Index: head/stand/i386/zfsboot/zfsboot.c =================================================================== --- head/stand/i386/zfsboot/zfsboot.c (revision 357496) +++ head/stand/i386/zfsboot/zfsboot.c (revision 357497) @@ -1,1179 +1,1190 @@ /*- * Copyright (c) 1998 Robert Nordier * All rights reserved. * * Redistribution and use in source and binary forms are freely * permitted provided that the above copyright notice and this * paragraph and the following disclaimer are duplicated in all * such forms. * * This software is provided "AS IS" and without any express or * implied warranties, including, without limitation, the implied * warranties of merchantability and fitness for a particular * purpose. */ #include __FBSDID("$FreeBSD$"); #include "stand.h" #include #include #include #ifdef GPT #include #endif #include #include #include #include #include #include #include #include #include #include "lib.h" #include "rbx.h" #include "drv.h" #include "edd.h" #include "cons.h" #include "bootargs.h" #include "paths.h" #include "libzfs.h" #define ARGS 0x900 #define NOPT 14 #define NDEV 3 #define BIOS_NUMDRIVES 0x475 #define DRV_HARD 0x80 #define DRV_MASK 0x7f #define TYPE_AD 0 #define TYPE_DA 1 #define TYPE_MAXHARD TYPE_DA #define TYPE_FD 2 #define DEV_GELIBOOT_BSIZE 4096 extern uint32_t _end; #ifdef GPT static const uuid_t freebsd_zfs_uuid = GPT_ENT_TYPE_FREEBSD_ZFS; #endif static const char optstr[NOPT] = "DhaCcdgmnpqrsv"; /* Also 'P', 'S' */ static const unsigned char flags[NOPT] = { RBX_DUAL, RBX_SERIAL, RBX_ASKNAME, RBX_CDROM, RBX_CONFIG, RBX_KDB, RBX_GDB, RBX_MUTE, RBX_NOINTR, RBX_PAUSE, RBX_QUIET, RBX_DFLTROOT, RBX_SINGLE, RBX_VERBOSE }; uint32_t opts; /* * Paths to try loading before falling back to the boot2 prompt. * * /boot/zfsloader must be tried before /boot/loader in order to remain * backward compatible with ZFS boot environments where /boot/loader exists * but does not have ZFS support, which was the case before FreeBSD 12. * * If no loader is found, try to load a kernel directly instead. */ static const struct string { const char *p; size_t len; } loadpath[] = { { PATH_LOADER_ZFS, sizeof(PATH_LOADER_ZFS) }, { PATH_LOADER, sizeof(PATH_LOADER) }, { PATH_KERNEL, sizeof(PATH_KERNEL) }, }; static const unsigned char dev_maj[NDEV] = {30, 4, 2}; static char cmd[512]; static char cmddup[512]; static char kname[1024]; static char rootname[256]; static int comspeed = SIOSPD; static struct bootinfo bootinfo; static uint32_t bootdev; static struct zfs_boot_args zfsargs; vm_offset_t high_heap_base; uint32_t bios_basemem, bios_extmem, high_heap_size; static struct bios_smap smap; /* * The minimum amount of memory to reserve in bios_extmem for the heap. */ #define HEAP_MIN (64 * 1024 * 1024) static char *heap_next; static char *heap_end; /* Buffers that must not span a 64k boundary. */ #define READ_BUF_SIZE 8192 struct dmadat { char rdbuf[READ_BUF_SIZE]; /* for reading large things */ char secbuf[READ_BUF_SIZE]; /* for MBR/disklabel */ }; static struct dmadat *dmadat; void exit(int); void reboot(void); static void load(void); static int parse_cmd(void); static void bios_getmem(void); int main(void); #ifdef LOADER_GELI_SUPPORT #include "geliboot.h" static char gelipw[GELI_PW_MAXLEN]; #endif struct zfsdsk { struct dsk dsk; #ifdef LOADER_GELI_SUPPORT struct geli_dev *gdev; #endif }; #include "zfsimpl.c" /* * Read from a dnode (which must be from a ZPL filesystem). */ static int zfs_read(spa_t *spa, const dnode_phys_t *dnode, off_t *offp, void *start, size_t size) { const znode_phys_t *zp = (const znode_phys_t *) dnode->dn_bonus; size_t n; int rc; n = size; if (*offp + n > zp->zp_size) n = zp->zp_size - *offp; rc = dnode_read(spa, dnode, *offp, start, n); if (rc) return (-1); *offp += n; return (n); } /* * Current ZFS pool */ static spa_t *spa; static spa_t *primary_spa; static vdev_t *primary_vdev; /* * A wrapper for dskread that doesn't have to worry about whether the * buffer pointer crosses a 64k boundary. */ static int vdev_read(void *xvdev, void *priv, off_t off, void *buf, size_t bytes) { char *p; daddr_t lba, alignlba; off_t diff; unsigned int nb, alignnb; struct zfsdsk *zdsk = (struct zfsdsk *) priv; if ((off & (DEV_BSIZE - 1)) || (bytes & (DEV_BSIZE - 1))) return -1; p = buf; lba = off / DEV_BSIZE; lba += zdsk->dsk.start; /* * Align reads to 4k else 4k sector GELIs will not decrypt. * Round LBA down to nearest multiple of DEV_GELIBOOT_BSIZE bytes. */ alignlba = rounddown2(off, DEV_GELIBOOT_BSIZE) / DEV_BSIZE; /* * The read must be aligned to DEV_GELIBOOT_BSIZE bytes relative to the * start of the GELI partition, not the start of the actual disk. */ alignlba += zdsk->dsk.start; diff = (lba - alignlba) * DEV_BSIZE; while (bytes > 0) { nb = bytes / DEV_BSIZE; /* * Ensure that the read size plus the leading offset does not * exceed the size of the read buffer. */ if (nb > (READ_BUF_SIZE - diff) / DEV_BSIZE) nb = (READ_BUF_SIZE - diff) / DEV_BSIZE; /* * Round the number of blocks to read up to the nearest multiple * of DEV_GELIBOOT_BSIZE. */ alignnb = roundup2(nb * DEV_BSIZE + diff, DEV_GELIBOOT_BSIZE) / DEV_BSIZE; if (zdsk->dsk.size > 0 && alignlba + alignnb > zdsk->dsk.size + zdsk->dsk.start) { printf("Shortening read at %lld from %d to %lld\n", alignlba, alignnb, (zdsk->dsk.size + zdsk->dsk.start) - alignlba); alignnb = (zdsk->dsk.size + zdsk->dsk.start) - alignlba; } if (drvread(&zdsk->dsk, dmadat->rdbuf, alignlba, alignnb)) return -1; #ifdef LOADER_GELI_SUPPORT /* decrypt */ if (zdsk->gdev != NULL) { if (geli_read(zdsk->gdev, ((alignlba - zdsk->dsk.start) * DEV_BSIZE), dmadat->rdbuf, alignnb * DEV_BSIZE)) return (-1); } #endif memcpy(p, dmadat->rdbuf + diff, nb * DEV_BSIZE); p += nb * DEV_BSIZE; lba += nb; alignlba += alignnb; bytes -= nb * DEV_BSIZE; /* Don't need the leading offset after the first block. */ diff = 0; } return 0; } /* Match the signature exactly due to signature madness */ static int vdev_read2(vdev_t *vdev, void *priv, off_t off, void *buf, size_t bytes) { return vdev_read(vdev, priv, off, buf, bytes); } static int vdev_write(vdev_t *vdev, void *priv, off_t off, void *buf, size_t bytes) { char *p; daddr_t lba; unsigned int nb; struct zfsdsk *zdsk = (struct zfsdsk *) priv; if ((off & (DEV_BSIZE - 1)) || (bytes & (DEV_BSIZE - 1))) return -1; p = buf; lba = off / DEV_BSIZE; lba += zdsk->dsk.start; while (bytes > 0) { nb = bytes / DEV_BSIZE; if (nb > READ_BUF_SIZE / DEV_BSIZE) nb = READ_BUF_SIZE / DEV_BSIZE; memcpy(dmadat->rdbuf, p, nb * DEV_BSIZE); if (drvwrite(&zdsk->dsk, dmadat->rdbuf, lba, nb)) return -1; p += nb * DEV_BSIZE; lba += nb; bytes -= nb * DEV_BSIZE; } return 0; } static int xfsread(const dnode_phys_t *dnode, off_t *offp, void *buf, size_t nbyte) { if ((size_t)zfs_read(spa, dnode, offp, buf, nbyte) != nbyte) { printf("Invalid format\n"); return -1; } return 0; } /* * Read Pad2 (formerly "Boot Block Header") area of the first * vdev label of the given vdev. */ static int vdev_read_pad2(vdev_t *vdev, char *buf, size_t size) { blkptr_t bp; - char *tmp = zap_scratch; + char *tmp; off_t off = offsetof(vdev_label_t, vl_pad2); + int rc; if (size > VDEV_PAD_SIZE) size = VDEV_PAD_SIZE; + tmp = malloc(size); + if (tmp == NULL) + return (ENOMEM); + BP_ZERO(&bp); BP_SET_LSIZE(&bp, VDEV_PAD_SIZE); BP_SET_PSIZE(&bp, VDEV_PAD_SIZE); BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL); BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF); DVA_SET_OFFSET(BP_IDENTITY(&bp), off); - if (vdev_read_phys(vdev, &bp, tmp, off, 0)) - return (EIO); - memcpy(buf, tmp, size); - return (0); + rc = vdev_read_phys(vdev, &bp, tmp, off, 0); + if (rc == 0) + memcpy(buf, tmp, size); + free(buf); + return (rc); } static int vdev_clear_pad2(vdev_t *vdev) { - char *zeroes = zap_scratch; + char *zeroes; uint64_t *end; off_t off = offsetof(vdev_label_t, vl_pad2); + int rc; + zeroes = malloc(VDEV_PAD_SIZE); + if (zeroes == NULL) + return (ENOMEM); + memset(zeroes, 0, VDEV_PAD_SIZE); end = (uint64_t *)(zeroes + VDEV_PAD_SIZE); /* ZIO_CHECKSUM_LABEL magic and pre-calcualted checksum for all zeros */ end[-5] = 0x0210da7ab10c7a11; end[-4] = 0x97f48f807f6e2a3f; end[-3] = 0xaf909f1658aacefc; end[-2] = 0xcbd1ea57ff6db48b; end[-1] = 0x6ec692db0d465fab; - if (vdev_write(vdev, vdev->v_read_priv, off, zeroes, VDEV_PAD_SIZE)) - return (EIO); - return (0); + rc = vdev_write(vdev, vdev->v_read_priv, off, zeroes, VDEV_PAD_SIZE); + free(zeroes); + return (rc); } static void bios_getmem(void) { uint64_t size; /* Parse system memory map */ v86.ebx = 0; do { v86.ctl = V86_FLAGS; v86.addr = 0x15; /* int 0x15 function 0xe820*/ v86.eax = 0xe820; v86.ecx = sizeof(struct bios_smap); v86.edx = SMAP_SIG; v86.es = VTOPSEG(&smap); v86.edi = VTOPOFF(&smap); v86int(); if (V86_CY(v86.efl) || (v86.eax != SMAP_SIG)) break; /* look for a low-memory segment that's large enough */ if ((smap.type == SMAP_TYPE_MEMORY) && (smap.base == 0) && (smap.length >= (512 * 1024))) bios_basemem = smap.length; /* look for the first segment in 'extended' memory */ if ((smap.type == SMAP_TYPE_MEMORY) && (smap.base == 0x100000)) { bios_extmem = smap.length; } /* * Look for the largest segment in 'extended' memory beyond * 1MB but below 4GB. */ if ((smap.type == SMAP_TYPE_MEMORY) && (smap.base > 0x100000) && (smap.base < 0x100000000ull)) { size = smap.length; /* * If this segment crosses the 4GB boundary, truncate it. */ if (smap.base + size > 0x100000000ull) size = 0x100000000ull - smap.base; if (size > high_heap_size) { high_heap_size = size; high_heap_base = smap.base; } } } while (v86.ebx != 0); /* Fall back to the old compatibility function for base memory */ if (bios_basemem == 0) { v86.ctl = 0; v86.addr = 0x12; /* int 0x12 */ v86int(); bios_basemem = (v86.eax & 0xffff) * 1024; } /* Fall back through several compatibility functions for extended memory */ if (bios_extmem == 0) { v86.ctl = V86_FLAGS; v86.addr = 0x15; /* int 0x15 function 0xe801*/ v86.eax = 0xe801; v86int(); if (!V86_CY(v86.efl)) { bios_extmem = ((v86.ecx & 0xffff) + ((v86.edx & 0xffff) * 64)) * 1024; } } if (bios_extmem == 0) { v86.ctl = 0; v86.addr = 0x15; /* int 0x15 function 0x88*/ v86.eax = 0x8800; v86int(); bios_extmem = (v86.eax & 0xffff) * 1024; } /* * If we have extended memory and did not find a suitable heap * region in the SMAP, use the last 3MB of 'extended' memory as a * high heap candidate. */ if (bios_extmem >= HEAP_MIN && high_heap_size < HEAP_MIN) { high_heap_size = HEAP_MIN; high_heap_base = bios_extmem + 0x100000 - HEAP_MIN; } } /* * Try to detect a device supported by the legacy int13 BIOS */ static int int13probe(int drive) { v86.ctl = V86_FLAGS; v86.addr = 0x13; v86.eax = 0x800; v86.edx = drive; v86int(); if (!V86_CY(v86.efl) && /* carry clear */ ((v86.edx & 0xff) != (drive & DRV_MASK))) { /* unit # OK */ if ((v86.ecx & 0x3f) == 0) { /* absurd sector size */ return(0); /* skip device */ } return (1); } return(0); } /* * We call this when we find a ZFS vdev - ZFS consumes the dsk * structure so we must make a new one. */ static struct zfsdsk * copy_dsk(struct zfsdsk *zdsk) { struct zfsdsk *newdsk; newdsk = malloc(sizeof(struct zfsdsk)); *newdsk = *zdsk; return (newdsk); } /* * Get disk size from GPT. */ static uint64_t drvsize_gpt(struct dsk *dskp) { #ifdef GPT struct gpt_hdr hdr; char *sec; sec = dmadat->secbuf; if (drvread(dskp, sec, 1, 1)) return (0); memcpy(&hdr, sec, sizeof(hdr)); if (memcmp(hdr.hdr_sig, GPT_HDR_SIG, sizeof(hdr.hdr_sig)) != 0 || hdr.hdr_lba_self != 1 || hdr.hdr_revision < 0x00010000 || hdr.hdr_entsz < sizeof(struct gpt_ent) || DEV_BSIZE % hdr.hdr_entsz != 0) { return (0); } return (hdr.hdr_lba_alt + 1); #else return (0); #endif } /* * Get disk size from eax=0x800 and 0x4800. We need to probe both * because 0x4800 may not be available and we would like to get more * or less correct disk size - if it is possible at all. * Note we do not really want to touch drv.c because that code is shared * with boot2 and we can not afford to grow that code. */ static uint64_t drvsize_ext(struct zfsdsk *zdsk) { struct dsk *dskp; uint64_t size, tmp; int cyl, hds, sec; dskp = &zdsk->dsk; /* Try to read disk size from GPT */ size = drvsize_gpt(dskp); if (size != 0) return (size); v86.ctl = V86_FLAGS; v86.addr = 0x13; v86.eax = 0x800; v86.edx = dskp->drive; v86int(); /* Don't error out if we get bad sector number, try EDD as well */ if (V86_CY(v86.efl) || /* carry set */ (v86.edx & 0xff) <= (unsigned)(dskp->drive & 0x7f)) /* unit # bad */ return (0); cyl = ((v86.ecx & 0xc0) << 2) + ((v86.ecx & 0xff00) >> 8) + 1; /* Convert max head # -> # of heads */ hds = ((v86.edx & 0xff00) >> 8) + 1; sec = v86.ecx & 0x3f; size = (uint64_t)cyl * hds * sec; /* Determine if we can use EDD with this device. */ v86.ctl = V86_FLAGS; v86.addr = 0x13; v86.eax = 0x4100; v86.edx = dskp->drive; v86.ebx = 0x55aa; v86int(); if (V86_CY(v86.efl) || /* carry set */ (v86.ebx & 0xffff) != 0xaa55 || /* signature */ (v86.ecx & EDD_INTERFACE_FIXED_DISK) == 0) return (size); tmp = drvsize(dskp); if (tmp > size) size = tmp; return (size); } /* * The "layered" ioctl to read disk/partition size. Unfortunately * the zfsboot case is hardest, because we do not have full software * stack available, so we need to do some manual work here. */ uint64_t ldi_get_size(void *priv) { struct zfsdsk *zdsk = priv; uint64_t size = zdsk->dsk.size; if (zdsk->dsk.start == 0) size = drvsize_ext(zdsk); return (size * DEV_BSIZE); } static void probe_drive(struct zfsdsk *zdsk) { #ifdef GPT struct gpt_hdr hdr; struct gpt_ent *ent; unsigned part, entries_per_sec; daddr_t slba; #endif #if defined(GPT) || defined(LOADER_GELI_SUPPORT) daddr_t elba; #endif struct dos_partition *dp; char *sec; unsigned i; #ifdef LOADER_GELI_SUPPORT /* * Taste the disk, if it is GELI encrypted, decrypt it then dig out the * partition table and probe each slice/partition in turn for a vdev or * GELI encrypted vdev. */ elba = drvsize_ext(zdsk); if (elba > 0) { elba--; } zdsk->gdev = geli_taste(vdev_read, zdsk, elba, "disk%u:0:"); if ((zdsk->gdev != NULL) && (geli_havekey(zdsk->gdev) == 0)) geli_passphrase(zdsk->gdev, gelipw); #endif /* LOADER_GELI_SUPPORT */ sec = dmadat->secbuf; zdsk->dsk.start = 0; #ifdef GPT /* * First check for GPT. */ if (drvread(&zdsk->dsk, sec, 1, 1)) { return; } memcpy(&hdr, sec, sizeof(hdr)); if (memcmp(hdr.hdr_sig, GPT_HDR_SIG, sizeof(hdr.hdr_sig)) != 0 || hdr.hdr_lba_self != 1 || hdr.hdr_revision < 0x00010000 || hdr.hdr_entsz < sizeof(*ent) || DEV_BSIZE % hdr.hdr_entsz != 0) { goto trymbr; } /* * Probe all GPT partitions for the presence of ZFS pools. We * return the spa_t for the first we find (if requested). This * will have the effect of booting from the first pool on the * disk. * * If no vdev is found, GELI decrypting the device and try again */ entries_per_sec = DEV_BSIZE / hdr.hdr_entsz; slba = hdr.hdr_lba_table; elba = slba + hdr.hdr_entries / entries_per_sec; while (slba < elba) { zdsk->dsk.start = 0; if (drvread(&zdsk->dsk, sec, slba, 1)) return; for (part = 0; part < entries_per_sec; part++) { ent = (struct gpt_ent *)(sec + part * hdr.hdr_entsz); if (memcmp(&ent->ent_type, &freebsd_zfs_uuid, sizeof(uuid_t)) == 0) { zdsk->dsk.start = ent->ent_lba_start; zdsk->dsk.size = ent->ent_lba_end - ent->ent_lba_start + 1; zdsk->dsk.slice = part + 1; zdsk->dsk.part = 255; if (vdev_probe(vdev_read2, zdsk, NULL) == 0) { /* * This slice had a vdev. We need a new dsk * structure now since the vdev now owns this one. */ zdsk = copy_dsk(zdsk); } #ifdef LOADER_GELI_SUPPORT else if ((zdsk->gdev = geli_taste(vdev_read, zdsk, ent->ent_lba_end - ent->ent_lba_start, "disk%up%u:", zdsk->dsk.unit, zdsk->dsk.slice)) != NULL) { if (geli_havekey(zdsk->gdev) == 0 || geli_passphrase(zdsk->gdev, gelipw) == 0) { /* * This slice has GELI, check it for ZFS. */ if (vdev_probe(vdev_read2, zdsk, NULL) == 0) { /* * This slice had a vdev. We need a new dsk * structure now since the vdev now owns this one. */ zdsk = copy_dsk(zdsk); } break; } } #endif /* LOADER_GELI_SUPPORT */ } } slba++; } return; trymbr: #endif /* GPT */ if (drvread(&zdsk->dsk, sec, DOSBBSECTOR, 1)) return; dp = (void *)(sec + DOSPARTOFF); for (i = 0; i < NDOSPART; i++) { if (!dp[i].dp_typ) continue; zdsk->dsk.start = dp[i].dp_start; zdsk->dsk.size = dp[i].dp_size; zdsk->dsk.slice = i + 1; if (vdev_probe(vdev_read2, zdsk, NULL) == 0) { zdsk = copy_dsk(zdsk); } #ifdef LOADER_GELI_SUPPORT else if ((zdsk->gdev = geli_taste(vdev_read, zdsk, dp[i].dp_size - dp[i].dp_start, "disk%us%u:")) != NULL) { if (geli_havekey(zdsk->gdev) == 0 || geli_passphrase(zdsk->gdev, gelipw) == 0) { /* * This slice has GELI, check it for ZFS. */ if (vdev_probe(vdev_read2, zdsk, NULL) == 0) { /* * This slice had a vdev. We need a new dsk * structure now since the vdev now owns this one. */ zdsk = copy_dsk(zdsk); } break; } } #endif /* LOADER_GELI_SUPPORT */ } } int main(void) { dnode_phys_t dn; off_t off; struct zfsdsk *zdsk; int autoboot, i; int nextboot; int rc; dmadat = (void *)(roundup2(__base + (int32_t)&_end, 0x10000) - __base); bios_getmem(); if (high_heap_size > 0) { heap_end = PTOV(high_heap_base + high_heap_size); heap_next = PTOV(high_heap_base); } else { heap_next = (char *)dmadat + sizeof(*dmadat); heap_end = (char *)PTOV(bios_basemem); } setheap(heap_next, heap_end); zdsk = calloc(1, sizeof(struct zfsdsk)); zdsk->dsk.drive = *(uint8_t *)PTOV(ARGS); zdsk->dsk.type = zdsk->dsk.drive & DRV_HARD ? TYPE_AD : TYPE_FD; zdsk->dsk.unit = zdsk->dsk.drive & DRV_MASK; zdsk->dsk.slice = *(uint8_t *)PTOV(ARGS + 1) + 1; zdsk->dsk.part = 0; zdsk->dsk.start = 0; zdsk->dsk.size = drvsize_ext(zdsk); bootinfo.bi_version = BOOTINFO_VERSION; bootinfo.bi_size = sizeof(bootinfo); bootinfo.bi_basemem = bios_basemem / 1024; bootinfo.bi_extmem = bios_extmem / 1024; bootinfo.bi_memsizes_valid++; bootinfo.bi_bios_dev = zdsk->dsk.drive; bootdev = MAKEBOOTDEV(dev_maj[zdsk->dsk.type], zdsk->dsk.slice, zdsk->dsk.unit, zdsk->dsk.part); /* Process configuration file */ autoboot = 1; zfs_init(); /* * Probe the boot drive first - we will try to boot from whatever * pool we find on that drive. */ probe_drive(zdsk); /* * Probe the rest of the drives that the bios knows about. This * will find any other available pools and it may fill in missing * vdevs for the boot pool. */ #ifndef VIRTUALBOX for (i = 0; i < *(unsigned char *)PTOV(BIOS_NUMDRIVES); i++) #else for (i = 0; i < MAXBDDEV; i++) #endif { if ((i | DRV_HARD) == *(uint8_t *)PTOV(ARGS)) continue; if (!int13probe(i | DRV_HARD)) break; zdsk = calloc(1, sizeof(struct zfsdsk)); zdsk->dsk.drive = i | DRV_HARD; zdsk->dsk.type = zdsk->dsk.drive & TYPE_AD; zdsk->dsk.unit = i; zdsk->dsk.slice = 0; zdsk->dsk.part = 0; zdsk->dsk.start = 0; zdsk->dsk.size = drvsize_ext(zdsk); probe_drive(zdsk); } /* * The first discovered pool, if any, is the pool. */ spa = spa_get_primary(); if (!spa) { printf("%s: No ZFS pools located, can't boot\n", BOOTPROG); for (;;) ; } primary_spa = spa; primary_vdev = spa_get_primary_vdev(spa); nextboot = 0; rc = vdev_read_pad2(primary_vdev, cmd, sizeof(cmd)); if (vdev_clear_pad2(primary_vdev)) printf("failed to clear pad2 area of primary vdev\n"); if (rc == 0) { if (*cmd) { /* * We could find an old-style ZFS Boot Block header here. * Simply ignore it. */ if (*(uint64_t *)cmd != 0x2f5b007b10c) { /* * Note that parse() is destructive to cmd[] and we also want * to honor RBX_QUIET option that could be present in cmd[]. */ nextboot = 1; memcpy(cmddup, cmd, sizeof(cmd)); if (parse_cmd()) { printf("failed to parse pad2 area of primary vdev\n"); reboot(); } if (!OPT_CHECK(RBX_QUIET)) printf("zfs nextboot: %s\n", cmddup); } /* Do not process this command twice */ *cmd = 0; } } else printf("failed to read pad2 area of primary vdev\n"); /* Mount ZFS only if it's not already mounted via nextboot parsing. */ if (zfsmount.spa == NULL && (zfs_spa_init(spa) != 0 || zfs_mount(spa, 0, &zfsmount) != 0)) { printf("%s: failed to mount default pool %s\n", BOOTPROG, spa->spa_name); autoboot = 0; } else if (zfs_lookup(&zfsmount, PATH_CONFIG, &dn) == 0 || zfs_lookup(&zfsmount, PATH_DOTCONFIG, &dn) == 0) { off = 0; zfs_read(spa, &dn, &off, cmd, sizeof(cmd)); } if (*cmd) { /* * Note that parse_cmd() is destructive to cmd[] and we also want * to honor RBX_QUIET option that could be present in cmd[]. */ memcpy(cmddup, cmd, sizeof(cmd)); if (parse_cmd()) autoboot = 0; if (!OPT_CHECK(RBX_QUIET)) printf("%s: %s\n", PATH_CONFIG, cmddup); /* Do not process this command twice */ *cmd = 0; } /* Do not risk waiting at the prompt forever. */ if (nextboot && !autoboot) reboot(); if (autoboot && !*kname) { /* * Iterate through the list of loader and kernel paths, trying to load. * If interrupted by a keypress, or in case of failure, drop the user * to the boot2 prompt. */ for (i = 0; i < nitems(loadpath); i++) { memcpy(kname, loadpath[i].p, loadpath[i].len); if (keyhit(3)) break; load(); } } /* Present the user with the boot2 prompt. */ for (;;) { if (!autoboot || !OPT_CHECK(RBX_QUIET)) { printf("\nFreeBSD/x86 boot\n"); if (zfs_rlookup(spa, zfsmount.rootobj, rootname) != 0) printf("Default: %s/<0x%llx>:%s\n" "boot: ", spa->spa_name, zfsmount.rootobj, kname); else if (rootname[0] != '\0') printf("Default: %s/%s:%s\n" "boot: ", spa->spa_name, rootname, kname); else printf("Default: %s:%s\n" "boot: ", spa->spa_name, kname); } if (ioctrl & IO_SERIAL) sio_flush(); if (!autoboot || keyhit(5)) getstr(cmd, sizeof(cmd)); else if (!autoboot || !OPT_CHECK(RBX_QUIET)) putchar('\n'); autoboot = 0; if (parse_cmd()) putchar('\a'); else load(); } } /* XXX - Needed for btxld to link the boot2 binary; do not remove. */ void exit(int x) { __exit(x); } void reboot(void) { __exit(0); } static void load(void) { union { struct exec ex; Elf32_Ehdr eh; } hdr; static Elf32_Phdr ep[2]; static Elf32_Shdr es[2]; caddr_t p; dnode_phys_t dn; off_t off; uint32_t addr, x; int fmt, i, j; if (zfs_lookup(&zfsmount, kname, &dn)) { printf("\nCan't find %s\n", kname); return; } off = 0; if (xfsread(&dn, &off, &hdr, sizeof(hdr))) return; if (N_GETMAGIC(hdr.ex) == ZMAGIC) fmt = 0; else if (IS_ELF(hdr.eh)) fmt = 1; else { printf("Invalid %s\n", "format"); return; } if (fmt == 0) { addr = hdr.ex.a_entry & 0xffffff; p = PTOV(addr); off = PAGE_SIZE; if (xfsread(&dn, &off, p, hdr.ex.a_text)) return; p += roundup2(hdr.ex.a_text, PAGE_SIZE); if (xfsread(&dn, &off, p, hdr.ex.a_data)) return; p += hdr.ex.a_data + roundup2(hdr.ex.a_bss, PAGE_SIZE); bootinfo.bi_symtab = VTOP(p); memcpy(p, &hdr.ex.a_syms, sizeof(hdr.ex.a_syms)); p += sizeof(hdr.ex.a_syms); if (hdr.ex.a_syms) { if (xfsread(&dn, &off, p, hdr.ex.a_syms)) return; p += hdr.ex.a_syms; if (xfsread(&dn, &off, p, sizeof(int))) return; x = *(uint32_t *)p; p += sizeof(int); x -= sizeof(int); if (xfsread(&dn, &off, p, x)) return; p += x; } } else { off = hdr.eh.e_phoff; for (j = i = 0; i < hdr.eh.e_phnum && j < 2; i++) { if (xfsread(&dn, &off, ep + j, sizeof(ep[0]))) return; if (ep[j].p_type == PT_LOAD) j++; } for (i = 0; i < 2; i++) { p = PTOV(ep[i].p_paddr & 0xffffff); off = ep[i].p_offset; if (xfsread(&dn, &off, p, ep[i].p_filesz)) return; } p += roundup2(ep[1].p_memsz, PAGE_SIZE); bootinfo.bi_symtab = VTOP(p); if (hdr.eh.e_shnum == hdr.eh.e_shstrndx + 3) { off = hdr.eh.e_shoff + sizeof(es[0]) * (hdr.eh.e_shstrndx + 1); if (xfsread(&dn, &off, &es, sizeof(es))) return; for (i = 0; i < 2; i++) { memcpy(p, &es[i].sh_size, sizeof(es[i].sh_size)); p += sizeof(es[i].sh_size); off = es[i].sh_offset; if (xfsread(&dn, &off, p, es[i].sh_size)) return; p += es[i].sh_size; } } addr = hdr.eh.e_entry & 0xffffff; } bootinfo.bi_esymtab = VTOP(p); bootinfo.bi_kernelname = VTOP(kname); zfsargs.size = sizeof(zfsargs); zfsargs.pool = zfsmount.spa->spa_guid; zfsargs.root = zfsmount.rootobj; zfsargs.primary_pool = primary_spa->spa_guid; #ifdef LOADER_GELI_SUPPORT explicit_bzero(gelipw, sizeof(gelipw)); export_geli_boot_data(&zfsargs.gelidata); #endif if (primary_vdev != NULL) zfsargs.primary_vdev = primary_vdev->v_guid; else printf("failed to detect primary vdev\n"); /* * Note that the zfsargs struct is passed by value, not by pointer. Code in * btxldr.S copies the values from the entry stack to a fixed location * within loader(8) at startup due to the presence of KARGS_FLAGS_EXTARG. */ __exec((caddr_t)addr, RB_BOOTINFO | (opts & RBX_MASK), bootdev, KARGS_FLAGS_ZFS | KARGS_FLAGS_EXTARG, (uint32_t) spa->spa_guid, (uint32_t) (spa->spa_guid >> 32), VTOP(&bootinfo), zfsargs); } static int zfs_mount_ds(char *dsname) { uint64_t newroot; spa_t *newspa; char *q; q = strchr(dsname, '/'); if (q) *q++ = '\0'; newspa = spa_find_by_name(dsname); if (newspa == NULL) { printf("\nCan't find ZFS pool %s\n", dsname); return -1; } if (zfs_spa_init(newspa)) return -1; newroot = 0; if (q) { if (zfs_lookup_dataset(newspa, q, &newroot)) { printf("\nCan't find dataset %s in ZFS pool %s\n", q, newspa->spa_name); return -1; } } if (zfs_mount(newspa, newroot, &zfsmount)) { printf("\nCan't mount ZFS dataset\n"); return -1; } spa = newspa; return (0); } static int parse_cmd(void) { char *arg = cmd; char *ep, *p, *q; const char *cp; int c, i, j; while ((c = *arg++)) { if (c == ' ' || c == '\t' || c == '\n') continue; for (p = arg; *p && *p != '\n' && *p != ' ' && *p != '\t'; p++); ep = p; if (*p) *p++ = 0; if (c == '-') { while ((c = *arg++)) { if (c == 'P') { if (*(uint8_t *)PTOV(0x496) & 0x10) { cp = "yes"; } else { opts |= OPT_SET(RBX_DUAL) | OPT_SET(RBX_SERIAL); cp = "no"; } printf("Keyboard: %s\n", cp); continue; } else if (c == 'S') { j = 0; while ((unsigned int)(i = *arg++ - '0') <= 9) j = j * 10 + i; if (j > 0 && i == -'0') { comspeed = j; break; } /* Fall through to error below ('S' not in optstr[]). */ } for (i = 0; c != optstr[i]; i++) if (i == NOPT - 1) return -1; opts ^= OPT_SET(flags[i]); } ioctrl = OPT_CHECK(RBX_DUAL) ? (IO_SERIAL|IO_KEYBOARD) : OPT_CHECK(RBX_SERIAL) ? IO_SERIAL : IO_KEYBOARD; if (ioctrl & IO_SERIAL) { if (sio_init(115200 / comspeed) != 0) ioctrl &= ~IO_SERIAL; } } if (c == '?') { dnode_phys_t dn; if (zfs_lookup(&zfsmount, arg, &dn) == 0) { zap_list(spa, &dn); } return -1; } else { arg--; /* * Report pool status if the comment is 'status'. Lets * hope no-one wants to load /status as a kernel. */ if (!strcmp(arg, "status")) { spa_all_status(); return -1; } /* * If there is "zfs:" prefix simply ignore it. */ if (strncmp(arg, "zfs:", 4) == 0) arg += 4; /* * If there is a colon, switch pools. */ q = strchr(arg, ':'); if (q) { *q++ = '\0'; if (zfs_mount_ds(arg) != 0) return -1; arg = q; } if ((i = ep - arg)) { if ((size_t)i >= sizeof(kname)) return -1; memcpy(kname, arg, i + 1); } } arg = p; } return 0; } Index: head/stand/libsa/zfs/zfsimpl.c =================================================================== --- head/stand/libsa/zfs/zfsimpl.c (revision 357496) +++ head/stand/libsa/zfs/zfsimpl.c (revision 357497) @@ -1,3556 +1,3668 @@ /*- * Copyright (c) 2007 Doug Rabson * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); /* * Stand-alone ZFS file reader. */ #include #include #include #include #include #include "zfsimpl.h" #include "zfssubr.c" struct zfsmount { const spa_t *spa; objset_phys_t objset; uint64_t rootobj; }; static struct zfsmount zfsmount __unused; /* * The indirect_child_t represents the vdev that we will read from, when we * need to read all copies of the data (e.g. for scrub or reconstruction). * For plain (non-mirror) top-level vdevs (i.e. is_vdev is not a mirror), * ic_vdev is the same as is_vdev. However, for mirror top-level vdevs, * ic_vdev is a child of the mirror. */ typedef struct indirect_child { void *ic_data; vdev_t *ic_vdev; } indirect_child_t; /* * The indirect_split_t represents one mapped segment of an i/o to the * indirect vdev. For non-split (contiguously-mapped) blocks, there will be * only one indirect_split_t, with is_split_offset==0 and is_size==io_size. * For split blocks, there will be several of these. */ typedef struct indirect_split { list_node_t is_node; /* link on iv_splits */ /* * is_split_offset is the offset into the i/o. * This is the sum of the previous splits' is_size's. */ uint64_t is_split_offset; vdev_t *is_vdev; /* top-level vdev */ uint64_t is_target_offset; /* offset on is_vdev */ uint64_t is_size; int is_children; /* number of entries in is_child[] */ /* * is_good_child is the child that we are currently using to * attempt reconstruction. */ int is_good_child; indirect_child_t is_child[1]; /* variable-length */ } indirect_split_t; /* * The indirect_vsd_t is associated with each i/o to the indirect vdev. * It is the "Vdev-Specific Data" in the zio_t's io_vsd. */ typedef struct indirect_vsd { boolean_t iv_split_block; boolean_t iv_reconstruct; list_t iv_splits; /* list of indirect_split_t's */ } indirect_vsd_t; /* * List of all vdevs, chained through v_alllink. */ static vdev_list_t zfs_vdevs; /* * List of ZFS features supported for read */ static const char *features_for_read[] = { "org.illumos:lz4_compress", "com.delphix:hole_birth", "com.delphix:extensible_dataset", "com.delphix:embedded_data", "org.open-zfs:large_blocks", "org.illumos:sha512", "org.illumos:skein", "org.zfsonlinux:large_dnode", "com.joyent:multi_vdev_crash_dump", "com.delphix:spacemap_histogram", "com.delphix:zpool_checkpoint", "com.delphix:spacemap_v2", "com.datto:encryption", "org.zfsonlinux:allocation_classes", "com.datto:resilver_defer", "com.delphix:device_removal", "com.delphix:obsolete_counts", "com.intel:allocation_classes", NULL }; /* * List of all pools, chained through spa_link. */ static spa_list_t zfs_pools; static const dnode_phys_t *dnode_cache_obj; static uint64_t dnode_cache_bn; static char *dnode_cache_buf; -static char *zap_scratch; static char *zfs_temp_buf, *zfs_temp_end, *zfs_temp_ptr; #define TEMP_SIZE (1024 * 1024) static int zio_read(const spa_t *spa, const blkptr_t *bp, void *buf); static int zfs_get_root(const spa_t *spa, uint64_t *objid); static int zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result); static int zap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t integer_size, uint64_t num_integers, void *value); static int objset_get_dnode(const spa_t *, const objset_phys_t *, uint64_t, dnode_phys_t *); static int dnode_read(const spa_t *, const dnode_phys_t *, off_t, void *, size_t); static int vdev_indirect_read(vdev_t *, const blkptr_t *, void *, off_t, size_t); static int vdev_mirror_read(vdev_t *, const blkptr_t *, void *, off_t, size_t); vdev_indirect_mapping_t *vdev_indirect_mapping_open(spa_t *, objset_phys_t *, uint64_t); vdev_indirect_mapping_entry_phys_t * vdev_indirect_mapping_duplicate_adjacent_entries(vdev_t *, uint64_t, uint64_t, uint64_t *); static void zfs_init(void) { STAILQ_INIT(&zfs_vdevs); STAILQ_INIT(&zfs_pools); zfs_temp_buf = malloc(TEMP_SIZE); zfs_temp_end = zfs_temp_buf + TEMP_SIZE; zfs_temp_ptr = zfs_temp_buf; dnode_cache_buf = malloc(SPA_MAXBLOCKSIZE); - zap_scratch = malloc(SPA_MAXBLOCKSIZE); zfs_init_crc(); } static void * zfs_alloc(size_t size) { char *ptr; if (zfs_temp_ptr + size > zfs_temp_end) { panic("ZFS: out of temporary buffer space"); } ptr = zfs_temp_ptr; zfs_temp_ptr += size; return (ptr); } static void zfs_free(void *ptr, size_t size) { zfs_temp_ptr -= size; if (zfs_temp_ptr != ptr) { panic("ZFS: zfs_alloc()/zfs_free() mismatch"); } } static int xdr_int(const unsigned char **xdr, int *ip) { *ip = be32dec(*xdr); (*xdr) += 4; return (0); } static int xdr_u_int(const unsigned char **xdr, u_int *ip) { *ip = be32dec(*xdr); (*xdr) += 4; return (0); } static int xdr_uint64_t(const unsigned char **xdr, uint64_t *lp) { u_int hi, lo; xdr_u_int(xdr, &hi); xdr_u_int(xdr, &lo); *lp = (((uint64_t)hi) << 32) | lo; return (0); } static int nvlist_find(const unsigned char *nvlist, const char *name, int type, int *elementsp, void *valuep) { const unsigned char *p, *pair; int junk; int encoded_size, decoded_size; p = nvlist; xdr_int(&p, &junk); xdr_int(&p, &junk); pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); while (encoded_size && decoded_size) { int namelen, pairtype, elements; const char *pairname; xdr_int(&p, &namelen); pairname = (const char *)p; p += roundup(namelen, 4); xdr_int(&p, &pairtype); if (memcmp(name, pairname, namelen) == 0 && type == pairtype) { xdr_int(&p, &elements); if (elementsp) *elementsp = elements; if (type == DATA_TYPE_UINT64) { xdr_uint64_t(&p, (uint64_t *)valuep); return (0); } else if (type == DATA_TYPE_STRING) { int len; xdr_int(&p, &len); (*(const char **)valuep) = (const char *)p; return (0); } else if (type == DATA_TYPE_NVLIST || type == DATA_TYPE_NVLIST_ARRAY) { (*(const unsigned char **)valuep) = (const unsigned char *)p; return (0); } else { return (EIO); } } else { /* * Not the pair we are looking for, skip to the * next one. */ p = pair + encoded_size; } pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); } return (EIO); } static int nvlist_check_features_for_read(const unsigned char *nvlist) { const unsigned char *p, *pair; int junk; int encoded_size, decoded_size; int rc; rc = 0; p = nvlist; xdr_int(&p, &junk); xdr_int(&p, &junk); pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); while (encoded_size && decoded_size) { int namelen, pairtype; const char *pairname; int i, found; found = 0; xdr_int(&p, &namelen); pairname = (const char *)p; p += roundup(namelen, 4); xdr_int(&p, &pairtype); for (i = 0; features_for_read[i] != NULL; i++) { if (memcmp(pairname, features_for_read[i], namelen) == 0) { found = 1; break; } } if (!found) { printf("ZFS: unsupported feature: %s\n", pairname); rc = EIO; } p = pair + encoded_size; pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); } return (rc); } /* * Return the next nvlist in an nvlist array. */ static const unsigned char * nvlist_next(const unsigned char *nvlist) { const unsigned char *p, *pair; int junk; int encoded_size, decoded_size; p = nvlist; xdr_int(&p, &junk); xdr_int(&p, &junk); pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); while (encoded_size && decoded_size) { p = pair + encoded_size; pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); } return (p); } #ifdef TEST static const unsigned char * nvlist_print(const unsigned char *nvlist, unsigned int indent) { static const char *typenames[] = { "DATA_TYPE_UNKNOWN", "DATA_TYPE_BOOLEAN", "DATA_TYPE_BYTE", "DATA_TYPE_INT16", "DATA_TYPE_UINT16", "DATA_TYPE_INT32", "DATA_TYPE_UINT32", "DATA_TYPE_INT64", "DATA_TYPE_UINT64", "DATA_TYPE_STRING", "DATA_TYPE_BYTE_ARRAY", "DATA_TYPE_INT16_ARRAY", "DATA_TYPE_UINT16_ARRAY", "DATA_TYPE_INT32_ARRAY", "DATA_TYPE_UINT32_ARRAY", "DATA_TYPE_INT64_ARRAY", "DATA_TYPE_UINT64_ARRAY", "DATA_TYPE_STRING_ARRAY", "DATA_TYPE_HRTIME", "DATA_TYPE_NVLIST", "DATA_TYPE_NVLIST_ARRAY", "DATA_TYPE_BOOLEAN_VALUE", "DATA_TYPE_INT8", "DATA_TYPE_UINT8", "DATA_TYPE_BOOLEAN_ARRAY", "DATA_TYPE_INT8_ARRAY", "DATA_TYPE_UINT8_ARRAY" }; unsigned int i, j; const unsigned char *p, *pair; int junk; int encoded_size, decoded_size; p = nvlist; xdr_int(&p, &junk); xdr_int(&p, &junk); pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); while (encoded_size && decoded_size) { int namelen, pairtype, elements; const char *pairname; xdr_int(&p, &namelen); pairname = (const char *)p; p += roundup(namelen, 4); xdr_int(&p, &pairtype); for (i = 0; i < indent; i++) printf(" "); printf("%s %s", typenames[pairtype], pairname); xdr_int(&p, &elements); switch (pairtype) { case DATA_TYPE_UINT64: { uint64_t val; xdr_uint64_t(&p, &val); printf(" = 0x%jx\n", (uintmax_t)val); break; } case DATA_TYPE_STRING: { int len; xdr_int(&p, &len); printf(" = \"%s\"\n", p); break; } case DATA_TYPE_NVLIST: printf("\n"); nvlist_print(p, indent + 1); break; case DATA_TYPE_NVLIST_ARRAY: for (j = 0; j < elements; j++) { printf("[%d]\n", j); p = nvlist_print(p, indent + 1); if (j != elements - 1) { for (i = 0; i < indent; i++) printf(" "); printf("%s %s", typenames[pairtype], pairname); } } break; default: printf("\n"); } p = pair + encoded_size; pair = p; xdr_int(&p, &encoded_size); xdr_int(&p, &decoded_size); } return (p); } #endif static int vdev_read_phys(vdev_t *vdev, const blkptr_t *bp, void *buf, off_t offset, size_t size) { size_t psize; int rc; if (!vdev->v_phys_read) return (EIO); if (bp) { psize = BP_GET_PSIZE(bp); } else { psize = size; } rc = vdev->v_phys_read(vdev, vdev->v_read_priv, offset, buf, psize); if (rc == 0) { if (bp != NULL) rc = zio_checksum_verify(vdev->v_spa, bp, buf); } return (rc); } typedef struct remap_segment { vdev_t *rs_vd; uint64_t rs_offset; uint64_t rs_asize; uint64_t rs_split_offset; list_node_t rs_node; } remap_segment_t; static remap_segment_t * rs_alloc(vdev_t *vd, uint64_t offset, uint64_t asize, uint64_t split_offset) { remap_segment_t *rs = malloc(sizeof (remap_segment_t)); if (rs != NULL) { rs->rs_vd = vd; rs->rs_offset = offset; rs->rs_asize = asize; rs->rs_split_offset = split_offset; } return (rs); } vdev_indirect_mapping_t * vdev_indirect_mapping_open(spa_t *spa, objset_phys_t *os, uint64_t mapping_object) { vdev_indirect_mapping_t *vim; vdev_indirect_mapping_phys_t *vim_phys; int rc; vim = calloc(1, sizeof (*vim)); if (vim == NULL) return (NULL); vim->vim_dn = calloc(1, sizeof (*vim->vim_dn)); if (vim->vim_dn == NULL) { free(vim); return (NULL); } rc = objset_get_dnode(spa, os, mapping_object, vim->vim_dn); if (rc != 0) { free(vim->vim_dn); free(vim); return (NULL); } vim->vim_spa = spa; vim->vim_phys = malloc(sizeof (*vim->vim_phys)); if (vim->vim_phys == NULL) { free(vim->vim_dn); free(vim); return (NULL); } vim_phys = (vdev_indirect_mapping_phys_t *)DN_BONUS(vim->vim_dn); *vim->vim_phys = *vim_phys; vim->vim_objset = os; vim->vim_object = mapping_object; vim->vim_entries = NULL; vim->vim_havecounts = (vim->vim_dn->dn_bonuslen > VDEV_INDIRECT_MAPPING_SIZE_V0); return (vim); } /* * Compare an offset with an indirect mapping entry; there are three * possible scenarios: * * 1. The offset is "less than" the mapping entry; meaning the * offset is less than the source offset of the mapping entry. In * this case, there is no overlap between the offset and the * mapping entry and -1 will be returned. * * 2. The offset is "greater than" the mapping entry; meaning the * offset is greater than the mapping entry's source offset plus * the entry's size. In this case, there is no overlap between * the offset and the mapping entry and 1 will be returned. * * NOTE: If the offset is actually equal to the entry's offset * plus size, this is considered to be "greater" than the entry, * and this case applies (i.e. 1 will be returned). Thus, the * entry's "range" can be considered to be inclusive at its * start, but exclusive at its end: e.g. [src, src + size). * * 3. The last case to consider is if the offset actually falls * within the mapping entry's range. If this is the case, the * offset is considered to be "equal to" the mapping entry and * 0 will be returned. * * NOTE: If the offset is equal to the entry's source offset, * this case applies and 0 will be returned. If the offset is * equal to the entry's source plus its size, this case does * *not* apply (see "NOTE" above for scenario 2), and 1 will be * returned. */ static int dva_mapping_overlap_compare(const void *v_key, const void *v_array_elem) { const uint64_t *key = v_key; const vdev_indirect_mapping_entry_phys_t *array_elem = v_array_elem; uint64_t src_offset = DVA_MAPPING_GET_SRC_OFFSET(array_elem); if (*key < src_offset) { return (-1); } else if (*key < src_offset + DVA_GET_ASIZE(&array_elem->vimep_dst)) { return (0); } else { return (1); } } /* * Return array entry. */ static vdev_indirect_mapping_entry_phys_t * vdev_indirect_mapping_entry(vdev_indirect_mapping_t *vim, uint64_t index) { uint64_t size; off_t offset = 0; int rc; if (vim->vim_phys->vimp_num_entries == 0) return (NULL); if (vim->vim_entries == NULL) { uint64_t bsize; bsize = vim->vim_dn->dn_datablkszsec << SPA_MINBLOCKSHIFT; size = vim->vim_phys->vimp_num_entries * sizeof (*vim->vim_entries); if (size > bsize) { size = bsize / sizeof (*vim->vim_entries); size *= sizeof (*vim->vim_entries); } vim->vim_entries = malloc(size); if (vim->vim_entries == NULL) return (NULL); vim->vim_num_entries = size / sizeof (*vim->vim_entries); offset = index * sizeof (*vim->vim_entries); } /* We have data in vim_entries */ if (offset == 0) { if (index >= vim->vim_entry_offset && index <= vim->vim_entry_offset + vim->vim_num_entries) { index -= vim->vim_entry_offset; return (&vim->vim_entries[index]); } offset = index * sizeof (*vim->vim_entries); } vim->vim_entry_offset = index; size = vim->vim_num_entries * sizeof (*vim->vim_entries); rc = dnode_read(vim->vim_spa, vim->vim_dn, offset, vim->vim_entries, size); if (rc != 0) { /* Read error, invalidate vim_entries. */ free(vim->vim_entries); vim->vim_entries = NULL; return (NULL); } index -= vim->vim_entry_offset; return (&vim->vim_entries[index]); } /* * Returns the mapping entry for the given offset. * * It's possible that the given offset will not be in the mapping table * (i.e. no mapping entries contain this offset), in which case, the * return value value depends on the "next_if_missing" parameter. * * If the offset is not found in the table and "next_if_missing" is * B_FALSE, then NULL will always be returned. The behavior is intended * to allow consumers to get the entry corresponding to the offset * parameter, iff the offset overlaps with an entry in the table. * * If the offset is not found in the table and "next_if_missing" is * B_TRUE, then the entry nearest to the given offset will be returned, * such that the entry's source offset is greater than the offset * passed in (i.e. the "next" mapping entry in the table is returned, if * the offset is missing from the table). If there are no entries whose * source offset is greater than the passed in offset, NULL is returned. */ static vdev_indirect_mapping_entry_phys_t * vdev_indirect_mapping_entry_for_offset(vdev_indirect_mapping_t *vim, uint64_t offset) { ASSERT(vim->vim_phys->vimp_num_entries > 0); vdev_indirect_mapping_entry_phys_t *entry; uint64_t last = vim->vim_phys->vimp_num_entries - 1; uint64_t base = 0; /* * We don't define these inside of the while loop because we use * their value in the case that offset isn't in the mapping. */ uint64_t mid; int result; while (last >= base) { mid = base + ((last - base) >> 1); entry = vdev_indirect_mapping_entry(vim, mid); if (entry == NULL) break; result = dva_mapping_overlap_compare(&offset, entry); if (result == 0) { break; } else if (result < 0) { last = mid - 1; } else { base = mid + 1; } } return (entry); } /* * Given an indirect vdev and an extent on that vdev, it duplicates the * physical entries of the indirect mapping that correspond to the extent * to a new array and returns a pointer to it. In addition, copied_entries * is populated with the number of mapping entries that were duplicated. * * Finally, since we are doing an allocation, it is up to the caller to * free the array allocated in this function. */ vdev_indirect_mapping_entry_phys_t * vdev_indirect_mapping_duplicate_adjacent_entries(vdev_t *vd, uint64_t offset, uint64_t asize, uint64_t *copied_entries) { vdev_indirect_mapping_entry_phys_t *duplicate_mappings = NULL; vdev_indirect_mapping_t *vim = vd->v_mapping; uint64_t entries = 0; vdev_indirect_mapping_entry_phys_t *first_mapping = vdev_indirect_mapping_entry_for_offset(vim, offset); ASSERT3P(first_mapping, !=, NULL); vdev_indirect_mapping_entry_phys_t *m = first_mapping; while (asize > 0) { uint64_t size = DVA_GET_ASIZE(&m->vimep_dst); uint64_t inner_offset = offset - DVA_MAPPING_GET_SRC_OFFSET(m); uint64_t inner_size = MIN(asize, size - inner_offset); offset += inner_size; asize -= inner_size; entries++; m++; } size_t copy_length = entries * sizeof (*first_mapping); duplicate_mappings = malloc(copy_length); if (duplicate_mappings != NULL) bcopy(first_mapping, duplicate_mappings, copy_length); else entries = 0; *copied_entries = entries; return (duplicate_mappings); } static vdev_t * vdev_lookup_top(spa_t *spa, uint64_t vdev) { vdev_t *rvd; vdev_list_t *vlist; vlist = &spa->spa_root_vdev->v_children; STAILQ_FOREACH(rvd, vlist, v_childlink) if (rvd->v_id == vdev) break; return (rvd); } /* * This is a callback for vdev_indirect_remap() which allocates an * indirect_split_t for each split segment and adds it to iv_splits. */ static void vdev_indirect_gather_splits(uint64_t split_offset, vdev_t *vd, uint64_t offset, uint64_t size, void *arg) { int n = 1; zio_t *zio = arg; indirect_vsd_t *iv = zio->io_vsd; if (vd->v_read == vdev_indirect_read) return; if (vd->v_read == vdev_mirror_read) n = vd->v_nchildren; indirect_split_t *is = malloc(offsetof(indirect_split_t, is_child[n])); if (is == NULL) { zio->io_error = ENOMEM; return; } bzero(is, offsetof(indirect_split_t, is_child[n])); is->is_children = n; is->is_size = size; is->is_split_offset = split_offset; is->is_target_offset = offset; is->is_vdev = vd; /* * Note that we only consider multiple copies of the data for * *mirror* vdevs. We don't for "replacing" or "spare" vdevs, even * though they use the same ops as mirror, because there's only one * "good" copy under the replacing/spare. */ if (vd->v_read == vdev_mirror_read) { int i = 0; vdev_t *kid; STAILQ_FOREACH(kid, &vd->v_children, v_childlink) { is->is_child[i++].ic_vdev = kid; } } else { is->is_child[0].ic_vdev = vd; } list_insert_tail(&iv->iv_splits, is); } static void vdev_indirect_remap(vdev_t *vd, uint64_t offset, uint64_t asize, void *arg) { list_t stack; spa_t *spa = vd->v_spa; zio_t *zio = arg; remap_segment_t *rs; list_create(&stack, sizeof (remap_segment_t), offsetof(remap_segment_t, rs_node)); rs = rs_alloc(vd, offset, asize, 0); if (rs == NULL) { printf("vdev_indirect_remap: out of memory.\n"); zio->io_error = ENOMEM; } for (; rs != NULL; rs = list_remove_head(&stack)) { vdev_t *v = rs->rs_vd; uint64_t num_entries = 0; /* vdev_indirect_mapping_t *vim = v->v_mapping; */ vdev_indirect_mapping_entry_phys_t *mapping = vdev_indirect_mapping_duplicate_adjacent_entries(v, rs->rs_offset, rs->rs_asize, &num_entries); if (num_entries == 0) zio->io_error = ENOMEM; for (uint64_t i = 0; i < num_entries; i++) { vdev_indirect_mapping_entry_phys_t *m = &mapping[i]; uint64_t size = DVA_GET_ASIZE(&m->vimep_dst); uint64_t dst_offset = DVA_GET_OFFSET(&m->vimep_dst); uint64_t dst_vdev = DVA_GET_VDEV(&m->vimep_dst); uint64_t inner_offset = rs->rs_offset - DVA_MAPPING_GET_SRC_OFFSET(m); uint64_t inner_size = MIN(rs->rs_asize, size - inner_offset); vdev_t *dst_v = vdev_lookup_top(spa, dst_vdev); if (dst_v->v_read == vdev_indirect_read) { remap_segment_t *o; o = rs_alloc(dst_v, dst_offset + inner_offset, inner_size, rs->rs_split_offset); if (o == NULL) { printf("vdev_indirect_remap: " "out of memory.\n"); zio->io_error = ENOMEM; break; } list_insert_head(&stack, o); } vdev_indirect_gather_splits(rs->rs_split_offset, dst_v, dst_offset + inner_offset, inner_size, arg); /* * vdev_indirect_gather_splits can have memory * allocation error, we can not recover from it. */ if (zio->io_error != 0) break; rs->rs_offset += inner_size; rs->rs_asize -= inner_size; rs->rs_split_offset += inner_size; } free(mapping); free(rs); if (zio->io_error != 0) break; } list_destroy(&stack); } static void vdev_indirect_map_free(zio_t *zio) { indirect_vsd_t *iv = zio->io_vsd; indirect_split_t *is; while ((is = list_head(&iv->iv_splits)) != NULL) { for (int c = 0; c < is->is_children; c++) { indirect_child_t *ic = &is->is_child[c]; free(ic->ic_data); } list_remove(&iv->iv_splits, is); free(is); } free(iv); } static int vdev_indirect_read(vdev_t *vdev, const blkptr_t *bp, void *buf, off_t offset, size_t bytes) { zio_t zio; spa_t *spa = vdev->v_spa; indirect_vsd_t *iv; indirect_split_t *first; int rc = EIO; iv = calloc(1, sizeof(*iv)); if (iv == NULL) return (ENOMEM); list_create(&iv->iv_splits, sizeof (indirect_split_t), offsetof(indirect_split_t, is_node)); bzero(&zio, sizeof(zio)); zio.io_spa = spa; zio.io_bp = (blkptr_t *)bp; zio.io_data = buf; zio.io_size = bytes; zio.io_offset = offset; zio.io_vd = vdev; zio.io_vsd = iv; if (vdev->v_mapping == NULL) { vdev_indirect_config_t *vic; vic = &vdev->vdev_indirect_config; vdev->v_mapping = vdev_indirect_mapping_open(spa, &spa->spa_mos, vic->vic_mapping_object); } vdev_indirect_remap(vdev, offset, bytes, &zio); if (zio.io_error != 0) return (zio.io_error); first = list_head(&iv->iv_splits); if (first->is_size == zio.io_size) { /* * This is not a split block; we are pointing to the entire * data, which will checksum the same as the original data. * Pass the BP down so that the child i/o can verify the * checksum, and try a different location if available * (e.g. on a mirror). * * While this special case could be handled the same as the * general (split block) case, doing it this way ensures * that the vast majority of blocks on indirect vdevs * (which are not split) are handled identically to blocks * on non-indirect vdevs. This allows us to be less strict * about performance in the general (but rare) case. */ rc = first->is_vdev->v_read(first->is_vdev, zio.io_bp, zio.io_data, first->is_target_offset, bytes); } else { iv->iv_split_block = B_TRUE; /* * Read one copy of each split segment, from the * top-level vdev. Since we don't know the * checksum of each split individually, the child * zio can't ensure that we get the right data. * E.g. if it's a mirror, it will just read from a * random (healthy) leaf vdev. We have to verify * the checksum in vdev_indirect_io_done(). */ for (indirect_split_t *is = list_head(&iv->iv_splits); is != NULL; is = list_next(&iv->iv_splits, is)) { char *ptr = zio.io_data; rc = is->is_vdev->v_read(is->is_vdev, zio.io_bp, ptr + is->is_split_offset, is->is_target_offset, is->is_size); } if (zio_checksum_verify(spa, zio.io_bp, zio.io_data)) rc = ECKSUM; else rc = 0; } vdev_indirect_map_free(&zio); if (rc == 0) rc = zio.io_error; return (rc); } static int vdev_disk_read(vdev_t *vdev, const blkptr_t *bp, void *buf, off_t offset, size_t bytes) { return (vdev_read_phys(vdev, bp, buf, offset + VDEV_LABEL_START_SIZE, bytes)); } static int vdev_mirror_read(vdev_t *vdev, const blkptr_t *bp, void *buf, off_t offset, size_t bytes) { vdev_t *kid; int rc; rc = EIO; STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) { if (kid->v_state != VDEV_STATE_HEALTHY) continue; rc = kid->v_read(kid, bp, buf, offset, bytes); if (!rc) return (0); } return (rc); } static int vdev_replacing_read(vdev_t *vdev, const blkptr_t *bp, void *buf, off_t offset, size_t bytes) { vdev_t *kid; /* * Here we should have two kids: * First one which is the one we are replacing and we can trust * only this one to have valid data, but it might not be present. * Second one is that one we are replacing with. It is most likely * healthy, but we can't trust it has needed data, so we won't use it. */ kid = STAILQ_FIRST(&vdev->v_children); if (kid == NULL) return (EIO); if (kid->v_state != VDEV_STATE_HEALTHY) return (EIO); return (kid->v_read(kid, bp, buf, offset, bytes)); } static vdev_t * vdev_find(uint64_t guid) { vdev_t *vdev; STAILQ_FOREACH(vdev, &zfs_vdevs, v_alllink) if (vdev->v_guid == guid) return (vdev); return (0); } static vdev_t * vdev_create(uint64_t guid, vdev_read_t *_read) { vdev_t *vdev; vdev_indirect_config_t *vic; vdev = calloc(1, sizeof(vdev_t)); if (vdev != NULL) { STAILQ_INIT(&vdev->v_children); vdev->v_guid = guid; vdev->v_read = _read; /* * root vdev has no read function, we use this fact to * skip setting up data we do not need for root vdev. * We only point root vdev from spa. */ if (_read != NULL) { vic = &vdev->vdev_indirect_config; vic->vic_prev_indirect_vdev = UINT64_MAX; STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink); } } return (vdev); } static void vdev_set_initial_state(vdev_t *vdev, const unsigned char *nvlist) { uint64_t is_offline, is_faulted, is_degraded, is_removed, isnt_present; uint64_t is_log; is_offline = is_removed = is_faulted = is_degraded = isnt_present = 0; is_log = 0; (void) nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, NULL, &is_offline); (void) nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, NULL, &is_removed); (void) nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, NULL, &is_faulted); (void) nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, NULL, &is_degraded); (void) nvlist_find(nvlist, ZPOOL_CONFIG_NOT_PRESENT, DATA_TYPE_UINT64, NULL, &isnt_present); (void) nvlist_find(nvlist, ZPOOL_CONFIG_IS_LOG, DATA_TYPE_UINT64, NULL, &is_log); if (is_offline != 0) vdev->v_state = VDEV_STATE_OFFLINE; else if (is_removed != 0) vdev->v_state = VDEV_STATE_REMOVED; else if (is_faulted != 0) vdev->v_state = VDEV_STATE_FAULTED; else if (is_degraded != 0) vdev->v_state = VDEV_STATE_DEGRADED; else if (isnt_present != 0) vdev->v_state = VDEV_STATE_CANT_OPEN; vdev->v_islog = is_log != 0; } static int vdev_init(uint64_t guid, const unsigned char *nvlist, vdev_t **vdevp) { uint64_t id, ashift, asize, nparity; const char *path; const char *type; vdev_t *vdev; if (nvlist_find(nvlist, ZPOOL_CONFIG_ID, DATA_TYPE_UINT64, NULL, &id) || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE, DATA_TYPE_STRING, NULL, &type)) { return (ENOENT); } if (strcmp(type, VDEV_TYPE_MIRROR) != 0 && strcmp(type, VDEV_TYPE_DISK) != 0 && #ifdef ZFS_TEST strcmp(type, VDEV_TYPE_FILE) != 0 && #endif strcmp(type, VDEV_TYPE_RAIDZ) != 0 && strcmp(type, VDEV_TYPE_INDIRECT) != 0 && strcmp(type, VDEV_TYPE_REPLACING) != 0) { printf("ZFS: can only boot from disk, mirror, raidz1, " "raidz2 and raidz3 vdevs\n"); return (EIO); } if (strcmp(type, VDEV_TYPE_MIRROR) == 0) vdev = vdev_create(guid, vdev_mirror_read); else if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) vdev = vdev_create(guid, vdev_raidz_read); else if (strcmp(type, VDEV_TYPE_REPLACING) == 0) vdev = vdev_create(guid, vdev_replacing_read); else if (strcmp(type, VDEV_TYPE_INDIRECT) == 0) { vdev_indirect_config_t *vic; vdev = vdev_create(guid, vdev_indirect_read); if (vdev != NULL) { vdev->v_state = VDEV_STATE_HEALTHY; vic = &vdev->vdev_indirect_config; nvlist_find(nvlist, ZPOOL_CONFIG_INDIRECT_OBJECT, DATA_TYPE_UINT64, NULL, &vic->vic_mapping_object); nvlist_find(nvlist, ZPOOL_CONFIG_INDIRECT_BIRTHS, DATA_TYPE_UINT64, NULL, &vic->vic_births_object); nvlist_find(nvlist, ZPOOL_CONFIG_PREV_INDIRECT_VDEV, DATA_TYPE_UINT64, NULL, &vic->vic_prev_indirect_vdev); } } else { vdev = vdev_create(guid, vdev_disk_read); } if (vdev == NULL) return (ENOMEM); vdev_set_initial_state(vdev, nvlist); vdev->v_id = id; if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT, DATA_TYPE_UINT64, NULL, &ashift) == 0) vdev->v_ashift = ashift; if (nvlist_find(nvlist, ZPOOL_CONFIG_ASIZE, DATA_TYPE_UINT64, NULL, &asize) == 0) { vdev->v_psize = asize + VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE; } if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY, DATA_TYPE_UINT64, NULL, &nparity) == 0) vdev->v_nparity = nparity; if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH, DATA_TYPE_STRING, NULL, &path) == 0) { if (strncmp(path, "/dev/", 5) == 0) path += 5; vdev->v_name = strdup(path); } else { char *name; name = NULL; if (strcmp(type, "raidz") == 0) { if (vdev->v_nparity < 1 || vdev->v_nparity > 3) { printf("ZFS: invalid raidz parity: %d\n", vdev->v_nparity); return (EIO); } (void) asprintf(&name, "%s%d-%" PRIu64, type, vdev->v_nparity, id); } else { (void) asprintf(&name, "%s-%" PRIu64, type, id); } vdev->v_name = name; } *vdevp = vdev; return (0); } /* * Find slot for vdev. We return either NULL to signal to use * STAILQ_INSERT_HEAD, or we return link element to be used with * STAILQ_INSERT_AFTER. */ static vdev_t * vdev_find_previous(vdev_t *top_vdev, vdev_t *vdev) { vdev_t *v, *previous; if (STAILQ_EMPTY(&top_vdev->v_children)) return (NULL); previous = NULL; STAILQ_FOREACH(v, &top_vdev->v_children, v_childlink) { if (v->v_id > vdev->v_id) return (previous); if (v->v_id == vdev->v_id) return (v); if (v->v_id < vdev->v_id) previous = v; } return (previous); } static size_t vdev_child_count(vdev_t *vdev) { vdev_t *v; size_t count; count = 0; STAILQ_FOREACH(v, &vdev->v_children, v_childlink) { count++; } return (count); } /* * Insert vdev into top_vdev children list. List is ordered by v_id. */ static void vdev_insert(vdev_t *top_vdev, vdev_t *vdev) { vdev_t *previous; size_t count; /* * The top level vdev can appear in random order, depending how * the firmware is presenting the disk devices. * However, we will insert vdev to create list ordered by v_id, * so we can use either STAILQ_INSERT_HEAD or STAILQ_INSERT_AFTER * as STAILQ does not have insert before. */ previous = vdev_find_previous(top_vdev, vdev); if (previous == NULL) { STAILQ_INSERT_HEAD(&top_vdev->v_children, vdev, v_childlink); } else if (previous->v_id == vdev->v_id) { /* * This vdev was configured from label config, * do not insert duplicate. */ return; } else { STAILQ_INSERT_AFTER(&top_vdev->v_children, previous, vdev, v_childlink); } count = vdev_child_count(top_vdev); if (top_vdev->v_nchildren < count) top_vdev->v_nchildren = count; } static int vdev_from_nvlist(spa_t *spa, uint64_t top_guid, const unsigned char *nvlist) { vdev_t *top_vdev, *vdev; const unsigned char *kids; int rc, nkids; /* Get top vdev. */ top_vdev = vdev_find(top_guid); if (top_vdev == NULL) { rc = vdev_init(top_guid, nvlist, &top_vdev); if (rc != 0) return (rc); top_vdev->v_spa = spa; top_vdev->v_top = top_vdev; vdev_insert(spa->spa_root_vdev, top_vdev); } /* Add children if there are any. */ rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN, DATA_TYPE_NVLIST_ARRAY, &nkids, &kids); if (rc == 0) { for (int i = 0; i < nkids; i++) { uint64_t guid; rc = nvlist_find(kids, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64, NULL, &guid); if (rc != 0) return (rc); rc = vdev_init(guid, kids, &vdev); if (rc != 0) return (rc); vdev->v_spa = spa; vdev->v_top = top_vdev; vdev_insert(top_vdev, vdev); kids = nvlist_next(kids); } } else { /* * When there are no children, nvlist_find() does return * error, reset it because leaf devices have no children. */ rc = 0; } return (rc); } static int vdev_init_from_label(spa_t *spa, const unsigned char *nvlist) { uint64_t pool_guid, top_guid; const unsigned char *vdevs; if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_GUID, DATA_TYPE_UINT64, NULL, &pool_guid) || nvlist_find(nvlist, ZPOOL_CONFIG_TOP_GUID, DATA_TYPE_UINT64, NULL, &top_guid) || nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_TREE, DATA_TYPE_NVLIST, NULL, &vdevs)) { printf("ZFS: can't find vdev details\n"); return (ENOENT); } return (vdev_from_nvlist(spa, top_guid, vdevs)); } static void vdev_set_state(vdev_t *vdev) { vdev_t *kid; int good_kids; int bad_kids; STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) { vdev_set_state(kid); } /* * A mirror or raidz is healthy if all its kids are healthy. A * mirror is degraded if any of its kids is healthy; a raidz * is degraded if at most nparity kids are offline. */ if (STAILQ_FIRST(&vdev->v_children)) { good_kids = 0; bad_kids = 0; STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) { if (kid->v_state == VDEV_STATE_HEALTHY) good_kids++; else bad_kids++; } if (bad_kids == 0) { vdev->v_state = VDEV_STATE_HEALTHY; } else { if (vdev->v_read == vdev_mirror_read) { if (good_kids) { vdev->v_state = VDEV_STATE_DEGRADED; } else { vdev->v_state = VDEV_STATE_OFFLINE; } } else if (vdev->v_read == vdev_raidz_read) { if (bad_kids > vdev->v_nparity) { vdev->v_state = VDEV_STATE_OFFLINE; } else { vdev->v_state = VDEV_STATE_DEGRADED; } } } } } static int vdev_update_from_nvlist(uint64_t top_guid, const unsigned char *nvlist) { vdev_t *vdev; const unsigned char *kids; int rc, nkids; /* Update top vdev. */ vdev = vdev_find(top_guid); if (vdev != NULL) vdev_set_initial_state(vdev, nvlist); /* Update children if there are any. */ rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN, DATA_TYPE_NVLIST_ARRAY, &nkids, &kids); if (rc == 0) { for (int i = 0; i < nkids; i++) { uint64_t guid; rc = nvlist_find(kids, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64, NULL, &guid); if (rc != 0) break; vdev = vdev_find(guid); if (vdev != NULL) vdev_set_initial_state(vdev, kids); kids = nvlist_next(kids); } } else { rc = 0; } return (rc); } static int vdev_init_from_nvlist(spa_t *spa, const unsigned char *nvlist) { uint64_t pool_guid, vdev_children; const unsigned char *vdevs, *kids; int rc, nkids; if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_GUID, DATA_TYPE_UINT64, NULL, &pool_guid) || nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_CHILDREN, DATA_TYPE_UINT64, NULL, &vdev_children) || nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_TREE, DATA_TYPE_NVLIST, NULL, &vdevs)) { printf("ZFS: can't find vdev details\n"); return (ENOENT); } /* Wrong guid?! */ if (spa->spa_guid != pool_guid) return (EINVAL); spa->spa_root_vdev->v_nchildren = vdev_children; rc = nvlist_find(vdevs, ZPOOL_CONFIG_CHILDREN, DATA_TYPE_NVLIST_ARRAY, &nkids, &kids); /* * MOS config has at least one child for root vdev. */ if (rc != 0) return (rc); for (int i = 0; i < nkids; i++) { uint64_t guid; vdev_t *vdev; rc = nvlist_find(kids, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64, NULL, &guid); if (rc != 0) break; vdev = vdev_find(guid); /* * Top level vdev is missing, create it. */ if (vdev == NULL) rc = vdev_from_nvlist(spa, guid, kids); else rc = vdev_update_from_nvlist(guid, kids); if (rc != 0) break; kids = nvlist_next(kids); } /* * Re-evaluate top-level vdev state. */ vdev_set_state(spa->spa_root_vdev); return (rc); } static spa_t * spa_find_by_guid(uint64_t guid) { spa_t *spa; STAILQ_FOREACH(spa, &zfs_pools, spa_link) if (spa->spa_guid == guid) return (spa); return (NULL); } static spa_t * spa_find_by_name(const char *name) { spa_t *spa; STAILQ_FOREACH(spa, &zfs_pools, spa_link) if (strcmp(spa->spa_name, name) == 0) return (spa); return (NULL); } #ifdef BOOT2 static spa_t * spa_get_primary(void) { return (STAILQ_FIRST(&zfs_pools)); } static vdev_t * spa_get_primary_vdev(const spa_t *spa) { vdev_t *vdev; vdev_t *kid; if (spa == NULL) spa = spa_get_primary(); if (spa == NULL) return (NULL); vdev = spa->spa_root_vdev; if (vdev == NULL) return (NULL); for (kid = STAILQ_FIRST(&vdev->v_children); kid != NULL; kid = STAILQ_FIRST(&vdev->v_children)) vdev = kid; return (vdev); } #endif static spa_t * spa_create(uint64_t guid, const char *name) { spa_t *spa; if ((spa = calloc(1, sizeof(spa_t))) == NULL) return (NULL); if ((spa->spa_name = strdup(name)) == NULL) { free(spa); return (NULL); } spa->spa_guid = guid; spa->spa_root_vdev = vdev_create(guid, NULL); if (spa->spa_root_vdev == NULL) { free(spa->spa_name); free(spa); return (NULL); } spa->spa_root_vdev->v_name = strdup("root"); STAILQ_INSERT_TAIL(&zfs_pools, spa, spa_link); return (spa); } static const char * state_name(vdev_state_t state) { static const char *names[] = { "UNKNOWN", "CLOSED", "OFFLINE", "REMOVED", "CANT_OPEN", "FAULTED", "DEGRADED", "ONLINE" }; return (names[state]); } #ifdef BOOT2 #define pager_printf printf #else static int pager_printf(const char *fmt, ...) { char line[80]; va_list args; va_start(args, fmt); vsnprintf(line, sizeof(line), fmt, args); va_end(args); return (pager_output(line)); } #endif #define STATUS_FORMAT " %s %s\n" static int print_state(int indent, const char *name, vdev_state_t state) { int i; char buf[512]; buf[0] = 0; for (i = 0; i < indent; i++) strcat(buf, " "); strcat(buf, name); return (pager_printf(STATUS_FORMAT, buf, state_name(state))); } static int vdev_status(vdev_t *vdev, int indent) { vdev_t *kid; int ret; if (vdev->v_islog) { (void) pager_output(" logs\n"); indent++; } ret = print_state(indent, vdev->v_name, vdev->v_state); if (ret != 0) return (ret); STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) { ret = vdev_status(kid, indent + 1); if (ret != 0) return (ret); } return (ret); } static int spa_status(spa_t *spa) { static char bootfs[ZFS_MAXNAMELEN]; uint64_t rootid; vdev_list_t *vlist; vdev_t *vdev; int good_kids, bad_kids, degraded_kids, ret; vdev_state_t state; ret = pager_printf(" pool: %s\n", spa->spa_name); if (ret != 0) return (ret); if (zfs_get_root(spa, &rootid) == 0 && zfs_rlookup(spa, rootid, bootfs) == 0) { if (bootfs[0] == '\0') ret = pager_printf("bootfs: %s\n", spa->spa_name); else ret = pager_printf("bootfs: %s/%s\n", spa->spa_name, bootfs); if (ret != 0) return (ret); } ret = pager_printf("config:\n\n"); if (ret != 0) return (ret); ret = pager_printf(STATUS_FORMAT, "NAME", "STATE"); if (ret != 0) return (ret); good_kids = 0; degraded_kids = 0; bad_kids = 0; vlist = &spa->spa_root_vdev->v_children; STAILQ_FOREACH(vdev, vlist, v_childlink) { if (vdev->v_state == VDEV_STATE_HEALTHY) good_kids++; else if (vdev->v_state == VDEV_STATE_DEGRADED) degraded_kids++; else bad_kids++; } state = VDEV_STATE_CLOSED; if (good_kids > 0 && (degraded_kids + bad_kids) == 0) state = VDEV_STATE_HEALTHY; else if ((good_kids + degraded_kids) > 0) state = VDEV_STATE_DEGRADED; ret = print_state(0, spa->spa_name, state); if (ret != 0) return (ret); STAILQ_FOREACH(vdev, vlist, v_childlink) { ret = vdev_status(vdev, 1); if (ret != 0) return (ret); } return (ret); } static int spa_all_status(void) { spa_t *spa; int first = 1, ret = 0; STAILQ_FOREACH(spa, &zfs_pools, spa_link) { if (!first) { ret = pager_printf("\n"); if (ret != 0) return (ret); } first = 0; ret = spa_status(spa); if (ret != 0) return (ret); } return (ret); } static uint64_t vdev_label_offset(uint64_t psize, int l, uint64_t offset) { uint64_t label_offset; if (l < VDEV_LABELS / 2) label_offset = 0; else label_offset = psize - VDEV_LABELS * sizeof (vdev_label_t); return (offset + l * sizeof (vdev_label_t) + label_offset); } static int vdev_uberblock_compare(const uberblock_t *ub1, const uberblock_t *ub2) { unsigned int seq1 = 0; unsigned int seq2 = 0; int cmp = AVL_CMP(ub1->ub_txg, ub2->ub_txg); if (cmp != 0) return (cmp); cmp = AVL_CMP(ub1->ub_timestamp, ub2->ub_timestamp); if (cmp != 0) return (cmp); if (MMP_VALID(ub1) && MMP_SEQ_VALID(ub1)) seq1 = MMP_SEQ(ub1); if (MMP_VALID(ub2) && MMP_SEQ_VALID(ub2)) seq2 = MMP_SEQ(ub2); return (AVL_CMP(seq1, seq2)); } static int uberblock_verify(uberblock_t *ub) { if (ub->ub_magic == BSWAP_64((uint64_t)UBERBLOCK_MAGIC)) { byteswap_uint64_array(ub, sizeof (uberblock_t)); } if (ub->ub_magic != UBERBLOCK_MAGIC || !SPA_VERSION_IS_SUPPORTED(ub->ub_version)) return (EINVAL); return (0); } static int vdev_label_read(vdev_t *vd, int l, void *buf, uint64_t offset, size_t size) { blkptr_t bp; off_t off; off = vdev_label_offset(vd->v_psize, l, offset); BP_ZERO(&bp); BP_SET_LSIZE(&bp, size); BP_SET_PSIZE(&bp, size); BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL); BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF); DVA_SET_OFFSET(BP_IDENTITY(&bp), off); ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0); return (vdev_read_phys(vd, &bp, buf, off, size)); } static unsigned char * vdev_label_read_config(vdev_t *vd, uint64_t txg) { vdev_phys_t *label; uint64_t best_txg = 0; uint64_t label_txg = 0; uint64_t asize; unsigned char *nvl; size_t nvl_size; int error; label = malloc(sizeof (vdev_phys_t)); if (label == NULL) return (NULL); nvl_size = VDEV_PHYS_SIZE - sizeof (zio_eck_t) - 4; nvl = malloc(nvl_size); if (nvl == NULL) goto done; for (int l = 0; l < VDEV_LABELS; l++) { const unsigned char *nvlist; if (vdev_label_read(vd, l, label, offsetof(vdev_label_t, vl_vdev_phys), sizeof (vdev_phys_t))) continue; if (label->vp_nvlist[0] != NV_ENCODE_XDR) continue; nvlist = (const unsigned char *) label->vp_nvlist + 4; error = nvlist_find(nvlist, ZPOOL_CONFIG_POOL_TXG, DATA_TYPE_UINT64, NULL, &label_txg); if (error != 0 || label_txg == 0) { memcpy(nvl, nvlist, nvl_size); goto done; } if (label_txg <= txg && label_txg > best_txg) { best_txg = label_txg; memcpy(nvl, nvlist, nvl_size); /* * Use asize from pool config. We need this * because we can get bad value from BIOS. */ if (nvlist_find(nvlist, ZPOOL_CONFIG_ASIZE, DATA_TYPE_UINT64, NULL, &asize) == 0) { vd->v_psize = asize + VDEV_LABEL_START_SIZE + VDEV_LABEL_END_SIZE; } } } if (best_txg == 0) { free(nvl); nvl = NULL; } done: free(label); return (nvl); } static void vdev_uberblock_load(vdev_t *vd, uberblock_t *ub) { uberblock_t *buf; buf = malloc(VDEV_UBERBLOCK_SIZE(vd)); if (buf == NULL) return; for (int l = 0; l < VDEV_LABELS; l++) { for (int n = 0; n < VDEV_UBERBLOCK_COUNT(vd); n++) { if (vdev_label_read(vd, l, buf, VDEV_UBERBLOCK_OFFSET(vd, n), VDEV_UBERBLOCK_SIZE(vd))) continue; if (uberblock_verify(buf) != 0) continue; if (vdev_uberblock_compare(buf, ub) > 0) *ub = *buf; } } free(buf); } static int vdev_probe(vdev_phys_read_t *_read, void *read_priv, spa_t **spap) { vdev_t vtmp; spa_t *spa; vdev_t *vdev; unsigned char *nvlist; uint64_t val; uint64_t guid, vdev_children; uint64_t pool_txg, pool_guid; const char *pool_name; const unsigned char *features; int rc; /* * Load the vdev label and figure out which * uberblock is most current. */ memset(&vtmp, 0, sizeof(vtmp)); vtmp.v_phys_read = _read; vtmp.v_read_priv = read_priv; vtmp.v_psize = P2ALIGN(ldi_get_size(read_priv), (uint64_t)sizeof (vdev_label_t)); /* Test for minimum device size. */ if (vtmp.v_psize < SPA_MINDEVSIZE) return (EIO); nvlist = vdev_label_read_config(&vtmp, UINT64_MAX); if (nvlist == NULL) return (EIO); if (nvlist_find(nvlist, ZPOOL_CONFIG_VERSION, DATA_TYPE_UINT64, NULL, &val) != 0) { free(nvlist); return (EIO); } if (!SPA_VERSION_IS_SUPPORTED(val)) { printf("ZFS: unsupported ZFS version %u (should be %u)\n", (unsigned)val, (unsigned)SPA_VERSION); free(nvlist); return (EIO); } /* Check ZFS features for read */ if (nvlist_find(nvlist, ZPOOL_CONFIG_FEATURES_FOR_READ, DATA_TYPE_NVLIST, NULL, &features) == 0 && nvlist_check_features_for_read(features) != 0) { free(nvlist); return (EIO); } if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_STATE, DATA_TYPE_UINT64, NULL, &val) != 0) { free(nvlist); return (EIO); } if (val == POOL_STATE_DESTROYED) { /* We don't boot only from destroyed pools. */ free(nvlist); return (EIO); } if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_TXG, DATA_TYPE_UINT64, NULL, &pool_txg) != 0 || nvlist_find(nvlist, ZPOOL_CONFIG_POOL_GUID, DATA_TYPE_UINT64, NULL, &pool_guid) != 0 || nvlist_find(nvlist, ZPOOL_CONFIG_POOL_NAME, DATA_TYPE_STRING, NULL, &pool_name) != 0) { /* * Cache and spare devices end up here - just ignore * them. */ free(nvlist); return (EIO); } /* * Create the pool if this is the first time we've seen it. */ spa = spa_find_by_guid(pool_guid); if (spa == NULL) { nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_CHILDREN, DATA_TYPE_UINT64, NULL, &vdev_children); spa = spa_create(pool_guid, pool_name); if (spa == NULL) { free(nvlist); return (ENOMEM); } spa->spa_root_vdev->v_nchildren = vdev_children; } if (pool_txg > spa->spa_txg) spa->spa_txg = pool_txg; /* * Get the vdev tree and create our in-core copy of it. * If we already have a vdev with this guid, this must * be some kind of alias (overlapping slices, dangerously dedicated * disks etc). */ if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64, NULL, &guid) != 0) { free(nvlist); return (EIO); } vdev = vdev_find(guid); /* Has this vdev already been inited? */ if (vdev && vdev->v_phys_read) { free(nvlist); return (EIO); } rc = vdev_init_from_label(spa, nvlist); free(nvlist); if (rc != 0) return (rc); /* * We should already have created an incomplete vdev for this * vdev. Find it and initialise it with our read proc. */ vdev = vdev_find(guid); if (vdev != NULL) { vdev->v_phys_read = _read; vdev->v_read_priv = read_priv; vdev->v_psize = vtmp.v_psize; /* * If no other state is set, mark vdev healthy. */ if (vdev->v_state == VDEV_STATE_UNKNOWN) vdev->v_state = VDEV_STATE_HEALTHY; } else { printf("ZFS: inconsistent nvlist contents\n"); return (EIO); } if (vdev->v_islog) spa->spa_with_log = vdev->v_islog; /* * Re-evaluate top-level vdev state. */ vdev_set_state(vdev->v_top); /* * Ok, we are happy with the pool so far. Lets find * the best uberblock and then we can actually access * the contents of the pool. */ vdev_uberblock_load(vdev, &spa->spa_uberblock); if (spap != NULL) *spap = spa; return (0); } static int ilog2(int n) { int v; for (v = 0; v < 32; v++) if (n == (1 << v)) return (v); return (-1); } static int zio_read_gang(const spa_t *spa, const blkptr_t *bp, void *buf) { blkptr_t gbh_bp; zio_gbh_phys_t zio_gb; char *pbuf; int i; /* Artificial BP for gang block header. */ gbh_bp = *bp; BP_SET_PSIZE(&gbh_bp, SPA_GANGBLOCKSIZE); BP_SET_LSIZE(&gbh_bp, SPA_GANGBLOCKSIZE); BP_SET_CHECKSUM(&gbh_bp, ZIO_CHECKSUM_GANG_HEADER); BP_SET_COMPRESS(&gbh_bp, ZIO_COMPRESS_OFF); for (i = 0; i < SPA_DVAS_PER_BP; i++) DVA_SET_GANG(&gbh_bp.blk_dva[i], 0); /* Read gang header block using the artificial BP. */ if (zio_read(spa, &gbh_bp, &zio_gb)) return (EIO); pbuf = buf; for (i = 0; i < SPA_GBH_NBLKPTRS; i++) { blkptr_t *gbp = &zio_gb.zg_blkptr[i]; if (BP_IS_HOLE(gbp)) continue; if (zio_read(spa, gbp, pbuf)) return (EIO); pbuf += BP_GET_PSIZE(gbp); } if (zio_checksum_verify(spa, bp, buf)) return (EIO); return (0); } static int zio_read(const spa_t *spa, const blkptr_t *bp, void *buf) { int cpfunc = BP_GET_COMPRESS(bp); uint64_t align, size; void *pbuf; int i, error; /* * Process data embedded in block pointer */ if (BP_IS_EMBEDDED(bp)) { ASSERT(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA); size = BPE_GET_PSIZE(bp); ASSERT(size <= BPE_PAYLOAD_SIZE); if (cpfunc != ZIO_COMPRESS_OFF) pbuf = zfs_alloc(size); else pbuf = buf; decode_embedded_bp_compressed(bp, pbuf); error = 0; if (cpfunc != ZIO_COMPRESS_OFF) { error = zio_decompress_data(cpfunc, pbuf, size, buf, BP_GET_LSIZE(bp)); zfs_free(pbuf, size); } if (error != 0) printf("ZFS: i/o error - unable to decompress " "block pointer data, error %d\n", error); return (error); } error = EIO; for (i = 0; i < SPA_DVAS_PER_BP; i++) { const dva_t *dva = &bp->blk_dva[i]; vdev_t *vdev; vdev_list_t *vlist; uint64_t vdevid; off_t offset; if (!dva->dva_word[0] && !dva->dva_word[1]) continue; vdevid = DVA_GET_VDEV(dva); offset = DVA_GET_OFFSET(dva); vlist = &spa->spa_root_vdev->v_children; STAILQ_FOREACH(vdev, vlist, v_childlink) { if (vdev->v_id == vdevid) break; } if (!vdev || !vdev->v_read) continue; size = BP_GET_PSIZE(bp); if (vdev->v_read == vdev_raidz_read) { align = 1ULL << vdev->v_ashift; if (P2PHASE(size, align) != 0) size = P2ROUNDUP(size, align); } if (size != BP_GET_PSIZE(bp) || cpfunc != ZIO_COMPRESS_OFF) pbuf = zfs_alloc(size); else pbuf = buf; if (DVA_GET_GANG(dva)) error = zio_read_gang(spa, bp, pbuf); else error = vdev->v_read(vdev, bp, pbuf, offset, size); if (error == 0) { if (cpfunc != ZIO_COMPRESS_OFF) error = zio_decompress_data(cpfunc, pbuf, BP_GET_PSIZE(bp), buf, BP_GET_LSIZE(bp)); else if (size != BP_GET_PSIZE(bp)) bcopy(pbuf, buf, BP_GET_PSIZE(bp)); } if (buf != pbuf) zfs_free(pbuf, size); if (error == 0) break; } if (error != 0) printf("ZFS: i/o error - all block copies unavailable\n"); return (error); } static int dnode_read(const spa_t *spa, const dnode_phys_t *dnode, off_t offset, void *buf, size_t buflen) { int ibshift = dnode->dn_indblkshift - SPA_BLKPTRSHIFT; int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; int nlevels = dnode->dn_nlevels; int i, rc; if (bsize > SPA_MAXBLOCKSIZE) { printf("ZFS: I/O error - blocks larger than %llu are not " "supported\n", SPA_MAXBLOCKSIZE); return (EIO); } /* * Note: bsize may not be a power of two here so we need to do an * actual divide rather than a bitshift. */ while (buflen > 0) { uint64_t bn = offset / bsize; int boff = offset % bsize; int ibn; const blkptr_t *indbp; blkptr_t bp; if (bn > dnode->dn_maxblkid) return (EIO); if (dnode == dnode_cache_obj && bn == dnode_cache_bn) goto cached; indbp = dnode->dn_blkptr; for (i = 0; i < nlevels; i++) { /* * Copy the bp from the indirect array so that * we can re-use the scratch buffer for multi-level * objects. */ ibn = bn >> ((nlevels - i - 1) * ibshift); ibn &= ((1 << ibshift) - 1); bp = indbp[ibn]; if (BP_IS_HOLE(&bp)) { memset(dnode_cache_buf, 0, bsize); break; } rc = zio_read(spa, &bp, dnode_cache_buf); if (rc) return (rc); indbp = (const blkptr_t *) dnode_cache_buf; } dnode_cache_obj = dnode; dnode_cache_bn = bn; cached: /* * The buffer contains our data block. Copy what we * need from it and loop. */ i = bsize - boff; if (i > buflen) i = buflen; memcpy(buf, &dnode_cache_buf[boff], i); buf = ((char *)buf) + i; offset += i; buflen -= i; } return (0); } /* * Lookup a value in a microzap directory. Assumes that the zap * scratch buffer contains the directory contents. */ static int -mzap_lookup(const dnode_phys_t *dnode, const char *name, uint64_t *value) +mzap_lookup(const mzap_phys_t *mz, size_t size, const char *name, + uint64_t *value) { - const mzap_phys_t *mz; const mzap_ent_phys_t *mze; - size_t size; int chunks, i; /* * Microzap objects use exactly one block. Read the whole * thing. */ - size = dnode->dn_datablkszsec * 512; - - mz = (const mzap_phys_t *) zap_scratch; chunks = size / MZAP_ENT_LEN - 1; - for (i = 0; i < chunks; i++) { mze = &mz->mz_chunk[i]; if (strcmp(mze->mze_name, name) == 0) { *value = mze->mze_value; return (0); } } return (ENOENT); } /* * Compare a name with a zap leaf entry. Return non-zero if the name * matches. */ static int fzap_name_equal(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, const char *name) { size_t namelen; const zap_leaf_chunk_t *nc; const char *p; namelen = zc->l_entry.le_name_numints; nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk); p = name; while (namelen > 0) { size_t len; len = namelen; if (len > ZAP_LEAF_ARRAY_BYTES) len = ZAP_LEAF_ARRAY_BYTES; if (memcmp(p, nc->l_array.la_array, len)) return (0); p += len; namelen -= len; nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next); } return (1); } /* * Extract a uint64_t value from a zap leaf entry. */ static uint64_t fzap_leaf_value(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc) { const zap_leaf_chunk_t *vc; int i; uint64_t value; const uint8_t *p; vc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_value_chunk); for (i = 0, value = 0, p = vc->l_array.la_array; i < 8; i++) { value = (value << 8) | p[i]; } return (value); } static void stv(int len, void *addr, uint64_t value) { switch (len) { case 1: *(uint8_t *)addr = value; return; case 2: *(uint16_t *)addr = value; return; case 4: *(uint32_t *)addr = value; return; case 8: *(uint64_t *)addr = value; return; } } /* * Extract a array from a zap leaf entry. */ static void fzap_leaf_array(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, uint64_t integer_size, uint64_t num_integers, void *buf) { uint64_t array_int_len = zc->l_entry.le_value_intlen; uint64_t value = 0; uint64_t *u64 = buf; char *p = buf; int len = MIN(zc->l_entry.le_value_numints, num_integers); int chunk = zc->l_entry.le_value_chunk; int byten = 0; if (integer_size == 8 && len == 1) { *u64 = fzap_leaf_value(zl, zc); return; } while (len > 0) { struct zap_leaf_array *la = &ZAP_LEAF_CHUNK(zl, chunk).l_array; int i; ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(zl)); for (i = 0; i < ZAP_LEAF_ARRAY_BYTES && len > 0; i++) { value = (value << 8) | la->la_array[i]; byten++; if (byten == array_int_len) { stv(integer_size, p, value); byten = 0; len--; if (len == 0) return; p += integer_size; } } chunk = la->la_next; } } static int fzap_check_size(uint64_t integer_size, uint64_t num_integers) { switch (integer_size) { case 1: case 2: case 4: case 8: break; default: return (EINVAL); } if (integer_size * num_integers > ZAP_MAXVALUELEN) return (E2BIG); return (0); } -/* - * Lookup a value in a fatzap directory. Assumes that the zap scratch - * buffer contains the directory header. - */ +static void +zap_leaf_free(zap_leaf_t *leaf) +{ + free(leaf->l_phys); + free(leaf); +} + static int -fzap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name, - uint64_t integer_size, uint64_t num_integers, void *value) +zap_get_leaf_byblk(fat_zap_t *zap, uint64_t blk, zap_leaf_t **lp) { - int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; - zap_phys_t zh = *(zap_phys_t *)zap_scratch; - fat_zap_t z; - uint64_t *ptrtbl; - uint64_t hash; - int rc; + int bs = FZAP_BLOCK_SHIFT(zap); + int err; - if (zh.zap_magic != ZAP_MAGIC) - return (EIO); + *lp = malloc(sizeof(**lp)); + if (*lp == NULL) + return (ENOMEM); - if ((rc = fzap_check_size(integer_size, num_integers)) != 0) - return (rc); + (*lp)->l_bs = bs; + (*lp)->l_phys = malloc(1 << bs); - z.zap_block_shift = ilog2(bsize); - z.zap_phys = (zap_phys_t *)zap_scratch; + if ((*lp)->l_phys == NULL) { + free(*lp); + return (ENOMEM); + } + err = dnode_read(zap->zap_spa, zap->zap_dnode, blk << bs, (*lp)->l_phys, + 1 << bs); + if (err != 0) { + zap_leaf_free(*lp); + } + return (err); +} - /* - * Figure out where the pointer table is and read it in if necessary. - */ - if (zh.zap_ptrtbl.zt_blk) { - rc = dnode_read(spa, dnode, zh.zap_ptrtbl.zt_blk * bsize, - zap_scratch, bsize); - if (rc) - return (rc); - ptrtbl = (uint64_t *)zap_scratch; +static int +zap_table_load(fat_zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, + uint64_t *valp) +{ + int bs = FZAP_BLOCK_SHIFT(zap); + uint64_t blk = idx >> (bs - 3); + uint64_t off = idx & ((1 << (bs - 3)) - 1); + uint64_t *buf; + int rc; + + buf = malloc(1 << zap->zap_block_shift); + if (buf == NULL) + return (ENOMEM); + rc = dnode_read(zap->zap_spa, zap->zap_dnode, (tbl->zt_blk + blk) << bs, + buf, 1 << zap->zap_block_shift); + if (rc == 0) + *valp = buf[off]; + free(buf); + return (rc); +} + +static int +zap_idx_to_blk(fat_zap_t *zap, uint64_t idx, uint64_t *valp) +{ + if (zap->zap_phys->zap_ptrtbl.zt_numblks == 0) { + *valp = ZAP_EMBEDDED_PTRTBL_ENT(zap, idx); + return (0); } else { - ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0); + return (zap_table_load(zap, &zap->zap_phys->zap_ptrtbl, + idx, valp)); } +} - hash = zap_hash(zh.zap_salt, name); +#define ZAP_HASH_IDX(hash, n) (((n) == 0) ? 0 : ((hash) >> (64 - (n)))) +static int +zap_deref_leaf(fat_zap_t *zap, uint64_t h, zap_leaf_t **lp) +{ + uint64_t idx, blk; + int err; - zap_leaf_t zl; - zl.l_bs = z.zap_block_shift; + idx = ZAP_HASH_IDX(h, zap->zap_phys->zap_ptrtbl.zt_shift); + err = zap_idx_to_blk(zap, idx, &blk); + if (err != 0) + return (err); + return (zap_get_leaf_byblk(zap, blk, lp)); +} - off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs; - zap_leaf_chunk_t *zc; +#define CHAIN_END 0xffff /* end of the chunk chain */ +#define LEAF_HASH(l, h) \ + ((ZAP_LEAF_HASH_NUMENTRIES(l)-1) & \ + ((h) >> \ + (64 - ZAP_LEAF_HASH_SHIFT(l) - (l)->l_phys->l_hdr.lh_prefix_len))) +#define LEAF_HASH_ENTPTR(l, h) (&(l)->l_phys->l_hash[LEAF_HASH(l, h)]) - rc = dnode_read(spa, dnode, off, zap_scratch, bsize); - if (rc) - return (rc); +static int +zap_leaf_lookup(zap_leaf_t *zl, uint64_t hash, const char *name, + uint64_t integer_size, uint64_t num_integers, void *value) +{ + int rc; + uint16_t *chunkp; + struct zap_leaf_entry *le; - zl.l_phys = (zap_leaf_phys_t *)zap_scratch; - /* * Make sure this chunk matches our hash. */ - if (zl.l_phys->l_hdr.lh_prefix_len > 0 && - zl.l_phys->l_hdr.lh_prefix != - hash >> (64 - zl.l_phys->l_hdr.lh_prefix_len)) - return (ENOENT); + if (zl->l_phys->l_hdr.lh_prefix_len > 0 && + zl->l_phys->l_hdr.lh_prefix != + hash >> (64 - zl->l_phys->l_hdr.lh_prefix_len)) + return (EIO); - /* - * Hash within the chunk to find our entry. - */ - int shift = (64 - ZAP_LEAF_HASH_SHIFT(&zl) - - zl.l_phys->l_hdr.lh_prefix_len); - int h = (hash >> shift) & ((1 << ZAP_LEAF_HASH_SHIFT(&zl)) - 1); - h = zl.l_phys->l_hash[h]; - if (h == 0xffff) - return (ENOENT); - zc = &ZAP_LEAF_CHUNK(&zl, h); - while (zc->l_entry.le_hash != hash) { - if (zc->l_entry.le_next == 0xffff) - return (ENOENT); - zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next); + rc = ENOENT; + for (chunkp = LEAF_HASH_ENTPTR(zl, hash); + *chunkp != CHAIN_END; chunkp = &le->le_next) { + zap_leaf_chunk_t *zc; + uint16_t chunk = *chunkp; + + le = ZAP_LEAF_ENTRY(zl, chunk); + if (le->le_hash != hash) + continue; + zc = &ZAP_LEAF_CHUNK(zl, chunk); + if (fzap_name_equal(zl, zc, name)) { + if (zc->l_entry.le_value_intlen > integer_size) { + rc = EINVAL; + } else { + fzap_leaf_array(zl, zc, integer_size, + num_integers, value); + rc = 0; + } + break; + } } - if (fzap_name_equal(&zl, zc, name)) { - if (zc->l_entry.le_value_intlen * zc->l_entry.le_value_numints > - integer_size * num_integers) - return (E2BIG); - fzap_leaf_array(&zl, zc, integer_size, num_integers, value); - return (0); - } + return (rc); +} - return (ENOENT); +/* + * Lookup a value in a fatzap directory. + */ +static int +fzap_lookup(const spa_t *spa, const dnode_phys_t *dnode, zap_phys_t *zh, + const char *name, uint64_t integer_size, uint64_t num_integers, + void *value) +{ + int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; + fat_zap_t z; + zap_leaf_t *zl; + uint64_t hash; + int rc; + + if (zh->zap_magic != ZAP_MAGIC) + return (EIO); + + if ((rc = fzap_check_size(integer_size, num_integers)) != 0) + return (rc); + + z.zap_block_shift = ilog2(bsize); + z.zap_phys = zh; + z.zap_spa = spa; + z.zap_dnode = dnode; + + hash = zap_hash(zh->zap_salt, name); + rc = zap_deref_leaf(&z, hash, &zl); + if (rc != 0) + return (rc); + + rc = zap_leaf_lookup(zl, hash, name, integer_size, num_integers, value); + + zap_leaf_free(zl); + return (rc); } /* * Lookup a name in a zap object and return its value as a uint64_t. */ static int zap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t integer_size, uint64_t num_integers, void *value) { int rc; - uint64_t zap_type; + zap_phys_t *zap; size_t size = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; - rc = dnode_read(spa, dnode, 0, zap_scratch, size); + zap = malloc(size); + if (zap == NULL) + return (ENOMEM); + + rc = dnode_read(spa, dnode, 0, zap, size); if (rc) - return (rc); + goto done; - zap_type = *(uint64_t *)zap_scratch; - if (zap_type == ZBT_MICRO) - return (mzap_lookup(dnode, name, value)); - else if (zap_type == ZBT_HEADER) { - return (fzap_lookup(spa, dnode, name, integer_size, - num_integers, value)); + switch (zap->zap_block_type) { + case ZBT_MICRO: + rc = mzap_lookup((const mzap_phys_t *)zap, size, name, value); + break; + case ZBT_HEADER: + rc = fzap_lookup(spa, dnode, zap, name, integer_size, + num_integers, value); + break; + default: + printf("ZFS: invalid zap_type=%" PRIx64 "\n", + zap->zap_block_type); + rc = EIO; } - printf("ZFS: invalid zap_type=%d\n", (int)zap_type); - return (EIO); +done: + free(zap); + return (rc); } /* - * List a microzap directory. Assumes that the zap scratch buffer contains - * the directory contents. + * List a microzap directory. */ static int -mzap_list(const dnode_phys_t *dnode, int (*callback)(const char *, uint64_t)) +mzap_list(const mzap_phys_t *mz, size_t size, + int (*callback)(const char *, uint64_t)) { - const mzap_phys_t *mz; const mzap_ent_phys_t *mze; - size_t size; int chunks, i, rc; /* * Microzap objects use exactly one block. Read the whole * thing. */ - size = dnode->dn_datablkszsec * 512; - mz = (const mzap_phys_t *) zap_scratch; + rc = 0; chunks = size / MZAP_ENT_LEN - 1; - for (i = 0; i < chunks; i++) { mze = &mz->mz_chunk[i]; if (mze->mze_name[0]) { rc = callback(mze->mze_name, mze->mze_value); if (rc != 0) - return (rc); + break; } } - return (0); + return (rc); } /* - * List a fatzap directory. Assumes that the zap scratch buffer contains - * the directory header. + * List a fatzap directory. */ static int -fzap_list(const spa_t *spa, const dnode_phys_t *dnode, +fzap_list(const spa_t *spa, const dnode_phys_t *dnode, zap_phys_t *zh, int (*callback)(const char *, uint64_t)) { int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; - zap_phys_t zh = *(zap_phys_t *)zap_scratch; fat_zap_t z; - int i, j, rc; + uint64_t i; + int j, rc; - if (zh.zap_magic != ZAP_MAGIC) + if (zh->zap_magic != ZAP_MAGIC) return (EIO); z.zap_block_shift = ilog2(bsize); - z.zap_phys = (zap_phys_t *)zap_scratch; + z.zap_phys = zh; /* * This assumes that the leaf blocks start at block 1. The * documentation isn't exactly clear on this. */ zap_leaf_t zl; zl.l_bs = z.zap_block_shift; - for (i = 0; i < zh.zap_num_leafs; i++) { + zl.l_phys = malloc(bsize); + if (zl.l_phys == NULL) + return (ENOMEM); + + for (i = 0; i < zh->zap_num_leafs; i++) { off_t off = ((off_t)(i + 1)) << zl.l_bs; char name[256], *p; uint64_t value; - if (dnode_read(spa, dnode, off, zap_scratch, bsize)) + if (dnode_read(spa, dnode, off, zl.l_phys, bsize)) { + free(zl.l_phys); return (EIO); + } - zl.l_phys = (zap_leaf_phys_t *)zap_scratch; - for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) { zap_leaf_chunk_t *zc, *nc; int namelen; zc = &ZAP_LEAF_CHUNK(&zl, j); if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY) continue; namelen = zc->l_entry.le_name_numints; if (namelen > sizeof(name)) namelen = sizeof(name); /* * Paste the name back together. */ nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk); p = name; while (namelen > 0) { int len; len = namelen; if (len > ZAP_LEAF_ARRAY_BYTES) len = ZAP_LEAF_ARRAY_BYTES; memcpy(p, nc->l_array.la_array, len); p += len; namelen -= len; nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next); } /* * Assume the first eight bytes of the value are * a uint64_t. */ value = fzap_leaf_value(&zl, zc); /* printf("%s 0x%jx\n", name, (uintmax_t)value); */ rc = callback((const char *)name, value); - if (rc != 0) + if (rc != 0) { + free(zl.l_phys); return (rc); + } } } + free(zl.l_phys); return (0); } static int zfs_printf(const char *name, uint64_t value __unused) { printf("%s\n", name); return (0); } /* * List a zap directory. */ static int zap_list(const spa_t *spa, const dnode_phys_t *dnode) { - uint64_t zap_type; + zap_phys_t *zap; size_t size = dnode->dn_datablkszsec * 512; + int rc; - if (dnode_read(spa, dnode, 0, zap_scratch, size)) - return (EIO); + zap = malloc(size); + if (zap == NULL) + return (ENOMEM); - zap_type = *(uint64_t *)zap_scratch; - if (zap_type == ZBT_MICRO) - return (mzap_list(dnode, zfs_printf)); - else - return (fzap_list(spa, dnode, zfs_printf)); + rc = dnode_read(spa, dnode, 0, zap, size); + if (rc == 0) { + if (zap->zap_block_type == ZBT_MICRO) + rc = mzap_list((const mzap_phys_t *)zap, size, + zfs_printf); + else + rc = fzap_list(spa, dnode, zap, zfs_printf); + } + free(zap); + return (rc); } static int objset_get_dnode(const spa_t *spa, const objset_phys_t *os, uint64_t objnum, dnode_phys_t *dnode) { off_t offset; offset = objnum * sizeof(dnode_phys_t); return dnode_read(spa, &os->os_meta_dnode, offset, dnode, sizeof(dnode_phys_t)); } +/* + * Lookup a name in a microzap directory. + */ static int -mzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, - uint64_t value) +mzap_rlookup(const mzap_phys_t *mz, size_t size, char *name, uint64_t value) { - const mzap_phys_t *mz; const mzap_ent_phys_t *mze; - size_t size; int chunks, i; /* * Microzap objects use exactly one block. Read the whole * thing. */ - size = dnode->dn_datablkszsec * 512; - - mz = (const mzap_phys_t *)zap_scratch; chunks = size / MZAP_ENT_LEN - 1; - for (i = 0; i < chunks; i++) { mze = &mz->mz_chunk[i]; if (value == mze->mze_value) { strcpy(name, mze->mze_name); return (0); } } return (ENOENT); } static void fzap_name_copy(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, char *name) { size_t namelen; const zap_leaf_chunk_t *nc; char *p; namelen = zc->l_entry.le_name_numints; nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk); p = name; while (namelen > 0) { size_t len; len = namelen; if (len > ZAP_LEAF_ARRAY_BYTES) len = ZAP_LEAF_ARRAY_BYTES; memcpy(p, nc->l_array.la_array, len); p += len; namelen -= len; nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next); } *p = '\0'; } static int -fzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, - uint64_t value) +fzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, zap_phys_t *zh, + char *name, uint64_t value) { int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; - zap_phys_t zh = *(zap_phys_t *)zap_scratch; fat_zap_t z; - int i, j; + uint64_t i; + int j, rc; - if (zh.zap_magic != ZAP_MAGIC) + if (zh->zap_magic != ZAP_MAGIC) return (EIO); z.zap_block_shift = ilog2(bsize); - z.zap_phys = (zap_phys_t *)zap_scratch; + z.zap_phys = zh; /* * This assumes that the leaf blocks start at block 1. The * documentation isn't exactly clear on this. */ zap_leaf_t zl; zl.l_bs = z.zap_block_shift; - for (i = 0; i < zh.zap_num_leafs; i++) { + zl.l_phys = malloc(bsize); + if (zl.l_phys == NULL) + return (ENOMEM); + + for (i = 0; i < zh->zap_num_leafs; i++) { off_t off = ((off_t)(i + 1)) << zl.l_bs; - if (dnode_read(spa, dnode, off, zap_scratch, bsize)) - return (EIO); + rc = dnode_read(spa, dnode, off, zl.l_phys, bsize); + if (rc != 0) + goto done; - zl.l_phys = (zap_leaf_phys_t *)zap_scratch; - for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) { zap_leaf_chunk_t *zc; zc = &ZAP_LEAF_CHUNK(&zl, j); if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY) continue; if (zc->l_entry.le_value_intlen != 8 || zc->l_entry.le_value_numints != 1) continue; if (fzap_leaf_value(&zl, zc) == value) { fzap_name_copy(&zl, zc, name); - return (0); + goto done; } } } - return (ENOENT); + rc = ENOENT; +done: + free(zl.l_phys); + return (rc); } static int zap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value) { - int rc; - uint64_t zap_type; + zap_phys_t *zap; size_t size = dnode->dn_datablkszsec * 512; + int rc; - rc = dnode_read(spa, dnode, 0, zap_scratch, size); - if (rc) - return (rc); + zap = malloc(size); + if (zap == NULL) + return (ENOMEM); - zap_type = *(uint64_t *)zap_scratch; - if (zap_type == ZBT_MICRO) - return (mzap_rlookup(spa, dnode, name, value)); - else - return (fzap_rlookup(spa, dnode, name, value)); + rc = dnode_read(spa, dnode, 0, zap, size); + if (rc == 0) { + if (zap->zap_block_type == ZBT_MICRO) + rc = mzap_rlookup((const mzap_phys_t *)zap, size, + name, value); + else + rc = fzap_rlookup(spa, dnode, zap, name, value); + } + free(zap); + return (rc); } static int zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result) { char name[256]; char component[256]; uint64_t dir_obj, parent_obj, child_dir_zapobj; dnode_phys_t child_dir_zap, dataset, dir, parent; dsl_dir_phys_t *dd; dsl_dataset_phys_t *ds; char *p; int len; p = &name[sizeof(name) - 1]; *p = '\0'; if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) { printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); return (EIO); } ds = (dsl_dataset_phys_t *)&dataset.dn_bonus; dir_obj = ds->ds_dir_obj; for (;;) { if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir) != 0) return (EIO); dd = (dsl_dir_phys_t *)&dir.dn_bonus; /* Actual loop condition. */ parent_obj = dd->dd_parent_obj; if (parent_obj == 0) break; if (objset_get_dnode(spa, &spa->spa_mos, parent_obj, &parent) != 0) return (EIO); dd = (dsl_dir_phys_t *)&parent.dn_bonus; child_dir_zapobj = dd->dd_child_dir_zapobj; if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) return (EIO); if (zap_rlookup(spa, &child_dir_zap, component, dir_obj) != 0) return (EIO); len = strlen(component); p -= len; memcpy(p, component, len); --p; *p = '/'; /* Actual loop iteration. */ dir_obj = parent_obj; } if (*p != '\0') ++p; strcpy(result, p); return (0); } static int zfs_lookup_dataset(const spa_t *spa, const char *name, uint64_t *objnum) { char element[256]; uint64_t dir_obj, child_dir_zapobj; dnode_phys_t child_dir_zap, dir; dsl_dir_phys_t *dd; const char *p, *q; if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) return (EIO); if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, sizeof (dir_obj), 1, &dir_obj)) return (EIO); p = name; for (;;) { if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir)) return (EIO); dd = (dsl_dir_phys_t *)&dir.dn_bonus; while (*p == '/') p++; /* Actual loop condition #1. */ if (*p == '\0') break; q = strchr(p, '/'); if (q) { memcpy(element, p, q - p); element[q - p] = '\0'; p = q + 1; } else { strcpy(element, p); p += strlen(p); } child_dir_zapobj = dd->dd_child_dir_zapobj; if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) return (EIO); /* Actual loop condition #2. */ if (zap_lookup(spa, &child_dir_zap, element, sizeof (dir_obj), 1, &dir_obj) != 0) return (ENOENT); } *objnum = dd->dd_head_dataset_obj; return (0); } #ifndef BOOT2 static int zfs_list_dataset(const spa_t *spa, uint64_t objnum/*, int pos, char *entry*/) { uint64_t dir_obj, child_dir_zapobj; dnode_phys_t child_dir_zap, dir, dataset; dsl_dataset_phys_t *ds; dsl_dir_phys_t *dd; if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) { printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); return (EIO); } ds = (dsl_dataset_phys_t *)&dataset.dn_bonus; dir_obj = ds->ds_dir_obj; if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir)) { printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj); return (EIO); } dd = (dsl_dir_phys_t *)&dir.dn_bonus; child_dir_zapobj = dd->dd_child_dir_zapobj; if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) { printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj); return (EIO); } return (zap_list(spa, &child_dir_zap) != 0); } int zfs_callback_dataset(const spa_t *spa, uint64_t objnum, int (*callback)(const char *, uint64_t)) { - uint64_t dir_obj, child_dir_zapobj, zap_type; + uint64_t dir_obj, child_dir_zapobj; dnode_phys_t child_dir_zap, dir, dataset; dsl_dataset_phys_t *ds; dsl_dir_phys_t *dd; + zap_phys_t *zap; + size_t size; int err; err = objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset); if (err != 0) { printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); return (err); } ds = (dsl_dataset_phys_t *)&dataset.dn_bonus; dir_obj = ds->ds_dir_obj; err = objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir); if (err != 0) { printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj); return (err); } dd = (dsl_dir_phys_t *)&dir.dn_bonus; child_dir_zapobj = dd->dd_child_dir_zapobj; err = objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap); if (err != 0) { printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj); return (err); } - err = dnode_read(spa, &child_dir_zap, 0, zap_scratch, - child_dir_zap.dn_datablkszsec * 512); - if (err != 0) - return (err); + size = child_dir_zap.dn_datablkszsec * 512; + zap = malloc(size); + if (zap != NULL) { + err = dnode_read(spa, &child_dir_zap, 0, zap, size); + if (err != 0) + goto done; - zap_type = *(uint64_t *)zap_scratch; - if (zap_type == ZBT_MICRO) - return (mzap_list(&child_dir_zap, callback)); - else - return (fzap_list(spa, &child_dir_zap, callback)); + if (zap->zap_block_type == ZBT_MICRO) + err = mzap_list((const mzap_phys_t *)zap, size, + callback); + else + err = fzap_list(spa, &child_dir_zap, zap, callback); + } +done: + free(zap); + return (err); } #endif /* * Find the object set given the object number of its dataset object * and return its details in *objset */ static int zfs_mount_dataset(const spa_t *spa, uint64_t objnum, objset_phys_t *objset) { dnode_phys_t dataset; dsl_dataset_phys_t *ds; if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) { printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); return (EIO); } ds = (dsl_dataset_phys_t *)&dataset.dn_bonus; if (zio_read(spa, &ds->ds_bp, objset)) { printf("ZFS: can't read object set for dataset %ju\n", (uintmax_t)objnum); return (EIO); } return (0); } /* * Find the object set pointed to by the BOOTFS property or the root * dataset if there is none and return its details in *objset */ static int zfs_get_root(const spa_t *spa, uint64_t *objid) { dnode_phys_t dir, propdir; uint64_t props, bootfs, root; *objid = 0; /* * Start with the MOS directory object. */ if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) { printf("ZFS: can't read MOS object directory\n"); return (EIO); } /* * Lookup the pool_props and see if we can find a bootfs. */ if (zap_lookup(spa, &dir, DMU_POOL_PROPS, sizeof(props), 1, &props) == 0 && objset_get_dnode(spa, &spa->spa_mos, props, &propdir) == 0 && zap_lookup(spa, &propdir, "bootfs", sizeof(bootfs), 1, &bootfs) == 0 && bootfs != 0) { *objid = bootfs; return (0); } /* * Lookup the root dataset directory */ if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, sizeof(root), 1, &root) || objset_get_dnode(spa, &spa->spa_mos, root, &dir)) { printf("ZFS: can't find root dsl_dir\n"); return (EIO); } /* * Use the information from the dataset directory's bonus buffer * to find the dataset object and from that the object set itself. */ dsl_dir_phys_t *dd = (dsl_dir_phys_t *)&dir.dn_bonus; *objid = dd->dd_head_dataset_obj; return (0); } static int zfs_mount(const spa_t *spa, uint64_t rootobj, struct zfsmount *mount) { mount->spa = spa; /* * Find the root object set if not explicitly provided */ if (rootobj == 0 && zfs_get_root(spa, &rootobj)) { printf("ZFS: can't find root filesystem\n"); return (EIO); } if (zfs_mount_dataset(spa, rootobj, &mount->objset)) { printf("ZFS: can't open root filesystem\n"); return (EIO); } mount->rootobj = rootobj; return (0); } /* * callback function for feature name checks. */ static int check_feature(const char *name, uint64_t value) { int i; if (value == 0) return (0); if (name[0] == '\0') return (0); for (i = 0; features_for_read[i] != NULL; i++) { if (strcmp(name, features_for_read[i]) == 0) return (0); } printf("ZFS: unsupported feature: %s\n", name); return (EIO); } /* * Checks whether the MOS features that are active are supported. */ static int check_mos_features(const spa_t *spa) { dnode_phys_t dir; - uint64_t objnum, zap_type; + zap_phys_t *zap; + uint64_t objnum; size_t size; int rc; if ((rc = objset_get_dnode(spa, &spa->spa_mos, DMU_OT_OBJECT_DIRECTORY, &dir)) != 0) return (rc); if ((rc = zap_lookup(spa, &dir, DMU_POOL_FEATURES_FOR_READ, sizeof (objnum), 1, &objnum)) != 0) { /* * It is older pool without features. As we have already * tested the label, just return without raising the error. */ return (0); } if ((rc = objset_get_dnode(spa, &spa->spa_mos, objnum, &dir)) != 0) return (rc); if (dir.dn_type != DMU_OTN_ZAP_METADATA) return (EIO); size = dir.dn_datablkszsec * 512; - if (dnode_read(spa, &dir, 0, zap_scratch, size)) + zap = malloc(size); + if (zap == NULL) + return (ENOMEM); + + if (dnode_read(spa, &dir, 0, zap, size)) { + free(zap); return (EIO); + } - zap_type = *(uint64_t *)zap_scratch; - if (zap_type == ZBT_MICRO) - rc = mzap_list(&dir, check_feature); + if (zap->zap_block_type == ZBT_MICRO) + rc = mzap_list((const mzap_phys_t *)zap, size, check_feature); else - rc = fzap_list(spa, &dir, check_feature); + rc = fzap_list(spa, &dir, zap, check_feature); + free(zap); return (rc); } static int load_nvlist(spa_t *spa, uint64_t obj, unsigned char **value) { dnode_phys_t dir; size_t size; int rc; unsigned char *nv; *value = NULL; if ((rc = objset_get_dnode(spa, &spa->spa_mos, obj, &dir)) != 0) return (rc); if (dir.dn_type != DMU_OT_PACKED_NVLIST && dir.dn_bonustype != DMU_OT_PACKED_NVLIST_SIZE) { return (EIO); } if (dir.dn_bonuslen != sizeof (uint64_t)) return (EIO); size = *(uint64_t *)DN_BONUS(&dir); nv = malloc(size); if (nv == NULL) return (ENOMEM); rc = dnode_read(spa, &dir, 0, nv, size); if (rc != 0) { free(nv); nv = NULL; return (rc); } *value = nv; return (rc); } static int zfs_spa_init(spa_t *spa) { dnode_phys_t dir; uint64_t config_object; unsigned char *nvlist; int rc; if (zio_read(spa, &spa->spa_uberblock.ub_rootbp, &spa->spa_mos)) { printf("ZFS: can't read MOS of pool %s\n", spa->spa_name); return (EIO); } if (spa->spa_mos.os_type != DMU_OST_META) { printf("ZFS: corrupted MOS of pool %s\n", spa->spa_name); return (EIO); } if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) { printf("ZFS: failed to read pool %s directory object\n", spa->spa_name); return (EIO); } /* this is allowed to fail, older pools do not have salt */ rc = zap_lookup(spa, &dir, DMU_POOL_CHECKSUM_SALT, 1, sizeof (spa->spa_cksum_salt.zcs_bytes), spa->spa_cksum_salt.zcs_bytes); rc = check_mos_features(spa); if (rc != 0) { printf("ZFS: pool %s is not supported\n", spa->spa_name); return (rc); } rc = zap_lookup(spa, &dir, DMU_POOL_CONFIG, sizeof (config_object), 1, &config_object); if (rc != 0) { printf("ZFS: can not read MOS %s\n", DMU_POOL_CONFIG); return (EIO); } rc = load_nvlist(spa, config_object, &nvlist); if (rc != 0) return (rc); /* * Update vdevs from MOS config. Note, we do skip encoding bytes * here. See also vdev_label_read_config(). */ rc = vdev_init_from_nvlist(spa, nvlist + 4); free(nvlist); return (rc); } static int zfs_dnode_stat(const spa_t *spa, dnode_phys_t *dn, struct stat *sb) { if (dn->dn_bonustype != DMU_OT_SA) { znode_phys_t *zp = (znode_phys_t *)dn->dn_bonus; sb->st_mode = zp->zp_mode; sb->st_uid = zp->zp_uid; sb->st_gid = zp->zp_gid; sb->st_size = zp->zp_size; } else { sa_hdr_phys_t *sahdrp; int hdrsize; size_t size = 0; void *buf = NULL; if (dn->dn_bonuslen != 0) sahdrp = (sa_hdr_phys_t *)DN_BONUS(dn); else { if ((dn->dn_flags & DNODE_FLAG_SPILL_BLKPTR) != 0) { blkptr_t *bp = DN_SPILL_BLKPTR(dn); int error; size = BP_GET_LSIZE(bp); buf = zfs_alloc(size); error = zio_read(spa, bp, buf); if (error != 0) { zfs_free(buf, size); return (error); } sahdrp = buf; } else { return (EIO); } } hdrsize = SA_HDR_SIZE(sahdrp); sb->st_mode = *(uint64_t *)((char *)sahdrp + hdrsize + SA_MODE_OFFSET); sb->st_uid = *(uint64_t *)((char *)sahdrp + hdrsize + SA_UID_OFFSET); sb->st_gid = *(uint64_t *)((char *)sahdrp + hdrsize + SA_GID_OFFSET); sb->st_size = *(uint64_t *)((char *)sahdrp + hdrsize + SA_SIZE_OFFSET); if (buf != NULL) zfs_free(buf, size); } return (0); } static int zfs_dnode_readlink(const spa_t *spa, dnode_phys_t *dn, char *path, size_t psize) { int rc = 0; if (dn->dn_bonustype == DMU_OT_SA) { sa_hdr_phys_t *sahdrp = NULL; size_t size = 0; void *buf = NULL; int hdrsize; char *p; if (dn->dn_bonuslen != 0) sahdrp = (sa_hdr_phys_t *)DN_BONUS(dn); else { blkptr_t *bp; if ((dn->dn_flags & DNODE_FLAG_SPILL_BLKPTR) == 0) return (EIO); bp = DN_SPILL_BLKPTR(dn); size = BP_GET_LSIZE(bp); buf = zfs_alloc(size); rc = zio_read(spa, bp, buf); if (rc != 0) { zfs_free(buf, size); return (rc); } sahdrp = buf; } hdrsize = SA_HDR_SIZE(sahdrp); p = (char *)((uintptr_t)sahdrp + hdrsize + SA_SYMLINK_OFFSET); memcpy(path, p, psize); if (buf != NULL) zfs_free(buf, size); return (0); } /* * Second test is purely to silence bogus compiler * warning about accessing past the end of dn_bonus. */ if (psize + sizeof(znode_phys_t) <= dn->dn_bonuslen && sizeof(znode_phys_t) <= sizeof(dn->dn_bonus)) { memcpy(path, &dn->dn_bonus[sizeof(znode_phys_t)], psize); } else { rc = dnode_read(spa, dn, 0, path, psize); } return (rc); } struct obj_list { uint64_t objnum; STAILQ_ENTRY(obj_list) entry; }; /* * Lookup a file and return its dnode. */ static int zfs_lookup(const struct zfsmount *mount, const char *upath, dnode_phys_t *dnode) { int rc; uint64_t objnum; const spa_t *spa; dnode_phys_t dn; const char *p, *q; char element[256]; char path[1024]; int symlinks_followed = 0; struct stat sb; struct obj_list *entry, *tentry; STAILQ_HEAD(, obj_list) on_cache = STAILQ_HEAD_INITIALIZER(on_cache); spa = mount->spa; if (mount->objset.os_type != DMU_OST_ZFS) { printf("ZFS: unexpected object set type %ju\n", (uintmax_t)mount->objset.os_type); return (EIO); } if ((entry = malloc(sizeof(struct obj_list))) == NULL) return (ENOMEM); /* * Get the root directory dnode. */ rc = objset_get_dnode(spa, &mount->objset, MASTER_NODE_OBJ, &dn); if (rc) { free(entry); return (rc); } rc = zap_lookup(spa, &dn, ZFS_ROOT_OBJ, sizeof(objnum), 1, &objnum); if (rc) { free(entry); return (rc); } entry->objnum = objnum; STAILQ_INSERT_HEAD(&on_cache, entry, entry); rc = objset_get_dnode(spa, &mount->objset, objnum, &dn); if (rc != 0) goto done; p = upath; while (p && *p) { rc = objset_get_dnode(spa, &mount->objset, objnum, &dn); if (rc != 0) goto done; while (*p == '/') p++; if (*p == '\0') break; q = p; while (*q != '\0' && *q != '/') q++; /* skip dot */ if (p + 1 == q && p[0] == '.') { p++; continue; } /* double dot */ if (p + 2 == q && p[0] == '.' && p[1] == '.') { p += 2; if (STAILQ_FIRST(&on_cache) == STAILQ_LAST(&on_cache, obj_list, entry)) { rc = ENOENT; goto done; } entry = STAILQ_FIRST(&on_cache); STAILQ_REMOVE_HEAD(&on_cache, entry); free(entry); objnum = (STAILQ_FIRST(&on_cache))->objnum; continue; } if (q - p + 1 > sizeof(element)) { rc = ENAMETOOLONG; goto done; } memcpy(element, p, q - p); element[q - p] = 0; p = q; if ((rc = zfs_dnode_stat(spa, &dn, &sb)) != 0) goto done; if (!S_ISDIR(sb.st_mode)) { rc = ENOTDIR; goto done; } rc = zap_lookup(spa, &dn, element, sizeof (objnum), 1, &objnum); if (rc) goto done; objnum = ZFS_DIRENT_OBJ(objnum); if ((entry = malloc(sizeof(struct obj_list))) == NULL) { rc = ENOMEM; goto done; } entry->objnum = objnum; STAILQ_INSERT_HEAD(&on_cache, entry, entry); rc = objset_get_dnode(spa, &mount->objset, objnum, &dn); if (rc) goto done; /* * Check for symlink. */ rc = zfs_dnode_stat(spa, &dn, &sb); if (rc) goto done; if (S_ISLNK(sb.st_mode)) { if (symlinks_followed > 10) { rc = EMLINK; goto done; } symlinks_followed++; /* * Read the link value and copy the tail of our * current path onto the end. */ if (sb.st_size + strlen(p) + 1 > sizeof(path)) { rc = ENAMETOOLONG; goto done; } strcpy(&path[sb.st_size], p); rc = zfs_dnode_readlink(spa, &dn, path, sb.st_size); if (rc != 0) goto done; /* * Restart with the new path, starting either at * the root or at the parent depending whether or * not the link is relative. */ p = path; if (*p == '/') { while (STAILQ_FIRST(&on_cache) != STAILQ_LAST(&on_cache, obj_list, entry)) { entry = STAILQ_FIRST(&on_cache); STAILQ_REMOVE_HEAD(&on_cache, entry); free(entry); } } else { entry = STAILQ_FIRST(&on_cache); STAILQ_REMOVE_HEAD(&on_cache, entry); free(entry); } objnum = (STAILQ_FIRST(&on_cache))->objnum; } } *dnode = dn; done: STAILQ_FOREACH_SAFE(entry, &on_cache, entry, tentry) free(entry); return (rc); } Index: head/sys/cddl/boot/zfs/zfsimpl.h =================================================================== --- head/sys/cddl/boot/zfs/zfsimpl.h (revision 357496) +++ head/sys/cddl/boot/zfs/zfsimpl.h (revision 357497) @@ -1,1808 +1,1813 @@ /*- * Copyright (c) 2002 McAfee, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Marshall * Kirk McKusick and McAfee Research,, the Security Research Division of * McAfee, Inc. under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as * part of the DARPA CHATS research program * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * 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 2013 by Saso Kiselkov. All rights reserved. */ /* * Copyright (c) 2013 by Delphix. All rights reserved. */ #define MAXNAMELEN 256 #define _NOTE(s) /* * AVL comparator helpers */ #define AVL_ISIGN(a) (((a) > 0) - ((a) < 0)) #define AVL_CMP(a, b) (((a) > (b)) - ((a) < (b))) #define AVL_PCMP(a, b) \ (((uintptr_t)(a) > (uintptr_t)(b)) - ((uintptr_t)(a) < (uintptr_t)(b))) typedef enum { B_FALSE, B_TRUE } boolean_t; /* CRC64 table */ #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ /* * Macros for various sorts of alignment and rounding when the alignment * is known to be a power of 2. */ #define P2ALIGN(x, align) ((x) & -(align)) #define P2PHASE(x, align) ((x) & ((align) - 1)) #define P2NPHASE(x, align) (-(x) & ((align) - 1)) #define P2ROUNDUP(x, align) (-(-(x) & -(align))) #define P2END(x, align) (-(~(x) & -(align))) #define P2PHASEUP(x, align, phase) ((phase) - (((phase) - (x)) & -(align))) #define P2BOUNDARY(off, len, align) (((off) ^ ((off) + (len) - 1)) > (align) - 1) /* * General-purpose 32-bit and 64-bit bitfield encodings. */ #define BF32_DECODE(x, low, len) P2PHASE((x) >> (low), 1U << (len)) #define BF64_DECODE(x, low, len) P2PHASE((x) >> (low), 1ULL << (len)) #define BF32_ENCODE(x, low, len) (P2PHASE((x), 1U << (len)) << (low)) #define BF64_ENCODE(x, low, len) (P2PHASE((x), 1ULL << (len)) << (low)) #define BF32_GET(x, low, len) BF32_DECODE(x, low, len) #define BF64_GET(x, low, len) BF64_DECODE(x, low, len) #define BF32_SET(x, low, len, val) \ ((x) ^= BF32_ENCODE((x >> low) ^ (val), low, len)) #define BF64_SET(x, low, len, val) \ ((x) ^= BF64_ENCODE((x >> low) ^ (val), low, len)) #define BF32_GET_SB(x, low, len, shift, bias) \ ((BF32_GET(x, low, len) + (bias)) << (shift)) #define BF64_GET_SB(x, low, len, shift, bias) \ ((BF64_GET(x, low, len) + (bias)) << (shift)) #define BF32_SET_SB(x, low, len, shift, bias, val) \ BF32_SET(x, low, len, ((val) >> (shift)) - (bias)) #define BF64_SET_SB(x, low, len, shift, bias, val) \ BF64_SET(x, low, len, ((val) >> (shift)) - (bias)) /* * Macros to reverse byte order */ #define BSWAP_8(x) ((x) & 0xff) #define BSWAP_16(x) ((BSWAP_8(x) << 8) | BSWAP_8((x) >> 8)) #define BSWAP_32(x) ((BSWAP_16(x) << 16) | BSWAP_16((x) >> 16)) #define BSWAP_64(x) ((BSWAP_32(x) << 32) | BSWAP_32((x) >> 32)) #define SPA_MINBLOCKSHIFT 9 #define SPA_OLDMAXBLOCKSHIFT 17 #define SPA_MAXBLOCKSHIFT 24 #define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT) #define SPA_OLDMAXBLOCKSIZE (1ULL << SPA_OLDMAXBLOCKSHIFT) #define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT) /* * The DVA size encodings for LSIZE and PSIZE support blocks up to 32MB. * The ASIZE encoding should be at least 64 times larger (6 more bits) * to support up to 4-way RAID-Z mirror mode with worst-case gang block * overhead, three DVAs per bp, plus one more bit in case we do anything * else that expands the ASIZE. */ #define SPA_LSIZEBITS 16 /* LSIZE up to 32M (2^16 * 512) */ #define SPA_PSIZEBITS 16 /* PSIZE up to 32M (2^16 * 512) */ #define SPA_ASIZEBITS 24 /* ASIZE up to 64 times larger */ /* * All SPA data is represented by 128-bit data virtual addresses (DVAs). * The members of the dva_t should be considered opaque outside the SPA. */ typedef struct dva { uint64_t dva_word[2]; } dva_t; /* * Each block has a 256-bit checksum -- strong enough for cryptographic hashes. */ typedef struct zio_cksum { uint64_t zc_word[4]; } zio_cksum_t; /* * Some checksums/hashes need a 256-bit initialization salt. This salt is kept * secret and is suitable for use in MAC algorithms as the key. */ typedef struct zio_cksum_salt { uint8_t zcs_bytes[32]; } zio_cksum_salt_t; /* * Each block is described by its DVAs, time of birth, checksum, etc. * The word-by-word, bit-by-bit layout of the blkptr is as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | vdev1 | GRID | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 1 |G| offset1 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 2 | vdev2 | GRID | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 3 |G| offset2 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 4 | vdev3 | GRID | ASIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 5 |G| offset3 | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 6 |BDX|lvl| type | cksum |E| comp| PSIZE | LSIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 7 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 8 | padding | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 9 | physical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * a | logical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * b | fill count | * +-------+-------+-------+-------+-------+-------+-------+-------+ * c | checksum[0] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * d | checksum[1] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * e | checksum[2] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * f | checksum[3] | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * Legend: * * vdev virtual device ID * offset offset into virtual device * LSIZE logical size * PSIZE physical size (after compression) * ASIZE allocated size (including RAID-Z parity and gang block headers) * GRID RAID-Z layout information (reserved for future use) * cksum checksum function * comp compression function * G gang block indicator * B byteorder (endianness) * D dedup * X encryption (on version 30, which is not supported) * E blkptr_t contains embedded data (see below) * lvl level of indirection * type DMU object type * phys birth txg of block allocation; zero if same as logical birth txg * log. birth transaction group in which the block was logically born * fill count number of non-zero blocks under this bp * checksum[4] 256-bit checksum of the data this bp describes */ /* * "Embedded" blkptr_t's don't actually point to a block, instead they * have a data payload embedded in the blkptr_t itself. See the comment * in blkptr.c for more details. * * The blkptr_t is laid out as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | payload | * 1 | payload | * 2 | payload | * 3 | payload | * 4 | payload | * 5 | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 6 |BDX|lvl| type | etype |E| comp| PSIZE| LSIZE | * +-------+-------+-------+-------+-------+-------+-------+-------+ * 7 | payload | * 8 | payload | * 9 | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * a | logical birth txg | * +-------+-------+-------+-------+-------+-------+-------+-------+ * b | payload | * c | payload | * d | payload | * e | payload | * f | payload | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * Legend: * * payload contains the embedded data * B (byteorder) byteorder (endianness) * D (dedup) padding (set to zero) * X encryption (set to zero; see above) * E (embedded) set to one * lvl indirection level * type DMU object type * etype how to interpret embedded data (BP_EMBEDDED_TYPE_*) * comp compression function of payload * PSIZE size of payload after compression, in bytes * LSIZE logical size of payload, in bytes * note that 25 bits is enough to store the largest * "normal" BP's LSIZE (2^16 * 2^9) in bytes * log. birth transaction group in which the block was logically born * * Note that LSIZE and PSIZE are stored in bytes, whereas for non-embedded * bp's they are stored in units of SPA_MINBLOCKSHIFT. * Generally, the generic BP_GET_*() macros can be used on embedded BP's. * The B, D, X, lvl, type, and comp fields are stored the same as with normal * BP's so the BP_SET_* macros can be used with them. etype, PSIZE, LSIZE must * be set with the BPE_SET_* macros. BP_SET_EMBEDDED() should be called before * other macros, as they assert that they are only used on BP's of the correct * "embedded-ness". */ #define BPE_GET_ETYPE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET((bp)->blk_prop, 40, 8)) #define BPE_SET_ETYPE(bp, t) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET((bp)->blk_prop, 40, 8, t); \ _NOTE(CONSTCOND) } while (0) #define BPE_GET_LSIZE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET_SB((bp)->blk_prop, 0, 25, 0, 1)) #define BPE_SET_LSIZE(bp, x) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, 0, 25, 0, 1, x); \ _NOTE(CONSTCOND) } while (0) #define BPE_GET_PSIZE(bp) \ (ASSERT(BP_IS_EMBEDDED(bp)), \ BF64_GET_SB((bp)->blk_prop, 25, 7, 0, 1)) #define BPE_SET_PSIZE(bp, x) do { \ ASSERT(BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, 25, 7, 0, 1, x); \ _NOTE(CONSTCOND) } while (0) typedef enum bp_embedded_type { BP_EMBEDDED_TYPE_DATA, BP_EMBEDDED_TYPE_RESERVED, /* Reserved for an unintegrated feature. */ NUM_BP_EMBEDDED_TYPES = BP_EMBEDDED_TYPE_RESERVED } bp_embedded_type_t; #define BPE_NUM_WORDS 14 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) #define BPE_IS_PAYLOADWORD(bp, wp) \ ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) #define SPA_BLKPTRSHIFT 7 /* blkptr_t is 128 bytes */ #define SPA_DVAS_PER_BP 3 /* Number of DVAs in a bp */ typedef struct blkptr { dva_t blk_dva[SPA_DVAS_PER_BP]; /* Data Virtual Addresses */ uint64_t blk_prop; /* size, compression, type, etc */ uint64_t blk_pad[2]; /* Extra space for the future */ uint64_t blk_phys_birth; /* txg when block was allocated */ uint64_t blk_birth; /* transaction group at birth */ uint64_t blk_fill; /* fill count */ zio_cksum_t blk_cksum; /* 256-bit checksum */ } blkptr_t; /* * Macros to get and set fields in a bp or DVA. */ #define DVA_GET_ASIZE(dva) \ BF64_GET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, SPA_MINBLOCKSHIFT, 0) #define DVA_SET_ASIZE(dva, x) \ BF64_SET_SB((dva)->dva_word[0], 0, SPA_ASIZEBITS, \ SPA_MINBLOCKSHIFT, 0, x) #define DVA_GET_GRID(dva) BF64_GET((dva)->dva_word[0], 24, 8) #define DVA_SET_GRID(dva, x) BF64_SET((dva)->dva_word[0], 24, 8, x) #define DVA_GET_VDEV(dva) BF64_GET((dva)->dva_word[0], 32, 32) #define DVA_SET_VDEV(dva, x) BF64_SET((dva)->dva_word[0], 32, 32, x) #define DVA_GET_OFFSET(dva) \ BF64_GET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0) #define DVA_SET_OFFSET(dva, x) \ BF64_SET_SB((dva)->dva_word[1], 0, 63, SPA_MINBLOCKSHIFT, 0, x) #define DVA_GET_GANG(dva) BF64_GET((dva)->dva_word[1], 63, 1) #define DVA_SET_GANG(dva, x) BF64_SET((dva)->dva_word[1], 63, 1, x) #define BP_GET_LSIZE(bp) \ (BP_IS_EMBEDDED(bp) ? \ (BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA ? BPE_GET_LSIZE(bp) : 0): \ BF64_GET_SB((bp)->blk_prop, 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1)) #define BP_SET_LSIZE(bp, x) do { \ ASSERT(!BP_IS_EMBEDDED(bp)); \ BF64_SET_SB((bp)->blk_prop, \ 0, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x); \ _NOTE(CONSTCOND) } while (0) #define BP_GET_PSIZE(bp) \ BF64_GET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1) #define BP_SET_PSIZE(bp, x) \ BF64_SET_SB((bp)->blk_prop, 16, SPA_LSIZEBITS, SPA_MINBLOCKSHIFT, 1, x) #define BP_GET_COMPRESS(bp) BF64_GET((bp)->blk_prop, 32, 7) #define BP_SET_COMPRESS(bp, x) BF64_SET((bp)->blk_prop, 32, 7, x) #define BP_GET_CHECKSUM(bp) BF64_GET((bp)->blk_prop, 40, 8) #define BP_SET_CHECKSUM(bp, x) BF64_SET((bp)->blk_prop, 40, 8, x) #define BP_GET_TYPE(bp) BF64_GET((bp)->blk_prop, 48, 8) #define BP_SET_TYPE(bp, x) BF64_SET((bp)->blk_prop, 48, 8, x) #define BP_GET_LEVEL(bp) BF64_GET((bp)->blk_prop, 56, 5) #define BP_SET_LEVEL(bp, x) BF64_SET((bp)->blk_prop, 56, 5, x) #define BP_IS_EMBEDDED(bp) BF64_GET((bp)->blk_prop, 39, 1) #define BP_GET_DEDUP(bp) BF64_GET((bp)->blk_prop, 62, 1) #define BP_SET_DEDUP(bp, x) BF64_SET((bp)->blk_prop, 62, 1, x) #define BP_GET_BYTEORDER(bp) BF64_GET((bp)->blk_prop, 63, 1) #define BP_SET_BYTEORDER(bp, x) BF64_SET((bp)->blk_prop, 63, 1, x) #define BP_PHYSICAL_BIRTH(bp) \ ((bp)->blk_phys_birth ? (bp)->blk_phys_birth : (bp)->blk_birth) #define BP_GET_ASIZE(bp) \ (DVA_GET_ASIZE(&(bp)->blk_dva[0]) + DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ DVA_GET_ASIZE(&(bp)->blk_dva[2])) #define BP_GET_UCSIZE(bp) \ ((BP_GET_LEVEL(bp) > 0 || dmu_ot[BP_GET_TYPE(bp)].ot_metadata) ? \ BP_GET_PSIZE(bp) : BP_GET_LSIZE(bp)); #define BP_GET_NDVAS(bp) \ (!!DVA_GET_ASIZE(&(bp)->blk_dva[0]) + \ !!DVA_GET_ASIZE(&(bp)->blk_dva[1]) + \ !!DVA_GET_ASIZE(&(bp)->blk_dva[2])) #define DVA_EQUAL(dva1, dva2) \ ((dva1)->dva_word[1] == (dva2)->dva_word[1] && \ (dva1)->dva_word[0] == (dva2)->dva_word[0]) #define ZIO_CHECKSUM_EQUAL(zc1, zc2) \ (0 == (((zc1).zc_word[0] - (zc2).zc_word[0]) | \ ((zc1).zc_word[1] - (zc2).zc_word[1]) | \ ((zc1).zc_word[2] - (zc2).zc_word[2]) | \ ((zc1).zc_word[3] - (zc2).zc_word[3]))) #define DVA_IS_VALID(dva) (DVA_GET_ASIZE(dva) != 0) #define ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3) \ { \ (zcp)->zc_word[0] = w0; \ (zcp)->zc_word[1] = w1; \ (zcp)->zc_word[2] = w2; \ (zcp)->zc_word[3] = w3; \ } #define BP_IDENTITY(bp) (&(bp)->blk_dva[0]) #define BP_IS_GANG(bp) DVA_GET_GANG(BP_IDENTITY(bp)) #define DVA_IS_EMPTY(dva) ((dva)->dva_word[0] == 0ULL && \ (dva)->dva_word[1] == 0ULL) #define BP_IS_HOLE(bp) DVA_IS_EMPTY(BP_IDENTITY(bp)) #define BP_IS_OLDER(bp, txg) (!BP_IS_HOLE(bp) && (bp)->blk_birth < (txg)) #define BP_ZERO(bp) \ { \ (bp)->blk_dva[0].dva_word[0] = 0; \ (bp)->blk_dva[0].dva_word[1] = 0; \ (bp)->blk_dva[1].dva_word[0] = 0; \ (bp)->blk_dva[1].dva_word[1] = 0; \ (bp)->blk_dva[2].dva_word[0] = 0; \ (bp)->blk_dva[2].dva_word[1] = 0; \ (bp)->blk_prop = 0; \ (bp)->blk_pad[0] = 0; \ (bp)->blk_pad[1] = 0; \ (bp)->blk_phys_birth = 0; \ (bp)->blk_birth = 0; \ (bp)->blk_fill = 0; \ ZIO_SET_CHECKSUM(&(bp)->blk_cksum, 0, 0, 0, 0); \ } #if BYTE_ORDER == _BIG_ENDIAN #define ZFS_HOST_BYTEORDER (0ULL) #else #define ZFS_HOST_BYTEORDER (1ULL) #endif #define BP_SHOULD_BYTESWAP(bp) (BP_GET_BYTEORDER(bp) != ZFS_HOST_BYTEORDER) #define BPE_NUM_WORDS 14 #define BPE_PAYLOAD_SIZE (BPE_NUM_WORDS * sizeof (uint64_t)) #define BPE_IS_PAYLOADWORD(bp, wp) \ ((wp) != &(bp)->blk_prop && (wp) != &(bp)->blk_birth) /* * Embedded checksum */ #define ZEC_MAGIC 0x210da7ab10c7a11ULL typedef struct zio_eck { uint64_t zec_magic; /* for validation, endianness */ zio_cksum_t zec_cksum; /* 256-bit checksum */ } zio_eck_t; /* * Gang block headers are self-checksumming and contain an array * of block pointers. */ #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t)) / sizeof (blkptr_t)) #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t) - \ (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ sizeof (uint64_t)) typedef struct zio_gbh { blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; uint64_t zg_filler[SPA_GBH_FILLER]; zio_eck_t zg_tail; } zio_gbh_phys_t; #define VDEV_RAIDZ_MAXPARITY 3 #define VDEV_PAD_SIZE (8 << 10) /* 2 padding areas (vl_pad1 and vl_pad2) to skip */ #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 #define VDEV_PHYS_SIZE (112 << 10) #define VDEV_UBERBLOCK_RING (128 << 10) /* * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock * ring when MMP is enabled. */ #define MMP_BLOCKS_PER_LABEL 1 /* The largest uberblock we support is 8k. */ #define MAX_UBERBLOCK_SHIFT (13) #define VDEV_UBERBLOCK_SHIFT(vd) \ MIN(MAX((vd)->v_top->v_ashift, UBERBLOCK_SHIFT), MAX_UBERBLOCK_SHIFT) #define VDEV_UBERBLOCK_COUNT(vd) \ (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) #define VDEV_UBERBLOCK_OFFSET(vd, n) \ offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) typedef struct vdev_phys { char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; zio_eck_t vp_zbt; } vdev_phys_t; typedef struct vdev_label { char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ char vl_pad2[VDEV_PAD_SIZE]; /* 8K */ vdev_phys_t vl_vdev_phys; /* 112K */ char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ } vdev_label_t; /* 256K total */ /* * vdev_dirty() flags */ #define VDD_METASLAB 0x01 #define VDD_DTL 0x02 /* * Size and offset of embedded boot loader region on each label. * The total size of the first two labels plus the boot area is 4MB. */ #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ /* * Size of label regions at the start and end of each leaf device. */ #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) #define VDEV_LABELS 4 enum zio_checksum { ZIO_CHECKSUM_INHERIT = 0, ZIO_CHECKSUM_ON, ZIO_CHECKSUM_OFF, ZIO_CHECKSUM_LABEL, ZIO_CHECKSUM_GANG_HEADER, ZIO_CHECKSUM_ZILOG, ZIO_CHECKSUM_FLETCHER_2, ZIO_CHECKSUM_FLETCHER_4, ZIO_CHECKSUM_SHA256, ZIO_CHECKSUM_ZILOG2, ZIO_CHECKSUM_NOPARITY, ZIO_CHECKSUM_SHA512, ZIO_CHECKSUM_SKEIN, ZIO_CHECKSUM_EDONR, ZIO_CHECKSUM_FUNCTIONS }; #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON enum zio_compress { ZIO_COMPRESS_INHERIT = 0, ZIO_COMPRESS_ON, ZIO_COMPRESS_OFF, ZIO_COMPRESS_LZJB, ZIO_COMPRESS_EMPTY, ZIO_COMPRESS_GZIP_1, ZIO_COMPRESS_GZIP_2, ZIO_COMPRESS_GZIP_3, ZIO_COMPRESS_GZIP_4, ZIO_COMPRESS_GZIP_5, ZIO_COMPRESS_GZIP_6, ZIO_COMPRESS_GZIP_7, ZIO_COMPRESS_GZIP_8, ZIO_COMPRESS_GZIP_9, ZIO_COMPRESS_ZLE, ZIO_COMPRESS_LZ4, ZIO_COMPRESS_FUNCTIONS }; #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF /* nvlist pack encoding */ #define NV_ENCODE_NATIVE 0 #define NV_ENCODE_XDR 1 typedef enum { DATA_TYPE_UNKNOWN = 0, DATA_TYPE_BOOLEAN, DATA_TYPE_BYTE, DATA_TYPE_INT16, DATA_TYPE_UINT16, DATA_TYPE_INT32, DATA_TYPE_UINT32, DATA_TYPE_INT64, DATA_TYPE_UINT64, DATA_TYPE_STRING, DATA_TYPE_BYTE_ARRAY, DATA_TYPE_INT16_ARRAY, DATA_TYPE_UINT16_ARRAY, DATA_TYPE_INT32_ARRAY, DATA_TYPE_UINT32_ARRAY, DATA_TYPE_INT64_ARRAY, DATA_TYPE_UINT64_ARRAY, DATA_TYPE_STRING_ARRAY, DATA_TYPE_HRTIME, DATA_TYPE_NVLIST, DATA_TYPE_NVLIST_ARRAY, DATA_TYPE_BOOLEAN_VALUE, DATA_TYPE_INT8, DATA_TYPE_UINT8, DATA_TYPE_BOOLEAN_ARRAY, DATA_TYPE_INT8_ARRAY, DATA_TYPE_UINT8_ARRAY } data_type_t; /* * On-disk version number. */ #define SPA_VERSION_1 1ULL #define SPA_VERSION_2 2ULL #define SPA_VERSION_3 3ULL #define SPA_VERSION_4 4ULL #define SPA_VERSION_5 5ULL #define SPA_VERSION_6 6ULL #define SPA_VERSION_7 7ULL #define SPA_VERSION_8 8ULL #define SPA_VERSION_9 9ULL #define SPA_VERSION_10 10ULL #define SPA_VERSION_11 11ULL #define SPA_VERSION_12 12ULL #define SPA_VERSION_13 13ULL #define SPA_VERSION_14 14ULL #define SPA_VERSION_15 15ULL #define SPA_VERSION_16 16ULL #define SPA_VERSION_17 17ULL #define SPA_VERSION_18 18ULL #define SPA_VERSION_19 19ULL #define SPA_VERSION_20 20ULL #define SPA_VERSION_21 21ULL #define SPA_VERSION_22 22ULL #define SPA_VERSION_23 23ULL #define SPA_VERSION_24 24ULL #define SPA_VERSION_25 25ULL #define SPA_VERSION_26 26ULL #define SPA_VERSION_27 27ULL #define SPA_VERSION_28 28ULL #define SPA_VERSION_5000 5000ULL /* * When bumping up SPA_VERSION, make sure GRUB ZFS understands the on-disk * format change. Go to usr/src/grub/grub-0.97/stage2/{zfs-include/, fsys_zfs*}, * and do the appropriate changes. Also bump the version number in * usr/src/grub/capability. */ #define SPA_VERSION SPA_VERSION_5000 #define SPA_VERSION_STRING "5000" /* * Symbolic names for the changes that caused a SPA_VERSION switch. * Used in the code when checking for presence or absence of a feature. * Feel free to define multiple symbolic names for each version if there * were multiple changes to on-disk structures during that version. * * NOTE: When checking the current SPA_VERSION in your code, be sure * to use spa_version() since it reports the version of the * last synced uberblock. Checking the in-flight version can * be dangerous in some cases. */ #define SPA_VERSION_INITIAL SPA_VERSION_1 #define SPA_VERSION_DITTO_BLOCKS SPA_VERSION_2 #define SPA_VERSION_SPARES SPA_VERSION_3 #define SPA_VERSION_RAID6 SPA_VERSION_3 #define SPA_VERSION_BPLIST_ACCOUNT SPA_VERSION_3 #define SPA_VERSION_RAIDZ_DEFLATE SPA_VERSION_3 #define SPA_VERSION_DNODE_BYTES SPA_VERSION_3 #define SPA_VERSION_ZPOOL_HISTORY SPA_VERSION_4 #define SPA_VERSION_GZIP_COMPRESSION SPA_VERSION_5 #define SPA_VERSION_BOOTFS SPA_VERSION_6 #define SPA_VERSION_SLOGS SPA_VERSION_7 #define SPA_VERSION_DELEGATED_PERMS SPA_VERSION_8 #define SPA_VERSION_FUID SPA_VERSION_9 #define SPA_VERSION_REFRESERVATION SPA_VERSION_9 #define SPA_VERSION_REFQUOTA SPA_VERSION_9 #define SPA_VERSION_UNIQUE_ACCURATE SPA_VERSION_9 #define SPA_VERSION_L2CACHE SPA_VERSION_10 #define SPA_VERSION_NEXT_CLONES SPA_VERSION_11 #define SPA_VERSION_ORIGIN SPA_VERSION_11 #define SPA_VERSION_DSL_SCRUB SPA_VERSION_11 #define SPA_VERSION_SNAP_PROPS SPA_VERSION_12 #define SPA_VERSION_USED_BREAKDOWN SPA_VERSION_13 #define SPA_VERSION_PASSTHROUGH_X SPA_VERSION_14 #define SPA_VERSION_USERSPACE SPA_VERSION_15 #define SPA_VERSION_STMF_PROP SPA_VERSION_16 #define SPA_VERSION_RAIDZ3 SPA_VERSION_17 #define SPA_VERSION_USERREFS SPA_VERSION_18 #define SPA_VERSION_HOLES SPA_VERSION_19 #define SPA_VERSION_ZLE_COMPRESSION SPA_VERSION_20 #define SPA_VERSION_DEDUP SPA_VERSION_21 #define SPA_VERSION_RECVD_PROPS SPA_VERSION_22 #define SPA_VERSION_SLIM_ZIL SPA_VERSION_23 #define SPA_VERSION_SA SPA_VERSION_24 #define SPA_VERSION_SCAN SPA_VERSION_25 #define SPA_VERSION_DIR_CLONES SPA_VERSION_26 #define SPA_VERSION_DEADLISTS SPA_VERSION_26 #define SPA_VERSION_FAST_SNAP SPA_VERSION_27 #define SPA_VERSION_MULTI_REPLACE SPA_VERSION_28 #define SPA_VERSION_BEFORE_FEATURES SPA_VERSION_28 #define SPA_VERSION_FEATURES SPA_VERSION_5000 #define SPA_VERSION_IS_SUPPORTED(v) \ (((v) >= SPA_VERSION_INITIAL && (v) <= SPA_VERSION_BEFORE_FEATURES) || \ ((v) >= SPA_VERSION_FEATURES && (v) <= SPA_VERSION)) /* * The following are configuration names used in the nvlist describing a pool's * configuration. */ #define ZPOOL_CONFIG_VERSION "version" #define ZPOOL_CONFIG_POOL_NAME "name" #define ZPOOL_CONFIG_POOL_STATE "state" #define ZPOOL_CONFIG_POOL_TXG "txg" #define ZPOOL_CONFIG_POOL_GUID "pool_guid" #define ZPOOL_CONFIG_CREATE_TXG "create_txg" #define ZPOOL_CONFIG_TOP_GUID "top_guid" #define ZPOOL_CONFIG_VDEV_TREE "vdev_tree" #define ZPOOL_CONFIG_TYPE "type" #define ZPOOL_CONFIG_CHILDREN "children" #define ZPOOL_CONFIG_ID "id" #define ZPOOL_CONFIG_GUID "guid" #define ZPOOL_CONFIG_INDIRECT_OBJECT "com.delphix:indirect_object" #define ZPOOL_CONFIG_INDIRECT_BIRTHS "com.delphix:indirect_births" #define ZPOOL_CONFIG_PREV_INDIRECT_VDEV "com.delphix:prev_indirect_vdev" #define ZPOOL_CONFIG_PATH "path" #define ZPOOL_CONFIG_DEVID "devid" #define ZPOOL_CONFIG_METASLAB_ARRAY "metaslab_array" #define ZPOOL_CONFIG_METASLAB_SHIFT "metaslab_shift" #define ZPOOL_CONFIG_ASHIFT "ashift" #define ZPOOL_CONFIG_ASIZE "asize" #define ZPOOL_CONFIG_DTL "DTL" #define ZPOOL_CONFIG_STATS "stats" #define ZPOOL_CONFIG_WHOLE_DISK "whole_disk" #define ZPOOL_CONFIG_ERRCOUNT "error_count" #define ZPOOL_CONFIG_NOT_PRESENT "not_present" #define ZPOOL_CONFIG_SPARES "spares" #define ZPOOL_CONFIG_IS_SPARE "is_spare" #define ZPOOL_CONFIG_NPARITY "nparity" #define ZPOOL_CONFIG_HOSTID "hostid" #define ZPOOL_CONFIG_HOSTNAME "hostname" #define ZPOOL_CONFIG_IS_LOG "is_log" #define ZPOOL_CONFIG_TIMESTAMP "timestamp" /* not stored on disk */ #define ZPOOL_CONFIG_FEATURES_FOR_READ "features_for_read" #define ZPOOL_CONFIG_VDEV_CHILDREN "vdev_children" /* * The persistent vdev state is stored as separate values rather than a single * 'vdev_state' entry. This is because a device can be in multiple states, such * as offline and degraded. */ #define ZPOOL_CONFIG_OFFLINE "offline" #define ZPOOL_CONFIG_FAULTED "faulted" #define ZPOOL_CONFIG_DEGRADED "degraded" #define ZPOOL_CONFIG_REMOVED "removed" #define ZPOOL_CONFIG_FRU "fru" #define ZPOOL_CONFIG_AUX_STATE "aux_state" #define VDEV_TYPE_ROOT "root" #define VDEV_TYPE_MIRROR "mirror" #define VDEV_TYPE_REPLACING "replacing" #define VDEV_TYPE_RAIDZ "raidz" #define VDEV_TYPE_DISK "disk" #define VDEV_TYPE_FILE "file" #define VDEV_TYPE_MISSING "missing" #define VDEV_TYPE_HOLE "hole" #define VDEV_TYPE_SPARE "spare" #define VDEV_TYPE_LOG "log" #define VDEV_TYPE_L2CACHE "l2cache" #define VDEV_TYPE_INDIRECT "indirect" /* * This is needed in userland to report the minimum necessary device size. */ #define SPA_MINDEVSIZE (64ULL << 20) /* * The location of the pool configuration repository, shared between kernel and * userland. */ #define ZPOOL_CACHE "/boot/zfs/zpool.cache" /* * vdev states are ordered from least to most healthy. * A vdev that's CANT_OPEN or below is considered unusable. */ typedef enum vdev_state { VDEV_STATE_UNKNOWN = 0, /* Uninitialized vdev */ VDEV_STATE_CLOSED, /* Not currently open */ VDEV_STATE_OFFLINE, /* Not allowed to open */ VDEV_STATE_REMOVED, /* Explicitly removed from system */ VDEV_STATE_CANT_OPEN, /* Tried to open, but failed */ VDEV_STATE_FAULTED, /* External request to fault device */ VDEV_STATE_DEGRADED, /* Replicated vdev with unhealthy kids */ VDEV_STATE_HEALTHY /* Presumed good */ } vdev_state_t; /* * vdev aux states. When a vdev is in the CANT_OPEN state, the aux field * of the vdev stats structure uses these constants to distinguish why. */ typedef enum vdev_aux { VDEV_AUX_NONE, /* no error */ VDEV_AUX_OPEN_FAILED, /* ldi_open_*() or vn_open() failed */ VDEV_AUX_CORRUPT_DATA, /* bad label or disk contents */ VDEV_AUX_NO_REPLICAS, /* insufficient number of replicas */ VDEV_AUX_BAD_GUID_SUM, /* vdev guid sum doesn't match */ VDEV_AUX_TOO_SMALL, /* vdev size is too small */ VDEV_AUX_BAD_LABEL, /* the label is OK but invalid */ VDEV_AUX_VERSION_NEWER, /* on-disk version is too new */ VDEV_AUX_VERSION_OLDER, /* on-disk version is too old */ VDEV_AUX_SPARED /* hot spare used in another pool */ } vdev_aux_t; /* * pool state. The following states are written to disk as part of the normal * SPA lifecycle: ACTIVE, EXPORTED, DESTROYED, SPARE. The remaining states are * software abstractions used at various levels to communicate pool state. */ typedef enum pool_state { POOL_STATE_ACTIVE = 0, /* In active use */ POOL_STATE_EXPORTED, /* Explicitly exported */ POOL_STATE_DESTROYED, /* Explicitly destroyed */ POOL_STATE_SPARE, /* Reserved for hot spare use */ POOL_STATE_UNINITIALIZED, /* Internal spa_t state */ POOL_STATE_UNAVAIL, /* Internal libzfs state */ POOL_STATE_POTENTIALLY_ACTIVE /* Internal libzfs state */ } pool_state_t; /* * The uberblock version is incremented whenever an incompatible on-disk * format change is made to the SPA, DMU, or ZAP. * * Note: the first two fields should never be moved. When a storage pool * is opened, the uberblock must be read off the disk before the version * can be checked. If the ub_version field is moved, we may not detect * version mismatch. If the ub_magic field is moved, applications that * expect the magic number in the first word won't work. */ #define UBERBLOCK_MAGIC 0x00bab10c /* oo-ba-bloc! */ #define UBERBLOCK_SHIFT 10 /* up to 1K */ #define MMP_MAGIC 0xa11cea11 /* all-see-all */ #define MMP_INTERVAL_VALID_BIT 0x01 #define MMP_SEQ_VALID_BIT 0x02 #define MMP_FAIL_INT_VALID_BIT 0x04 #define MMP_VALID(ubp) (ubp->ub_magic == UBERBLOCK_MAGIC && \ ubp->ub_mmp_magic == MMP_MAGIC) #define MMP_INTERVAL_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ MMP_INTERVAL_VALID_BIT)) #define MMP_SEQ_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ MMP_SEQ_VALID_BIT)) #define MMP_FAIL_INT_VALID(ubp) (MMP_VALID(ubp) && (ubp->ub_mmp_config & \ MMP_FAIL_INT_VALID_BIT)) #define MMP_INTERVAL(ubp) ((ubp->ub_mmp_config & 0x00000000FFFFFF00) \ >> 8) #define MMP_SEQ(ubp) ((ubp->ub_mmp_config & 0x0000FFFF00000000) \ >> 32) #define MMP_FAIL_INT(ubp) ((ubp->ub_mmp_config & 0xFFFF000000000000) \ >> 48) typedef struct uberblock { uint64_t ub_magic; /* UBERBLOCK_MAGIC */ uint64_t ub_version; /* SPA_VERSION */ uint64_t ub_txg; /* txg of last sync */ uint64_t ub_guid_sum; /* sum of all vdev guids */ uint64_t ub_timestamp; /* UTC time of last sync */ blkptr_t ub_rootbp; /* MOS objset_phys_t */ /* highest SPA_VERSION supported by software that wrote this txg */ uint64_t ub_software_version; /* Maybe missing in uberblocks we read, but always written */ uint64_t ub_mmp_magic; /* * If ub_mmp_delay == 0 and ub_mmp_magic is valid, MMP is off. * Otherwise, nanosec since last MMP write. */ uint64_t ub_mmp_delay; /* * The ub_mmp_config contains the multihost write interval, multihost * fail intervals, sequence number for sub-second granularity, and * valid bit mask. This layout is as follows: * * 64 56 48 40 32 24 16 8 0 * +-------+-------+-------+-------+-------+-------+-------+-------+ * 0 | Fail Intervals| Seq | Write Interval (ms) | VALID | * +-------+-------+-------+-------+-------+-------+-------+-------+ * * This allows a write_interval of (2^24/1000)s, over 4.5 hours * * VALID Bits: * - 0x01 - Write Interval (ms) * - 0x02 - Sequence number exists * - 0x04 - Fail Intervals * - 0xf8 - Reserved */ uint64_t ub_mmp_config; /* * ub_checkpoint_txg indicates two things about the current uberblock: * * 1] If it is not zero then this uberblock is a checkpoint. If it is * zero, then this uberblock is not a checkpoint. * * 2] On checkpointed uberblocks, the value of ub_checkpoint_txg is * the ub_txg that the uberblock had at the time we moved it to * the MOS config. * * The field is set when we checkpoint the uberblock and continues to * hold that value even after we've rewound (unlike the ub_txg that * is reset to a higher value). * * Besides checks used to determine whether we are reopening the * pool from a checkpointed uberblock [see spa_ld_select_uberblock()], * the value of the field is used to determine which ZIL blocks have * been allocated according to the ms_sm when we are rewinding to a * checkpoint. Specifically, if blk_birth > ub_checkpoint_txg, then * the ZIL block is not allocated [see uses of spa_min_claim_txg()]. */ uint64_t ub_checkpoint_txg; } uberblock_t; /* * Flags. */ #define DNODE_MUST_BE_ALLOCATED 1 #define DNODE_MUST_BE_FREE 2 /* * Fixed constants. */ #define DNODE_SHIFT 9 /* 512 bytes */ #define DN_MIN_INDBLKSHIFT 12 /* 4k */ #define DN_MAX_INDBLKSHIFT 17 /* 128k */ #define DNODE_BLOCK_SHIFT 14 /* 16k */ #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ /* * Derived constants. */ #define DNODE_MIN_SIZE (1 << DNODE_SHIFT) #define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT) #define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT) #define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT) #define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT) #define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \ (1 << SPA_BLKPTRSHIFT)) #define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT) #define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE)) #define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> \ SPA_BLKPTRSHIFT) #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) #define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1) #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) /* The +2 here is a cheesy way to round up */ #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ #define DNODE_FLAG_USED_BYTES (1<<0) #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) /* Does dnode have a SA spill blkptr in bonus? */ #define DNODE_FLAG_SPILL_BLKPTR (1<<2) typedef struct dnode_phys { uint8_t dn_type; /* dmu_object_type_t */ uint8_t dn_indblkshift; /* ln2(indirect block size) */ uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ uint8_t dn_nblkptr; /* length of dn_blkptr */ uint8_t dn_bonustype; /* type of data in bonus buffer */ uint8_t dn_checksum; /* ZIO_CHECKSUM type */ uint8_t dn_compress; /* ZIO_COMPRESS type */ uint8_t dn_flags; /* DNODE_FLAG_* */ uint16_t dn_datablkszsec; /* data block size in 512b sectors */ uint16_t dn_bonuslen; /* length of dn_bonus */ uint8_t dn_extra_slots; /* # of subsequent slots consumed */ uint8_t dn_pad2[3]; /* accounting is protected by dn_dirty_mtx */ uint64_t dn_maxblkid; /* largest allocated block ID */ uint64_t dn_used; /* bytes (or sectors) of disk space */ uint64_t dn_pad3[4]; /* * The tail region is 448 bytes for a 512 byte dnode, and * correspondingly larger for larger dnode sizes. The spill * block pointer, when present, is always at the end of the tail * region. There are three ways this space may be used, using * a 512 byte dnode for this diagram: * * 0 64 128 192 256 320 384 448 (offset) * +---------------+---------------+---------------+-------+ * | dn_blkptr[0] | dn_blkptr[1] | dn_blkptr[2] | / | * +---------------+---------------+---------------+-------+ * | dn_blkptr[0] | dn_bonus[0..319] | * +---------------+-----------------------+---------------+ * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill | * +---------------+-----------------------+---------------+ */ union { blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)]; struct { blkptr_t __dn_ignore1; uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN]; }; struct { blkptr_t __dn_ignore2; uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN - sizeof (blkptr_t)]; blkptr_t dn_spill; }; }; } dnode_phys_t; #define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \ (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT)) typedef enum dmu_object_byteswap { DMU_BSWAP_UINT8, DMU_BSWAP_UINT16, DMU_BSWAP_UINT32, DMU_BSWAP_UINT64, DMU_BSWAP_ZAP, DMU_BSWAP_DNODE, DMU_BSWAP_OBJSET, DMU_BSWAP_ZNODE, DMU_BSWAP_OLDACL, DMU_BSWAP_ACL, /* * Allocating a new byteswap type number makes the on-disk format * incompatible with any other format that uses the same number. * * Data can usually be structured to work with one of the * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. */ DMU_BSWAP_NUMFUNCS } dmu_object_byteswap_t; #define DMU_OT_NEWTYPE 0x80 #define DMU_OT_METADATA 0x40 #define DMU_OT_BYTESWAP_MASK 0x3f /* * Defines a uint8_t object type. Object types specify if the data * in the object is metadata (boolean) and how to byteswap the data * (dmu_object_byteswap_t). */ #define DMU_OT(byteswap, metadata) \ (DMU_OT_NEWTYPE | \ ((metadata) ? DMU_OT_METADATA : 0) | \ ((byteswap) & DMU_OT_BYTESWAP_MASK)) typedef enum dmu_object_type { DMU_OT_NONE, /* general: */ DMU_OT_OBJECT_DIRECTORY, /* ZAP */ DMU_OT_OBJECT_ARRAY, /* UINT64 */ DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ DMU_OT_BPLIST, /* UINT64 */ DMU_OT_BPLIST_HDR, /* UINT64 */ /* spa: */ DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ DMU_OT_SPACE_MAP, /* UINT64 */ /* zil: */ DMU_OT_INTENT_LOG, /* UINT64 */ /* dmu: */ DMU_OT_DNODE, /* DNODE */ DMU_OT_OBJSET, /* OBJSET */ /* dsl: */ DMU_OT_DSL_DIR, /* UINT64 */ DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ DMU_OT_DSL_PROPS, /* ZAP */ DMU_OT_DSL_DATASET, /* UINT64 */ /* zpl: */ DMU_OT_ZNODE, /* ZNODE */ DMU_OT_OLDACL, /* Old ACL */ DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ DMU_OT_MASTER_NODE, /* ZAP */ DMU_OT_UNLINKED_SET, /* ZAP */ /* zvol: */ DMU_OT_ZVOL, /* UINT8 */ DMU_OT_ZVOL_PROP, /* ZAP */ /* other; for testing only! */ DMU_OT_PLAIN_OTHER, /* UINT8 */ DMU_OT_UINT64_OTHER, /* UINT64 */ DMU_OT_ZAP_OTHER, /* ZAP */ /* new object types: */ DMU_OT_ERROR_LOG, /* ZAP */ DMU_OT_SPA_HISTORY, /* UINT8 */ DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ DMU_OT_POOL_PROPS, /* ZAP */ DMU_OT_DSL_PERMS, /* ZAP */ DMU_OT_ACL, /* ACL */ DMU_OT_SYSACL, /* SYSACL */ DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ DMU_OT_NEXT_CLONES, /* ZAP */ DMU_OT_SCAN_QUEUE, /* ZAP */ DMU_OT_USERGROUP_USED, /* ZAP */ DMU_OT_USERGROUP_QUOTA, /* ZAP */ DMU_OT_USERREFS, /* ZAP */ DMU_OT_DDT_ZAP, /* ZAP */ DMU_OT_DDT_STATS, /* ZAP */ DMU_OT_SA, /* System attr */ DMU_OT_SA_MASTER_NODE, /* ZAP */ DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ DMU_OT_SCAN_XLATE, /* ZAP */ DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ DMU_OT_NUMTYPES, /* * Names for valid types declared with DMU_OT(). */ DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE) } dmu_object_type_t; typedef enum dmu_objset_type { DMU_OST_NONE, DMU_OST_META, DMU_OST_ZFS, DMU_OST_ZVOL, DMU_OST_OTHER, /* For testing only! */ DMU_OST_ANY, /* Be careful! */ DMU_OST_NUMTYPES } dmu_objset_type_t; -#define ZAP_MAXVALUELEN (1024 * 8) +#define ZAP_MAXVALUELEN (1024 * 8) /* * header for all bonus and spill buffers. * The header has a fixed portion with a variable number * of "lengths" depending on the number of variable sized * attribues which are determined by the "layout number" */ #define SA_MAGIC 0x2F505A /* ZFS SA */ typedef struct sa_hdr_phys { uint32_t sa_magic; uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ /* ... Data follows the lengths. */ } sa_hdr_phys_t; /* * sa_hdr_phys -> sa_layout_info * * 16 10 0 * +--------+-------+ * | hdrsz |layout | * +--------+-------+ * * Bits 0-10 are the layout number * Bits 11-16 are the size of the header. * The hdrsize is the number * 8 * * For example. * hdrsz of 1 ==> 8 byte header * 2 ==> 16 byte header * */ #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0) #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ { \ BF32_SET_SB(x, 10, 6, 3, 0, size); \ BF32_SET(x, 0, 10, num); \ } #define SA_MODE_OFFSET 0 #define SA_SIZE_OFFSET 8 #define SA_GEN_OFFSET 16 #define SA_UID_OFFSET 24 #define SA_GID_OFFSET 32 #define SA_PARENT_OFFSET 40 #define SA_SYMLINK_OFFSET 160 #define ZIO_OBJSET_MAC_LEN 32 /* * Intent log header - this on disk structure holds fields to manage * the log. All fields are 64 bit to easily handle cross architectures. */ typedef struct zil_header { uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ uint64_t zh_replay_seq; /* highest replayed sequence number */ blkptr_t zh_log; /* log chain */ uint64_t zh_claim_seq; /* highest claimed sequence number */ uint64_t zh_pad[5]; } zil_header_t; #define OBJSET_PHYS_SIZE_V2 2048 #define OBJSET_PHYS_SIZE_V3 4096 typedef struct objset_phys { dnode_phys_t os_meta_dnode; zil_header_t os_zil_header; uint64_t os_type; uint64_t os_flags; uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN]; uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN]; char os_pad0[OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)*3 - sizeof (zil_header_t) - sizeof (uint64_t)*2 - 2*ZIO_OBJSET_MAC_LEN]; dnode_phys_t os_userused_dnode; dnode_phys_t os_groupused_dnode; dnode_phys_t os_projectused_dnode; char os_pad1[OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)]; } objset_phys_t; typedef struct dsl_dir_phys { uint64_t dd_creation_time; /* not actually used */ uint64_t dd_head_dataset_obj; uint64_t dd_parent_obj; uint64_t dd_clone_parent_obj; uint64_t dd_child_dir_zapobj; /* * how much space our children are accounting for; for leaf * datasets, == physical space used by fs + snaps */ uint64_t dd_used_bytes; uint64_t dd_compressed_bytes; uint64_t dd_uncompressed_bytes; /* Administrative quota setting */ uint64_t dd_quota; /* Administrative reservation setting */ uint64_t dd_reserved; uint64_t dd_props_zapobj; uint64_t dd_pad[21]; /* pad out to 256 bytes for good measure */ } dsl_dir_phys_t; typedef struct dsl_dataset_phys { uint64_t ds_dir_obj; uint64_t ds_prev_snap_obj; uint64_t ds_prev_snap_txg; uint64_t ds_next_snap_obj; uint64_t ds_snapnames_zapobj; /* zap obj of snaps; ==0 for snaps */ uint64_t ds_num_children; /* clone/snap children; ==0 for head */ uint64_t ds_creation_time; /* seconds since 1970 */ uint64_t ds_creation_txg; uint64_t ds_deadlist_obj; uint64_t ds_used_bytes; uint64_t ds_compressed_bytes; uint64_t ds_uncompressed_bytes; uint64_t ds_unique_bytes; /* only relevant to snapshots */ /* * The ds_fsid_guid is a 56-bit ID that can change to avoid * collisions. The ds_guid is a 64-bit ID that will never * change, so there is a small probability that it will collide. */ uint64_t ds_fsid_guid; uint64_t ds_guid; uint64_t ds_flags; blkptr_t ds_bp; uint64_t ds_pad[8]; /* pad out to 320 bytes for good measure */ } dsl_dataset_phys_t; /* * The names of zap entries in the DIRECTORY_OBJECT of the MOS. */ #define DMU_POOL_DIRECTORY_OBJECT 1 #define DMU_POOL_CONFIG "config" #define DMU_POOL_FEATURES_FOR_READ "features_for_read" #define DMU_POOL_ROOT_DATASET "root_dataset" #define DMU_POOL_SYNC_BPLIST "sync_bplist" #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" #define DMU_POOL_ERRLOG_LAST "errlog_last" #define DMU_POOL_SPARES "spares" #define DMU_POOL_DEFLATE "deflate" #define DMU_POOL_HISTORY "history" #define DMU_POOL_PROPS "pool_props" #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" #define DMU_POOL_REMOVING "com.delphix:removing" #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" #define ZAP_MAGIC 0x2F52AB2ABULL #define FZAP_BLOCK_SHIFT(zap) ((zap)->zap_block_shift) #define ZAP_MAXCD (uint32_t)(-1) #define ZAP_HASHBITS 28 #define MZAP_ENT_LEN 64 #define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2) -#define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT -#define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT) +#define MZAP_MAX_BLKSZ SPA_OLD_MAXBLOCKSIZE typedef struct mzap_ent_phys { uint64_t mze_value; uint32_t mze_cd; uint16_t mze_pad; /* in case we want to chain them someday */ char mze_name[MZAP_NAME_LEN]; } mzap_ent_phys_t; typedef struct mzap_phys { uint64_t mz_block_type; /* ZBT_MICRO */ uint64_t mz_salt; - uint64_t mz_pad[6]; + uint64_t mz_normflags; + uint64_t mz_pad[5]; mzap_ent_phys_t mz_chunk[1]; /* actually variable size depending on block size */ } mzap_phys_t; /* * The (fat) zap is stored in one object. It is an array of * 1<= 6] [zap_leaf_t] [ptrtbl] ... * */ #define ZBT_LEAF ((1ULL << 63) + 0) #define ZBT_HEADER ((1ULL << 63) + 1) #define ZBT_MICRO ((1ULL << 63) + 3) /* any other values are ptrtbl blocks */ /* * the embedded pointer table takes up half a block: * block size / entry size (2^3) / 2 */ #define ZAP_EMBEDDED_PTRTBL_SHIFT(zap) (FZAP_BLOCK_SHIFT(zap) - 3 - 1) /* * The embedded pointer table starts half-way through the block. Since * the pointer table itself is half the block, it starts at (64-bit) * word number (1<zap_phys) \ [(idx) + (1<l_bs) - hash entry size (2) * number of hash * entries - header space (2*chunksize) */ #define ZAP_LEAF_NUMCHUNKS(l) \ (((1<<(l)->l_bs) - 2*ZAP_LEAF_HASH_NUMENTRIES(l)) / \ ZAP_LEAF_CHUNKSIZE - 2) /* * The amount of space within the chunk available for the array is: * chunk size - space for type (1) - space for next pointer (2) */ #define ZAP_LEAF_ARRAY_BYTES (ZAP_LEAF_CHUNKSIZE - 3) #define ZAP_LEAF_ARRAY_NCHUNKS(bytes) \ (((bytes)+ZAP_LEAF_ARRAY_BYTES-1)/ZAP_LEAF_ARRAY_BYTES) /* * Low water mark: when there are only this many chunks free, start * growing the ptrtbl. Ideally, this should be larger than a * "reasonably-sized" entry. 20 chunks is more than enough for the * largest directory entry (MAXNAMELEN (256) byte name, 8-byte value), * while still being only around 3% for 16k blocks. */ #define ZAP_LEAF_LOW_WATER (20) /* * The leaf hash table has block size / 2^5 (32) number of entries, * which should be more than enough for the maximum number of entries, * which is less than block size / CHUNKSIZE (24) / minimum number of * chunks per entry (3). */ #define ZAP_LEAF_HASH_SHIFT(l) ((l)->l_bs - 5) #define ZAP_LEAF_HASH_NUMENTRIES(l) (1 << ZAP_LEAF_HASH_SHIFT(l)) /* * The chunks start immediately after the hash table. The end of the * hash table is at l_hash + HASH_NUMENTRIES, which we simply cast to a * chunk_t. */ #define ZAP_LEAF_CHUNK(l, idx) \ ((zap_leaf_chunk_t *) \ ((l)->l_phys->l_hash + ZAP_LEAF_HASH_NUMENTRIES(l)))[idx] #define ZAP_LEAF_ENTRY(l, idx) (&ZAP_LEAF_CHUNK(l, idx).l_entry) typedef enum zap_chunk_type { ZAP_CHUNK_FREE = 253, ZAP_CHUNK_ENTRY = 252, ZAP_CHUNK_ARRAY = 251, ZAP_CHUNK_TYPE_MAX = 250 } zap_chunk_type_t; /* * TAKE NOTE: * If zap_leaf_phys_t is modified, zap_leaf_byteswap() must be modified. */ typedef struct zap_leaf_phys { struct zap_leaf_header { uint64_t lh_block_type; /* ZBT_LEAF */ uint64_t lh_pad1; uint64_t lh_prefix; /* hash prefix of this leaf */ uint32_t lh_magic; /* ZAP_LEAF_MAGIC */ uint16_t lh_nfree; /* number free chunks */ uint16_t lh_nentries; /* number of entries */ uint16_t lh_prefix_len; /* num bits used to id this */ /* above is accessable to zap, below is zap_leaf private */ uint16_t lh_freelist; /* chunk head of free list */ uint8_t lh_pad2[12]; } l_hdr; /* 2 24-byte chunks */ /* * The header is followed by a hash table with * ZAP_LEAF_HASH_NUMENTRIES(zap) entries. The hash table is * followed by an array of ZAP_LEAF_NUMCHUNKS(zap) * zap_leaf_chunk structures. These structures are accessed * with the ZAP_LEAF_CHUNK() macro. */ uint16_t l_hash[1]; } zap_leaf_phys_t; typedef union zap_leaf_chunk { struct zap_leaf_entry { uint8_t le_type; /* always ZAP_CHUNK_ENTRY */ uint8_t le_value_intlen; /* size of ints */ uint16_t le_next; /* next entry in hash chain */ uint16_t le_name_chunk; /* first chunk of the name */ uint16_t le_name_numints; /* bytes in name, incl null */ uint16_t le_value_chunk; /* first chunk of the value */ uint16_t le_value_numints; /* value length in ints */ uint32_t le_cd; /* collision differentiator */ uint64_t le_hash; /* hash value of the name */ } l_entry; struct zap_leaf_array { uint8_t la_type; /* always ZAP_CHUNK_ARRAY */ uint8_t la_array[ZAP_LEAF_ARRAY_BYTES]; uint16_t la_next; /* next blk or CHAIN_END */ } l_array; struct zap_leaf_free { uint8_t lf_type; /* always ZAP_CHUNK_FREE */ uint8_t lf_pad[ZAP_LEAF_ARRAY_BYTES]; uint16_t lf_next; /* next in free list, or CHAIN_END */ } l_free; } zap_leaf_chunk_t; typedef struct zap_leaf { int l_bs; /* block size shift */ zap_leaf_phys_t *l_phys; } zap_leaf_t; /* * Define special zfs pflags */ #define ZFS_XATTR 0x1 /* is an extended attribute */ #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */ #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */ #define MASTER_NODE_OBJ 1 /* * special attributes for master node. */ #define ZFS_FSID "FSID" #define ZFS_UNLINKED_SET "DELETE_QUEUE" #define ZFS_ROOT_OBJ "ROOT" #define ZPL_VERSION_OBJ "VERSION" #define ZFS_PROP_BLOCKPERPAGE "BLOCKPERPAGE" #define ZFS_PROP_NOGROWBLOCKS "NOGROWBLOCKS" #define ZFS_FLAG_BLOCKPERPAGE 0x1 #define ZFS_FLAG_NOGROWBLOCKS 0x2 /* * ZPL version - rev'd whenever an incompatible on-disk format change * occurs. Independent of SPA/DMU/ZAP versioning. */ #define ZPL_VERSION 1ULL /* * The directory entry has the type (currently unused on Solaris) in the * top 4 bits, and the object number in the low 48 bits. The "middle" * 12 bits are unused. */ #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4) #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48) #define ZFS_DIRENT_MAKE(type, obj) (((uint64_t)type << 60) | obj) typedef struct ace { uid_t a_who; /* uid or gid */ uint32_t a_access_mask; /* read,write,... */ uint16_t a_flags; /* see below */ uint16_t a_type; /* allow or deny */ } ace_t; #define ACE_SLOT_CNT 6 typedef struct zfs_znode_acl { uint64_t z_acl_extern_obj; /* ext acl pieces */ uint32_t z_acl_count; /* Number of ACEs */ uint16_t z_acl_version; /* acl version */ uint16_t z_acl_pad; /* pad */ ace_t z_ace_data[ACE_SLOT_CNT]; /* 6 standard ACEs */ } zfs_znode_acl_t; /* * This is the persistent portion of the znode. It is stored * in the "bonus buffer" of the file. Short symbolic links * are also stored in the bonus buffer. */ typedef struct znode_phys { uint64_t zp_atime[2]; /* 0 - last file access time */ uint64_t zp_mtime[2]; /* 16 - last file modification time */ uint64_t zp_ctime[2]; /* 32 - last file change time */ uint64_t zp_crtime[2]; /* 48 - creation time */ uint64_t zp_gen; /* 64 - generation (txg of creation) */ uint64_t zp_mode; /* 72 - file mode bits */ uint64_t zp_size; /* 80 - size of file */ uint64_t zp_parent; /* 88 - directory parent (`..') */ uint64_t zp_links; /* 96 - number of links to file */ uint64_t zp_xattr; /* 104 - DMU object for xattrs */ uint64_t zp_rdev; /* 112 - dev_t for VBLK & VCHR files */ uint64_t zp_flags; /* 120 - persistent flags */ uint64_t zp_uid; /* 128 - file owner */ uint64_t zp_gid; /* 136 - owning group */ uint64_t zp_pad[4]; /* 144 - future */ zfs_znode_acl_t zp_acl; /* 176 - 263 ACL */ /* * Data may pad out any remaining bytes in the znode buffer, eg: * * |<---------------------- dnode_phys (512) ------------------------>| * |<-- dnode (192) --->|<----------- "bonus" buffer (320) ---------->| * |<---- znode (264) ---->|<---- data (56) ---->| * * At present, we only use this space to store symbolic links. */ } znode_phys_t; /* * In-core vdev representation. */ struct vdev; struct spa; typedef int vdev_phys_read_t(struct vdev *vdev, void *priv, off_t offset, void *buf, size_t bytes); typedef int vdev_read_t(struct vdev *vdev, const blkptr_t *bp, void *buf, off_t offset, size_t bytes); typedef STAILQ_HEAD(vdev_list, vdev) vdev_list_t; typedef struct vdev_indirect_mapping_entry_phys { /* * Decode with DVA_MAPPING_* macros. * Contains: * the source offset (low 63 bits) * the one-bit "mark", used for garbage collection (by zdb) */ uint64_t vimep_src; /* * Note: the DVA's asize is 24 bits, and can thus store ranges * up to 8GB. */ dva_t vimep_dst; } vdev_indirect_mapping_entry_phys_t; #define DVA_MAPPING_GET_SRC_OFFSET(vimep) \ BF64_GET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0) #define DVA_MAPPING_SET_SRC_OFFSET(vimep, x) \ BF64_SET_SB((vimep)->vimep_src, 0, 63, SPA_MINBLOCKSHIFT, 0, x) typedef struct vdev_indirect_mapping_entry { vdev_indirect_mapping_entry_phys_t vime_mapping; uint32_t vime_obsolete_count; list_node_t vime_node; } vdev_indirect_mapping_entry_t; /* * This is stored in the bonus buffer of the mapping object, see comment of * vdev_indirect_config for more details. */ typedef struct vdev_indirect_mapping_phys { uint64_t vimp_max_offset; uint64_t vimp_bytes_mapped; uint64_t vimp_num_entries; /* number of v_i_m_entry_phys_t's */ /* * For each entry in the mapping object, this object contains an * entry representing the number of bytes of that mapping entry * that were no longer in use by the pool at the time this indirect * vdev was last condensed. */ uint64_t vimp_counts_object; } vdev_indirect_mapping_phys_t; #define VDEV_INDIRECT_MAPPING_SIZE_V0 (3 * sizeof (uint64_t)) typedef struct vdev_indirect_mapping { uint64_t vim_object; boolean_t vim_havecounts; /* vim_entries segment offset currently in memory. */ uint64_t vim_entry_offset; /* vim_entries segment size. */ size_t vim_num_entries; /* Needed by dnode_read() */ const void *vim_spa; dnode_phys_t *vim_dn; /* * An ordered array of mapping entries, sorted by source offset. * Note that vim_entries is needed during a removal (and contains * mappings that have been synced to disk so far) to handle frees * from the removing device. */ vdev_indirect_mapping_entry_phys_t *vim_entries; objset_phys_t *vim_objset; vdev_indirect_mapping_phys_t *vim_phys; } vdev_indirect_mapping_t; /* * On-disk indirect vdev state. * * An indirect vdev is described exclusively in the MOS config of a pool. * The config for an indirect vdev includes several fields, which are * accessed in memory by a vdev_indirect_config_t. */ typedef struct vdev_indirect_config { /* * Object (in MOS) which contains the indirect mapping. This object * contains an array of vdev_indirect_mapping_entry_phys_t ordered by * vimep_src. The bonus buffer for this object is a * vdev_indirect_mapping_phys_t. This object is allocated when a vdev * removal is initiated. * * Note that this object can be empty if none of the data on the vdev * has been copied yet. */ uint64_t vic_mapping_object; /* * Object (in MOS) which contains the birth times for the mapping * entries. This object contains an array of * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus * buffer for this object is a vdev_indirect_birth_phys_t. This object * is allocated when a vdev removal is initiated. * * Note that this object can be empty if none of the vdev has yet been * copied. */ uint64_t vic_births_object; /* * This is the vdev ID which was removed previous to this vdev, or * UINT64_MAX if there are no previously removed vdevs. */ uint64_t vic_prev_indirect_vdev; } vdev_indirect_config_t; typedef struct vdev { STAILQ_ENTRY(vdev) v_childlink; /* link in parent's child list */ STAILQ_ENTRY(vdev) v_alllink; /* link in global vdev list */ vdev_list_t v_children; /* children of this vdev */ const char *v_name; /* vdev name */ uint64_t v_guid; /* vdev guid */ uint64_t v_id; /* index in parent */ uint64_t v_psize; /* physical device capacity */ int v_ashift; /* offset to block shift */ int v_nparity; /* # parity for raidz */ struct vdev *v_top; /* parent vdev */ size_t v_nchildren; /* # children */ vdev_state_t v_state; /* current state */ vdev_phys_read_t *v_phys_read; /* read from raw leaf vdev */ vdev_read_t *v_read; /* read from vdev */ void *v_read_priv; /* private data for read function */ boolean_t v_islog; struct spa *v_spa; /* link to spa */ /* * Values stored in the config for an indirect or removing vdev. */ vdev_indirect_config_t vdev_indirect_config; vdev_indirect_mapping_t *v_mapping; } vdev_t; /* * In-core pool representation. */ typedef STAILQ_HEAD(spa_list, spa) spa_list_t; typedef struct spa { STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ char *spa_name; /* pool name */ uint64_t spa_guid; /* pool guid */ uint64_t spa_txg; /* most recent transaction */ struct uberblock spa_uberblock; /* best uberblock so far */ vdev_t *spa_root_vdev; /* toplevel vdev container */ objset_phys_t spa_mos; /* MOS for this pool */ zio_cksum_salt_t spa_cksum_salt; /* secret salt for cksum */ void *spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS]; boolean_t spa_with_log; /* this pool has log */ } spa_t; /* IO related arguments. */ typedef struct zio { spa_t *io_spa; blkptr_t *io_bp; void *io_data; uint64_t io_size; uint64_t io_offset; /* Stuff for the vdev stack */ vdev_t *io_vd; void *io_vsd; int io_error; } zio_t; static void decode_embedded_bp_compressed(const blkptr_t *, void *);