Re: [PATCH 1/5] oom: select_bad_process: check PF_KTHREAD instead of !mm to skip kthreads

[David Rientjes <rientjes@google.com>]

On Wed, 2 Jun 2010, KOSAKI Motohiro wrote:

> > Again, the question is whether or not the fix is rc material or not, 
> > otherwise there's no difference in the route that it gets upstream: the 
> > patch is duplicated in both series.  If you feel that this minor issue 
> > (which has never been reported in at least the last three years and 
> > doesn't have any side effects other than a couple of millisecond delay 
> > until unuse_mm() when the oom killer will kill something else) should be 
> > addressed in 2.6.35-rc2, then that's a conversation to be had with Andrew.
> 
> Well, we have bugfix-at-first development rule. Why do you refuse our
> development process?
> 

This isn't a bugfix, it simply prevents a recall to the oom killer after 
the kthread has called unuse_mm().  Please show where any side effects of 
oom killing a kthread, which cannot exit, as a result of use_mm() causes a 
problem _anywhere_.

If that's the definition you have for a "bugfix," then I could certainly 
argue that some of my patches like "oom: filter tasks not sharing the same 
cpuset" is a bugfix because it allows needlessly killing tasks that won't 
free memory for current, or "oom: avoid oom killer for lowmem allocations" 
is a bugfix because it allows killing a task that won't free lowmem, etc.

I agree that this is a nice patch to have to avoid that recall later, 
which is why I merged it into my patchset, but let's please be accurate 
about its impact.