On Thu, 2007-04-19 at 13:20 -0600, Eric W. Biederman wrote:
> Trond Myklebust <trond.myklebust@fys.uio.no> writes:
> 
> > On Thu, 2007-04-19 at 01:58 -0600, Eric W. Biederman wrote:
> >> From: Eric W. Biederman <ebiederm@xmission.com>
> >> 
> >> Start the reclaimer thread using kthread_run instead
> >> of a combination of kernel_thread and daemonize.
> >> The small amount of signal handling code is also removed
> >> as it makes no sense and is a maintenance problem to handle
> >> signals in kernel threads.
> >
> > Vetoed. Removing stuff just because it doesn't make sense to you is not
> > acceptable.
> >
> > Signal handling in reclaimer threads is there in order to allow
> > administrators to deal with the case where the server never comes up
> > again.
> 
> Doesn't unmount handle that?

On a pinned filesystem?

> Regardless kernel threads should be an implementation detail
> not a part of the user interface.  If kernel threads are part
> of the user interface it makes them very hard to change.
> 
> So it isn't that it doesn't make sense to me it is that it looks
> fundamentally broken and like a maintenance nightmare.
> 
> I would rather kill kernel threads then try and simulate them
> when the kernel implementation has changed and kernel threads
> are not visible.
> 
> If I could be convinced that signal handling in kernel threads
> is not something that will impede code modifications and refactoring
> I would have less of a problem, and might not care.

Tough. You're the one proposing to change existing code.

> With pid namespaces all kernel threads will disappear so how do
> we cope with the problem when the sysadmin can not see the kernel
> threads?

Then you have a usability problem. How does the sysadmin reboot the
system if there is no way to shut down the processes that are hanging on
an unresponsive filesystem?

Trond

_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers