David C. Hansen wrote:
> Adam J. Richter wrote:
>> The was a small change to to kernel/exit.c
>> in 2.5.8-pre1 which deleted a kernel_lock() call. Restoring that line
>> resulted in a kernel that booted fine.
>>
> I take it you don't have a copy of the BUG(). I was going to ask if
> preemption was enabled, but I see that it was from another message. I
> was guessing that preemption contributed to this, but now I know. The
> lock_kernel() has 2 different effects here. It locks the kernel_flag,
> AND it disables preemption. The correct fix here will probably be to
> disable preemption, rather than readd the lock_kernel().
I've replicated the problem too.
I've diabled preemption in the area where it used to be disabled because
of the old lock_kernel(). I'm sending this message from a machine with
that patch applied, so the patch does fix it. As the comment says, this
is something that the preempt experts need to take a look at.
Linus, this is a hack, and there is probably still a window where
preemption can happen. But, it is a band-aid until we find the real
problem.
-- Dave Hansen haveblue@us.ibm.com
--- linux-2.5.8-pre1-clean/kernel/exit.c Thu Apr 4 08:58:31 2002 +++ linux/kernel/exit.c Thu Apr 4 10:19:37 2002 @@ -499,6 +499,11 @@ acct_process(code); __exit_mm(tsk); + /* I removed the lock_kernel() from here. It caused preempt kernels + to oops. This fixes it for now, but the real cause needs to + be found. + - Dave Hansen <haveblue@us.ibm.com> 04-04-2002 */ + preempt_disable(); sem_exit(); __exit_files(tsk); __exit_fs(tsk); @@ -515,6 +520,7 @@ tsk->exit_code = code; exit_notify(); + preempt_enable_no_resched(); /* partner of above preempt_disable(); */ schedule(); BUG(); /*
- 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/
This archive was generated by hypermail 2b29 : Sun Apr 07 2002 - 22:00:14 EST