events above. */ zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL); if ((zio->io_error || zio->io_reexecute) && IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio && !(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE))) zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp); zio_gang_tree_free(&zio->io_gang_tree); /* * Godfather I/Os should never suspend. */ if ((zio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) zio->io_reexecute &= ~ZIO_REEXECUTE_SUSPEND; if (zio->io_reexecute) { /* * This is a logical I/O that wants to reexecute. * * Reexecute is top-down. When an i/o fails, if it's not * the root, it simply notifies its parent and sticks around. * The parent, seeing that it still has children in zio_done(), * does the same. This percolates all the way up to the root. * The root i/o will reexecute or suspend the entire tree. * * This approach ensures that zio_reexecute() honors * all the original i/o dependency relationships, e.g. * parents not executing until children are ready. */ ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); zio->io_gang_leader = NULL; mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); /* * "The Godfather" I/O monitors its children but is * not a true parent to them. It will track them through * the pipeline but severs its ties whenever they get into * trouble (e.g. suspended). This allows "The Godfather" * I/O to return status without blocking. */ zl = NULL; for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) { zio_link_t *remove_zl = zl; pio_next = zio_walk_parents(zio, &zl); if ((pio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) { zio_remove_child(pio, zio, remove_zl); /* * This is a rare code path, so we don't * bother with "next_to_execute". */ zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL); } } if ((pio = zio_unique_parent(zio)) != NULL) { /* * We're not a root i/o, so there's nothing to do * but notify our parent. Don't propagate errors * upward since we haven't permanently failed yet. */ ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); zio->io_flags |= ZIO_FLAG_DONT_PROPAGATE; /* * This is a rare code path, so we don't bother with * "next_to_execute". */ zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL); } else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) { /* * We'd fail again if we reexecuted now, so suspend * until conditions improve (e.g. device comes online). */ zio_suspend(zio->io_spa, zio, ZIO_SUSPEND_IOERR); } else { /* * Reexecution is potentially a huge amount of work. * Hand it off to the otherwise-unused claim taskq. */ ASSERT(taskq_empty_ent(&zio->io_tqent)); spa_taskq_dispatch_ent(zio->io_spa, ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE, (task_func_t *)zio_reexecute, zio, 0, &zio->io_tqent); } return (NULL); } ASSERT(zio->io_child_count == 0); ASSERT(zio->io_reexecute == 0); ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL)); /* * Report any checksum errors, since the I/O is complete. */ while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, NULL); zfs_ereport_free_checksum(zcr); } if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp && !BP_IS_HOLE(zio->io_bp) && !BP_IS_EMBEDDED(zio->io_bp) && !(zio->io_flags & ZIO_FLAG_NOPWRITE)) { metaslab_fastwrite_unmark(zio->io_spa, zio->io_bp); } /* * It is the responsibility of the done callback to ensure that this * particular zio is no longer discoverable for adoption, and as * such, cannot acquire any new parents. */ if (zio->io_done) zio->io_done(zio); mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); /* * We are done executing this zio. We may want to execute a parent * next. See the comment in zio_notify_parent(). */ zio_t *next_to_execute = NULL; zl = NULL; for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) { zio_link_t *remove_zl = zl; pio_next = zio_walk_parents(zio, &zl); zio_remove_child(pio, zio, remove_zl); zio_notify_parent(pio, zio, ZIO_WAIT_DONE, &next_to_execute); } if (zio->io_waiter != NULL) { mutex_enter(&zio->io_lock); zio->io_executor = NULL; cv_broadcast(&zio->io_cv); mutex_exit(&zio->io_lock); } else { zio_destroy(zio); } return (next_to_execute); } /* * ========================================================================== * I/O pipeline definition * ========================================================================== */ static zio_pipe_stage_t *zio_pipeline[] = { NULL, zio_read_bp_init, zio_write_bp_init, zio_free_bp_init, zio_issue_async, zio_write_compress, zio_encrypt, zio_checksum_generate, zio_nop_write, zio_ddt_read_start, zio_ddt_read_done, zio_ddt_write, zio_ddt_free, zio_gang_assemble, zio_gang_issue, zio_dva_throttle, zio_dva_allocate, zio_dva_free, zio_dva_claim, zio_ready, zio_vdev_io_start, zio_vdev_io_done, zio_vdev_io_assess, zio_checksum_verify, zio_done }; /* * Compare two zbookmark_phys_t's to see which we would reach first in a * pre-order traversal of the object tree. * * This is simple in every case aside from the meta-dnode object. For all other * objects, we traverse them in order (object 1 before object 2, and so on). * However, all of these objects are traversed while traversing object 0, since * the data it points to is the list of objects. Thus, we need to convert to a * canonical representation so we can compare meta-dnode bookmarks to * non-meta-dnode bookmarks. * * We do this by calculating "equivalents" for each field of the zbookmark. * zbookmarks outside of the meta-dnode use their own object and level, and * calculate the level 0 equivalent (the first L0 blkid that is contained in the * blocks this bookmark refers to) by multiplying their blkid by their span * (the number of L0 blocks contained within one block at their level). * zbookmarks inside the meta-dnode calculate their object equivalent * (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use * level + 1<<31 (any value larger than a level could ever be) for their level. * This causes them to always compare before a bookmark in their object * equivalent, compare appropriately to bookmarks in other objects, and to * compare appropriately to other bookmarks in the meta-dnode. */ int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2) { /* * These variables represent the "equivalent" values for the zbookmark, * after converting zbookmarks inside the meta dnode to their * normal-object equivalents. */ uint64_t zb1obj, zb2obj; uint64_t zb1L0, zb2L0; uint64_t zb1level, zb2level; if (zb1->zb_object == zb2->zb_object && zb1->zb_level == zb2->zb_level && zb1->zb_blkid == zb2->zb_blkid) return (0); /* * BP_SPANB calculates the span in blocks. */ zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level); zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level); if (zb1->zb_object == DMU_META_DNODE_OBJECT) { zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb1L0 = 0; zb1level = zb1->zb_level + COMPARE_META_LEVEL; } else { zb1obj = zb1->zb_object; zb1level = zb1->zb_level; } if (zb2->zb_object == DMU_META_DNODE_OBJECT) { zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb2L0 = 0; zb2level = zb2->zb_level + COMPARE_META_LEVEL; } else { zb2obj = zb2->zb_object; zb2level = zb2->zb_level; } /* Now that we have a canonical representation, do the comparison. */ if (zb1obj != zb2obj) return (zb1obj < zb2obj ? -1 : 1); else if (zb1L0 != zb2L0) return (zb1L0 < zb2L0 ? -1 : 1); else if (zb1level != zb2level) return (zb1level > zb2level ? -1 : 1); /* * This can (theoretically) happen if the bookmarks have the same object * and level, but different blkids, if the block sizes are not the same. * There is presently no way to change the indirect block sizes */ return (0); } /* * This function checks the following: given that last_block is the place that * our traversal stopped last time, does that guarantee that we've visited * every node under subtree_root? Therefore, we can't just use the raw output * of zbookmark_compare. We have to pass in a modified version of * subtree_root; by incrementing the block id, and then checking whether * last_block is before or equal to that, we can tell whether or not having * visited last_block implies that all of subtree_root's children have been * visited. */ boolean_t zbookmark_subtree_completed(const dnode_phys_t *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block) { zbookmark_phys_t mod_zb = *subtree_root; mod_zb.zb_blkid++; ASSERT(last_block->zb_level == 0); /* The objset_phys_t isn't before anything. */ if (dnp == NULL) return (B_FALSE); /* * We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the * data block size in sectors, because that variable is only used if * the bookmark refers to a block in the meta-dnode. Since we don't * know without examining it what object it refers to, and there's no * harm in passing in this value in other cases, we always pass it in. * * We pass in 0 for the indirect block size shift because zb2 must be * level 0. The indirect block size is only used to calculate the span * of the bookmark, but since the bookmark must be level 0, the span is * always 1, so the math works out. * * If you make changes to how the zbookmark_compare code works, be sure * to make sure that this code still works afterwards. */ return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift, 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb, last_block) <= 0); } #if defined(_KERNEL) EXPORT_SYMBOL(zio_type_name); EXPORT_SYMBOL(zio_buf_alloc); EXPORT_SYMBOL(zio_data_buf_alloc); EXPORT_SYMBOL(zio_buf_free); EXPORT_SYMBOL(zio_data_buf_free); module_param(zio_delay_max, int, 0644); MODULE_PARM_DESC(zio_delay_max, "Max zio millisec delay before posting event"); module_param(zio_requeue_io_start_cut_in_line, int, 0644); MODULE_PARM_DESC(zio_requeue_io_start_cut_in_line, "Prioritize requeued I/O"); module_param(zfs_sync_pass_deferred_free, int, 0644); MODULE_PARM_DESC(zfs_sync_pass_deferred_free, "Defer frees starting in this pass"); module_param(zfs_sync_pass_dont_compress, int, 0644); MODULE_PARM_DESC(zfs_sync_pass_dont_compress, "Don't compress starting in this pass"); module_param(zfs_sync_pass_rewrite, int, 0644); MODULE_PARM_DESC(zfs_sync_pass_rewrite, "Rewrite new bps starting in this pass"); module_param(zio_dva_throttle_enabled, int, 0644); MODULE_PARM_DESC(zio_dva_throttle_enabled, "Throttle block allocations in the ZIO pipeline"); #endif diff --git a/module/zfs/zio_compress.c b/module/zfs/zio_compress.c index 971e8de8b5a0..f5cbc3e8218a 100644 --- a/module/zfs/zio_compress.c +++ b/module/zfs/zio_compress.c @@ -1,152 +1,168 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2009 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ /* * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. */ /* - * Copyright (c) 2013, 2016 by Delphix. All rights reserved. + * Copyright (c) 2013, 2018 by Delphix. All rights reserved. */ #include #include #include #include #include +/* + * If nonzero, every 1/X decompression attempts will fail, simulating + * an undetected memory error. + */ +unsigned long zio_decompress_fail_fraction = 0; + /* * Compression vectors. */ zio_compress_info_t zio_compress_table[ZIO_COMPRESS_FUNCTIONS] = { {"inherit", 0, NULL, NULL}, {"on", 0, NULL, NULL}, {"uncompressed", 0, NULL, NULL}, {"lzjb", 0, lzjb_compress, lzjb_decompress}, {"empty", 0, NULL, NULL}, {"gzip-1", 1, gzip_compress, gzip_decompress}, {"gzip-2", 2, gzip_compress, gzip_decompress}, {"gzip-3", 3, gzip_compress, gzip_decompress}, {"gzip-4", 4, gzip_compress, gzip_decompress}, {"gzip-5", 5, gzip_compress, gzip_decompress}, {"gzip-6", 6, gzip_compress, gzip_decompress}, {"gzip-7", 7, gzip_compress, gzip_decompress}, {"gzip-8", 8, gzip_compress, gzip_decompress}, {"gzip-9", 9, gzip_compress, gzip_decompress}, {"zle", 64, zle_compress, zle_decompress}, {"lz4", 0, lz4_compress_zfs, lz4_decompress_zfs} }; enum zio_compress zio_compress_select(spa_t *spa, enum zio_compress child, enum zio_compress parent) { enum zio_compress result; ASSERT(child < ZIO_COMPRESS_FUNCTIONS); ASSERT(parent < ZIO_COMPRESS_FUNCTIONS); ASSERT(parent != ZIO_COMPRESS_INHERIT); result = child; if (result == ZIO_COMPRESS_INHERIT) result = parent; if (result == ZIO_COMPRESS_ON) { if (spa_feature_is_active(spa, SPA_FEATURE_LZ4_COMPRESS)) result = ZIO_COMPRESS_LZ4_ON_VALUE; else result = ZIO_COMPRESS_LEGACY_ON_VALUE; } return (result); } /*ARGSUSED*/ static int zio_compress_zeroed_cb(void *data, size_t len, void *private) { uint64_t *end = (uint64_t *)((char *)data + len); for (uint64_t *word = (uint64_t *)data; word < end; word++) if (*word != 0) return (1); return (0); } size_t zio_compress_data(enum zio_compress c, abd_t *src, void *dst, size_t s_len) { size_t c_len, d_len; zio_compress_info_t *ci = &zio_compress_table[c]; ASSERT((uint_t)c < ZIO_COMPRESS_FUNCTIONS); ASSERT((uint_t)c == ZIO_COMPRESS_EMPTY || ci->ci_compress != NULL); /* * If the data is all zeroes, we don't even need to allocate * a block for it. We indicate this by returning zero size. */ if (abd_iterate_func(src, 0, s_len, zio_compress_zeroed_cb, NULL) == 0) return (0); if (c == ZIO_COMPRESS_EMPTY) return (s_len); /* Compress at least 12.5% */ d_len = s_len - (s_len >> 3); /* No compression algorithms can read from ABDs directly */ void *tmp = abd_borrow_buf_copy(src, s_len); c_len = ci->ci_compress(tmp, dst, s_len, d_len, ci->ci_level); abd_return_buf(src, tmp, s_len); if (c_len > d_len) return (s_len); ASSERT3U(c_len, <=, d_len); return (c_len); } int zio_decompress_data_buf(enum zio_compress c, void *src, void *dst, size_t s_len, size_t d_len) { zio_compress_info_t *ci = &zio_compress_table[c]; if ((uint_t)c >= ZIO_COMPRESS_FUNCTIONS || ci->ci_decompress == NULL) return (SET_ERROR(EINVAL)); return (ci->ci_decompress(src, dst, s_len, d_len, ci->ci_level)); } int zio_decompress_data(enum zio_compress c, abd_t *src, void *dst, size_t s_len, size_t d_len) { void *tmp = abd_borrow_buf_copy(src, s_len); int ret = zio_decompress_data_buf(c, tmp, dst, s_len, d_len); abd_return_buf(src, tmp, s_len); + /* + * Decompression shouldn't fail, because we've already verifyied + * the checksum. However, for extra protection (e.g. against bitflips + * in non-ECC RAM), we handle this error (and test it). + */ + ASSERT0(ret); + if (zio_decompress_fail_fraction != 0 && + spa_get_random(zio_decompress_fail_fraction) == 0) + ret = SET_ERROR(EINVAL); + return (ret); } diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run index a9594aeffd01..0a64e8cd5ebc 100644 --- a/tests/runfiles/linux.run +++ b/tests/runfiles/linux.run @@ -1,856 +1,856 @@ # # This file and its contents are supplied under the terms of the # Common Development and Distribution License ("CDDL"), version 1.0. # You may only use this file in accordance with the terms of version # 1.0 of the CDDL. # # A full copy of the text of the CDDL should have accompanied this # source. A copy of the CDDL is also available via the Internet at # http://www.illumos.org/license/CDDL. # [DEFAULT] pre = setup quiet = False pre_user = root user = root timeout = 600 post_user = root post = cleanup outputdir = /var/tmp/test_results tags = ['functional'] [tests/functional/acl/posix] tests = ['posix_001_pos', 'posix_002_pos', 'posix_003_pos'] tags = ['functional', 'acl', 'posix'] [tests/functional/arc] tests = ['dbufstats_001_pos', 'dbufstats_002_pos'] tags = ['functional', 'arc'] [tests/functional/atime] tests = ['atime_001_pos', 'atime_002_neg', 'atime_003_pos'] tags = ['functional', 'atime'] [tests/functional/bootfs] tests = ['bootfs_001_pos', 'bootfs_002_neg', 'bootfs_003_pos', 'bootfs_004_neg', 'bootfs_005_neg', 'bootfs_006_pos', 'bootfs_007_pos', 'bootfs_008_pos'] tags = ['functional', 'bootfs'] [tests/functional/cache] tests = ['cache_001_pos', 'cache_002_pos', 'cache_003_pos', 'cache_004_neg', 'cache_005_neg', 'cache_006_pos', 'cache_007_neg', 'cache_008_neg', 'cache_009_pos', 'cache_010_neg', 'cache_011_pos'] tags = ['functional', 'cache'] [tests/functional/cachefile] tests = ['cachefile_001_pos', 'cachefile_002_pos', 'cachefile_003_pos', 'cachefile_004_pos'] tags = ['functional', 'cachefile'] [tests/functional/casenorm] tests = ['case_all_values', 'norm_all_values', 'mixed_create_failure', 'sensitive_none_lookup', 'sensitive_none_delete', 'sensitive_formd_lookup', 'sensitive_formd_delete', 'insensitive_none_lookup', 'insensitive_none_delete', 'insensitive_formd_lookup', 'insensitive_formd_delete', 'mixed_none_lookup', 'mixed_none_lookup_ci', 'mixed_none_delete', 'mixed_formd_lookup', 'mixed_formd_lookup_ci', 'mixed_formd_delete'] tags = ['functional', 'casenorm'] [tests/functional/channel_program/lua_core] tests = ['tst.args_to_lua', 'tst.divide_by_zero', 'tst.exists', 'tst.integer_illegal', 'tst.integer_overflow', 'tst.language_functions_neg', 'tst.language_functions_pos', 'tst.large_prog', 'tst.libraries', 'tst.memory_limit', 'tst.nested_neg', 'tst.nested_pos', 'tst.nvlist_to_lua', 'tst.recursive_neg', 'tst.recursive_pos', 'tst.return_large', 'tst.return_nvlist_neg', 'tst.return_nvlist_pos', 'tst.return_recursive_table', 'tst.timeout'] tags = ['functional', 'channel_program', 'lua_core'] [tests/functional/channel_program/synctask_core] tests = ['tst.destroy_fs', 'tst.destroy_snap', 'tst.get_count_and_limit', 'tst.get_index_props', 'tst.get_mountpoint', 'tst.get_neg', 'tst.get_number_props', 'tst.get_string_props', 'tst.get_type', 'tst.get_userquota', 'tst.get_written', 'tst.list_children', 'tst.list_clones', 'tst.list_snapshots', 'tst.list_system_props', 'tst.list_user_props', 'tst.parse_args_neg','tst.promote_conflict', 'tst.promote_multiple', 'tst.promote_simple', 'tst.rollback_mult', 'tst.rollback_one', 'tst.snapshot_destroy', 'tst.snapshot_neg', 'tst.snapshot_recursive', 'tst.snapshot_simple'] tags = ['functional', 'channel_program', 'synctask_core'] [tests/functional/chattr] tests = ['chattr_001_pos', 'chattr_002_neg'] tags = ['functional', 'chattr'] [tests/functional/checksum] tests = ['run_edonr_test', 'run_sha2_test', 'run_skein_test', 'filetest_001_pos'] tags = ['functional', 'checksum'] [tests/functional/clean_mirror] tests = [ 'clean_mirror_001_pos', 'clean_mirror_002_pos', 'clean_mirror_003_pos', 'clean_mirror_004_pos'] tags = ['functional', 'clean_mirror'] [tests/functional/cli_root/zdb] tests = ['zdb_001_neg', 'zdb_002_pos', 'zdb_003_pos', 'zdb_004_pos', 'zdb_005_pos', 'zdb_006_pos'] pre = post = tags = ['functional', 'cli_root', 'zdb'] [tests/functional/cli_root/zfs] tests = ['zfs_001_neg', 'zfs_002_pos', 'zfs_003_neg'] tags = ['functional', 'cli_root', 'zfs'] [tests/functional/cli_root/zfs_bookmark] tests = ['zfs_bookmark_cliargs'] tags = ['functional', 'cli_root', 'zfs_bookmark'] [tests/functional/cli_root/zfs_change-key] tests = ['zfs_change-key', 'zfs_change-key_child', 'zfs_change-key_format', 'zfs_change-key_inherit', 'zfs_change-key_load', 'zfs_change-key_location', 'zfs_change-key_pbkdf2iters'] tags = ['functional', 'cli_root', 'zfs_change-key'] [tests/functional/cli_root/zfs_clone] tests = ['zfs_clone_001_neg', 'zfs_clone_002_pos', 'zfs_clone_003_pos', 'zfs_clone_004_pos', 'zfs_clone_005_pos', 'zfs_clone_006_pos', 'zfs_clone_007_pos', 'zfs_clone_008_neg', 'zfs_clone_009_neg', 'zfs_clone_010_pos', 'zfs_clone_encrypted', 'zfs_clone_deeply_nested'] tags = ['functional', 'cli_root', 'zfs_clone'] [tests/functional/cli_root/zfs_copies] tests = ['zfs_copies_001_pos', 'zfs_copies_002_pos', 'zfs_copies_003_pos', 'zfs_copies_004_neg', 'zfs_copies_005_neg', 'zfs_copies_006_pos'] tags = ['functional', 'cli_root', 'zfs_copies'] [tests/functional/cli_root/zfs_create] tests = ['zfs_create_001_pos', 'zfs_create_002_pos', 'zfs_create_003_pos', 'zfs_create_004_pos', 'zfs_create_005_pos', 'zfs_create_006_pos', 'zfs_create_007_pos', 'zfs_create_008_neg', 'zfs_create_009_neg', 'zfs_create_010_neg', 'zfs_create_011_pos', 'zfs_create_012_pos', 'zfs_create_013_pos', 'zfs_create_014_pos', 'zfs_create_encrypted', 'zfs_create_crypt_combos'] tags = ['functional', 'cli_root', 'zfs_create'] [tests/functional/cli_root/zfs_destroy] tests = ['zfs_destroy_001_pos', 'zfs_destroy_002_pos', 'zfs_destroy_003_pos', 'zfs_destroy_004_pos', 'zfs_destroy_005_neg', 'zfs_destroy_006_neg', 'zfs_destroy_007_neg', 'zfs_destroy_008_pos', 'zfs_destroy_009_pos', 'zfs_destroy_010_pos', 'zfs_destroy_011_pos', 'zfs_destroy_012_pos', 'zfs_destroy_013_neg', 'zfs_destroy_014_pos', 'zfs_destroy_015_pos', 'zfs_destroy_016_pos'] tags = ['functional', 'cli_root', 'zfs_destroy'] [tests/functional/cli_root/zfs_diff] tests = ['zfs_diff_changes', 'zfs_diff_cliargs', 'zfs_diff_timestamp', 'zfs_diff_types', 'zfs_diff_encrypted'] tags = ['functional', 'cli_root', 'zfs_diff'] [tests/functional/cli_root/zfs_get] tests = ['zfs_get_001_pos', 'zfs_get_002_pos', 'zfs_get_003_pos', 'zfs_get_004_pos', 'zfs_get_005_neg', 'zfs_get_006_neg', 'zfs_get_007_neg', 'zfs_get_008_pos', 'zfs_get_009_pos', 'zfs_get_010_neg'] tags = ['functional', 'cli_root', 'zfs_get'] [tests/functional/cli_root/zfs_inherit] tests = ['zfs_inherit_001_neg', 'zfs_inherit_002_neg', 'zfs_inherit_003_pos'] tags = ['functional', 'cli_root', 'zfs_inherit'] [tests/functional/cli_root/zfs_load-key] tests = ['zfs_load-key', 'zfs_load-key_all', 'zfs_load-key_file', 'zfs_load-key_location', 'zfs_load-key_noop', 'zfs_load-key_recursive'] tags = ['functional', 'cli_root', 'zfs_load-key'] [tests/functional/cli_root/zfs_mount] tests = ['zfs_mount_001_pos', 'zfs_mount_002_pos', 'zfs_mount_003_pos', 'zfs_mount_004_pos', 'zfs_mount_005_pos', 'zfs_mount_006_pos', 'zfs_mount_007_pos', 'zfs_mount_008_pos', 'zfs_mount_009_neg', 'zfs_mount_010_neg', 'zfs_mount_011_neg', 'zfs_mount_012_neg', 'zfs_mount_all_001_pos', 'zfs_mount_encrypted', 'zfs_mount_remount', 'zfs_multi_mount'] tags = ['functional', 'cli_root', 'zfs_mount'] [tests/functional/cli_root/zfs_program] tests = ['zfs_program_json'] tags = ['functional', 'cli_root', 'zfs_program'] [tests/functional/cli_root/zfs_promote] tests = ['zfs_promote_001_pos', 'zfs_promote_002_pos', 'zfs_promote_003_pos', 'zfs_promote_004_pos', 'zfs_promote_005_pos', 'zfs_promote_006_neg', 'zfs_promote_007_neg', 'zfs_promote_008_pos', 'zfs_promote_encryptionroot'] tags = ['functional', 'cli_root', 'zfs_promote'] [tests/functional/cli_root/zfs_property] tests = ['zfs_written_property_001_pos'] tags = ['functional', 'cli_root', 'zfs_property'] [tests/functional/cli_root/zfs_receive] tests = ['zfs_receive_001_pos', 'zfs_receive_002_pos', 'zfs_receive_003_pos', 'zfs_receive_004_neg', 'zfs_receive_005_neg', 'zfs_receive_006_pos', 'zfs_receive_007_neg', 'zfs_receive_008_pos', 'zfs_receive_009_neg', 'zfs_receive_010_pos', 'zfs_receive_011_pos', 'zfs_receive_012_pos', 'zfs_receive_013_pos', 'zfs_receive_014_pos', 'zfs_receive_015_pos', 'receive-o-x_props_override', 'zfs_receive_from_encrypted', 'zfs_receive_to_encrypted', 'zfs_receive_raw', 'zfs_receive_raw_incremental'] tags = ['functional', 'cli_root', 'zfs_receive'] [tests/functional/cli_root/zfs_remap] tests = ['zfs_remap_cliargs', 'zfs_remap_obsolete_counts'] tags = ['functional', 'cli_root', 'zfs_remap'] [tests/functional/cli_root/zfs_rename] tests = ['zfs_rename_001_pos', 'zfs_rename_002_pos', 'zfs_rename_003_pos', 'zfs_rename_004_neg', 'zfs_rename_005_neg', 'zfs_rename_006_pos', 'zfs_rename_007_pos', 'zfs_rename_008_pos', 'zfs_rename_009_neg', 'zfs_rename_010_neg', 'zfs_rename_011_pos', 'zfs_rename_012_neg', 'zfs_rename_013_pos', 'zfs_rename_014_neg', 'zfs_rename_encrypted_child', 'zfs_rename_to_encrypted'] tags = ['functional', 'cli_root', 'zfs_rename'] [tests/functional/cli_root/zfs_reservation] tests = ['zfs_reservation_001_pos', 'zfs_reservation_002_pos'] tags = ['functional', 'cli_root', 'zfs_reservation'] [tests/functional/cli_root/zfs_rollback] tests = ['zfs_rollback_001_pos', 'zfs_rollback_002_pos', 'zfs_rollback_003_neg', 'zfs_rollback_004_neg'] tags = ['functional', 'cli_root', 'zfs_rollback'] [tests/functional/cli_root/zfs_send] tests = ['zfs_send_001_pos', 'zfs_send_002_pos', 'zfs_send_003_pos', 'zfs_send_004_neg', 'zfs_send_005_pos', 'zfs_send_006_pos', 'zfs_send_007_pos', 'zfs_send_encrypted', 'zfs_send_raw', 'zfs_send_sparse', 'zfs_send-b'] tags = ['functional', 'cli_root', 'zfs_send'] [tests/functional/cli_root/zfs_set] tests = ['cache_001_pos', 'cache_002_neg', 'canmount_001_pos', 'canmount_002_pos', 'canmount_003_pos', 'canmount_004_pos', 'checksum_001_pos', 'compression_001_pos', 'mountpoint_001_pos', 'mountpoint_002_pos', 'reservation_001_neg', 'user_property_002_pos', 'share_mount_001_neg', 'snapdir_001_pos', 'onoffs_001_pos', 'user_property_001_pos', 'user_property_003_neg', 'readonly_001_pos', 'user_property_004_pos', 'version_001_neg', 'zfs_set_001_neg', 'zfs_set_002_neg', 'zfs_set_003_neg', 'property_alias_001_pos', 'mountpoint_003_pos', 'ro_props_001_pos', 'zfs_set_keylocation'] tags = ['functional', 'cli_root', 'zfs_set'] [tests/functional/cli_root/zfs_share] tests = ['zfs_share_001_pos', 'zfs_share_002_pos', 'zfs_share_003_pos', 'zfs_share_004_pos', 'zfs_share_005_pos', 'zfs_share_006_pos', 'zfs_share_007_neg', 'zfs_share_008_neg', 'zfs_share_009_neg', 'zfs_share_010_neg', 'zfs_share_011_pos'] tags = ['functional', 'cli_root', 'zfs_share'] [tests/functional/cli_root/zfs_snapshot] tests = ['zfs_snapshot_001_neg', 'zfs_snapshot_002_neg', 'zfs_snapshot_003_neg', 'zfs_snapshot_004_neg', 'zfs_snapshot_005_neg', 'zfs_snapshot_006_pos', 'zfs_snapshot_007_neg', 'zfs_snapshot_008_neg', 'zfs_snapshot_009_pos'] tags = ['functional', 'cli_root', 'zfs_snapshot'] [tests/functional/cli_root/zfs_unload-key] tests = ['zfs_unload-key', 'zfs_unload-key_all', 'zfs_unload-key_recursive'] tags = ['functional', 'cli_root', 'zfs_unload-key'] [tests/functional/cli_root/zfs_unmount] tests = ['zfs_unmount_001_pos', 'zfs_unmount_002_pos', 'zfs_unmount_003_pos', 'zfs_unmount_004_pos', 'zfs_unmount_005_pos', 'zfs_unmount_006_pos', 'zfs_unmount_007_neg', 'zfs_unmount_008_neg', 'zfs_unmount_009_pos', 'zfs_unmount_all_001_pos'] tags = ['functional', 'cli_root', 'zfs_unmount'] [tests/functional/cli_root/zfs_unshare] tests = ['zfs_unshare_001_pos', 'zfs_unshare_002_pos', 'zfs_unshare_003_pos', 'zfs_unshare_004_neg', 'zfs_unshare_005_neg', 'zfs_unshare_006_pos'] tags = ['functional', 'cli_root', 'zfs_unshare'] [tests/functional/cli_root/zfs_upgrade] tests = ['zfs_upgrade_001_pos', 'zfs_upgrade_002_pos', 'zfs_upgrade_003_pos', 'zfs_upgrade_004_pos', 'zfs_upgrade_005_pos', 'zfs_upgrade_006_neg', 'zfs_upgrade_007_neg'] tags = ['functional', 'cli_root', 'zfs_upgrade'] [tests/functional/cli_root/zpool] tests = ['zpool_001_neg', 'zpool_002_pos', 'zpool_003_pos'] tags = ['functional', 'cli_root', 'zpool'] [tests/functional/cli_root/zpool_add] tests = ['zpool_add_001_pos', 'zpool_add_002_pos', 'zpool_add_003_pos', 'zpool_add_004_pos', 'zpool_add_005_pos', 'zpool_add_006_pos', 'zpool_add_007_neg', 'zpool_add_008_neg', 'zpool_add_009_neg', 'zpool_add_010_pos', 'add-o_ashift', 'add_prop_ashift', 'add_nested_replacing_spare'] tags = ['functional', 'cli_root', 'zpool_add'] [tests/functional/cli_root/zpool_attach] tests = ['zpool_attach_001_neg', 'attach-o_ashift'] tags = ['functional', 'cli_root', 'zpool_attach'] [tests/functional/cli_root/zpool_clear] tests = ['zpool_clear_001_pos', 'zpool_clear_002_neg', 'zpool_clear_003_neg', 'zpool_clear_readonly'] tags = ['functional', 'cli_root', 'zpool_clear'] [tests/functional/cli_root/zpool_create] tests = ['zpool_create_001_pos', 'zpool_create_002_pos', 'zpool_create_003_pos', 'zpool_create_004_pos', 'zpool_create_005_pos', 'zpool_create_006_pos', 'zpool_create_007_neg', 'zpool_create_008_pos', 'zpool_create_009_neg', 'zpool_create_010_neg', 'zpool_create_011_neg', 'zpool_create_012_neg', 'zpool_create_014_neg', 'zpool_create_015_neg', 'zpool_create_016_pos', 'zpool_create_017_neg', 'zpool_create_018_pos', 'zpool_create_019_pos', 'zpool_create_020_pos', 'zpool_create_021_pos', 'zpool_create_022_pos', 'zpool_create_023_neg', 'zpool_create_024_pos', 'zpool_create_encrypted', 'zpool_create_crypt_combos', 'zpool_create_features_001_pos', 'zpool_create_features_002_pos', 'zpool_create_features_003_pos', 'zpool_create_features_004_neg', 'zpool_create_features_005_pos', 'create-o_ashift', 'zpool_create_tempname'] tags = ['functional', 'cli_root', 'zpool_create'] [tests/functional/cli_root/zpool_destroy] tests = ['zpool_destroy_001_pos', 'zpool_destroy_002_pos', 'zpool_destroy_003_neg'] pre = post = tags = ['functional', 'cli_root', 'zpool_destroy'] [tests/functional/cli_root/zpool_detach] tests = ['zpool_detach_001_neg'] tags = ['functional', 'cli_root', 'zpool_detach'] [tests/functional/cli_root/zpool_events] tests = ['zpool_events_clear', 'zpool_events_cliargs', 'zpool_events_follow', 'zpool_events_poolname'] tags = ['functional', 'cli_root', 'zpool_events'] [tests/functional/cli_root/zpool_expand] tests = ['zpool_expand_001_pos', 'zpool_expand_002_pos', 'zpool_expand_003_neg', 'zpool_expand_004_pos', 'zpool_expand_005_pos'] tags = ['functional', 'cli_root', 'zpool_expand'] [tests/functional/cli_root/zpool_export] tests = ['zpool_export_001_pos', 'zpool_export_002_pos', 'zpool_export_003_neg', 'zpool_export_004_pos'] tags = ['functional', 'cli_root', 'zpool_export'] [tests/functional/cli_root/zpool_get] tests = ['zpool_get_001_pos', 'zpool_get_002_pos', 'zpool_get_003_pos', 'zpool_get_004_neg'] tags = ['functional', 'cli_root', 'zpool_get'] [tests/functional/cli_root/zpool_history] tests = ['zpool_history_001_neg', 'zpool_history_002_pos'] tags = ['functional', 'cli_root', 'zpool_history'] [tests/functional/cli_root/zpool_import] tests = ['zpool_import_001_pos', 'zpool_import_002_pos', 'zpool_import_003_pos', 'zpool_import_004_pos', 'zpool_import_005_pos', 'zpool_import_006_pos', 'zpool_import_007_pos', 'zpool_import_008_pos', 'zpool_import_009_neg', 'zpool_import_010_pos', 'zpool_import_011_neg', 'zpool_import_012_pos', 'zpool_import_013_neg', 'zpool_import_014_pos', 'zpool_import_015_pos', 'zpool_import_features_001_pos', 'zpool_import_features_002_neg', 'zpool_import_features_003_pos', 'zpool_import_missing_001_pos', 'zpool_import_missing_002_pos', 'zpool_import_missing_003_pos', 'zpool_import_rename_001_pos', 'zpool_import_all_001_pos', 'zpool_import_encrypted', 'zpool_import_encrypted_load', 'zpool_import_errata3', 'import_cachefile_device_added', 'import_cachefile_device_removed', 'import_cachefile_device_replaced', 'import_cachefile_mirror_attached', 'import_cachefile_mirror_detached', 'import_cachefile_shared_device', 'import_devices_missing', 'import_paths_changed', 'import_rewind_config_changed', 'import_rewind_device_replaced'] tags = ['functional', 'cli_root', 'zpool_import'] [tests/functional/cli_root/zpool_labelclear] tests = ['zpool_labelclear_active', 'zpool_labelclear_exported'] pre = post = tags = ['functional', 'cli_root', 'zpool_labelclear'] [tests/functional/cli_root/zpool_offline] tests = ['zpool_offline_001_pos', 'zpool_offline_002_neg', 'zpool_offline_003_pos'] tags = ['functional', 'cli_root', 'zpool_offline'] [tests/functional/cli_root/zpool_online] tests = ['zpool_online_001_pos', 'zpool_online_002_neg'] tags = ['functional', 'cli_root', 'zpool_online'] [tests/functional/cli_root/zpool_remove] tests = ['zpool_remove_001_neg', 'zpool_remove_002_pos', 'zpool_remove_003_pos'] tags = ['functional', 'cli_root', 'zpool_remove'] [tests/functional/cli_root/zpool_reopen] tests = ['zpool_reopen_001_pos', 'zpool_reopen_002_pos', 'zpool_reopen_003_pos', 'zpool_reopen_004_pos', 'zpool_reopen_005_pos', 'zpool_reopen_006_neg', 'zpool_reopen_007_pos'] tags = ['functional', 'cli_root', 'zpool_reopen'] [tests/functional/cli_root/zpool_replace] tests = ['zpool_replace_001_neg', 'replace-o_ashift', 'replace_prop_ashift'] tags = ['functional', 'cli_root', 'zpool_replace'] [tests/functional/cli_root/zpool_scrub] tests = ['zpool_scrub_001_neg', 'zpool_scrub_002_pos', 'zpool_scrub_003_pos', 'zpool_scrub_004_pos', 'zpool_scrub_005_pos', 'zpool_scrub_encrypted_unloaded', 'zpool_scrub_offline_device'] tags = ['functional', 'cli_root', 'zpool_scrub'] [tests/functional/cli_root/zpool_set] tests = ['zpool_set_001_pos', 'zpool_set_002_neg', 'zpool_set_003_neg', 'zpool_set_ashift', 'zpool_set_features'] tags = ['functional', 'cli_root', 'zpool_set'] [tests/functional/cli_root/zpool_split] tests = ['zpool_split_cliargs', 'zpool_split_devices', 'zpool_split_encryption', 'zpool_split_props', 'zpool_split_vdevs'] tags = ['functional', 'cli_root', 'zpool_split'] [tests/functional/cli_root/zpool_status] tests = ['zpool_status_001_pos', 'zpool_status_002_pos','zpool_status_003_pos', 'zpool_status_-c_disable', 'zpool_status_-c_homedir', 'zpool_status_-c_searchpath'] user = tags = ['functional', 'cli_root', 'zpool_status'] [tests/functional/cli_root/zpool_sync] tests = ['zpool_sync_001_pos', 'zpool_sync_002_neg'] tags = ['functional', 'cli_root', 'zpool_sync'] [tests/functional/cli_root/zpool_upgrade] tests = ['zpool_upgrade_001_pos', 'zpool_upgrade_002_pos', 'zpool_upgrade_003_pos', 'zpool_upgrade_004_pos', 'zpool_upgrade_005_neg', 'zpool_upgrade_006_neg', 'zpool_upgrade_007_pos', 'zpool_upgrade_008_pos', 'zpool_upgrade_009_neg'] tags = ['functional', 'cli_root', 'zpool_upgrade'] [tests/functional/cli_user/misc] tests = ['zdb_001_neg', 'zfs_001_neg', 'zfs_allow_001_neg', 'zfs_clone_001_neg', 'zfs_create_001_neg', 'zfs_destroy_001_neg', 'zfs_get_001_neg', 'zfs_inherit_001_neg', 'zfs_mount_001_neg', 'zfs_promote_001_neg', 'zfs_receive_001_neg', 'zfs_rename_001_neg', 'zfs_rollback_001_neg', 'zfs_send_001_neg', 'zfs_set_001_neg', 'zfs_share_001_neg', 'zfs_snapshot_001_neg', 'zfs_unallow_001_neg', 'zfs_unmount_001_neg', 'zfs_unshare_001_neg', 'zfs_upgrade_001_neg', 'zpool_001_neg', 'zpool_add_001_neg', 'zpool_attach_001_neg', 'zpool_clear_001_neg', 'zpool_create_001_neg', 'zpool_destroy_001_neg', 'zpool_detach_001_neg', 'zpool_export_001_neg', 'zpool_get_001_neg', 'zpool_history_001_neg', 'zpool_import_001_neg', 'zpool_import_002_neg', 'zpool_offline_001_neg', 'zpool_online_001_neg', 'zpool_remove_001_neg', 'zpool_replace_001_neg', 'zpool_scrub_001_neg', 'zpool_set_001_neg', 'zpool_status_001_neg', 'zpool_upgrade_001_neg', 'arcstat_001_pos', 'arc_summary_001_pos', 'arc_summary_002_neg', 'arc_summary3_001_pos', 'dbufstat_001_pos'] user = tags = ['functional', 'cli_user', 'misc'] [tests/functional/cli_user/zfs_list] tests = ['zfs_list_001_pos', 'zfs_list_002_pos', 'zfs_list_003_pos', 'zfs_list_004_neg', 'zfs_list_007_pos', 'zfs_list_008_neg'] user = tags = ['functional', 'cli_user', 'zfs_list'] [tests/functional/cli_user/zpool_iostat] tests = ['zpool_iostat_001_neg', 'zpool_iostat_002_pos', 'zpool_iostat_003_neg', 'zpool_iostat_004_pos', 'zpool_iostat_005_pos', 'zpool_iostat_-c_disable', 'zpool_iostat_-c_homedir', 'zpool_iostat_-c_searchpath'] user = tags = ['functional', 'cli_user', 'zpool_iostat'] [tests/functional/cli_user/zpool_list] tests = ['zpool_list_001_pos', 'zpool_list_002_neg'] user = tags = ['functional', 'cli_user', 'zpool_list'] [tests/functional/compression] tests = ['compress_001_pos', 'compress_002_pos', 'compress_003_pos', 'compress_004_pos'] tags = ['functional', 'compression'] [tests/functional/cp_files] tests = ['cp_files_001_pos'] tags = ['functional', 'cp_files'] [tests/functional/ctime] tests = ['ctime_001_pos' ] tags = ['functional', 'ctime'] [tests/functional/deadman] tests = ['deadman_sync', 'deadman_zio'] pre = post = tags = ['functional', 'deadman'] [tests/functional/delegate] tests = ['zfs_allow_001_pos', 'zfs_allow_002_pos', 'zfs_allow_004_pos', 'zfs_allow_005_pos', 'zfs_allow_006_pos', 'zfs_allow_007_pos', 'zfs_allow_008_pos', 'zfs_allow_009_neg', 'zfs_allow_010_pos', 'zfs_allow_011_neg', 'zfs_allow_012_neg', 'zfs_unallow_001_pos', 'zfs_unallow_002_pos', 'zfs_unallow_003_pos', 'zfs_unallow_004_pos', 'zfs_unallow_005_pos', 'zfs_unallow_006_pos', 'zfs_unallow_007_neg', 'zfs_unallow_008_neg'] tags = ['functional', 'delegate'] [tests/functional/devices] tests = ['devices_001_pos', 'devices_002_neg', 'devices_003_pos'] tags = ['functional', 'devices'] [tests/functional/events] tests = ['events_001_pos', 'events_002_pos', 'zed_rc_filter'] tags = ['functional', 'events'] [tests/functional/exec] tests = ['exec_001_pos', 'exec_002_neg'] tags = ['functional', 'exec'] [tests/functional/fault] tests = ['auto_online_001_pos', 'auto_replace_001_pos', 'auto_spare_001_pos', 'auto_spare_002_pos', 'auto_spare_ashift', 'auto_spare_multiple', - 'scrub_after_resilver', 'decrypt_fault'] + 'scrub_after_resilver', 'decrypt_fault', 'decompress_fault'] tags = ['functional', 'fault'] [tests/functional/features/async_destroy] tests = ['async_destroy_001_pos'] tags = ['functional', 'features', 'async_destroy'] [tests/functional/features/large_dnode] tests = ['large_dnode_001_pos', 'large_dnode_002_pos', 'large_dnode_003_pos', 'large_dnode_004_neg', 'large_dnode_005_pos', 'large_dnode_006_pos', 'large_dnode_007_neg', 'large_dnode_008_pos', 'large_dnode_009_pos'] tags = ['functional', 'features', 'large_dnode'] [tests/functional/grow] pre = post = tests = ['grow_pool_001_pos', 'grow_replicas_001_pos'] tags = ['functional', 'grow'] [tests/functional/history] tests = ['history_001_pos', 'history_002_pos', 'history_003_pos', 'history_004_pos', 'history_005_neg', 'history_006_neg', 'history_007_pos', 'history_008_pos', 'history_009_pos', 'history_010_pos'] tags = ['functional', 'history'] [tests/functional/hkdf] tests = ['run_hkdf_test'] tags = ['functional', 'hkdf'] [tests/functional/inheritance] tests = ['inherit_001_pos'] pre = tags = ['functional', 'inheritance'] [tests/functional/io] tests = ['sync', 'psync', 'libaio', 'posixaio', 'mmap'] tags = ['functional', 'io'] [tests/functional/inuse] tests = ['inuse_001_pos', 'inuse_003_pos', 'inuse_004_pos', 'inuse_005_pos', 'inuse_006_pos', 'inuse_007_pos', 'inuse_008_pos', 'inuse_009_pos'] post = tags = ['functional', 'inuse'] [tests/functional/kstat] tests = ['state'] tags = ['functional', 'kstat'] [tests/functional/large_files] tests = ['large_files_001_pos', 'large_files_002_pos'] tags = ['functional', 'large_files'] [tests/functional/largest_pool] tests = ['largest_pool_001_pos'] pre = post = tags = ['functional', 'largest_pool'] [tests/functional/link_count] tests = ['link_count_001'] tags = ['functional', 'link_count'] [tests/functional/migration] tests = ['migration_001_pos', 'migration_002_pos', 'migration_003_pos', 'migration_004_pos', 'migration_005_pos', 'migration_006_pos', 'migration_007_pos', 'migration_008_pos', 'migration_009_pos', 'migration_010_pos', 'migration_011_pos', 'migration_012_pos'] tags = ['functional', 'migration'] [tests/functional/mmap] tests = ['mmap_write_001_pos', 'mmap_read_001_pos', 'mmap_libaio_001_pos'] tags = ['functional', 'mmap'] [tests/functional/mmp] tests = ['mmp_on_thread', 'mmp_on_uberblocks', 'mmp_on_off', 'mmp_interval', 'mmp_active_import', 'mmp_inactive_import', 'mmp_exported_import', 'mmp_write_uberblocks', 'mmp_reset_interval', 'multihost_history', 'mmp_on_zdb'] tags = ['functional', 'mmp'] [tests/functional/mount] tests = ['umount_001', 'umountall_001'] tags = ['functional', 'mount'] [tests/functional/mv_files] tests = ['mv_files_001_pos', 'mv_files_002_pos', 'random_creation'] tags = ['functional', 'mv_files'] [tests/functional/nestedfs] tests = ['nestedfs_001_pos'] tags = ['functional', 'nestedfs'] [tests/functional/no_space] tests = ['enospc_001_pos', 'enospc_002_pos', 'enospc_003_pos'] tags = ['functional', 'no_space'] [tests/functional/nopwrite] tests = ['nopwrite_copies', 'nopwrite_mtime', 'nopwrite_negative', 'nopwrite_promoted_clone', 'nopwrite_recsize', 'nopwrite_sync', 'nopwrite_varying_compression', 'nopwrite_volume'] tags = ['functional', 'nopwrite'] [tests/functional/online_offline] tests = ['online_offline_001_pos', 'online_offline_002_neg', 'online_offline_003_neg'] tags = ['functional', 'online_offline'] [tests/functional/pool_checkpoint] tests = ['checkpoint_after_rewind', 'checkpoint_big_rewind', 'checkpoint_capacity', 'checkpoint_conf_change', 'checkpoint_discard', 'checkpoint_discard_busy', 'checkpoint_discard_many', 'checkpoint_indirect', 'checkpoint_invalid', 'checkpoint_lun_expsz', 'checkpoint_open', 'checkpoint_removal', 'checkpoint_rewind', 'checkpoint_ro_rewind', 'checkpoint_sm_scale', 'checkpoint_twice', 'checkpoint_vdev_add', 'checkpoint_zdb', 'checkpoint_zhack_feat'] tags = ['functional', 'pool_checkpoint'] timeout = 1800 [tests/functional/pool_names] tests = ['pool_names_001_pos', 'pool_names_002_neg'] pre = post = tags = ['functional', 'pool_names'] [tests/functional/poolversion] tests = ['poolversion_001_pos', 'poolversion_002_pos'] tags = ['functional', 'poolversion'] [tests/functional/privilege] tests = ['privilege_001_pos', 'privilege_002_pos'] tags = ['functional', 'privilege'] [tests/functional/projectquota] tests = ['projectid_001_pos', 'projectid_002_pos', 'projectid_003_pos', 'projectquota_001_pos', 'projectquota_002_pos', 'projectquota_003_pos', 'projectquota_004_neg', 'projectquota_005_pos', 'projectquota_006_pos', 'projectquota_007_pos', 'projectquota_008_pos', 'projectquota_009_pos', 'projectspace_001_pos', 'projectspace_002_pos', 'projectspace_003_pos', 'projectspace_004_pos', 'projecttree_001_pos', 'projecttree_002_pos', 'projecttree_003_neg' ] tags = ['functional', 'projectquota'] [tests/functional/pyzfs] tests = ['pyzfs_unittest'] pre = post = tags = ['functional', 'pyzfs'] [tests/functional/quota] tests = ['quota_001_pos', 'quota_002_pos', 'quota_003_pos', 'quota_004_pos', 'quota_005_pos', 'quota_006_neg'] tags = ['functional', 'quota'] [tests/functional/raidz] tests = ['raidz_001_neg', 'raidz_002_pos'] tags = ['functional', 'raidz'] [tests/functional/redundancy] tests = ['redundancy_001_pos', 'redundancy_002_pos', 'redundancy_003_pos', 'redundancy_004_neg'] tags = ['functional', 'redundancy'] [tests/functional/refquota] tests = ['refquota_001_pos', 'refquota_002_pos', 'refquota_003_pos', 'refquota_004_pos', 'refquota_005_pos', 'refquota_006_neg'] tags = ['functional', 'refquota'] [tests/functional/refreserv] tests = ['refreserv_001_pos', 'refreserv_002_pos', 'refreserv_003_pos', 'refreserv_004_pos', 'refreserv_005_pos'] tags = ['functional', 'refreserv'] [tests/functional/removal] pre = tests = ['removal_all_vdev', 'removal_check_space', 'removal_condense_export', 'removal_multiple_indirection', 'removal_remap', 'removal_remap_deadlists', 'removal_resume_export', 'removal_sanity', 'removal_with_add', 'removal_with_create_fs', 'removal_with_dedup', 'removal_with_export', 'removal_with_ganging', 'removal_with_remap', 'removal_with_remove', 'removal_with_scrub', 'removal_with_send', 'removal_with_send_recv', 'removal_with_snapshot', 'removal_with_write', 'removal_with_zdb', 'remove_expanded', 'remove_mirror', 'remove_mirror_sanity', 'remove_raidz'] tags = ['functional', 'removal'] [tests/functional/rename_dirs] tests = ['rename_dirs_001_pos'] tags = ['functional', 'rename_dirs'] [tests/functional/replacement] tests = ['replacement_001_pos', 'replacement_002_pos', 'replacement_003_pos'] tags = ['functional', 'replacement'] [tests/functional/reservation] tests = ['reservation_001_pos', 'reservation_002_pos', 'reservation_003_pos', 'reservation_004_pos', 'reservation_005_pos', 'reservation_006_pos', 'reservation_007_pos', 'reservation_008_pos', 'reservation_009_pos', 'reservation_010_pos', 'reservation_011_pos', 'reservation_012_pos', 'reservation_013_pos', 'reservation_014_pos', 'reservation_015_pos', 'reservation_016_pos', 'reservation_017_pos', 'reservation_018_pos', 'reservation_019_pos', 'reservation_020_pos', 'reservation_021_neg', 'reservation_022_pos'] tags = ['functional', 'reservation'] [tests/functional/rootpool] tests = ['rootpool_002_neg', 'rootpool_003_neg', 'rootpool_007_pos'] tags = ['functional', 'rootpool'] [tests/functional/rsend] tests = ['rsend_001_pos', 'rsend_002_pos', 'rsend_003_pos', 'rsend_004_pos', 'rsend_005_pos', 'rsend_006_pos', 'rsend_007_pos', 'rsend_008_pos', 'rsend_009_pos', 'rsend_010_pos', 'rsend_011_pos', 'rsend_012_pos', 'rsend_013_pos', 'rsend_014_pos', 'rsend_019_pos', 'rsend_020_pos', 'rsend_021_pos', 'rsend_022_pos', 'rsend_024_pos', 'send-c_verify_ratio', 'send-c_verify_contents', 'send-c_props', 'send-c_incremental', 'send-c_volume', 'send-c_zstreamdump', 'send-c_lz4_disabled', 'send-c_recv_lz4_disabled', 'send-c_mixed_compression', 'send-c_stream_size_estimate', 'send-cD', 'send-c_embedded_blocks', 'send-c_resume', 'send-cpL_varied_recsize', 'send-c_recv_dedup', 'send_encrypted_files', 'send_encrypted_heirarchy', 'send_encrypted_props', 'send_freeobjects', 'send_realloc_dnode_size', 'send_hole_birth', 'send-wDR_encrypted_zvol'] tags = ['functional', 'rsend'] [tests/functional/scrub_mirror] tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos', 'scrub_mirror_003_pos', 'scrub_mirror_004_pos'] tags = ['functional', 'scrub_mirror'] [tests/functional/slog] tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos', 'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg', 'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg', 'slog_013_pos', 'slog_014_pos', 'slog_015_neg', 'slog_replay_fs', 'slog_replay_volume'] tags = ['functional', 'slog'] [tests/functional/snapshot] tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos', 'rollback_003_pos', 'snapshot_001_pos', 'snapshot_002_pos', 'snapshot_003_pos', 'snapshot_004_pos', 'snapshot_005_pos', 'snapshot_006_pos', 'snapshot_007_pos', 'snapshot_008_pos', 'snapshot_009_pos', 'snapshot_010_pos', 'snapshot_011_pos', 'snapshot_012_pos', 'snapshot_013_pos', 'snapshot_014_pos', 'snapshot_015_pos', 'snapshot_016_pos', 'snapshot_017_pos'] tags = ['functional', 'snapshot'] [tests/functional/snapused] tests = ['snapused_001_pos', 'snapused_002_pos', 'snapused_003_pos', 'snapused_004_pos', 'snapused_005_pos'] tags = ['functional', 'snapused'] [tests/functional/sparse] tests = ['sparse_001_pos'] tags = ['functional', 'sparse'] [tests/functional/threadsappend] tests = ['threadsappend_001_pos'] tags = ['functional', 'threadsappend'] [tests/functional/tmpfile] tests = ['tmpfile_001_pos', 'tmpfile_002_pos', 'tmpfile_003_pos'] tags = ['functional', 'tmpfile'] [tests/functional/truncate] tests = ['truncate_001_pos', 'truncate_002_pos', 'truncate_timestamps'] tags = ['functional', 'truncate'] [tests/functional/upgrade] tests = ['upgrade_userobj_001_pos', 'upgrade_projectquota_001_pos'] tags = ['functional', 'upgrade'] [tests/functional/user_namespace] tests = ['user_namespace_001'] tags = ['functional', 'user_namespace'] [tests/functional/userquota] tests = [ 'userquota_001_pos', 'userquota_002_pos', 'userquota_003_pos', 'userquota_004_pos', 'userquota_005_neg', 'userquota_006_pos', 'userquota_007_pos', 'userquota_008_pos', 'userquota_009_pos', 'userquota_010_pos', 'userquota_011_pos', 'userquota_012_neg', 'userquota_013_pos', 'userspace_001_pos', 'userspace_002_pos', 'userspace_003_pos', 'groupspace_001_pos', 'groupspace_002_pos', 'groupspace_003_pos' ] tags = ['functional', 'userquota'] [tests/functional/vdev_zaps] tests = ['vdev_zaps_001_pos', 'vdev_zaps_002_pos', 'vdev_zaps_003_pos', 'vdev_zaps_004_pos', 'vdev_zaps_005_pos', 'vdev_zaps_006_pos', 'vdev_zaps_007_pos'] tags = ['functional', 'vdev_zaps'] [tests/functional/write_dirs] tests = ['write_dirs_001_pos', 'write_dirs_002_pos'] tags = ['functional', 'write_dirs'] [tests/functional/xattr] tests = ['xattr_001_pos', 'xattr_002_neg', 'xattr_003_neg', 'xattr_004_pos', 'xattr_005_pos', 'xattr_006_pos', 'xattr_007_neg', 'xattr_008_pos', 'xattr_009_neg', 'xattr_010_neg', 'xattr_011_pos', 'xattr_012_pos', 'xattr_013_pos'] tags = ['functional', 'xattr'] [tests/functional/zvol/zvol_ENOSPC] tests = ['zvol_ENOSPC_001_pos'] tags = ['functional', 'zvol', 'zvol_ENOSPC'] [tests/functional/zvol/zvol_cli] tests = ['zvol_cli_001_pos', 'zvol_cli_002_pos', 'zvol_cli_003_neg'] tags = ['functional', 'zvol', 'zvol_cli'] [tests/functional/zvol/zvol_misc] tests = ['zvol_misc_001_neg', 'zvol_misc_002_pos', 'zvol_misc_003_neg', 'zvol_misc_004_pos', 'zvol_misc_005_neg', 'zvol_misc_006_pos', 'zvol_misc_snapdev', 'zvol_misc_volmode', 'zvol_misc_zil'] tags = ['functional', 'zvol', 'zvol_misc'] [tests/functional/zvol/zvol_swap] tests = ['zvol_swap_001_pos', 'zvol_swap_002_pos', 'zvol_swap_003_pos', 'zvol_swap_004_pos', 'zvol_swap_005_pos', 'zvol_swap_006_pos'] tags = ['functional', 'zvol', 'zvol_swap'] [tests/functional/libzfs] tests = ['many_fds'] tags = ['functional', 'libzfs'] diff --git a/tests/zfs-tests/tests/functional/fault/Makefile.am b/tests/zfs-tests/tests/functional/fault/Makefile.am index 1153ad8d660b..285e331a1a1d 100644 --- a/tests/zfs-tests/tests/functional/fault/Makefile.am +++ b/tests/zfs-tests/tests/functional/fault/Makefile.am @@ -1,15 +1,16 @@ pkgdatadir = $(datadir)/@PACKAGE@/zfs-tests/tests/functional/fault dist_pkgdata_SCRIPTS = \ setup.ksh \ cleanup.ksh \ auto_online_001_pos.ksh \ auto_replace_001_pos.ksh \ auto_spare_001_pos.ksh \ auto_spare_002_pos.ksh \ auto_spare_ashift.ksh \ auto_spare_multiple.ksh \ decrypt_fault.ksh \ + decompress_fault.ksh \ scrub_after_resilver.ksh dist_pkgdata_DATA = \ fault.cfg diff --git a/tests/zfs-tests/tests/functional/fault/decrypt_fault.ksh b/tests/zfs-tests/tests/functional/fault/decompress_fault.ksh similarity index 51% copy from tests/zfs-tests/tests/functional/fault/decrypt_fault.ksh copy to tests/zfs-tests/tests/functional/fault/decompress_fault.ksh index 10008f22eee3..ea831efdf452 100755 --- a/tests/zfs-tests/tests/functional/fault/decrypt_fault.ksh +++ b/tests/zfs-tests/tests/functional/fault/decompress_fault.ksh @@ -1,55 +1,55 @@ #!/bin/ksh -p # # This file and its contents are supplied under the terms of the # Common Development and Distribution License ("CDDL"), version 1.0. # You may only use this file in accordance with the terms of version # 1.0 of the CDDL. # # A full copy of the text of the CDDL should have accompanied this # source. A copy of the CDDL is also available via the Internet at # http://www.illumos.org/license/CDDL. # # -# Copyright (c) 2018 by Lawrence Livermore National Security, LLC. +# Copyright (c) 2018 by Datto Inc. # All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/fault/fault.cfg # # DESCRIPTION: -# Test that injected decryption errors are handled correctly. +# Test that injected decompression errors are handled correctly. # # STRATEGY: -# 1. Create an encrypted dataset with an test file -# 2. Inject decryption errors on the file 20% of the time +# 1. Create an compressed dataset with a test file +# 2. Inject decompression errors on the file 20% of the time # 3. Read the file to confirm that errors are handled correctly -# 4. Confirm that the decryption injection was added to the ZED logs +# 4. Confirm that the decompression injection was added to the ZED logs # -log_assert "Testing that injected decryption errors are handled correctly" +log_assert "Testing that injected decompression errors are handled correctly" function cleanup { + log_must set_tunable64 zfs_compressed_arc_enabled 1 log_must zinject -c all default_cleanup_noexit } log_onexit cleanup default_mirror_setup_noexit $DISK1 $DISK2 -log_must eval "echo 'password' | zfs create -o encryption=on \ - -o keyformat=passphrase -o keylocation=prompt $TESTPOOL/fs" +log_must set_tunable64 zfs_compressed_arc_enabled 0 +log_must zfs create -o compression=on $TESTPOOL/fs mntpt=$(get_prop mountpoint $TESTPOOL/fs) -log_must mkfile 32M $mntpt/file1 - -log_must zinject -a -t data -e decrypt -f 20 $mntpt/file1 +write_compressible $mntpt 32m 1 0 "testfile" +log_must sync log_must zfs umount $TESTPOOL/fs log_must zfs mount $TESTPOOL/fs +log_must zinject -a -t data -e decompress -f 20 $mntpt/testfile.0 +log_mustnot eval "cat $mntpt/testfile.0 > /dev/null" +log_must eval "zpool events $TESTPOOL | grep -q 'data'" -log_mustnot eval "cat $mntpt/file1 > /dev/null" -log_must eval "zpool events $TESTPOOL | grep -q 'authentication'" - -log_pass "Injected decryption errors are handled correctly" +log_pass "Injected decompression errors are handled correctly" diff --git a/tests/zfs-tests/tests/functional/fault/decrypt_fault.ksh b/tests/zfs-tests/tests/functional/fault/decrypt_fault.ksh index 10008f22eee3..ca698f778367 100755 --- a/tests/zfs-tests/tests/functional/fault/decrypt_fault.ksh +++ b/tests/zfs-tests/tests/functional/fault/decrypt_fault.ksh @@ -1,55 +1,55 @@ #!/bin/ksh -p # # This file and its contents are supplied under the terms of the # Common Development and Distribution License ("CDDL"), version 1.0. # You may only use this file in accordance with the terms of version # 1.0 of the CDDL. # # A full copy of the text of the CDDL should have accompanied this # source. A copy of the CDDL is also available via the Internet at # http://www.illumos.org/license/CDDL. # # -# Copyright (c) 2018 by Lawrence Livermore National Security, LLC. +# Copyright (c) 2018 by Datto Inc. # All rights reserved. # . $STF_SUITE/include/libtest.shlib . $STF_SUITE/tests/functional/fault/fault.cfg # # DESCRIPTION: # Test that injected decryption errors are handled correctly. # # STRATEGY: -# 1. Create an encrypted dataset with an test file +# 1. Create an encrypted dataset with a test file # 2. Inject decryption errors on the file 20% of the time # 3. Read the file to confirm that errors are handled correctly # 4. Confirm that the decryption injection was added to the ZED logs # log_assert "Testing that injected decryption errors are handled correctly" function cleanup { log_must zinject -c all default_cleanup_noexit } log_onexit cleanup default_mirror_setup_noexit $DISK1 $DISK2 log_must eval "echo 'password' | zfs create -o encryption=on \ -o keyformat=passphrase -o keylocation=prompt $TESTPOOL/fs" mntpt=$(get_prop mountpoint $TESTPOOL/fs) log_must mkfile 32M $mntpt/file1 log_must zinject -a -t data -e decrypt -f 20 $mntpt/file1 log_must zfs umount $TESTPOOL/fs log_must zfs mount $TESTPOOL/fs log_mustnot eval "cat $mntpt/file1 > /dev/null" log_must eval "zpool events $TESTPOOL | grep -q 'authentication'" log_pass "Injected decryption errors are handled correctly"