Re: [RFC][PATCH 0/7] nested sleeps, fixes and debug infra

[Peter Zijlstra]

On Wed, Aug 06, 2014 at 11:51:29AM +0400, Ilya Dryomov wrote:

> OK, this one is a bit different.
> 
> WARNING: CPU: 1 PID: 1744 at kernel/sched/core.c:7104 __might_sleep+0x58/0x90()
> do not call blocking ops when !TASK_RUNNING; state=1 set at [<ffffffff81070e10>] prepare_to_wait+0x50 /0xa0

>  [<ffffffff8105bc38>] __might_sleep+0x58/0x90
>  [<ffffffff8148c671>] lock_sock_nested+0x31/0xb0
>  [<ffffffff81498aaa>] sk_stream_wait_memory+0x18a/0x2d0

Urgh, tedious. Its not an actual bug as is. Due to the condition check
in sk_wait_event() we can call lock_sock() with ->state != TASK_RUNNING.

I'm not entirely sure what the cleanest way is to make this go away.
Possibly something like so:

---
 include/net/sock.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/include/net/sock.h b/include/net/sock.h
index 156350745700..37902176c5ab 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -886,6 +886,7 @@ static inline void sock_rps_reset_rxhash(struct sock *sk)
 		if (!__rc) {						\
 			*(__timeo) = schedule_timeout(*(__timeo));	\
 		}							\
+		__set_current_state(TASK_RUNNING);			\
 		lock_sock(__sk);					\
 		__rc = __condition;					\
 		__rc;							\

Attachment: pgpJFqjT6yI01.pgp
Description: PGP signature