When switching from Raid0 to Raid5(4) a special procedure must be applied when storing meta on disks.
IMSM do not support Raid4 layout thus it will be converted back to Raid0 layout with raid diskNum-1.
The conversion is performed on the separate buffer right before actual disk write operation.
In such a case mdmon is still aware of Raid4 personality/layout.
Backwards compatibility conversion (when assembling an array) will be added along with
IMSM reshape functionality. If no reshape is detected during an array assembly no
conversion will be made.
---
super-intel.c | 151 +++++++++++++++++++++++++++++++++++++++++++++++++++++++--
1 files changed, 147 insertions(+), 4 deletions(-)
diff --git a/super-intel.c b/super-intel.c
index 32b29c8..9b0992f 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -3344,14 +3344,132 @@ static int write_super_imsm_spares(struct intel_super *super, int doclose)
return 0;
}
-static int write_super_imsm(struct intel_super *super, int doclose)
+static int make_mpb_imsm_compatible(struct supertype *st, struct imsm_super *mpb)
{
+ struct intel_super *super = st->sb;
+ struct imsm_map *map;
+ struct active_array *a;
+ __u8 *src, *dst;
+ __u32 orig_mpb_size = mpb->mpb_size;
+ __u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
+ __u32 sum;
+ int i, bytes_to_copy, spares = 0;
+ struct dl *d;
+
+ /* Make sure that there is missing disk */
+ if (super->missing == NULL)
+ return -1;
+
+ /* Decrease the number of disks -
+ missing disk will not be exposed to the user*/
+ mpb->num_disks--;
+
+ /* Create MBP disk table */
+ mpb_size += sizeof(struct imsm_disk) * mpb->num_disks;
+ for (d = super->disks; d; d = d->next) {
+ if (d->index == -1)
+ spares++;
+ else
+ mpb->disk[d->index] = d->disk;
+ }
+
+ /* Iterate through the imsm devices */
+ for (i = 0; i < mpb->num_raid_devs; i++) {
+
+ struct imsm_dev *dev = __get_imsm_dev(mpb, i);
+ imsm_copy_dev(dev, get_imsm_dev(super, i));
+
+ /* Find layout of the current aray */
+ for (a = st->arrays; a; a = a->next)
+ if (a->info.container_member == i)
+ break;
+
+ /* Process the first map */
+ map = get_imsm_map(dev, 0);
+ if ((map->raid_level == 5) && (a->info.array.layout == ALGORITHM_PARITY_N)) {
+
+ /* get the pointer to the rest of the metadata */
+ src = (__u8 *)map + sizeof_imsm_map(map);
+
+ /* change the level and disk number to be compatible with IMSM */
+ map->raid_level = 0;
+ map->num_members--;
+
+ /* get the updated pointer to the rest of the metadata */
+ dst = (__u8 *)map + sizeof_imsm_map(map);
+ /* Now move the rest of the metadata to be properly aligned */
+ bytes_to_copy = orig_mpb_size - (src - (__u8 *)mpb);
+ if (bytes_to_copy > 0)
+ memmove(dst, src, bytes_to_copy);
+ }
+
+ /* Process the second map if needed */
+ if (dev->vol.migr_state != IMSM_T_STATE_NORMAL) {
+
+ map = get_imsm_map(dev, 1);
+ if ((map->raid_level == 5) && (a->info.array.layout == ALGORITHM_PARITY_N)) {
+
+ /* change the level and disk number to be compatible with IMSM */
+ map->raid_level = 0;
+ map->num_members--;
+ }
+ }
+ 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);
+ mpb->check_sum = __cpu_to_le32(sum);
+
+ return 0;
+}
+
+static int is_mpb_imsm_compatible(struct supertype *st, struct imsm_super *mpb)
+{
+ int i;
+
+ /* Iterate through the imsm devices */
+ for (i = 0; i < mpb->num_raid_devs; i++) {
+ struct imsm_dev *dev;
+ struct imsm_map *map;
+ struct active_array *a;
+ dev = __get_imsm_dev(mpb, i);
+
+ /* Find layout of the current aray */
+ for (a = st->arrays; a; a = a->next) {
+ if (a->info.container_member == i)
+ break;
+ }
+ /* Get first map */
+ map = get_imsm_map(dev, 0);
+ if ((map->raid_level == 5) && (a->info.array.layout == ALGORITHM_PARITY_N)) {
+ /* This leve/layout is not compatible with IMSM */
+ return 0;
+ }
+ if (dev->vol.migr_state != IMSM_T_STATE_NORMAL) {
+ /* Get second map */
+ map = get_imsm_map(dev, 1);
+ if ((map->raid_level == 5) && (a->info.array.layout == ALGORITHM_PARITY_N)) {
+ /* This leve/layout is not compatible with IMSM */
+ return 0;
+ }
+ }
+ }
+ return 1;
+}
+
+static int write_super_imsm(struct supertype *st, int doclose)
+{
+ struct intel_super *super = st->sb;
struct imsm_super *mpb = super->anchor;
+ struct imsm_super *tmp_mpb = NULL;
struct dl *d;
__u32 generation;
__u32 sum;
int spares = 0;
- int i;
+ int i, buflen;
__u32 mpb_size = sizeof(struct imsm_super) - sizeof(struct imsm_disk);
/* 'generation' is incremented everytime the metadata is written */
@@ -3388,6 +3506,27 @@ static int write_super_imsm(struct intel_super *super, int doclose)
sum = __gen_imsm_checksum(mpb);
mpb->check_sum = __cpu_to_le32(sum);
+ /* The mpb is generated and ready to be flushed to disks.
+ * Check configuration for compatibility with IMSM
+ * The following configurations should be handled here:
+ * N-drive Raid 4 => will be written as N-1 Raid0
+ */
+ if (!is_mpb_imsm_compatible(st, mpb)) {
+ buflen = ROUND_UP(super->anchor->mpb_size, 512);
+ if (posix_memalign((void **)&tmp_mpb, 512, buflen) != 0) {
+ fprintf(stderr,
+ Name ": unable to allocate %zu byte mpb buffer\n",
+ super->len);
+ return 1;
+ }
+ /* Copy the current mpb */
+ memcpy(tmp_mpb, mpb, buflen);
+ /* Translate mpb to be IMSM compatible */
+ make_mpb_imsm_compatible(st, tmp_mpb);
+ /* From now mpb will point to the new meta */
+ mpb = tmp_mpb;
+ }
+
/* write the mpb for disks that compose raid devices */
for (d = super->disks; d ; d = d->next) {
if (d->index < 0)
@@ -3401,6 +3540,10 @@ static int write_super_imsm(struct intel_super *super, int doclose)
}
}
+ /* Free temporary mpb */
+ if (tmp_mpb)
+ free(tmp_mpb);
+
if (spares)
return write_super_imsm_spares(super, doclose);
@@ -3496,7 +3639,7 @@ static int write_init_super_imsm(struct supertype *st)
return rv;
} else
- return write_super_imsm(st->sb, 1);
+ return write_super_imsm(st, 1);
}
#endif
@@ -4569,7 +4712,7 @@ static void imsm_sync_metadata(struct supertype *container)
if (!super->updates_pending)
return;
- write_super_imsm(super, 0);
+ write_super_imsm(container, 0);
super->updates_pending = 0;
}
��.n��������+%������w��{.n�����{����w��ܨ}���Ơz�j:+v�����w����ޙ��&�)ߡ�a����z�ޗ���ݢj��w�f
