lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87tt2t9lkp.fsf@DESKTOP-5N7EMDA>
Date: Thu, 31 Jul 2025 09:48:54 +0800
From: "Huang, Ying" <ying.huang@...ux.alibaba.com>
To: Joshua Hahn <joshua.hahnjy@...il.com>
Cc: Andrew Morton <akpm@...ux-foundation.org>,  David Hildenbrand
 <david@...hat.com>,  Johannes Weiner <hannes@...xchg.org>,  Zi Yan
 <ziy@...dia.com>,  Matthew Brost <matthew.brost@...el.com>,  Rakie Kim
 <rakie.kim@...com>,  Byungchul Park <byungchul@...com>,  Gregory Price
 <gourry@...rry.net>,  Alistair Popple <apopple@...dia.com>,
  linux-kernel@...r.kernel.org,  linux-mm@...ck.org,  kernel-team@...a.com,
  Dave Hansen <dave.hansen@...ux.intel.com>
Subject: Re: [PATCH] mempolicy: Clarify what RECLAIM_ZONE means

Joshua Hahn <joshua.hahnjy@...il.com> writes:

> On Tue, 29 Jul 2025 08:58:49 +0800 "Huang, Ying" <ying.huang@...ux.alibaba.com> wrote:
>
>> Joshua Hahn <joshua.hahnjy@...il.com> writes:
>> 
>> > On Mon, 28 Jul 2025 09:44:06 +0800 "Huang, Ying" <ying.huang@...ux.alibaba.com> wrote:
>> >
>> >> Hi, Joshua,
>> >> 
>> >> Joshua Hahn <joshua.hahnjy@...il.com> writes:
>> >> 
>> >> > The zone_reclaim_mode API controls reclaim behavior when a node runs out of
>> >> > memory. Contrary to its user-facing name, it is internally referred to as
>> >> > "node_reclaim_mode". This is slightly confusing but there is not much we can
>> >> > do given that it has already been exposed to userspace (since at least 2.6).
>> >> >
>> >> > However, what we can do is to make sure the internal description of what the
>> >> > bits inside zone_reclaim_mode aligns with what it does in practice.
>> >> > Setting RECLAIM_ZONE does indeed run shrink_inactive_list, but a more holistic
>> >> > description would be to explain that zone reclaim modulates whether page
>> >> > allocation (and khugepaged collapsing) prefers reclaiming & attempting to
>> >> > allocate locally or should fall back to the next node in the zonelist.
>> >> >
>> >> > Change the description to clarify what zone reclaim entails.
>> >> >
>> >> > Signed-off-by: Joshua Hahn <joshua.hahnjy@...il.com>
>> >> > ---
>> >> >  include/uapi/linux/mempolicy.h | 2 +-
>> >> >  1 file changed, 1 insertion(+), 1 deletion(-)
>> >> >
>> >> > diff --git a/include/uapi/linux/mempolicy.h b/include/uapi/linux/mempolicy.h
>> >> > index 1f9bb10d1a47..24083809d920 100644
>> >> > --- a/include/uapi/linux/mempolicy.h
>> >> > +++ b/include/uapi/linux/mempolicy.h
>> >> > @@ -69,7 +69,7 @@ enum {
>> >> >   * These bit locations are exposed in the vm.zone_reclaim_mode sysctl
>> >> >   * ABI.  New bits are OK, but existing bits can never change.
>> >> >   */
>> >> > -#define RECLAIM_ZONE	(1<<0)	/* Run shrink_inactive_list on the zone */
>> >> > +#define RECLAIM_ZONE	(1<<0)	/* Prefer reclaiming & allocating locally */
>> >> >  #define RECLAIM_WRITE	(1<<1)	/* Writeout pages during reclaim */
>> >> >  #define RECLAIM_UNMAP	(1<<2)	/* Unmap pages during reclaim */
>> >> >  
>> >> >
>> >> > base-commit: 25fae0b93d1d7ddb25958bcb90c3c0e5e0e202bd
>> >
>> > Hi Ying, thanks for your review, as always!
>> >
>> >> Please consider the document of zone_reclaim_mode in
>> >> Documentation/admin-guide/sysctl/vm.rst too.
>> >
>> > Yes, will do. Along with SJ's comment, I think that the information in the
>> > admin-guide should be sufficient enough to explain what these bits do, so
>> > I think my patch is not very necessary.
>> >
>> >> And, IIUC, RECLAIM_ZONE doesn't mean "locally" exactly.  It's legal to
>> >> bind to some node other than "local node".
>> >
>> > You are correct, it seems you can also reclaim on non-local nodes once you
>> > go further down in the zonelist. I think my intent with the new comment was just
>> > to indicate a preference to reclaim and allocate on the *current* node, as
>> > opposed to falling back to the next node in the zonelist.
>> >
>> > With that said, I think your comment along with SJ's feedback have gotten me
>> > to understand that we proably don't need this change : -) 
>> 
>> TBH, I think that it's good to make some change to the comments.
>> Because IMHO, the original comments are bound to some specific
>> implementation details.  Some more general words may be better for the
>> user space API description.
>
> Hi Ying, sorry for the late reply.
>
> I think that is a good point. Then maybe in that case, we can take SJ's comment
> and leave information about both the implementation detail (i.e. that it will
> perform shrink inactive_list on the zone), and that it will prefer this over
> allocating on the next node as a general description of what happens?

Yes.  Something like this, or

Try to reclaim in the current node/zone before allocating on the fallback.

> On that note, one thing that I felt was slightly undercaptured in
> Documentation/admin-guide is what "zone reclaim" actually means. What it does
> is of course well captured by its name, but it misses the nuance of preferring
> reclaim over fallback allocation.
>
> Actually the whole motivation behind all of this conversation is because I saw
> zone reclaim preventing allocation into a second node in a 2-NUMA node system
> and was a bit confused until I understood what the implication of having
> zone reclaim was.

Yes.  It's good to improve the document.  If it makes you confusing, it
may make others confusing too.

> Anyways, I can probably spin the patch to include information about what
> zone reclaim is, in the comment block above the bits.
>
> But please feel free to correct me if you feel that the descriptions available
> in both the mempolicy.h uapi file or the Documentation/admin-guide is already
> enough.

Thanks for doing this.

---
Best Regards,
Huang, Ying

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ