Re: [PATCH] Btrfs: deal with unexpected return value in flush_space

[Liu Bo <bo.li.liu@oracle.com>]

On Thu, Jul 28, 2016 at 03:36:53PM +0000, Holger Hoffstätte wrote:
> On Wed, 27 Jul 2016 18:42:03 -0700, Liu Bo wrote:
> 
> > Function start_transaction() can return ERR_PTR(1) when flush is
> > BTRFS_RESERVE_FLUSH_LIMIT, so the call graph is
> > 
> > start_transaction (return ERR_PTR(1))
> >   -> btrfs_block_rsv_add (return 1)
> >      -> reserve_metadata_bytes (return 1)
> >         -> flush_space (return 1)
> >            -> do_chunk_alloc  (return 1)
> > 
> > With BTRFS_RESERVE_FLUSH_LIMIT, if flush_space is already on the
> > flush_state of ALLOC_CHUNK and it successfully allocates a new
> > chunk, then instead of trying to reserve space again,
> > reserve_metadata_bytes returns 1 immediately.
> > 
> > Eventually the callers who call start_transaction() usually just
> > do the IS_ERR() check which ERR_PTR(1) can pass, then it'll get
> > a panic when dereferencing a pointer which is ERR_PTR(1).
> > 
> > This makes flush_space() translate 'ret = 1' to 'ret = 0'.
> > 
> > Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
> > ---
> > We found this 'NULL pointer dereference' on an old 3.8 kernel but
> > it's not going to happen on the upstream since there is no caller
> > of btrfs_start_transaction_lflush().
> > 
> >  fs/btrfs/extent-tree.c | 11 ++++++++++-
> >  1 file changed, 10 insertions(+), 1 deletion(-)
> > 
> > diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
> > index 7a35c9d..a00fb67 100644
> > --- a/fs/btrfs/extent-tree.c
> > +++ b/fs/btrfs/extent-tree.c
> > @@ -4457,6 +4457,15 @@ void check_system_chunk(struct btrfs_trans_handle *trans,
> >  	}
> >  }
> >  
> > +/*
> > + * If force is CHUNK_ALLOC_FORCE:
> > + *    - return 1 if it successfully allocates a chunk,
> > + *    - return errors including -ENOSPC otherwise.
> > + * If force is NOT CHUNK_ALLOC_FORCE:
> > + *    - return 0 if it doesn't need to allocate a new chunk,
> > + *    - return 1 if it successfully allocates a chunk,
> > + *    - return errors including -ENOSPC otherwise.
> > + */
> >  static int do_chunk_alloc(struct btrfs_trans_handle *trans,
> >  			  struct btrfs_root *extent_root, u64 flags, int force)
> >  {
> > @@ -4857,7 +4866,7 @@ static int flush_space(struct btrfs_root *root,
> >  				     btrfs_get_alloc_profile(root, 0),
> >  				     CHUNK_ALLOC_NO_FORCE);
> >  		btrfs_end_transaction(trans, root);
> > -		if (ret == -ENOSPC)
> > +		if (ret == -ENOSPC || ret == 1)
> >  			ret = 0;
> >  		break;
> >  	case COMMIT_TRANS:
> > -- 
> > 2.5.5
> 
> For reviewers - this came up before here:
> https://patchwork.kernel.org/patch/7778651/
> 
> Same fix basically.

Aha, I've given it my Reviewed-by.

Taking either one works for me, I can make the clarifying comment into a
seperate patch if we need to.

Thanks,

-liubo