Index: head/sys/boot/efi/boot1/zfs_module.c =================================================================== --- head/sys/boot/efi/boot1/zfs_module.c (revision 316584) +++ head/sys/boot/efi/boot1/zfs_module.c (revision 316585) @@ -1,234 +1,243 @@ /*- * Copyright (c) 2015 Eric McCorkle * 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. * * $FreeBSD$ */ #include #include #include #include #include #include #include #include "boot_module.h" #include "libzfs.h" #include "zfsimpl.c" static dev_info_t *devices; +uint64_t +ldi_get_size(void *priv) +{ + dev_info_t *devinfo = priv; + + return (devinfo->dev->Media->BlockSize * + (devinfo->dev->Media->LastBlock + 1)); +} + static int vdev_read(vdev_t *vdev, void *priv, off_t off, void *buf, size_t bytes) { dev_info_t *devinfo; uint64_t lba; size_t size, remainder, rb_size, blksz; char *bouncebuf = NULL, *rb_buf; EFI_STATUS status; devinfo = (dev_info_t *)priv; lba = off / devinfo->dev->Media->BlockSize; remainder = off % devinfo->dev->Media->BlockSize; rb_buf = buf; rb_size = bytes; /* * If we have remainder from off, we need to add remainder part. * Since buffer must be multiple of the BlockSize, round it all up. */ size = roundup2(bytes + remainder, devinfo->dev->Media->BlockSize); blksz = size; if (remainder != 0 || size != bytes) { rb_size = devinfo->dev->Media->BlockSize; bouncebuf = malloc(rb_size); if (bouncebuf == NULL) { printf("vdev_read: out of memory\n"); return (-1); } rb_buf = bouncebuf; blksz = rb_size - remainder; } while (bytes > 0) { status = devinfo->dev->ReadBlocks(devinfo->dev, devinfo->dev->Media->MediaId, lba, rb_size, rb_buf); if (EFI_ERROR(status)) goto error; if (bytes < blksz) blksz = bytes; if (bouncebuf != NULL) memcpy(buf, rb_buf + remainder, blksz); buf = (void *)((uintptr_t)buf + blksz); bytes -= blksz; lba++; remainder = 0; blksz = rb_size; } free(bouncebuf); return (0); error: free(bouncebuf); DPRINTF("vdev_read: failed dev: %p, id: %u, lba: %ju, size: %zu," " rb_size: %zu, status: %lu\n", devinfo->dev, devinfo->dev->Media->MediaId, (uintmax_t)lba, bytes, rb_size, EFI_ERROR_CODE(status)); return (-1); } static EFI_STATUS probe(dev_info_t *dev) { spa_t *spa; dev_info_t *tdev; EFI_STATUS status; /* ZFS consumes the dev on success so we need a copy. */ if ((status = bs->AllocatePool(EfiLoaderData, sizeof(*dev), (void**)&tdev)) != EFI_SUCCESS) { DPRINTF("Failed to allocate tdev (%lu)\n", EFI_ERROR_CODE(status)); return (status); } memcpy(tdev, dev, sizeof(*dev)); if (vdev_probe(vdev_read, tdev, &spa) != 0) { (void)bs->FreePool(tdev); return (EFI_UNSUPPORTED); } dev->devdata = spa; add_device(&devices, dev); return (EFI_SUCCESS); } static EFI_STATUS load(const char *filepath, dev_info_t *devinfo, void **bufp, size_t *bufsize) { spa_t *spa; struct zfsmount zfsmount; dnode_phys_t dn; struct stat st; int err; void *buf; EFI_STATUS status; spa = devinfo->devdata; DPRINTF("load: '%s' spa: '%s', devpath: %s\n", filepath, spa->spa_name, devpath_str(devinfo->devpath)); if ((err = zfs_spa_init(spa)) != 0) { DPRINTF("Failed to load pool '%s' (%d)\n", spa->spa_name, err); return (EFI_NOT_FOUND); } if ((err = zfs_mount(spa, 0, &zfsmount)) != 0) { DPRINTF("Failed to mount pool '%s' (%d)\n", spa->spa_name, err); return (EFI_NOT_FOUND); } if ((err = zfs_lookup(&zfsmount, filepath, &dn)) != 0) { if (err == ENOENT) { DPRINTF("Failed to find '%s' on pool '%s' (%d)\n", filepath, spa->spa_name, err); return (EFI_NOT_FOUND); } printf("Failed to lookup '%s' on pool '%s' (%d)\n", filepath, spa->spa_name, err); return (EFI_INVALID_PARAMETER); } if ((err = zfs_dnode_stat(spa, &dn, &st)) != 0) { printf("Failed to stat '%s' on pool '%s' (%d)\n", filepath, spa->spa_name, err); return (EFI_INVALID_PARAMETER); } if ((status = bs->AllocatePool(EfiLoaderData, (UINTN)st.st_size, &buf)) != EFI_SUCCESS) { printf("Failed to allocate load buffer %zd for pool '%s' for '%s' " "(%lu)\n", st.st_size, spa->spa_name, filepath, EFI_ERROR_CODE(status)); return (EFI_INVALID_PARAMETER); } if ((err = dnode_read(spa, &dn, 0, buf, st.st_size)) != 0) { printf("Failed to read node from %s (%d)\n", spa->spa_name, err); (void)bs->FreePool(buf); return (EFI_INVALID_PARAMETER); } *bufsize = st.st_size; *bufp = buf; return (EFI_SUCCESS); } static void status(void) { spa_t *spa; spa = STAILQ_FIRST(&zfs_pools); if (spa == NULL) { printf("%s found no pools\n", zfs_module.name); return; } printf("%s found the following pools:", zfs_module.name); STAILQ_FOREACH(spa, &zfs_pools, spa_link) printf(" %s", spa->spa_name); printf("\n"); } static void init(void) { zfs_init(); } static dev_info_t * _devices(void) { return (devices); } const boot_module_t zfs_module = { .name = "ZFS", .init = init, .probe = probe, .load = load, .status = status, .devices = _devices }; Index: head/sys/boot/efi/loader/main.c =================================================================== --- head/sys/boot/efi/loader/main.c (revision 316584) +++ head/sys/boot/efi/loader/main.c (revision 316585) @@ -1,841 +1,852 @@ /*- * Copyright (c) 2008-2010 Rui Paulo * Copyright (c) 2006 Marcel Moolenaar * 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 ``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 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("$FreeBSD$"); +#include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef EFI_ZFS_BOOT #include #endif #include "loader_efi.h" extern char bootprog_info[]; struct arch_switch archsw; /* MI/MD interface boundary */ EFI_GUID acpi = ACPI_TABLE_GUID; EFI_GUID acpi20 = ACPI_20_TABLE_GUID; EFI_GUID devid = DEVICE_PATH_PROTOCOL; EFI_GUID imgid = LOADED_IMAGE_PROTOCOL; EFI_GUID mps = MPS_TABLE_GUID; EFI_GUID netid = EFI_SIMPLE_NETWORK_PROTOCOL; EFI_GUID smbios = SMBIOS_TABLE_GUID; EFI_GUID dxe = DXE_SERVICES_TABLE_GUID; EFI_GUID hoblist = HOB_LIST_TABLE_GUID; EFI_GUID memtype = MEMORY_TYPE_INFORMATION_TABLE_GUID; EFI_GUID debugimg = DEBUG_IMAGE_INFO_TABLE_GUID; EFI_GUID fdtdtb = FDT_TABLE_GUID; EFI_GUID inputid = SIMPLE_TEXT_INPUT_PROTOCOL; #ifdef EFI_ZFS_BOOT static void efi_zfs_probe(void); static uint64_t pool_guid; #endif static int has_keyboard(void) { EFI_STATUS status; EFI_DEVICE_PATH *path; EFI_HANDLE *hin, *hin_end, *walker; UINTN sz; int retval = 0; /* * Find all the handles that support the SIMPLE_TEXT_INPUT_PROTOCOL and * do the typical dance to get the right sized buffer. */ sz = 0; hin = NULL; status = BS->LocateHandle(ByProtocol, &inputid, 0, &sz, 0); if (status == EFI_BUFFER_TOO_SMALL) { hin = (EFI_HANDLE *)malloc(sz); status = BS->LocateHandle(ByProtocol, &inputid, 0, &sz, hin); if (EFI_ERROR(status)) free(hin); } if (EFI_ERROR(status)) return retval; /* * Look at each of the handles. If it supports the device path protocol, * use it to get the device path for this handle. Then see if that * device path matches either the USB device path for keyboards or the * legacy device path for keyboards. */ hin_end = &hin[sz / sizeof(*hin)]; for (walker = hin; walker < hin_end; walker++) { status = BS->HandleProtocol(*walker, &devid, (VOID **)&path); if (EFI_ERROR(status)) continue; while (!IsDevicePathEnd(path)) { /* * Check for the ACPI keyboard node. All PNP3xx nodes * are keyboards of different flavors. Note: It is * unclear of there's always a keyboard node when * there's a keyboard controller, or if there's only one * when a keyboard is detected at boot. */ if (DevicePathType(path) == ACPI_DEVICE_PATH && (DevicePathSubType(path) == ACPI_DP || DevicePathSubType(path) == ACPI_EXTENDED_DP)) { ACPI_HID_DEVICE_PATH *acpi; acpi = (ACPI_HID_DEVICE_PATH *)(void *)path; if ((EISA_ID_TO_NUM(acpi->HID) & 0xff00) == 0x300 && (acpi->HID & 0xffff) == PNP_EISA_ID_CONST) { retval = 1; goto out; } /* * Check for USB keyboard node, if present. Unlike a * PS/2 keyboard, these definitely only appear when * connected to the system. */ } else if (DevicePathType(path) == MESSAGING_DEVICE_PATH && DevicePathSubType(path) == MSG_USB_CLASS_DP) { USB_CLASS_DEVICE_PATH *usb; usb = (USB_CLASS_DEVICE_PATH *)(void *)path; if (usb->DeviceClass == 3 && /* HID */ usb->DeviceSubClass == 1 && /* Boot devices */ usb->DeviceProtocol == 1) { /* Boot keyboards */ retval = 1; goto out; } } path = NextDevicePathNode(path); } } out: free(hin); return retval; } static void set_devdesc_currdev(struct devsw *dev, int unit) { struct devdesc currdev; char *devname; currdev.d_dev = dev; currdev.d_type = currdev.d_dev->dv_type; currdev.d_unit = unit; currdev.d_opendata = NULL; devname = efi_fmtdev(&currdev); env_setenv("currdev", EV_VOLATILE, devname, efi_setcurrdev, env_nounset); env_setenv("loaddev", EV_VOLATILE, devname, env_noset, env_nounset); } static int find_currdev(EFI_LOADED_IMAGE *img) { pdinfo_list_t *pdi_list; pdinfo_t *dp, *pp; EFI_DEVICE_PATH *devpath, *copy; EFI_HANDLE h; char *devname; struct devsw *dev; int unit; uint64_t extra; #ifdef EFI_ZFS_BOOT /* Did efi_zfs_probe() detect the boot pool? */ if (pool_guid != 0) { struct zfs_devdesc currdev; currdev.d_dev = &zfs_dev; currdev.d_unit = 0; currdev.d_type = currdev.d_dev->dv_type; currdev.d_opendata = NULL; currdev.pool_guid = pool_guid; currdev.root_guid = 0; devname = efi_fmtdev(&currdev); env_setenv("currdev", EV_VOLATILE, devname, efi_setcurrdev, env_nounset); env_setenv("loaddev", EV_VOLATILE, devname, env_noset, env_nounset); init_zfs_bootenv(devname); return (0); } #endif /* EFI_ZFS_BOOT */ /* We have device lists for hd, cd, fd, walk them all. */ pdi_list = efiblk_get_pdinfo_list(&efipart_hddev); STAILQ_FOREACH(dp, pdi_list, pd_link) { struct disk_devdesc currdev; currdev.d_dev = &efipart_hddev; currdev.d_type = currdev.d_dev->dv_type; currdev.d_unit = dp->pd_unit; currdev.d_opendata = NULL; currdev.d_slice = -1; currdev.d_partition = -1; if (dp->pd_handle == img->DeviceHandle) { devname = efi_fmtdev(&currdev); env_setenv("currdev", EV_VOLATILE, devname, efi_setcurrdev, env_nounset); env_setenv("loaddev", EV_VOLATILE, devname, env_noset, env_nounset); return (0); } /* Assuming GPT partitioning. */ STAILQ_FOREACH(pp, &dp->pd_part, pd_link) { if (pp->pd_handle == img->DeviceHandle) { currdev.d_slice = pp->pd_unit; currdev.d_partition = 255; devname = efi_fmtdev(&currdev); env_setenv("currdev", EV_VOLATILE, devname, efi_setcurrdev, env_nounset); env_setenv("loaddev", EV_VOLATILE, devname, env_noset, env_nounset); return (0); } } } pdi_list = efiblk_get_pdinfo_list(&efipart_cddev); STAILQ_FOREACH(dp, pdi_list, pd_link) { if (dp->pd_handle == img->DeviceHandle || dp->pd_alias == img->DeviceHandle) { set_devdesc_currdev(&efipart_cddev, dp->pd_unit); return (0); } } pdi_list = efiblk_get_pdinfo_list(&efipart_fddev); STAILQ_FOREACH(dp, pdi_list, pd_link) { if (dp->pd_handle == img->DeviceHandle) { set_devdesc_currdev(&efipart_fddev, dp->pd_unit); return (0); } } /* * Try the device handle from our loaded image first. If that * fails, use the device path from the loaded image and see if * any of the nodes in that path match one of the enumerated * handles. */ if (efi_handle_lookup(img->DeviceHandle, &dev, &unit, &extra) == 0) { set_devdesc_currdev(dev, unit); return (0); } copy = NULL; devpath = efi_lookup_image_devpath(IH); while (devpath != NULL) { h = efi_devpath_handle(devpath); if (h == NULL) break; free(copy); copy = NULL; if (efi_handle_lookup(h, &dev, &unit, &extra) == 0) { set_devdesc_currdev(dev, unit); return (0); } devpath = efi_lookup_devpath(h); if (devpath != NULL) { copy = efi_devpath_trim(devpath); devpath = copy; } } free(copy); return (ENOENT); } EFI_STATUS main(int argc, CHAR16 *argv[]) { char var[128]; EFI_LOADED_IMAGE *img; EFI_GUID *guid; int i, j, vargood, howto; UINTN k; int has_kbd; char buf[40]; archsw.arch_autoload = efi_autoload; archsw.arch_getdev = efi_getdev; archsw.arch_copyin = efi_copyin; archsw.arch_copyout = efi_copyout; archsw.arch_readin = efi_readin; #ifdef EFI_ZFS_BOOT /* Note this needs to be set before ZFS init. */ archsw.arch_zfs_probe = efi_zfs_probe; #endif /* Init the time source */ efi_time_init(); has_kbd = has_keyboard(); /* * XXX Chicken-and-egg problem; we want to have console output * early, but some console attributes may depend on reading from * eg. the boot device, which we can't do yet. We can use * printf() etc. once this is done. */ cons_probe(); /* * Initialise the block cache. Set the upper limit. */ bcache_init(32768, 512); /* * Parse the args to set the console settings, etc * boot1.efi passes these in, if it can read /boot.config or /boot/config * or iPXE may be setup to pass these in. * * Loop through the args, and for each one that contains an '=' that is * not the first character, add it to the environment. This allows * loader and kernel env vars to be passed on the command line. Convert * args from UCS-2 to ASCII (16 to 8 bit) as they are copied. */ howto = 0; for (i = 1; i < argc; i++) { if (argv[i][0] == '-') { for (j = 1; argv[i][j] != 0; j++) { int ch; ch = argv[i][j]; switch (ch) { case 'a': howto |= RB_ASKNAME; break; case 'd': howto |= RB_KDB; break; case 'D': howto |= RB_MULTIPLE; break; case 'h': howto |= RB_SERIAL; break; case 'm': howto |= RB_MUTE; break; case 'p': howto |= RB_PAUSE; break; case 'P': if (!has_kbd) howto |= RB_SERIAL | RB_MULTIPLE; break; case 'r': howto |= RB_DFLTROOT; break; case 's': howto |= RB_SINGLE; break; case 'S': if (argv[i][j + 1] == 0) { if (i + 1 == argc) { setenv("comconsole_speed", "115200", 1); } else { cpy16to8(&argv[i + 1][0], var, sizeof(var)); setenv("comconsole_speedspeed", var, 1); } i++; break; } else { cpy16to8(&argv[i][j + 1], var, sizeof(var)); setenv("comconsole_speed", var, 1); break; } case 'v': howto |= RB_VERBOSE; break; } } } else { vargood = 0; for (j = 0; argv[i][j] != 0; j++) { if (j == sizeof(var)) { vargood = 0; break; } if (j > 0 && argv[i][j] == '=') vargood = 1; var[j] = (char)argv[i][j]; } if (vargood) { var[j] = 0; putenv(var); } } } for (i = 0; howto_names[i].ev != NULL; i++) if (howto & howto_names[i].mask) setenv(howto_names[i].ev, "YES", 1); if (howto & RB_MULTIPLE) { if (howto & RB_SERIAL) setenv("console", "comconsole efi" , 1); else setenv("console", "efi comconsole" , 1); } else if (howto & RB_SERIAL) { setenv("console", "comconsole" , 1); } if (efi_copy_init()) { printf("failed to allocate staging area\n"); return (EFI_BUFFER_TOO_SMALL); } /* * March through the device switch probing for things. */ for (i = 0; devsw[i] != NULL; i++) if (devsw[i]->dv_init != NULL) (devsw[i]->dv_init)(); /* Get our loaded image protocol interface structure. */ BS->HandleProtocol(IH, &imgid, (VOID**)&img); printf("Command line arguments:"); for (i = 0; i < argc; i++) printf(" %S", argv[i]); printf("\n"); printf("Image base: 0x%lx\n", (u_long)img->ImageBase); printf("EFI version: %d.%02d\n", ST->Hdr.Revision >> 16, ST->Hdr.Revision & 0xffff); printf("EFI Firmware: %S (rev %d.%02d)\n", ST->FirmwareVendor, ST->FirmwareRevision >> 16, ST->FirmwareRevision & 0xffff); printf("\n%s", bootprog_info); /* * Disable the watchdog timer. By default the boot manager sets * the timer to 5 minutes before invoking a boot option. If we * want to return to the boot manager, we have to disable the * watchdog timer and since we're an interactive program, we don't * want to wait until the user types "quit". The timer may have * fired by then. We don't care if this fails. It does not prevent * normal functioning in any way... */ BS->SetWatchdogTimer(0, 0, 0, NULL); if (find_currdev(img) != 0) return (EFI_NOT_FOUND); efi_init_environment(); setenv("LINES", "24", 1); /* optional */ for (k = 0; k < ST->NumberOfTableEntries; k++) { guid = &ST->ConfigurationTable[k].VendorGuid; if (!memcmp(guid, &smbios, sizeof(EFI_GUID))) { snprintf(buf, sizeof(buf), "%p", ST->ConfigurationTable[k].VendorTable); setenv("hint.smbios.0.mem", buf, 1); smbios_detect(ST->ConfigurationTable[k].VendorTable); break; } } interact(NULL); /* doesn't return */ return (EFI_SUCCESS); /* keep compiler happy */ } COMMAND_SET(reboot, "reboot", "reboot the system", command_reboot); static int command_reboot(int argc, char *argv[]) { int i; for (i = 0; devsw[i] != NULL; ++i) if (devsw[i]->dv_cleanup != NULL) (devsw[i]->dv_cleanup)(); RS->ResetSystem(EfiResetCold, EFI_SUCCESS, 23, (CHAR16 *)"Reboot from the loader"); /* NOTREACHED */ return (CMD_ERROR); } COMMAND_SET(quit, "quit", "exit the loader", command_quit); static int command_quit(int argc, char *argv[]) { exit(0); return (CMD_OK); } COMMAND_SET(memmap, "memmap", "print memory map", command_memmap); static int command_memmap(int argc, char *argv[]) { UINTN sz; EFI_MEMORY_DESCRIPTOR *map, *p; UINTN key, dsz; UINT32 dver; EFI_STATUS status; int i, ndesc; char line[80]; static char *types[] = { "Reserved", "LoaderCode", "LoaderData", "BootServicesCode", "BootServicesData", "RuntimeServicesCode", "RuntimeServicesData", "ConventionalMemory", "UnusableMemory", "ACPIReclaimMemory", "ACPIMemoryNVS", "MemoryMappedIO", "MemoryMappedIOPortSpace", "PalCode" }; sz = 0; status = BS->GetMemoryMap(&sz, 0, &key, &dsz, &dver); if (status != EFI_BUFFER_TOO_SMALL) { printf("Can't determine memory map size\n"); return (CMD_ERROR); } map = malloc(sz); status = BS->GetMemoryMap(&sz, map, &key, &dsz, &dver); if (EFI_ERROR(status)) { printf("Can't read memory map\n"); return (CMD_ERROR); } ndesc = sz / dsz; snprintf(line, sizeof(line), "%23s %12s %12s %8s %4s\n", "Type", "Physical", "Virtual", "#Pages", "Attr"); pager_open(); if (pager_output(line)) { pager_close(); return (CMD_OK); } for (i = 0, p = map; i < ndesc; i++, p = NextMemoryDescriptor(p, dsz)) { printf("%23s %012jx %012jx %08jx ", types[p->Type], (uintmax_t)p->PhysicalStart, (uintmax_t)p->VirtualStart, (uintmax_t)p->NumberOfPages); if (p->Attribute & EFI_MEMORY_UC) printf("UC "); if (p->Attribute & EFI_MEMORY_WC) printf("WC "); if (p->Attribute & EFI_MEMORY_WT) printf("WT "); if (p->Attribute & EFI_MEMORY_WB) printf("WB "); if (p->Attribute & EFI_MEMORY_UCE) printf("UCE "); if (p->Attribute & EFI_MEMORY_WP) printf("WP "); if (p->Attribute & EFI_MEMORY_RP) printf("RP "); if (p->Attribute & EFI_MEMORY_XP) printf("XP "); if (pager_output("\n")) break; } pager_close(); return (CMD_OK); } COMMAND_SET(configuration, "configuration", "print configuration tables", command_configuration); static const char * guid_to_string(EFI_GUID *guid) { static char buf[40]; sprintf(buf, "%08x-%04x-%04x-%02x%02x-%02x%02x%02x%02x%02x%02x", guid->Data1, guid->Data2, guid->Data3, guid->Data4[0], guid->Data4[1], guid->Data4[2], guid->Data4[3], guid->Data4[4], guid->Data4[5], guid->Data4[6], guid->Data4[7]); return (buf); } static int command_configuration(int argc, char *argv[]) { char line[80]; UINTN i; snprintf(line, sizeof(line), "NumberOfTableEntries=%lu\n", (unsigned long)ST->NumberOfTableEntries); pager_open(); if (pager_output(line)) { pager_close(); return (CMD_OK); } for (i = 0; i < ST->NumberOfTableEntries; i++) { EFI_GUID *guid; printf(" "); guid = &ST->ConfigurationTable[i].VendorGuid; if (!memcmp(guid, &mps, sizeof(EFI_GUID))) printf("MPS Table"); else if (!memcmp(guid, &acpi, sizeof(EFI_GUID))) printf("ACPI Table"); else if (!memcmp(guid, &acpi20, sizeof(EFI_GUID))) printf("ACPI 2.0 Table"); else if (!memcmp(guid, &smbios, sizeof(EFI_GUID))) printf("SMBIOS Table %p", ST->ConfigurationTable[i].VendorTable); else if (!memcmp(guid, &dxe, sizeof(EFI_GUID))) printf("DXE Table"); else if (!memcmp(guid, &hoblist, sizeof(EFI_GUID))) printf("HOB List Table"); else if (!memcmp(guid, &memtype, sizeof(EFI_GUID))) printf("Memory Type Information Table"); else if (!memcmp(guid, &debugimg, sizeof(EFI_GUID))) printf("Debug Image Info Table"); else if (!memcmp(guid, &fdtdtb, sizeof(EFI_GUID))) printf("FDT Table"); else printf("Unknown Table (%s)", guid_to_string(guid)); snprintf(line, sizeof(line), " at %p\n", ST->ConfigurationTable[i].VendorTable); if (pager_output(line)) break; } pager_close(); return (CMD_OK); } COMMAND_SET(mode, "mode", "change or display EFI text modes", command_mode); static int command_mode(int argc, char *argv[]) { UINTN cols, rows; unsigned int mode; int i; char *cp; char rowenv[8]; EFI_STATUS status; SIMPLE_TEXT_OUTPUT_INTERFACE *conout; extern void HO(void); conout = ST->ConOut; if (argc > 1) { mode = strtol(argv[1], &cp, 0); if (cp[0] != '\0') { printf("Invalid mode\n"); return (CMD_ERROR); } status = conout->QueryMode(conout, mode, &cols, &rows); if (EFI_ERROR(status)) { printf("invalid mode %d\n", mode); return (CMD_ERROR); } status = conout->SetMode(conout, mode); if (EFI_ERROR(status)) { printf("couldn't set mode %d\n", mode); return (CMD_ERROR); } sprintf(rowenv, "%u", (unsigned)rows); setenv("LINES", rowenv, 1); HO(); /* set cursor */ return (CMD_OK); } printf("Current mode: %d\n", conout->Mode->Mode); for (i = 0; i <= conout->Mode->MaxMode; i++) { status = conout->QueryMode(conout, i, &cols, &rows); if (EFI_ERROR(status)) continue; printf("Mode %d: %u columns, %u rows\n", i, (unsigned)cols, (unsigned)rows); } if (i != 0) printf("Select a mode with the command \"mode \"\n"); return (CMD_OK); } #ifdef EFI_ZFS_BOOT COMMAND_SET(lszfs, "lszfs", "list child datasets of a zfs dataset", command_lszfs); static int command_lszfs(int argc, char *argv[]) { int err; if (argc != 2) { command_errmsg = "wrong number of arguments"; return (CMD_ERROR); } err = zfs_list(argv[1]); if (err != 0) { command_errmsg = strerror(err); return (CMD_ERROR); } return (CMD_OK); } COMMAND_SET(reloadbe, "reloadbe", "refresh the list of ZFS Boot Environments", command_reloadbe); static int command_reloadbe(int argc, char *argv[]) { int err; char *root; if (argc > 2) { command_errmsg = "wrong number of arguments"; return (CMD_ERROR); } if (argc == 2) { err = zfs_bootenv(argv[1]); } else { root = getenv("zfs_be_root"); if (root == NULL) { return (CMD_OK); } err = zfs_bootenv(root); } if (err != 0) { command_errmsg = strerror(err); return (CMD_ERROR); } return (CMD_OK); } #endif #ifdef LOADER_FDT_SUPPORT extern int command_fdt_internal(int argc, char *argv[]); /* * Since proper fdt command handling function is defined in fdt_loader_cmd.c, * and declaring it as extern is in contradiction with COMMAND_SET() macro * (which uses static pointer), we're defining wrapper function, which * calls the proper fdt handling routine. */ static int command_fdt(int argc, char *argv[]) { return (command_fdt_internal(argc, argv)); } COMMAND_SET(fdt, "fdt", "flattened device tree handling", command_fdt); #endif #ifdef EFI_ZFS_BOOT static void efi_zfs_probe(void) { pdinfo_list_t *hdi; pdinfo_t *hd, *pd = NULL; EFI_GUID imgid = LOADED_IMAGE_PROTOCOL; EFI_LOADED_IMAGE *img; EFI_HANDLE boot_disk = NULL; char devname[SPECNAMELEN + 1]; uint64_t *guidp = NULL; BS->HandleProtocol(IH, &imgid, (VOID**)&img); /* Find the handle for the boot disk. */ hdi = efiblk_get_pdinfo_list(&efipart_hddev); STAILQ_FOREACH(hd, hdi, pd_link) { STAILQ_FOREACH(pd, &hd->pd_part, pd_link) { if (pd->pd_handle == img->DeviceHandle) boot_disk = hd->pd_handle; } } /* * We provide non-NULL guid pointer if the disk was used for boot, * and reset after the first found pool. * Technically this solution is not very correct, we assume the boot * pool is the first pool on this disk. */ STAILQ_FOREACH(hd, hdi, pd_link) { if (hd->pd_handle == boot_disk) guidp = &pool_guid; STAILQ_FOREACH(pd, &hd->pd_part, pd_link) { snprintf(devname, sizeof(devname), "%s%dp%d:", efipart_hddev.dv_name, hd->pd_unit, pd->pd_unit); (void) zfs_probe_dev(devname, guidp); if (guidp != NULL && pool_guid != 0) guidp = NULL; } } +} + +uint64_t +ldi_get_size(void *priv) +{ + int fd = (uintptr_t) priv; + uint64_t size; + + ioctl(fd, DIOCGMEDIASIZE, &size); + return (size); } #endif Index: head/sys/boot/i386/common/drv.h =================================================================== --- head/sys/boot/i386/common/drv.h (revision 316584) +++ head/sys/boot/i386/common/drv.h (revision 316585) @@ -1,48 +1,48 @@ /*- * Copyright (c) 2010 Pawel Jakub Dawidek * 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 AUTHORS 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 AUTHORS 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. * * $FreeBSD$ */ #ifndef _DRV_H_ #define _DRV_H_ struct dsk { unsigned int drive; unsigned int type; unsigned int unit; unsigned int slice; int part; daddr_t start; - int init; + uint64_t size; }; int drvread(struct dsk *dskp, void *buf, daddr_t lba, unsigned nblk); #if defined(GPT) || defined(ZFS) int drvwrite(struct dsk *dskp, void *buf, daddr_t lba, unsigned nblk); #endif /* GPT || ZFS */ uint64_t drvsize(struct dsk *dskp); #endif /* !_DRV_H_ */ Index: head/sys/boot/i386/loader/main.c =================================================================== --- head/sys/boot/i386/loader/main.c (revision 316584) +++ head/sys/boot/i386/loader/main.c (revision 316585) @@ -1,466 +1,477 @@ /*- * Copyright (c) 1998 Michael Smith * 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("$FreeBSD$"); /* * MD bootstrap main() and assorted miscellaneous * commands. */ #include #include #include #include #include #include +#include #include #include #include "bootstrap.h" #include "common/bootargs.h" #include "libi386/libi386.h" #include "libi386/smbios.h" #include "btxv86.h" #ifdef LOADER_ZFS_SUPPORT #include "../zfs/libzfs.h" #endif CTASSERT(sizeof(struct bootargs) == BOOTARGS_SIZE); CTASSERT(offsetof(struct bootargs, bootinfo) == BA_BOOTINFO); CTASSERT(offsetof(struct bootargs, bootflags) == BA_BOOTFLAGS); CTASSERT(offsetof(struct bootinfo, bi_size) == BI_SIZE); /* Arguments passed in from the boot1/boot2 loader */ static struct bootargs *kargs; static u_int32_t initial_howto; static u_int32_t initial_bootdev; static struct bootinfo *initial_bootinfo; struct arch_switch archsw; /* MI/MD interface boundary */ static void extract_currdev(void); static int isa_inb(int port); static void isa_outb(int port, int value); void exit(int code); #ifdef LOADER_GELI_SUPPORT #include "geliboot.h" struct geli_boot_args *gargs; #endif #ifdef LOADER_ZFS_SUPPORT struct zfs_boot_args *zargs; static void i386_zfs_probe(void); #endif /* from vers.c */ extern char bootprog_info[]; /* XXX debugging */ extern char end[]; static void *heap_top; static void *heap_bottom; int main(void) { int i; /* Pick up arguments */ kargs = (void *)__args; initial_howto = kargs->howto; initial_bootdev = kargs->bootdev; initial_bootinfo = kargs->bootinfo ? (struct bootinfo *)PTOV(kargs->bootinfo) : NULL; /* Initialize the v86 register set to a known-good state. */ bzero(&v86, sizeof(v86)); v86.efl = PSL_RESERVED_DEFAULT | PSL_I; /* * Initialise the heap as early as possible. Once this is done, malloc() is usable. */ bios_getmem(); #if defined(LOADER_BZIP2_SUPPORT) || defined(LOADER_FIREWIRE_SUPPORT) || \ defined(LOADER_GPT_SUPPORT) || defined(LOADER_ZFS_SUPPORT) if (high_heap_size > 0) { heap_top = PTOV(high_heap_base + high_heap_size); heap_bottom = PTOV(high_heap_base); if (high_heap_base < memtop_copyin) memtop_copyin = high_heap_base; } else #endif { heap_top = (void *)PTOV(bios_basemem); heap_bottom = (void *)end; } setheap(heap_bottom, heap_top); /* * XXX Chicken-and-egg problem; we want to have console output early, but some * console attributes may depend on reading from eg. the boot device, which we * can't do yet. * * We can use printf() etc. once this is done. * If the previous boot stage has requested a serial console, prefer that. */ bi_setboothowto(initial_howto); if (initial_howto & RB_MULTIPLE) { if (initial_howto & RB_SERIAL) setenv("console", "comconsole vidconsole", 1); else setenv("console", "vidconsole comconsole", 1); } else if (initial_howto & RB_SERIAL) setenv("console", "comconsole", 1); else if (initial_howto & RB_MUTE) setenv("console", "nullconsole", 1); cons_probe(); /* * Initialise the block cache. Set the upper limit. */ bcache_init(32768, 512); /* * Special handling for PXE and CD booting. */ if (kargs->bootinfo == 0) { /* * We only want the PXE disk to try to init itself in the below * walk through devsw if we actually booted off of PXE. */ if (kargs->bootflags & KARGS_FLAGS_PXE) pxe_enable(kargs->pxeinfo ? PTOV(kargs->pxeinfo) : NULL); else if (kargs->bootflags & KARGS_FLAGS_CD) bc_add(initial_bootdev); } archsw.arch_autoload = i386_autoload; archsw.arch_getdev = i386_getdev; archsw.arch_copyin = i386_copyin; archsw.arch_copyout = i386_copyout; archsw.arch_readin = i386_readin; archsw.arch_isainb = isa_inb; archsw.arch_isaoutb = isa_outb; #ifdef LOADER_ZFS_SUPPORT archsw.arch_zfs_probe = i386_zfs_probe; #ifdef LOADER_GELI_SUPPORT if ((kargs->bootflags & KARGS_FLAGS_EXTARG) != 0) { zargs = (struct zfs_boot_args *)(kargs + 1); if (zargs != NULL && zargs->size >= offsetof(struct zfs_boot_args, gelipw)) { if (zargs->size >= offsetof(struct zfs_boot_args, keybuf_sentinel) && zargs->keybuf_sentinel == KEYBUF_SENTINEL) { geli_save_keybuf(zargs->keybuf); } if (zargs->gelipw[0] != '\0') { setenv("kern.geom.eli.passphrase", zargs->gelipw, 1); explicit_bzero(zargs->gelipw, sizeof(zargs->gelipw)); } } } #endif /* LOADER_GELI_SUPPORT */ #else /* !LOADER_ZFS_SUPPORT */ #ifdef LOADER_GELI_SUPPORT if ((kargs->bootflags & KARGS_FLAGS_EXTARG) != 0) { gargs = (struct geli_boot_args *)(kargs + 1); if (gargs != NULL && gargs->size >= offsetof(struct geli_boot_args, gelipw)) { if (gargs->keybuf_sentinel == KEYBUF_SENTINEL) { geli_save_keybuf(gargs->keybuf); } if (gargs->gelipw[0] != '\0') { setenv("kern.geom.eli.passphrase", gargs->gelipw, 1); explicit_bzero(gargs->gelipw, sizeof(gargs->gelipw)); } } } #endif /* LOADER_GELI_SUPPORT */ #endif /* LOADER_ZFS_SUPPORT */ /* * March through the device switch probing for things. */ for (i = 0; devsw[i] != NULL; i++) if (devsw[i]->dv_init != NULL) (devsw[i]->dv_init)(); printf("BIOS %dkB/%dkB available memory\n", bios_basemem / 1024, bios_extmem / 1024); if (initial_bootinfo != NULL) { initial_bootinfo->bi_basemem = bios_basemem / 1024; initial_bootinfo->bi_extmem = bios_extmem / 1024; } /* detect ACPI for future reference */ biosacpi_detect(); /* detect SMBIOS for future reference */ smbios_detect(NULL); /* detect PCI BIOS for future reference */ biospci_detect(); printf("\n%s", bootprog_info); extract_currdev(); /* set $currdev and $loaddev */ setenv("LINES", "24", 1); /* optional */ bios_getsmap(); interact(NULL); /* if we ever get here, it is an error */ return (1); } /* * Set the 'current device' by (if possible) recovering the boot device as * supplied by the initial bootstrap. * * XXX should be extended for netbooting. */ static void extract_currdev(void) { struct i386_devdesc new_currdev; #ifdef LOADER_ZFS_SUPPORT char buf[20]; #endif int biosdev = -1; /* Assume we are booting from a BIOS disk by default */ new_currdev.d_dev = &biosdisk; /* new-style boot loaders such as pxeldr and cdldr */ if (kargs->bootinfo == 0) { if ((kargs->bootflags & KARGS_FLAGS_CD) != 0) { /* we are booting from a CD with cdboot */ new_currdev.d_dev = &bioscd; new_currdev.d_unit = bc_bios2unit(initial_bootdev); } else if ((kargs->bootflags & KARGS_FLAGS_PXE) != 0) { /* we are booting from pxeldr */ new_currdev.d_dev = &pxedisk; new_currdev.d_unit = 0; } else { /* we don't know what our boot device is */ new_currdev.d_kind.biosdisk.slice = -1; new_currdev.d_kind.biosdisk.partition = 0; biosdev = -1; } #ifdef LOADER_ZFS_SUPPORT } else if ((kargs->bootflags & KARGS_FLAGS_ZFS) != 0) { zargs = NULL; /* check for new style extended argument */ if ((kargs->bootflags & KARGS_FLAGS_EXTARG) != 0) zargs = (struct zfs_boot_args *)(kargs + 1); if (zargs != NULL && zargs->size >= offsetof(struct zfs_boot_args, primary_pool)) { /* sufficient data is provided */ new_currdev.d_kind.zfs.pool_guid = zargs->pool; new_currdev.d_kind.zfs.root_guid = zargs->root; if (zargs->size >= sizeof(*zargs) && zargs->primary_vdev != 0) { sprintf(buf, "%llu", zargs->primary_pool); setenv("vfs.zfs.boot.primary_pool", buf, 1); sprintf(buf, "%llu", zargs->primary_vdev); setenv("vfs.zfs.boot.primary_vdev", buf, 1); } } else { /* old style zfsboot block */ new_currdev.d_kind.zfs.pool_guid = kargs->zfspool; new_currdev.d_kind.zfs.root_guid = 0; } new_currdev.d_dev = &zfs_dev; #endif } else if ((initial_bootdev & B_MAGICMASK) != B_DEVMAGIC) { /* The passed-in boot device is bad */ new_currdev.d_kind.biosdisk.slice = -1; new_currdev.d_kind.biosdisk.partition = 0; biosdev = -1; } else { new_currdev.d_kind.biosdisk.slice = B_SLICE(initial_bootdev) - 1; new_currdev.d_kind.biosdisk.partition = B_PARTITION(initial_bootdev); biosdev = initial_bootinfo->bi_bios_dev; /* * If we are booted by an old bootstrap, we have to guess at the BIOS * unit number. We will lose if there is more than one disk type * and we are not booting from the lowest-numbered disk type * (ie. SCSI when IDE also exists). */ if ((biosdev == 0) && (B_TYPE(initial_bootdev) != 2)) /* biosdev doesn't match major */ biosdev = 0x80 + B_UNIT(initial_bootdev); /* assume harddisk */ } new_currdev.d_type = new_currdev.d_dev->dv_type; /* * If we are booting off of a BIOS disk and we didn't succeed in determining * which one we booted off of, just use disk0: as a reasonable default. */ if ((new_currdev.d_type == biosdisk.dv_type) && ((new_currdev.d_unit = bd_bios2unit(biosdev)) == -1)) { printf("Can't work out which disk we are booting from.\n" "Guessed BIOS device 0x%x not found by probes, defaulting to disk0:\n", biosdev); new_currdev.d_unit = 0; } #ifdef LOADER_ZFS_SUPPORT if (new_currdev.d_type == DEVT_ZFS) init_zfs_bootenv(zfs_fmtdev(&new_currdev)); #endif env_setenv("currdev", EV_VOLATILE, i386_fmtdev(&new_currdev), i386_setcurrdev, env_nounset); env_setenv("loaddev", EV_VOLATILE, i386_fmtdev(&new_currdev), env_noset, env_nounset); } COMMAND_SET(reboot, "reboot", "reboot the system", command_reboot); static int command_reboot(int argc, char *argv[]) { int i; for (i = 0; devsw[i] != NULL; ++i) if (devsw[i]->dv_cleanup != NULL) (devsw[i]->dv_cleanup)(); printf("Rebooting...\n"); delay(1000000); __exit(0); } /* provide this for panic, as it's not in the startup code */ void exit(int code) { __exit(code); } COMMAND_SET(heap, "heap", "show heap usage", command_heap); static int command_heap(int argc, char *argv[]) { mallocstats(); printf("heap base at %p, top at %p, upper limit at %p\n", heap_bottom, sbrk(0), heap_top); return(CMD_OK); } #ifdef LOADER_ZFS_SUPPORT COMMAND_SET(lszfs, "lszfs", "list child datasets of a zfs dataset", command_lszfs); static int command_lszfs(int argc, char *argv[]) { int err; if (argc != 2) { command_errmsg = "wrong number of arguments"; return (CMD_ERROR); } err = zfs_list(argv[1]); if (err != 0) { command_errmsg = strerror(err); return (CMD_ERROR); } return (CMD_OK); } COMMAND_SET(reloadbe, "reloadbe", "refresh the list of ZFS Boot Environments", command_reloadbe); static int command_reloadbe(int argc, char *argv[]) { int err; char *root; if (argc > 2) { command_errmsg = "wrong number of arguments"; return (CMD_ERROR); } if (argc == 2) { err = zfs_bootenv(argv[1]); } else { root = getenv("zfs_be_root"); if (root == NULL) { /* There does not appear to be a ZFS pool here, exit without error */ return (CMD_OK); } err = zfs_bootenv(root); } if (err != 0) { command_errmsg = strerror(err); return (CMD_ERROR); } return (CMD_OK); } #endif /* ISA bus access functions for PnP. */ static int isa_inb(int port) { return (inb(port)); } static void isa_outb(int port, int value) { outb(port, value); } #ifdef LOADER_ZFS_SUPPORT static void i386_zfs_probe(void) { char devname[32]; int unit; /* * Open all the disks we can find and see if we can reconstruct * ZFS pools from them. */ for (unit = 0; unit < MAXBDDEV; unit++) { if (bd_unit2bios(unit) == -1) break; sprintf(devname, "disk%d:", unit); zfs_probe_dev(devname, NULL); } +} + +uint64_t +ldi_get_size(void *priv) +{ + int fd = (uintptr_t) priv; + uint64_t size; + + ioctl(fd, DIOCGMEDIASIZE, &size); + return (size); } #endif Index: head/sys/boot/i386/zfsboot/zfsboot.c =================================================================== --- head/sys/boot/i386/zfsboot/zfsboot.c (revision 316584) +++ head/sys/boot/i386/zfsboot/zfsboot.c (revision 316585) @@ -1,1080 +1,1099 @@ /*- * 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 #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 "util.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; 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; static struct zfsmount zfsmount; 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); void *malloc(size_t n); void free(void *ptr); void * malloc(size_t n) { char *p = heap_next; if (p + n > heap_end) { printf("malloc failure\n"); for (;;) ; /* NOTREACHED */ return (0); } heap_next += n; return (p); } void free(void *ptr) { return; } static char * strdup(const char *s) { char *p = malloc(strlen(s) + 1); strcpy(p, s); return (p); } #ifdef LOADER_GELI_SUPPORT #include "geliboot.c" static char gelipw[GELI_PW_MAXLEN]; static struct keybuf *gelibuf; #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(vdev_t *vdev, void *priv, off_t off, void *buf, size_t bytes) { char *p; daddr_t lba, alignlba; off_t diff; unsigned int nb, alignnb; struct dsk *dsk = (struct dsk *) priv; if ((off & (DEV_BSIZE - 1)) || (bytes & (DEV_BSIZE - 1))) return -1; p = buf; lba = off / DEV_BSIZE; lba += 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 += 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 (drvread(dsk, dmadat->rdbuf, alignlba, alignnb)) return -1; #ifdef LOADER_GELI_SUPPORT /* decrypt */ if (is_geli(dsk) == 0) { if (geli_read(dsk, ((alignlba - 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; } 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 dsk *dsk = (struct dsk *) priv; if ((off & (DEV_BSIZE - 1)) || (bytes & (DEV_BSIZE - 1))) return -1; p = buf; lba = off / DEV_BSIZE; lba += 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(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; off_t off = offsetof(vdev_label_t, vl_pad2); if (size > VDEV_PAD_SIZE) size = VDEV_PAD_SIZE; 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); } static int vdev_clear_pad2(vdev_t *vdev) { char *zeroes = zap_scratch; uint64_t *end; off_t off = offsetof(vdev_label_t, vl_pad2); 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); } 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 dsk * copy_dsk(struct dsk *dsk) { struct dsk *newdsk; newdsk = malloc(sizeof(struct dsk)); *newdsk = *dsk; return (newdsk); } +/* + * 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 dsk *dskp = priv; + uint64_t size = dskp->size; + + if (dskp->start == 0) + size = drvsize(dskp); + + return (size * DEV_BSIZE); +} + static void probe_drive(struct dsk *dsk) { #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; /* * If we find a vdev on the whole disk, stop here. */ if (vdev_probe(vdev_read, dsk, NULL) == 0) return; #ifdef LOADER_GELI_SUPPORT /* * Taste the disk, if it is GELI encrypted, decrypt it and check to see if * it is a usable vdev then. Otherwise dig * out the partition table and probe each slice/partition * in turn for a vdev or GELI encrypted vdev. */ elba = drvsize(dsk); if (elba > 0) { elba--; } if (geli_taste(vdev_read, dsk, elba) == 0) { if (geli_havekey(dsk) == 0 || geli_passphrase(&gelipw, dsk->unit, ':', 0, dsk) == 0) { if (vdev_probe(vdev_read, dsk, NULL) == 0) { return; } } } #endif /* LOADER_GELI_SUPPORT */ sec = dmadat->secbuf; dsk->start = 0; #ifdef GPT /* * First check for GPT. */ if (drvread(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) { dsk->start = 0; if (drvread(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) { dsk->start = ent->ent_lba_start; + dsk->size = ent->ent_lba_end - ent->ent_lba_start + 1; dsk->slice = part + 1; dsk->part = 255; if (vdev_probe(vdev_read, dsk, NULL) == 0) { /* * This slice had a vdev. We need a new dsk * structure now since the vdev now owns this one. */ dsk = copy_dsk(dsk); } #ifdef LOADER_GELI_SUPPORT else if (geli_taste(vdev_read, dsk, ent->ent_lba_end - ent->ent_lba_start) == 0) { if (geli_havekey(dsk) == 0 || geli_passphrase(&gelipw, dsk->unit, 'p', dsk->slice, dsk) == 0) { /* * This slice has GELI, check it for ZFS. */ if (vdev_probe(vdev_read, dsk, NULL) == 0) { /* * This slice had a vdev. We need a new dsk * structure now since the vdev now owns this one. */ dsk = copy_dsk(dsk); } break; } } #endif /* LOADER_GELI_SUPPORT */ } } slba++; } return; trymbr: #endif /* GPT */ if (drvread(dsk, sec, DOSBBSECTOR, 1)) return; dp = (void *)(sec + DOSPARTOFF); for (i = 0; i < NDOSPART; i++) { if (!dp[i].dp_typ) continue; dsk->start = dp[i].dp_start; + dsk->size = dp[i].dp_size; dsk->slice = i + 1; if (vdev_probe(vdev_read, dsk, NULL) == 0) { dsk = copy_dsk(dsk); } #ifdef LOADER_GELI_SUPPORT else if (geli_taste(vdev_read, dsk, dp[i].dp_size - dp[i].dp_start) == 0) { if (geli_havekey(dsk) == 0 || geli_passphrase(&gelipw, dsk->unit, 's', i, dsk) == 0) { /* * This slice has GELI, check it for ZFS. */ if (vdev_probe(vdev_read, dsk, NULL) == 0) { /* * This slice had a vdev. We need a new dsk * structure now since the vdev now owns this one. */ dsk = copy_dsk(dsk); } break; } } #endif /* LOADER_GELI_SUPPORT */ } } int main(void) { dnode_phys_t dn; off_t off; struct dsk *dsk; 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); } dsk = malloc(sizeof(struct dsk)); dsk->drive = *(uint8_t *)PTOV(ARGS); dsk->type = dsk->drive & DRV_HARD ? TYPE_AD : TYPE_FD; dsk->unit = dsk->drive & DRV_MASK; dsk->slice = *(uint8_t *)PTOV(ARGS + 1) + 1; dsk->part = 0; dsk->start = 0; - dsk->init = 0; + dsk->size = 0; 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 = dsk->drive; bootdev = MAKEBOOTDEV(dev_maj[dsk->type], dsk->slice, dsk->unit, dsk->part); /* Process configuration file */ autoboot = 1; #ifdef LOADER_GELI_SUPPORT geli_init(); #endif zfs_init(); /* * Probe the boot drive first - we will try to boot from whatever * pool we find on that drive. */ probe_drive(dsk); /* * 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; dsk = malloc(sizeof(struct dsk)); dsk->drive = i | DRV_HARD; dsk->type = dsk->drive & TYPE_AD; dsk->unit = i; dsk->slice = 0; dsk->part = 0; dsk->start = 0; - dsk->init = 0; + dsk->size = 0; probe_drive(dsk); } /* * 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(); /* * Try to exec /boot/loader. If interrupted by a keypress, * or in case of failure, try to load a kernel directly instead. */ if (autoboot && !*kname) { memcpy(kname, PATH_LOADER_ZFS, sizeof(PATH_LOADER_ZFS)); if (!keyhit(3)) { load(); memcpy(kname, PATH_KERNEL, sizeof(PATH_KERNEL)); } } /* 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)); gelibuf = malloc(sizeof(struct keybuf) + (GELI_MAX_KEYS * sizeof(struct keybuf_ent))); geli_fill_keybuf(gelibuf); zfsargs.notapw = '\0'; zfsargs.keybuf_sentinel = KEYBUF_SENTINEL; zfsargs.keybuf = gelibuf; #else zfsargs.gelipw[0] = '\0'; #endif if (primary_vdev != NULL) zfsargs.primary_vdev = primary_vdev->v_guid; else printf("failed to detect primary vdev\n"); __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/sys/boot/zfs/libzfs.h =================================================================== --- head/sys/boot/zfs/libzfs.h (revision 316584) +++ head/sys/boot/zfs/libzfs.h (revision 316585) @@ -1,92 +1,93 @@ /*- * Copyright (c) 2012 Andriy Gapon * 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. * * $FreeBSD$ */ #ifndef _BOOT_LIBZFS_H_ #define _BOOT_LIBZFS_H_ #define ZFS_MAXNAMELEN 256 /* * ZFS fully-qualified device descriptor. * Note, this must match the 'struct devdesc' declaration in bootstrap.h. * Arch-specific device descriptors should be binary compatible with this * structure if they are to support ZFS. */ struct zfs_devdesc { struct devsw *d_dev; int d_type; int d_unit; void *d_opendata; uint64_t pool_guid; uint64_t root_guid; }; #ifdef LOADER_GELI_SUPPORT #include #endif struct zfs_boot_args { uint32_t size; uint32_t reserved; uint64_t pool; uint64_t root; uint64_t primary_pool; uint64_t primary_vdev; union { char gelipw[256]; struct { char notapw; /* * single null byte to stop keybuf * being interpreted as a password */ uint32_t keybuf_sentinel; #ifdef LOADER_GELI_SUPPORT struct keybuf *keybuf; #else void *keybuf; #endif }; }; }; int zfs_parsedev(struct zfs_devdesc *dev, const char *devspec, const char **path); char *zfs_fmtdev(void *vdev); int zfs_probe_dev(const char *devname, uint64_t *pool_guid); int zfs_list(const char *name); +uint64_t ldi_get_size(void *); void init_zfs_bootenv(char *currdev); int zfs_bootenv(const char *name); int zfs_belist_add(const char *name, uint64_t __unused); int zfs_set_env(void); extern struct devsw zfs_dev; extern struct fs_ops zfs_fsops; #endif /*_BOOT_LIBZFS_H_*/ Index: head/sys/boot/zfs/zfsimpl.c =================================================================== --- head/sys/boot/zfs/zfsimpl.c (revision 316584) +++ head/sys/boot/zfs/zfsimpl.c (revision 316585) @@ -1,2487 +1,2525 @@ /*- * 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 "zfsimpl.h" #include "zfssubr.c" struct zfsmount { const spa_t *spa; objset_phys_t objset; uint64_t rootobj; }; /* * 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", NULL }; /* * List of all pools, chained through spa_link. */ static spa_list_t zfs_pools; -static const dnode_phys_t *dnode_cache_obj = NULL; +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 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) { printf("ZFS: out of temporary buffer space\n"); for (;;) ; } 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) { printf("ZFS: zfs_alloc()/zfs_free() mismatch\n"); for (;;) ; } } static int xdr_int(const unsigned char **xdr, int *ip) { *ip = ((*xdr)[0] << 24) | ((*xdr)[1] << 16) | ((*xdr)[2] << 8) | ((*xdr)[3] << 0); (*xdr) += 4; return (0); } static int xdr_u_int(const unsigned char **xdr, u_int *ip) { *ip = ((*xdr)[0] << 24) | ((*xdr)[1] << 16) | ((*xdr)[2] << 8) | ((*xdr)[3] << 0); (*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) && 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)) { 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; } /*printf("ZFS: reading %d bytes at 0x%jx to %p\n", psize, (uintmax_t)offset, buf);*/ rc = vdev->v_phys_read(vdev, vdev->v_read_priv, offset, buf, psize); if (rc) return (rc); if (bp && zio_checksum_verify(vdev->spa, bp, buf)) return (EIO); return (0); } 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 = malloc(sizeof(vdev_t)); memset(vdev, 0, sizeof(vdev_t)); STAILQ_INIT(&vdev->v_children); vdev->v_guid = guid; vdev->v_state = VDEV_STATE_OFFLINE; vdev->v_read = _read; vdev->v_phys_read = 0; vdev->v_read_priv = 0; STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink); return (vdev); } static int vdev_init_from_nvlist(const unsigned char *nvlist, vdev_t *pvdev, vdev_t **vdevp, int is_newer) { int rc; uint64_t guid, id, ashift, nparity; const char *type; const char *path; vdev_t *vdev, *kid; const unsigned char *kids; int nkids, i, is_new; uint64_t is_offline, is_faulted, is_degraded, is_removed, isnt_present; - if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID, - DATA_TYPE_UINT64, 0, &guid) - || nvlist_find(nvlist, ZPOOL_CONFIG_ID, - DATA_TYPE_UINT64, 0, &id) - || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE, - DATA_TYPE_STRING, 0, &type)) { + if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64, + NULL, &guid) + || nvlist_find(nvlist, ZPOOL_CONFIG_ID, DATA_TYPE_UINT64, NULL, &id) + || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE, DATA_TYPE_STRING, + NULL, &type)) { printf("ZFS: can't find vdev details\n"); return (ENOENT); } if (strcmp(type, VDEV_TYPE_MIRROR) && strcmp(type, VDEV_TYPE_DISK) #ifdef ZFS_TEST && strcmp(type, VDEV_TYPE_FILE) #endif && strcmp(type, VDEV_TYPE_RAIDZ) && strcmp(type, VDEV_TYPE_REPLACING)) { printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n"); return (EIO); } is_offline = is_removed = is_faulted = is_degraded = isnt_present = 0; - nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, 0, + nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, NULL, &is_offline); - nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, 0, + nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, NULL, &is_removed); - nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, 0, + nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, NULL, &is_faulted); - nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, 0, + nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, NULL, &is_degraded); - nvlist_find(nvlist, ZPOOL_CONFIG_NOT_PRESENT, DATA_TYPE_UINT64, 0, + nvlist_find(nvlist, ZPOOL_CONFIG_NOT_PRESENT, DATA_TYPE_UINT64, NULL, &isnt_present); vdev = vdev_find(guid); if (!vdev) { is_new = 1; if (!strcmp(type, VDEV_TYPE_MIRROR)) vdev = vdev_create(guid, vdev_mirror_read); else if (!strcmp(type, VDEV_TYPE_RAIDZ)) vdev = vdev_create(guid, vdev_raidz_read); else if (!strcmp(type, VDEV_TYPE_REPLACING)) vdev = vdev_create(guid, vdev_replacing_read); else vdev = vdev_create(guid, vdev_disk_read); vdev->v_id = id; vdev->v_top = pvdev != NULL ? pvdev : vdev; if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT, - DATA_TYPE_UINT64, 0, &ashift) == 0) + DATA_TYPE_UINT64, NULL, &ashift) == 0) { vdev->v_ashift = ashift; - else + } else { vdev->v_ashift = 0; + } if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY, - DATA_TYPE_UINT64, 0, &nparity) == 0) + DATA_TYPE_UINT64, NULL, &nparity) == 0) { vdev->v_nparity = nparity; - else + } else { vdev->v_nparity = 0; + } if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH, - DATA_TYPE_STRING, 0, &path) == 0) { + DATA_TYPE_STRING, NULL, &path) == 0) { if (strncmp(path, "/dev/", 5) == 0) path += 5; vdev->v_name = strdup(path); } else { if (!strcmp(type, "raidz")) { if (vdev->v_nparity == 1) vdev->v_name = "raidz1"; else if (vdev->v_nparity == 2) vdev->v_name = "raidz2"; else if (vdev->v_nparity == 3) vdev->v_name = "raidz3"; else { printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n"); return (EIO); } } else { vdev->v_name = strdup(type); } } } else { is_new = 0; } if (is_new || is_newer) { /* * This is either new vdev or we've already seen this vdev, * but from an older vdev label, so let's refresh its state * from the newer label. */ if (is_offline) vdev->v_state = VDEV_STATE_OFFLINE; else if (is_removed) vdev->v_state = VDEV_STATE_REMOVED; else if (is_faulted) vdev->v_state = VDEV_STATE_FAULTED; else if (is_degraded) vdev->v_state = VDEV_STATE_DEGRADED; else if (isnt_present) vdev->v_state = VDEV_STATE_CANT_OPEN; } - rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN, - DATA_TYPE_NVLIST_ARRAY, &nkids, &kids); + rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN, DATA_TYPE_NVLIST_ARRAY, + &nkids, &kids); /* * Its ok if we don't have any kids. */ if (rc == 0) { vdev->v_nchildren = nkids; for (i = 0; i < nkids; i++) { rc = vdev_init_from_nvlist(kids, vdev, &kid, is_newer); if (rc) return (rc); if (is_new) STAILQ_INSERT_TAIL(&vdev->v_children, kid, v_childlink); kids = nvlist_next(kids); } } else { vdev->v_nchildren = 0; } if (vdevp) *vdevp = vdev; return (0); } static void vdev_set_state(vdev_t *vdev) { vdev_t *kid; int good_kids; int bad_kids; /* * 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 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 (0); } 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)) return (spa); return (0); } #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 = STAILQ_FIRST(&spa->spa_vdevs); 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) +spa_create(uint64_t guid, const char *name) { spa_t *spa; - spa = malloc(sizeof(spa_t)); + if ((spa = malloc(sizeof(spa_t))) == NULL) + return (NULL); memset(spa, 0, sizeof(spa_t)); + if ((spa->spa_name = strdup(name)) == NULL) { + free(spa); + return (NULL); + } STAILQ_INIT(&spa->spa_vdevs); spa->spa_guid = guid; 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); vsprintf(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; 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_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; STAILQ_FOREACH(vdev, &spa->spa_vdevs, 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, &spa->spa_vdevs, 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); } +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_probe(vdev_phys_read_t *_read, void *read_priv, spa_t **spap) { vdev_t vtmp; vdev_phys_t *vdev_label = (vdev_phys_t *) zap_scratch; + vdev_phys_t *tmp_label = zfs_alloc(sizeof(vdev_phys_t)); spa_t *spa; vdev_t *vdev, *top_vdev, *pool_vdev; off_t off; blkptr_t bp; - const unsigned char *nvlist; + const unsigned char *nvlist = NULL; uint64_t val; uint64_t guid; + uint64_t best_txg = 0; uint64_t pool_txg, pool_guid; - uint64_t is_log; + uint64_t psize; const char *pool_name; const unsigned char *vdevs; const unsigned char *features; - int i, rc, is_newer; + int i, l, rc, is_newer; char *upbuf; const struct uberblock *up; /* * 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; - off = offsetof(vdev_label_t, vl_vdev_phys); - BP_ZERO(&bp); - BP_SET_LSIZE(&bp, sizeof(vdev_phys_t)); - BP_SET_PSIZE(&bp, sizeof(vdev_phys_t)); - 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); - if (vdev_read_phys(&vtmp, &bp, vdev_label, off, 0)) - return (EIO); + psize = P2ALIGN(ldi_get_size(read_priv), + (uint64_t)sizeof (vdev_label_t)); - if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR) { - return (EIO); + for (l = 0; l < VDEV_LABELS; l++) { + off = vdev_label_offset(psize, l, + offsetof(vdev_label_t, vl_vdev_phys)); + + BP_ZERO(&bp); + BP_SET_LSIZE(&bp, sizeof(vdev_phys_t)); + BP_SET_PSIZE(&bp, sizeof(vdev_phys_t)); + 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); + + if (vdev_read_phys(&vtmp, &bp, tmp_label, off, 0)) + continue; + + if (tmp_label->vp_nvlist[0] != NV_ENCODE_XDR) + continue; + + nvlist = (const unsigned char *) tmp_label->vp_nvlist + 4; + if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_TXG, + DATA_TYPE_UINT64, NULL, &pool_txg) != 0) + continue; + + if (best_txg <= pool_txg) { + best_txg = pool_txg; + memcpy(vdev_label, tmp_label, sizeof (vdev_phys_t)); + } } + zfs_free(tmp_label, sizeof (vdev_phys_t)); + + if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR) + return (EIO); + nvlist = (const unsigned char *) vdev_label->vp_nvlist + 4; - if (nvlist_find(nvlist, - ZPOOL_CONFIG_VERSION, - DATA_TYPE_UINT64, 0, &val)) { + if (nvlist_find(nvlist, ZPOOL_CONFIG_VERSION, DATA_TYPE_UINT64, + NULL, &val) != 0) { return (EIO); } if (!SPA_VERSION_IS_SUPPORTED(val)) { printf("ZFS: unsupported ZFS version %u (should be %u)\n", (unsigned) val, (unsigned) SPA_VERSION); return (EIO); } /* Check ZFS features for read */ - if (nvlist_find(nvlist, - ZPOOL_CONFIG_FEATURES_FOR_READ, - DATA_TYPE_NVLIST, 0, &features) == 0 - && nvlist_check_features_for_read(features) != 0) + if (nvlist_find(nvlist, ZPOOL_CONFIG_FEATURES_FOR_READ, + DATA_TYPE_NVLIST, NULL, &features) == 0 && + nvlist_check_features_for_read(features) != 0) { return (EIO); + } - if (nvlist_find(nvlist, - ZPOOL_CONFIG_POOL_STATE, - DATA_TYPE_UINT64, 0, &val)) { + if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_STATE, DATA_TYPE_UINT64, + NULL, &val) != 0) { return (EIO); } if (val == POOL_STATE_DESTROYED) { /* We don't boot only from destroyed pools. */ return (EIO); } - if (nvlist_find(nvlist, - ZPOOL_CONFIG_POOL_TXG, - DATA_TYPE_UINT64, 0, &pool_txg) - || nvlist_find(nvlist, - ZPOOL_CONFIG_POOL_GUID, - DATA_TYPE_UINT64, 0, &pool_guid) - || nvlist_find(nvlist, - ZPOOL_CONFIG_POOL_NAME, - DATA_TYPE_STRING, 0, &pool_name)) { + 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. */ /*printf("ZFS: can't find pool details\n");*/ return (EIO); } - is_log = 0; - (void) nvlist_find(nvlist, ZPOOL_CONFIG_IS_LOG, DATA_TYPE_UINT64, 0, - &is_log); - if (is_log) + if (nvlist_find(nvlist, ZPOOL_CONFIG_IS_LOG, DATA_TYPE_UINT64, + NULL, &val) == 0 && val != 0) { return (EIO); + } /* * Create the pool if this is the first time we've seen it. */ spa = spa_find_by_guid(pool_guid); - if (!spa) { - spa = spa_create(pool_guid); - spa->spa_name = strdup(pool_name); + if (spa == NULL) { + spa = spa_create(pool_guid, pool_name); + if (spa == NULL) + return (ENOMEM); } if (pool_txg > spa->spa_txg) { spa->spa_txg = pool_txg; is_newer = 1; - } else + } else { is_newer = 0; + } /* * 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, 0, &guid)) { + if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64, + NULL, &guid) != 0) { return (EIO); } vdev = vdev_find(guid); if (vdev && vdev->v_phys_read) /* Has this vdev already been inited? */ return (EIO); - if (nvlist_find(nvlist, - ZPOOL_CONFIG_VDEV_TREE, - DATA_TYPE_NVLIST, 0, &vdevs)) { + if (nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_TREE, DATA_TYPE_NVLIST, + NULL, &vdevs)) { return (EIO); } rc = vdev_init_from_nvlist(vdevs, NULL, &top_vdev, is_newer); - if (rc) + if (rc != 0) return (rc); /* * Add the toplevel vdev to the pool if its not already there. */ STAILQ_FOREACH(pool_vdev, &spa->spa_vdevs, v_childlink) if (top_vdev == pool_vdev) break; if (!pool_vdev && top_vdev) { top_vdev->spa = spa; STAILQ_INSERT_TAIL(&spa->spa_vdevs, top_vdev, v_childlink); } /* * 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) { vdev->v_phys_read = _read; vdev->v_read_priv = read_priv; vdev->v_state = VDEV_STATE_HEALTHY; } else { printf("ZFS: inconsistent nvlist contents\n"); return (EIO); } /* * Re-evaluate top-level vdev state. */ vdev_set_state(top_vdev); /* * 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. */ upbuf = zfs_alloc(VDEV_UBERBLOCK_SIZE(vdev)); up = (const struct uberblock *)upbuf; - for (i = 0; - i < VDEV_UBERBLOCK_COUNT(vdev); - i++) { - off = VDEV_UBERBLOCK_OFFSET(vdev, i); - BP_ZERO(&bp); - DVA_SET_OFFSET(&bp.blk_dva[0], off); - BP_SET_LSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev)); - BP_SET_PSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev)); - BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL); - BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF); - ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0); + for (l = 0; l < VDEV_LABELS; l++) { + for (i = 0; i < VDEV_UBERBLOCK_COUNT(vdev); i++) { + off = vdev_label_offset(psize, l, + VDEV_UBERBLOCK_OFFSET(vdev, i)); + BP_ZERO(&bp); + DVA_SET_OFFSET(&bp.blk_dva[0], off); + BP_SET_LSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev)); + BP_SET_PSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev)); + BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL); + BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF); + ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0); - if (vdev_read_phys(vdev, &bp, upbuf, off, 0)) - continue; + if (vdev_read_phys(vdev, &bp, upbuf, off, 0)) + continue; - if (up->ub_magic != UBERBLOCK_MAGIC) - continue; - if (up->ub_txg < spa->spa_txg) - continue; - if (up->ub_txg > spa->spa_uberblock.ub_txg) { - spa->spa_uberblock = *up; - } else if (up->ub_txg == spa->spa_uberblock.ub_txg) { - if (up->ub_timestamp > spa->spa_uberblock.ub_timestamp) + if (up->ub_magic != UBERBLOCK_MAGIC) + continue; + if (up->ub_txg < spa->spa_txg) + continue; + if (up->ub_txg > spa->spa_uberblock.ub_txg || + (up->ub_txg == spa->spa_uberblock.ub_txg && + up->ub_timestamp > + spa->spa_uberblock.ub_timestamp)) { spa->spa_uberblock = *up; + } } } zfs_free(upbuf, VDEV_UBERBLOCK_SIZE(vdev)); vdev->spa = spa; - if (spap) + 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; int vdevid; off_t offset; if (!dva->dva_word[0] && !dva->dva_word[1]) continue; vdevid = DVA_GET_VDEV(dva); offset = DVA_GET_OFFSET(dva); STAILQ_FOREACH(vdev, &spa->spa_vdevs, 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_top->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) { 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)) { *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; } } /* * Lookup a value in a fatzap directory. Assumes that the zap scratch * buffer contains the directory header. */ 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) { 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; if (zh.zap_magic != ZAP_MAGIC) return (EIO); z.zap_block_shift = ilog2(bsize); z.zap_phys = (zap_phys_t *) zap_scratch; /* * 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; } else { ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0); } hash = zap_hash(zh.zap_salt, name); zap_leaf_t zl; zl.l_bs = z.zap_block_shift; off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs; zap_leaf_chunk_t *zc; rc = dnode_read(spa, dnode, off, zap_scratch, bsize); if (rc) return (rc); 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); /* * 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) { zc = NULL; break; } zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next); } 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 (ENOENT); } /* * 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; size_t size = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; rc = dnode_read(spa, dnode, 0, zap_scratch, size); if (rc) return (rc); 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); } printf("ZFS: invalid zap_type=%d\n", (int)zap_type); return (EIO); } /* * List a microzap directory. Assumes that the zap scratch buffer contains * the directory contents. */ static int mzap_list(const dnode_phys_t *dnode, 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; 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); } } return (0); } /* * List a fatzap directory. Assumes that the zap scratch buffer contains * the directory header. */ static int fzap_list(const spa_t *spa, const dnode_phys_t *dnode, 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; if (zh.zap_magic != ZAP_MAGIC) return (EIO); z.zap_block_shift = ilog2(bsize); z.zap_phys = (zap_phys_t *) zap_scratch; /* * 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++) { off_t off = (i + 1) << zl.l_bs; char name[256], *p; uint64_t value; if (dnode_read(spa, dnode, off, zap_scratch, bsize)) 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) return (rc); } } 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; size_t size = dnode->dn_datablkszsec * 512; if (dnode_read(spa, dnode, 0, zap_scratch, size)) return (EIO); 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); } 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)); } static int mzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, 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) { int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; zap_phys_t zh = *(zap_phys_t *) zap_scratch; fat_zap_t z; int i, j; if (zh.zap_magic != ZAP_MAGIC) return (EIO); z.zap_block_shift = ilog2(bsize); z.zap_phys = (zap_phys_t *) zap_scratch; /* * 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++) { off_t off = (i + 1) << zl.l_bs; if (dnode_read(spa, dnode, off, zap_scratch, bsize)) 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; 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); } } } return (ENOENT); } static int zap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value) { int rc; uint64_t zap_type; size_t size = dnode->dn_datablkszsec * 512; rc = dnode_read(spa, dnode, 0, zap_scratch, size); if (rc) return (rc); 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); } 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; dnode_phys_t child_dir_zap, dir, dataset; dsl_dataset_phys_t *ds; dsl_dir_phys_t *dd; 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); 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); } #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; 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)) return (EIO); zap_type = *(uint64_t *) zap_scratch; if (zap_type == ZBT_MICRO) rc = mzap_list(&dir, check_feature); else rc = fzap_list(spa, &dir, check_feature); return (rc); } static int zfs_spa_init(spa_t *spa) { dnode_phys_t dir; 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); } 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->dn_spill; 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->dn_spill; 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); }