* 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
*/
+#define HAVE_STDINT_H 1
#include "mdadm.h"
#include "mdmon.h"
+#include "sha1.h"
#include <values.h>
#include <scsi/sg.h>
#include <ctype.h>
__u8 num_members; /* number of member disks */
__u8 reserved[3];
__u32 filler[7]; /* expansion area */
+#define IMSM_ORD_REBUILD (1 << 24)
__u32 disk_ord_tbl[1]; /* disk_ord_tbl[num_members],
- top byte special */
+ * top byte contains some flags
+ */
} __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;
struct imsm_super *anchor; /* immovable parameters */
};
size_t len; /* size of the 'buf' allocation */
+ void *next_buf; /* for realloc'ing buf from the manager */
+ size_t next_len;
int updates_pending; /* count of pending updates for mdmon */
int creating_imsm; /* flag to indicate container creation */
int current_vol; /* index of raid device undergoing creation */
struct imsm_disk disk;
int fd;
} *disks;
- struct bbm_log *bbm_log;
+ 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;
};
struct extent {
enum imsm_update_type {
update_activate_spare,
update_create_array,
+ update_add_disk,
};
struct imsm_update_activate_spare {
struct imsm_update_create_array {
enum imsm_update_type type;
- struct imsm_dev dev;
int dev_idx;
+ struct imsm_dev dev;
+};
+
+struct imsm_update_add_disk {
+ enum imsm_update_type type;
};
static int imsm_env_devname_as_serial(void)
return st;
}
+#ifndef MDASSEMBLE
static __u8 *get_imsm_version(struct imsm_super *mpb)
{
return &mpb->sig[MPB_SIG_LEN];
}
+#endif
/* retrieve a disk directly from the anchor when the anchor is known to be
* up-to-date, currently only at load time
return &mpb->disk[index];
}
+#ifndef MDASSEMBLE
/* retrieve a disk from the parsed metadata */
static struct imsm_disk *get_imsm_disk(struct intel_super *super, __u8 index)
{
return NULL;
}
+#endif
/* generate a checksum directly from the anchor when the anchor is known to be
* up-to-date, currently only at load or write_super after coalescing
return super->dev_tbl[index];
}
-static __u32 get_imsm_disk_idx(struct imsm_map *map, int slot)
+static __u32 get_imsm_ord_tbl_ent(struct imsm_dev *dev, int slot)
+{
+ struct imsm_map *map;
+
+ if (dev->vol.migr_state)
+ map = get_imsm_map(dev, 1);
+ else
+ map = get_imsm_map(dev, 0);
+
+ /* top byte identifies disk under rebuild */
+ return __le32_to_cpu(map->disk_ord_tbl[slot]);
+}
+
+#define ord_to_idx(ord) (((ord) << 8) >> 8)
+static __u32 get_imsm_disk_idx(struct imsm_dev *dev, int slot)
{
- __u32 *ord_tbl = &map->disk_ord_tbl[slot];
+ __u32 ord = get_imsm_ord_tbl_ent(dev, slot);
- /* top byte is 'special' */
- return __le32_to_cpu(*ord_tbl & ~(0xff << 24));
+ return ord_to_idx(ord);
+}
+
+static void set_imsm_ord_tbl_ent(struct imsm_map *map, int slot, __u32 ord)
+{
+ map->disk_ord_tbl[slot] = __cpu_to_le32(ord);
}
static int get_imsm_raid_level(struct imsm_map *map)
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);
struct imsm_map *map = get_imsm_map(dev, 0);
for (j = 0; j < map->num_members; j++) {
- __u32 index = get_imsm_disk_idx(map, j);
+ __u32 index = get_imsm_disk_idx(dev, j);
if (index == dl->index)
memberships++;
struct imsm_map *map = get_imsm_map(dev, 0);
for (j = 0; j < map->num_members; j++) {
- __u32 index = get_imsm_disk_idx(map, j);
+ __u32 index = get_imsm_disk_idx(dev, j);
if (index == dl->index) {
e->start = __le32_to_cpu(map->pba_of_lba0);
}
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 print_imsm_dev(struct imsm_dev *dev, int index)
{
__u64 sz;
int slot;
struct imsm_map *map = get_imsm_map(dev, 0);
+ __u32 ord;
printf("\n");
printf("[%s]:\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++)
- if (index == get_imsm_disk_idx(map, slot))
+ if (index == get_imsm_disk_idx(dev, slot))
break;
- if (slot < map->num_members)
- printf(" This Slot : %d\n", slot);
- else
+ if (slot < map->num_members) {
+ ord = get_imsm_ord_tbl_ent(dev, slot);
+ printf(" This Slot : %d%s\n", slot,
+ ord & IMSM_ORD_REBUILD ? " (out-of-sync)" : "");
+ } else
printf(" This Slot : ?\n");
sz = __le32_to_cpu(dev->size_high);
sz <<= 32;
printf(" Map State : %s", map_state_str[map->map_state]);
if (dev->vol.migr_state) {
struct imsm_map *map = get_imsm_map(dev, 1);
- printf(", %s", map_state_str[map->map_state]);
+ printf(" <-- %s", map_state_str[map->map_state]);
}
printf("\n");
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];
+ char str[MAX_RAID_SERIAL_LEN + 1];
__u32 s;
__u64 sz;
return;
printf("\n");
- snprintf(str, MAX_RAID_SERIAL_LEN, "%s", disk->serial);
+ snprintf(str, MAX_RAID_SERIAL_LEN + 1, "%s", disk->serial);
printf(" Disk%02d Serial : %s\n", index, str);
s = __le32_to_cpu(disk->status);
printf(" State :%s%s%s%s\n", s&SPARE_DISK ? " spare" : "",
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));
}
char str[MAX_SIGNATURE_LENGTH];
int i;
__u32 sum;
+ __u32 reserved = imsm_reserved_sectors(super, super->disks);
snprintf(str, MPB_SIG_LEN, "%s", mpb->sig);
printf(" Magic : %s\n", str);
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 getinfo_super_imsm(struct supertype *st, struct mdinfo *info);
+
static void brief_examine_super_imsm(struct supertype *st)
{
- printf("ARRAY /dev/imsm metadata=imsm\n");
+ /* We just write a generic DDF ARRAY entry
+ */
+ struct mdinfo info;
+ char nbuf[64];
+
+ getinfo_super_imsm(st, &info);
+ fname_from_uuid(st, &info, nbuf,'-');
+ printf("ARRAY /dev/imsm metadata=imsm UUID=%s\n", nbuf + 5);
}
static void detail_super_imsm(struct supertype *st, char *homehost)
static void brief_detail_super_imsm(struct supertype *st)
{
- printf("%s\n", __FUNCTION__);
+ struct mdinfo info;
+ char nbuf[64];
+ getinfo_super_imsm(st, &info);
+ fname_from_uuid(st, &info, nbuf,'-');
+ printf(" UUID=%s", nbuf + 5);
}
#endif
{
printf("%s\n", __FUNCTION__);
- return 0;
+ return -1;
}
static void uuid_from_super_imsm(struct supertype *st, int uuid[4])
{
- printf("%s\n", __FUNCTION__);
+ /* The uuid returned here is used for:
+ * uuid to put into bitmap file (Create, Grow)
+ * uuid for backup header when saving critical section (Grow)
+ * comparing uuids when re-adding a device into an array
+ * In these cases the uuid required is that of the data-array,
+ * not the device-set.
+ * uuid to recognise same set when adding a missing device back
+ * to an array. This is a uuid for the device-set.
+ *
+ * For each of these we can make do with a truncated
+ * or hashed uuid rather than the original, as long as
+ * everyone agrees.
+ * In each case the uuid required is that of the data-array,
+ * not the device-set.
+ */
+ /* imsm does not track uuid's so we synthesis one using sha1 on
+ * - The signature (Which is constant for all imsm array, but no matter)
+ * - the family_num of the container
+ * - the index number of the volume
+ * - the 'serial' number of the volume.
+ * Hopefully these are all constant.
+ */
+ struct intel_super *super = st->sb;
+
+ char buf[20];
+ struct sha1_ctx ctx;
+ struct imsm_dev *dev = NULL;
+
+ sha1_init_ctx(&ctx);
+ sha1_process_bytes(super->anchor->sig, MAX_SIGNATURE_LENGTH, &ctx);
+ sha1_process_bytes(&super->anchor->family_num, sizeof(__u32), &ctx);
+ if (super->current_vol >= 0)
+ dev = get_imsm_dev(super, super->current_vol);
+ if (dev) {
+ __u32 vol = super->current_vol;
+ sha1_process_bytes(&vol, sizeof(vol), &ctx);
+ sha1_process_bytes(dev->volume, MAX_RAID_SERIAL_LEN, &ctx);
+ }
+ sha1_finish_ctx(&ctx, buf);
+ memcpy(uuid, buf, 4*4);
}
#if 0
case 6:
return ALGORITHM_LEFT_ASYMMETRIC;
case 10:
- return 0x102; //FIXME is this correct?
+ return 0x102;
}
return -1;
}
info->array.md_minor = -1;
info->array.ctime = 0;
info->array.utime = 0;
- info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip * 512);
+ info->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
+ info->array.state = !dev->vol.dirty;
+
+ info->disk.major = 0;
+ info->disk.minor = 0;
info->data_offset = __le32_to_cpu(map->pba_of_lba0);
info->component_size = __le32_to_cpu(map->blocks_per_member);
+ memset(info->uuid, 0, sizeof(info->uuid));
- info->disk.major = 0;
- info->disk.minor = 0;
+ 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;
+ strncpy(info->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
+ info->name[MAX_RAID_SERIAL_LEN] = 0;
+
+ info->array.major_version = -1;
+ info->array.minor_version = -2;
sprintf(info->text_version, "/%s/%d",
devnum2devname(st->container_dev),
info->container_member);
+ info->safe_mode_delay = 4000; /* 4 secs like the Matrix driver */
+ uuid_from_super_imsm(st, info->uuid);
}
info->disk.minor = 0;
info->disk.raid_disk = -1;
info->reshape_active = 0;
+ info->array.major_version = -1;
+ info->array.minor_version = -2;
strcpy(info->text_version, "imsm");
+ info->safe_mode_delay = 0;
info->disk.number = -1;
info->disk.state = 0;
+ 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;
}
+ uuid_from_super_imsm(st, info->uuid);
}
static int update_super_imsm(struct supertype *st, struct mdinfo *info,
sec->anchor->num_raid_devs > 0) {
if (first->anchor->family_num != sec->anchor->family_num)
return 3;
- if (first->anchor->mpb_size != sec->anchor->mpb_size)
- return 3;
- if (first->anchor->check_sum != sec->anchor->check_sum)
- return 3;
+ }
+
+ /* if 'first' is a spare promote it to a populated mpb with sec's
+ * family number
+ */
+ if (first->anchor->num_raid_devs == 0 &&
+ sec->anchor->num_raid_devs > 0) {
+ first->anchor->num_raid_devs = sec->anchor->num_raid_devs;
+ first->anchor->family_num = sec->anchor->family_num;
}
return 0;
unsigned char scsi_serial[255];
int rv;
int rsp_len;
- int i, cnt;
+ int len;
+ char *c, *rsp_buf;
memset(scsi_serial, 0, sizeof(scsi_serial));
- if (imsm_env_devname_as_serial()) {
- char name[MAX_RAID_SERIAL_LEN];
-
- fd2devname(fd, name);
- strcpy((char *) serial, name);
+ rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial));
+
+ if (rv && imsm_env_devname_as_serial()) {
+ memset(serial, 0, MAX_RAID_SERIAL_LEN);
+ fd2devname(fd, (char *) serial);
return 0;
}
- rv = scsi_get_serial(fd, scsi_serial, sizeof(scsi_serial));
-
if (rv != 0) {
if (devname)
fprintf(stderr,
return rv;
}
+ /* trim leading whitespace */
rsp_len = scsi_serial[3];
- for (i = 0, cnt = 0; i < rsp_len; i++) {
- if (!isspace(scsi_serial[4 + i]))
- serial[cnt++] = scsi_serial[4 + i];
- if (cnt == MAX_RAID_SERIAL_LEN)
- break;
- }
+ 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);
- serial[MAX_RAID_SERIAL_LEN - 1] = '\0';
+ /* trim trailing whitespace starting with the last character copied */
+ c = (char *) &serial[len - 1];
+ while (isspace(*c) || *c == '\0')
+ *c-- = '\0';
return 0;
}
+static int serialcmp(__u8 *s1, __u8 *s2)
+{
+ return strncmp((char *) s1, (char *) s2, MAX_RAID_SERIAL_LEN);
+}
+
+static void serialcpy(__u8 *dest, __u8 *src)
+{
+ strncpy((char *) dest, (char *) src, MAX_RAID_SERIAL_LEN);
+}
+
static int
load_imsm_disk(int fd, struct intel_super *super, char *devname, int keep_fd)
{
* check if we need to update dl->index
*/
for (dl = super->disks; dl; dl = dl->next)
- if (memcmp(dl->serial, serial, MAX_RAID_SERIAL_LEN) == 0)
+ if (serialcmp(dl->serial, serial) == 0)
break;
if (!dl)
dl->next = super->disks;
dl->fd = keep_fd ? fd : -1;
dl->devname = devname ? strdup(devname) : NULL;
- strncpy((char *) dl->serial, (char *) serial, MAX_RAID_SERIAL_LEN);
+ serialcpy(dl->serial, serial);
+ dl->index = -2;
} else if (keep_fd) {
close(dl->fd);
dl->fd = fd;
disk_iter = __get_imsm_disk(super->anchor, i);
- if (memcmp(disk_iter->serial, dl->serial,
- MAX_RAID_SERIAL_LEN) == 0) {
+ if (serialcmp(disk_iter->serial, dl->serial) == 0) {
__u32 status;
dl->disk = *disk_iter;
/* only set index on disks that are a member of a
* populated contianer, i.e. one with raid_devs
*/
- if (status & SPARE_DISK)
+ if (status & FAILED_DISK)
+ dl->index = -2;
+ else if (status & SPARE_DISK)
dl->index = -1;
else
dl->index = i;
+
break;
}
}
- if (i == super->anchor->num_disks && alloc) {
- if (devname)
- fprintf(stderr,
- Name ": failed to load disk with serial \'%s\' for %s\n",
- dl->serial, devname);
- free(dl);
- return 1;
- }
+ /* no match, maybe a stale failed drive */
if (i == super->anchor->num_disks && dl->index >= 0) {
- if (devname)
- fprintf(stderr,
- Name ": confused... disk %d with serial \'%s\' "
- "is not listed in the current anchor\n",
- dl->index, dl->serial);
- return 1;
+ 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;
-
+
return 0;
}
memcpy(dest, src, sizeof_imsm_dev(src, 0));
}
-static void dup_map(struct imsm_dev *dev)
+#ifndef MDASSEMBLE
+/* When migrating map0 contains the 'destination' state while map1
+ * contains the current state. When not migrating map0 contains the
+ * current state. This routine assumes that map[0].map_state is set to
+ * the current array state before being called.
+ *
+ * Migration is indicated by one of the following states
+ * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
+ * 2/ Initialize (migr_state=1 migr_type=0 map0state=normal
+ * map1state=unitialized)
+ * 3/ Verify (Resync) (migr_state=1 migr_type=1 map0state=normal
+ * map1state=normal)
+ * 4/ Rebuild (migr_state=1 migr_type=1 map0state=normal
+ * map1state=degraded)
+ */
+static void migrate(struct imsm_dev *dev, __u8 to_state, int rebuild_resync)
{
- struct imsm_map *dest = get_imsm_map(dev, 1);
+ struct imsm_map *dest;
struct imsm_map *src = get_imsm_map(dev, 0);
+ 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)
{
int i;
struct imsm_dev *dev_new;
- size_t len;
+ size_t len, len_migr;
+ size_t space_needed = 0;
+ struct imsm_super *mpb = super->anchor;
for (i = 0; i < super->anchor->num_raid_devs; i++) {
struct imsm_dev *dev_iter = __get_imsm_dev(super->anchor, i);
- len = sizeof_imsm_dev(dev_iter, 1);
- dev_new = malloc(len);
+ len = sizeof_imsm_dev(dev_iter, 0);
+ len_migr = sizeof_imsm_dev(dev_iter, 1);
+ if (len_migr > len)
+ space_needed += len_migr - len;
+
+ dev_new = malloc(len_migr);
if (!dev_new)
return 1;
imsm_copy_dev(dev_new, dev_iter);
super->dev_tbl[i] = dev_new;
}
+ /* ensure that super->buf is large enough when all raid devices
+ * are migrating
+ */
+ if (__le32_to_cpu(mpb->mpb_size) + space_needed > super->len) {
+ void *buf;
+
+ len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + space_needed, 512);
+ if (posix_memalign(&buf, 512, len) != 0)
+ return 1;
+
+ memcpy(buf, super->buf, len);
+ free(super->buf);
+ super->buf = buf;
+ super->len = len;
+ }
+
return 0;
}
}
__free_imsm(super, 0);
- super->len = __le32_to_cpu(anchor->mpb_size);
super->len = ROUND_UP(anchor->mpb_size, 512);
if (posix_memalign(&super->buf, 512, super->len) != 0) {
if (devname)
rc = load_imsm_disk(fd, super, devname, 0);
if (rc == 0)
rc = parse_raid_devices(super);
+
return rc;
}
+static void __free_imsm_disk(struct dl *d)
+{
+ if (d->fd >= 0)
+ close(d->fd);
+ if (d->devname)
+ free(d->devname);
+ free(d);
+
+}
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;
- if (d->fd >= 0)
- close(d->fd);
- if (d->devname)
- free(d->devname);
- free(d);
+ __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);
}
*sbp = super;
+ st->container_dev = fd2devnum(fd);
if (st->ss == NULL) {
st->ss = &super_imsm;
st->minor_version = 0;
st->max_devs = IMSM_MAX_DEVICES;
- st->container_dev = fd2devnum(fd);
}
+ st->loaded_container = 1;
return 0;
}
st->minor_version = 0;
st->max_devs = IMSM_MAX_DEVICES;
}
+ st->loaded_container = 0;
return 0;
}
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);
map->num_members = info->raid_disks;
for (i = 0; i < map->num_members; i++) {
/* initialized in add_to_super */
- map->disk_ord_tbl[i] = __cpu_to_le32(0);
+ set_imsm_ord_tbl_ent(map, i, 0);
}
mpb->num_raid_devs++;
super->dev_tbl[super->current_vol] = dev;
return 1;
}
+#ifndef MDASSEMBLE
static void add_to_super_imsm_volume(struct supertype *st, mdu_disk_info_t *dk,
int fd, char *devname)
{
dl->index = super->anchor->num_disks;
super->anchor->num_disks++;
}
- map->disk_ord_tbl[dk->number] = __cpu_to_le32(dl->index);
+ set_imsm_ord_tbl_ent(map, dk->number, dl->index);
status = CONFIGURED_DISK | USABLE_DISK;
dl->disk.status = __cpu_to_le32(status);
dd->minor = minor(stb.st_rdev);
dd->index = -1;
dd->devname = devname ? strdup(devname) : NULL;
- dd->next = super->disks;
dd->fd = fd;
rv = imsm_read_serial(fd, devname, dd->serial);
if (rv) {
get_dev_size(fd, NULL, &size);
size /= 512;
status = USABLE_DISK | SPARE_DISK;
- strcpy((char *) dd->disk.serial, (char *) dd->serial);
+ serialcpy(dd->disk.serial, dd->serial);
dd->disk.total_blocks = __cpu_to_le32(size);
dd->disk.status = __cpu_to_le32(status);
if (sysfs_disk_to_scsi_id(fd, &id) == 0)
dd->disk.scsi_id = __cpu_to_le32(id);
else
dd->disk.scsi_id = __cpu_to_le32(0);
- super->disks = dd;
+
+ if (st->update_tail) {
+ dd->next = super->add;
+ super->add = dd;
+ } else {
+ dd->next = super->disks;
+ super->disks = dd;
+ }
}
static int store_imsm_mpb(int fd, struct intel_super *super);
mpb->generation_num = __cpu_to_le32(1UL);
for (d = super->disks; d; d = d->next) {
- if (d->index >= 0)
+ if (d->index != -1)
continue;
mpb->disk[0] = d->disk;
fprintf(stderr, "%s: failed for device %d:%d %s\n",
__func__, d->major, d->minor, strerror(errno));
*mpb = mpb_save;
- return 0;
+ return 1;
}
if (doclose) {
close(d->fd);
}
*mpb = mpb_save;
- return 1;
+ return 0;
}
static int write_super_imsm(struct intel_super *super, int doclose)
__u32 generation;
__u32 sum;
int spares = 0;
- int raid_disks = 0;
int i;
+ __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
/* 'generation' is incremented everytime the metadata is written */
generation = __le32_to_cpu(mpb->generation_num);
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 < 0)
+ if (d->index == -1)
spares++;
- else {
- raid_disks++;
+ else
mpb->disk[d->index] = d->disk;
- }
- }
- if (raid_disks != mpb->num_disks) {
- fprintf(stderr, "%s: expected %d disks only found %d\n",
- __func__, mpb->num_disks, raid_disks);
- return 0;
}
+ 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);
imsm_copy_dev(dev, super->dev_tbl[i]);
+ mpb_size += sizeof_imsm_dev(dev, 0);
}
+ mpb_size += __le32_to_cpu(mpb->bbm_log_size);
+ mpb->mpb_size = __cpu_to_le32(mpb_size);
/* recalculate checksum */
sum = __gen_imsm_checksum(mpb);
for (d = super->disks; d ; d = d->next) {
if (d->index < 0)
continue;
- if (store_imsm_mpb(d->fd, super)) {
+ if (store_imsm_mpb(d->fd, super))
fprintf(stderr, "%s: failed for device %d:%d %s\n",
__func__, d->major, d->minor, strerror(errno));
- return 0;
- }
if (doclose) {
close(d->fd);
d->fd = -1;
if (spares)
return write_super_imsm_spares(super, doclose);
- return 1;
+ return 0;
+}
+
+
+static int create_array(struct supertype *st)
+{
+ size_t len;
+ struct imsm_update_create_array *u;
+ struct intel_super *super = st->sb;
+ struct imsm_dev *dev = get_imsm_dev(super, super->current_vol);
+
+ len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0);
+ u = malloc(len);
+ if (!u) {
+ fprintf(stderr, "%s: failed to allocate update buffer\n",
+ __func__);
+ return 1;
+ }
+
+ u->type = update_create_array;
+ u->dev_idx = super->current_vol;
+ imsm_copy_dev(&u->dev, dev);
+ append_metadata_update(st, u, len);
+
+ return 0;
+}
+
+static int _add_disk(struct supertype *st)
+{
+ struct intel_super *super = st->sb;
+ size_t len;
+ struct imsm_update_add_disk *u;
+
+ if (!super->add)
+ return 0;
+
+ len = sizeof(*u);
+ u = malloc(len);
+ if (!u) {
+ fprintf(stderr, "%s: failed to allocate update buffer\n",
+ __func__);
+ return 1;
+ }
+
+ u->type = update_add_disk;
+ append_metadata_update(st, u, len);
+
+ return 0;
}
static int write_init_super_imsm(struct supertype *st)
{
if (st->update_tail) {
- /* queue the recently created array as a metadata update */
- size_t len;
- struct imsm_update_create_array *u;
+ /* queue the recently created array / added disk
+ * as a metadata update */
struct intel_super *super = st->sb;
- struct imsm_dev *dev;
struct dl *d;
+ int rv;
- if (super->current_vol < 0 ||
- !(dev = get_imsm_dev(super, super->current_vol))) {
- fprintf(stderr, "%s: could not determine sub-array\n",
- __func__);
- return 1;
- }
-
-
- len = sizeof(*u) - sizeof(*dev) + sizeof_imsm_dev(dev, 0);
- u = malloc(len);
- if (!u) {
- fprintf(stderr, "%s: failed to allocate update buffer\n",
- __func__);
- return 1;
- }
-
- u->type = update_create_array;
- u->dev_idx = super->current_vol;
- imsm_copy_dev(&u->dev, dev);
- append_metadata_update(st, u, len);
+ /* determine if we are creating a volume or adding a disk */
+ if (super->current_vol < 0) {
+ /* in the add disk case we are running in mdmon
+ * context, so don't close fd's
+ */
+ return _add_disk(st);
+ } else
+ rv = create_array(st);
for (d = super->disks; d ; d = d->next) {
close(d->fd);
d->fd = -1;
}
- return 0;
+ return rv;
} else
return write_super_imsm(st->sb, 1);
}
+#endif
static int store_zero_imsm(struct supertype *st, int fd)
{
return 0;
}
+static int imsm_bbm_log_size(struct imsm_super *mpb)
+{
+ return __le32_to_cpu(mpb->bbm_log_size);
+}
+
+#ifndef MDASSEMBLE
static int validate_geometry_imsm_container(struct supertype *st, int level,
int layout, int raiddisks, int chunk,
unsigned long long size, char *dev,
return 1;
}
-int imsm_bbm_log_size(struct imsm_super *mpb)
-{
- return __le32_to_cpu(mpb->bbm_log_size);
-}
-
static int validate_geometry_imsm(struct supertype *st, int level, int layout,
int raiddisks, int chunk, unsigned long long size,
char *dev, unsigned long long *freesize,
return 1;
}
+#endif /* MDASSEMBLE */
static struct mdinfo *container_content_imsm(struct supertype *st)
{
for (i = 0; i < mpb->num_raid_devs; i++) {
struct imsm_dev *dev = get_imsm_dev(super, i);
- struct imsm_vol *vol = &dev->vol;
struct imsm_map *map = get_imsm_map(dev, 0);
struct mdinfo *this;
int slot;
this = malloc(sizeof(*this));
memset(this, 0, sizeof(*this));
this->next = rest;
- rest = this;
-
- this->array.level = get_imsm_raid_level(map);
- this->array.raid_disks = map->num_members;
- this->array.layout = imsm_level_to_layout(this->array.level);
- this->array.md_minor = -1;
- this->array.ctime = 0;
- this->array.utime = 0;
- this->array.chunk_size = __le16_to_cpu(map->blocks_per_strip) << 9;
- this->array.state = !vol->dirty;
- this->container_member = i;
- if (map->map_state == IMSM_T_STATE_UNINITIALIZED || dev->vol.dirty)
- this->resync_start = 0;
- else
- this->resync_start = ~0ULL;
-
- strncpy(this->name, (char *) dev->volume, MAX_RAID_SERIAL_LEN);
- this->name[MAX_RAID_SERIAL_LEN] = 0;
-
- sprintf(this->text_version, "/%s/%d",
- devnum2devname(st->container_dev),
- this->container_member);
-
- memset(this->uuid, 0, sizeof(this->uuid));
-
- this->component_size = __le32_to_cpu(map->blocks_per_member);
+ super->current_vol = i;
+ getinfo_super_imsm_volume(st, this);
for (slot = 0 ; slot < map->num_members; slot++) {
struct mdinfo *info_d;
struct dl *d;
int idx;
+ int skip;
__u32 s;
+ __u32 ord;
- idx = get_imsm_disk_idx(map, slot);
+ skip = 0;
+ idx = get_imsm_disk_idx(dev, slot);
+ ord = get_imsm_ord_tbl_ent(dev, slot);
for (d = super->disks; d ; d = d->next)
if (d->index == idx)
break;
if (d == NULL)
- break; /* shouldn't this be continue ?? */
+ skip = 1;
+
+ s = d ? __le32_to_cpu(d->disk.status) : 0;
+ if (s & FAILED_DISK)
+ skip = 1;
+ if (!(s & USABLE_DISK))
+ skip = 1;
+ if (ord & IMSM_ORD_REBUILD)
+ skip = 1;
+
+ /*
+ * if we skip some disks the array will be assmebled degraded;
+ * reset resync start to avoid a dirty-degraded situation
+ *
+ * FIXME handle dirty degraded
+ */
+ if (skip && !dev->vol.dirty)
+ this->resync_start = ~0ULL;
+ if (skip)
+ continue;
info_d = malloc(sizeof(*info_d));
- if (!info_d)
- break; /* ditto ?? */
+ if (!info_d) {
+ fprintf(stderr, Name ": failed to allocate disk"
+ " for volume %s\n", (char *) dev->volume);
+ free(this);
+ this = rest;
+ break;
+ }
memset(info_d, 0, sizeof(*info_d));
info_d->next = this->devs;
this->devs = info_d;
- s = __le32_to_cpu(d->disk.status);
-
info_d->disk.number = d->index;
info_d->disk.major = d->major;
info_d->disk.minor = d->minor;
info_d->disk.raid_disk = slot;
- info_d->disk.state = s & CONFIGURED_DISK ? (1 << MD_DISK_ACTIVE) : 0;
- info_d->disk.state |= s & FAILED_DISK ? (1 << MD_DISK_FAULTY) : 0;
- info_d->disk.state |= s & USABLE_DISK ? (1 << MD_DISK_SYNC) : 0;
this->array.working_disks++;
if (d->devname)
strcpy(info_d->name, d->devname);
}
+ rest = this;
}
return rest;
}
+#ifndef MDASSEMBLE
static int imsm_open_new(struct supertype *c, struct active_array *a,
char *inst)
{
return 0;
}
-static __u8 imsm_check_degraded(struct intel_super *super, int n, int failed)
+static __u8 imsm_check_degraded(struct intel_super *super, struct imsm_dev *dev, int failed)
{
- struct imsm_dev *dev = get_imsm_dev(super, n);
struct imsm_map *map = get_imsm_map(dev, 0);
if (!failed)
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 failed = 0;
int i;
+ int insync;
for (i = 0; i < map->num_members; i++) {
- int idx = get_imsm_disk_idx(map, 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 (__le32_to_cpu(disk->status) & FAILED_DISK)
- failed++;
+ /* 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 (failed >= 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 'failed' for next mirror set */
- if (!((i + 1) % device_per_mirror))
- failed = 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 map->map_state;
}
-static int imsm_count_failed(struct intel_super *super, struct imsm_map *map)
+static int imsm_count_failed(struct intel_super *super, struct imsm_dev *dev)
{
int i;
int failed = 0;
struct imsm_disk *disk;
+ struct imsm_map *map = get_imsm_map(dev, 0);
for (i = 0; i < map->num_members; i++) {
- int idx = get_imsm_disk_idx(map, i);
+ __u32 ord = get_imsm_ord_tbl_ent(dev, i);
+ int idx = ord_to_idx(ord);
disk = get_imsm_disk(super, idx);
- if (__le32_to_cpu(disk->status) & FAILED_DISK)
+ if (!disk ||
+ __le32_to_cpu(disk->status) & FAILED_DISK ||
+ ord & IMSM_ORD_REBUILD)
failed++;
}
return failed;
}
-static void imsm_set_array_state(struct active_array *a, int consistent)
+static int is_resyncing(struct imsm_dev *dev)
+{
+ struct imsm_map *migr_map;
+
+ if (!dev->vol.migr_state)
+ return 0;
+
+ if (dev->vol.migr_type == 0)
+ return 1;
+
+ migr_map = get_imsm_map(dev, 1);
+
+ if (migr_map->map_state == IMSM_T_STATE_NORMAL)
+ return 1;
+ else
+ return 0;
+}
+
+static int is_rebuilding(struct imsm_dev *dev)
+{
+ struct imsm_map *migr_map;
+
+ if (!dev->vol.migr_state)
+ return 0;
+
+ if (dev->vol.migr_type == 0)
+ return 0;
+
+ migr_map = get_imsm_map(dev, 1);
+
+ if (migr_map->map_state == IMSM_T_STATE_DEGRADED)
+ return 1;
+ else
+ 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
+ * 'degraded' state for the array.
+ */
+static int imsm_set_array_state(struct active_array *a, int consistent)
{
int inst = a->info.container_member;
struct intel_super *super = a->container->sb;
struct imsm_dev *dev = get_imsm_dev(super, inst);
struct imsm_map *map = get_imsm_map(dev, 0);
- int dirty = !consistent;
- int failed;
- __u8 map_state;
+ int failed = imsm_count_failed(super, dev);
+ __u8 map_state = imsm_check_degraded(super, dev, failed);
- failed = imsm_count_failed(super, map);
- map_state = imsm_check_degraded(super, inst, failed);
+ /* before we activate this array handle any missing disks */
+ if (consistent == 2 && super->missing) {
+ struct dl *dl;
- if (a->resync_start == ~0ULL) {
- /* complete recovery or initial resync */
- if (map->map_state != map_state) {
- dprintf("imsm: map_state %d: %d\n",
- inst, map_state);
- map->map_state = map_state;
- super->updates_pending++;
- }
- if (dev->vol.migr_state) {
- dprintf("imsm: mark resync complete\n");
- dev->vol.migr_state = 0;
- dev->vol.migr_type = 0;
+ 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 &&
+ (!is_resync_complete(a) ||
+ map_state != IMSM_T_STATE_NORMAL ||
+ dev->vol.migr_state))
+ consistent = 0;
+
+ if (is_resync_complete(a)) {
+ /* complete intialization / resync,
+ * recovery is completed in ->set_disk
+ */
+ if (is_resyncing(dev)) {
+ dprintf("imsm: mark resync done\n");
+ end_migration(dev, map_state);
super->updates_pending++;
}
- } else if (!dev->vol.migr_state) {
- dprintf("imsm: mark '%s' (%llu)\n",
- failed ? "rebuild" : "initializing", a->resync_start);
- /* mark that we are rebuilding */
- map->map_state = failed ? map_state : IMSM_T_STATE_NORMAL;
- dev->vol.migr_state = 1;
- dev->vol.migr_type = failed ? 1 : 0;
- dup_map(dev);
+ } else if (!is_resyncing(dev) && !failed) {
+ /* mark the start of the init process if nothing is failed */
+ dprintf("imsm: mark resync start (%llu)\n", a->resync_start);
+ map->map_state = map_state;
+ migrate(dev, IMSM_T_STATE_NORMAL,
+ map->map_state == IMSM_T_STATE_NORMAL);
+ 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 (dirty != dev->vol.dirty) {
+ if (dev->vol.dirty != !consistent) {
dprintf("imsm: mark '%s' (%llu)\n",
- dirty ? "dirty" : "clean", a->resync_start);
- dev->vol.dirty = dirty;
+ consistent ? "clean" : "dirty", a->resync_start);
+ if (consistent)
+ dev->vol.dirty = 0;
+ else
+ dev->vol.dirty = 1;
super->updates_pending++;
}
+ return consistent;
}
static void imsm_set_disk(struct active_array *a, int n, int state)
struct imsm_dev *dev = get_imsm_dev(super, inst);
struct imsm_map *map = get_imsm_map(dev, 0);
struct imsm_disk *disk;
+ int failed;
__u32 status;
- int failed = 0;
- int new_failure = 0;
+ __u32 ord;
+ __u8 map_state;
if (n > map->num_members)
fprintf(stderr, "imsm: set_disk %d out of range 0..%d\n",
dprintf("imsm: set_disk %d:%x\n", n, state);
- disk = get_imsm_disk(super, get_imsm_disk_idx(map, n));
+ ord = get_imsm_ord_tbl_ent(dev, n);
+ disk = get_imsm_disk(super, ord_to_idx(ord));
/* 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);
- new_failure = 1;
+ mark_failure(disk);
super->updates_pending++;
}
+
/* check if in_sync */
- if ((state & DS_INSYNC) && !(status & USABLE_DISK)) {
- status |= USABLE_DISK;
- disk->status = __cpu_to_le32(status);
+ if (state & DS_INSYNC && ord & IMSM_ORD_REBUILD) {
+ struct imsm_map *migr_map = get_imsm_map(dev, 1);
+
+ set_imsm_ord_tbl_ent(migr_map, n, ord_to_idx(ord));
super->updates_pending++;
}
- /* the number of failures have changed, count up 'failed' to determine
- * degraded / failed status
- */
- if (new_failure && map->map_state != IMSM_T_STATE_FAILED)
- failed = imsm_count_failed(super, map);
-
- /* determine map_state based on failed or in_sync count */
- if (failed)
- map->map_state = imsm_check_degraded(super, inst, failed);
- else if (map->map_state == IMSM_T_STATE_DEGRADED) {
- struct mdinfo *d;
- int working = 0;
-
- for (d = a->info.devs ; d ; d = d->next)
- if (d->curr_state & DS_INSYNC)
- working++;
+ failed = imsm_count_failed(super, dev);
+ map_state = imsm_check_degraded(super, dev, failed);
- if (working == a->info.array.raid_disks) {
- map->map_state = IMSM_T_STATE_NORMAL;
- dev->vol.migr_state = 0;
- dev->vol.migr_type = 0;
- super->updates_pending++;
- }
+ /* check if recovery complete, newly degraded, or failed */
+ if (map_state == IMSM_T_STATE_NORMAL && is_rebuilding(dev)) {
+ end_migration(dev, map_state);
+ super->updates_pending++;
+ } else if (map_state == IMSM_T_STATE_DEGRADED &&
+ map->map_state != map_state &&
+ !dev->vol.migr_state) {
+ dprintf("imsm: mark degraded\n");
+ map->map_state = map_state;
+ super->updates_pending++;
+ } else if (map_state == IMSM_T_STATE_FAILED &&
+ map->map_state != map_state) {
+ dprintf("imsm: mark failed\n");
+ end_migration(dev, map_state);
+ super->updates_pending++;
}
}
super->updates_pending = 0;
}
+static struct dl *imsm_readd(struct intel_super *super, int idx, struct active_array *a)
+{
+ struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
+ int i = get_imsm_disk_idx(dev, idx);
+ struct dl *dl;
+
+ for (dl = super->disks; dl; dl = dl->next)
+ if (dl->index == i)
+ break;
+
+ if (dl && __le32_to_cpu(dl->disk.status) & FAILED_DISK)
+ dl = NULL;
+
+ if (dl)
+ dprintf("%s: found %x:%x\n", __func__, dl->major, dl->minor);
+
+ return dl;
+}
+
+static struct dl *imsm_add_spare(struct intel_super *super, int slot, struct active_array *a)
+{
+ struct imsm_dev *dev = get_imsm_dev(super, a->info.container_member);
+ int idx = get_imsm_disk_idx(dev, slot);
+ struct imsm_map *map = get_imsm_map(dev, 0);
+ unsigned long long esize;
+ unsigned long long pos;
+ struct mdinfo *d;
+ struct extent *ex;
+ int j;
+ int found;
+ __u32 array_start;
+ __u32 status;
+ struct dl *dl;
+
+ for (dl = super->disks; dl; dl = dl->next) {
+ /* If in this array, skip */
+ for (d = a->info.devs ; d ; d = d->next)
+ if (d->state_fd >= 0 &&
+ d->disk.major == dl->major &&
+ d->disk.minor == dl->minor) {
+ dprintf("%x:%x already in array\n", dl->major, dl->minor);
+ break;
+ }
+ if (d)
+ continue;
+
+ /* skip in use or failed drives */
+ status = __le32_to_cpu(dl->disk.status);
+ if (status & FAILED_DISK || idx == dl->index) {
+ dprintf("%x:%x status ( %s%s)\n",
+ dl->major, dl->minor,
+ status & FAILED_DISK ? "failed " : "",
+ idx == dl->index ? "in use " : "");
+ continue;
+ }
+
+ /* Does this unused device have the requisite free space?
+ * We need a->info.component_size sectors
+ */
+ ex = get_extents(super, dl);
+ if (!ex) {
+ dprintf("cannot get extents\n");
+ continue;
+ }
+ found = 0;
+ j = 0;
+ pos = 0;
+ array_start = __le32_to_cpu(map->pba_of_lba0);
+
+ do {
+ /* check that we can start at pba_of_lba0 with
+ * a->info.component_size of space
+ */
+ esize = ex[j].start - pos;
+ if (array_start >= pos &&
+ array_start + a->info.component_size < ex[j].start) {
+ found = 1;
+ break;
+ }
+ pos = ex[j].start + ex[j].size;
+ j++;
+
+ } while (ex[j-1].size);
+
+ free(ex);
+ if (!found) {
+ dprintf("%x:%x does not have %llu at %d\n",
+ dl->major, dl->minor,
+ a->info.component_size,
+ __le32_to_cpu(map->pba_of_lba0));
+ /* No room */
+ continue;
+ } else
+ break;
+ }
+
+ return dl;
+}
+
static struct mdinfo *imsm_activate_spare(struct active_array *a,
struct metadata_update **updates)
{
dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
inst, failed, a->info.array.raid_disks, a->info.array.level);
- if (imsm_check_degraded(super, inst, failed) != IMSM_T_STATE_DEGRADED)
+ if (imsm_check_degraded(super, dev, failed) != IMSM_T_STATE_DEGRADED)
return NULL;
/* For each slot, if it is not working, find a spare */
- dl = super->disks;
for (i = 0; i < a->info.array.raid_disks; i++) {
for (d = a->info.devs ; d ; d = d->next)
if (d->disk.raid_disk == i)
if (d && (d->state_fd >= 0))
continue;
- /* OK, this device needs recovery. Find a spare */
- for ( ; dl ; dl = dl->next) {
- unsigned long long esize;
- unsigned long long pos;
- struct mdinfo *d2;
- struct extent *ex;
- int j;
- int found;
- __u32 array_start;
-
- /* If in this array, skip */
- for (d2 = a->info.devs ; d2 ; d2 = d2->next)
- if (d2->disk.major == dl->major &&
- d2->disk.minor == dl->minor) {
- dprintf("%x:%x already in array\n", dl->major, dl->minor);
- break;
- }
- if (d2)
- continue;
-
- /* Does this unused device have the requisite free space?
- * We need a->info.component_size sectors
- */
- ex = get_extents(super, dl);
- if (!ex) {
- dprintf("cannot get extents\n");
- continue;
- }
- found = 0;
- j = 0;
- pos = 0;
- array_start = __le32_to_cpu(map->pba_of_lba0);
-
- do {
- /* check that we can start at pba_of_lba0 with
- * a->info.component_size of space
- */
- esize = ex[j].start - pos;
- if (array_start >= pos &&
- array_start + a->info.component_size < ex[j].start) {
- found = 1;
- break;
- }
- pos = ex[j].start + ex[j].size;
- j++;
-
- } while (ex[j-1].size);
-
- free(ex);
- if (!found) {
- dprintf("%x:%x does not have %llu at %d\n",
- dl->major, dl->minor,
- a->info.component_size,
- __le32_to_cpu(map->pba_of_lba0));
- /* No room */
- continue;
- }
-
- /* found a usable disk with enough space */
- di = malloc(sizeof(*di));
- memset(di, 0, sizeof(*di));
+ /*
+ * OK, this device needs recovery. Try to re-add the previous
+ * occupant of this slot, if this fails add a new spare
+ */
+ dl = imsm_readd(super, i, a);
+ if (!dl)
+ dl = imsm_add_spare(super, i, a);
+ if (!dl)
+ continue;
+
+ /* 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
- * pristine spare. imsm_process_update() will create a
- * new index in this case. Once a disk is found to be
- * failed in all member arrays it is kicked from the
- * metadata
- */
- di->disk.number = dl->index;
+ /* dl->index will be -1 in the case we are activating a
+ * pristine spare. imsm_process_update() will create a
+ * new index in this case. Once a disk is found to be
+ * failed in all member arrays it is kicked from the
+ * metadata
+ */
+ di->disk.number = dl->index;
- /* (ab)use di->devs to store a pointer to the device
- * we chose
- */
- di->devs = (struct mdinfo *) dl;
-
- di->disk.raid_disk = i;
- di->disk.major = dl->major;
- di->disk.minor = dl->minor;
- di->disk.state = 0;
- di->data_offset = array_start;
- di->component_size = a->info.component_size;
- di->container_member = inst;
- di->next = rv;
- rv = di;
- num_spares++;
- dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
- i, pos);
+ /* (ab)use di->devs to store a pointer to the device
+ * we chose
+ */
+ di->devs = (struct mdinfo *) dl;
+
+ di->disk.raid_disk = i;
+ di->disk.major = dl->major;
+ di->disk.minor = dl->minor;
+ di->disk.state = 0;
+ di->data_offset = __le32_to_cpu(map->pba_of_lba0);
+ di->component_size = a->info.component_size;
+ di->container_member = inst;
+ di->next = rv;
+ rv = di;
+ num_spares++;
+ dprintf("%x:%x to be %d at %llu\n", dl->major, dl->minor,
+ i, di->data_offset);
- break;
- }
+ break;
}
if (!rv)
* 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;
return rv;
}
-static int disks_overlap(struct imsm_map *m1, struct imsm_map *m2)
+static int disks_overlap(struct imsm_dev *d1, struct imsm_dev *d2)
{
+ struct imsm_map *m1 = get_imsm_map(d1, 0);
+ struct imsm_map *m2 = get_imsm_map(d2, 0);
int i;
int j;
int idx;
for (i = 0; i < m1->num_members; i++) {
- idx = get_imsm_disk_idx(m1, i);
+ idx = get_imsm_disk_idx(d1, i);
for (j = 0; j < m2->num_members; j++)
- if (idx == get_imsm_disk_idx(m2, j))
+ if (idx == get_imsm_disk_idx(d2, j))
return 1;
}
return 0;
}
+static void imsm_delete(struct intel_super *super, struct dl **dlp, int index);
+
static void imsm_process_update(struct supertype *st,
struct metadata_update *update)
{
* flag
*/
struct intel_super *super = st->sb;
- struct imsm_super *mpb = super->anchor;
+ struct imsm_super *mpb;
enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
+ /* update requires a larger buf but the allocation failed */
+ if (super->next_len && !super->next_buf) {
+ super->next_len = 0;
+ return;
+ }
+
+ if (super->next_buf) {
+ memcpy(super->next_buf, super->buf, super->len);
+ free(super->buf);
+ super->len = super->next_len;
+ super->buf = super->next_buf;
+
+ super->next_len = 0;
+ super->next_buf = NULL;
+ }
+
+ mpb = super->anchor;
+
switch (type) {
case update_activate_spare: {
struct imsm_update_activate_spare *u = (void *) update->buf;
struct imsm_dev *dev = get_imsm_dev(super, u->array);
struct imsm_map *map = get_imsm_map(dev, 0);
+ struct imsm_map *migr_map;
struct active_array *a;
struct imsm_disk *disk;
__u32 status;
+ __u8 to_state;
struct dl *dl;
unsigned int found;
- int victim;
+ int failed;
+ int victim = get_imsm_disk_idx(dev, u->slot);
int i;
for (dl = super->disks; dl; dl = dl->next)
if (!dl) {
fprintf(stderr, "error: imsm_activate_spare passed "
- "an unknown disk (index: %d serial: %s)\n",
- u->dl->index, u->dl->serial);
+ "an unknown disk (index: %d)\n",
+ u->dl->index);
return;
}
super->updates_pending++;
+ /* count failures (excluding rebuilds and the victim)
+ * to determine map[0] state
+ */
+ failed = 0;
+ for (i = 0; i < map->num_members; i++) {
+ if (i == u->slot)
+ continue;
+ disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i));
+ if (!disk ||
+ __le32_to_cpu(disk->status) & FAILED_DISK)
+ failed++;
+ }
+
/* adding a pristine spare, assign a new index */
if (dl->index < 0) {
dl->index = super->anchor->num_disks;
super->anchor->num_disks++;
}
- victim = get_imsm_disk_idx(map, u->slot);
- map->disk_ord_tbl[u->slot] = __cpu_to_le32(dl->index);
disk = &dl->disk;
status = __le32_to_cpu(disk->status);
status |= CONFIGURED_DISK;
- status &= ~(SPARE_DISK | USABLE_DISK);
+ status &= ~SPARE_DISK;
disk->status = __cpu_to_le32(status);
+ /* mark rebuild */
+ to_state = imsm_check_degraded(super, dev, failed);
+ map->map_state = IMSM_T_STATE_DEGRADED;
+ migrate(dev, to_state, 1);
+ migr_map = get_imsm_map(dev, 1);
+ set_imsm_ord_tbl_ent(map, u->slot, dl->index);
+ set_imsm_ord_tbl_ent(migr_map, u->slot, dl->index | IMSM_ORD_REBUILD);
+
/* count arrays using the victim in the metadata */
found = 0;
for (a = st->arrays; a ; a = a->next) {
dev = get_imsm_dev(super, a->info.container_member);
- map = get_imsm_map(dev, 0);
for (i = 0; i < map->num_members; i++)
- if (victim == get_imsm_disk_idx(map, i))
+ if (victim == get_imsm_disk_idx(dev, i))
found++;
}
- /* clear some flags if the victim is no longer being
+ /* delete the victim if it is no longer being
* utilized anywhere
*/
if (!found) {
- disk = get_imsm_disk(super, victim);
- status = __le32_to_cpu(disk->status);
- status &= ~(CONFIGURED_DISK | USABLE_DISK);
- disk->status = __cpu_to_le32(status);
- /* at this point the disk can be removed from the
- * metadata, however we need to guarantee that we do
- * not race with any manager thread routine that relies
- * on dl->index or map->disk_ord_tbl
- */
+ 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;
+
+ /* 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;
}
/* check update is next in sequence */
if (u->dev_idx != mpb->num_raid_devs) {
- dprintf("%s: can not create arrays out of sequence\n",
- __func__);
+ dprintf("%s: can not create array %d expected index %d\n",
+ __func__, u->dev_idx, mpb->num_raid_devs);
return;
}
if ((new_start >= start && new_start <= end) ||
(start >= new_start && start <= new_end))
overlap = 1;
- if (overlap && disks_overlap(map, new_map)) {
+ if (overlap && disks_overlap(dev, &u->dev)) {
dprintf("%s: arrays overlap\n", __func__);
return;
}
super->updates_pending++;
dev = update->space;
+ map = get_imsm_map(dev, 0);
update->space = NULL;
imsm_copy_dev(dev, &u->dev);
+ map = get_imsm_map(dev, 0);
super->dev_tbl[u->dev_idx] = dev;
mpb->num_raid_devs++;
- /* fix up flags, if arrays overlap then the drives can not be
- * spares
- */
+ /* fix up flags */
for (i = 0; i < map->num_members; i++) {
struct imsm_disk *disk;
__u32 status;
- disk = get_imsm_disk(super, get_imsm_disk_idx(map, i));
+ disk = get_imsm_disk(super, get_imsm_disk_idx(dev, i));
status = __le32_to_cpu(disk->status);
status |= CONFIGURED_DISK;
- if (overlap)
- status &= ~SPARE_DISK;
+ status &= ~SPARE_DISK;
disk->status = __cpu_to_le32(status);
}
break;
}
+ case update_add_disk:
+
+ /* we may be able to repair some arrays if disks are
+ * being added */
+ if (super->add) {
+ struct active_array *a;
+ for (a = st->arrays; a; a = a->next)
+ a->check_degraded = 1;
+ }
+ /* add some spares to the metadata */
+ while (super->add) {
+ struct dl *al;
+
+ al = super->add;
+ super->add = al->next;
+ al->next = super->disks;
+ super->disks = al;
+ dprintf("%s: added %x:%x\n",
+ __func__, al->major, al->minor);
+ }
+
+ break;
}
}
struct metadata_update *update)
{
/**
- * Allocate space to hold new disk entries, raid-device entries or a
- * new mpb if necessary. We currently maintain an mpb large enough to
- * hold 2 subarrays for the given number of disks. This may not be
- * sufficient when reshaping.
- *
- * FIX ME handle the reshape case.
- *
- * The monitor will be able to safely change super->mpb by arranging
- * for it to be freed in check_update_queue(). I.e. the monitor thread
- * will start using the new pointer and the manager can continue to use
- * the old value until check_update_queue() runs.
+ * Allocate space to hold new disk entries, raid-device entries or a new
+ * mpb if necessary. The manager synchronously waits for updates to
+ * complete in the monitor, so new mpb buffers allocated here can be
+ * integrated by the monitor thread without worrying about live pointers
+ * in the manager thread.
*/
enum imsm_update_type type = *(enum imsm_update_type *) update->buf;
+ struct intel_super *super = st->sb;
+ struct imsm_super *mpb = super->anchor;
+ size_t buf_len;
+ size_t len = 0;
switch (type) {
case update_create_array: {
struct imsm_update_create_array *u = (void *) update->buf;
- size_t len = sizeof_imsm_dev(&u->dev, 1);
+ len = sizeof_imsm_dev(&u->dev, 1);
update->space = malloc(len);
break;
default:
}
}
- return;
+ /* check if we need a larger metadata buffer */
+ if (super->next_buf)
+ buf_len = super->next_len;
+ else
+ buf_len = super->len;
+
+ if (__le32_to_cpu(mpb->mpb_size) + len > buf_len) {
+ /* ok we need a larger buf than what is currently allocated
+ * if this allocation fails process_update will notice that
+ * ->next_len is set and ->next_buf is NULL
+ */
+ buf_len = ROUND_UP(__le32_to_cpu(mpb->mpb_size) + len, 512);
+ if (super->next_buf)
+ free(super->next_buf);
+
+ super->next_len = buf_len;
+ if (posix_memalign(&super->next_buf, buf_len, 512) != 0)
+ super->next_buf = NULL;
+ }
+}
+
+/* must be called while manager is quiesced */
+static void imsm_delete(struct intel_super *super, struct dl **dlp, int index)
+{
+ struct imsm_super *mpb = super->anchor;
+ struct dl *iter;
+ struct imsm_dev *dev;
+ struct imsm_map *map;
+ int i, j, num_members;
+ __u32 ord;
+
+ dprintf("%s: deleting device[%d] from imsm_super\n",
+ __func__, index);
+
+ /* shift all indexes down one */
+ 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);
+ map = get_imsm_map(dev, 0);
+ num_members = map->num_members;
+ for (j = 0; j < num_members; j++) {
+ /* update ord entries being careful not to propagate
+ * ord-flags to the first map
+ */
+ ord = get_imsm_ord_tbl_ent(dev, j);
+
+ if (ord_to_idx(ord) <= index)
+ continue;
+
+ map = get_imsm_map(dev, 0);
+ set_imsm_ord_tbl_ent(map, j, ord_to_idx(ord - 1));
+ map = get_imsm_map(dev, 1);
+ if (map)
+ set_imsm_ord_tbl_ent(map, j, ord - 1);
+ }
+ }
+
+ mpb->num_disks--;
+ super->updates_pending++;
+ if (*dlp) {
+ struct dl *dl = *dlp;
+
+ *dlp = (*dlp)->next;
+ __free_imsm_disk(dl);
+ }
}
+#endif /* MDASSEMBLE */
struct superswitch super_imsm = {
#ifndef MDASSEMBLE
.detail_super = detail_super_imsm,
.brief_detail_super = brief_detail_super_imsm,
.write_init_super = write_init_super_imsm,
+ .validate_geometry = validate_geometry_imsm,
+ .add_to_super = add_to_super_imsm,
#endif
.match_home = match_home_imsm,
.uuid_from_super= uuid_from_super_imsm,
.load_super = load_super_imsm,
.init_super = init_super_imsm,
- .add_to_super = add_to_super_imsm,
.store_super = store_zero_imsm,
.free_super = free_super_imsm,
.match_metadata_desc = match_metadata_desc_imsm,
.container_content = container_content_imsm,
- .validate_geometry = validate_geometry_imsm,
.external = 1,
+#ifndef MDASSEMBLE
/* for mdmon */
.open_new = imsm_open_new,
.load_super = load_super_imsm,
.activate_spare = imsm_activate_spare,
.process_update = imsm_process_update,
.prepare_update = imsm_prepare_update,
+#endif /* MDASSEMBLE */
};