Skip to contentSkip to conversation

dm snapshot: use unsigned integer chunk size

PATCH In review Mike Snitzer 4 participants 7 messages Sep 30, 2009 – Dec 2, 2009 Device mapper

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 dm snapshot: use unsigned integer chunk size

Use unsigned integer chunk size.

Maximum chunk size is 512kB, there won't ever be need to use 4GB chunk size, so the number can be 32-bit. This fixes compiler failure on 32-bit systems with large block devices.

---
NOTE: this patch was made to go _before_ the snapshot exception patchset; I'll be sending [v2] of 2 patches from that series. They needed small adjustments now that this chunk_size patch comes before them.

---
 drivers/md/dm-exception-store.c |   23 +++++++++++------------
 drivers/md/dm-exception-store.h |    8 ++++----
 drivers/md/dm-snap-persistent.c |   16 ++++++++--------
 drivers/md/dm-snap.c            |    4 ++--
 4 files changed, 25 insertions(+), 26 deletions(-)
drivers/md/dm-snap.c +2 -2
Index: linux-2.6/drivers/md/dm-snap.c
===================================================================
@@ -527,7 +527,7 @@ static int dm_add_exception(void *contex
 static sector_t __minimum_chunk_size(struct origin *o)
 {
 	struct dm_snapshot *snap;
-	sector_t chunk_size = 0;
+	unsigned chunk_size = 0;
 
 	if (o)
 		list_for_each_entry(snap, &o->snapshots, list)
@@ -975,7 +975,7 @@ static void start_copy(struct dm_snap_pe
 
 	src.bdev = bdev;
 	src.sector = chunk_to_sector(s->store, pe->e.old_chunk);
-	src.count = min(s->store->chunk_size, dev_size - src.sector);
+	src.count = min((sector_t)s->store->chunk_size, dev_size - src.sector);
 
 	dest.bdev = s->store->cow->bdev;
 	dest.sector = chunk_to_sector(s->store, pe->e.new_chunk);
drivers/md/dm-exception-store.c +11 -12
Index: linux-2.6/drivers/md/dm-exception-store.c
===================================================================
@@ -141,48 +141,47 @@ EXPORT_SYMBOL(dm_exception_store_type_un
 static int set_chunk_size(struct dm_exception_store *store,
 			  const char *chunk_size_arg, char **error)
 {
-	unsigned long chunk_size_ulong;
+	unsigned chunk_size;
 	char *value;
 
-	chunk_size_ulong = simple_strtoul(chunk_size_arg, &value, 10);
+	chunk_size = simple_strtoul(chunk_size_arg, &value, 10);
 	if (*chunk_size_arg == '\0' || *value != '\0') {
 		*error = "Invalid chunk size";
 		return -EINVAL;
 	}
 
-	if (!chunk_size_ulong) {
+	if (!chunk_size) {
 		store->chunk_size = store->chunk_mask = store->chunk_shift = 0;
 		return 0;
 	}
 
-	return dm_exception_store_set_chunk_size(store, chunk_size_ulong,
-						 error);
+	return dm_exception_store_set_chunk_size(store, chunk_size, error);
 }
 
 int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
-				      unsigned long chunk_size_ulong,
+				      unsigned chunk_size,
 				      char **error)
 {
 	/* Check chunk_size is a power of 2 */
-	if (!is_power_of_2(chunk_size_ulong)) {
+	if (!is_power_of_2(chunk_size)) {
 		*error = "Chunk size is not a power of 2";
 		return -EINVAL;
 	}
 
 	/* Validate the chunk size against the device block size */
-	if (chunk_size_ulong % (bdev_logical_block_size(store->cow->bdev) >> 9)) {
+	if (chunk_size % (bdev_logical_block_size(store->cow->bdev) >> 9)) {
 		*error = "Chunk size is not a multiple of device blocksize";
 		return -EINVAL;
 	}
 
-	if (chunk_size_ulong > INT_MAX >> SECTOR_SHIFT) {
+	if (chunk_size > INT_MAX >> SECTOR_SHIFT) {
 		*error = "Chunk size is too high";
 		return -EINVAL;
 	}
 
-	store->chunk_size = chunk_size_ulong;
-	store->chunk_mask = chunk_size_ulong - 1;
-	store->chunk_shift = ffs(chunk_size_ulong) - 1;
+	store->chunk_size = chunk_size;
+	store->chunk_mask = chunk_size - 1;
+	store->chunk_shift = ffs(chunk_size) - 1;
 
 	return 0;
 }
drivers/md/dm-exception-store.h +4 -4
Index: linux-2.6/drivers/md/dm-exception-store.h
===================================================================
@@ -101,9 +101,9 @@ struct dm_exception_store {
 	struct dm_dev *cow;
 
 	/* Size of data blocks saved - must be a power of 2 */
-	chunk_t chunk_size;
-	chunk_t chunk_mask;
-	chunk_t chunk_shift;
+	unsigned chunk_size;
+	unsigned chunk_mask;
+	unsigned chunk_shift;
 
 	void *context;
 };
@@ -169,7 +169,7 @@ int dm_exception_store_type_register(str
 int dm_exception_store_type_unregister(struct dm_exception_store_type *type);
 
 int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
-				      unsigned long chunk_size_ulong,
+				      unsigned chunk_size,
 				      char **error);
 
 int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
drivers/md/dm-snap-persistent.c +8 -8
Index: linux-2.6/drivers/md/dm-snap-persistent.c
===================================================================
@@ -284,12 +284,13 @@ static int read_header(struct pstore *ps
 {
 	int r;
 	struct disk_header *dh;
-	chunk_t chunk_size;
+	unsigned chunk_size;
 	int chunk_size_supplied = 1;
 	char *chunk_err;
 
 	/*
-	 * Use default chunk size (or hardsect_size, if larger) if none supplied
+	 * Use default chunk size (or logical_block_size, if larger)
+	 * if none supplied
 	 */
 	if (!ps->store->chunk_size) {
 		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
@@ -334,10 +335,9 @@ static int read_header(struct pstore *ps
 		return 0;
 
 	if (chunk_size_supplied)
-		DMWARN("chunk size %llu in device metadata overrides "
-		       "table chunk size of %llu.",
-		       (unsigned long long)chunk_size,
-		       (unsigned long long)ps->store->chunk_size);
+		DMWARN("chunk size %u in device metadata overrides "
+		       "table chunk size of %u.",
+		       chunk_size, ps->store->chunk_size);
 
 	/* We had a bogus chunk_size. Fix stuff up. */
 	free_area(ps);
@@ -345,8 +345,8 @@ static int read_header(struct pstore *ps
 	r = dm_exception_store_set_chunk_size(ps->store, chunk_size,
 					      &chunk_err);
 	if (r) {
-		DMERR("invalid on-disk chunk size %llu: %s.",
-		      (unsigned long long)chunk_size, chunk_err);
+		DMERR("invalid on-disk chunk size %u: %s.",
+		      chunk_size, chunk_err);
 		return r;
 	}
 
#1 lore raw
Quoted 1 line · Mike Snitzer

Maximum chunk size is 512kB,

Can you confirm that a supplied chunk size of 4294971392, for example, still gets rejected?

drivers/md/dm-exception-store.c EXPORT_SYMBOL(dm_exception_store_type_un · line 147

Quoted 9 lines
drivers/md/dm-exception-store.c +2 -2
-	unsigned long chunk_size_ulong;
+	unsigned chunk_size;
 
-	chunk_size_ulong = simple_strtoul(chunk_size_arg, &value, 10);
+	chunk_size = simple_strtoul(chunk_size_arg, &value, 10);

Alasdair

maintainer
#2 lore raw

PATCH 1/1 fix bug on invalid chunksize

Quoted 5 lines · Alasdair G Kergon
Quoted 1 line · Mike Snitzer

Maximum chunk size is 512kB,

Can you confirm that a supplied chunk size of 4294971392, for example, still gets rejected?

It gets rejected but there's a bug that device is left open after a failure. See this patch:

Properly close the device if failing because of an invalid chunk size.

---
 drivers/md/dm-exception-store.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/md/dm-exception-store.c +1 -1
Index: linux-2.6.31-fast-new/drivers/md/dm-exception-store.c
===================================================================
@@ -237,7 +237,7 @@ int dm_exception_store_create(struct dm_
 
 	r = set_chunk_size(tmp_store, argv[2], &ti->error);
 	if (r)
-		goto bad_cow;
+		goto bad_ctr;
 
 	r = type->ctr(tmp_store, 0, NULL);
 	if (r) {
maintainer
#4 lore raw

PATCH 1/1 dm snapshot: don't use zero chunksize

In reply to Alasdair G Kergon

When playing with invalid chunksizes, I have found another bug: under certain conditions (too small devices or too big chunksize) the calculated hash table size is zero. This triggers invalid C operation (a shift by -1) and causes "out of memory" messages --- when there's really nothing out of memory.

This patch changes it to use minimum 64 hashsize.

Mikulas

---

Under some special conditions (too big chunk size or zero-sized device), the resulting hash_size is calculated as zero.

rounddown_pow_of_two(0) is undefined operation (it expands to shift by -1). And init_exception_table with zero argument would fail with -ENOMEM.

This patch makes minimum chunk size 64, just like for pending exception table.

---
 drivers/md/dm-snap.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/md/dm-snap.c +2 -0
Index: linux-2.6.31-fast-new/drivers/md/dm-snap.c
===================================================================
@@ -570,6 +570,8 @@ static int init_hash_tables(struct dm_sn
 	hash_size = min(origin_dev_size, cow_dev_size) >> s->store->chunk_shift;
 	hash_size = min(hash_size, max_buckets);
 
+	if (hash_size < 64)
+		hash_size = 64;
 	hash_size = rounddown_pow_of_two(hash_size);
 	if (init_exception_table(&s->complete, hash_size,
 				 DM_CHUNK_CONSECUTIVE_BITS))
Reviewed-by on v1 1/1
#3 lore raw

Re: [PATCH] fix bug on invalid chunksize

In reply to Mikulas Patocka

good catch.

brassow

Quoted 36 lines · Mikulas Patocka
Quoted 5 lines · Alasdair G Kergon
Quoted 1 line · Mike Snitzer

Maximum chunk size is 512kB,

Can you confirm that a supplied chunk size of 4294971392, for example, still gets rejected?

It gets rejected but there's a bug that device is left open after a failure. See this patch:

Properly close the device if failing because of an invalid chunk size.

---
drivers/md/dm-exception-store.c | 2 +-

1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux-2.6.31-fast-new/drivers/md/dm-exception-store.c
===================================================================
--- linux-2.6.31-fast-new.orig/drivers/md/dm-exception-store.c

2009-10-07 00:05:28.000000000 +0200

+++ linux-2.6.31-fast-new/drivers/md/dm-exception-store.c 2009-10-07
00:05:35.000000000 +0200

drivers/md/dm-exception-store.c +1 -1
@@ -237,7 +237,7 @@ int dm_exception_store_create(struct dm_
 
 	r = set_chunk_size(tmp_store, argv[2], &ti->error);
 	if (r)
-		goto bad_cow;
+		goto bad_ctr;
 
 	r = type->ctr(tmp_store, 0, NULL);
 	if (r) {
#5 lore raw

Re: [PATCH] dm snapshot: don't use zero chunksize

In reply to Mikulas Patocka
Quoted 7 lines · Mikulas Patocka

Under some special conditions (too big chunk size or zero-sized device), the resulting hash_size is calculated as zero.

rounddown_pow_of_two(0) is undefined operation (it expands to shift by -1). And init_exception_table with zero argument would fail with -ENOMEM.

This patch makes minimum chunk size 64, just like for pending exception table.

Could we have some more specific information in this patch header?

How does a "too big" chunk size happen? And what is "too big"?

How does a zero-sized device happen and how can this code do anything sensible if it encounters one - shouldn't it fail with an error instead? Or do we still have a problem with the sequence in which dm ioctls are being issued?

Alasdairs

maintainer
#6 lore raw

Re: [PATCH] dm snapshot: don't use zero chunksize

Quoted 4 lines · Alasdair G Kergon
Quoted 3 lines · Mikulas Patocka

Under some special conditions (too big chunk size or zero-sized device), the resulting hash_size is calculated as zero.

Quoted 8 lines
Quoted 4 lines

rounddown_pow_of_two(0) is undefined operation (it expands to shift by -1). And init_exception_table with zero argument would fail with -ENOMEM.

This patch makes minimum chunk size 64, just like for pending exception table.

Could we have some more specific information in this patch header?

How does a "too big" chunk size happen? And what is "too big"?

The problem is if you have bigger chunk size than the origin device (this is valid, but useless). Then, hash_size is set to zero and false allocation failure is reported.

I think bug 502965 is caused by this --- Milan tried to create a 10k volume with 16k chunksize.

Quoted 4 lines

How does a zero-sized device happen and how can this code do anything sensible if it encounters one - shouldn't it fail with an error instead? Or do we still have a problem with the sequence in which dm ioctls are being issued?

I don't know if zero-sized device can happen with lvm. I think not. It could be only created manually with dmsetup. Snapshot of a zero-sized device may be still considered allowed.

Quoted 1 line

Alasdairs

Mikulas