__u64 reshape_position; /* next address in array-space for reshape */
__u32 delta_disks; /* change in number of raid_disks */
__u32 new_layout; /* new layout */
- __u32 new_chunk; /* new chunk size (bytes) */
+ __u32 new_chunk; /* new chunk size (sectors) */
__u32 new_offset; /* signed number to add to data_offset in new
* layout. 0 == no-change. This can be
* different on each device in the array.
if (sb->bblog_size && sb->bblog_offset) {
printf(" Bad Block Log : %d entries available at offset %ld sectors",
__le16_to_cpu(sb->bblog_size)*512/8,
- (long)__le32_to_cpu(sb->bblog_offset));
+ (long)(int32_t)__le32_to_cpu(sb->bblog_offset));
if (sb->feature_map &
__cpu_to_le32(MD_FEATURE_BAD_BLOCKS))
printf(" - bad blocks present.");
printf("\n");
}
-
if (calc_sb_1_csum(sb) == sb->sb_csum)
printf(" Checksum : %x - correct\n", __le32_to_cpu(sb->sb_csum));
else
printf("\n");
}
-
static void brief_examine_super1(struct supertype *st, int verbose)
{
struct mdp_superblock_1 *sb = st->sb;
role = __le16_to_cpu(sb->dev_roles[__le32_to_cpu(sb->dev_number)]);
super_offset = __le64_to_cpu(sb->super_offset);
- data_size = __le64_to_cpu(sb->size);
+ if (info->array.level <= 0)
+ data_size = __le64_to_cpu(sb->data_size);
+ else
+ data_size = __le64_to_cpu(sb->size);
if (info->data_offset < super_offset) {
unsigned long long end;
info->space_before = info->data_offset;
info->disk.state |= 1 << MD_DISK_REPLACEMENT;
}
-
if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET))
info->recovery_start = __le32_to_cpu(sb->recovery_offset);
else
if (sb->feature_map & __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE)) {
info->reshape_active = 1;
- if (info->array.level == 10)
+ if ((sb->feature_map & __le32_to_cpu(MD_FEATURE_NEW_OFFSET)) &&
+ sb->new_offset != 0)
info->reshape_active |= RESHAPE_NO_BACKUP;
info->reshape_progress = __le64_to_cpu(sb->reshape_position);
info->new_level = __le32_to_cpu(sb->new_level);
*/
unsigned long long sb_offset = __le64_to_cpu(sb->super_offset);
unsigned long long data_offset = __le64_to_cpu(sb->data_offset);
- long bitmap_offset = (long)__le64_to_cpu(sb->bitmap_offset);
+ long bitmap_offset = (long)(int32_t)__le32_to_cpu(sb->bitmap_offset);
long bm_sectors = 0;
long space;
misc->device_size - __le64_to_cpu(sb->data_offset));
printf("Size is %llu\n", (unsigned long long)
__le64_to_cpu(sb->data_size));
+ } else if (strcmp(update, "revert-reshape") == 0) {
+ rv = -2;
+ if (!(sb->feature_map & __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE)))
+ pr_err("No active reshape to revert on %s\n",
+ devname);
+ else {
+ __u32 temp;
+ unsigned long long reshape_sectors;
+ long reshape_chunk;
+ rv = 0;
+ /* reshape_position is a little messy.
+ * Its value must be a multiple of the larger
+ * chunk size, and of the "after" data disks.
+ * So when reverting we need to change it to
+ * be a multiple of the new "after" data disks,
+ * which is the old "before".
+ * If it isn't already a multiple of 'before',
+ * the only thing we could do would be
+ * copy some block around on the disks, which
+ * is easy to get wrong.
+ * So we reject a revert-reshape unless the
+ * alignment is good.
+ */
+ if (__le32_to_cpu(sb->level) >= 4 &&
+ __le32_to_cpu(sb->level) <= 6) {
+ reshape_sectors = __le64_to_cpu(sb->reshape_position);
+ reshape_chunk = __le32_to_cpu(sb->new_chunk);
+ reshape_chunk *= __le32_to_cpu(sb->raid_disks) - __le32_to_cpu(sb->delta_disks) -
+ (__le32_to_cpu(sb->level)==6 ? 2 : 1);
+ if (reshape_sectors % reshape_chunk) {
+ pr_err("Reshape position is not suitably aligned.\n");
+ pr_err("Try normal assembly and stop again\n");
+ return -2;
+ }
+ }
+ sb->raid_disks = __cpu_to_le32(__le32_to_cpu(sb->raid_disks) -
+ __le32_to_cpu(sb->delta_disks));
+ if (sb->delta_disks == 0)
+ sb->feature_map ^= __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS);
+ else
+ sb->delta_disks = __cpu_to_le32(-__le32_to_cpu(sb->delta_disks));
+
+ temp = sb->new_layout;
+ sb->new_layout = sb->layout;
+ sb->layout = temp;
+
+ temp = sb->new_chunk;
+ sb->new_chunk = sb->chunksize;
+ sb->chunksize = temp;
+
+ if (sb->feature_map & __cpu_to_le32(MD_FEATURE_NEW_OFFSET)) {
+ long offset_delta = (int32_t)__le32_to_cpu(sb->new_offset);
+ sb->data_offset = __cpu_to_le64(__le64_to_cpu(sb->data_offset) + offset_delta);
+ sb->new_offset = __cpu_to_le32(-offset_delta);
+ sb->data_size = __cpu_to_le64(__le64_to_cpu(sb->data_size) - offset_delta);
+ }
+ }
} else if (strcmp(update, "_reshape_progress")==0)
sb->reshape_position = __cpu_to_le64(info->reshape_progress);
else if (strcmp(update, "writemostly")==0)
return -EINVAL;
}
-
-
if (sb_offset != __le64_to_cpu(sb->super_offset) &&
0 != __le64_to_cpu(sb->super_offset)
) {
int rfd;
int rv = 0;
unsigned long long bm_space;
- unsigned long long reserved;
struct devinfo *di;
unsigned long long dsize, array_size;
- unsigned long long sb_offset, headroom;
+ unsigned long long sb_offset;
unsigned long long data_offset;
for (di = st->info; di; di = di->next) {
sb->events = 0;
refst = dup_super(st);
- if (load_super1(refst, di->fd, NULL)==0) {
+ if (load_super1(refst, di->fd, NULL)==0) {
struct mdp_superblock_1 *refsb = refst->sb;
memcpy(sb->device_uuid, refsb->device_uuid, 16);
goto error_out;
}
-
/*
* Calculate the position of the superblock.
* It is always aligned to a 4K boundary and
* 0: At least 8K, but less than 12K, from end of device
* 1: At start of device
* 2: 4K from start of device.
- * Depending on the array size, we might leave extra space
- * for a bitmap.
- * Also leave 4K for bad-block log.
+ * data_offset has already been set.
*/
array_size = __le64_to_cpu(sb->size);
/* work out how much space we left for a bitmap,
* Add 8 sectors for bad block log */
bm_space = choose_bm_space(array_size) + 8;
- /* We try to leave 0.1% at the start for reshape
- * operations, but limit this to 128Meg (0.1% of 10Gig)
- * which is plenty for efficient reshapes
- * However we make it at least 2 chunks as one chunk
- * is minimum needed for reshape.
- */
- headroom = 128 * 1024 * 2;
- while (headroom << 10 > array_size &&
- headroom/2 >= __le32_to_cpu(sb->chunksize) * 2)
- headroom >>= 1;
-
data_offset = di->data_offset;
+ if (data_offset == INVALID_SECTORS)
+ data_offset = st->data_offset;
switch(st->minor_version) {
case 0:
+ if (data_offset == INVALID_SECTORS)
+ data_offset = 0;
sb_offset = dsize;
sb_offset -= 8*2;
sb_offset &= ~(4*2-1);
+ sb->data_offset = __cpu_to_le64(data_offset);
sb->super_offset = __cpu_to_le64(sb_offset);
- if (data_offset == INVALID_SECTORS)
- sb->data_offset = 0;
if (sb_offset < array_size + bm_space)
bm_space = sb_offset - array_size;
sb->data_size = __cpu_to_le64(sb_offset - bm_space);
break;
case 1:
sb->super_offset = __cpu_to_le64(0);
- if (data_offset == INVALID_SECTORS) {
- reserved = bm_space + 4*2;
- if (reserved < headroom)
- reserved = headroom;
- if (reserved + array_size > dsize)
- reserved = dsize - array_size;
- /* Try for multiple of 1Meg so it is nicely aligned */
- #define ONE_MEG (2*1024)
- if (reserved > ONE_MEG)
- reserved = (reserved/ONE_MEG) * ONE_MEG;
-
- /* force 4K alignment */
- reserved &= ~7ULL;
-
- } else
- reserved = data_offset;
-
- sb->data_offset = __cpu_to_le64(reserved);
- sb->data_size = __cpu_to_le64(dsize - reserved);
- if (reserved >= 8 + 32*2 + 8) {
+ if (data_offset == INVALID_SECTORS)
+ data_offset = 16;
+
+ sb->data_offset = __cpu_to_le64(data_offset);
+ sb->data_size = __cpu_to_le64(dsize - data_offset);
+ if (data_offset >= 8 + 32*2 + 8) {
sb->bblog_size = __cpu_to_le16(8);
sb->bblog_offset = __cpu_to_le32(8 + 32*2);
- } else if (reserved >= 16) {
+ } else if (data_offset >= 16) {
sb->bblog_size = __cpu_to_le16(8);
- sb->bblog_offset = __cpu_to_le32(reserved-8);
+ sb->bblog_offset = __cpu_to_le32(data_offset-8);
}
break;
case 2:
sb_offset = 4*2;
- sb->super_offset = __cpu_to_le64(4*2);
- if (data_offset == INVALID_SECTORS) {
- if (4*2 + 4*2 + bm_space + array_size
- > dsize)
- bm_space = dsize - array_size
- - 4*2 - 4*2;
-
- reserved = bm_space + 4*2 + 4*2;
- if (reserved < headroom)
- reserved = headroom;
- if (reserved + array_size > dsize)
- reserved = dsize - array_size;
- /* Try for multiple of 1Meg so it is nicely aligned */
- #define ONE_MEG (2*1024)
- if (reserved > ONE_MEG)
- reserved = (reserved/ONE_MEG) * ONE_MEG;
-
- /* force 4K alignment */
- reserved &= ~7ULL;
-
- } else
- reserved = data_offset;
-
- sb->data_offset = __cpu_to_le64(reserved);
- sb->data_size = __cpu_to_le64(dsize - reserved);
- if (reserved >= 16 + 32*2 + 8) {
+ sb->super_offset = __cpu_to_le64(sb_offset);
+ if (data_offset == INVALID_SECTORS)
+ data_offset = 24;
+
+ sb->data_offset = __cpu_to_le64(data_offset);
+ sb->data_size = __cpu_to_le64(dsize - data_offset);
+ if (data_offset >= 16 + 32*2 + 8) {
sb->bblog_size = __cpu_to_le16(8);
sb->bblog_offset = __cpu_to_le32(8 + 32*2);
- } else if (reserved >= 16+16) {
+ } else if (data_offset >= 16+16) {
sb->bblog_size = __cpu_to_le16(8);
/* '8' sectors for the bblog, and another '8'
* because we want offset from superblock, not
* start of device.
*/
- sb->bblog_offset = __cpu_to_le32(reserved-8-8);
+ sb->bblog_offset = __cpu_to_le32(data_offset-8-8);
}
break;
default:
misc = (struct misc_dev_info*) (((char*)super)+MAX_SB_SIZE+BM_SUPER_SIZE);
misc->device_size = dsize;
+ if (st->data_offset == INVALID_SECTORS)
+ st->data_offset = __le64_to_cpu(super->data_offset);
/* Now check on the bitmap superblock */
if ((__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) == 0)
return 0;
}
-
static struct supertype *match_metadata_desc1(char *arg)
{
struct supertype *st = xcalloc(1, sizeof(*st));
st->ss = &super1;
st->max_devs = MAX_DEVS;
st->sb = NULL;
+ st->data_offset = INVALID_SECTORS;
/* leading zeros can be safely ignored. --detail generates them. */
while (*arg == '0')
arg++;
* superblock type st, and reserving 'reserve' sectors for
* a possible bitmap
*/
-static __u64 _avail_size1(struct supertype *st, __u64 devsize,
- unsigned long long data_offset, int chunksize)
+static __u64 avail_size1(struct supertype *st, __u64 devsize,
+ unsigned long long data_offset)
{
struct mdp_superblock_1 *super = st->sb;
int bmspace = 0;
+ int bbspace = 0;
if (devsize < 24)
return 0;
- if (super == NULL)
- /* creating: allow suitable space for bitmap */
- bmspace = choose_bm_space(devsize);
#ifndef MDASSEMBLE
- else if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
+ if (__le32_to_cpu(super->feature_map)&MD_FEATURE_BITMAP_OFFSET) {
/* hot-add. allow for actual size of bitmap */
struct bitmap_super_s *bsb;
bsb = (struct bitmap_super_s *)(((char*)super)+MAX_SB_SIZE);
}
#endif
/* Allow space for bad block log */
- if (super && super->bblog_size)
- devsize -= __le16_to_cpu(super->bblog_size);
- else
- devsize -= 8;
-
+ if (super->bblog_size)
+ bbspace = __le16_to_cpu(super->bblog_size);
if (st->minor_version < 0)
/* not specified, so time to set default */
st->minor_version = 2;
+ if (data_offset == INVALID_SECTORS)
+ data_offset = st->data_offset;
+
if (data_offset != INVALID_SECTORS)
switch(st->minor_version) {
case 0:
- return devsize - data_offset - 8*2;
+ return devsize - data_offset - 8*2 - bbspace;
case 1:
case 2:
return devsize - data_offset;
devsize -= bmspace;
- if (super == NULL && st->minor_version > 0) {
- /* haven't committed to a size yet, so allow some
- * slack for space for reshape.
- * Limit slack to 128M, but aim for about 0.1%
- */
- unsigned long long headroom = 128*1024*2;
- while ((headroom << 10) > devsize &&
- (chunksize == 0 ||
- headroom / 2 >= ((unsigned)chunksize*2)*2))
- headroom >>= 1;
- devsize -= headroom;
- }
switch(st->minor_version) {
case 0:
/* at end */
- return ((devsize - 8*2 ) & ~(4*2-1));
+ return ((devsize - 8*2 - bbspace ) & ~(4*2-1));
case 1:
/* at start, 4K for superblock and possible bitmap */
- return devsize - 4*2;
+ return devsize - 4*2 - bbspace;
case 2:
/* 4k from start, 4K for superblock and possible bitmap */
- return devsize - (4+4)*2;
+ return devsize - (4+4)*2 - bbspace;
}
return 0;
}
-static __u64 avail_size1(struct supertype *st, __u64 devsize,
- unsigned long long data_offset)
-{
- return _avail_size1(st, devsize, data_offset, 0);
-}
static int
add_internal_bitmap1(struct supertype *st,
bitmap_super_t *bms = (bitmap_super_t*)(((char*)sb) + MAX_SB_SIZE);
int uuid[4];
-
if (__le64_to_cpu(sb->data_size) == 0)
/* Must be creating the array, else data_size would be non-zero */
creating = 1;
char *subdev, unsigned long long *freesize,
int verbose)
{
- unsigned long long ldsize;
+ unsigned long long ldsize, devsize;
+ int bmspace;
+ unsigned long long headroom;
int fd;
if (level == LEVEL_CONTAINER) {
pr_err("1.x metadata does not support containers\n");
return 0;
}
- if (chunk && *chunk == UnSet)
+ if (*chunk == UnSet)
*chunk = DEFAULT_CHUNK;
if (!subdev)
return 1;
+ if (st->minor_version < 0)
+ /* not specified, so time to set default */
+ st->minor_version = 2;
+
fd = open(subdev, O_RDONLY|O_EXCL, 0);
if (fd < 0) {
if (verbose)
}
close(fd);
- *freesize = _avail_size1(st, ldsize >> 9, data_offset, *chunk);
+ devsize = ldsize >> 9;
+ if (devsize < 24) {
+ *freesize = 0;
+ return 0;
+ }
+
+ /* creating: allow suitable space for bitmap */
+ bmspace = choose_bm_space(devsize);
+
+ if (data_offset == INVALID_SECTORS)
+ data_offset = st->data_offset;
+ if (data_offset == INVALID_SECTORS)
+ switch (st->minor_version) {
+ case 0:
+ data_offset = 0;
+ break;
+ case 1:
+ case 2:
+ /* Choose data offset appropriate for this device
+ * and use as default for whole array.
+ * The data_offset must allow for bitmap space
+ * and base metadata, should allow for some headroom
+ * for reshape, and should be rounded to multiple
+ * of 1M.
+ * Headroom is limited to 128M, but aim for about 0.1%
+ */
+ headroom = 128*1024*2;
+ while ((headroom << 10) > devsize &&
+ (*chunk == 0 ||
+ headroom / 2 >= ((unsigned)(*chunk)*2)*2))
+ headroom >>= 1;
+ data_offset = 12*2 + bmspace + headroom;
+ #define ONE_MEG (2*1024)
+ if (data_offset > ONE_MEG)
+ data_offset = (data_offset / ONE_MEG) * ONE_MEG;
+ break;
+ }
+ if (st->data_offset == INVALID_SECTORS)
+ st->data_offset = data_offset;
+ switch(st->minor_version) {
+ case 0: /* metadata at end. Round down and subtract space to reserve */
+ devsize = (devsize & ~(4ULL*2-1));
+ /* space for metadata, bblog, bitmap */
+ devsize -= 8*2 + 8 + bmspace;
+ break;
+ case 1:
+ case 2:
+ devsize -= data_offset;
+ break;
+ }
+ *freesize = devsize;
return 1;
}
#endif /* MDASSEMBLE */
+void *super1_make_v0(struct supertype *st, struct mdinfo *info, mdp_super_t *sb0)
+{
+ /* Create a v1.0 superblock based on 'info'*/
+ void *ret;
+ struct mdp_superblock_1 *sb;
+ int i;
+ int rfd;
+ unsigned long long offset;
+
+ if (posix_memalign(&ret, 4096, 1024) != 0)
+ return NULL;
+ sb = ret;
+ memset(ret, 0, 1024);
+ sb->magic = __cpu_to_le32(MD_SB_MAGIC);
+ sb->major_version = __cpu_to_le32(1);
+
+ copy_uuid(sb->set_uuid, info->uuid, super1.swapuuid);
+ sprintf(sb->set_name, "%d", sb0->md_minor);
+ sb->ctime = __cpu_to_le32(info->array.ctime+1);
+ sb->level = __cpu_to_le32(info->array.level);
+ sb->layout = __cpu_to_le32(info->array.layout);
+ sb->size = __cpu_to_le64(info->component_size);
+ sb->chunksize = __cpu_to_le32(info->array.chunk_size/512);
+ sb->raid_disks = __cpu_to_le32(info->array.raid_disks);
+ if (info->array.level > 0)
+ sb->data_size = sb->size;
+ else
+ sb->data_size = st->ss->avail_size(st, st->devsize/512, 0);
+ sb->resync_offset = MaxSector;
+ sb->max_dev = __cpu_to_le32(MD_SB_DISKS);
+ sb->dev_number = __cpu_to_le32(info->disk.number);
+ sb->utime = __cpu_to_le64(info->array.utime);
+
+ offset = st->devsize/512 - 8*2;
+ offset &= ~(4*2-1);
+ sb->super_offset = __cpu_to_le64(offset);
+ //*(__u64*)(st->other + 128 + 8 + 8) = __cpu_to_le64(offset);
+
+ if ((rfd = open("/dev/urandom", O_RDONLY)) < 0 ||
+ read(rfd, sb->device_uuid, 16) != 16) {
+ __u32 r[4] = {random(), random(), random(), random()};
+ memcpy(sb->device_uuid, r, 16);
+ }
+ if (rfd >= 0)
+ close(rfd);
+
+ for (i = 0; i < MD_SB_DISKS; i++) {
+ int state = sb0->disks[i].state;
+ sb->dev_roles[i] = 0xFFFF;
+ if ((state & (1<<MD_DISK_SYNC)) &&
+ !(state & (1<<MD_DISK_FAULTY)))
+ sb->dev_roles[i] = __cpu_to_le16(sb0->disks[i].raid_disk);
+ }
+ sb->sb_csum = calc_sb_1_csum(sb);
+ return ret;
+}
+
struct superswitch super1 = {
#ifndef MDASSEMBLE
.examine_super = examine_super1,