[PATCH] nfsd: Fix slot wake up race in the nfsv4.1 callback code

[PATCH] nfsd: Fix slot wake up race in the nfsv4.1 callback code

[Trond Myklebust]

The currect code for nfsd41_cb_get_slot() and nfsd4_cb_done() has no
locking in order to guarantee atomicity, and so allows for races of
the form.

Task 1                                  Task 2
======                                  ======
if (test_and_set_bit(0) != 0) {
                                        clear_bit(0)
                                        rpc_wake_up_next(queue)
        rpc_sleep_on(queue)
        return false;
}

This patch breaks the race condition by adding a retest of the bit
after the call to rpc_sleep_on().

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: stable@vger.kernel.org
---
 fs/nfsd/nfs4callback.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
index ed2b1151b171..7cbdf1b2e4ab 100644
--- a/fs/nfsd/nfs4callback.c
+++ b/fs/nfsd/nfs4callback.c
@@ -774,8 +774,12 @@ static bool nfsd41_cb_get_slot(struct nfs4_client *clp, struct rpc_task *task)
 {
 	if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
 		rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
-		dprintk("%s slot is busy\n", __func__);
-		return false;
+		/* Race breaker */
+		if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
+			dprintk("%s slot is busy\n", __func__);
+			return false;
+		}
+		rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
 	}
 	return true;
 }
-- 
1.9.3

Re: [PATCH] nfsd: Fix slot wake up race in the nfsv4.1 callback code

[Bruce Fields]

On Wed, Nov 19, 2014 at 12:47:50PM -0500, Trond Myklebust wrote:
> The currect code for nfsd41_cb_get_slot() and nfsd4_cb_done() has no
> locking in order to guarantee atomicity, and so allows for races of
> the form.

Makes sense, thanks.  Applying for 3.18 and stable.

--b.

> 
> Task 1                                  Task 2
> ======                                  ======
> if (test_and_set_bit(0) != 0) {
>                                         clear_bit(0)
>                                         rpc_wake_up_next(queue)
>         rpc_sleep_on(queue)
>         return false;
> }
> 
> This patch breaks the race condition by adding a retest of the bit
> after the call to rpc_sleep_on().
> 
> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
> Cc: stable@vger.kernel.org
> ---
>  fs/nfsd/nfs4callback.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
> index ed2b1151b171..7cbdf1b2e4ab 100644
> --- a/fs/nfsd/nfs4callback.c
> +++ b/fs/nfsd/nfs4callback.c
> @@ -774,8 +774,12 @@ static bool nfsd41_cb_get_slot(struct nfs4_client *clp, struct rpc_task *task)
>  {
>  	if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
>  		rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
> -		dprintk("%s slot is busy\n", __func__);
> -		return false;
> +		/* Race breaker */
> +		if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
> +			dprintk("%s slot is busy\n", __func__);
> +			return false;
> +		}
> +		rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
>  	}
>  	return true;
>  }
> -- 
> 1.9.3
>