Killing a process with a kernel module

Killing a process with a kernel module

[Patrick Barnes]

Hi there,

RedHat recently released RHEL4 U2, and this update has some
unfortunate consequences, mainly
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=170087

As stated in the bugzilla report above, useradd gets stuck in a R
state and ignores all signals.

I have a critical system which I can't take down for another week or
so, but the extra load generated by the rogue useradd process is
slowing things down a little too much.

I was wondering if it was possible to safely terminate the process
using a more direct method (perhaps by manipulating kernel data
structures.)

What I'd like to know is if it is actually possible before attempting
to write the code to do it.

I know this is a long shot, but it's probably worth trying. :)

Thanks for your time. I look forward to any replies.

Regards,
Patrick

Re: Killing a process with a kernel module

[Pavel Machek]

Hi!

> RedHat recently released RHEL4 U2, and this update has some
> unfortunate consequences, mainly
> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=170087
> 
> As stated in the bugzilla report above, useradd gets stuck in a R
> state and ignores all signals.
> 
> I have a critical system which I can't take down for another week or
> so, but the extra load generated by the rogue useradd process is
> slowing things down a little too much.
> 
> I was wondering if it was possible to safely terminate the process
> using a more direct method (perhaps by manipulating kernel data
> structures.)
> 
> What I'd like to know is if it is actually possible before attempting
> to write the code to do it.
> 
> I know this is a long shot, but it's probably worth trying. :)

Maybe doable, but I'd certainly not do it on important machine.

								Pavel
-- 
Thanks, Sharp!

Re: Killing a process with a kernel module

[Robert Hancock]

Patrick Barnes wrote:
> Hi there,
> 
> RedHat recently released RHEL4 U2, and this update has some
> unfortunate consequences, mainly
> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=170087
> 
> As stated in the bugzilla report above, useradd gets stuck in a R
> state and ignores all signals.
> 
> I have a critical system which I can't take down for another week or
> so, but the extra load generated by the rogue useradd process is
> slowing things down a little too much.
> 
> I was wondering if it was possible to safely terminate the process
> using a more direct method (perhaps by manipulating kernel data
> structures.)
> 
> What I'd like to know is if it is actually possible before attempting
> to write the code to do it.
> 
> I know this is a long shot, but it's probably worth trying. :)
> 
> Thanks for your time. I look forward to any replies.

If it's not responding to signals I wouldn't count on being able to do 
this safely. Can't you nice the process down to lowest priority?


-- 
Robert Hancock      Saskatoon, SK, Canada
To email, remove "nospam" from hancockr@nospamshaw.ca
Home Page: http://www.roberthancock.com/