Re: How to diagnose a process stuck in D state?

[Daniel J Blueman]

Stephen,

On 22 Feb, 23:10, "Stephen Oberholtzer" <oliverklozoff@xxxxxxxxx> wrote:
> First off: I'm not subscribed to the list (I don't think I could
> handle the volume), so please make sure you CC me if you reply.
>
> I run an application on one of my machines; it often hangs, with the
> process stuck D state.  When this happens, the process sticks around
> until I reboot the machine.
>
> I have tried the following to start diagnosing the problem:
>
> * Running 'ps -axl' shows "-" in the wchan column.
> * The contents of /proc/pid/wchan say "_stext".
> * 'strace -p pid' says "Process pid attached - interrupt to quit" and
> stops responding.  Sending SIGINT and SIGTERM have no effect on the
> strace process, although kill -11 (SIGSEGV, my personal favorite) does
> work.
>
> This is very confusing.  I would greatly appreciate it if someone
> could tell me how a process can enter D state without being in a
> syscall, and what I can do to start tracking down the cause.

wchan shows where the process is sleeping - probably a kernel mutex or
lock here. This doesn't help much, but the stack signature will.

Enable and use the sys-request mechanism (via terminal, keyboard or
serial) eg sysrq-T to dump the stack frames of all processes,
including this.

Unsurprisingly, it's documented in the kernel Documentation directory.

> (By the way: This is on amd64, 2.6.23. I'm updating to 2.6.24.2 right
> now, on the off chance that whatever was causing the problem has been
> fixed.)

Daniel
-- 
Daniel J Blueman
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/