bsd accounting lockups on smp 2.6.x machines

bsd accounting lockups on smp 2.6.x machines

[John M Flinchbaugh]

[-- Attachment #1: Type: text/plain, Size: 725 bytes --]

i originally reported kernel oopses and locks here against 2.6.1:
http://marc.theaimsgroup.com/?l=linux-kernel&m=107488761716255&w=2

then another person reported it against 2.6.2:
http://marc.theaimsgroup.com/?l=linux-kernel&m=107697407107875&w=2

i just saw it happen against 2.6.3, but i couldn't capture an
oops.  it's still out there.  does anyone have any leads on what
causes it?  disabling  bsd accounting seems to aleviate the
crashes.

running tomcat's stop script (java processes) triggers it at
times for me.
-- 
____________________}John Flinchbaugh{______________________
| glynis@hjsoft.com         http://www.hjsoft.com/~glynis/ |
~~Powered by Linux: Reboots are for hardware upgrades only~~

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: bsd accounting lockups on smp 2.6.x machines

[Andrew Morton]

John M Flinchbaugh <glynis@butterfly.hjsoft.com> wrote:
>
> i originally reported kernel oopses and locks here against 2.6.1:
> http://marc.theaimsgroup.com/?l=linux-kernel&m=107488761716255&w=2
> 
> then another person reported it against 2.6.2:
> http://marc.theaimsgroup.com/?l=linux-kernel&m=107697407107875&w=2
> 
> i just saw it happen against 2.6.3, but i couldn't capture an
> oops.  it's still out there.  does anyone have any leads on what
> causes it?  disabling  bsd accounting seems to aleviate the
> crashes.
> 
> running tomcat's stop script (java processes) triggers it at
> times for me.

Guys, are either of you still able to reproduce this oops in BSD accounting?

If so, would you be able to determine whether this fixes it?   Thanks.

diff -puN kernel/acct.c~acct-oops-fix kernel/acct.c
--- 25/kernel/acct.c~acct-oops-fix	2004-04-09 03:52:59.816330248 -0700
+++ 25-akpm/kernel/acct.c	2004-04-09 03:54:08.671862616 -0700
@@ -347,7 +347,10 @@ static void do_acct_process(long exitcod
 	/* we really need to bite the bullet and change layout */
 	ac.ac_uid = current->uid;
 	ac.ac_gid = current->gid;
+
+	read_lock(&tasklist_lock);	/* pin current->tty */
 	ac.ac_tty = current->tty ? old_encode_dev(tty_devnum(current->tty)) : 0;
+	read_unlock(&tasklist_lock);
 
 	ac.ac_flag = 0;
 	if (current->flags & PF_FORKNOEXEC)

_