From:	 David Woodhouse <dwmw2@infradead.org>
To:	 Alan Cox <alan@lxorguk.ukuu.org.uk>
Subject: Re: Poke in the eye regarding sleep_on
Date:	 Thu, 29 Nov 2001 11:09:11 +0000
Cc:	 tao@acc.umu.se (David Weinehall),
	 linux-kernel@vger.kernel.org (Linux-Kernel Mailing List),
	 torvalds@transmeta.com


alan@lxorguk.ukuu.org.uk said:
> > Since the 2.5 work now has begun, I should remind you Alan that you
> > promised on this listto take care of getting rid of sleep_on for good
> > just as soon as this tree opened.

> That was David Woodhouse.. but he is right

There are still cases where it's just about OK to use sleep_on(). If you
_and_ your waker will hold the BKL, or if you call it with IRQs disabled and
are woken from an ISR, which is horrible but does currently work. Also if
you just don't _care_ if you miss a wakeup sometimes.

ISTR much filesystem code falls into the former category at the moment.
 
We _should_ remove sleep_on and friends, but let's start with this patch, 
which allows us to clean up the filesystem code later...

Note that I haven't put the check in the _timeout versions. Maybe we should.

Index: kernel/sched.c
===================================================================
RCS file: /inst/cvs/linux/kernel/sched.c,v
retrieving revision 1.4.2.57
diff -u -r1.4.2.57 sched.c
--- kernel/sched.c	22 Nov 2001 08:41:45 -0000	1.4.2.57
+++ kernel/sched.c	29 Nov 2001 10:57:48 -0000
@@ -777,6 +777,12 @@
 	wait_queue_t wait;			\
 	init_waitqueue_entry(&wait, current);
 
+#define SLEEP_ON_CHECK										\
+	if (current->lock_depth == -1) {							\
+		printk(KERN_WARNING __FUNCTION__ " called without BKL. Probable race.\n");	\
+		BUG();										\
+	}	
+		
 #define	SLEEP_ON_HEAD					\
 	wq_write_lock_irqsave(&q->lock,flags);		\
 	__add_wait_queue(q, &wait);			\
@@ -790,6 +796,7 @@
 void interruptible_sleep_on(wait_queue_head_t *q)
 {
 	SLEEP_ON_VAR
+	SLEEP_ON_CHECK
 
 	current->state = TASK_INTERRUPTIBLE;
 
@@ -814,6 +821,7 @@
 void sleep_on(wait_queue_head_t *q)
 {
 	SLEEP_ON_VAR
+	SLEEP_ON_CHECK
 	
 	current->state = TASK_UNINTERRUPTIBLE;
 



--
dwmw2


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/