diff --git a/stand/i386/loader/main.c b/stand/i386/loader/main.c index 4d5a0a98a08e..c15c97c961b7 100644 --- a/stand/i386/loader/main.c +++ b/stand/i386/loader/main.c @@ -1,466 +1,466 @@ /*- * 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 #include "btxv86.h" #ifdef LOADER_ZFS_SUPPORT #include #include "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 uint32_t initial_howto; static uint32_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; struct geli_boot_data *gbdata; #endif #ifdef LOADER_ZFS_SUPPORT struct zfs_boot_args *zargs; static void i386_zfs_probe(void); #endif /* XXX debugging */ extern char end[]; static void *heap_top; static void *heap_bottom; caddr_t ptov(uintptr_t x) { return (PTOV(x)); } 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); /* * Now that malloc is usable, allocate a buffer for tslog and start * logging timestamps during the boot process. */ tslog_init(); /* * detect ACPI for future reference. This may set console to comconsole * if we do have ACPI SPCR table. */ biosacpi_detect(); /* * 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(); /* Set up currdev variable to have hooks in place. */ env_setenv("currdev", EV_VOLATILE | EV_NOHOOK, "", i386_setcurrdev, env_nounset); /* * 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; archsw.arch_hypervisor = x86_hypervisor; #ifdef LOADER_ZFS_SUPPORT archsw.arch_zfs_probe = i386_zfs_probe; /* * zfsboot and gptzfsboot have always passed KARGS_FLAGS_ZFS, * so if that is set along with KARGS_FLAGS_EXTARG we know we * can interpret the extarg data as a struct zfs_boot_args. */ #define KARGS_EXTARGS_ZFS (KARGS_FLAGS_EXTARG | KARGS_FLAGS_ZFS) if ((kargs->bootflags & KARGS_EXTARGS_ZFS) == KARGS_EXTARGS_ZFS) { zargs = (struct zfs_boot_args *)(kargs + 1); } #endif /* LOADER_ZFS_SUPPORT */ #ifdef LOADER_GELI_SUPPORT /* * If we decided earlier that we have zfs_boot_args extarg data, * and it is big enough to contain the embedded geli data * (the early zfs_boot_args structs weren't), then init the gbdata * pointer accordingly. If there is extarg data which isn't * zfs_boot_args data, determine whether it is geli_boot_args data. * Recent versions of gptboot set KARGS_FLAGS_GELI to indicate that. * Earlier versions didn't, but we presume that's what we * have if the extarg size exactly matches the size of the * geli_boot_args struct during that pre-flag era. */ #define LEGACY_GELI_ARGS_SIZE 260 /* This can never change */ #ifdef LOADER_ZFS_SUPPORT if (zargs != NULL) { if (zargs->size > offsetof(struct zfs_boot_args, gelidata)) { gbdata = &zargs->gelidata; } } else #endif /* LOADER_ZFS_SUPPORT */ if ((kargs->bootflags & KARGS_FLAGS_EXTARG) != 0) { gargs = (struct geli_boot_args *)(kargs + 1); if ((kargs->bootflags & KARGS_FLAGS_GELI) || gargs->size == LEGACY_GELI_ARGS_SIZE) { gbdata = &gargs->gelidata; } } if (gbdata != NULL) import_geli_boot_data(gbdata); #endif /* LOADER_GELI_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 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 */ autoload_font(true); bios_getsmap(); interact(); /* 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]; char *bootonce; #endif int biosdev = -1; /* Assume we are booting from a BIOS disk by default */ new_currdev.dd.d_dev = &bioshd; /* 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.dd.d_dev = &bioscd; new_currdev.dd.d_unit = bd_bios2unit(initial_bootdev); } else if ((kargs->bootflags & KARGS_FLAGS_PXE) != 0) { /* we are booting from pxeldr */ new_currdev.dd.d_dev = &pxedisk; new_currdev.dd.d_unit = 0; } else { /* we don't know what our boot device is */ new_currdev.disk.d_slice = -1; new_currdev.disk.d_partition = 0; biosdev = -1; } #ifdef LOADER_ZFS_SUPPORT } else if ((kargs->bootflags & KARGS_FLAGS_ZFS) != 0) { /* * zargs was set in main() if we have new style extended * argument */ if (zargs != NULL && zargs->size >= offsetof(struct zfs_boot_args, primary_pool)) { /* sufficient data is provided */ new_currdev.zfs.pool_guid = zargs->pool; new_currdev.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.zfs.pool_guid = kargs->zfspool; new_currdev.zfs.root_guid = 0; } new_currdev.dd.d_dev = &zfs_dev; if ((bootonce = malloc(VDEV_PAD_SIZE)) != NULL) { if (zfs_get_bootonce(&new_currdev, OS_BOOTONCE_USED, bootonce, VDEV_PAD_SIZE) == 0) { setenv("zfs-bootonce", bootonce, 1); } free(bootonce); (void) zfs_attach_nvstore(&new_currdev); } #endif } else if ((initial_bootdev & B_MAGICMASK) != B_DEVMAGIC) { /* The passed-in boot device is bad */ new_currdev.disk.d_slice = -1; new_currdev.disk.d_partition = 0; biosdev = -1; } else { new_currdev.disk.d_slice = B_SLICE(initial_bootdev) - 1; new_currdev.disk.d_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, assume harddisk */ biosdev = 0x80 + B_UNIT(initial_bootdev); } } /* * 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.dd.d_dev->dv_type == bioshd.dv_type) && ((new_currdev.dd.d_unit = bd_bios2unit(biosdev)) == -1)) { printf("Can't work out which disk we are booting " "from.\nGuessed BIOS device 0x%x not found by " "probes, defaulting to disk0:\n", biosdev); new_currdev.dd.d_unit = 0; } #ifdef LOADER_ZFS_SUPPORT if (new_currdev.dd.d_dev->dv_type == DEVT_ZFS) - init_zfs_boot_options(zfs_fmtdev(&new_currdev.dd)); + init_zfs_boot_options(devformat(&new_currdev.dd)); #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); } /* 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]; struct i386_devdesc dev; /* * Open all the disks we can find and see if we can reconstruct * ZFS pools from them. */ dev.dd.d_dev = &bioshd; for (dev.dd.d_unit = 0; bd_unit2bios(&dev) >= 0; dev.dd.d_unit++) { snprintf(devname, sizeof(devname), "%s%d:", bioshd.dv_name, dev.dd.d_unit); zfs_probe_dev(devname, NULL); } } #endif diff --git a/stand/i386/zfsboot/zfsboot.c b/stand/i386/zfsboot/zfsboot.c index 52fd41688a17..930a6fb530af 100644 --- a/stand/i386/zfsboot/zfsboot.c +++ b/stand/i386/zfsboot/zfsboot.c @@ -1,720 +1,720 @@ /*- * 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 #include #ifdef GPT #include #endif #include #include #ifdef LOADER_ZFS_SUPPORT #include #endif #include #include #include #include #include #include #include "bootstrap.h" #include "libi386.h" #include #include "lib.h" #include "rbx.h" #include "cons.h" #include "bootargs.h" #include "disk.h" #include "part.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 extern uint32_t _end; static const char optstr[NOPT] = "DhaCcdgmnpqrsv"; /* Also 'P', 'S' */ static const unsigned char flags[NOPT] = { RBX_DUAL, RBX_SERIAL, RBX_ASKNAME, RBX_CDROM, RBX_CONFIG, RBX_KDB, RBX_GDB, RBX_MUTE, RBX_NOINTR, RBX_PAUSE, RBX_QUIET, RBX_DFLTROOT, RBX_SINGLE, RBX_VERBOSE }; uint32_t opts; /* * Paths to try loading before falling back to the boot2 prompt. * * /boot/zfsloader must be tried before /boot/loader in order to remain * backward compatible with ZFS boot environments where /boot/loader exists * but does not have ZFS support, which was the case before FreeBSD 12. * * If no loader is found, try to load a kernel directly instead. */ static const struct string { const char *p; size_t len; } loadpath[] = { { PATH_LOADER_ZFS, sizeof(PATH_LOADER_ZFS) }, { PATH_LOADER, sizeof(PATH_LOADER) }, { PATH_KERNEL, sizeof(PATH_KERNEL) }, }; static const unsigned char dev_maj[NDEV] = {30, 4, 2}; static struct i386_devdesc *bdev; static char cmd[512]; static char cmddup[512]; static char kname[1024]; static int comspeed = SIOSPD; static struct bootinfo bootinfo; static uint32_t bootdev; static struct zfs_boot_args zfsargs; #ifdef LOADER_GELI_SUPPORT static struct geli_boot_args geliargs; #endif extern vm_offset_t high_heap_base; extern uint32_t bios_basemem, bios_extmem, high_heap_size; static char *heap_top; static char *heap_bottom; void exit(int); static void i386_zfs_probe(void); static void load(void); static int parse_cmd(void); #ifdef LOADER_GELI_SUPPORT #include "geliboot.h" static char gelipw[GELI_PW_MAXLEN]; #endif struct arch_switch archsw; /* MI/MD interface boundary */ static char boot_devname[2 * ZFS_MAXNAMELEN + 8]; /* disk or pool:dataset */ struct devsw *devsw[] = { &bioshd, #if defined(LOADER_ZFS_SUPPORT) &zfs_dev, #endif NULL }; struct fs_ops *file_system[] = { #if defined(LOADER_ZFS_SUPPORT) &zfs_fsops, #endif #if defined(LOADER_UFS_SUPPORT) &ufs_fsops, #endif NULL }; caddr_t ptov(uintptr_t x) { return (PTOV(x)); } int main(void); int main(void) { unsigned i; int auto_boot, fd, nextboot = 0; struct disk_devdesc devdesc; bios_getmem(); if (high_heap_size > 0) { heap_top = PTOV(high_heap_base + high_heap_size); heap_bottom = PTOV(high_heap_base); } else { heap_bottom = (char *) (roundup2(__base + (int32_t)&_end, 0x10000) - __base); heap_top = (char *)PTOV(bios_basemem); } setheap(heap_bottom, heap_top); /* * Initialise the block cache. Set the upper limit. */ bcache_init(32768, 512); archsw.arch_autoload = NULL; archsw.arch_getdev = i386_getdev; archsw.arch_copyin = NULL; archsw.arch_copyout = NULL; archsw.arch_readin = NULL; archsw.arch_isainb = NULL; archsw.arch_isaoutb = NULL; archsw.arch_zfs_probe = i386_zfs_probe; 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 = *(uint8_t *)PTOV(ARGS); /* Set up fall back device name. */ snprintf(boot_devname, sizeof (boot_devname), "disk%d:", bd_bios2unit(bootinfo.bi_bios_dev)); /* Set up currdev variable to have hooks in place. */ env_setenv("currdev", EV_VOLATILE, "", i386_setcurrdev, env_nounset); for (i = 0; devsw[i] != NULL; i++) if (devsw[i]->dv_init != NULL) (devsw[i]->dv_init)(); disk_parsedev(&devdesc, boot_devname + 4, NULL); bootdev = MAKEBOOTDEV(dev_maj[DEVT_DISK], devdesc.d_slice + 1, devdesc.dd.d_unit, devdesc.d_partition >= 0 ? devdesc.d_partition : 0xff); /* - * zfs_fmtdev() can be called only after dv_init + * devformat() can be called only after dv_init */ if (bdev != NULL && bdev->dd.d_dev->dv_type == DEVT_ZFS) { /* set up proper device name string for ZFS */ - strncpy(boot_devname, zfs_fmtdev(&bdev->dd), sizeof (boot_devname)); + strncpy(boot_devname, devformat(&bdev->dd), sizeof (boot_devname)); if (zfs_get_bootonce(bdev, OS_BOOTONCE, cmd, sizeof(cmd)) == 0) { nvlist_t *benv; nextboot = 1; memcpy(cmddup, cmd, sizeof(cmd)); if (parse_cmd()) { if (!OPT_CHECK(RBX_QUIET)) printf("failed to parse bootonce " "command\n"); exit(0); } if (!OPT_CHECK(RBX_QUIET)) printf("zfs bootonce: %s\n", cmddup); if (zfs_get_bootenv(bdev, &benv) == 0) { nvlist_add_string(benv, OS_BOOTONCE_USED, cmddup); zfs_set_bootenv(bdev, benv); } /* Do not process this command twice */ *cmd = 0; } } /* now make sure we have bdev on all cases */ free(bdev); i386_getdev((void **)&bdev, boot_devname, NULL); env_setenv("currdev", EV_VOLATILE, boot_devname, i386_setcurrdev, env_nounset); /* Process configuration file */ auto_boot = 1; fd = open(PATH_CONFIG, O_RDONLY); if (fd == -1) fd = open(PATH_DOTCONFIG, O_RDONLY); if (fd != -1) { ssize_t cmdlen; if ((cmdlen = read(fd, cmd, sizeof(cmd))) > 0) cmd[cmdlen] = '\0'; else *cmd = '\0'; close(fd); } 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()) auto_boot = 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 && !auto_boot) exit(0); if (auto_boot && !*kname) { /* * Iterate through the list of loader and kernel paths, * trying to load. If interrupted by a keypress, or in case of * failure, drop the user to the boot2 prompt. */ for (i = 0; i < nitems(loadpath); i++) { memcpy(kname, loadpath[i].p, loadpath[i].len); if (keyhit(3)) break; load(); } } /* Present the user with the boot2 prompt. */ for (;;) { if (!auto_boot || !OPT_CHECK(RBX_QUIET)) { printf("\nFreeBSD/x86 boot\n"); printf("Default: %s%s\nboot: ", boot_devname, kname); } if (ioctrl & IO_SERIAL) sio_flush(); if (!auto_boot || keyhit(5)) getstr(cmd, sizeof(cmd)); else if (!auto_boot || !OPT_CHECK(RBX_QUIET)) putchar('\n'); auto_boot = 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); } 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; uint32_t addr, x; int fd, fmt, i, j; ssize_t size; if ((fd = open(kname, O_RDONLY)) == -1) { printf("\nCan't find %s\n", kname); return; } size = sizeof(hdr); if (read(fd, &hdr, sizeof (hdr)) != size) { close(fd); return; } if (N_GETMAGIC(hdr.ex) == ZMAGIC) { fmt = 0; } else if (IS_ELF(hdr.eh)) { fmt = 1; } else { printf("Invalid %s\n", "format"); close(fd); return; } if (fmt == 0) { addr = hdr.ex.a_entry & 0xffffff; p = PTOV(addr); lseek(fd, PAGE_SIZE, SEEK_SET); size = hdr.ex.a_text; if (read(fd, p, hdr.ex.a_text) != size) { close(fd); return; } p += roundup2(hdr.ex.a_text, PAGE_SIZE); size = hdr.ex.a_data; if (read(fd, p, hdr.ex.a_data) != size) { close(fd); 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) { size = hdr.ex.a_syms; if (read(fd, p, hdr.ex.a_syms) != size) { close(fd); return; } p += hdr.ex.a_syms; size = sizeof (int); if (read(fd, p, sizeof (int)) != size) { close(fd); return; } x = *(uint32_t *)p; p += sizeof(int); x -= sizeof(int); size = x; if (read(fd, p, x) != size) { close(fd); return; } p += x; } } else { lseek(fd, hdr.eh.e_phoff, SEEK_SET); for (j = i = 0; i < hdr.eh.e_phnum && j < 2; i++) { size = sizeof (ep[0]); if (read(fd, ep + j, sizeof (ep[0])) != size) { close(fd); return; } if (ep[j].p_type == PT_LOAD) j++; } for (i = 0; i < 2; i++) { p = PTOV(ep[i].p_paddr & 0xffffff); lseek(fd, ep[i].p_offset, SEEK_SET); size = ep[i].p_filesz; if (read(fd, p, ep[i].p_filesz) != size) { close(fd); return; } } p += roundup2(ep[1].p_memsz, PAGE_SIZE); bootinfo.bi_symtab = VTOP(p); if (hdr.eh.e_shnum == hdr.eh.e_shstrndx + 3) { lseek(fd, hdr.eh.e_shoff + sizeof (es[0]) * (hdr.eh.e_shstrndx + 1), SEEK_SET); size = sizeof(es); if (read(fd, &es, sizeof (es)) != size) { close(fd); return; } for (i = 0; i < 2; i++) { memcpy(p, &es[i].sh_size, sizeof(es[i].sh_size)); p += sizeof(es[i].sh_size); lseek(fd, es[i].sh_offset, SEEK_SET); size = es[i].sh_size; if (read(fd, p, es[i].sh_size) != size) { close(fd); return; } p += es[i].sh_size; } } addr = hdr.eh.e_entry & 0xffffff; } close(fd); bootinfo.bi_esymtab = VTOP(p); bootinfo.bi_kernelname = VTOP(kname); #ifdef LOADER_GELI_SUPPORT explicit_bzero(gelipw, sizeof(gelipw)); #endif if (bdev->dd.d_dev->dv_type == DEVT_ZFS) { zfsargs.size = sizeof(zfsargs); zfsargs.pool = bdev->zfs.pool_guid; zfsargs.root = bdev->zfs.root_guid; #ifdef LOADER_GELI_SUPPORT export_geli_boot_data(&zfsargs.gelidata); #endif /* * Note that the zfsargs struct is passed by value, not by * pointer. Code in btxldr.S copies the values from the entry * stack to a fixed location within loader(8) at startup due * to the presence of KARGS_FLAGS_EXTARG. */ __exec((caddr_t)addr, RB_BOOTINFO | (opts & RBX_MASK), bootdev, KARGS_FLAGS_ZFS | KARGS_FLAGS_EXTARG, (uint32_t)bdev->zfs.pool_guid, (uint32_t)(bdev->zfs.pool_guid >> 32), VTOP(&bootinfo), zfsargs); } else { #ifdef LOADER_GELI_SUPPORT geliargs.size = sizeof(geliargs); export_geli_boot_data(&geliargs.gelidata); #endif /* * Note that the geliargs struct is passed by value, not by * pointer. Code in btxldr.S copies the values from the entry * stack to a fixed location within loader(8) at startup due * to the presence of the KARGS_FLAGS_EXTARG flag. */ __exec((caddr_t)addr, RB_BOOTINFO | (opts & RBX_MASK), bootdev, #ifdef LOADER_GELI_SUPPORT KARGS_FLAGS_GELI | KARGS_FLAGS_EXTARG, 0, 0, VTOP(&bootinfo), geliargs #else 0, 0, 0, VTOP(&bootinfo) #endif ); } } static int mount_root(char *arg) { char *root; struct i386_devdesc *ddesc; uint8_t part; if (asprintf(&root, "%s:", arg) < 0) return (1); if (i386_getdev((void **)&ddesc, root, NULL)) { free(root); return (1); } /* we should have new device descriptor, free old and replace it. */ free(bdev); bdev = ddesc; if (bdev->dd.d_dev->dv_type == DEVT_DISK) { if (bdev->disk.d_partition == -1) part = 0xff; else part = bdev->disk.d_partition; bootdev = MAKEBOOTDEV(dev_maj[bdev->dd.d_dev->dv_type], bdev->disk.d_slice + 1, bdev->dd.d_unit, part); bootinfo.bi_bios_dev = bd_unit2bios(bdev); } strncpy(boot_devname, root, sizeof (boot_devname)); setenv("currdev", root, 1); free(root); return (0); } static void fs_list(char *arg) { int fd; struct dirent *d; char line[80]; fd = open(arg, O_RDONLY); if (fd < 0) return; pager_open(); while ((d = readdirfd(fd)) != NULL) { sprintf(line, "%s\n", d->d_name); if (pager_output(line)) break; } pager_close(); close(fd); } static int parse_cmd(void) { char *arg = cmd; char *ep, *p, *q; const char *cp; char line[80]; 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); opts |= 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 == '?') { printf("\n"); if (*arg == '\0') arg = (char *)"/"; fs_list(arg); zfs_list(arg); return (-1); } else { char *ptr; printf("\n"); arg--; /* * Report pool status if the comment is 'status'. Lets * hope no-one wants to load /status as a kernel. */ if (strcmp(arg, "status") == 0) { pager_open(); for (i = 0; devsw[i] != NULL; i++) { if (devsw[i]->dv_print != NULL) { if (devsw[i]->dv_print(1)) break; } else { snprintf(line, sizeof(line), "%s: (unknown)\n", devsw[i]->dv_name); if (pager_output(line)) break; } } pager_close(); return (-1); } /* * If there is "zfs:" prefix simply ignore it. */ ptr = arg; if (strncmp(ptr, "zfs:", 4) == 0) ptr += 4; /* * If there is a colon, switch pools. */ q = strchr(ptr, ':'); if (q) { *q++ = '\0'; if (mount_root(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); } /* * Probe all disks to discover ZFS pools. The idea is to walk all possible * disk devices, however, we also need to identify possible boot pool. * For boot pool detection we have boot disk passed us from BIOS, recorded * in bootinfo.bi_bios_dev. */ static void i386_zfs_probe(void) { char devname[32]; int boot_unit; struct i386_devdesc dev; uint64_t pool_guid = 0; dev.dd.d_dev = &bioshd; /* Translate bios dev to our unit number. */ boot_unit = bd_bios2unit(bootinfo.bi_bios_dev); /* * Open all the disks we can find and see if we can reconstruct * ZFS pools from them. */ for (dev.dd.d_unit = 0; bd_unit2bios(&dev) >= 0; dev.dd.d_unit++) { snprintf(devname, sizeof (devname), "%s%d:", bioshd.dv_name, dev.dd.d_unit); /* If this is not boot disk, use generic probe. */ if (dev.dd.d_unit != boot_unit) zfs_probe_dev(devname, NULL); else zfs_probe_dev(devname, &pool_guid); if (pool_guid != 0 && bdev == NULL) { bdev = malloc(sizeof (struct i386_devdesc)); bzero(bdev, sizeof (struct i386_devdesc)); bdev->zfs.dd.d_dev = &zfs_dev; bdev->zfs.pool_guid = pool_guid; } } } diff --git a/stand/libsa/zfs/zfs.c b/stand/libsa/zfs/zfs.c index 0ac67d750fcd..b525858ffc3c 100644 --- a/stand/libsa/zfs/zfs.c +++ b/stand/libsa/zfs/zfs.c @@ -1,2043 +1,2043 @@ /*- * 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. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); /* * Stand-alone file reading package. */ #include #include #include #include #include #include #include #include #include #include #include "libzfs.h" #include "zfsimpl.c" /* Define the range of indexes to be populated with ZFS Boot Environments */ #define ZFS_BE_FIRST 4 #define ZFS_BE_LAST 8 static int zfs_open(const char *path, struct open_file *f); static int zfs_close(struct open_file *f); static int zfs_read(struct open_file *f, void *buf, size_t size, size_t *resid); static off_t zfs_seek(struct open_file *f, off_t offset, int where); static int zfs_stat(struct open_file *f, struct stat *sb); static int zfs_readdir(struct open_file *f, struct dirent *d); static int zfs_mount(const char *dev, const char *path, void **data); static int zfs_unmount(const char *dev, void *data); static void zfs_bootenv_initial(const char *envname, spa_t *spa, const char *name, const char *dsname, int checkpoint); static void zfs_checkpoints_initial(spa_t *spa, const char *name, const char *dsname); struct devsw zfs_dev; struct fs_ops zfs_fsops = { .fs_name = "zfs", .fo_open = zfs_open, .fo_close = zfs_close, .fo_read = zfs_read, .fo_write = null_write, .fo_seek = zfs_seek, .fo_stat = zfs_stat, .fo_readdir = zfs_readdir, .fo_mount = zfs_mount, .fo_unmount = zfs_unmount }; /* * In-core open file. */ struct file { off_t f_seekp; /* seek pointer */ dnode_phys_t f_dnode; uint64_t f_zap_type; /* zap type for readdir */ uint64_t f_num_leafs; /* number of fzap leaf blocks */ zap_leaf_phys_t *f_zap_leaf; /* zap leaf buffer */ }; static int zfs_env_index; static int zfs_env_count; SLIST_HEAD(zfs_be_list, zfs_be_entry) zfs_be_head = SLIST_HEAD_INITIALIZER(zfs_be_head); struct zfs_be_list *zfs_be_headp; struct zfs_be_entry { char *name; SLIST_ENTRY(zfs_be_entry) entries; } *zfs_be, *zfs_be_tmp; /* * Open a file. */ static int zfs_open(const char *upath, struct open_file *f) { struct zfsmount *mount = (struct zfsmount *)f->f_devdata; struct file *fp; int rc; if (f->f_dev != &zfs_dev) return (EINVAL); /* allocate file system specific data structure */ fp = calloc(1, sizeof(struct file)); if (fp == NULL) return (ENOMEM); f->f_fsdata = fp; rc = zfs_lookup(mount, upath, &fp->f_dnode); fp->f_seekp = 0; if (rc) { f->f_fsdata = NULL; free(fp); } return (rc); } static int zfs_close(struct open_file *f) { struct file *fp = (struct file *)f->f_fsdata; dnode_cache_obj = NULL; f->f_fsdata = NULL; free(fp); return (0); } /* * Copy a portion of a file into kernel memory. * Cross block boundaries when necessary. */ static int zfs_read(struct open_file *f, void *start, size_t size, size_t *resid /* out */) { const spa_t *spa = ((struct zfsmount *)f->f_devdata)->spa; struct file *fp = (struct file *)f->f_fsdata; struct stat sb; size_t n; int rc; rc = zfs_stat(f, &sb); if (rc) return (rc); n = size; if (fp->f_seekp + n > sb.st_size) n = sb.st_size - fp->f_seekp; rc = dnode_read(spa, &fp->f_dnode, fp->f_seekp, start, n); if (rc) return (rc); if (0) { int i; for (i = 0; i < n; i++) putchar(((char*) start)[i]); } fp->f_seekp += n; if (resid) *resid = size - n; return (0); } static off_t zfs_seek(struct open_file *f, off_t offset, int where) { struct file *fp = (struct file *)f->f_fsdata; switch (where) { case SEEK_SET: fp->f_seekp = offset; break; case SEEK_CUR: fp->f_seekp += offset; break; case SEEK_END: { struct stat sb; int error; error = zfs_stat(f, &sb); if (error != 0) { errno = error; return (-1); } fp->f_seekp = sb.st_size - offset; break; } default: errno = EINVAL; return (-1); } return (fp->f_seekp); } static int zfs_stat(struct open_file *f, struct stat *sb) { const spa_t *spa = ((struct zfsmount *)f->f_devdata)->spa; struct file *fp = (struct file *)f->f_fsdata; return (zfs_dnode_stat(spa, &fp->f_dnode, sb)); } static int zfs_readdir(struct open_file *f, struct dirent *d) { const spa_t *spa = ((struct zfsmount *)f->f_devdata)->spa; struct file *fp = (struct file *)f->f_fsdata; mzap_ent_phys_t mze; struct stat sb; size_t bsize = fp->f_dnode.dn_datablkszsec << SPA_MINBLOCKSHIFT; int rc; rc = zfs_stat(f, &sb); if (rc) return (rc); if (!S_ISDIR(sb.st_mode)) return (ENOTDIR); /* * If this is the first read, get the zap type. */ if (fp->f_seekp == 0) { rc = dnode_read(spa, &fp->f_dnode, 0, &fp->f_zap_type, sizeof(fp->f_zap_type)); if (rc) return (rc); if (fp->f_zap_type == ZBT_MICRO) { fp->f_seekp = offsetof(mzap_phys_t, mz_chunk); } else { rc = dnode_read(spa, &fp->f_dnode, offsetof(zap_phys_t, zap_num_leafs), &fp->f_num_leafs, sizeof(fp->f_num_leafs)); if (rc) return (rc); fp->f_seekp = bsize; fp->f_zap_leaf = malloc(bsize); if (fp->f_zap_leaf == NULL) return (ENOMEM); rc = dnode_read(spa, &fp->f_dnode, fp->f_seekp, fp->f_zap_leaf, bsize); if (rc) return (rc); } } if (fp->f_zap_type == ZBT_MICRO) { mzap_next: if (fp->f_seekp >= bsize) return (ENOENT); rc = dnode_read(spa, &fp->f_dnode, fp->f_seekp, &mze, sizeof(mze)); if (rc) return (rc); fp->f_seekp += sizeof(mze); if (!mze.mze_name[0]) goto mzap_next; d->d_fileno = ZFS_DIRENT_OBJ(mze.mze_value); d->d_type = ZFS_DIRENT_TYPE(mze.mze_value); strcpy(d->d_name, mze.mze_name); d->d_namlen = strlen(d->d_name); return (0); } else { zap_leaf_t zl; zap_leaf_chunk_t *zc, *nc; int chunk; size_t namelen; char *p; uint64_t value; /* * Initialise this so we can use the ZAP size * calculating macros. */ zl.l_bs = ilog2(bsize); zl.l_phys = fp->f_zap_leaf; /* * Figure out which chunk we are currently looking at * and consider seeking to the next leaf. We use the * low bits of f_seekp as a simple chunk index. */ fzap_next: chunk = fp->f_seekp & (bsize - 1); if (chunk == ZAP_LEAF_NUMCHUNKS(&zl)) { fp->f_seekp = rounddown2(fp->f_seekp, bsize) + bsize; chunk = 0; /* * Check for EOF and read the new leaf. */ if (fp->f_seekp >= bsize * fp->f_num_leafs) return (ENOENT); rc = dnode_read(spa, &fp->f_dnode, fp->f_seekp, fp->f_zap_leaf, bsize); if (rc) return (rc); } zc = &ZAP_LEAF_CHUNK(&zl, chunk); fp->f_seekp++; if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY) goto fzap_next; namelen = zc->l_entry.le_name_numints; if (namelen > sizeof(d->d_name)) namelen = sizeof(d->d_name); /* * Paste the name back together. */ nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk); p = d->d_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); } d->d_name[sizeof(d->d_name) - 1] = 0; /* * Assume the first eight bytes of the value are * a uint64_t. */ value = fzap_leaf_value(&zl, zc); d->d_fileno = ZFS_DIRENT_OBJ(value); d->d_type = ZFS_DIRENT_TYPE(value); d->d_namlen = strlen(d->d_name); return (0); } } /* * if path is NULL, create mount structure, but do not add it to list. */ static int zfs_mount(const char *dev, const char *path, void **data) { struct zfs_devdesc *zfsdev; spa_t *spa; struct zfsmount *mnt; int rv; errno = 0; zfsdev = malloc(sizeof(*zfsdev)); if (zfsdev == NULL) return (errno); rv = zfs_parsedev(zfsdev, dev + 3, NULL); if (rv != 0) { free(zfsdev); return (rv); } spa = spa_find_by_dev(zfsdev); if (spa == NULL) return (ENXIO); mnt = calloc(1, sizeof(*mnt)); if (mnt != NULL && path != NULL) mnt->path = strdup(path); rv = errno; if (mnt != NULL) rv = zfs_mount_impl(spa, zfsdev->root_guid, mnt); free(zfsdev); if (rv == 0 && mnt != NULL && mnt->objset.os_type != DMU_OST_ZFS) { printf("Unexpected object set type %ju\n", (uintmax_t)mnt->objset.os_type); rv = EIO; } if (rv != 0) { if (mnt != NULL) free(mnt->path); free(mnt); return (rv); } if (mnt != NULL) { *data = mnt; if (path != NULL) STAILQ_INSERT_TAIL(&zfsmount, mnt, next); } return (rv); } static int zfs_unmount(const char *dev, void *data) { struct zfsmount *mnt = data; STAILQ_REMOVE(&zfsmount, mnt, zfsmount, next); free(mnt->path); free(mnt); return (0); } static int vdev_read(vdev_t *vdev, void *priv, off_t offset, void *buf, size_t bytes) { int fd, ret; size_t res, head, tail, total_size, full_sec_size; unsigned secsz, do_tail_read; off_t start_sec; char *outbuf, *bouncebuf; fd = (uintptr_t) priv; outbuf = (char *) buf; bouncebuf = NULL; ret = ioctl(fd, DIOCGSECTORSIZE, &secsz); if (ret != 0) return (ret); /* * Handling reads of arbitrary offset and size - multi-sector case * and single-sector case. * * Multi-sector Case * (do_tail_read = true if tail > 0) * * |<----------------------total_size--------------------->| * | | * |<--head-->|<--------------bytes------------>|<--tail-->| * | | | | * | | |<~full_sec_size~>| | | * +------------------+ +------------------+ * | |0101010| . . . |0101011| | * +------------------+ +------------------+ * start_sec start_sec + n * * * Single-sector Case * (do_tail_read = false) * * |<------total_size = secsz----->| * | | * |<-head->|<---bytes--->|<-tail->| * +-------------------------------+ * | |0101010101010| | * +-------------------------------+ * start_sec */ start_sec = offset / secsz; head = offset % secsz; total_size = roundup2(head + bytes, secsz); tail = total_size - (head + bytes); do_tail_read = ((tail > 0) && (head + bytes > secsz)); full_sec_size = total_size; if (head > 0) full_sec_size -= secsz; if (do_tail_read) full_sec_size -= secsz; /* Return of partial sector data requires a bounce buffer. */ if ((head > 0) || do_tail_read || bytes < secsz) { bouncebuf = malloc(secsz); if (bouncebuf == NULL) { printf("vdev_read: out of memory\n"); return (ENOMEM); } } if (lseek(fd, start_sec * secsz, SEEK_SET) == -1) { ret = errno; goto error; } /* Partial data return from first sector */ if (head > 0) { res = read(fd, bouncebuf, secsz); if (res != secsz) { ret = EIO; goto error; } memcpy(outbuf, bouncebuf + head, min(secsz - head, bytes)); outbuf += min(secsz - head, bytes); } /* * Full data return from read sectors. * Note, there is still corner case where we read * from sector boundary, but less than sector size, e.g. reading 512B * from 4k sector. */ if (full_sec_size > 0) { if (bytes < full_sec_size) { res = read(fd, bouncebuf, secsz); if (res != secsz) { ret = EIO; goto error; } memcpy(outbuf, bouncebuf, bytes); } else { res = read(fd, outbuf, full_sec_size); if (res != full_sec_size) { ret = EIO; goto error; } outbuf += full_sec_size; } } /* Partial data return from last sector */ if (do_tail_read) { res = read(fd, bouncebuf, secsz); if (res != secsz) { ret = EIO; goto error; } memcpy(outbuf, bouncebuf, secsz - tail); } ret = 0; error: free(bouncebuf); return (ret); } static int vdev_write(vdev_t *vdev, off_t offset, void *buf, size_t bytes) { int fd, ret; size_t head, tail, total_size, full_sec_size; unsigned secsz, do_tail_write; off_t start_sec; ssize_t res; char *outbuf, *bouncebuf; fd = (uintptr_t)vdev->v_priv; outbuf = (char *)buf; bouncebuf = NULL; ret = ioctl(fd, DIOCGSECTORSIZE, &secsz); if (ret != 0) return (ret); start_sec = offset / secsz; head = offset % secsz; total_size = roundup2(head + bytes, secsz); tail = total_size - (head + bytes); do_tail_write = ((tail > 0) && (head + bytes > secsz)); full_sec_size = total_size; if (head > 0) full_sec_size -= secsz; if (do_tail_write) full_sec_size -= secsz; /* Partial sector write requires a bounce buffer. */ if ((head > 0) || do_tail_write || bytes < secsz) { bouncebuf = malloc(secsz); if (bouncebuf == NULL) { printf("vdev_write: out of memory\n"); return (ENOMEM); } } if (lseek(fd, start_sec * secsz, SEEK_SET) == -1) { ret = errno; goto error; } /* Partial data for first sector */ if (head > 0) { res = read(fd, bouncebuf, secsz); if ((unsigned)res != secsz) { ret = EIO; goto error; } memcpy(bouncebuf + head, outbuf, min(secsz - head, bytes)); (void) lseek(fd, -secsz, SEEK_CUR); res = write(fd, bouncebuf, secsz); if ((unsigned)res != secsz) { ret = EIO; goto error; } outbuf += min(secsz - head, bytes); } /* * Full data write to sectors. * Note, there is still corner case where we write * to sector boundary, but less than sector size, e.g. write 512B * to 4k sector. */ if (full_sec_size > 0) { if (bytes < full_sec_size) { res = read(fd, bouncebuf, secsz); if ((unsigned)res != secsz) { ret = EIO; goto error; } memcpy(bouncebuf, outbuf, bytes); (void) lseek(fd, -secsz, SEEK_CUR); res = write(fd, bouncebuf, secsz); if ((unsigned)res != secsz) { ret = EIO; goto error; } } else { res = write(fd, outbuf, full_sec_size); if ((unsigned)res != full_sec_size) { ret = EIO; goto error; } outbuf += full_sec_size; } } /* Partial data write to last sector */ if (do_tail_write) { res = read(fd, bouncebuf, secsz); if ((unsigned)res != secsz) { ret = EIO; goto error; } memcpy(bouncebuf, outbuf, secsz - tail); (void) lseek(fd, -secsz, SEEK_CUR); res = write(fd, bouncebuf, secsz); if ((unsigned)res != secsz) { ret = EIO; goto error; } } ret = 0; error: free(bouncebuf); return (ret); } static int zfs_dev_init(void) { spa_t *spa; spa_t *next; spa_t *prev; zfs_init(); if (archsw.arch_zfs_probe == NULL) return (ENXIO); archsw.arch_zfs_probe(); prev = NULL; spa = STAILQ_FIRST(&zfs_pools); while (spa != NULL) { next = STAILQ_NEXT(spa, spa_link); if (zfs_spa_init(spa)) { if (prev == NULL) STAILQ_REMOVE_HEAD(&zfs_pools, spa_link); else STAILQ_REMOVE_AFTER(&zfs_pools, prev, spa_link); } else prev = spa; spa = next; } return (0); } struct zfs_probe_args { int fd; const char *devname; uint64_t *pool_guid; u_int secsz; }; static int zfs_diskread(void *arg, void *buf, size_t blocks, uint64_t offset) { struct zfs_probe_args *ppa; ppa = (struct zfs_probe_args *)arg; return (vdev_read(NULL, (void *)(uintptr_t)ppa->fd, offset * ppa->secsz, buf, blocks * ppa->secsz)); } static int zfs_probe(int fd, uint64_t *pool_guid) { spa_t *spa; int ret; spa = NULL; ret = vdev_probe(vdev_read, vdev_write, (void *)(uintptr_t)fd, &spa); if (ret == 0 && pool_guid != NULL) if (*pool_guid == 0) *pool_guid = spa->spa_guid; return (ret); } static int zfs_probe_partition(void *arg, const char *partname, const struct ptable_entry *part) { struct zfs_probe_args *ppa, pa; struct ptable *table; char devname[32]; int ret; /* Probe only freebsd-zfs and freebsd partitions */ if (part->type != PART_FREEBSD && part->type != PART_FREEBSD_ZFS) return (0); ppa = (struct zfs_probe_args *)arg; strncpy(devname, ppa->devname, strlen(ppa->devname) - 1); devname[strlen(ppa->devname) - 1] = '\0'; snprintf(devname, sizeof(devname), "%s%s:", devname, partname); pa.fd = open(devname, O_RDWR); if (pa.fd == -1) return (0); ret = zfs_probe(pa.fd, ppa->pool_guid); if (ret == 0) return (0); /* Do we have BSD label here? */ if (part->type == PART_FREEBSD) { pa.devname = devname; pa.pool_guid = ppa->pool_guid; pa.secsz = ppa->secsz; table = ptable_open(&pa, part->end - part->start + 1, ppa->secsz, zfs_diskread); if (table != NULL) { ptable_iterate(table, &pa, zfs_probe_partition); ptable_close(table); } } close(pa.fd); return (0); } /* * Return bootenv nvlist from pool label. */ int zfs_get_bootenv(void *vdev, nvlist_t **benvp) { struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; nvlist_t *benv = NULL; vdev_t *vd; spa_t *spa; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); if (spa->spa_bootenv == NULL) { STAILQ_FOREACH(vd, &spa->spa_root_vdev->v_children, v_childlink) { benv = vdev_read_bootenv(vd); if (benv != NULL) break; } spa->spa_bootenv = benv; } else { benv = spa->spa_bootenv; } if (benv == NULL) return (ENOENT); *benvp = benv; return (0); } /* * Store nvlist to pool label bootenv area. Also updates cached pointer in spa. */ int zfs_set_bootenv(void *vdev, nvlist_t *benv) { struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; spa_t *spa; vdev_t *vd; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); STAILQ_FOREACH(vd, &spa->spa_root_vdev->v_children, v_childlink) { vdev_write_bootenv(vd, benv); } spa->spa_bootenv = benv; return (0); } /* * Get bootonce value by key. The bootonce pair is removed * from the bootenv nvlist and the remaining nvlist is committed back to disk. */ int zfs_get_bootonce(void *vdev, const char *key, char *buf, size_t size) { nvlist_t *benv; char *result = NULL; int result_size, rv; if ((rv = zfs_get_bootenv(vdev, &benv)) != 0) return (rv); if ((rv = nvlist_find(benv, key, DATA_TYPE_STRING, NULL, &result, &result_size)) == 0) { if (result_size == 0) { /* ignore empty string */ rv = ENOENT; } else { size = MIN((size_t)result_size + 1, size); strlcpy(buf, result, size); } (void) nvlist_remove(benv, key, DATA_TYPE_STRING); (void) zfs_set_bootenv(vdev, benv); } return (rv); } /* * nvstore backend. */ static int zfs_nvstore_setter(void *, int, const char *, const void *, size_t); static int zfs_nvstore_setter_str(void *, const char *, const char *, const char *); static int zfs_nvstore_unset_impl(void *, const char *, bool); static int zfs_nvstore_setenv(void *, void *); /* * nvstore is only present for current rootfs pool. */ static int zfs_nvstore_sethook(struct env_var *ev, int flags __unused, const void *value) { struct zfs_devdesc *dev; int rv; archsw.arch_getdev((void **)&dev, NULL, NULL); if (dev == NULL) return (ENXIO); rv = zfs_nvstore_setter_str(dev, NULL, ev->ev_name, value); free(dev); return (rv); } /* * nvstore is only present for current rootfs pool. */ static int zfs_nvstore_unsethook(struct env_var *ev) { struct zfs_devdesc *dev; int rv; archsw.arch_getdev((void **)&dev, NULL, NULL); if (dev == NULL) return (ENXIO); rv = zfs_nvstore_unset_impl(dev, ev->ev_name, false); free(dev); return (rv); } static int zfs_nvstore_getter(void *vdev, const char *name, void **data) { struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; spa_t *spa; nvlist_t *nv; char *str, **ptr; int size; int rv; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); if (spa->spa_bootenv == NULL) return (ENXIO); if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST, NULL, &nv, NULL) != 0) return (ENOENT); rv = nvlist_find(nv, name, DATA_TYPE_STRING, NULL, &str, &size); if (rv == 0) { ptr = (char **)data; asprintf(ptr, "%.*s", size, str); if (*data == NULL) rv = ENOMEM; } nvlist_destroy(nv); return (rv); } static int zfs_nvstore_setter(void *vdev, int type, const char *name, const void *data, size_t size) { struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; spa_t *spa; nvlist_t *nv; int rv; bool env_set = true; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); if (spa->spa_bootenv == NULL) return (ENXIO); if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST, NULL, &nv, NULL) != 0) { nv = nvlist_create(NV_UNIQUE_NAME); if (nv == NULL) return (ENOMEM); } rv = 0; switch (type) { case DATA_TYPE_INT8: if (size != sizeof (int8_t)) { rv = EINVAL; break; } rv = nvlist_add_int8(nv, name, *(int8_t *)data); break; case DATA_TYPE_INT16: if (size != sizeof (int16_t)) { rv = EINVAL; break; } rv = nvlist_add_int16(nv, name, *(int16_t *)data); break; case DATA_TYPE_INT32: if (size != sizeof (int32_t)) { rv = EINVAL; break; } rv = nvlist_add_int32(nv, name, *(int32_t *)data); break; case DATA_TYPE_INT64: if (size != sizeof (int64_t)) { rv = EINVAL; break; } rv = nvlist_add_int64(nv, name, *(int64_t *)data); break; case DATA_TYPE_BYTE: if (size != sizeof (uint8_t)) { rv = EINVAL; break; } rv = nvlist_add_byte(nv, name, *(int8_t *)data); break; case DATA_TYPE_UINT8: if (size != sizeof (uint8_t)) { rv = EINVAL; break; } rv = nvlist_add_uint8(nv, name, *(int8_t *)data); break; case DATA_TYPE_UINT16: if (size != sizeof (uint16_t)) { rv = EINVAL; break; } rv = nvlist_add_uint16(nv, name, *(uint16_t *)data); break; case DATA_TYPE_UINT32: if (size != sizeof (uint32_t)) { rv = EINVAL; break; } rv = nvlist_add_uint32(nv, name, *(uint32_t *)data); break; case DATA_TYPE_UINT64: if (size != sizeof (uint64_t)) { rv = EINVAL; break; } rv = nvlist_add_uint64(nv, name, *(uint64_t *)data); break; case DATA_TYPE_STRING: rv = nvlist_add_string(nv, name, data); break; case DATA_TYPE_BOOLEAN_VALUE: if (size != sizeof (boolean_t)) { rv = EINVAL; break; } rv = nvlist_add_boolean_value(nv, name, *(boolean_t *)data); break; default: rv = EINVAL; break; } if (rv == 0) { rv = nvlist_add_nvlist(spa->spa_bootenv, OS_NVSTORE, nv); if (rv == 0) { rv = zfs_set_bootenv(vdev, spa->spa_bootenv); } if (rv == 0) { if (env_set) { rv = zfs_nvstore_setenv(vdev, nvpair_find(nv, name)); } else { env_discard(env_getenv(name)); rv = 0; } } } nvlist_destroy(nv); return (rv); } static int get_int64(const char *data, int64_t *ip) { char *end; int64_t val; errno = 0; val = strtoll(data, &end, 0); if (errno != 0 || *data == '\0' || *end != '\0') return (EINVAL); *ip = val; return (0); } static int get_uint64(const char *data, uint64_t *ip) { char *end; uint64_t val; errno = 0; val = strtoull(data, &end, 0); if (errno != 0 || *data == '\0' || *end != '\0') return (EINVAL); *ip = val; return (0); } /* * Translate textual data to data type. If type is not set, and we are * creating new pair, use DATA_TYPE_STRING. */ static int zfs_nvstore_setter_str(void *vdev, const char *type, const char *name, const char *data) { struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; spa_t *spa; nvlist_t *nv; int rv; data_type_t dt; int64_t val; uint64_t uval; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); if (spa->spa_bootenv == NULL) return (ENXIO); if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST, NULL, &nv, NULL) != 0) { nv = NULL; } if (type == NULL) { nvp_header_t *nvh; /* * if there is no existing pair, default to string. * Otherwise, use type from existing pair. */ nvh = nvpair_find(nv, name); if (nvh == NULL) { dt = DATA_TYPE_STRING; } else { nv_string_t *nvp_name; nv_pair_data_t *nvp_data; nvp_name = (nv_string_t *)(nvh + 1); nvp_data = (nv_pair_data_t *)(&nvp_name->nv_data[0] + NV_ALIGN4(nvp_name->nv_size)); dt = nvp_data->nv_type; } } else { dt = nvpair_type_from_name(type); } nvlist_destroy(nv); rv = 0; switch (dt) { case DATA_TYPE_INT8: rv = get_int64(data, &val); if (rv == 0) { int8_t v = val; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } break; case DATA_TYPE_INT16: rv = get_int64(data, &val); if (rv == 0) { int16_t v = val; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } break; case DATA_TYPE_INT32: rv = get_int64(data, &val); if (rv == 0) { int32_t v = val; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } break; case DATA_TYPE_INT64: rv = get_int64(data, &val); if (rv == 0) { rv = zfs_nvstore_setter(vdev, dt, name, &val, sizeof (val)); } break; case DATA_TYPE_BYTE: rv = get_uint64(data, &uval); if (rv == 0) { uint8_t v = uval; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } break; case DATA_TYPE_UINT8: rv = get_uint64(data, &uval); if (rv == 0) { uint8_t v = uval; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } break; case DATA_TYPE_UINT16: rv = get_uint64(data, &uval); if (rv == 0) { uint16_t v = uval; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } break; case DATA_TYPE_UINT32: rv = get_uint64(data, &uval); if (rv == 0) { uint32_t v = uval; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } break; case DATA_TYPE_UINT64: rv = get_uint64(data, &uval); if (rv == 0) { rv = zfs_nvstore_setter(vdev, dt, name, &uval, sizeof (uval)); } break; case DATA_TYPE_STRING: rv = zfs_nvstore_setter(vdev, dt, name, data, strlen(data) + 1); break; case DATA_TYPE_BOOLEAN_VALUE: rv = get_int64(data, &val); if (rv == 0) { boolean_t v = val; rv = zfs_nvstore_setter(vdev, dt, name, &v, sizeof (v)); } default: rv = EINVAL; } return (rv); } static int zfs_nvstore_unset_impl(void *vdev, const char *name, bool unset_env) { struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; spa_t *spa; nvlist_t *nv; int rv; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); if (spa->spa_bootenv == NULL) return (ENXIO); if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST, NULL, &nv, NULL) != 0) return (ENOENT); rv = nvlist_remove(nv, name, DATA_TYPE_UNKNOWN); if (rv == 0) { if (nvlist_next_nvpair(nv, NULL) == NULL) { rv = nvlist_remove(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST); } else { rv = nvlist_add_nvlist(spa->spa_bootenv, OS_NVSTORE, nv); } if (rv == 0) rv = zfs_set_bootenv(vdev, spa->spa_bootenv); } if (unset_env) env_discard(env_getenv(name)); return (rv); } static int zfs_nvstore_unset(void *vdev, const char *name) { return (zfs_nvstore_unset_impl(vdev, name, true)); } static int zfs_nvstore_print(void *vdev __unused, void *ptr) { nvpair_print(ptr, 0); return (0); } /* * Create environment variable from nvpair. * set hook will update nvstore with new value, unset hook will remove * variable from nvstore. */ static int zfs_nvstore_setenv(void *vdev __unused, void *ptr) { nvp_header_t *nvh = ptr; nv_string_t *nvp_name, *nvp_value; nv_pair_data_t *nvp_data; char *name, *value; int rv = 0; if (nvh == NULL) return (ENOENT); nvp_name = (nv_string_t *)(nvh + 1); nvp_data = (nv_pair_data_t *)(&nvp_name->nv_data[0] + NV_ALIGN4(nvp_name->nv_size)); if ((name = nvstring_get(nvp_name)) == NULL) return (ENOMEM); value = NULL; switch (nvp_data->nv_type) { case DATA_TYPE_BYTE: case DATA_TYPE_UINT8: (void) asprintf(&value, "%uc", *(unsigned *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_INT8: (void) asprintf(&value, "%c", *(int *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_INT16: (void) asprintf(&value, "%hd", *(short *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_UINT16: (void) asprintf(&value, "%hu", *(unsigned short *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_BOOLEAN_VALUE: case DATA_TYPE_INT32: (void) asprintf(&value, "%d", *(int *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_UINT32: (void) asprintf(&value, "%u", *(unsigned *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_INT64: (void) asprintf(&value, "%jd", (intmax_t)*(int64_t *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_UINT64: (void) asprintf(&value, "%ju", (uintmax_t)*(uint64_t *)&nvp_data->nv_data[0]); if (value == NULL) rv = ENOMEM; break; case DATA_TYPE_STRING: nvp_value = (nv_string_t *)&nvp_data->nv_data[0]; if ((value = nvstring_get(nvp_value)) == NULL) { rv = ENOMEM; break; } break; default: rv = EINVAL; break; } if (value != NULL) { rv = env_setenv(name, EV_VOLATILE | EV_NOHOOK, value, zfs_nvstore_sethook, zfs_nvstore_unsethook); free(value); } free(name); return (rv); } static int zfs_nvstore_iterate(void *vdev, int (*cb)(void *, void *)) { struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; spa_t *spa; nvlist_t *nv; nvp_header_t *nvh; int rv; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); if (spa->spa_bootenv == NULL) return (ENXIO); if (nvlist_find(spa->spa_bootenv, OS_NVSTORE, DATA_TYPE_NVLIST, NULL, &nv, NULL) != 0) return (ENOENT); rv = 0; nvh = NULL; while ((nvh = nvlist_next_nvpair(nv, nvh)) != NULL) { rv = cb(vdev, nvh); if (rv != 0) break; } return (rv); } nvs_callbacks_t nvstore_zfs_cb = { .nvs_getter = zfs_nvstore_getter, .nvs_setter = zfs_nvstore_setter, .nvs_setter_str = zfs_nvstore_setter_str, .nvs_unset = zfs_nvstore_unset, .nvs_print = zfs_nvstore_print, .nvs_iterate = zfs_nvstore_iterate }; int zfs_attach_nvstore(void *vdev) { struct zfs_devdesc *dev = vdev; spa_t *spa; uint64_t version; int rv; if (dev->dd.d_dev->dv_type != DEVT_ZFS) return (ENOTSUP); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); rv = nvlist_find(spa->spa_bootenv, BOOTENV_VERSION, DATA_TYPE_UINT64, NULL, &version, NULL); if (rv != 0 || version != VB_NVLIST) { return (ENXIO); } dev = malloc(sizeof (*dev)); if (dev == NULL) return (ENOMEM); memcpy(dev, vdev, sizeof (*dev)); rv = nvstore_init(spa->spa_name, &nvstore_zfs_cb, dev); if (rv != 0) free(dev); else rv = zfs_nvstore_iterate(dev, zfs_nvstore_setenv); return (rv); } int zfs_probe_dev(const char *devname, uint64_t *pool_guid) { struct ptable *table; struct zfs_probe_args pa; uint64_t mediasz; int ret; if (pool_guid) *pool_guid = 0; pa.fd = open(devname, O_RDWR); if (pa.fd == -1) return (ENXIO); /* Probe the whole disk */ ret = zfs_probe(pa.fd, pool_guid); if (ret == 0) return (0); /* Probe each partition */ ret = ioctl(pa.fd, DIOCGMEDIASIZE, &mediasz); if (ret == 0) ret = ioctl(pa.fd, DIOCGSECTORSIZE, &pa.secsz); if (ret == 0) { pa.devname = devname; pa.pool_guid = pool_guid; table = ptable_open(&pa, mediasz / pa.secsz, pa.secsz, zfs_diskread); if (table != NULL) { ptable_iterate(table, &pa, zfs_probe_partition); ptable_close(table); } } close(pa.fd); if (pool_guid && *pool_guid == 0) ret = ENXIO; return (ret); } /* * Print information about ZFS pools */ static int zfs_dev_print(int verbose) { spa_t *spa; char line[80]; int ret = 0; if (STAILQ_EMPTY(&zfs_pools)) return (0); printf("%s devices:", zfs_dev.dv_name); if ((ret = pager_output("\n")) != 0) return (ret); if (verbose) { return (spa_all_status()); } STAILQ_FOREACH(spa, &zfs_pools, spa_link) { snprintf(line, sizeof(line), " zfs:%s\n", spa->spa_name); ret = pager_output(line); if (ret != 0) break; } return (ret); } /* * Attempt to open the pool described by (dev) for use by (f). */ static int zfs_dev_open(struct open_file *f, ...) { va_list args; struct zfs_devdesc *dev; struct zfsmount *mount; spa_t *spa; int rv; va_start(args, f); dev = va_arg(args, struct zfs_devdesc *); va_end(args); if ((spa = spa_find_by_dev(dev)) == NULL) return (ENXIO); STAILQ_FOREACH(mount, &zfsmount, next) { if (spa->spa_guid == mount->spa->spa_guid) break; } rv = 0; /* This device is not set as currdev, mount us private copy. */ if (mount == NULL) - rv = zfs_mount(zfs_fmtdev(&dev->dd), NULL, (void **)&mount); + rv = zfs_mount(devformat(&dev->dd), NULL, (void **)&mount); if (rv == 0) { f->f_devdata = mount; free(dev); } return (rv); } static int zfs_dev_close(struct open_file *f) { struct zfsmount *mnt, *mount; mnt = f->f_devdata; STAILQ_FOREACH(mount, &zfsmount, next) { if (mnt->spa->spa_guid == mount->spa->spa_guid) break; } /* * devclose() will free f->f_devdata, but since we do have * pointer to zfsmount structure in f->f_devdata, and * zfs_unmount() will also free the zfsmount structure, * we will get double free. To prevent double free, * we must set f_devdata to NULL there. */ if (mount != NULL) f->f_devdata = NULL; return (0); } static int zfs_dev_strategy(void *devdata, int rw, daddr_t dblk, size_t size, char *buf, size_t *rsize) { return (ENOSYS); } struct devsw zfs_dev = { .dv_name = "zfs", .dv_type = DEVT_ZFS, .dv_init = zfs_dev_init, .dv_strategy = zfs_dev_strategy, .dv_open = zfs_dev_open, .dv_close = zfs_dev_close, .dv_ioctl = noioctl, .dv_print = zfs_dev_print, .dv_cleanup = nullsys, .dv_fmtdev = zfs_fmtdev, }; int zfs_parsedev(struct zfs_devdesc *dev, const char *devspec, const char **path) { static char rootname[ZFS_MAXNAMELEN]; static char poolname[ZFS_MAXNAMELEN]; spa_t *spa; const char *end; const char *np; const char *sep; int rv; np = devspec; if (*np != ':') return (EINVAL); np++; end = strrchr(np, ':'); if (end == NULL) return (EINVAL); sep = strchr(np, '/'); if (sep == NULL || sep >= end) sep = end; memcpy(poolname, np, sep - np); poolname[sep - np] = '\0'; if (sep < end) { sep++; memcpy(rootname, sep, end - sep); rootname[end - sep] = '\0'; } else rootname[0] = '\0'; spa = spa_find_by_name(poolname); if (!spa) return (ENXIO); dev->pool_guid = spa->spa_guid; rv = zfs_lookup_dataset(spa, rootname, &dev->root_guid); if (rv != 0) return (rv); if (path != NULL) *path = (*end == '\0') ? end : end + 1; dev->dd.d_dev = &zfs_dev; return (0); } char * zfs_fmtdev(struct devdesc *vdev) { static char rootname[ZFS_MAXNAMELEN]; static char buf[2 * ZFS_MAXNAMELEN + 8]; struct zfs_devdesc *dev = (struct zfs_devdesc *)vdev; spa_t *spa; buf[0] = '\0'; if (vdev->d_dev->dv_type != DEVT_ZFS) return (buf); /* Do we have any pools? */ spa = STAILQ_FIRST(&zfs_pools); if (spa == NULL) return (buf); if (dev->pool_guid == 0) dev->pool_guid = spa->spa_guid; else spa = spa_find_by_guid(dev->pool_guid); if (spa == NULL) { printf("ZFS: can't find pool by guid\n"); return (buf); } if (dev->root_guid == 0 && zfs_get_root(spa, &dev->root_guid)) { printf("ZFS: can't find root filesystem\n"); return (buf); } if (zfs_rlookup(spa, dev->root_guid, rootname)) { printf("ZFS: can't find filesystem by guid\n"); return (buf); } if (rootname[0] == '\0') snprintf(buf, sizeof(buf), "%s:%s:", dev->dd.d_dev->dv_name, spa->spa_name); else snprintf(buf, sizeof(buf), "%s:%s/%s:", dev->dd.d_dev->dv_name, spa->spa_name, rootname); return (buf); } static int split_devname(const char *name, char *poolname, size_t size, const char **dsnamep) { const char *dsname; size_t len; ASSERT(name != NULL); ASSERT(poolname != NULL); len = strlen(name); dsname = strchr(name, '/'); if (dsname != NULL) { len = dsname - name; dsname++; } else dsname = ""; if (len + 1 > size) return (EINVAL); strlcpy(poolname, name, len + 1); if (dsnamep != NULL) *dsnamep = dsname; return (0); } int zfs_list(const char *name) { static char poolname[ZFS_MAXNAMELEN]; uint64_t objid; spa_t *spa; const char *dsname; int rv; if (split_devname(name, poolname, sizeof(poolname), &dsname) != 0) return (EINVAL); spa = spa_find_by_name(poolname); if (!spa) return (ENXIO); rv = zfs_lookup_dataset(spa, dsname, &objid); if (rv != 0) return (rv); return (zfs_list_dataset(spa, objid)); } void init_zfs_boot_options(const char *currdev_in) { char poolname[ZFS_MAXNAMELEN]; char *beroot, *currdev; spa_t *spa; int currdev_len; const char *dsname; currdev = NULL; currdev_len = strlen(currdev_in); if (currdev_len == 0) return; if (strncmp(currdev_in, "zfs:", 4) != 0) return; currdev = strdup(currdev_in); if (currdev == NULL) return; /* Remove the trailing : */ currdev[currdev_len - 1] = '\0'; setenv("zfs_be_active", currdev, 1); setenv("zfs_be_currpage", "1", 1); /* Remove the last element (current bootenv) */ beroot = strrchr(currdev, '/'); if (beroot != NULL) beroot[0] = '\0'; beroot = strchr(currdev, ':') + 1; setenv("zfs_be_root", beroot, 1); if (split_devname(beroot, poolname, sizeof(poolname), &dsname) != 0) return; spa = spa_find_by_name(poolname); if (spa == NULL) return; zfs_bootenv_initial("bootenvs", spa, beroot, dsname, 0); zfs_checkpoints_initial(spa, beroot, dsname); free(currdev); } static void zfs_checkpoints_initial(spa_t *spa, const char *name, const char *dsname) { char envname[32]; if (spa->spa_uberblock_checkpoint.ub_checkpoint_txg != 0) { snprintf(envname, sizeof(envname), "zpool_checkpoint"); setenv(envname, name, 1); spa->spa_uberblock = &spa->spa_uberblock_checkpoint; spa->spa_mos = &spa->spa_mos_checkpoint; zfs_bootenv_initial("bootenvs_check", spa, name, dsname, 1); spa->spa_uberblock = &spa->spa_uberblock_master; spa->spa_mos = &spa->spa_mos_master; } } static void zfs_bootenv_initial(const char *envprefix, spa_t *spa, const char *rootname, const char *dsname, int checkpoint) { char envname[32], envval[256]; uint64_t objid; int bootenvs_idx, rv; SLIST_INIT(&zfs_be_head); zfs_env_count = 0; rv = zfs_lookup_dataset(spa, dsname, &objid); if (rv != 0) return; rv = zfs_callback_dataset(spa, objid, zfs_belist_add); bootenvs_idx = 0; /* Populate the initial environment variables */ SLIST_FOREACH_SAFE(zfs_be, &zfs_be_head, entries, zfs_be_tmp) { /* Enumerate all bootenvs for general usage */ snprintf(envname, sizeof(envname), "%s[%d]", envprefix, bootenvs_idx); snprintf(envval, sizeof(envval), "zfs:%s%s/%s", checkpoint ? "!" : "", rootname, zfs_be->name); rv = setenv(envname, envval, 1); if (rv != 0) break; bootenvs_idx++; } snprintf(envname, sizeof(envname), "%s_count", envprefix); snprintf(envval, sizeof(envval), "%d", bootenvs_idx); setenv(envname, envval, 1); /* Clean up the SLIST of ZFS BEs */ while (!SLIST_EMPTY(&zfs_be_head)) { zfs_be = SLIST_FIRST(&zfs_be_head); SLIST_REMOVE_HEAD(&zfs_be_head, entries); free(zfs_be->name); free(zfs_be); } } int zfs_bootenv(const char *name) { char poolname[ZFS_MAXNAMELEN], *root; const char *dsname; char becount[4]; uint64_t objid; spa_t *spa; int rv, pages, perpage, currpage; if (name == NULL) return (EINVAL); if ((root = getenv("zfs_be_root")) == NULL) return (EINVAL); if (strcmp(name, root) != 0) { if (setenv("zfs_be_root", name, 1) != 0) return (ENOMEM); } SLIST_INIT(&zfs_be_head); zfs_env_count = 0; if (split_devname(name, poolname, sizeof(poolname), &dsname) != 0) return (EINVAL); spa = spa_find_by_name(poolname); if (!spa) return (ENXIO); rv = zfs_lookup_dataset(spa, dsname, &objid); if (rv != 0) return (rv); rv = zfs_callback_dataset(spa, objid, zfs_belist_add); /* Calculate and store the number of pages of BEs */ perpage = (ZFS_BE_LAST - ZFS_BE_FIRST + 1); pages = (zfs_env_count / perpage) + ((zfs_env_count % perpage) > 0 ? 1 : 0); snprintf(becount, 4, "%d", pages); if (setenv("zfs_be_pages", becount, 1) != 0) return (ENOMEM); /* Roll over the page counter if it has exceeded the maximum */ currpage = strtol(getenv("zfs_be_currpage"), NULL, 10); if (currpage > pages) { if (setenv("zfs_be_currpage", "1", 1) != 0) return (ENOMEM); } /* Populate the menu environment variables */ zfs_set_env(); /* Clean up the SLIST of ZFS BEs */ while (!SLIST_EMPTY(&zfs_be_head)) { zfs_be = SLIST_FIRST(&zfs_be_head); SLIST_REMOVE_HEAD(&zfs_be_head, entries); free(zfs_be->name); free(zfs_be); } return (rv); } int zfs_belist_add(const char *name, uint64_t value __unused) { /* Skip special datasets that start with a $ character */ if (strncmp(name, "$", 1) == 0) { return (0); } /* Add the boot environment to the head of the SLIST */ zfs_be = malloc(sizeof(struct zfs_be_entry)); if (zfs_be == NULL) { return (ENOMEM); } zfs_be->name = strdup(name); if (zfs_be->name == NULL) { free(zfs_be); return (ENOMEM); } SLIST_INSERT_HEAD(&zfs_be_head, zfs_be, entries); zfs_env_count++; return (0); } int zfs_set_env(void) { char envname[32], envval[256]; char *beroot, *pagenum; int rv, page, ctr; beroot = getenv("zfs_be_root"); if (beroot == NULL) { return (1); } pagenum = getenv("zfs_be_currpage"); if (pagenum != NULL) { page = strtol(pagenum, NULL, 10); } else { page = 1; } ctr = 1; rv = 0; zfs_env_index = ZFS_BE_FIRST; SLIST_FOREACH_SAFE(zfs_be, &zfs_be_head, entries, zfs_be_tmp) { /* Skip to the requested page number */ if (ctr <= ((ZFS_BE_LAST - ZFS_BE_FIRST + 1) * (page - 1))) { ctr++; continue; } snprintf(envname, sizeof(envname), "bootenvmenu_caption[%d]", zfs_env_index); snprintf(envval, sizeof(envval), "%s", zfs_be->name); rv = setenv(envname, envval, 1); if (rv != 0) { break; } snprintf(envname, sizeof(envname), "bootenvansi_caption[%d]", zfs_env_index); rv = setenv(envname, envval, 1); if (rv != 0){ break; } snprintf(envname, sizeof(envname), "bootenvmenu_command[%d]", zfs_env_index); rv = setenv(envname, "set_bootenv", 1); if (rv != 0){ break; } snprintf(envname, sizeof(envname), "bootenv_root[%d]", zfs_env_index); snprintf(envval, sizeof(envval), "zfs:%s/%s", beroot, zfs_be->name); rv = setenv(envname, envval, 1); if (rv != 0){ break; } zfs_env_index++; if (zfs_env_index > ZFS_BE_LAST) { break; } } for (; zfs_env_index <= ZFS_BE_LAST; zfs_env_index++) { snprintf(envname, sizeof(envname), "bootenvmenu_caption[%d]", zfs_env_index); (void)unsetenv(envname); snprintf(envname, sizeof(envname), "bootenvansi_caption[%d]", zfs_env_index); (void)unsetenv(envname); snprintf(envname, sizeof(envname), "bootenvmenu_command[%d]", zfs_env_index); (void)unsetenv(envname); snprintf(envname, sizeof(envname), "bootenv_root[%d]", zfs_env_index); (void)unsetenv(envname); } return (rv); } diff --git a/stand/userboot/userboot/main.c b/stand/userboot/userboot/main.c index d3f12ab1675e..d323d1f9789b 100644 --- a/stand/userboot/userboot/main.c +++ b/stand/userboot/userboot/main.c @@ -1,347 +1,347 @@ /*- * Copyright (c) 1998 Michael Smith * Copyright (c) 1998,2000 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$"); #include #include #include #include #include #include "bootstrap.h" #include "disk.h" #include "libuserboot.h" #if defined(USERBOOT_ZFS_SUPPORT) #include "libzfs.h" static void userboot_zfs_probe(void); static int userboot_zfs_found; #endif /* Minimum version required */ #define USERBOOT_VERSION USERBOOT_VERSION_3 #define LOADER_PATH "/boot/loader" #define INTERP_MARKER "$Interpreter:" #define MALLOCSZ (64*1024*1024) struct loader_callbacks *callbacks; void *callbacks_arg; static jmp_buf jb; struct arch_switch archsw; /* MI/MD interface boundary */ static void extract_currdev(void); static void check_interpreter(void); void delay(int usec) { CALLBACK(delay, usec); } time_t getsecs(void) { /* * userboot can't do netboot, so this implementation isn't strictly * required. Defining it avoids issues with BIND_NOW, and it doesn't * hurt to do it. */ return (time(NULL)); } void exit(int v) { CALLBACK(exit, v); longjmp(jb, 1); } static void check_interpreter(void) { struct stat st; size_t marklen, rdsize; const char *guest_interp, *my_interp; char *buf; int fd; /* * If we can't stat(2) or open(2) LOADER_PATH, then we'll fail by * simply letting us roll on with whatever interpreter we were compiled * with. This is likely not going to be an issue in reality. */ buf = NULL; if (stat(LOADER_PATH, &st) != 0) return; if ((fd = open(LOADER_PATH, O_RDONLY)) < 0) return; rdsize = st.st_size; buf = malloc(rdsize); if (buf == NULL) goto out; if (read(fd, buf, rdsize) < rdsize) goto out; marklen = strlen(INTERP_MARKER); my_interp = bootprog_interp + marklen; /* * Here we make the assumption that a loader binary without the * interpreter marker is a 4th one. All loader binaries going forward * should have this properly specified, so our assumption should always * be a good one. */ if ((guest_interp = memmem(buf, rdsize, INTERP_MARKER, marklen)) != NULL) guest_interp += marklen; else guest_interp = "4th"; /* * The guest interpreter may not have a version of loader that * specifies the interpreter installed. If that's the case, we'll * assume it's legacy (4th) and request a swap to that if we're * a Lua-userboot. */ if (strcmp(my_interp, guest_interp) != 0) CALLBACK(swap_interpreter, guest_interp); out: free(buf); close(fd); return; } void loader_main(struct loader_callbacks *cb, void *arg, int version, int ndisks) { static char mallocbuf[MALLOCSZ]; char *var; int i; if (version < USERBOOT_VERSION) abort(); callbacks = cb; callbacks_arg = arg; userboot_disk_maxunit = ndisks; /* * initialise the heap as early as possible. Once this is done, * alloc() is usable. */ setheap((void *)mallocbuf, (void *)(mallocbuf + sizeof(mallocbuf))); /* * Hook up the console */ cons_probe(); /* Set up currdev variable to have hooks in place. */ env_setenv("currdev", EV_VOLATILE, "", userboot_setcurrdev, env_nounset); printf("\n%s", bootprog_info); #if 0 printf("Memory: %ld k\n", memsize() / 1024); #endif setenv("LINES", "24", 1); /* optional */ /* * Set custom environment variables */ i = 0; while (1) { var = CALLBACK(getenv, i++); if (var == NULL) break; putenv(var); } archsw.arch_autoload = userboot_autoload; archsw.arch_getdev = userboot_getdev; archsw.arch_copyin = userboot_copyin; archsw.arch_copyout = userboot_copyout; archsw.arch_readin = userboot_readin; #if defined(USERBOOT_ZFS_SUPPORT) archsw.arch_zfs_probe = userboot_zfs_probe; #endif /* * Initialise the block cache. Set the upper limit. */ bcache_init(32768, 512); /* * 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)(); extract_currdev(); /* * Checking the interpreter isn't worth the overhead unless we * actually have the swap_interpreter callback, so we actually version * check here rather than later on. */ if (version >= USERBOOT_VERSION_5) check_interpreter(); if (setjmp(jb)) return; interact(); /* doesn't return */ exit(0); } static void set_currdev(const char *devname) { env_setenv("currdev", EV_VOLATILE, devname, userboot_setcurrdev, env_nounset); env_setenv("loaddev", EV_VOLATILE, devname, env_noset, env_nounset); } /* * Set the 'current device' by (if possible) recovering the boot device as * supplied by the initial bootstrap. */ static void extract_currdev(void) { struct disk_devdesc dev; struct devdesc *dd; #if defined(USERBOOT_ZFS_SUPPORT) struct zfs_devdesc zdev; char *buf = NULL; if (userboot_zfs_found) { /* Leave the pool/root guid's unassigned */ bzero(&zdev, sizeof(zdev)); zdev.dd.d_dev = &zfs_dev; - init_zfs_boot_options(zfs_fmtdev(&zdev.dd)); + init_zfs_boot_options(devformat(&zdev.dd)); dd = &zdev.dd; } else #endif if (userboot_disk_maxunit > 0) { dev.dd.d_dev = &userboot_disk; dev.dd.d_unit = 0; dev.d_slice = D_SLICEWILD; dev.d_partition = D_PARTWILD; /* * If we cannot auto-detect the partition type then * access the disk as a raw device. */ if (dev.dd.d_dev->dv_open(NULL, &dev)) { dev.d_slice = D_SLICENONE; dev.d_partition = D_PARTNONE; } dd = &dev.dd; } else { dev.dd.d_dev = &host_dev; dev.dd.d_unit = 0; dd = &dev.dd; } set_currdev(userboot_fmtdev(dd)); #if defined(USERBOOT_ZFS_SUPPORT) if (userboot_zfs_found) { buf = malloc(VDEV_PAD_SIZE); if (buf != NULL) { if (zfs_get_bootonce(&zdev, OS_BOOTONCE, buf, VDEV_PAD_SIZE) == 0) { printf("zfs bootonce: %s\n", buf); set_currdev(buf); setenv("zfs-bootonce", buf, 1); } free(buf); (void) zfs_attach_nvstore(&zdev); } } #endif } #if defined(USERBOOT_ZFS_SUPPORT) static void userboot_zfs_probe(void) { char devname[32]; uint64_t pool_guid; int unit; /* * Open all the disks we can find and see if we can reconstruct * ZFS pools from them. Record if any were found. */ for (unit = 0; unit < userboot_disk_maxunit; unit++) { sprintf(devname, "disk%d:", unit); pool_guid = 0; zfs_probe_dev(devname, &pool_guid); if (pool_guid != 0) userboot_zfs_found = 1; } } #endif COMMAND_SET(quit, "quit", "exit the loader", command_quit); static int command_quit(int argc, char *argv[]) { exit(USERBOOT_EXIT_QUIT); return (CMD_OK); } COMMAND_SET(reboot, "reboot", "reboot the system", command_reboot); static int command_reboot(int argc, char *argv[]) { exit(USERBOOT_EXIT_REBOOT); return (CMD_OK); }