Message-ID: <1360111333.2621.22.camel@gandalf.local.home>
Date:	Tue, 05 Feb 2013 19:42:13 -0500
From:	Steven Rostedt <rostedt@...dmis.org>
To:	Frederic Weisbecker <fweisbec@...il.com>
Cc:	Andrew Morton <akpm@...ux-foundation.org>, Jan Kara <jack@...e.cz>,
	LKML <linux-kernel@...r.kernel.org>, jslaby@...e.cz,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Ingo Molnar <mingo@...nel.org>,
	Peter Zijlstra <peterz@...radead.org>
Subject: Re: [PATCH v2] printk: Avoid softlockups in console_unlock()

On Tue, 2013-02-05 at 23:56 +0100, Frederic Weisbecker wrote:
> 2013/2/5 Andrew Morton <akpm@...ux-foundation.org>:
>  callers to printk() or by a queued work.
> >
> > I still hate the patch :(

I hate it too.

> >
> >> ...
> >>
> >> +void console_unlock(void)
> >> +{
> >> +     if (__console_unlock()) {
> >> +             /* Let worker do the rest of printing */
> >> +             schedule_work(&printk_work);
> >> +     }
> >>  }
> >
> > This creates another place from where we cannot call printk(): anywhere
> > where worker_pool.lock is held.
> >
> > And as schedule_work() can do a wakeup it creates a third reason why
> > the sched code cannot call printk (along with rq->lock taken by
> > wake_up(klogd) and rq->lock taken by up(&console_sem).  Hence
> > printk_sched().  See the lkml thread "[GIT PULL] printk: Support for
> > full dynticks mode".
> 
> Agreed, I really wish we avoid that workqueue solution.

Yes, please do not use a workqueue here.

> 
> >
> > We already have machinery for doing async tickling in printk: the
> > printk_pending stuff.  Did you consider adding another
> > PRINTK_PENDING_foo in some fashion?
> 
> Yeah, that would delay until the next timer tick (small exception
> after my patchset, if tick is stopped this will trigger as soon as
> irqs are re-enabled through a self IPI) but we can probably improve
> that behaviour. And that won't mess up with locking scenarios. The
> printk tick (or irq work after my patchset) can also re-trigger itself
> until the next tick if the batch to send to the console driver is too
> big.

Actually, we could have printk_sched() do something similar. Right now,
printk_sched() loads a per cpu buffer (wasting space, as it is seldom
used), and it sets the PRINTK_PENDING_SCHED flag. Then the next timer
interrupt, it calls printk() with the contents in the buffer.

What we could do, is have printk_sched() instead load the printk buffer.
This requires a little tweaking of the logbuf_lock to keep it from being
held when doing any wake ups, otherwise we have the AB BA deadlock issue
between logbuf_lock and the rq locks.

This patch and printk_sched() can use the same flag. If the
PRINTK_PENDING_SCHED is set (we probably should rename that to
PRINTK_PENDING_OUTPUT) it will do the console_trylock()/unlock() to
finish the print to console.

The below patch makes printk_sched() use the printk buffer instead of
its own (and removing the 512 bytes per cpu that it uses), and does a
console_trylock()/console_unlock() instead. Jan's patch could easily be
based off of this one with very little to change.

-- Steve


printk: Remove separate printk_sched buffers and use printk buf instead

To prevent deadlocks with doing a printk inside the scheduler,
printk_sched() was created. The issue is that printk has a console_sem
that it can grab and release. The release does a wake up if there's a
task pending on the sem, and this wake up grabs the rq locks that is
held in the scheduler. This leads to a possible deadlock if the wake up
uses the same rq as the one with the rq lock held already.

What printk_sched() does is to save the printk write in a per cpu buffer
and sets the PRINTK_PENDING_SCHED flag. On a timer tick, if this flag is
set, the printk() is done against the buffer.

There's a couple of issues with this approach.

1) If two printk_sched()s are called before the tick, the second one
will overwrite the first one.

2) The temporary buffer is 512 bytes and is per cpu. This is a quite a
bit of space wasted for something that is seldom used.

In order to remove this, the printk_sched() can instead use the printk
buffer instead, and delay the console_trylock()/console_unlock() to the
tick.

Because printk_sched() would then be taking the logbuf_lock, the
logbuf_lock must not be held while doing anything that may call into the
scheduler functions, which includes wake ups. Unfortunately, printk()
also has a console_sem that it uses, and on release, the
up(&console_sem) may do a wake up of any pending waiters. This must be
avoided while holding the logbuf_lock.

Luckily, there's not many places that do the unlock, or hold the
logbuf_lock. By moving things around a little, the console_sem can be
released without ever holding the logbuf_lock, and we can safely have
printk_sched() use the printk buffer directly.

Signed-off-by: Steven Rostedt <rostedt@...dmis.org>

diff --git a/kernel/printk.c b/kernel/printk.c
index 357f714..ed630fa 100644
--- a/kernel/printk.c
+++ b/kernel/printk.c
@@ -219,6 +219,9 @@ struct log {
 /*
  * The logbuf_lock protects kmsg buffer, indices, counters. It is also
  * used in interesting ways to provide interlocking in console_unlock();
+ * This can be taken within the scheduler's rq lock. It must be released
+ * before calling console_unlock() or anything else that might wake up
+ * a process.
  */
 static DEFINE_RAW_SPINLOCK(logbuf_lock);
 
@@ -1343,30 +1346,46 @@ static inline int can_use_console(unsigned int cpu)
  * interrupts disabled. It should return with 'lockbuf_lock'
  * released but interrupts still disabled.
  */
-static int console_trylock_for_printk(unsigned int cpu)
+static int console_trylock_for_printk(unsigned int cpu, bool in_sched)
 	__releases(&logbuf_lock)
 {
 	int retval = 0, wake = 0;
 
-	if (console_trylock()) {
-		retval = 1;
+	/* if called from the scheduler, we can not call up() */
+	if (in_sched)
+		goto out;
 
-		/*
-		 * If we can't use the console, we need to release
-		 * the console semaphore by hand to avoid flushing
-		 * the buffer. We need to hold the console semaphore
-		 * in order to do this test safely.
-		 */
-		if (!can_use_console(cpu)) {
-			console_locked = 0;
-			wake = 1;
-			retval = 0;
-		}
+	if (down_trylock(&console_sem))
+		goto out;
+
+	/*
+	 * If we can't use the console, we need to release
+	 * the console semaphore by hand to avoid flushing
+	 * the buffer. We need to hold the console semaphore
+	 * in order to do this test safely.
+	 */
+	if (console_suspended || !can_use_console(cpu)) {
+		wake = 1;
+		goto out;
 	}
+
+	/* console is now locked */
+
+	console_locked = 1;
+	console_may_schedule = 0;
+	mutex_acquire(&console_lock_dep_map, 0, 1, _RET_IP_);
+
+	retval = 1;
+
+out:
 	logbuf_cpu = UINT_MAX;
+	/*
+	 * The logbuf_lock must not be held when doing a wake up,
+	 * which the up(&console_sem) can do.
+	 */
+	raw_spin_unlock(&logbuf_lock);
 	if (wake)
 		up(&console_sem);
-	raw_spin_unlock(&logbuf_lock);
 	return retval;
 }
 
@@ -1495,11 +1514,17 @@ asmlinkage int vprintk_emit(int facility, int level,
 	static int recursion_bug;
 	static char textbuf[LOG_LINE_MAX];
 	char *text = textbuf;
-	size_t text_len;
+	size_t text_len = 0;
 	enum log_flags lflags = 0;
 	unsigned long flags;
 	int this_cpu;
 	int printed_len = 0;
+	bool in_sched = false;
+
+	if (level == -2) {
+		level = -1;
+		in_sched = true;
+	}
 
 	boot_delay_msec(level);
 	printk_delay();
@@ -1545,7 +1570,12 @@ asmlinkage int vprintk_emit(int facility, int level,
 	 * The printf needs to come first; we need the syslog
 	 * prefix which might be passed-in as a parameter.
 	 */
-	text_len = vscnprintf(text, sizeof(textbuf), fmt, args);
+	if (in_sched)
+		text_len = scnprintf(text, sizeof(textbuf),
+				     KERN_WARNING "[sched_delayed] ");
+
+	text_len += vscnprintf(text + text_len,
+			       sizeof(textbuf) - text_len, fmt, args);
 
 	/* mark and strip a trailing newline */
 	if (text_len && text[text_len-1] == '\n') {
@@ -1620,7 +1650,7 @@ asmlinkage int vprintk_emit(int facility, int level,
 	 * The console_trylock_for_printk() function will release 'logbuf_lock'
 	 * regardless of whether it actually gets the console semaphore or not.
 	 */
-	if (console_trylock_for_printk(this_cpu))
+	if (console_trylock_for_printk(this_cpu, in_sched))
 		console_unlock();
 
 	lockdep_on();
@@ -1962,18 +1992,20 @@ int is_console_locked(void)
 #define PRINTK_BUF_SIZE		512
 
 #define PRINTK_PENDING_WAKEUP	0x01
-#define PRINTK_PENDING_SCHED	0x02
+#define PRINTK_PENDING_OUTPUT	0x02
 
 static DEFINE_PER_CPU(int, printk_pending);
-static DEFINE_PER_CPU(char [PRINTK_BUF_SIZE], printk_sched_buf);
 
 void printk_tick(void)
 {
 	if (__this_cpu_read(printk_pending)) {
 		int pending = __this_cpu_xchg(printk_pending, 0);
-		if (pending & PRINTK_PENDING_SCHED) {
-			char *buf = __get_cpu_var(printk_sched_buf);
-			printk(KERN_WARNING "[sched_delayed] %s", buf);
+		if (pending & PRINTK_PENDING_OUTPUT) {
+			if (console_trylock())
+				console_unlock();
+			else
+				/* Try again later */
+				__this_cpu_or(printk_pending, PRINTK_PENDING_OUTPUT);
 		}
 		if (pending & PRINTK_PENDING_WAKEUP)
 			wake_up_interruptible(&log_wait);
@@ -2458,20 +2490,14 @@ late_initcall(printk_late_init);
 
 int printk_sched(const char *fmt, ...)
 {
-	unsigned long flags;
 	va_list args;
-	char *buf;
 	int r;
 
-	local_irq_save(flags);
-	buf = __get_cpu_var(printk_sched_buf);
-
 	va_start(args, fmt);
-	r = vsnprintf(buf, PRINTK_BUF_SIZE, fmt, args);
+	r = vprintk_emit(0, -2, NULL, 0, fmt, args);
 	va_end(args);
 
-	__this_cpu_or(printk_pending, PRINTK_PENDING_SCHED);
-	local_irq_restore(flags);
+	__this_cpu_or(printk_pending, PRINTK_PENDING_OUTPUT);
 
 	return r;
 }


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

Powered by blists - more mailing lists