/*
* mdadm - Intel(R) Matrix Storage Manager Support
*
- * Copyright (C) 2002-2007 Intel Corporation
+ * Copyright (C) 2002-2008 Intel Corporation
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
} __attribute__ ((packed));
struct imsm_vol {
- __u32 reserved[2];
+ __u32 curr_migr_unit;
+ __u32 reserved;
__u8 migr_state; /* Normal or Migrating */
__u8 migr_type; /* Initializing, Rebuilding, ... */
__u8 dirty;
int fd;
} *disks;
struct dl *add; /* list of disks to add while mdmon active */
+ struct dl *missing; /* disks removed while we weren't looking */
struct bbm_log *bbm_log;
};
enum imsm_update_type type;
};
-static int imsm_env_devname_as_serial(void)
-{
- char *val = getenv("IMSM_DEVNAME_AS_SERIAL");
-
- if (val && atoi(val) == 1)
- return 1;
-
- return 0;
-}
-
-
static struct supertype *match_metadata_desc_imsm(char *arg)
{
struct supertype *st;
return map->raid_level;
}
-#ifndef MDASSEMBLE
static int cmp_extent(const void *av, const void *bv)
{
const struct extent *a = av;
struct extent *rv, *e;
int i, j;
int memberships = 0;
+ __u32 reservation = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
for (i = 0; i < super->anchor->num_raid_devs; i++) {
struct imsm_dev *dev = get_imsm_dev(super, i);
}
qsort(rv, memberships, sizeof(*rv), cmp_extent);
- e->start = __le32_to_cpu(dl->disk.total_blocks) -
- (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
+ /* determine the start of the metadata
+ * when no raid devices are defined use the default
+ * ...otherwise allow the metadata to truncate the value
+ * as is the case with older versions of imsm
+ */
+ if (memberships) {
+ struct extent *last = &rv[memberships - 1];
+ __u32 remainder;
+
+ remainder = __le32_to_cpu(dl->disk.total_blocks) -
+ (last->start + last->size);
+ if (reservation > remainder)
+ reservation = remainder;
+ }
+ e->start = __le32_to_cpu(dl->disk.total_blocks) - reservation;
e->size = 0;
return rv;
}
+/* try to determine how much space is reserved for metadata from
+ * the last get_extents() entry, otherwise fallback to the
+ * default
+ */
+static __u32 imsm_reserved_sectors(struct intel_super *super, struct dl *dl)
+{
+ struct extent *e;
+ int i;
+ __u32 rv;
+
+ /* for spares just return a minimal reservation which will grow
+ * once the spare is picked up by an array
+ */
+ if (dl->index == -1)
+ return MPB_SECTOR_CNT;
+
+ e = get_extents(super, dl);
+ if (!e)
+ return MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
+
+ /* scroll to last entry */
+ for (i = 0; e[i].size; i++)
+ continue;
+
+ rv = __le32_to_cpu(dl->disk.total_blocks) - e[i].start;
+
+ free(e);
+
+ return rv;
+}
+
+#ifndef MDASSEMBLE
+static void getinfo_super_imsm(struct supertype *st, struct mdinfo *info);
+
static void print_imsm_dev(struct imsm_dev *dev, int index)
{
__u64 sz;
__u32 ord;
printf("\n");
- printf("[%s]:\n", dev->volume);
+ printf("[%.16s]:\n", dev->volume);
printf(" RAID Level : %d\n", get_imsm_raid_level(map));
printf(" Members : %d\n", map->num_members);
for (slot = 0; slot < map->num_members; slot++)
printf(" Dirty State : %s\n", dev->vol.dirty ? "dirty" : "clean");
}
-static void print_imsm_disk(struct imsm_super *mpb, int index)
+static void print_imsm_disk(struct imsm_super *mpb, int index, __u32 reserved)
{
struct imsm_disk *disk = __get_imsm_disk(mpb, index);
char str[MAX_RAID_SERIAL_LEN + 1];
s&FAILED_DISK ? " failed" : "",
s&USABLE_DISK ? " usable" : "");
printf(" Id : %08x\n", __le32_to_cpu(disk->scsi_id));
- sz = __le32_to_cpu(disk->total_blocks) -
- (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS * mpb->num_raid_devs);
+ sz = __le32_to_cpu(disk->total_blocks) - reserved;
printf(" Usable Size : %llu%s\n", (unsigned long long)sz,
human_size(sz * 512));
}
struct imsm_super *mpb = super->anchor;
char str[MAX_SIGNATURE_LENGTH];
int i;
+ struct mdinfo info;
+ char nbuf[64];
__u32 sum;
+ __u32 reserved = imsm_reserved_sectors(super, super->disks);
+
snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
printf(" Magic : %s\n", str);
printf(" Version : %s\n", get_imsm_version(mpb));
printf(" Family : %08x\n", __le32_to_cpu(mpb->family_num));
printf(" Generation : %08x\n", __le32_to_cpu(mpb->generation_num));
+ getinfo_super_imsm(st, &info);
+ fname_from_uuid(st, &info, nbuf,'-');
+ printf(" UUID : %s\n", nbuf + 5);
sum = __le32_to_cpu(mpb->check_sum);
printf(" Checksum : %08x %s\n", sum,
__gen_imsm_checksum(mpb) == sum ? "correct" : "incorrect");
printf(" MPB Sectors : %d\n", mpb_sectors(mpb));
printf(" Disks : %d\n", mpb->num_disks);
printf(" RAID Devices : %d\n", mpb->num_raid_devs);
- print_imsm_disk(mpb, super->disks->index);
+ print_imsm_disk(mpb, super->disks->index, reserved);
if (super->bbm_log) {
struct bbm_log *log = super->bbm_log;
for (i = 0; i < mpb->num_disks; i++) {
if (i == super->disks->index)
continue;
- print_imsm_disk(mpb, i);
+ print_imsm_disk(mpb, i, reserved);
}
}
static void brief_examine_super_imsm(struct supertype *st)
{
- /* We just write a generic DDF ARRAY entry
- */
+ /* We just write a generic IMSM ARRAY entry */
struct mdinfo info;
char nbuf[64];
+ struct intel_super *super = st->sb;
+ int i;
+
+ if (!super->anchor->num_raid_devs)
+ return;
getinfo_super_imsm(st, &info);
fname_from_uuid(st, &info, nbuf,'-');
- printf("ARRAY /dev/imsm metadata=imsm UUID=%s\n", nbuf + 5);
+ printf("ARRAY /dev/imsm metadata=imsm auto=md UUID=%s\n", nbuf + 5);
+ for (i = 0; i < super->anchor->num_raid_devs; i++) {
+ struct imsm_dev *dev = get_imsm_dev(super, i);
+
+ super->current_vol = i;
+ getinfo_super_imsm(st, &info);
+ fname_from_uuid(st, &info, nbuf,'-');
+ printf("ARRAY /dev/md/%.16s container=/dev/imsm member=%d auto=mdp UUID=%s\n",
+ dev->volume, i, nbuf + 5);
+ }
}
static void detail_super_imsm(struct supertype *st, char *homehost)
case 6:
return ALGORITHM_LEFT_ASYMMETRIC;
case 10:
- return 0x102; //FIXME is this correct?
+ return 0x102;
}
return -1;
}
info->component_size = __le32_to_cpu(map->blocks_per_member);
memset(info->uuid, 0, sizeof(info->uuid));
- if (map->map_state == IMSM_T_STATE_UNINITIALIZED ||
- dev->vol.dirty || dev->vol.migr_state)
+ if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
info->resync_start = 0;
+ else if (dev->vol.migr_state)
+ info->resync_start = __le32_to_cpu(dev->vol.curr_migr_unit);
else
info->resync_start = ~0ULL;
info->name[0] = 0;
if (super->disks) {
+ __u32 reserved = imsm_reserved_sectors(super, super->disks);
+
disk = &super->disks->disk;
- info->disk.number = super->disks->index;
- info->disk.raid_disk = super->disks->index;
- info->data_offset = __le32_to_cpu(disk->total_blocks) -
- (MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS);
- info->component_size = MPB_SECTOR_CNT + IMSM_RESERVED_SECTORS;
+ info->data_offset = __le32_to_cpu(disk->total_blocks) - reserved;
+ info->component_size = reserved;
s = __le32_to_cpu(disk->status);
info->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
info->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
- info->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
+ info->disk.state |= s & SPARE_DISK ? 0 : (1 << MD_DISK_SYNC);
}
- uuid_from_super_imsm(st, info->uuid);
+
+ /* only call uuid_from_super_imsm when this disk is part of a populated container,
+ * ->compare_super may have updated the 'num_raid_devs' field for spares
+ */
+ if (info->disk.state & (1 << MD_DISK_SYNC) || super->anchor->num_raid_devs)
+ uuid_from_super_imsm(st, info->uuid);
+ else
+ memcpy(info->uuid, uuid_match_any, sizeof(int[4]));
}
static int update_super_imsm(struct supertype *st, struct mdinfo *info,
*/
if (first->anchor->num_raid_devs == 0 &&
sec->anchor->num_raid_devs > 0) {
+ int i;
+
+ /* we need to copy raid device info from sec if an allocation
+ * fails here we don't associate the spare
+ */
+ for (i = 0; i < sec->anchor->num_raid_devs; i++) {
+ first->dev_tbl[i] = malloc(sizeof(struct imsm_dev));
+ if (!first->dev_tbl) {
+ while (--i >= 0) {
+ free(first->dev_tbl[i]);
+ first->dev_tbl[i] = NULL;
+ }
+ fprintf(stderr, "imsm: failed to associate spare\n");
+ return 3;
+ }
+ *first->dev_tbl[i] = *sec->dev_tbl[i];
+ }
+
first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
first->anchor->family_num = sec->anchor->family_num;
}
rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial));
- if (rv && imsm_env_devname_as_serial()) {
+ if (rv && check_env("IMSM_DEVNAME_AS_SERIAL")) {
memset(serial, 0, MAX_RAID_SERIAL_LEN);
fd2devname(fd, (char *) serial);
return 0;
return rv;
}
- /* trim whitespace */
+ /* trim leading whitespace */
rsp_len = scsi_serial[3];
rsp_buf = (char *) &scsi_serial[4];
c = rsp_buf;
while (isspace(*c))
c++;
+
+ /* truncate len to the end of rsp_buf if necessary */
if (c + MAX_RAID_SERIAL_LEN > rsp_buf + rsp_len)
len = rsp_len - (c - rsp_buf);
else
len = MAX_RAID_SERIAL_LEN;
+
+ /* initialize the buffer and copy rsp_buf characters */
+ memset(serial, 0, MAX_RAID_SERIAL_LEN);
memcpy(serial, c, len);
+
+ /* trim trailing whitespace starting with the last character copied */
c = (char *) &serial[len - 1];
while (isspace(*c) || *c == '\0')
*c-- = '\0';
}
}
+ /* no match, maybe a stale failed drive */
+ if (i == super->anchor->num_disks && dl->index >= 0) {
+ dl->disk = *__get_imsm_disk(super->anchor, dl->index);
+ if (__le32_to_cpu(dl->disk.status) & FAILED_DISK)
+ dl->index = -2;
+ }
+
if (alloc)
super->disks = dl;
dev->vol.migr_state = 1;
dev->vol.migr_type = rebuild_resync;
+ dev->vol.curr_migr_unit = 0;
dest = get_imsm_map(dev, 1);
memcpy(dest, src, sizeof_imsm_map(src));
src->map_state = to_state;
}
+
+static void end_migration(struct imsm_dev *dev, __u8 map_state)
+{
+ struct imsm_map *map = get_imsm_map(dev, 0);
+
+ dev->vol.migr_state = 0;
+ dev->vol.curr_migr_unit = 0;
+ map->map_state = map_state;
+}
#endif
static int parse_raid_devices(struct intel_super *super)
}
static void free_imsm_disks(struct intel_super *super)
{
- while (super->disks) {
- struct dl *d = super->disks;
+ struct dl *d;
+ while (super->disks) {
+ d = super->disks;
super->disks = d->next;
__free_imsm_disk(d);
}
+ while (super->missing) {
+ d = super->missing;
+ super->missing = d->next;
+ __free_imsm_disk(d);
+ }
+
}
/* free all the pieces hanging off of a super pointer */
}
#ifndef MDASSEMBLE
+/* find_missing - helper routine for load_super_imsm_all that identifies
+ * disks that have disappeared from the system. This routine relies on
+ * the mpb being uptodate, which it is at load time.
+ */
+static int find_missing(struct intel_super *super)
+{
+ int i;
+ struct imsm_super *mpb = super->anchor;
+ struct dl *dl;
+ struct imsm_disk *disk;
+ __u32 status;
+
+ for (i = 0; i < mpb->num_disks; i++) {
+ disk = __get_imsm_disk(mpb, i);
+ for (dl = super->disks; dl; dl = dl->next)
+ if (serialcmp(dl->disk.serial, disk->serial) == 0)
+ break;
+ if (dl)
+ continue;
+ /* ok we have a 'disk' without a live entry in
+ * super->disks
+ */
+ status = __le32_to_cpu(disk->status);
+ if (status & FAILED_DISK || !(status & USABLE_DISK))
+ continue; /* never mind, already marked */
+
+ dl = malloc(sizeof(*dl));
+ if (!dl)
+ return 1;
+ dl->major = 0;
+ dl->minor = 0;
+ dl->fd = -1;
+ dl->devname = strdup("missing");
+ dl->index = i;
+ serialcpy(dl->serial, disk->serial);
+ dl->disk = *disk;
+ dl->next = super->missing;
+ super->missing = dl;
+ }
+
+ return 0;
+}
+
static int load_super_imsm_all(struct supertype *st, int fd, void **sbp,
char *devname, int keep_fd)
{
close(dfd);
}
+
+ if (find_missing(super) != 0) {
+ free_imsm(super);
+ return 2;
+ }
+
if (st->subarray[0]) {
if (atoi(st->subarray) <= super->anchor->num_raid_devs)
super->current_vol = atoi(st->subarray);
vol->migr_state = 0;
vol->migr_type = 0;
vol->dirty = 0;
+ vol->curr_migr_unit = 0;
for (i = 0; i < idx; i++) {
struct imsm_dev *prev = get_imsm_dev(super, i);
struct imsm_map *pmap = get_imsm_map(prev, 0);
generation++;
mpb->generation_num = __cpu_to_le32(generation);
+ mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
for (d = super->disks; d; d = d->next) {
if (d->index == -1)
spares++;
- else {
+ else
mpb->disk[d->index] = d->disk;
- mpb_size += sizeof(struct imsm_disk);
- }
}
+ for (d = super->missing; d; d = d->next)
+ mpb->disk[d->index] = d->disk;
for (i = 0; i < mpb->num_raid_devs; i++) {
struct imsm_dev *dev = __get_imsm_dev(mpb, i);
case 10:
{
/**
- * check to see if any mirrors have failed,
- * otherwise we are degraded
+ * check to see if any mirrors have failed, otherwise we
+ * are degraded. Even numbered slots are mirrored on
+ * slot+1
*/
- int device_per_mirror = 2; /* FIXME is this always the case?
- * and are they always adjacent?
- */
- int r10fail = 0;
int i;
+ /* gcc -Os complains that this is unused */
+ int insync = insync;
for (i = 0; i < map->num_members; i++) {
- int idx = get_imsm_disk_idx(dev, i);
- struct imsm_disk *disk = get_imsm_disk(super, idx);
+ __u32 ord = get_imsm_ord_tbl_ent(dev, i);
+ int idx = ord_to_idx(ord);
+ struct imsm_disk *disk;
- if (!disk)
- r10fail++;
- else if (__le32_to_cpu(disk->status) & FAILED_DISK)
- r10fail++;
+ /* reset the potential in-sync count on even-numbered
+ * slots. num_copies is always 2 for imsm raid10
+ */
+ if ((i & 1) == 0)
+ insync = 2;
- if (r10fail >= device_per_mirror)
- return IMSM_T_STATE_FAILED;
+ disk = get_imsm_disk(super, idx);
+ if (!disk ||
+ __le32_to_cpu(disk->status) & FAILED_DISK ||
+ ord & IMSM_ORD_REBUILD)
+ insync--;
- /* reset 'r10fail' for next mirror set */
- if (!((i + 1) % device_per_mirror))
- r10fail = 0;
+ /* no in-sync disks left in this mirror the
+ * array has failed
+ */
+ if (insync == 0)
+ return IMSM_T_STATE_FAILED;
}
return IMSM_T_STATE_DEGRADED;
return 0;
}
+static void mark_failure(struct imsm_disk *disk)
+{
+ __u32 status = __le32_to_cpu(disk->status);
+
+ if (status & FAILED_DISK)
+ return;
+ status |= FAILED_DISK;
+ disk->status = __cpu_to_le32(status);
+ disk->scsi_id = __cpu_to_le32(~(__u32)0);
+ memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
+}
+
/* Handle dirty -> clean transititions and resync. Degraded and rebuild
* states are handled in imsm_set_disk() with one exception, when a
* resync is stopped due to a new failure this routine will set the
int failed = imsm_count_failed(super, dev);
__u8 map_state = imsm_check_degraded(super, dev, failed);
+ /* before we activate this array handle any missing disks */
+ if (consistent == 2 && super->missing) {
+ struct dl *dl;
+
+ dprintf("imsm: mark missing\n");
+ end_migration(dev, map_state);
+ for (dl = super->missing; dl; dl = dl->next)
+ mark_failure(&dl->disk);
+ super->updates_pending++;
+ }
+
if (consistent == 2 &&
- (a->resync_start != ~0ULL ||
+ (!is_resync_complete(a) ||
map_state != IMSM_T_STATE_NORMAL ||
dev->vol.migr_state))
consistent = 0;
- if (a->resync_start == ~0ULL) {
+ if (is_resync_complete(a)) {
/* complete intialization / resync,
* recovery is completed in ->set_disk
*/
if (is_resyncing(dev)) {
dprintf("imsm: mark resync done\n");
- dev->vol.migr_state = 0;
- map->map_state = map_state;
+ end_migration(dev, map_state);
super->updates_pending++;
}
} else if (!is_resyncing(dev) && !failed) {
super->updates_pending++;
}
+ /* check if we can update the migration checkpoint */
+ if (dev->vol.migr_state &&
+ __le32_to_cpu(dev->vol.curr_migr_unit) != a->resync_start) {
+ dprintf("imsm: checkpoint migration (%llu)\n", a->resync_start);
+ dev->vol.curr_migr_unit = __cpu_to_le32(a->resync_start);
+ super->updates_pending++;
+ }
+
/* mark dirty / clean */
if (dev->vol.dirty != !consistent) {
dprintf("imsm: mark '%s' (%llu)\n",
/* check for new failures */
status = __le32_to_cpu(disk->status);
if ((state & DS_FAULTY) && !(status & FAILED_DISK)) {
- status |= FAILED_DISK;
- disk->status = __cpu_to_le32(status);
- disk->scsi_id = __cpu_to_le32(~(__u32)0);
- memmove(&disk->serial[0], &disk->serial[1], MAX_RAID_SERIAL_LEN - 1);
+ mark_failure(disk);
super->updates_pending++;
}
+
/* check if in_sync */
if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD) {
struct imsm_map *migr_map = get_imsm_map(dev, 1);
/* check if recovery complete, newly degraded, or failed */
if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) {
- map->map_state = map_state;
- dev->vol.migr_state = 0;
+ end_migration(dev, map_state);
super->updates_pending++;
} else if (map_state == IMSM_T_STATE_DEGRADED &&
map->map_state != map_state &&
} else if (map_state == IMSM_T_STATE_FAILED &&
map->map_state != map_state) {
dprintf("imsm: mark failed\n");
- dev->vol.migr_state = 0;
- map->map_state = map_state;
+ end_migration(dev, map_state);
super->updates_pending++;
}
}
/* found a usable disk with enough space */
di = malloc(sizeof(*di));
+ if (!di)
+ continue;
memset(di, 0, sizeof(*di));
/* dl->index will be -1 in the case we are activating a
* disk_ord_tbl for the array
*/
mu = malloc(sizeof(*mu));
- mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares);
+ if (mu) {
+ mu->buf = malloc(sizeof(struct imsm_update_activate_spare) * num_spares);
+ if (mu->buf == NULL) {
+ free(mu);
+ mu = NULL;
+ }
+ }
+ if (!mu) {
+ while (rv) {
+ struct mdinfo *n = rv->next;
+
+ free(rv);
+ rv = n;
+ }
+ return NULL;
+ }
+
mu->space = NULL;
mu->len = sizeof(struct imsm_update_activate_spare) * num_spares;
mu->next = *updates;
if (!found) {
struct dl **dlp;
+ /* We know that 'manager' isn't touching anything,
+ * so it is safe to delete
+ */
for (dlp = &super->disks; *dlp; dlp = &(*dlp)->next)
if ((*dlp)->index == victim)
break;
- /* We know that 'manager' isn't touching anything,
- * so it is safe to:
- */
+
+ /* victim may be on the missing list */
+ if (!*dlp)
+ for (dlp = &super->missing; *dlp; dlp = &(*dlp)->next)
+ if ((*dlp)->index == victim)
+ break;
imsm_delete(super, dlp, victim);
}
break;
* being added */
if (super->add) {
struct active_array *a;
+
+ super->updates_pending++;
for (a = st->arrays; a; a = a->next)
a->check_degraded = 1;
}
for (iter = super->disks; iter; iter = iter->next)
if (iter->index > index)
iter->index--;
+ for (iter = super->missing; iter; iter = iter->next)
+ if (iter->index > index)
+ iter->index--;
for (i = 0; i < mpb->num_raid_devs; i++) {
dev = get_imsm_dev(super, i);
projects / thirdparty / mdadm.git / blobdiff