lkml.org 
[lkml]   [2008]   [Oct]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PULL] module, param and stop_machine patches


On Sun, 26 Oct 2008, Rusty Russell wrote:
> On Sunday 26 October 2008 09:33:43 Linus Torvalds wrote:
> >
> > I'm not seeing any "tracked it down".
>
> He tracked it down to moving init_workqueues() too early, so he moved that
> back.

Yeah. I had done as well by my bisection. One of the reasons I didn't push
out my initial fix (just the plain revert) was exactly the fact that I
don't think that is sufficient. I want to _understand_ what was going on,
and get a feel from the commit message that makes me go "yeah, ok, makes
sense".

> Turns out it's the cpu_online_map difference. If init_workqueues() is called
> too early, only the boot cpu is set. We then only create_workqueue_thread()
> for the boot cpu.

Indeed. I figured it out yesterday, since I didn't get an explanation soon
enough for my impatient self. So I ended up committing the fix as the
minimal patch I sent out earlier, with a fairly minimal explanation. I'll
happily take more fixes, but I think it should be ok.

If S390 wants to do stop_machine() even before some of the basic
core_initcalls, I suspect it will have to change, but nothing I saw
indicated that it really needs to.

> If CONFIG_HOTPLUG_CPU=y, it's fine since the hotplug callback will create the
> workqueue threads for the other cpus as they come up. Without it, the kevent
> workqueues on non-boot cpus don't get processed.

.. and my default workstation config indeed hit this, since I normally
have a fairly minimal setup.

> > Nor do I see a sign-off from Heiko on it.
>
> I thought mine would be sufficient since we both work for IBM?

Technically, yes, since you all sign your life away and do this for work,
but he normally signs off his patches, so the lack of sign-off didn't make
me go "hmm, this looks odd from a legal copyright standpoint", but rather
"hmm, maybe Heiko doesn't feel comfy with the patch since he didn't sign
off".

To me, the legal side of the sign-off is actually secondary. The whole
thing was designed to be useful even in the absense of it. I've grown very
fond of seeing who has seen the patch (including who _should_ have seen it
but didn't react - the "cc" part ;)

Linus



\
 
 \ /
  Last update: 2008-10-26 17:33    [W:0.055 / U:1.408 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site