linux-kernel - Re: [PATCH] fs: nfs: fix a possible sleep-in-atomic-context bug in _pnfs_grab_empty

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <ff4e1443d70acc88bba68f87650c7b5118c63f2b.camel@hammerspace.com>
Date:   Tue, 17 Dec 2019 14:37:37 +0000
From:   Trond Myklebust <trondmy@...merspace.com>
To:     "baijiaju1990@...il.com" <baijiaju1990@...il.com>,
        "anna.schumaker@...app.com" <anna.schumaker@...app.com>
CC:     "linux-nfs@...r.kernel.org" <linux-nfs@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] fs: nfs: fix a possible sleep-in-atomic-context bug in
 _pnfs_grab_empty_layout()

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@...il.com>
> ---
>  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@...merspace.com