Skip to contentSkip to conversation

Btrfs: fix possible panic on unmount

PATCH In review Josef Bacik 3 participants 4 messages Nov 13, 2009 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Btrfs: fix possible panic on unmount

We can race with the unmount of an fs and the stopping of a kthread where we will free the block group before we're done using it. The reason for this is because we do not hold a reference on the block group while its caching, since the allocator drops its reference once it exits or moves on to the next block group. This patch fixes the problem by taking a reference to the block group before we start caching and dropping it when we're done to make sure all accesses to the block group are safe. Thanks,

---
 fs/btrfs/extent-tree.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)
fs/btrfs/extent-tree.c +3 -0
index 2a4cdce..197bc1b 100644
@@ -399,6 +399,8 @@ err:
 
 	put_caching_control(caching_ctl);
 	atomic_dec(&block_group->space_info->caching_threads);
+	btrfs_put_block_group(block_group);
+
 	return 0;
 }
 
@@ -439,6 +441,7 @@ static int cache_block_group(struct btrfs_block_group_cache *cache)
 	up_write(&fs_info->extent_commit_sem);
 
 	atomic_inc(&cache->space_info->caching_threads);
+	atomic_inc(&cache->count);
 
 	tsk = kthread_run(caching_kthread, cache, "btrfs-cache-%llu\n",
 			  cache->key.objectid);
Signature · 1 line
1.5.4.3
#1 lore raw
Quoted 1 line · Josef Bacik

We can race with the unmount of an fs and the stopping of a kthread w=

here we

Quoted 1 line

will free the block group before we're done using it. =A0The reason f=

or this is

Quoted 1 line

because we do not hold a reference on the block group while its cachi=

ng, since

Quoted 1 line

the allocator drops its reference once it exits or moves on to the ne=

xt block

Quoted 1 line

group. =A0This patch fixes the problem by taking a reference to the b=

lock group

Quoted 1 line

before we start caching and dropping it when we're done to make sure =

all

fs/btrfs/extent-tree.c static int cache_block_group(struct btrfs_block_g=

Quoted 21 lines

accesses to the block group are safe. =A0Thanks,

---
=A0fs/btrfs/extent-tree.c | =A0 =A03 +++ =A01 files changed, 3 insertions(+), 0 deletions(-)

fs/btrfs/extent-tree.c +2 -0
index 2a4cdce..197bc1b 100644
@@ -399,6 +399,8 @@ err:
 
 =A0 =A0 =A0 =A0put_caching_control(caching_ctl);
 =A0 =A0 =A0 =A0atomic_dec(&block_group->space_info->caching_threads);
+ =A0 =A0 =A0 btrfs_put_block_group(block_group);
+
 =A0 =A0 =A0 =A0return 0;
 =A0}
 
@@ -439,6 +441,7 @@ static int cache_block_group(struct btrfs_block_g=

roup_cache *cache)

fs/btrfs/extent-tree.c static int cache_block_group(struct btrfs_block_g= · line 444

Quoted 4 lines
fs/btrfs/extent-tree.c +1 -0
 =A0 =A0 =A0 =A0up_write(&fs_info->extent_commit_sem);
 
 =A0 =A0 =A0 =A0atomic_inc(&cache->space_info->caching_threads);
+ =A0 =A0 =A0 atomic_inc(&cache->count);

It would be slightly more informative/stylish if atomic_inc(&cache->count) is replaced by a cal to
static inline void btrfs_get_block_group(struct btrfs_block_group_cache=
*cache)

{
	atomic_inc(&cache->count);
}

Regards,
Andrey

Quoted 2 lines

=A0 =A0 =A0 =A0tsk =3D kthread_run(caching_kthread, cache, "btrfs-cac=

he-%llu\n",

Quoted 1 line

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0cache->key.objecti=

d);

Quoted 5 lines

--
1.5.4.3

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs=

" in

Quoted 3 lines

the body of a message to [email protected]
More majordomo info at =A0http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" = in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
In reply to Josef Bacik
Quoted 1 line · Josef Bacik

We can race with the unmount of an fs and the stopping of a kthread w=

here we

Quoted 1 line

will free the block group before we're done using it. =A0The reason f=

or this is

Quoted 1 line

because we do not hold a reference on the block group while its cachi=

ng, since

Quoted 1 line

the allocator drops its reference once it exits or moves on to the ne=

xt block

Quoted 1 line

group. =A0This patch fixes the problem by taking a reference to the b=

lock group

Quoted 1 line

before we start caching and dropping it when we're done to make sure =

all

fs/btrfs/extent-tree.c static int cache_block_group(struct btrfs_block_g=

Quoted 21 lines

accesses to the block group are safe. =A0Thanks,

---
=A0fs/btrfs/extent-tree.c | =A0 =A03 +++ =A01 files changed, 3 insertions(+), 0 deletions(-)

fs/btrfs/extent-tree.c +2 -0
index 2a4cdce..197bc1b 100644
@@ -399,6 +399,8 @@ err:
 
 =A0 =A0 =A0 =A0put_caching_control(caching_ctl);
 =A0 =A0 =A0 =A0atomic_dec(&block_group->space_info->caching_threads);
+ =A0 =A0 =A0 btrfs_put_block_group(block_group);
+
 =A0 =A0 =A0 =A0return 0;
 =A0}
 
@@ -439,6 +441,7 @@ static int cache_block_group(struct btrfs_block_g=

roup_cache *cache)

Quoted 6 lines
fs/btrfs/extent-tree.c +1 -0
 =A0 =A0 =A0 =A0up_write(&fs_info->extent_commit_sem);
 
 =A0 =A0 =A0 =A0atomic_inc(&cache->space_info->caching_threads);
+ =A0 =A0 =A0 atomic_inc(&cache->count);
 

=A0 =A0 =A0 =A0tsk =3D kthread_run(caching_kthread, cache, "btrfs-cac=

he-%llu\n",

Quoted 1 line

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0cache->key.objecti=

d);

Quoted 1 line

--

This won't work since btrfs_free_block_groups call kfree without checking the reference
count. I think the correct way to fix the race is waiting until caching threads exit

Yan, Zheng
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" = in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#3 lore raw
Quoted 1 line · Zheng

On Sat, Nov 14, 2009 at 1:06 AM, Josef Bacik <[email protected]> wrote=

:

Quoted 1 line
Quoted 1 line

We can race with the unmount of an fs and the stopping of a kthread=

where we

Quoted 1 line
Quoted 1 line

will free the block group before we're done using it. =A0The reason=

for this is

Quoted 1 line
Quoted 1 line

because we do not hold a reference on the block group while its cac=

hing, since

Quoted 1 line
Quoted 1 line

the allocator drops its reference once it exits or moves on to the =

next block

Quoted 1 line
Quoted 1 line

group. =A0This patch fixes the problem by taking a reference to the=

block group

Quoted 1 line
Quoted 1 line

before we start caching and dropping it when we're done to make sur=

e all

Quoted 7 lines
Quoted 7 lines

accesses to the block group are safe. =A0Thanks,

---
=A0fs/btrfs/extent-tree.c | =A0 =A03 +++ =A01 files changed, 3 insertions(+), 0 deletions(-)

Quoted 8 lines
Quoted 8 lines
fs/btrfs/extent-tree.c +0 -0
index 2a4cdce..197bc1b 100644
@@ -399,6 +399,8 @@ err:
 
 =A0 =A0 =A0 =A0put_caching_control(caching_ctl);

=A0 =A0 =A0 =A0atomic_dec(&block_group->space_info->caching_threads=

);

Quoted 6 lines

fs/btrfs/extent-tree.c static int cache_block_group(struct btrfs_block=

Quoted 6 lines

+ =A0 =A0 =A0 btrfs_put_block_group(block_group);
+
=A0 =A0 =A0 =A0return 0;
=A0}

fs/btrfs/extent-tree.c +0 -0
@@ -439,6 +441,7 @@ static int cache_block_group(struct btrfs_block=

_group_cache *cache)

Quoted 6 lines
Quoted 6 lines
fs/btrfs/extent-tree.c +1 -0
 =A0 =A0 =A0 =A0up_write(&fs_info->extent_commit_sem);
 
 =A0 =A0 =A0 =A0atomic_inc(&cache->space_info->caching_threads);
+ =A0 =A0 =A0 atomic_inc(&cache->count);
 

=A0 =A0 =A0 =A0tsk =3D kthread_run(caching_kthread, cache, "btrfs-c=

ache-%llu\n",

Quoted 1 line
Quoted 1 line

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0cache->key.objec=

tid);

Quoted 7 lines
Quoted 1 line

--

=20
This won't work since btrfs_free_block_groups call kfree without checking the reference
count. I think the correct way to fix the race is waiting until caching threads exit
=20

We do already do that, but the thing is we mark the block group as bein= g
finished, and then go on and do other things with it, hence the race. = The
correct thing to do is do a put in btrfs_free_block_groups so this prob= lem
doesn't happen. I will fix that. Thanks,

Josef
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" = in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html