Re: [RFC] sunrpc: Fix race between work-queue and rpc_killall_tasks.

[Ben Greear <greearb@candelatech.com>]

On 07/06/2011 04:45 PM, Trond Myklebust wrote:
> On Wed, 2011-07-06 at 15:49 -0700, greearb@candelatech.com wrote:
>> From: Ben Greear<greearb@candelatech.com>
>>
>> The rpc_killall_tasks logic is not locked against
>> the work-queue thread, but it still directly modifies
>> function pointers and data in the task objects.
>>
>> This patch changes the killall-tasks logic to set a flag
>> that tells the work-queue thread to terminate the task
>> instead of directly calling the terminate logic.
>>
>> Signed-off-by: Ben Greear<greearb@candelatech.com>
>> ---
>>
>> NOTE:  This needs review, as I am still struggling to understand
>> the rpc code, and it's quite possible this patch either doesn't
>> fully fix the problem or actually causes other issues.  That said,
>> my nfs stress test seems to run a bit more stable with this patch applied.
>
> Yes, but I don't see why you are adding a new flag, nor do I see why we
> want to keep checking for that flag in the rpc_execute() loop.
> rpc_killall_tasks() is not a frequent operation that we want to optimise
> for.
>
> How about the following instead?
>
> 8<----------------------------------------------------------------------------------
>  From ecb7244b661c3f9d2008ef6048733e5cea2f98ab Mon Sep 17 00:00:00 2001
> From: Trond Myklebust<Trond.Myklebust@netapp.com>
> Date: Wed, 6 Jul 2011 19:44:52 -0400
> Subject: [PATCH] SUNRPC: Fix a race between work-queue and rpc_killall_tasks
>
> Since rpc_killall_tasks may modify the rpc_task's tk_action field
> without any locking, we need to be careful when dereferencing it.
>
> Reported-by: Ben Greear<greearb@candelatech.com>
> Signed-off-by: Trond Myklebust<Trond.Myklebust@netapp.com>

I've been testing this for 4+ hours, and it seems to fix the problem.  We'll
continue to burn on it for a day or two just in case we're
getting (un)lucky in our testing.

Thanks,
Ben

> ---
>   net/sunrpc/sched.c |   27 +++++++++++----------------
>   1 files changed, 11 insertions(+), 16 deletions(-)
>
> diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
> index a27406b..4814e24 100644
> --- a/net/sunrpc/sched.c
> +++ b/net/sunrpc/sched.c
> @@ -616,30 +616,25 @@ static void __rpc_execute(struct rpc_task *task)
>   	BUG_ON(RPC_IS_QUEUED(task));
>
>   	for (;;) {
> +		void (*do_action)(struct rpc_task *);
>
>   		/*
> -		 * Execute any pending callback.
> +		 * Execute any pending callback first.
>   		 */
> -		if (task->tk_callback) {
> -			void (*save_callback)(struct rpc_task *);
> -
> -			/*
> -			 * We set tk_callback to NULL before calling it,
> -			 * in case it sets the tk_callback field itself:
> -			 */
> -			save_callback = task->tk_callback;
> -			task->tk_callback = NULL;
> -			save_callback(task);
> -		} else {
> +		do_action = task->tk_callback;
> +		task->tk_callback = NULL;
> +		if (do_action == NULL) {
>   			/*
>   			 * Perform the next FSM step.
> -			 * tk_action may be NULL when the task has been killed
> -			 * by someone else.
> +			 * tk_action may be NULL if the task has been killed.
> +			 * In particular, note that rpc_killall_tasks may
> +			 * do this at any time, so beware when dereferencing.
>   			 */
> -			if (task->tk_action == NULL)
> +			do_action = task->tk_action;
> +			if (do_action == NULL)
>   				break;
> -			task->tk_action(task);
>   		}
> +		do_action(task);
>
>   		/*
>   		 * Lockless check for whether task is sleeping or not.


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com