static int
zfs_probe(const char *devpath, uint64_t *outpoolguid)
{
nvlist_t *config = NULL;
int ret = FSUR_UNRECOGNIZED;
int fd;
uint64_t guid;
printf("+zfs_probe : devpath %s\n", devpath);
if (outpoolguid == NULL)
goto out;
if ((fd = open(devpath, O_RDONLY)) < 0) {
printf("Could not open devpath %s : fd %d\n", devpath, fd);
goto out;
}
if (zpool_read_label(fd, &config) != 0) {
(void) close(fd);
goto out;
}
(void) close(fd);
if (config != NULL) {
ret = FSUR_RECOGNIZED;
*outpoolguid = (nvlist_lookup_uint64(config,
ZPOOL_CONFIG_POOL_GUID, &guid) == 0) ? guid : 0;
nvlist_free(config);
}
out:
printf("-zfs_probe : ret %d\n", ret);
return (ret);
}
static void
zpool_open_func(void *arg)
{
rdsk_node_t *rn = arg;
struct stat64 statbuf;
nvlist_t *config;
int fd;
if (rn->rn_nozpool)
return;
if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
/* symlink to a device that's no longer there */
if (errno == ENOENT)
nozpool_all_slices(rn->rn_avl, rn->rn_name);
return;
}
/*
* Ignore failed stats. We only want regular
* files, character devs and block devs.
*/
if (fstat64(fd, &statbuf) != 0 ||
(!S_ISREG(statbuf.st_mode) &&
!S_ISCHR(statbuf.st_mode) &&
!S_ISBLK(statbuf.st_mode))) {
(void) close(fd);
return;
}
/* this file is too small to hold a zpool */
if (S_ISREG(statbuf.st_mode) &&
statbuf.st_size < SPA_MINDEVSIZE) {
(void) close(fd);
return;
} else if (!S_ISREG(statbuf.st_mode)) {
/*
* Try to read the disk label first so we don't have to
* open a bunch of minor nodes that can't have a zpool.
*/
check_slices(rn->rn_avl, fd, rn->rn_name);
}
if ((zpool_read_label(fd, &config)) != 0) {
(void) close(fd);
(void) no_memory(rn->rn_hdl);
return;
}
(void) close(fd);
rn->rn_config = config;
if (config != NULL) {
assert(rn->rn_nozpool == B_FALSE);
}
}
/*
* Determines if the pool is in use. If so, it returns true and the state of
* the pool as well as the name of the pool. Both strings are allocated and
* must be freed by the caller.
*/
int
zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
boolean_t *inuse)
{
nvlist_t *config;
char *name;
boolean_t ret;
uint64_t guid, vdev_guid;
zpool_handle_t *zhp;
nvlist_t *pool_config;
uint64_t stateval, isspare;
aux_cbdata_t cb = { 0 };
boolean_t isactive;
*inuse = B_FALSE;
if (zpool_read_label(fd, &config, NULL) != 0 && errno == ENOMEM) {
(void) no_memory(hdl);
return (-1);
}
if (config == NULL)
return (0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
&stateval) == 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
&vdev_guid) == 0);
if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
&name) == 0);
verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&guid) == 0);
}
switch (stateval) {
case POOL_STATE_EXPORTED:
/*
* A pool with an exported state may in fact be imported
* read-only, so check the in-core state to see if it's
* active and imported read-only. If it is, set
* its state to active.
*/
if (pool_active(hdl, name, guid, &isactive) == 0 && isactive &&
(zhp = zpool_open_canfail(hdl, name)) != NULL) {
if (zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL))
stateval = POOL_STATE_ACTIVE;
/*
* All we needed the zpool handle for is the
* readonly prop check.
*/
zpool_close(zhp);
}
ret = B_TRUE;
break;
case POOL_STATE_ACTIVE:
/*
* For an active pool, we have to determine if it's really part
* of a currently active pool (in which case the pool will exist
* and the guid will be the same), or whether it's part of an
* active pool that was disconnected without being explicitly
* exported.
*/
if (pool_active(hdl, name, guid, &isactive) != 0) {
nvlist_free(config);
return (-1);
}
if (isactive) {
/*
* Because the device may have been removed while
* offlined, we only report it as active if the vdev is
* still present in the config. Otherwise, pretend like
* it's not in use.
*/
if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
(pool_config = zpool_get_config(zhp, NULL))
!= NULL) {
nvlist_t *nvroot;
verify(nvlist_lookup_nvlist(pool_config,
ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
ret = find_guid(nvroot, vdev_guid);
} else {
ret = B_FALSE;
}
/*
* If this is an active spare within another pool, we
* treat it like an unused hot spare. This allows the
* user to create a pool with a hot spare that currently
* in use within another pool. Since we return B_TRUE,
* libdiskmgt will continue to prevent generic consumers
* from using the device.
*/
if (ret && nvlist_lookup_uint64(config,
ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
stateval = POOL_STATE_SPARE;
if (zhp != NULL)
zpool_close(zhp);
} else {
stateval = POOL_STATE_POTENTIALLY_ACTIVE;
ret = B_TRUE;
}
break;
case POOL_STATE_SPARE:
/*
* For a hot spare, it can be either definitively in use, or
* potentially active. To determine if it's in use, we iterate
* over all pools in the system and search for one with a spare
* with a matching guid.
*
* Due to the shared nature of spares, we don't actually report
* the potentially active case as in use. This means the user
* can freely create pools on the hot spares of exported pools,
* but to do otherwise makes the resulting code complicated, and
* we end up having to deal with this case anyway.
*/
cb.cb_zhp = NULL;
cb.cb_guid = vdev_guid;
cb.cb_type = ZPOOL_CONFIG_SPARES;
if (zpool_iter(hdl, find_aux, &cb) == 1) {
name = (char *)zpool_get_name(cb.cb_zhp);
ret = B_TRUE;
} else {
ret = B_FALSE;
}
break;
case POOL_STATE_L2CACHE:
/*
* Check if any pool is currently using this l2cache device.
*/
cb.cb_zhp = NULL;
cb.cb_guid = vdev_guid;
cb.cb_type = ZPOOL_CONFIG_L2CACHE;
if (zpool_iter(hdl, find_aux, &cb) == 1) {
name = (char *)zpool_get_name(cb.cb_zhp);
ret = B_TRUE;
} else {
ret = B_FALSE;
}
break;
default:
ret = B_FALSE;
}
if (ret) {
if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
if (cb.cb_zhp)
zpool_close(cb.cb_zhp);
nvlist_free(config);
return (-1);
}
*state = (pool_state_t)stateval;
}
if (cb.cb_zhp)
zpool_close(cb.cb_zhp);
nvlist_free(config);
*inuse = ret;
return (0);
}
/*
* Use libblkid to quickly search for zfs devices
*/
static int
zpool_find_import_blkid(libzfs_handle_t *hdl, pool_list_t *pools)
{
blkid_cache cache;
blkid_dev_iterate iter;
blkid_dev dev;
const char *devname;
nvlist_t *config;
int fd, err, num_labels;
err = blkid_get_cache(&cache, NULL);
if (err != 0) {
(void) zfs_error_fmt(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "blkid_get_cache() %d"), err);
goto err_blkid1;
}
err = blkid_probe_all(cache);
if (err != 0) {
(void) zfs_error_fmt(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "blkid_probe_all() %d"), err);
goto err_blkid2;
}
iter = blkid_dev_iterate_begin(cache);
if (iter == NULL) {
(void) zfs_error_fmt(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "blkid_dev_iterate_begin()"));
goto err_blkid2;
}
err = blkid_dev_set_search(iter, "TYPE", "zfs_member");
if (err != 0) {
(void) zfs_error_fmt(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "blkid_dev_set_search() %d"), err);
goto err_blkid3;
}
while (blkid_dev_next(iter, &dev) == 0) {
devname = blkid_dev_devname(dev);
if ((fd = open(devname, O_RDONLY)) < 0)
continue;
err = zpool_read_label(fd, &config, &num_labels);
(void) close(fd);
if (err != 0) {
(void) no_memory(hdl);
goto err_blkid3;
}
if (config != NULL) {
err = add_config(hdl, pools, devname, 0,
num_labels, config);
if (err != 0)
goto err_blkid3;
}
}
err_blkid3:
blkid_dev_iterate_end(iter);
err_blkid2:
blkid_put_cache(cache);
err_blkid1:
return (err);
}
static void
zpool_open_func(void *arg)
{
rdsk_node_t *rn = arg;
#ifdef __APPLE__
struct stat statbuf;
#else
struct stat64 statbuf;
#endif
nvlist_t *config;
int num_labels;
int fd;
if (rn->rn_nozpool)
return;
#if defined (__linux__) || defined (__APPLE__)
/*
* Skip devices with well known prefixes there can be side effects
* when opening devices which need to be avoided.
*
* core - Symlink to /proc/kcore
* fd* - Floppy interface.
* fuse - Fuse control device.
* hpet - High Precision Event Timer
* lp* - Printer interface.
* parport* - Parallel port interface.
* ppp - Generic PPP driver.
* random - Random device
* rtc - Real Time Clock
* tty* - Generic serial interface.
* urandom - Random device.
* usbmon* - USB IO monitor.
* vcs* - Virtual console memory.
* watchdog - Watchdog must be closed in a special way.
*/
if ((strncmp(rn->rn_name, "core", 4) == 0) ||
(strncmp(rn->rn_name, "fd", 2) == 0) ||
(strncmp(rn->rn_name, "fuse", 4) == 0) ||
(strncmp(rn->rn_name, "hpet", 4) == 0) ||
(strncmp(rn->rn_name, "lp", 2) == 0) ||
(strncmp(rn->rn_name, "parport", 7) == 0) ||
(strncmp(rn->rn_name, "ppp", 3) == 0) ||
(strncmp(rn->rn_name, "random", 6) == 0) ||
(strncmp(rn->rn_name, "rtc", 3) == 0) ||
(strncmp(rn->rn_name, "tty", 3) == 0) ||
(strncmp(rn->rn_name, "urandom", 7) == 0) ||
(strncmp(rn->rn_name, "usbmon", 6) == 0) ||
(strncmp(rn->rn_name, "vcs", 3) == 0) ||
#ifdef __APPLE__
(strncmp(rn->rn_name, "pty", 3) == 0) || // lots, skip for speed
(strncmp(rn->rn_name, "com", 3) == 0) || // /dev/com_digidesign_semiface
#endif
(strncmp(rn->rn_name, "watchdog", 8) == 0))
return;
/*
* Ignore failed stats. We only want regular files and block devices.
*/
if (fstatat64(rn->rn_dfd, rn->rn_name, &statbuf, 0) != 0 ||
(!S_ISREG(statbuf.st_mode) && !S_ISBLK(statbuf.st_mode)))
return;
#ifdef __APPLE__
/* It is desirable to skip optical media as well, as they are
* also called /dev/diskX
*/
if (is_optical_media((char *)rn->rn_name))
return;
#endif
if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
/* symlink to a device that's no longer there */
if (errno == ENOENT)
nozpool_all_slices(rn->rn_avl, rn->rn_name);
return;
}
#else /* LINUX, APPLE -> IllumOS */
if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
/* symlink to a device that's no longer there */
if (errno == ENOENT)
nozpool_all_slices(rn->rn_avl, rn->rn_name);
return;
}
/*
* Ignore failed stats. We only want regular
* files, character devs and block devs.
*/
if (fstat64(fd, &statbuf) != 0 ||
(!S_ISREG(statbuf.st_mode) &&
!S_ISCHR(statbuf.st_mode) &&
!S_ISBLK(statbuf.st_mode))) {
(void) close(fd);
return;
}
#endif
/* this file is too small to hold a zpool */
if (S_ISREG(statbuf.st_mode) &&
statbuf.st_size < SPA_MINDEVSIZE) {
(void) close(fd);
return;
} else if (!S_ISREG(statbuf.st_mode)) {
/*
* Try to read the disk label first so we don't have to
* open a bunch of minor nodes that can't have a zpool.
*/
check_slices(rn->rn_avl, fd, rn->rn_name);
}
#ifdef __APPLE__
int32_t blksz = 0;
if (S_ISBLK(statbuf.st_mode) &&
(ioctl(fd, DKIOCGETBLOCKSIZE, &blksz) || blksz == 0)) {
if (strncmp(rn->rn_name, "vn", 2) != 0)
fprintf(stderr, "device '%s' failed to report blocksize -- skipping\r\n",
rn->rn_name);
close(fd);
return;
}
struct sigaction sact;
sigemptyset(&sact.sa_mask);
sact.sa_flags = 0;
sact.sa_handler = signal_alarm;
sigaction(SIGALRM, &sact, NULL);
if (setjmp(buffer) != 0) {
printf("ZFS: Warning, timeout reading device '%s'\n", rn->rn_name);
close(fd);
return;
}
alarm(20);
#endif
if ((zpool_read_label(fd, &config, &num_labels)) != 0) {
#ifdef __APPLE__
alarm(0);
#endif
(void) close(fd);
(void) no_memory(rn->rn_hdl);
return;
}
#ifdef __APPLE__
alarm(0);
#endif
if (num_labels == 0) {
(void) close(fd);
nvlist_free(config);
return;
}
(void) close(fd);
rn->rn_config = config;
rn->rn_num_labels = num_labels;
}
/*
* Given a list of directories to search, find all pools stored on disk. This
* includes partial pools which are not available to import. If no args are
* given (argc is 0), then the default directory (/dev/dsk) is searched.
* poolname or guid (but not both) are provided by the caller when trying
* to import a specific pool.
*/
static nvlist_t *
zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
{
int i, num_labels, dirs = iarg->paths;
DIR *dirp = NULL;
struct dirent *dp;
char path[MAXPATHLEN];
char *end, **dir = iarg->path;
size_t pathleft;
struct stat statbuf;
nvlist_t *ret = NULL, *config;
int fd;
pool_list_t pools = { 0 };
pool_entry_t *pe, *penext;
vdev_entry_t *ve, *venext;
config_entry_t *ce, *cenext;
name_entry_t *ne, *nenext;
verify(iarg->poolname == NULL || iarg->guid == 0);
if (dirs == 0) {
#ifdef HAVE_LIBBLKID
/* Use libblkid to scan all device for their type */
if (zpool_find_import_blkid(hdl, &pools) == 0)
goto skip_scanning;
(void) zfs_error_fmt(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "blkid failure falling back "
"to manual probing"));
#endif /* HAVE_LIBBLKID */
dir = zpool_default_import_path;
dirs = DEFAULT_IMPORT_PATH_SIZE;
}
/*
* Go through and read the label configuration information from every
* possible device, organizing the information according to pool GUID
* and toplevel GUID.
*/
for (i = 0; i < dirs; i++) {
char *rdsk;
int dfd;
/* use realpath to normalize the path */
if (realpath(dir[i], path) == 0) {
/* it is safe to skip missing search paths */
if (errno == ENOENT)
continue;
zfs_error_aux(hdl, strerror(errno));
(void) zfs_error_fmt(hdl, EZFS_BADPATH,
dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
goto error;
}
end = &path[strlen(path)];
*end++ = '/';
*end = 0;
pathleft = &path[sizeof (path)] - end;
/*
* Using raw devices instead of block devices when we're
* reading the labels skips a bunch of slow operations during
* close(2) processing, so we replace /dev/dsk with /dev/rdsk.
*/
if (strcmp(path, "/dev/dsk/") == 0)
rdsk = "/dev/rdsk/";
else
rdsk = path;
if ((dfd = open(rdsk, O_RDONLY)) < 0 ||
(dirp = fdopendir(dfd)) == NULL) {
zfs_error_aux(hdl, strerror(errno));
(void) zfs_error_fmt(hdl, EZFS_BADPATH,
dgettext(TEXT_DOMAIN, "cannot open '%s'"),
rdsk);
goto error;
}
/*
* This is not MT-safe, but we have no MT consumers of libzfs
*/
while ((dp = readdir(dirp)) != NULL) {
const char *name = dp->d_name;
if (name[0] == '.' &&
(name[1] == 0 || (name[1] == '.' && name[2] == 0)))
continue;
/*
* Skip checking devices with well known prefixes:
* watchdog - A special close is required to avoid
* triggering it and resetting the system.
* fuse - Fuse control device.
* ppp - Generic PPP driver.
* tty* - Generic serial interface.
* vcs* - Virtual console memory.
* parport* - Parallel port interface.
* lp* - Printer interface.
* fd* - Floppy interface.
* hpet - High Precision Event Timer, crashes qemu
* when accessed from a virtual machine.
* core - Symlink to /proc/kcore, causes a crash
* when access from Xen dom0.
*/
if ((strncmp(name, "watchdog", 8) == 0) ||
(strncmp(name, "fuse", 4) == 0) ||
(strncmp(name, "ppp", 3) == 0) ||
(strncmp(name, "tty", 3) == 0) ||
(strncmp(name, "vcs", 3) == 0) ||
(strncmp(name, "parport", 7) == 0) ||
(strncmp(name, "lp", 2) == 0) ||
(strncmp(name, "fd", 2) == 0) ||
(strncmp(name, "hpet", 4) == 0) ||
#ifdef __APPLE__
(strncmp(name, "pty", 3) == 0) || // lots, skip for speed
(strncmp(name, "com", 3) == 0) || // /dev/com_digidesign_semiface
#endif
(strncmp(name, "core", 4) == 0))
continue;
/*
* Ignore failed stats. We only want regular
* files and block devices.
*/
if ((fstatat64(dfd, name, &statbuf, 0) != 0) ||
(!S_ISREG(statbuf.st_mode) &&
!S_ISBLK(statbuf.st_mode)))
continue;
#ifdef __APPLE__
/* It is desirable to skip optical media as well, as they are
* also called /dev/diskX
*/
if (is_optical_media((char *)name))
continue;
#endif
if ((fd = openat64(dfd, name, O_RDONLY)) < 0)
continue;
int32_t blksz = 0;
if (S_ISBLK(statbuf.st_mode) &&
(ioctl(fd, DKIOCGETBLOCKSIZE, &blksz) || blksz == 0)) {
if (strncmp(name, "vn", 2) != 0)
fprintf(stderr, "device '%s' failed to report blocksize -- skipping\r\n",
name);
close(fd);
continue;
}
#ifdef __APPLE__
struct sigaction sact;
sigemptyset(&sact.sa_mask);
sact.sa_flags = 0;
sact.sa_handler = signal_alarm;
sigaction(SIGALRM, &sact, NULL);
if (setjmp(buffer) != 0) {
printf("ZFS: Warning, timeout reading device '%s'\n", name);
close(fd);
continue;
}
alarm(20);
#endif
if ((zpool_read_label(fd, &config, NULL)) != 0) {
#ifdef __APPLE__
alarm(0);
#endif
(void) close(fd);
(void) no_memory(hdl);
goto error;
}
#ifdef __APPLE__
alarm(0);
#endif
(void) close(fd);
if (config != NULL) {
boolean_t matched = B_TRUE;
char *pname;
if ((iarg->poolname != NULL) &&
(nvlist_lookup_string(config,
ZPOOL_CONFIG_POOL_NAME, &pname) == 0)) {
if (strcmp(iarg->poolname, pname))
matched = B_FALSE;
} else if (iarg->guid != 0) {
uint64_t this_guid;
matched = nvlist_lookup_uint64(config,
ZPOOL_CONFIG_POOL_GUID,
&this_guid) == 0 &&
iarg->guid == this_guid;
}
if (!matched) {
nvlist_free(config);
config = NULL;
continue;
}
/* use the non-raw path for the config */
(void) strlcpy(end, name, pathleft);
if (add_config(hdl, &pools, path, i+1,
num_labels, config))
goto error;
}
}
(void) closedir(dirp);
dirp = NULL;
}
#ifdef HAVE_LIBBLKID
skip_scanning:
#endif
ret = get_configs(hdl, &pools, iarg->can_be_active);
error:
for (pe = pools.pools; pe != NULL; pe = penext) {
penext = pe->pe_next;
for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
venext = ve->ve_next;
for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
cenext = ce->ce_next;
if (ce->ce_config)
nvlist_free(ce->ce_config);
free(ce);
}
free(ve);
}
free(pe);
}
for (ne = pools.names; ne != NULL; ne = nenext) {
nenext = ne->ne_next;
if (ne->ne_name)
free(ne->ne_name);
free(ne);
}
if (dirp)
(void) closedir(dirp);
return (ret);
}
static void
zpool_open_func(void *arg)
{
rdsk_node_t *rn = arg;
struct stat64 statbuf;
nvlist_t *config;
int num_labels;
int fd;
if (rn->rn_nozpool)
return;
#ifdef __linux__
/*
* Skip devices with well known prefixes there can be side effects
* when opening devices which need to be avoided.
*
* core - Symlink to /proc/kcore
* fd* - Floppy interface.
* fuse - Fuse control device.
* hpet - High Precision Event Timer
* lp* - Printer interface.
* parport* - Parallel port interface.
* ppp - Generic PPP driver.
* random - Random device
* rtc - Real Time Clock
* tty* - Generic serial interface.
* urandom - Random device.
* usbmon* - USB IO monitor.
* vcs* - Virtual console memory.
* watchdog - Watchdog must be closed in a special way.
*/
if ((strncmp(rn->rn_name, "core", 4) == 0) ||
(strncmp(rn->rn_name, "fd", 2) == 0) ||
(strncmp(rn->rn_name, "fuse", 4) == 0) ||
(strncmp(rn->rn_name, "hpet", 4) == 0) ||
(strncmp(rn->rn_name, "lp", 2) == 0) ||
(strncmp(rn->rn_name, "parport", 7) == 0) ||
(strncmp(rn->rn_name, "ppp", 3) == 0) ||
(strncmp(rn->rn_name, "random", 6) == 0) ||
(strncmp(rn->rn_name, "rtc", 3) == 0) ||
(strncmp(rn->rn_name, "tty", 3) == 0) ||
(strncmp(rn->rn_name, "urandom", 7) == 0) ||
(strncmp(rn->rn_name, "usbmon", 6) == 0) ||
(strncmp(rn->rn_name, "vcs", 3) == 0) ||
(strncmp(rn->rn_name, "watchdog", 8) == 0))
return;
/*
* Ignore failed stats. We only want regular files and block devices.
*/
if (fstatat64(rn->rn_dfd, rn->rn_name, &statbuf, 0) != 0 ||
(!S_ISREG(statbuf.st_mode) && !S_ISBLK(statbuf.st_mode)))
return;
if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
/* symlink to a device that's no longer there */
if (errno == ENOENT)
nozpool_all_slices(rn->rn_avl, rn->rn_name);
return;
}
#else
if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
/* symlink to a device that's no longer there */
if (errno == ENOENT)
nozpool_all_slices(rn->rn_avl, rn->rn_name);
return;
}
/*
* Ignore failed stats. We only want regular
* files, character devs and block devs.
*/
if (fstat64(fd, &statbuf) != 0 ||
(!S_ISREG(statbuf.st_mode) &&
!S_ISCHR(statbuf.st_mode) &&
!S_ISBLK(statbuf.st_mode))) {
(void) close(fd);
return;
}
#endif
/* this file is too small to hold a zpool */
if (S_ISREG(statbuf.st_mode) &&
statbuf.st_size < SPA_MINDEVSIZE) {
(void) close(fd);
return;
} else if (!S_ISREG(statbuf.st_mode)) {
/*
* Try to read the disk label first so we don't have to
* open a bunch of minor nodes that can't have a zpool.
*/
check_slices(rn->rn_avl, fd, rn->rn_name);
}
if ((zpool_read_label(fd, &config, &num_labels)) != 0) {
(void) close(fd);
(void) no_memory(rn->rn_hdl);
return;
}
if (num_labels == 0) {
(void) close(fd);
nvlist_free(config);
return;
}
(void) close(fd);
rn->rn_config = config;
rn->rn_num_labels = num_labels;
}
/*
* Given a list of directories to search, find all pools stored on disk. This
* includes partial pools which are not available to import. If no args are
* given (argc is 0), then the default directory (/dev/dsk) is searched.
* poolname or guid (but not both) are provided by the caller when trying
* to import a specific pool.
*/
static nvlist_t *
zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg)
{
int i, dirs = iarg->paths;
DIR *dirp = NULL;
struct dirent64 *dp;
char path[MAXPATHLEN];
char *end, **dir = iarg->path;
size_t pathleft;
struct stat64 statbuf;
nvlist_t *ret = NULL, *config;
static char *default_dir = DISK_ROOT;
int fd;
pool_list_t pools = { 0 };
pool_entry_t *pe, *penext;
vdev_entry_t *ve, *venext;
config_entry_t *ce, *cenext;
name_entry_t *ne, *nenext;
verify(iarg->poolname == NULL || iarg->guid == 0);
if (dirs == 0) {
#ifdef HAVE_LIBBLKID
/* Use libblkid to scan all device for their type */
if (zpool_find_import_blkid(hdl, &pools) == 0)
goto skip_scanning;
(void) zfs_error_fmt(hdl, EZFS_BADCACHE,
dgettext(TEXT_DOMAIN, "blkid failure falling back "
"to manual probing"));
#endif /* HAVE_LIBBLKID */
dirs = 1;
dir = &default_dir;
}
/*
* Go through and read the label configuration information from every
* possible device, organizing the information according to pool GUID
* and toplevel GUID.
*/
for (i = 0; i < dirs; i++) {
char *rdsk;
int dfd;
/* use realpath to normalize the path */
if (realpath(dir[i], path) == 0) {
(void) zfs_error_fmt(hdl, EZFS_BADPATH,
dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
goto error;
}
end = &path[strlen(path)];
*end++ = '/';
*end = 0;
pathleft = &path[sizeof (path)] - end;
/*
* Using raw devices instead of block devices when we're
* reading the labels skips a bunch of slow operations during
* close(2) processing, so we replace /dev/dsk with /dev/rdsk.
*/
if (strcmp(path, "/dev/dsk/") == 0)
rdsk = "/dev/rdsk/";
else
rdsk = path;
if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
(dirp = fdopendir(dfd)) == NULL) {
zfs_error_aux(hdl, strerror(errno));
(void) zfs_error_fmt(hdl, EZFS_BADPATH,
dgettext(TEXT_DOMAIN, "cannot open '%s'"),
rdsk);
goto error;
}
/*
* This is not MT-safe, but we have no MT consumers of libzfs
*/
while ((dp = readdir64(dirp)) != NULL) {
const char *name = dp->d_name;
if (name[0] == '.' &&
(name[1] == 0 || (name[1] == '.' && name[2] == 0)))
continue;
/*
* Skip checking devices with well known prefixes:
* watchdog - A special close is required to avoid
* triggering it and resetting the system.
* fuse - Fuse control device.
* ppp - Generic PPP driver.
* tty* - Generic serial interface.
* vcs* - Virtual console memory.
* parport* - Parallel port interface.
* lp* - Printer interface.
* fd* - Floppy interface.
* hpet - High Precision Event Timer, crashes qemu
* when accessed from a virtual machine.
* core - Symlink to /proc/kcore, causes a crash
* when access from Xen dom0.
*/
if ((strncmp(name, "watchdog", 8) == 0) ||
(strncmp(name, "fuse", 4) == 0) ||
(strncmp(name, "ppp", 3) == 0) ||
(strncmp(name, "tty", 3) == 0) ||
(strncmp(name, "vcs", 3) == 0) ||
(strncmp(name, "parport", 7) == 0) ||
(strncmp(name, "lp", 2) == 0) ||
(strncmp(name, "fd", 2) == 0) ||
(strncmp(name, "hpet", 4) == 0) ||
(strncmp(name, "core", 4) == 0))
continue;
/*
* Ignore failed stats. We only want regular
* files and block devices.
*/
if ((fstatat64(dfd, name, &statbuf, 0) != 0) ||
(!S_ISREG(statbuf.st_mode) &&
!S_ISBLK(statbuf.st_mode)))
continue;
if ((fd = openat64(dfd, name, O_RDONLY)) < 0)
continue;
if ((zpool_read_label(fd, &config)) != 0) {
(void) close(fd);
(void) no_memory(hdl);
goto error;
}
(void) close(fd);
if (config != NULL) {
boolean_t matched = B_TRUE;
if (iarg->poolname != NULL) {
char *pname;
matched = nvlist_lookup_string(config,
ZPOOL_CONFIG_POOL_NAME,
&pname) == 0 &&
strcmp(iarg->poolname, pname) == 0;
} else if (iarg->guid != 0) {
uint64_t this_guid;
matched = nvlist_lookup_uint64(config,
ZPOOL_CONFIG_POOL_GUID,
&this_guid) == 0 &&
iarg->guid == this_guid;
}
if (!matched) {
nvlist_free(config);
config = NULL;
continue;
}
/* use the non-raw path for the config */
(void) strlcpy(end, name, pathleft);
if (add_config(hdl, &pools, path, config) != 0)
goto error;
}
}
(void) closedir(dirp);
dirp = NULL;
}
#ifdef HAVE_LIBBLKID
skip_scanning:
#endif
ret = get_configs(hdl, &pools, iarg->can_be_active);
error:
for (pe = pools.pools; pe != NULL; pe = penext) {
penext = pe->pe_next;
for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
venext = ve->ve_next;
for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
cenext = ce->ce_next;
if (ce->ce_config)
nvlist_free(ce->ce_config);
free(ce);
}
free(ve);
}
free(pe);
}
for (ne = pools.names; ne != NULL; ne = nenext) {
nenext = ne->ne_next;
if (ne->ne_name)
free(ne->ne_name);
free(ne);
}
if (dirp)
(void) closedir(dirp);
return (ret);
}
