* Incremental.c - support --incremental. Part of:
* mdadm - manage Linux "md" devices aka RAID arrays.
*
- * Copyright (C) 2006 Neil Brown <neilb@suse.de>
+ * Copyright (C) 2006-2009 Neil Brown <neilb@suse.de>
*
*
* This program is free software; you can redistribute it and/or modify
*/
#include "mdadm.h"
+#include <dirent.h>
+#include <ctype.h>
static int count_active(struct supertype *st, int mdfd, char **availp,
struct mdinfo *info);
static void find_reject(int mdfd, struct supertype *st, struct mdinfo *sra,
int number, __u64 events, int verbose,
char *array_name);
+static int try_spare(char *devname, int *dfdp, struct dev_policy *pol,
+ struct supertype *st, int verbose);
int Incremental(char *devname, int verbose, int runstop,
- struct supertype *st, char *homehost, int autof)
+ struct supertype *st, char *homehost, int require_homehost,
+ int autof)
{
/* Add this device to an array, creating the array if necessary
* and starting the array if sensible or - if runstop>0 - if possible.
* 2/ Find metadata, reject if none appropriate (check
* version/name from args)
* 3/ Check if there is a match in mdadm.conf
- * 3a/ if not, check for homehost match. If no match, reject.
+ * 3a/ if not, check for homehost match. If no match, assemble as
+ * a 'foreign' array.
* 4/ Determine device number.
* - If in mdadm.conf with std name, use that
* - UUID in /var/run/mdadm.map use that
* start the array (auto-readonly).
*/
struct stat stb;
- struct mdinfo info;
+ struct mdinfo info, dinfo;
struct mddev_ident_s *array_list, *match;
char chosen_name[1024];
- int rv;
- int devnum;
+ int rv = 1;
struct map_ent *mp, *map = NULL;
- int dfd, mdfd;
+ int dfd = -1, mdfd = -1;
char *avail;
int active_disks;
- int uuid_for_name = 0;
+ int trustworthy = FOREIGN;
char *name_to_use;
- char nbuf[64];
+ mdu_array_info_t ainf;
+ struct dev_policy *policy = NULL;
struct createinfo *ci = conf_get_create_info();
- if (autof == 0)
- autof = ci->autof;
/* 1/ Check if device is permitted by mdadm.conf */
fprintf(stderr, Name
": %s not permitted by mdadm.conf.\n",
devname);
- return 1;
+ goto out;
}
/* 2/ Find metadata, reject if none appropriate (check
if (verbose >= 0)
fprintf(stderr, Name ": cannot open %s: %s.\n",
devname, strerror(errno));
- return 1;
+ goto out;
}
if (fstat(dfd, &stb) < 0) {
if (verbose >= 0)
fprintf(stderr, Name ": fstat failed for %s: %s.\n",
devname, strerror(errno));
- close(dfd);
- return 1;
+ goto out;
}
if ((stb.st_mode & S_IFMT) != S_IFBLK) {
if (verbose >= 0)
fprintf(stderr, Name ": %s is not a block device.\n",
devname);
- close(dfd);
- return 1;
+ goto out;
}
+ dinfo.disk.major = major(stb.st_rdev);
+ dinfo.disk.minor = minor(stb.st_rdev);
+
+ policy = disk_policy(&dinfo);
+
if (st == NULL && (st = guess_super(dfd)) == NULL) {
if (verbose >= 0)
fprintf(stderr, Name
": no recognisable superblock on %s.\n",
devname);
- close(dfd);
- return 1;
+ rv = try_spare(devname, &dfd, policy, st, verbose);
+ goto out;
}
- if (st->ss->load_super(st, dfd, NULL)) {
+ if (st->ss->compare_super == NULL ||
+ st->ss->load_super(st, dfd, NULL)) {
if (verbose >= 0)
fprintf(stderr, Name ": no RAID superblock on %s.\n",
devname);
- close(dfd);
- return 1;
- }
- close (dfd);
-
- if (st->ss->container_content && st->loaded_container) {
- /* This is a pre-built container array, so we do something
- * rather different.
- */
- return Incremental_container(st, devname, verbose, runstop,
- autof);
+ rv = try_spare(devname, &dfd, policy, st, verbose);
+ free(st);
+ goto out;
}
+ close (dfd); dfd = -1;
memset(&info, 0, sizeof(info));
- st->ss->getinfo_super(st, &info);
+ st->ss->getinfo_super(st, &info, NULL);
/* 3/ Check if there is a match in mdadm.conf */
array_list = conf_get_ident(NULL);
if (array_list->uuid_set &&
same_uuid(array_list->uuid, info.uuid, st->ss->swapuuid)
== 0) {
- if (verbose >= 2)
+ if (verbose >= 2 && array_list->devname)
fprintf(stderr, Name
": UUID differs from %s.\n",
array_list->devname);
}
if (array_list->name[0] &&
strcasecmp(array_list->name, info.name) != 0) {
- if (verbose >= 2)
+ if (verbose >= 2 && array_list->devname)
fprintf(stderr, Name
": Name differs from %s.\n",
array_list->devname);
}
if (array_list->devices &&
!match_oneof(array_list->devices, devname)) {
- if (verbose >= 2)
+ if (verbose >= 2 && array_list->devname)
fprintf(stderr, Name
": Not a listed device for %s.\n",
array_list->devname);
}
if (array_list->super_minor != UnSet &&
array_list->super_minor != info.array.md_minor) {
- if (verbose >= 2)
+ if (verbose >= 2 && array_list->devname)
fprintf(stderr, Name
": Different super-minor to %s.\n",
array_list->devname);
!array_list->name[0] &&
!array_list->devices &&
array_list->super_minor == UnSet) {
- if (verbose >= 2)
+ if (verbose >= 2 && array_list->devname)
fprintf(stderr, Name
": %s doesn't have any identifying information.\n",
array_list->devname);
/* FIXME, should I check raid_disks and level too?? */
if (match) {
- if (verbose >= 0)
- fprintf(stderr, Name
+ if (verbose >= 0) {
+ if (match->devname && array_list->devname)
+ fprintf(stderr, Name
": we match both %s and %s - cannot decide which to use.\n",
- match->devname, array_list->devname);
- return 2;
+ match->devname, array_list->devname);
+ else
+ fprintf(stderr, Name
+ ": multiple lines in mdadm.conf match\n");
+ }
+ rv = 2;
+ goto out;
}
match = array_list;
}
- /* 3a/ if not, check for homehost match. If no match, reject. */
- if (!match) {
- if (homehost == NULL ||
- st->ss->match_home(st, homehost) != 1)
- uuid_for_name = 1;
- }
- /* 4/ Determine device number. */
- /* - If in mdadm.conf with std name, use that */
- /* - UUID in /var/run/mdadm.map use that */
- /* - If name is suggestive, use that. unless in use with */
- /* different uuid. */
- /* - Choose a free, high number. */
- /* - Use a partitioned device unless strong suggestion not to. */
- /* e.g. auto=md */
- mp = map_by_uuid(&map, info.uuid);
-
- if (uuid_for_name && ! mp) {
- name_to_use = fname_from_uuid(st, &info, nbuf, '-');
+ if (match && match->devname
+ && strcasecmp(match->devname, "<ignore>") == 0) {
if (verbose >= 0)
+ fprintf(stderr, Name ": array containing %s is explicitly"
+ " ignored by mdadm.conf\n",
+ devname);
+ goto out;
+ }
+
+ /* 3a/ if not, check for homehost match. If no match, continue
+ * but don't trust the 'name' in the array. Thus a 'random' minor
+ * number will be assigned, and the device name will be based
+ * on that. */
+ if (match)
+ trustworthy = LOCAL;
+ else if (st->ss->match_home(st, homehost) == 1)
+ trustworthy = LOCAL;
+ else if (st->ss->match_home(st, "any") == 1)
+ trustworthy = LOCAL_ANY;
+ else
+ trustworthy = FOREIGN;
+
+
+ if (!match && !conf_test_metadata(st->ss->name, policy,
+ (trustworthy == LOCAL))) {
+ if (verbose >= 1)
fprintf(stderr, Name
- ": not found in mdadm.conf and not identified by homehost"
- " - using uuid based name\n");
- } else
- name_to_use = info.name;
-
- if (match && is_standard(match->devname, &devnum))
- /* We have devnum now */;
- else if (mp != NULL)
- devnum = mp->devnum;
- else {
- /* Have to guess a bit. */
- int use_partitions = 1;
- char *np, *ep;
- char *nm, nbuf[1024];
- struct stat stb2;
-
- if ((autof&7) == 3 || (autof&7) == 5)
- use_partitions = 0;
- if (st->ss->external)
- use_partitions = 0;
- np = strchr(name_to_use, ':');
- if (np)
- np++;
- else
- np = name_to_use;
- devnum = strtoul(np, &ep, 10);
- if (ep > np && *ep == 0) {
- /* This is a number. Let check that it is unused. */
- if (mddev_busy(use_partitions ? (-1-devnum) : devnum))
- devnum = -1;
- } else
- devnum = -1;
+ ": %s has metadata type %s for which "
+ "auto-assembly is disabled\n",
+ devname, st->ss->name);
+ goto out;
+ }
+ if (trustworthy == LOCAL_ANY)
+ trustworthy = LOCAL;
+
+ /* There are three possible sources for 'autof': command line,
+ * ARRAY line in mdadm.conf, or CREATE line in mdadm.conf.
+ * ARRAY takes precedence, then command line, then
+ * CREATE.
+ */
+ if (match && match->autof)
+ autof = match->autof;
+ if (autof == 0)
+ autof = ci->autof;
- if (match)
- nm = match->devname;
+ if (st->ss->container_content && st->loaded_container) {
+ if ((runstop > 0 && info.container_enough >= 0) ||
+ info.container_enough > 0)
+ /* pass */;
else {
- sprintf(nbuf, "/dev/md/%s", np);
- nm = nbuf;
- }
- if (stat(nm, &stb2) == 0 &&
- S_ISBLK(stb2.st_mode) &&
- major(stb2.st_rdev) == (use_partitions ?
- get_mdp_major() : MD_MAJOR)) {
- if (use_partitions)
- devnum = minor(stb2.st_rdev) >> MdpMinorShift;
- else
- devnum = minor(stb2.st_rdev);
- if (mddev_busy(use_partitions ? (-1-devnum) : devnum))
- devnum = -1;
+ if (verbose)
+ fprintf(stderr, Name ": not enough devices to start the container\n");
+ rv = 0;
+ goto out;
}
- if (devnum < 0) {
- /* Haven't found anything yet, choose something free */
- devnum = find_free_devnum(use_partitions);
-
- if (devnum == NoMdDev) {
- fprintf(stderr, Name
- ": No spare md devices!!\n");
- return 2;
- }
- } else
- devnum = use_partitions ? (-1-devnum) : devnum;
+ /* This is a pre-built container array, so we do something
+ * rather different.
+ */
+ rv = Incremental_container(st, devname, verbose, runstop,
+ autof, trustworthy);
+ goto out;
}
- mdfd = open_mddev_devnum(match ? match->devname : mp ? mp->path : NULL,
- devnum,
- name_to_use,
- chosen_name, autof >> 3);
- if (mdfd < 0) {
- fprintf(stderr, Name ": failed to open %s: %s.\n",
- chosen_name, strerror(errno));
- return 2;
+ name_to_use = info.name;
+ if (name_to_use[0] == 0 &&
+ info.array.level == LEVEL_CONTAINER &&
+ trustworthy == LOCAL) {
+ name_to_use = info.text_version;
+ trustworthy = METADATA;
}
- sysfs_init(&info, mdfd, 0);
+ if (name_to_use[0] && trustworthy != LOCAL &&
+ ! require_homehost &&
+ conf_name_is_free(name_to_use))
+ trustworthy = LOCAL;
- /* 5/ Find out if array already exists */
- if (! mddev_busy(devnum)) {
- /* 5a/ if it does not */
- /* - choose a name, from mdadm.conf or 'name' field in array. */
- /* - create the array */
- /* - add the device */
+ /* strip "hostname:" prefix from name if we have decided
+ * to treat it as LOCAL
+ */
+ if (trustworthy == LOCAL && strchr(name_to_use, ':') != NULL)
+ name_to_use = strchr(name_to_use, ':')+1;
+
+ /* 4/ Check if array exists.
+ */
+ if (map_lock(&map))
+ fprintf(stderr, Name ": failed to get exclusive lock on "
+ "mapfile\n");
+ mp = map_by_uuid(&map, info.uuid);
+ if (mp)
+ mdfd = open_dev(mp->devnum);
+ else
+ mdfd = -1;
+
+ if (mdfd < 0) {
struct mdinfo *sra;
- struct mdinfo dinfo;
+
+ /* Couldn't find an existing array, maybe make a new one */
+ mdfd = create_mddev(match ? match->devname : NULL,
+ name_to_use, autof, trustworthy, chosen_name);
+
+ if (mdfd < 0)
+ goto out;
+
+ sysfs_init(&info, mdfd, 0);
if (set_array_info(mdfd, st, &info) != 0) {
fprintf(stderr, Name ": failed to set array info for %s: %s\n",
chosen_name, strerror(errno));
- close(mdfd);
- return 2;
+ rv = 2;
+ goto out;
}
dinfo = info;
fprintf(stderr, Name ": failed to add %s to %s: %s.\n",
devname, chosen_name, strerror(errno));
ioctl(mdfd, STOP_ARRAY, 0);
- close(mdfd);
- return 2;
+ rv = 2;
+ goto out;
}
- sra = sysfs_read(mdfd, devnum, GET_DEVS);
+ sra = sysfs_read(mdfd, fd2devnum(mdfd), GET_DEVS);
if (!sra || !sra->devs || sra->devs->disk.raid_disk >= 0) {
/* It really should be 'none' - must be old buggy
* kernel, and mdadm -I may not be able to complete.
fprintf(stderr, Name
": You have an old buggy kernel which cannot support\n"
" --incremental reliably. Aborting.\n");
- close(mdfd);
sysfs_free(sra);
- return 2;
+ rv = 2;
+ goto out;
}
info.array.working_disks = 1;
sysfs_free(sra);
+ /* 6/ Make sure /var/run/mdadm.map contains this array. */
+ map_update(&map, fd2devnum(mdfd),
+ info.text_version,
+ info.uuid, chosen_name);
} else {
/* 5b/ if it does */
/* - check one drive in array to make sure metadata is a reasonably */
struct mdinfo *sra;
struct supertype *st2;
struct mdinfo info2, *d;
- sra = sysfs_read(mdfd, devnum, (GET_DEVS | GET_STATE));
-
- sprintf(dn, "%d:%d", sra->devs->disk.major,
- sra->devs->disk.minor);
- dfd2 = dev_open(dn, O_RDONLY);
- st2 = dup_super(st);
- if (st2->ss->load_super(st2, dfd2, NULL) ||
- st->ss->compare_super(st, st2) != 0) {
- fprintf(stderr, Name
- ": metadata mismatch between %s and "
- "chosen array %s\n",
- devname, chosen_name);
- close(mdfd);
- close(dfd2);
- return 2;
+
+ if (mp->path)
+ strcpy(chosen_name, mp->path);
+ else
+ strcpy(chosen_name, devnum2devname(mp->devnum));
+
+ /* It is generally not OK to add non-spare drives to a
+ * running array as they are probably missing because
+ * they failed. However if runstop is 1, then the
+ * array was possibly started early and our best bet is
+ * to add this anyway.
+ * Also if action policy is re-add or better we allow
+ * re-add
+ */
+ if ((info.disk.state & (1<<MD_DISK_SYNC)) != 0
+ && ! policy_action_allows(policy, st->ss->name,
+ act_re_add)
+ && runstop < 1) {
+ int active = 0;
+
+ if (st->ss->external) {
+ char *devname = devnum2devname(fd2devnum(mdfd));
+
+ active = devname && is_container_active(devname);
+ free(devname);
+ } else if (ioctl(mdfd, GET_ARRAY_INFO, &ainf) == 0)
+ active = 1;
+ if (active) {
+ fprintf(stderr, Name
+ ": not adding %s to active array (without --run) %s\n",
+ devname, chosen_name);
+ rv = 2;
+ goto out;
+ }
}
- close(dfd2);
- memset(&info2, 0, sizeof(info2));
- st2->ss->getinfo_super(st2, &info2);
- st2->ss->free_super(st2);
- if (info.array.level != info2.array.level ||
- memcmp(info.uuid, info2.uuid, 16) != 0 ||
- info.array.raid_disks != info2.array.raid_disks) {
- fprintf(stderr, Name
- ": unexpected difference between %s and %s.\n",
- chosen_name, devname);
- close(mdfd);
- return 2;
+ sra = sysfs_read(mdfd, fd2devnum(mdfd), (GET_DEVS | GET_STATE));
+ if (!sra) {
+ rv = 2;
+ goto out;
+ }
+ if (sra->devs) {
+ sprintf(dn, "%d:%d", sra->devs->disk.major,
+ sra->devs->disk.minor);
+ dfd2 = dev_open(dn, O_RDONLY);
+ st2 = dup_super(st);
+ if (st2->ss->load_super(st2, dfd2, NULL) ||
+ st->ss->compare_super(st, st2) != 0) {
+ fprintf(stderr, Name
+ ": metadata mismatch between %s and "
+ "chosen array %s\n",
+ devname, chosen_name);
+ close(dfd2);
+ rv = 2;
+ goto out;
+ }
+ close(dfd2);
+ memset(&info2, 0, sizeof(info2));
+ st2->ss->getinfo_super(st2, &info2, NULL);
+ st2->ss->free_super(st2);
+ if (info.array.level != info2.array.level ||
+ memcmp(info.uuid, info2.uuid, 16) != 0 ||
+ info.array.raid_disks != info2.array.raid_disks) {
+ fprintf(stderr, Name
+ ": unexpected difference between %s and %s.\n",
+ chosen_name, devname);
+ rv = 2;
+ goto out;
+ }
}
info2.disk.major = major(stb.st_rdev);
info2.disk.minor = minor(stb.st_rdev);
/* add disk needs to know about containers */
if (st->ss->external)
sra->array.level = LEVEL_CONTAINER;
- err = add_disk(mdfd, st2, sra, &info2);
+ err = add_disk(mdfd, st, sra, &info2);
if (err < 0 && errno == EBUSY) {
/* could be another device present with the same
* disk.number. Find and reject any such
*/
find_reject(mdfd, st, sra, info.disk.number,
info.events, verbose, chosen_name);
- err = add_disk(mdfd, st2, sra, &info2);
+ err = add_disk(mdfd, st, sra, &info2);
}
if (err < 0) {
fprintf(stderr, Name ": failed to add %s to %s: %s.\n",
devname, chosen_name, strerror(errno));
- close(mdfd);
- return 2;
+ rv = 2;
+ goto out;
}
info.array.working_disks = 0;
for (d = sra->devs; d; d=d->next)
info.array.working_disks ++;
}
- /* 6/ Make sure /var/run/mdadm.map contains this array. */
- map_update(&map, devnum,
- info.text_version,
- info.uuid, chosen_name);
/* 7/ Is there enough devices to possibly start the array? */
/* 7a/ if not, finish with success. */
if (info.array.level == LEVEL_CONTAINER) {
/* Try to assemble within the container */
- close(mdfd);
+ map_unlock(&map);
+ sysfs_uevent(&info, "change");
if (verbose >= 0)
fprintf(stderr, Name
": container %s now has %d devices\n",
chosen_name, info.array.working_disks);
- return Incremental(chosen_name, verbose, runstop,
- NULL, homehost, autof);
+ wait_for(chosen_name, mdfd);
+ close(mdfd);
+ rv = Incremental(chosen_name, verbose, runstop,
+ NULL, homehost, require_homehost, autof);
+ if (rv == 1)
+ /* Don't fail the whole -I if a subarray didn't
+ * have enough devices to start yet
+ */
+ rv = 0;
+ return rv;
}
avail = NULL;
active_disks = count_active(st, mdfd, &avail, &info);
fprintf(stderr, Name
": %s attached to %s, not enough to start (%d).\n",
devname, chosen_name, active_disks);
- close(mdfd);
- return 0;
+ map_unlock(&map);
+ rv = 0;
+ goto out;
}
free(avail);
/* are enough, */
/* + add any bitmap file */
/* + start the array (auto-readonly). */
-{
- mdu_array_info_t ainf;
if (ioctl(mdfd, GET_ARRAY_INFO, &ainf) == 0) {
if (verbose >= 0)
fprintf(stderr, Name
": %s attached to %s which is already active.\n",
devname, chosen_name);
- close (mdfd);
- return 0;
+ map_unlock(&map);
+ rv = 0;
+ goto out;
}
-}
+
+ map_unlock(&map);
if (runstop > 0 || active_disks >= info.array.working_disks) {
- struct mdinfo *sra;
+ struct mdinfo *sra, *dsk;
/* Let's try to start it */
if (match && match->bitmap_file) {
int bmfd = open(match->bitmap_file, O_RDWR);
fprintf(stderr, Name
": Could not open bitmap file %s.\n",
match->bitmap_file);
- close(mdfd);
- return 1;
+ goto out;
}
if (ioctl(mdfd, SET_BITMAP_FILE, bmfd) != 0) {
close(bmfd);
fprintf(stderr, Name
": Failed to set bitmapfile for %s.\n",
chosen_name);
- close(mdfd);
- return 1;
+ goto out;
}
close(bmfd);
}
- sra = sysfs_read(mdfd, devnum, 0);
- if (sra == NULL || active_disks >= info.array.working_disks)
+ /* GET_* needed so add_disk works below */
+ sra = sysfs_read(mdfd, fd2devnum(mdfd),
+ GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE);
+ if ((sra == NULL || active_disks >= info.array.working_disks)
+ && trustworthy != FOREIGN)
rv = ioctl(mdfd, RUN_ARRAY, NULL);
else
rv = sysfs_set_str(sra, NULL,
if (rv == 0) {
if (verbose >= 0)
fprintf(stderr, Name
- ": %s attached to %s, which has been started.\n",
+ ": %s attached to %s, which has been started.\n",
devname, chosen_name);
rv = 0;
+ wait_for(chosen_name, mdfd);
+ /* We just started the array, so some devices
+ * might have been evicted from the array
+ * because their event counts were too old.
+ * If the action=re-add policy is in-force for
+ * those devices we should re-add them now.
+ */
+ for (dsk = sra->devs; dsk ; dsk = dsk->next) {
+ if (disk_action_allows(dsk, st->ss->name, act_re_add) &&
+ add_disk(mdfd, st, sra, dsk) == 0)
+ fprintf(stderr, Name
+ ": %s re-added to %s\n",
+ dsk->sys_name, chosen_name);
+ }
} else {
fprintf(stderr, Name
": %s attached to %s, but failed to start: %s.\n",
devname, chosen_name);
rv = 0;
}
- close(mdfd);
+out:
+ if (dfd >= 0)
+ close(dfd);
+ if (mdfd >= 0)
+ close(mdfd);
+ if (policy)
+ dev_policy_free(policy);
return rv;
}
close(dfd);
continue;
}
- st->ss->getinfo_super(st, &info);
+ st->ss->getinfo_super(st, &info, NULL);
st->ss->free_super(st);
close(dfd);
struct mdinfo *sra = sysfs_read(mdfd, -1, GET_DEVS | GET_STATE);
char *avail = NULL;
+ if (!sra)
+ return 0;
+
for (d = sra->devs ; d ; d = d->next) {
char dn[30];
int dfd;
close(dfd);
if (ok != 0)
continue;
- st->ss->getinfo_super(st, &info);
+ st->ss->getinfo_super(st, &info, NULL);
+ if (!avail) {
+ avail = malloc(info.array.raid_disks);
+ if (!avail) {
+ fprintf(stderr, Name ": out of memory.\n");
+ exit(1);
+ }
+ memset(avail, 0, info.array.raid_disks);
+ *availp = avail;
+ }
+
if (info.disk.state & (1<<MD_DISK_SYNC))
{
- if (avail == NULL) {
- avail = malloc(info.array.raid_disks);
- memset(avail, 0, info.array.raid_disks);
- }
if (cnt == 0) {
cnt++;
max_events = info.events;
avail[info.disk.raid_disk] = 2;
- st->ss->getinfo_super(st, bestinfo);
+ st->ss->getinfo_super(st, bestinfo, NULL);
} else if (info.events == max_events) {
cnt++;
avail[info.disk.raid_disk] = 2;
if (avail[i])
avail[i]--;
avail[info.disk.raid_disk] = 2;
- st->ss->getinfo_super(st, bestinfo);
+ st->ss->getinfo_super(st, bestinfo, NULL);
} else { /* info.events much bigger */
cnt = 1; cnt1 = 0;
memset(avail, 0, info.disk.raid_disk);
max_events = info.events;
- st->ss->getinfo_super(st, bestinfo);
+ avail[info.disk.raid_disk] = 2;
+ st->ss->getinfo_super(st, bestinfo, NULL);
}
}
st->ss->free_super(st);
return cnt + cnt1;
}
-void RebuildMap(void)
+static int array_try_spare(char *devname, int *dfdp, struct dev_policy *pol,
+ struct supertype *st, int verbose)
{
- struct mdstat_ent *mdstat = mdstat_read(0, 0);
- struct mdstat_ent *md;
- struct map_ent *map = NULL;
- int mdp = get_mdp_major();
-
- for (md = mdstat ; md ; md = md->next) {
- struct mdinfo *sra = sysfs_read(-1, md->devnum, GET_DEVS);
- struct mdinfo *sd;
-
- for (sd = sra->devs ; sd ; sd = sd->next) {
- char dn[30];
- int dfd;
- int ok;
- struct supertype *st;
- char *path;
- struct mdinfo info;
-
- sprintf(dn, "%d:%d", sd->disk.major, sd->disk.minor);
- dfd = dev_open(dn, O_RDONLY);
- if (dfd < 0)
+ /* This device doesn't have any md metadata
+ * If it is 'bare' and theh device policy allows 'spare' look for
+ * an array or container to attach it to.
+ * If st is set, then only arrays of that type are considered
+ * Return 0 on success, or some exit code on failure, probably 1.
+ */
+ int rv = -1;
+ struct stat stb;
+ struct map_ent *mp, *map = NULL;
+ struct mdinfo *chosen = NULL;
+ int dfd = *dfdp;
+
+ if (fstat(dfd, &stb) != 0)
+ return 1;
+
+ /*
+ * Now we need to find a suitable array to add this to.
+ * We only accept arrays that:
+ * - match 'st'
+ * - are in the same domains as the device
+ * - are of an size for which the device will be useful
+ * and we choose the one that is the most degraded
+ */
+
+ if (map_lock(&map)) {
+ fprintf(stderr, Name ": failed to get exclusive lock on "
+ "mapfile\n");
+ return 1;
+ }
+ for (mp = map ; mp ; mp = mp->next) {
+ struct supertype *st2;
+ struct domainlist *dl = NULL;
+ struct mdinfo *sra;
+ unsigned long long devsize;
+
+ if (is_subarray(mp->metadata))
+ continue;
+ if (st) {
+ st2 = st->ss->match_metadata_desc(mp->metadata);
+ if (!st2 ||
+ (st->minor_version >= 0 &&
+ st->minor_version != st2->minor_version)) {
+ if (verbose > 1)
+ fprintf(stderr, Name ": not adding %s to %s as metadata type doesn't match\n",
+ devname, mp->path);
+ free(st2);
continue;
- st = guess_super(dfd);
- if ( st == NULL)
- ok = -1;
- else
- ok = st->ss->load_super(st, dfd, NULL);
+ }
+ free(st2);
+ }
+ sra = sysfs_read(-1, mp->devnum,
+ GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE|
+ GET_DEGRADED|GET_COMPONENT|GET_VERSION);
+ if (!sra) {
+ /* Probably a container - no degraded info */
+ sra = sysfs_read(-1, mp->devnum,
+ GET_DEVS|GET_OFFSET|GET_SIZE|GET_STATE|
+ GET_COMPONENT|GET_VERSION);
+ if (sra)
+ sra->array.failed_disks = 0;
+ }
+ if (!sra)
+ continue;
+ if (st == NULL) {
+ int i;
+ st2 = NULL;
+ for(i=0; !st2 && superlist[i]; i++)
+ st2 = superlist[i]->match_metadata_desc(
+ sra->text_version);
+ } else
+ st2 = st;
+ get_dev_size(dfd, NULL, &devsize);
+ if (st2->ss->avail_size(st2, devsize) < sra->component_size) {
+ if (verbose > 1)
+ fprintf(stderr, Name ": not adding %s to %s as it is too small\n",
+ devname, mp->path);
+ goto next;
+ }
+ dl = domain_from_array(sra, st2->ss->name);
+ if (!domain_test(dl, pol, st2->ss->name)) {
+ /* domain test fails */
+ if (verbose > 1)
+ fprintf(stderr, Name ": not adding %s to %s as it is not in a compatible domain\n",
+ devname, mp->path);
+
+ goto next;
+ }
+ /* all tests passed, OK to add to this array */
+ if (!chosen) {
+ chosen = sra;
+ sra = NULL;
+ } else if (chosen->array.failed_disks < sra->array.failed_disks) {
+ sysfs_free(chosen);
+ chosen = sra;
+ sra = NULL;
+ }
+ next:
+ if (sra)
+ sysfs_free(sra);
+ if (st != st2)
+ free(st2);
+ if (dl)
+ domain_free(dl);
+ }
+ if (chosen) {
+ /* add current device to chosen array as a spare */
+ int mdfd = open_dev(devname2devnum(chosen->sys_name));
+ if (mdfd >= 0) {
+ struct mddev_dev_s devlist;
+ char devname[20];
+ devlist.next = NULL;
+ devlist.used = 0;
+ devlist.re_add = 0;
+ devlist.writemostly = 0;
+ devlist.devname = devname;
+ sprintf(devname, "%d:%d", major(stb.st_rdev),
+ minor(stb.st_rdev));
+ devlist.disposition = 'a';
close(dfd);
- if (ok != 0)
- continue;
- st->ss->getinfo_super(st, &info);
- if (md->devnum > 0)
- path = map_dev(MD_MAJOR, md->devnum, 0);
+ *dfdp = -1;
+ rv = Manage_subdevs(chosen->sys_name, mdfd, &devlist,
+ -1, 0);
+ close(mdfd);
+ }
+ if (verbose > 0) {
+ if (rv == 0)
+ fprintf(stderr, Name ": added %s as spare for %s\n",
+ devname, chosen->sys_name);
else
- path = map_dev(mdp, (-1-md->devnum)<< 6, 0);
- map_add(&map, md->devnum,
- info.text_version,
- info.uuid, path ? : "/unknown");
- st->ss->free_super(st);
- break;
+ fprintf(stderr, Name ": failed to add %s as spare for %s\n",
+ devname, chosen->sys_name);
+ }
+ sysfs_free(chosen);
+ }
+ return rv ? 0 : 1;
+}
+
+static int partition_try_spare(char *devname, int *dfdp, struct dev_policy *pol,
+ struct supertype *st, int verbose)
+{
+ /* we know that at least one partition virtual-metadata is
+ * allowed to incorporate spares like this device. We need to
+ * find a suitable device to copy partition information from.
+ *
+ * Getting a list of all disk (not partition) devices is
+ * slightly non-trivial. We could look at /sys/block, but
+ * that is theoretically due to be removed. Maybe best to use
+ * /dev/disk/by-path/?* and ignore names ending '-partNN' as
+ * we depend on this directory of 'path' info. But that fails
+ * to find loop devices and probably others. Maybe don't
+ * worry about that, they aren't the real target.
+ *
+ * So: check things in /dev/disk/by-path to see if they are in
+ * a compatible domain, then load the partition table and see
+ * if it is OK for the new device, and choose the largest
+ * partition table that fits.
+ */
+ DIR *dir;
+ struct dirent *de;
+ char *chosen = NULL;
+ unsigned long long chosen_size;
+ struct supertype *chosen_st = NULL;
+ int fd;
+
+ dir = opendir("/dev/disk/by-path");
+ if (!dir)
+ return 1;
+ while ((de = readdir(dir)) != NULL) {
+ char *ep;
+ struct dev_policy *pol2 = NULL;
+ struct domainlist *domlist = NULL;
+ int fd = -1;
+ struct mdinfo info;
+ struct supertype *st2 = NULL;
+ char *devname = NULL;
+ unsigned long long devsectors;
+
+ if (de->d_ino == 0 ||
+ de->d_name[0] == '.' ||
+ (de->d_type != DT_LNK && de->d_type != DT_UNKNOWN))
+ goto next;
+
+ ep = de->d_name + strlen(de->d_name);
+ while (ep > de->d_name &&
+ isdigit(ep[-1]))
+ ep--;
+ if (ep > de->d_name + 5 &&
+ strncmp(ep-5, "-part", 5) == 0)
+ /* This is a partition - skip it */
+ goto next;
+
+ pol2 = path_policy(de->d_name, type_disk);
+
+ domain_merge(&domlist, pol2, st ? st->ss->name : NULL);
+ if (domain_test(domlist, pol, st ? st->ss->name : NULL) == 0)
+ /* new device is incompatible with this device. */
+ goto next;
+
+ domain_free(domlist);
+ domlist = NULL;
+
+ asprintf(&devname, "/dev/disk/by-path/%s", de->d_name);
+ fd = open(devname, O_RDONLY);
+ if (fd < 0)
+ goto next;
+ if (get_dev_size(fd, devname, &devsectors) == 0)
+ goto next;
+ devsectors >>= 9;
+
+ if (st)
+ st2 = dup_super(st);
+ else
+ st2 = guess_super_type(fd, guess_partitions);
+ if (st2 == NULL ||
+ st2->ss->load_super(st2, fd, NULL) < 0)
+ goto next;
+
+ if (!st) {
+ /* Check domain policy again, this time referring to metadata */
+ domain_merge(&domlist, pol2, st2->ss->name);
+ if (domain_test(domlist, pol, st2->ss->name) == 0)
+ /* Incompatible devices for this metadata type */
+ goto next;
+ }
+
+ st2->ss->getinfo_super(st2, &info, NULL);
+ if (info.component_size > devsectors)
+ /* This partitioning doesn't fit in the device */
+ goto next;
+
+ /* This is an acceptable device to copy partition
+ * metadata from. We could just stop here, but I
+ * think I want to keep looking incase a larger
+ * metadata which makes better use of the device can
+ * be found.
+ */
+ if (chosen == NULL ||
+ chosen_size < info.component_size) {
+ chosen_size = info.component_size;
+ free(chosen);
+ chosen = devname;
+ devname = NULL;
+ if (chosen_st) {
+ chosen_st->ss->free_super(chosen_st);
+ free(chosen_st);
+ }
+ chosen_st = st2;
+ st2 = NULL;
}
+
+ next:
+ free(devname);
+ domain_free(domlist);
+ dev_policy_free(pol2);
+ if (st2)
+ st2->ss->free_super(st2);
+ free(st2);
+
+ if (fd >= 0)
+ close(fd);
+ }
+
+ if (!chosen)
+ return 1;
+
+ /* 'chosen' is the best device we can find. Let's write its
+ * metadata to devname dfd is read-only so don't use that
+ */
+ fd = open(devname, O_RDWR);
+ if (fd >= 0) {
+ chosen_st->ss->store_super(chosen_st, fd);
+ close(fd);
+ }
+ free(chosen);
+ chosen_st->ss->free_super(chosen_st);
+ free(chosen_st);
+ return 0;
+}
+
+
+/* adding a spare to a regular array is quite different from adding one to
+ * a set-of-partitions virtual array.
+ * This function determines which is worth trying and tries as appropriate.
+ * Arrays are given priority over partitions.
+ */
+static int try_spare(char *devname, int *dfdp, struct dev_policy *pol,
+ struct supertype *st, int verbose)
+{
+ int i;
+ int rv;
+ int arrays_ok = 0;
+ int partitions_ok = 0;
+ char bufpad[4096 + 4096];
+ char *buf = (char*)(((long)bufpad + 4096) & ~4095);
+ int dfd = *dfdp;
+
+ /* Can only add a spare if device has at least one domains */
+ if (pol_find(pol, pol_domain) == NULL)
+ return 1;
+ /* And only if some action allows spares */
+ if (!policy_action_allows(pol, st?st->ss->name:NULL, act_spare))
+ return 1;
+
+ /* Now check if the device is bare - we don't add non-bare devices
+ * yet even if action=-spare
+ */
+
+ if (lseek(dfd, 0, SEEK_SET) != 0 ||
+ read(dfd, buf, 4096) != 4096) {
+ not_bare:
+ if (verbose > 1)
+ fprintf(stderr, Name ": %s is not bare, so not considering as a spare\n",
+ devname);
+ return 1;
+ }
+ if (buf[0] != '\0' && buf[0] != '\x5a' && buf[0] != '\xff')
+ goto not_bare;
+ if (memcmp(buf, buf+1, 4095) != 0)
+ goto not_bare;
+
+ /* OK, first 4K appear blank, try the end. */
+ if (lseek(dfd, -4096, SEEK_END) < 0 ||
+ read(dfd, buf, 4096) != 4096)
+ goto not_bare;
+
+ if (buf[0] != '\0' && buf[0] != '\x5a' && buf[0] != '\xff')
+ goto not_bare;
+ if (memcmp(buf, buf+1, 4095) != 0)
+ goto not_bare;
+
+ /* This device passes our test for 'is bare'.
+ * Let's see what policy allows for such things.
+ */
+ if (st) {
+ /* just try try 'array' or 'partition' based on this metadata */
+ if (st->ss->add_to_super)
+ return array_try_spare(devname, dfdp, pol,
+ st, verbose);
+ else
+ return partition_try_spare(devname, dfdp, pol,
+ st, verbose);
}
- map_write(map);
- map_free(map);
+ /* Now see which metadata type support spare */
+ for (i = 0; (!arrays_ok || !partitions_ok) && superlist[i] ; i++) {
+ if (superlist[i]->add_to_super && !arrays_ok &&
+ policy_action_allows(pol, superlist[i]->name, act_spare))
+ arrays_ok = 1;
+ if (superlist[i]->add_to_super == NULL && !partitions_ok &&
+ policy_action_allows(pol, superlist[i]->name, act_spare))
+ partitions_ok = 1;
+ }
+ rv = 0;
+ if (arrays_ok)
+ rv = array_try_spare(devname, dfdp, pol, st, verbose);
+ if (rv == 0 && partitions_ok)
+ rv = partition_try_spare(devname, dfdp, pol, st, verbose);
+ return rv;
}
int IncrementalScan(int verbose)
devs = conf_get_ident(NULL);
for (me = mapl ; me ; me = me->next) {
- char path[1024];
mdu_array_info_t array;
mdu_bitmap_file_t bmf;
struct mdinfo *sra;
- int mdfd = open_mddev_devnum(me->path, me->devnum,
- NULL, path, 0);
+ int mdfd = open_dev(me->devnum);
+
if (mdfd < 0)
continue;
if (ioctl(mdfd, GET_ARRAY_INFO, &array) == 0 ||
}
/* Ok, we can try this one. Maybe it needs a bitmap */
for (mddev = devs ; mddev ; mddev = mddev->next)
- if (strcmp(mddev->devname, me->path) == 0)
+ if (mddev->devname && me->path
+ && devname_matches(mddev->devname, me->path))
break;
if (mddev && mddev->bitmap_file) {
/*
if (verbose >= 0)
fprintf(stderr, Name
": started array %s\n",
- me->path);
+ me->path ?: devnum2devname(me->devnum));
} else {
fprintf(stderr, Name
": failed to start array %s: %s\n",
- me->path, strerror(errno));
+ me->path ?: devnum2devname(me->devnum),
+ strerror(errno));
rv = 1;
}
}
static char *container2devname(char *devname)
{
- int fd = open(devname, O_RDONLY);
char *mdname = NULL;
- if (fd >= 0) {
- mdname = devnum2devname(fd2devnum(fd));
- close(fd);
+ if (devname[0] == '/') {
+ int fd = open(devname, O_RDONLY);
+ if (fd >= 0) {
+ mdname = devnum2devname(fd2devnum(fd));
+ close(fd);
+ }
+ } else {
+ int uuid[4];
+ struct map_ent *mp, *map = NULL;
+
+ if (!parse_uuid(devname, uuid))
+ return mdname;
+ mp = map_by_uuid(&map, uuid);
+ if (mp)
+ mdname = devnum2devname(mp->devnum);
+ map_free(map);
}
return mdname;
}
int Incremental_container(struct supertype *st, char *devname, int verbose,
- int runstop, int autof)
+ int runstop, int autof, int trustworthy)
{
/* Collect the contents of this container and for each
* array, choose a device name and assemble the array.
struct mdinfo *list = st->ss->container_content(st);
struct mdinfo *ra;
- char *mdname = container2devname(devname);
+ struct map_ent *map = NULL;
- if (!mdname) {
- fprintf(stderr, Name": failed to determine device name\n");
- return 2;
- }
+ if (map_lock(&map))
+ fprintf(stderr, Name ": failed to get exclusive lock on "
+ "mapfile\n");
for (ra = list ; ra ; ra = ra->next) {
- struct mdinfo *dev, *sra;
- int devnum = -1;
int mdfd;
char chosen_name[1024];
- int usepart = 1;
- char *n;
- int working = 0, preexist = 0;
- struct map_ent *mp, *map = NULL;
- char nbuf[64];
- char *name_to_use;
+ struct map_ent *mp;
struct mddev_ident_s *match = NULL;
- if ((autof&7) == 3 || (autof&7) == 5)
- usepart = 0;
-
mp = map_by_uuid(&map, ra->uuid);
- name_to_use = ra->name;
- if (! name_to_use ||
- ! *name_to_use ||
- (*devname != '/' || strncmp("UUID-", strrchr(devname,'/')+1,5) == 0)
- )
- name_to_use = fname_from_uuid(st, ra, nbuf, '-');
-
- if (!mp) {
+ if (mp) {
+ mdfd = open_dev(mp->devnum);
+ if (mp->path)
+ strcpy(chosen_name, mp->path);
+ else
+ strcpy(chosen_name, devnum2devname(mp->devnum));
+ } else {
- /* Check in mdadm.conf for devices == devname and
+ /* Check in mdadm.conf for container == devname and
* member == ra->text_version after second slash.
*/
char *sub = strchr(ra->text_version+1, '/');
} else
array_list = NULL;
for(; array_list ; array_list = array_list->next) {
- int fd;
char *dn;
if (array_list->member == NULL ||
array_list->container == NULL)
continue;
if (strcmp(array_list->member, sub) != 0)
continue;
- fd = open(array_list->container, O_RDONLY);
- if (fd < 0)
+ if (array_list->uuid_set &&
+ !same_uuid(ra->uuid, array_list->uuid, st->ss->swapuuid))
+ continue;
+ dn = container2devname(array_list->container);
+ if (dn == NULL)
continue;
- dn = devnum2devname(fd2devnum(fd));
- close(fd);
if (strncmp(dn, ra->text_version+1,
strlen(dn)) != 0 ||
ra->text_version[strlen(dn)+1] != '/') {
free(dn);
/* we have a match */
match = array_list;
+ if (verbose>0)
+ fprintf(stderr, Name ": match found for member %s\n",
+ array_list->member);
break;
}
- }
- if (match && is_standard(match->devname, &devnum))
- /* we have devnum now */;
- else if (mp)
- devnum = mp->devnum;
- else if (is_standard(name_to_use, &devnum))
- /* have devnum */;
- else {
- n = name_to_use;
- if (*n == 'd')
- n++;
- if (*n && devnum < 0) {
- devnum = strtoul(n, &n, 10);
- if (devnum >= 0 && (*n == 0 || *n == ' ')) {
- /* Use this devnum */
- usepart = (name_to_use[0] == 'd');
- if (mddev_busy(usepart ? (-1-devnum) : devnum))
- devnum = -1;
- } else
- devnum = -1;
- }
-
- if (devnum < 0) {
- char *nm = name_to_use;
- char nbuf[1024];
- struct stat stb;
- if (strchr(nm, ':'))
- nm = strchr(nm, ':')+1;
- sprintf(nbuf, "/dev/md/%s", nm);
-
- if (stat(nbuf, &stb) == 0 &&
- S_ISBLK(stb.st_mode) &&
- major(stb.st_rdev) == (usepart ?
- get_mdp_major() : MD_MAJOR)){
- if (usepart)
- devnum = minor(stb.st_rdev)
- >> MdpMinorShift;
- else
- devnum = minor(stb.st_rdev);
- if (mddev_busy(usepart ? (-1-devnum) : devnum))
- devnum = -1;
- }
+ if (match && match->devname &&
+ strcasecmp(match->devname, "<ignore>") == 0) {
+ if (verbose > 0)
+ fprintf(stderr, Name ": array %s/%s is "
+ "explicitly ignored by mdadm.conf\n",
+ match->container, match->member);
+ return 2;
}
-
- if (devnum >= 0)
- devnum = usepart ? (-1-devnum) : devnum;
- else
- devnum = find_free_devnum(usepart);
+ if (match)
+ trustworthy = LOCAL;
+
+ mdfd = create_mddev(match ? match->devname : NULL,
+ ra->name,
+ autof,
+ trustworthy,
+ chosen_name);
}
- mdfd = open_mddev_devnum(mp ? mp->path : match ? match->devname : NULL,
- devnum, name_to_use,
- chosen_name, autof>>3);
if (mdfd < 0) {
fprintf(stderr, Name ": failed to open %s: %s.\n",
return 2;
}
+ assemble_container_content(st, mdfd, ra, runstop,
+ chosen_name, verbose);
+ }
+ map_unlock(&map);
+ return 0;
+}
- sysfs_init(ra, mdfd, 0);
-
- sra = sysfs_read(mdfd, 0, GET_VERSION);
- if (sra == NULL || strcmp(sra->text_version, ra->text_version) != 0)
- if (sysfs_set_array(ra, md_get_version(mdfd)) != 0)
- return 1;
- if (sra)
- sysfs_free(sra);
+/*
+ * IncrementalRemove - Attempt to see if the passed in device belongs to any
+ * raid arrays, and if so first fail (if needed) and then remove the device.
+ *
+ * @devname - The device we want to remove
+ *
+ * Note: the device name must be a kernel name like "sda", so
+ * that we can find it in /proc/mdstat
+ */
+int IncrementalRemove(char *devname, int verbose)
+{
+ int mdfd;
+ int rv;
+ struct mdstat_ent *ent;
+ struct mddev_dev_s devlist;
- for (dev = ra->devs; dev; dev = dev->next)
- if (sysfs_add_disk(ra, dev) == 0)
- working++;
- else if (errno == EEXIST)
- preexist++;
- if (working == 0)
- /* Nothing new, don't try to start */ ;
- else if (runstop > 0 ||
- (working + preexist) >= ra->array.working_disks) {
- switch(ra->array.level) {
- case LEVEL_LINEAR:
- case LEVEL_MULTIPATH:
- case 0:
- sysfs_set_str(ra, NULL, "array_state",
- "active");
- break;
- default:
- sysfs_set_str(ra, NULL, "array_state",
- "readonly");
- /* start mdmon if needed. */
- if (!mdmon_running(st->container_dev))
- start_mdmon(st->container_dev);
- ping_monitor(devnum2devname(st->container_dev));
- break;
- }
- sysfs_set_safemode(ra, ra->safe_mode_delay);
- if (verbose >= 0) {
- fprintf(stderr, Name
- ": Started %s with %d devices",
- chosen_name, working + preexist);
- if (preexist)
- fprintf(stderr, " (%d new)", working);
- fprintf(stderr, "\n");
- }
- /* FIXME should have an O_EXCL and wait for read-auto */
- } else
- if (verbose >= 0)
- fprintf(stderr, Name
- ": %s assembled with %d devices but "
- "not started\n",
- chosen_name, working);
- close(mdfd);
- map_update(&map, devnum,
- ra->text_version,
- ra->uuid, chosen_name);
+ if (strchr(devname, '/')) {
+ fprintf(stderr, Name ": incremental removal requires a "
+ "kernel device name, not a file: %s\n", devname);
+ return 1;
}
- return 0;
+ ent = mdstat_by_component(devname);
+ if (!ent) {
+ fprintf(stderr, Name ": %s does not appear to be a component "
+ "of any array\n", devname);
+ return 1;
+ }
+ mdfd = open_dev(ent->devnum);
+ if (mdfd < 0) {
+ fprintf(stderr, Name ": Cannot open array %s!!\n", ent->dev);
+ return 1;
+ }
+ memset(&devlist, 0, sizeof(devlist));
+ devlist.devname = devname;
+ devlist.disposition = 'f';
+ Manage_subdevs(ent->dev, mdfd, &devlist, verbose, 0);
+ devlist.disposition = 'r';
+ rv = Manage_subdevs(ent->dev, mdfd, &devlist, verbose, 0);
+ close(mdfd);
+ return rv;
}