2015-02-20 17:13:54

by Manuel Reimer

[permalink] [raw]
Subject: "Unkillable processes" and INFO: task ... blocked for more than 120 seconds

Hello,

today, I had the following problem:

I was editing a few photos in GIMP and after I finished with that, I
tried to close GIMP which failed. KDE offered me to terminate the
application which also didn't work.

So I opened a terminal and tried "killall -9 gimp" --> No success

Next I tried "pgrep gimp" to get the PID of the hanging process. With
the result that pgrep now also got stuck with hangup of my whole shell.

A "strace pgrep gimp" told me that pgrep actually hung up on a read on
"/proc/18294/cmdline" so I guessed that this could be my hanging GIMP
process. But even "kill -9 18294" was not able to kill the process.

So I switched over to a VT shell and logged in as root which at first
hung up the login shell, but was fixable by pressing "Ctrl + C". Even as
root I was unable to kill GIMP.

The kernel didn't actually crash. My music player continued to play
nicely and my mail client was fully usable.

I saved some log output to pastebin: http://pastebin.com/FPDvyePJ

What has been happening here? Why is it possible that I actually get
non-killable processes?

If you need additional information, please ask.

Thank you very much in advance.

Manuel


2015-02-20 17:30:41

by Manuel Reimer

[permalink] [raw]
Subject: Re: "Unkillable processes" and INFO: task ... blocked for more than 120 seconds

Hello,

I can add that this problem seems to be reproducible for me so either
some change in one of the latest kernel updates caused this or something
on my system hardware is starting to fail...

If I open "a few" photos on GIMP (enough to get the application memory
usage above 40%) then I'm able to close GIMP without any problems but it
seems to get stuck in exactly the same way as it happened the first time.

My system kernel: Linux manuelspc 3.18.6-1-ARCH #1 SMP PREEMPT Sat Feb 7
08:44:05 CET 2015 x86_64 GNU/Linux


On 02/20/2015 06:13 PM, Manuel Reimer wrote:
> Hello,
>
> today, I had the following problem:
>
> I was editing a few photos in GIMP and after I finished with that, I
> tried to close GIMP which failed. KDE offered me to terminate the
> application which also didn't work.
>
> So I opened a terminal and tried "killall -9 gimp" --> No success
>
> Next I tried "pgrep gimp" to get the PID of the hanging process. With
> the result that pgrep now also got stuck with hangup of my whole shell.
>
> A "strace pgrep gimp" told me that pgrep actually hung up on a read on
> "/proc/18294/cmdline" so I guessed that this could be my hanging GIMP
> process. But even "kill -9 18294" was not able to kill the process.
>
> So I switched over to a VT shell and logged in as root which at first
> hung up the login shell, but was fixable by pressing "Ctrl + C". Even as
> root I was unable to kill GIMP.
>
> The kernel didn't actually crash. My music player continued to play
> nicely and my mail client was fully usable.
>
> I saved some log output to pastebin: http://pastebin.com/FPDvyePJ
>
> What has been happening here? Why is it possible that I actually get
> non-killable processes?
>
> If you need additional information, please ask.
>
> Thank you very much in advance.
>
> Manuel
>

2015-02-20 19:33:13

by Martin Steigerwald

[permalink] [raw]
Subject: Re: "Unkillable processes" and INFO: task ... blocked for more than 120 seconds

Am Freitag, 20. Februar 2015, 18:30:29 schrieb Manuel Reimer:
> Hello,

Hello Manuel,

> On 02/20/2015 06:13 PM, Manuel Reimer wrote:
> > Hello,
> >
> > today, I had the following problem:
> >
> > I was editing a few photos in GIMP and after I finished with that, I
> > tried to close GIMP which failed. KDE offered me to terminate the
> > application which also didn't work.
> >
> > So I opened a terminal and tried "killall -9 gimp" --> No success
> >
> > Next I tried "pgrep gimp" to get the PID of the hanging process. With
> > the result that pgrep now also got stuck with hangup of my whole
> > shell.
> >
> > A "strace pgrep gimp" told me that pgrep actually hung up on a read on
> > "/proc/18294/cmdline" so I guessed that this could be my hanging GIMP
> > process. But even "kill -9 18294" was not able to kill the process.
> >
> > So I switched over to a VT shell and logged in as root which at first
> > hung up the login shell, but was fixable by pressing "Ctrl + C". Even
> > as root I was unable to kill GIMP.
> >
> > The kernel didn't actually crash. My music player continued to play
> > nicely and my mail client was fully usable.
> >
> > I saved some log output to pastebin: http://pastebin.com/FPDvyePJ
> >
> > What has been happening here? Why is it possible that I actually get
> > non-killable processes?
> >
> > If you need additional information, please ask.
> >
> > Thank you very much in advance.

> I can add that this problem seems to be reproducible for me so either
> some change in one of the latest kernel updates caused this or something
> on my system hardware is starting to fail...
>
> If I open "a few" photos on GIMP (enough to get the application memory
> usage above 40%) then I'm able to close GIMP without any problems but it
> seems to get stuck in exactly the same way as it happened the first
> time.
>
> My system kernel: Linux manuelspc 3.18.6-1-ARCH #1 SMP PREEMPT Sat Feb 7
> 08:44:05 CET 2015 x86_64 GNU/Linux

Processes will be unkillable in Linux for as long as they remain in D
state (uninteruptible sleep). I suggest you read upon that topic.

ps aux | head -1 ; ps aux | grep "[g]imp"

should tell you whether thats the case.

In that case, there may be hints in kernel log about a task hung for more
than 120 seconds.

As to the reason? The process is stuck in a system function, usually I/O.

It can also happen on filesystem bugs.

If the process is not in D state, but still unkillable that would need
further exploration. Well zombie processes are unkillable as well then
there isn´t a real process anymore either just its structure. Well that
would be the Z state.

Whether its such a hot idea to have unkillable processes, well… thats the
way it currently is in Linux (and I think quite some other Unix like
operating systems).

Ciao,
--
Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA B82F 991B EAAC A599 84C7