Skip to contentSkip to conversation

printk: CON_PRINTBUFFER console registration is a bit racy

PATCH Superseded Sergey Senozhatsky 4 participants 7 messages Sep 14, 2018 – Sep 28, 2018 Printk

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 printk: CON_PRINTBUFFER console registration is a bit racy

CON_PRINTBUFFER console registration requires us to do several preparation steps:
- Rollback console_seq to replay logbuf messages which were already seen on other consoles;
- Set exclusive_console flag so console_unlock() will ->write() logbuf messages only to the exclusive_console driver.

The way we do it, however, is a bit racy

	logbuf_lock_irqsave(flags);
	console_seq = syslog_seq;
	console_idx = syslog_idx;
	logbuf_unlock_irqrestore(flags);
						<< preemption enabled
						<< irqs enabled
	exclusive_console = newcon;
	console_unlock();

We rollback console_seq under logbuf_lock with IRQs disabled, but we set exclusive_console with local IRQs enabled and logbuf unlocked. If the system oops-es or panic-s before we set exclusive_console - and given that we have IRQs and preemption enabled there is such a possibility - we will re-play all logbuf messages to every registered console, which may be a bit annoying and time consuming.

Move exclusive_console assignment under the same IRQs-disabled and logbuf_lock protected section where we rollback console_seq.

---
 kernel/printk/printk.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/printk/printk.c +1 -1
index 992bb6bb7ac2..c52a986a72b3 100644
@@ -2705,7 +2705,6 @@ void register_console(struct console *newcon)
 		logbuf_lock_irqsave(flags);
 		console_seq = syslog_seq;
 		console_idx = syslog_idx;
-		logbuf_unlock_irqrestore(flags);
 		/*
 		 * We're about to replay the log buffer.  Only do this to the
 		 * just-registered console to avoid excessive message spam to
@@ -2713,6 +2712,7 @@ void register_console(struct console *newcon)
 		 */
 		exclusive_console = newcon;
 		exclusive_console_stop_seq = console_seq;
+		logbuf_unlock_irqrestore(flags);
 	}
 	console_unlock();
 	console_sysfs_notify();
Signature · 1 line
2.19.0
maintainer
#1 lore raw
Quoted 46 lines · Sergey Senozhatsky

CON_PRINTBUFFER console registration requires us to do several preparation steps:
- Rollback console_seq to replay logbuf messages which were already seen on other consoles;
- Set exclusive_console flag so console_unlock() will ->write() logbuf messages only to the exclusive_console driver.

The way we do it, however, is a bit racy

	logbuf_lock_irqsave(flags);
	console_seq = syslog_seq;
	console_idx = syslog_idx;
	logbuf_unlock_irqrestore(flags);
						<< preemption enabled
						<< irqs enabled
	exclusive_console = newcon;
	console_unlock();

We rollback console_seq under logbuf_lock with IRQs disabled, but we set exclusive_console with local IRQs enabled and logbuf unlocked. If the system oops-es or panic-s before we set exclusive_console - and given that we have IRQs and preemption enabled there is such a possibility - we will re-play all logbuf messages to every registered console, which may be a bit annoying and time consuming.

Move exclusive_console assignment under the same IRQs-disabled and logbuf_lock protected section where we rollback console_seq.

---
 kernel/printk/printk.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/printk/printk.c +0 -1
index 992bb6bb7ac2..c52a986a72b3 100644
@@ -2705,7 +2705,6 @@ void register_console(struct console *newcon)
 		logbuf_lock_irqsave(flags);
 		console_seq = syslog_seq;
 		console_idx = syslog_idx;
-		logbuf_unlock_irqrestore(flags);
 		/*
 		 * We're about to replay the log buffer.  Only do this to the
 		 * just-registered console to avoid excessive message spam to

kernel/printk/printk.c void register_console(struct console *newcon) · line 2715

Quoted 4 lines
kernel/printk/printk.c +1 -0
@@ -2713,6 +2712,7 @@ void register_console(struct console *newcon)
 		 */
 		exclusive_console = newcon;
 		exclusive_console_stop_seq = console_seq;
+		logbuf_unlock_irqrestore(flags);

This would make a false feeling that these variables need to be synchronized using logbug_lock. It might either confuse people or it could get removed during a code clean up.

A better solution would be to explicitly disable preemption around the entire section and add a comment.

Well, I am not sure if it is worth the code complexity.

Best Regards,
Petr

#2 lore raw
Quoted 3 lines · Petr Mladek

Well, I am not sure if it is worth the code complexity.

Well, I don't think we need to bother that much here. Besides, exclusive_console is cleared under logbuf_lock with preemption disabled now. So we set it under logbuf_lock and !irq and we clear it under logbuf_lock and !irq. Looks quite OK to me.

	-ss
#3 lore raw
Quoted 9 lines · Sergey Senozhatsky
Quoted 3 lines · Petr Mladek

Well, I am not sure if it is worth the code complexity.

Well, I don't think we need to bother that much here. Besides, exclusive_console is cleared under logbuf_lock with preemption disabled now. So we set it under logbuf_lock and !irq and we clear it under logbuf_lock and !irq. Looks quite OK to me.

Petr, Steven,
any opinions? The patch is rather trivial.

	-ss
maintainer
#4 lore raw
In reply to Sergey Senozhatsky
Quoted 9 lines · Sergey Senozhatsky
Quoted 3 lines · Petr Mladek

Well, I am not sure if it is worth the code complexity.

Well, I don't think we need to bother that much here. Besides, exclusive_console is cleared under logbuf_lock with preemption disabled now. So we set it under logbuf_lock and !irq and we clear it under logbuf_lock and !irq. Looks quite OK to me.

I wanted to say that you moved exclusive_console handling under a locked section from a reason. This reason is far from clear from the code.

If you really want this change, please add a comment, for example:

/*
* Set exclusive_console still with disabled interrupts to
* reduce race window with eventual console_flush_on_panic()
* that ignores console_lock.
*/

I am not against the change. It makes some sense and it does not break anything. It is just not obvious and might either get easily lost again or it might cause confusion. I mean that it might cause false feeling that exclusive_console is synchronized by logbuf_lock.

printk code already is complex enough without this subtle tricks.

Best Regards,
Petr

#5 lore raw
Quoted 8 lines · On Wed, 26 Sep 2018 13:37:15 +0200 Petr Mladek <[email protected]> wrote:

If you really want this change, please add a comment, for example:

/*
* Set exclusive_console still with disabled interrupts to
* reduce race window with eventual console_flush_on_panic()
* that ignores console_lock.
*/

Quoted 8 lines

I am not against the change. It makes some sense and it does not break anything. It is just not obvious and might either get easily lost again or it might cause confusion. I mean that it might cause false feeling that exclusive_console is synchronized by logbuf_lock.

printk code already is complex enough without this subtle tricks.

I'm fine with the change and agree that a comment should be added.

-- Steve

#6 lore raw
In reply to Petr Mladek
Quoted 5 lines · Petr Mladek

/*
* Set exclusive_console still with disabled interrupts to
* reduce race window with eventual console_flush_on_panic()
* that ignores console_lock.
*/

Looks good to me.

Quoted 2 lines

I am not against the change. It makes some sense and it does not break anything. It is just not obvious

OK. Thanks.
Will send v2.

	-ss