Re: [PATCH] fs: nfs: fix a possible sleep-in-atomic-context bug in _pnfs_grab_empty_layout()

[Trond Myklebust]

On Tue, 2019-12-17 at 21:33 +0800, Jia-Ju Bai wrote:
> The filesystem may sleep while holding a spinlock.
> The function call path (from bottom to top) in Linux 4.19 is:
> 
> fs/nfs/pnfs.c, 2052: 
> 	pnfs_find_alloc_layout(GFP_KERNEL) in _pnfs_grab_empty_layout
> fs/nfs/pnfs.c, 2051: 
> 	spin_lock in _pnfs_grab_empty_layout
> 
> pnfs_find_alloc_layout(GFP_KERNEL) can sleep at runtime.
> 
> To fix this possible bug, GFP_KERNEL is replaced with GFP_ATOMIC for
> pnfs_find_alloc_layout().
> 
> This bug is found by a static analysis tool STCheck written by
> myself.
> 
> Signed-off-by: Jia-Ju Bai <baijiaju1990@xxxxxxxxx>
> ---
>  fs/nfs/pnfs.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
> index cec3070ab577..cfbe170f0651 100644
> --- a/fs/nfs/pnfs.c
> +++ b/fs/nfs/pnfs.c
> @@ -2138,7 +2138,7 @@ _pnfs_grab_empty_layout(struct inode *ino,
> struct nfs_open_context *ctx)
>  	struct pnfs_layout_hdr *lo;
>  
>  	spin_lock(&ino->i_lock);
> -	lo = pnfs_find_alloc_layout(ino, ctx, GFP_KERNEL);
> +	lo = pnfs_find_alloc_layout(ino, ctx, GFP_ATOMIC);
>  	if (!lo)
>  		goto out_unlock;
>  	if (!test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags))

I'm not seeing why this is necessary. As far as I can see,
pnfs_find_alloc_layout() will release the ino->i_lock before sleeping.

False positive?

-- 
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@xxxxxxxxxxxxxxx