[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <alpine.DEB.2.10.1604141307580.6593@chino.kir.corp.google.com>
Date: Thu, 14 Apr 2016 13:08:38 -0700 (PDT)
From: David Rientjes <rientjes@...gle.com>
To: Jason Baron <jbaron@...mai.com>
cc: akpm@...ux-foundation.org, kirill.shutemov@...ux.intel.com,
aarcange@...hat.com, mgorman@...hsingularity.net, mhocko@...e.com,
hannes@...xchg.org, vbabka@...e.cz, linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/1] mm: update min_free_kbytes from khugepaged after
core initialization
On Tue, 12 Apr 2016, Jason Baron wrote:
> Khugepaged attempts to raise min_free_kbytes if its set too low. However,
> on boot khugepaged sets min_free_kbytes first from subsys_initcall(), and
> then the mm 'core' over-rides min_free_kbytes after from
> init_per_zone_wmark_min(), via a module_init() call.
>
> Khugepaged used to use a late_initcall() to set min_free_kbytes (such that
> it occurred after the core initialization), however this was removed when
> the initialization of min_free_kbytes was integrated into the starting of
> the khugepaged thread.
>
> The fix here is simply to invoke the core initialization using a
> core_initcall() instead of module_init(), such that the previous
> initialization ordering is restored. I didn't restore the late_initcall()
> since start_stop_khugepaged() already sets min_free_kbytes via
> set_recommended_min_free_kbytes().
>
> This was noticed when we had a number of page allocation failures when
> moving a workload to a kernel with this new initialization ordering. On an
> 8GB system this restores min_free_kbytes back to 67584 from 11365 when
> CONFIG_TRANSPARENT_HUGEPAGE=y is set and either
> CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y or
> CONFIG_TRANSPARENT_HUGEPAGE_MADVISE=y.
>
> Fixes: 79553da293d3 ("thp: cleanup khugepaged startup")
> Signed-off-by: Jason Baron <jbaron@...mai.com>
Acked-by: David Rientjes <rientjes@...gle.com>
I assume it could also be fixed by not setting min_free_kbytes lower in
init_per_zone_wmark_min(), but if the ordering is correct this is less
code.
Powered by blists - more mailing lists