[PATCH] Btrfs: fix extent state leak on failed nodatasum reads

[PATCH] Btrfs: fix extent state leak on failed nodatasum reads

[Jan Schmidt]

When encountering an EIO while reading from a nodatasum extent, we
insert an error record into the inode's failure tree.
btrfs_readpage_end_io_hook returns early for nodatasum inodes. We'd
better clear the failure tree in that case, otherwise the kernel
complains about

	BUG extent_state: Objects remaining on kmem_cache_close()

on rmmod.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
---
 fs/btrfs/inode.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 870869a..9444551 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1970,7 +1970,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 	}
 
 	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
-		return 0;
+		goto good;
 
 	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
 	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
-- 
1.7.3.4

Re: [PATCH] Btrfs: fix extent state leak on failed nodatasum reads

[Jan Schmidt]

Hi Chris,

On 04.05.2011 16:18, Jan Schmidt wrote:
> When encountering an EIO while reading from a nodatasum extent, we
> insert an error record into the inode's failure tree.
> btrfs_readpage_end_io_hook returns early for nodatasum inodes. We'd
> better clear the failure tree in that case, otherwise the kernel
> complains about
> 
> 	BUG extent_state: Objects remaining on kmem_cache_close()
> 
> on rmmod.
> 
> Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
> ---
>  fs/btrfs/inode.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> index 870869a..9444551 100644
> --- a/fs/btrfs/inode.c
> +++ b/fs/btrfs/inode.c
> @@ -1970,7 +1970,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
>  	}
>  
>  	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
> -		return 0;
> +		goto good;
>  
>  	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
>  	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {

was there a specific reason for not pulling this one in? I find it neat,
straight forward and quite to the point ;-)

Just ran into that bug again and had to trace it down once more after I
rebased my upcoming patches to the for-linus branch... (Hey, it's been a
month!)

-Jan

Re: [PATCH] Btrfs: fix extent state leak on failed nodatasum reads

[Chris Mason]

Excerpts from Jan Schmidt's message of 2011-06-09 12:16:06 -0400:
> Hi Chris,
> 
> On 04.05.2011 16:18, Jan Schmidt wrote:
> > When encountering an EIO while reading from a nodatasum extent, we
> > insert an error record into the inode's failure tree.
> > btrfs_readpage_end_io_hook returns early for nodatasum inodes. We'd
> > better clear the failure tree in that case, otherwise the kernel
> > complains about
> > 
> >     BUG extent_state: Objects remaining on kmem_cache_close()
> > 
> > on rmmod.
> > 
> > Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
> > ---
> >  fs/btrfs/inode.c |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> > 
> > diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
> > index 870869a..9444551 100644
> > --- a/fs/btrfs/inode.c
> > +++ b/fs/btrfs/inode.c
> > @@ -1970,7 +1970,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
> >      }
> >  
> >      if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
> > -        return 0;
> > +        goto good;
> >  
> >      if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
> >          test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
> 
> was there a specific reason for not pulling this one in? I find it neat,
> straight forward and quite to the point ;-)
> 
> Just ran into that bug again and had to trace it down once more after I
> rebased my upcoming patches to the for-linus branch... (Hey, it's been a
> month!)

Sorry, I'll put this into the pull for the next rc.

-chris