/*
* Convert from a devid string to a path.
*/
static char *
devid_to_path(char *devid_str)
{
ddi_devid_t devid;
char *minor;
char *path;
devid_nmlist_t *list = NULL;
int ret;
if (devid_str_decode(devid_str, &devid, &minor) != 0)
return (NULL);
ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
devid_str_free(minor);
devid_free(devid);
if (ret != 0)
return (NULL);
if ((path = strdup(list[0].devname)) == NULL)
return (NULL);
devid_free_nmlist(list);
return (path);
}
/*
* Given a /devices path, lookup the corresponding devid for each minor node,
* and find any vdevs with matching devids. Doing this straight up would be
* rather inefficient, O(minor nodes * vdevs in system), so we take advantage of
* the fact that each devid ends with "/<minornode>". Once we find any valid
* minor node, we chop off the portion after the last slash, and then search for
* matching vdevs, which is O(vdevs in system).
*/
static boolean_t
devid_iter(const char *devpath, zfs_process_func_t func, boolean_t wholedisk)
{
size_t len = strlen(devpath) + sizeof ("/devices") +
sizeof (PHYS_PATH) - 1;
char *fullpath;
int fd;
ddi_devid_t devid;
char *devidstr, *fulldevid;
dev_data_t data = { 0 };
/*
* Try to open a known minor node.
*/
fullpath = alloca(len);
(void) snprintf(fullpath, len, "/devices%s%s", devpath, PHYS_PATH);
if ((fd = open(fullpath, O_RDONLY)) < 0)
return (B_FALSE);
/*
* Determine the devid as a string, with no trailing slash for the minor
* node.
*/
if (devid_get(fd, &devid) != 0) {
(void) close(fd);
return (B_FALSE);
}
(void) close(fd);
if ((devidstr = devid_str_encode(devid, NULL)) == NULL) {
devid_free(devid);
return (B_FALSE);
}
len = strlen(devidstr) + 2;
fulldevid = alloca(len);
(void) snprintf(fulldevid, len, "%s/", devidstr);
data.dd_compare = fulldevid;
data.dd_func = func;
data.dd_prop = ZPOOL_CONFIG_DEVID;
data.dd_found = B_FALSE;
data.dd_isdisk = wholedisk;
(void) zpool_iter(g_zfshdl, zfs_iter_pool, &data);
devid_str_free(devidstr);
devid_free(devid);
return (data.dd_found);
}
static char *
get_devid(const char *path)
{
int fd;
ddi_devid_t devid;
char *minor, *ret;
if ((fd = open(path, O_RDONLY)) < 0)
return (NULL);
minor = NULL;
ret = NULL;
if (devid_get(fd, &devid) == 0) {
if (devid_get_minor_name(fd, &minor) == 0)
ret = devid_str_encode(devid, minor);
if (minor != NULL)
devid_str_free(minor);
devid_free(devid);
}
(void) close(fd);
return (ret);
}
/*
* Go through and fix up any path and/or devid information for the given vdev
* configuration.
*/
static int
fix_paths(nvlist_t *nv, name_entry_t *names)
{
nvlist_t **child;
uint_t c, children;
uint64_t guid;
name_entry_t *ne, *best;
char *path, *devid;
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++)
if (fix_paths(child[c], names) != 0)
return (-1);
return (0);
}
/*
* This is a leaf (file or disk) vdev. In either case, go through
* the name list and see if we find a matching guid. If so, replace
* the path and see if we can calculate a new devid.
*
* There may be multiple names associated with a particular guid, in
* which case we have overlapping partitions or multiple paths to the
* same disk. In this case we prefer to use the path name which
* matches the ZPOOL_CONFIG_PATH. If no matching entry is found we
* use the lowest order device which corresponds to the first match
* while traversing the ZPOOL_IMPORT_PATH search path.
*/
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
path = NULL;
best = NULL;
for (ne = names; ne != NULL; ne = ne->ne_next) {
if (ne->ne_guid == guid) {
if (path == NULL) {
best = ne;
break;
}
if ((strlen(path) == strlen(ne->ne_name)) &&
strncmp(path, ne->ne_name, strlen(path)) == 0) {
best = ne;
break;
}
if (best == NULL) {
best = ne;
continue;
}
/* Prefer paths with move vdev labels. */
if (ne->ne_num_labels > best->ne_num_labels) {
best = ne;
continue;
}
/* Prefer paths earlier in the search order. */
if (ne->ne_num_labels == best->ne_num_labels &&
ne->ne_order < best->ne_order) {
best = ne;
continue;
}
}
}
if (best == NULL)
return (0);
if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
return (-1);
if ((devid = get_devid(best->ne_name)) == NULL) {
(void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
} else {
if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) {
devid_str_free(devid);
return (-1);
}
devid_str_free(devid);
}
return (0);
}
/*
* Given a vdev, return the name to display in iostat. If the vdev has a path,
* we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
* We also check if this is a whole disk, in which case we strip off the
* trailing 's0' slice name.
*
* This routine is also responsible for identifying when disks have been
* reconfigured in a new location. The kernel will have opened the device by
* devid, but the path will still refer to the old location. To catch this, we
* first do a path -> devid translation (which is fast for the common case). If
* the devid matches, we're done. If not, we do a reverse devid -> path
* translation and issue the appropriate ioctl() to update the path of the vdev.
* If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
* of these checks.
*/
char *
zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv)
{
char *path, *devid;
uint64_t value;
char buf[64];
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
&value) == 0) {
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
&value) == 0);
(void) snprintf(buf, sizeof (buf), "%llu", value);
path = buf;
} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
if (zhp != NULL &&
nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
/*
* Determine if the current path is correct.
*/
char *newdevid = path_to_devid(path);
if (newdevid == NULL ||
strcmp(devid, newdevid) != 0) {
char *newpath;
if ((newpath = devid_to_path(devid)) != NULL) {
/*
* Update the path appropriately.
*/
set_path(zhp, nv, newpath);
if (nvlist_add_string(nv,
ZPOOL_CONFIG_PATH, newpath) == 0)
verify(nvlist_lookup_string(nv,
ZPOOL_CONFIG_PATH,
&path) == 0);
free(newpath);
}
}
if (newdevid)
devid_str_free(newdevid);
}
if (strncmp(path, "/dev/dsk/", 9) == 0)
path += 9;
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
&value) == 0 && value) {
char *tmp = zfs_strdup(hdl, path);
if (tmp == NULL)
return (NULL);
tmp[strlen(path) - 2] = '\0';
return (tmp);
}
} else {
verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
/*
* If it's a raidz device, we need to stick in the parity level.
*/
if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
&value) == 0);
(void) snprintf(buf, sizeof (buf), "%s%llu", path,
value);
path = buf;
}
}
return (zfs_strdup(hdl, path));
}
/*
* Go through and fix up any path and/or devid information for the given vdev
* configuration.
*/
static int
fix_paths(nvlist_t *nv, name_entry_t *names)
{
nvlist_t **child;
uint_t c, children;
uint64_t guid;
name_entry_t *ne, *best;
char *path, *devid;
int matched;
if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++)
if (fix_paths(child[c], names) != 0)
return (-1);
return (0);
}
/*
* This is a leaf (file or disk) vdev. In either case, go through
* the name list and see if we find a matching guid. If so, replace
* the path and see if we can calculate a new devid.
*
* There may be multiple names associated with a particular guid, in
* which case we have overlapping slices or multiple paths to the same
* disk. If this is the case, then we want to pick the path that is
* the most similar to the original, where "most similar" is the number
* of matching characters starting from the end of the path. This will
* preserve slice numbers even if the disks have been reorganized, and
* will also catch preferred disk names if multiple paths exist.
*/
verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
path = NULL;
matched = 0;
best = NULL;
for (ne = names; ne != NULL; ne = ne->ne_next) {
if (ne->ne_guid == guid) {
const char *src, *dst;
int count;
if (path == NULL) {
best = ne;
break;
}
src = ne->ne_name + strlen(ne->ne_name) - 1;
dst = path + strlen(path) - 1;
for (count = 0; src >= ne->ne_name && dst >= path;
src--, dst--, count++)
if (*src != *dst)
break;
/*
* At this point, 'count' is the number of characters
* matched from the end.
*/
if (count > matched || best == NULL) {
best = ne;
matched = count;
}
}
}
if (best == NULL)
return (0);
if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
return (-1);
if ((devid = get_devid(best->ne_name)) == NULL) {
(void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
} else {
if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) {
devid_str_free(devid);
return (-1);
}
devid_str_free(devid);
}
return (0);
}
static int
mpxio_nvl_boilerplate(di_node_t curnode)
{
int rv;
char *strdevid;
ddi_devid_t curdevid;
nvlist_t *newnvl;
for (; curnode != DI_NODE_NIL; curnode = di_drv_next_node(curnode)) {
errno = 0;
curdevid = NULL;
get_devid(curnode, &curdevid);
if (curdevid == NULL)
/*
* There's no devid registered for this device
* so it's not cool enough to play with us
*/
continue;
strdevid = devid_str_encode(curdevid, NULL);
/* does this exist in the on-disk cache? */
rv = nvlist_lookup_nvlist(mapnvl, strdevid, &newnvl);
if (rv == ENOENT) {
logmsg(MSG_INFO, "nvlist for %s not found\n", strdevid);
/* no, so alloc a new nvl to store it */
if (nvlist_alloc(&newnvl, NV_UNIQUE_NAME, 0) != 0) {
logmsg(MSG_ERROR,
gettext("Unable to allocate space for "
"a devid property list: %s\n"),
strerror(errno));
return (-1);
}
} else {
if ((rv != ENOTSUP) && (rv != EINVAL))
logmsg(MSG_INFO,
"%s exists in ondisknvl, verifying\n",
strdevid);
}
if (popcheck_devnvl(curnode, newnvl, strdevid) != 0) {
logmsg(MSG_ERROR,
gettext("Unable to populate devid nvpair "
"for device with devid %s\n"),
strdevid);
devid_str_free(strdevid);
nvlist_free(newnvl);
return (-1);
}
/* Now add newnvl into our cache. */
errno = 0;
rv = nvlist_add_nvlist(mapnvl, strdevid, newnvl);
if (rv) {
logmsg(MSG_ERROR,
gettext("Unable to add device (devid %s) "
"to in-kernel nvl: %s (%d)\n"),
strdevid, strerror(rv), rv);
devid_str_free(strdevid);
nvlist_free(newnvl);
return (-1);
}
logmsg(MSG_INFO,
gettext("added device (devid %s) to mapnvl\n\n"),
strdevid);
devid_str_free(strdevid);
}
return (0);
}
static int
get_attrs(descriptor_t *dp, int fd, nvlist_t *attrs)
{
struct dk_minfo minfo;
int status;
int data_format = FMT_UNKNOWN;
int snum = -1;
int error;
struct extvtoc vtoc;
struct dk_gpt *efip;
struct dk_cinfo dkinfo;
int cooked_fd;
struct stat buf;
if (fd < 0) {
return (ENODEV);
}
/* First make sure media is inserted and spun up. */
if (!media_read_info(fd, &minfo)) {
return (ENODEV);
}
if ((status = read_extvtoc(fd, &vtoc)) >= 0) {
data_format = FMT_VTOC;
} else if (status == VT_ENOTSUP && efi_alloc_and_read(fd, &efip) >= 0) {
data_format = FMT_EFI;
if (nvlist_add_boolean(attrs, DM_EFI) != 0) {
efi_free(efip);
return (ENOMEM);
}
}
if (data_format == FMT_UNKNOWN) {
return (ENODEV);
}
if (ioctl(fd, DKIOCINFO, &dkinfo) >= 0) {
snum = dkinfo.dki_partition;
}
/* check the slice */
if (data_format == FMT_VTOC) {
if (snum < 0 || snum >= vtoc.v_nparts ||
vtoc.v_part[snum].p_size == 0) {
return (ENODEV);
}
} else { /* data_format == FMT_EFI */
if (snum < 0 || snum >= efip->efi_nparts ||
efip->efi_parts[snum].p_size == 0) {
efi_free(efip);
return (ENODEV);
}
}
/* the slice exists */
if (nvlist_add_uint32(attrs, DM_INDEX, snum) != 0) {
if (data_format == FMT_EFI) {
efi_free(efip);
}
return (ENOMEM);
}
if (data_format == FMT_VTOC) {
if (nvlist_add_uint64(attrs, DM_START, vtoc.v_part[snum].p_start)
!= 0) {
return (ENOMEM);
}
if (nvlist_add_uint64(attrs, DM_SIZE, vtoc.v_part[snum].p_size)
!= 0) {
return (ENOMEM);
}
if (nvlist_add_uint32(attrs, DM_TAG, vtoc.v_part[snum].p_tag)
!= 0) {
return (ENOMEM);
}
if (nvlist_add_uint32(attrs, DM_FLAG, vtoc.v_part[snum].p_flag)
!= 0) {
return (ENOMEM);
}
} else { /* data_format == FMT_EFI */
if (nvlist_add_uint64(attrs, DM_START,
efip->efi_parts[snum].p_start) != 0) {
efi_free(efip);
return (ENOMEM);
}
if (nvlist_add_uint64(attrs, DM_SIZE, efip->efi_parts[snum].p_size)
!= 0) {
efi_free(efip);
return (ENOMEM);
}
if (efip->efi_parts[snum].p_name[0] != 0) {
char label[EFI_PART_NAME_LEN + 1];
(void) snprintf(label, sizeof (label), "%.*s",
EFI_PART_NAME_LEN, efip->efi_parts[snum].p_name);
if (nvlist_add_string(attrs, DM_EFI_NAME, label) != 0) {
efi_free(efip);
return (ENOMEM);
}
}
}
if (data_format == FMT_EFI) {
efi_free(efip);
}
if (inuse_mnt(dp->name, attrs, &error)) {
if (error != 0)
return (error);
}
if (fstat(fd, &buf) != -1) {
if (nvlist_add_uint64(attrs, DM_DEVT, buf.st_rdev) != 0) {
return (ENOMEM);
}
}
/*
* We need to open the cooked slice (not the raw one) to get the
* correct devid.
*/
cooked_fd = open(dp->name, O_RDONLY|O_NDELAY);
if (cooked_fd >= 0) {
int no_mem = 0;
ddi_devid_t devid;
if (devid_get(cooked_fd, &devid) == 0) {
char *minor;
if (devid_get_minor_name(cooked_fd, &minor) == 0) {
char *devidstr;
if ((devidstr = devid_str_encode(devid, minor)) != 0) {
if (nvlist_add_string(attrs, DM_DEVICEID, devidstr)
!= 0) {
no_mem = 1;
}
devid_str_free(devidstr);
}
devid_str_free(minor);
}
devid_free(devid);
}
(void) close(cooked_fd);
if (no_mem) {
return (ENOMEM);
}
}
return (0);
}
/*
* Create a leaf vdev. Determine if this is a file or a device. If it's a
* device, fill in the device id to make a complete nvlist. Valid forms for a
* leaf vdev are:
*
* /dev/dsk/xxx Complete disk path
* /xxx Full path to file
* xxx Shorthand for /dev/dsk/xxx
*/
nvlist_t *
make_leaf_vdev(const char *arg)
{
char path[MAXPATHLEN];
struct stat statbuf;
nvlist_t *vdev = NULL;
char *type = NULL;
boolean_t wholedisk = B_FALSE;
/*
* Determine what type of vdev this is, and put the full path into
* 'path'. We detect whether this is a device of file afterwards by
* checking the st_mode of the file.
*/
if (arg[0] == '/') {
/*
* Complete device or file path. Exact type is determined by
* examining the file descriptor afterwards.
*/
if (is_whole_disk(arg, &statbuf)) {
wholedisk = B_TRUE;
} else if (stat(arg, &statbuf) != 0) {
(void) fprintf(stderr,
gettext("cannot open '%s': %s\n"),
arg, strerror(errno));
return (NULL);
}
(void) strlcpy(path, arg, sizeof (path));
} else {
/*
* This may be a short path for a device, or it could be total
* gibberish. Check to see if it's a known device in
* /dev/dsk/. As part of this check, see if we've been given a
* an entire disk (minus the slice number).
*/
(void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT,
arg);
if (is_whole_disk(path, &statbuf)) {
wholedisk = B_TRUE;
} else if (stat(path, &statbuf) != 0) {
/*
* If we got ENOENT, then the user gave us
* gibberish, so try to direct them with a
* reasonable error message. Otherwise,
* regurgitate strerror() since it's the best we
* can do.
*/
if (errno == ENOENT) {
(void) fprintf(stderr,
gettext("cannot open '%s': no such "
"device in %s\n"), arg, DISK_ROOT);
(void) fprintf(stderr,
gettext("must be a full path or "
"shorthand device name\n"));
return (NULL);
} else {
(void) fprintf(stderr,
gettext("cannot open '%s': %s\n"),
path, strerror(errno));
return (NULL);
}
}
}
/*
* Determine whether this is a device or a file.
*/
if (S_ISBLK(statbuf.st_mode)) {
type = VDEV_TYPE_DISK;
} else if (S_ISREG(statbuf.st_mode)) {
type = VDEV_TYPE_FILE;
} else {
(void) fprintf(stderr, gettext("cannot use '%s': must be a "
"block device or regular file\n"), path);
return (NULL);
}
/*
* Finally, we have the complete device or file, and we know that it is
* acceptable to use. Construct the nvlist to describe this vdev. All
* vdevs have a 'path' element, and devices also have a 'devid' element.
*/
verify(nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) == 0);
verify(nvlist_add_string(vdev, ZPOOL_CONFIG_PATH, path) == 0);
verify(nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, type) == 0);
if (strcmp(type, VDEV_TYPE_DISK) == 0)
verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK,
(uint64_t)wholedisk) == 0);
/*
* For a whole disk, defer getting its devid until after labeling it.
*/
if (S_ISBLK(statbuf.st_mode) && !wholedisk) {
/*
* Get the devid for the device.
*/
int fd;
ddi_devid_t devid;
char *minor = NULL, *devid_str = NULL;
if ((fd = open(path, O_RDONLY)) < 0) {
(void) fprintf(stderr, gettext("cannot open '%s': "
"%s\n"), path, strerror(errno));
nvlist_free(vdev);
return (NULL);
}
if (devid_get(fd, &devid) == 0) {
if (devid_get_minor_name(fd, &minor) == 0 &&
(devid_str = devid_str_encode(devid, minor)) !=
NULL) {
verify(nvlist_add_string(vdev,
ZPOOL_CONFIG_DEVID, devid_str) == 0);
}
if (devid_str != NULL)
devid_str_free(devid_str);
if (minor != NULL)
devid_str_free(minor);
devid_free(devid);
}
(void) close(fd);
}
return (vdev);
}