On Thu, Apr 03, 2014 at 01:34:23PM +0800, Liu Bo wrote:
> On Wed, Apr 02, 2014 at 07:13:00PM +0200, David Sterba wrote:
> > Commit fae7f21cece9a4c181 ("btrfs: Use WARN_ON()'s return value in place of
> > WARN_ON(1)") cleaned up WARN_ON usage and in one place reversed the condition
> > that led to loads of warnings that were not supposed to occur.
> >
> > WARN_ON will trigger because it sees 'ret' though in the previous code
> > did not reach the WARN_ON below. The correct pattern is
> >
> > - if (condition)
> > + if (WARN_ON(condition))
> >
> > CC: Dulshani Gunawardhana <dulshani.gunawardhana89@xxxxxxxxx>
> > CC: <stable@xxxxxxxxxxxxxxx> # 3.13
> > Reported-by: Liu Bo <bo.li.liu@xxxxxxxxxx>
> > Signed-off-by: David Sterba <dsterba@xxxxxxx>
> > ---
> > fs/btrfs/delayed-inode.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
> > index 451b00c86f6c..098af20abd88 100644
> > --- a/fs/btrfs/delayed-inode.c
> > +++ b/fs/btrfs/delayed-inode.c
> > @@ -649,7 +649,7 @@ static int btrfs_delayed_inode_reserve_metadata(
> > goto out;
> >
> > ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
> > - if (!WARN_ON(ret))
> > + if (WARN_ON(!ret))
> > goto out;
>
> Oh sorry, I'd have to get my Reviewed-by back and give a NACK instead.
>
> With this patch, (ret = 0) triggers the WARNING, which is not right.
Thanks for catching this, you're right, my patch was wrong. I must say
the patch (fae7f21ce) made the code harder to read at some places, I
don't see much help in removing plain WARN_ON(1) at this cost.
Back to the warning flood you observed, the comment under the warning
says:
655 /*
656 * Ok this is a problem, let's just steal from the global rsv
657 * since this really shouldn't happen that often.
658 */
659 ret = btrfs_block_rsv_migrate(&root->fs_info->global_block_rsv,
660 dst_rsv, num_bytes);
so the question is why it does happen so often.
A WARN_ON_ONCE hides the severity of the problem, so I'd rather suggest
to put it under enospc_debug option so we can debug it and it does not
bother users. As this is closer to the way you were going to fix that,
I'm not sending a patch, take this as a review comment.
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html