[PATCH] vt: fix a potential race in the VT_WAITACTIVE handler

[PATCH] vt: fix a potential race in the VT_WAITACTIVE handler

[Antonino A. Daplas]

From: Michal Januszewski <spock@gentoo.org>

On a multiprocessor machine the VT_WAITACTIVE ioctl call may return 0
if fg_console has already been updated in redraw_screen(), but the
console switch itself hasn't been completed. Fix this by checking
fg_console in vt_waitactive() with the console sem held.

Signed-off-by: Michal Januszewski <spock@gentoo.org>
Signed-off-by: Antonino Daplas <adaplas@gmail.com>
---

 drivers/char/vt_ioctl.c |    6 +++++-
 1 files changed, 5 insertions(+), 1 deletions(-)

diff --git a/drivers/char/vt_ioctl.c b/drivers/char/vt_ioctl.c
index 1fa2da8..6db3088 100644
--- a/drivers/char/vt_ioctl.c
+++ b/drivers/char/vt_ioctl.c
@@ -1041,8 +1041,12 @@ int vt_waitactive(int vt)
 	for (;;) {
 		set_current_state(TASK_INTERRUPTIBLE);
 		retval = 0;
-		if (vt == fg_console)
+		acquire_console_sem();
+		if (vt == fg_console) {
+			release_console_sem();
 			break;
+		}
+		release_console_sem();
 		retval = -EINTR;
 		if (signal_pending(current))
 			break;


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

[PATCH] vt: fix a potential race in the VT_WAITACTIVE handler

[Michal Januszewski]

On a multiprocessor machine the VT_WAITACTIVE ioctl call may return 0
if fg_console has already been updated in redraw_screen(), but the 
console switch itself hasn't been completed. Fix this by checking
fg_console in vt_waitactive() with the console sem held.

Signed-off-by: Michal Januszewski <spock@gentoo.org>

---
diff --git a/drivers/char/vt_ioctl.c b/drivers/char/vt_ioctl.c
index 3a5d301..00b5b34 100644
--- a/drivers/char/vt_ioctl.c
+++ b/drivers/char/vt_ioctl.c
@@ -1041,8 +1041,12 @@ int vt_waitactive(int vt)
 	for (;;) {
 		set_current_state(TASK_INTERRUPTIBLE);
 		retval = 0;
-		if (vt == fg_console)
+		acquire_console_sem();
+		if (vt == fg_console) {
+			release_console_sem();
 			break;
+		}
+		release_console_sem();
 		retval = -EINTR;
 		if (signal_pending(current))
 			break;

Re: [PATCH] vt: fix a potential race in the VT_WAITACTIVE handler

[Andrew Morton]

On Thu, 15 Mar 2007 15:10:23 +0100 Michal Januszewski <spock@gentoo.org> wrote:

> On a multiprocessor machine the VT_WAITACTIVE ioctl call may return 0
> if fg_console has already been updated in redraw_screen(), but the 
> console switch itself hasn't been completed. Fix this by checking
> fg_console in vt_waitactive() with the console sem held.
> 
> Signed-off-by: Michal Januszewski <spock@gentoo.org>
> 
> ---
> diff --git a/drivers/char/vt_ioctl.c b/drivers/char/vt_ioctl.c
> index 3a5d301..00b5b34 100644
> --- a/drivers/char/vt_ioctl.c
> +++ b/drivers/char/vt_ioctl.c
> @@ -1041,8 +1041,12 @@ int vt_waitactive(int vt)
>  	for (;;) {
>  		set_current_state(TASK_INTERRUPTIBLE);
>  		retval = 0;
> -		if (vt == fg_console)
> +		acquire_console_sem();
> +		if (vt == fg_console) {
> +			release_console_sem();
>  			break;
> +		}
> +		release_console_sem();
>  		retval = -EINTR;
>  		if (signal_pending(current))
>  			break;
> 

OK.  I think.  It's hard to tell.  I assume that the acquire_console_sem()
in here is to synchronise against some other function which also takes
acquire_console_sem(), but it is not clear which.

So could you please redo this with a comment which tells the reader exactly
what's being protected against what, and why?

Also, I always feel a bit worried by:

	set_current_state(TASK_INTERRUPTIBLE);
	down(...);

because if it hits contention, the down() will undo the
set_curremt_state().  Now that's normally OK because we loop, and because
the semaphore won't normally be 100% contended all the time.  Unless
someone reimplements down() so it happens to return in state TASK_RUNNING
all the time, which they could legitimately do (although this would
probably break stuff such as the above).


But still, it is nicer to do

	down(...);
	set_current_state(TASK_INTERRUPTIBLE);

if possible, and I think it is possible here.

Thanks.