Re: [PATCH] move put_task_struct() reaping into a thread [Re: 2.6.18-rt1]

[Bill Huey (hui) <billh@gnuppy.monkey.org>]

On Thu, Sep 21, 2006 at 09:16:24AM +0200, Ingo Molnar wrote:
> * Bill Huey <billh@gnuppy.monkey.org> wrote:
> > It's correct from the standpoint of it being reaped in another thread, 
> > so it fixed those crashes. But I pushed it down into another thread at 
> > the request of Esben and his private discussion with Paul McKenney, 
> > since a summary from Esben felt that call_rcu() was somehow less than 
> > ideal to do that.
> 
> but it _is_ already being reaped in another thread: softirq-rcu. 
> Splitting that up any further will only fragment the context-switching 
> and increases cache footprint - it wont (or rather, shouldnt) have any 
> functional effect. (As a sidenote, i'm considering the unification of 
> all 'same default priority' softirq threads into a single thread per 
> CPU, to further reduce this cost of 'spreadout'.)

I overloaded another reaping thread that was doing largely similar
functionality in that it was also reaping, so I don't think it's that bad.
I did it from a cleanliness point of view with the code tree. It's the
"desched_thread" in fork.c that I'm using. It seems to be the right
thing to do. I'm sure Esben will follow up on this.

> > > that you saw crashes under 2.6.17 - but did you manage to figure out 
> > > what the reason is for those crashes, and do those reasons really 
> > > necessiate the pushing of task-reapdown into yet another set of 
> > > kernel threads?
> > 
> > Unfortunately no. I even used Robert's .config on my machine. I added 
> > a disk controller and networking device driver just to boot into his 
> > configuration and I still couldn't replicated any of his kjournald 
> > problems at all. If I had his hardware I'd have a better way of 
> > replicating those problems and pound it out.
> 
> ok, then i guess what we have left is to wait and see whether it still 
> triggers with the current 2.6.18-rt codebase - maybe it triggers for 
> someone in a scenario that is easier to debug.

bill