[PATCH 2/9 v2] xfs: introduce and use KM_NOLOCKDEP to silence reclaim lockdep false positives

[Michal Hocko]

On Fri 16-12-16 11:37:50, Brian Foster wrote:
> On Fri, Dec 16, 2016 at 04:40:41PM +0100, Michal Hocko wrote:
> > Updated patch after Mike noticed a BUG_ON when KM_NOLOCKDEP is used.
> > ---
> > From 1497e713e11639157aef21cae29052cb3dc7ab44 Mon Sep 17 00:00:00 2001
> > From: Michal Hocko <mhocko at suse.com>
> > Date: Thu, 15 Dec 2016 13:06:43 +0100
> > Subject: [PATCH] xfs: introduce and use KM_NOLOCKDEP to silence reclaim
> >  lockdep false positives
> > 
> > Now that the page allocator offers __GFP_NOLOCKDEP let's introduce
> > KM_NOLOCKDEP alias for the xfs allocation APIs. While we are at it
> > also change KM_NOFS users introduced by b17cb364dbbb ("xfs: fix missing
> > KM_NOFS tags to keep lockdep happy") and use the new flag for them
> > instead. There is really no reason to make these allocations contexts
> > weaker just because of the lockdep which even might not be enabled
> > in most cases.
> > 
> 
> Hi Michal,
> 
> I haven't gone back to fully grok b17cb364dbbb ("xfs: fix missing
> KM_NOFS tags to keep lockdep happy"), so I'm not really familiar with
> the original problem. FWIW, there was another KM_NOFS instance added by
> that commit in xlog_cil_prepare_log_vecs() that is now in
> xlog_cil_alloc_shadow_bufs(). Perhaps Dave can confirm whether the
> original issue still applies..?

Yes, I've noticed that but the reworked code looked sufficiently
different that I didn't dare to simply convert it.
-- 
Michal Hocko
SUSE Labs