[PATCH] Btrfs: fix wrong failed mirror_num of read-repair on raid56

[PATCH] Btrfs: fix wrong failed mirror_num of read-repair on raid56

[Liu Bo]

In raid56 senario, after trying parity recovery, we didn't set
mirror_num for btrfs_bio with failed mirror_num, hence
end_bio_extent_readpage() will report a random mirror_num in dmesg
log.

Cc: David Sterba <dsterba@suse.cz>
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
---
 fs/btrfs/volumes.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 73d56ee..be64e4a 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -6197,6 +6197,7 @@ int btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
 		} else {
 			ret = raid56_parity_recover(fs_info, bio, bbio,
 						    map_length, mirror_num, 1);
+			btrfs_io_bio(bio)->mirror_num = mirror_num;
 		}
 
 		btrfs_bio_counter_dec(fs_info);
-- 
2.5.5

Re: [PATCH] Btrfs: fix wrong failed mirror_num of read-repair on raid56

[David Sterba]

On Fri, Mar 24, 2017 at 12:13:42PM -0700, Liu Bo wrote:
> In raid56 senario, after trying parity recovery, we didn't set
> mirror_num for btrfs_bio with failed mirror_num, hence
> end_bio_extent_readpage() will report a random mirror_num in dmesg
> log.
> 
> Cc: David Sterba <dsterba@suse.cz>
> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> ---
>  fs/btrfs/volumes.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
> index 73d56ee..be64e4a 100644
> --- a/fs/btrfs/volumes.c
> +++ b/fs/btrfs/volumes.c
> @@ -6197,6 +6197,7 @@ int btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
>  		} else {
>  			ret = raid56_parity_recover(fs_info, bio, bbio,
>  						    map_length, mirror_num, 1);
> +			btrfs_io_bio(bio)->mirror_num = mirror_num;

Should the mirror be set inside raid56_parity_recover? There's another
caller, scrub_submit_raid56_bio_wait, that does not update the
bio->mirror_num. I am not sure if this is the same case though.

>  		}
>  
>  		btrfs_bio_counter_dec(fs_info);

Re: [PATCH] Btrfs: fix wrong failed mirror_num of read-repair on raid56

[Liu Bo]

On Mon, Mar 27, 2017 at 07:07:15PM +0200, David Sterba wrote:
> On Fri, Mar 24, 2017 at 12:13:42PM -0700, Liu Bo wrote:
> > In raid56 senario, after trying parity recovery, we didn't set
> > mirror_num for btrfs_bio with failed mirror_num, hence
> > end_bio_extent_readpage() will report a random mirror_num in dmesg
> > log.
> > 
> > Cc: David Sterba <dsterba@suse.cz>
> > Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> > ---
> >  fs/btrfs/volumes.c | 1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
> > index 73d56ee..be64e4a 100644
> > --- a/fs/btrfs/volumes.c
> > +++ b/fs/btrfs/volumes.c
> > @@ -6197,6 +6197,7 @@ int btrfs_map_bio(struct btrfs_fs_info *fs_info, struct bio *bio,
> >  		} else {
> >  			ret = raid56_parity_recover(fs_info, bio, bbio,
> >  						    map_length, mirror_num, 1);
> > +			btrfs_io_bio(bio)->mirror_num = mirror_num;
> 
> Should the mirror be set inside raid56_parity_recover? There's another
> caller, scrub_submit_raid56_bio_wait, that does not update the
> bio->mirror_num. I am not sure if this is the same case though.
>

scrub_submit_raid56_bio_wait doesn't need to, but yes, we could set the
mirror inside raid56_parity_recover iff @generic_io is 1.

Thanks,

-liubo

> >  		}
> >  
> >  		btrfs_bio_counter_dec(fs_info);

[PATCH v2] Btrfs: fix wrong failed mirror_num of read-repair on raid56

[Liu Bo]

In raid56 scenario, after trying parity recovery, we didn't set
mirror_num for btrfs_bio with failed mirror_num, hence
end_bio_extent_readpage() will report a random mirror_num in dmesg
log.

Cc: David Sterba <dsterba@suse.cz>
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
---
v2: Set mirror_num inside raid56_parity_recover so that all callers
can get a correct mirror_num if they need.

 fs/btrfs/raid56.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c
index 1571bf2..a333acc 100644
--- a/fs/btrfs/raid56.c
+++ b/fs/btrfs/raid56.c
@@ -2118,6 +2118,11 @@ int raid56_parity_recover(struct btrfs_fs_info *fs_info, struct bio *bio,
 	struct btrfs_raid_bio *rbio;
 	int ret;
 
+	if (generic_io) {
+		ASSERT(bbio->mirror_num == mirror_num);
+		btrfs_io_bio(bio)->mirror_num = mirror_num;
+	}
+
 	rbio = alloc_rbio(fs_info, bbio, stripe_len);
 	if (IS_ERR(rbio)) {
 		if (generic_io)
-- 
2.5.5

Re: [PATCH v2] Btrfs: fix wrong failed mirror_num of read-repair on raid56

[David Sterba]

On Wed, Mar 29, 2017 at 10:54:26AM -0700, Liu Bo wrote:
> In raid56 scenario, after trying parity recovery, we didn't set
> mirror_num for btrfs_bio with failed mirror_num, hence
> end_bio_extent_readpage() will report a random mirror_num in dmesg
> log.
> 
> Cc: David Sterba <dsterba@suse.cz>
> Signed-off-by: Liu Bo <bo.li.liu@oracle.com>

Reviewed-by: David Sterba <dsterba@suse.com>