Re: stuck on TLB IPI wait

From: Jeff V. Merkey (jmerkey@timpanogas.com)
Date: Wed Jun 21 2000 - 19:42:38 EST


Greg,

I have seen this on 2.2.14 and 2.2.15 if you have a spinlock_irqsave()
call that gets stuck in a deadlock. 2.2.16 does not seems to do it (or
at least the error manifests itself differently). If you can download
ksymoops and/or get your hands on kdb from SGI's website, you may be
able to determine which spinlock call is deadlocking in the kernel.

You can also use gdb to do a post mortem dump and try to find out where
the code got hung in a loop.

Syntax is:

# gdb /usr/src/linux/vmlinux /proc/kcore

and then you can use gdb to grep around and attempt to locate the
offending code. One technique I use to run down deadlocks with
spinlocks is to modify the spinlock code in /usr/src/linux/include and
put in a counter for the spinlock, and if it spins for say 2,000,000,000
iterations, then purposely dereference a NULL pointer:

i.e.

char *crash = -1;

*crash = 0; // will generate a page fault error and Oops.

Then at least you can use GDB to run down where in the code the kernel
got stuck. Chances are good it's in a driver somewhere, or yu may have
some bad memory in the system or some other hardware problem.

Hope this helps....

:-)

Jeff

Greg Strelzoff wrote:
>
> Hi,
>
> My computer with the Linux operating system is having system down
> crashes.
>
> I am getting the following error message in my message file at
> the time of the crash:
> Jun 21 03:01:00 mars kernel: stuck on TLB IPI wait (CPU#1)
>
> I have tried all these kernels - 2.2.5, 2.2.9, 2.2.14 and I am on
> 2.2.15 now.
>
> The computer is a dual CPU p2 computer with 512 Megabytes of RAM.
>
> The computer had been up since June 6th and crashed with this message on
> June 21st.
>
> This server is running Intershop3.0 with apache 1.3 and
> openssl support.
>
> The web site is getting about 20,000 page views day [not that much].
>
> So, despite upgrading the kernel, I am getting this same error
> message still.
>
> Please CC on the replies,
>
> Thank you in advance,
> Greg
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.rutgers.edu
> Please read the FAQ at http://www.tux.org/lkml/

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



This archive was generated by hypermail 2b29 : Fri Jun 23 2000 - 21:00:23 EST