Skip to contentSkip to conversation

fix hard-coded disk header chunk size for persistent snapshot

PATCH In review Tomohiro Kusumi 3 participants 4 messages May 12, 2010 Device mapper

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fix hard-coded disk header chunk size for persistent snapshot

Hi,

This patch fixes hard-coded value for the size of a chunk that includes disk header for persistent snapshot. It should be changed to existing macro NUM_SNAPSHOT_HDR_CHUNKS instead of using hard-coded value 1.

Thanks,
Tomohiro Kusumi

---

drivers/md/dm-snap-persistent.c 2010-05-12 09:21:50.000000000 +0900 renamed from drivers/md/dm-snap-persistent.c 2010-05-10 10:36:28.000000000 +0900 +1 -1
diff -aNur linux-2.6.34-rc7.org/drivers/md/dm-snap-persistent.c linux-2.6.34-rc7/drivers/md/dm-snap-persistent.c
@@ -266,7 +266,7 @@
  */
 static chunk_t area_location(struct pstore *ps, chunk_t area)
 {
-       return 1 + ((ps->exceptions_per_area + 1) * area);
+       return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area);
 }
 
 /*
#1 lore raw
Quoted 3 lines · Tomohiro Kusumi

This patch fixes hard-coded value for the size of a chunk that includes disk header for persistent snapshot. It should be changed to existing macro NUM_SNAPSHOT_HDR_CHUNKS instead of using hard-coded value 1.

Any more places too? (E.g. persistent_commit_merge)

Alasdair

maintainer
#2 lore raw
Quoted 8 lines · Alasdair G Kergon
Quoted 3 lines · Tomohiro Kusumi

This patch fixes hard-coded value for the size of a chunk that includes disk header for persistent snapshot. It should be changed to existing macro NUM_SNAPSHOT_HDR_CHUNKS instead of using hard-coded value 1.

Any more places too? (E.g. persistent_commit_merge)

Alasdair

I think the expression in persistent_commit_merge should be changed to:
ps->next_free = area_location(ps, ps->current_area) +

                ps->current_committed + 1;

It is numerically the same, but will avoid trouble if NUM_SNAPSHOT_HDR_CHUNKS is increased.

Mikulas

#3 lore raw
Quoted 13 lines · Mikulas Patocka
Quoted 8 lines · Alasdair G Kergon
Quoted 3 lines · Tomohiro Kusumi

This patch fixes hard-coded value for the size of a chunk that includes disk header for persistent snapshot. It should be changed to existing macro NUM_SNAPSHOT_HDR_CHUNKS instead of using hard-coded value 1.

Any more places too? (E.g. persistent_commit_merge)

Alasdair

Quoted 10 lines

I think the expression in persistent_commit_merge should be changed to:
ps->next_free = area_location(ps, ps->current_area) +

                 ps->current_committed + 1;

It is numerically the same, but will avoid trouble if NUM_SNAPSHOT_HDR_CHUNKS is increased.

Mikulas

agree,

a single chunk is large enough to contain 16byte disk header
which does not seem necessary to increase NUM_SNAPSHOT_HDR_CHUNKS in the future (unless whole structure get totally changed),
but it should also get fixed to avoid trouble.

I merged it to the patch with your names signed off.

Thanks,
Tomohiro Kusumi

---

drivers/md/dm-snap-persistent.c +3 -3
diff -aNur linux-2.6.34-rc7.org/drivers/md/dm-snap-persistent.c linux-2.6.34-rc7/drivers/md/dm-snap-persistent.c
@@ -266,7 +266,7 @@
  */
 static chunk_t area_location(struct pstore *ps, chunk_t area)
 {
-	return 1 + ((ps->exceptions_per_area + 1) * area);
+	return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area);
 }
 
 /*
@@ -780,8 +780,8 @@
 	 * ps->current_area does not get reduced by prepare_merge() until
 	 * after commit_merge() has removed the nr_merged previous exceptions.
 	 */
-	ps->next_free = (area_location(ps, ps->current_area) - 1) +
-			(ps->current_committed + 1) + NUM_SNAPSHOT_HDR_CHUNKS;
+	ps->next_free = area_location(ps, ps->current_area) +
+			ps->current_committed + 1;
 
 	return 0;
 }