linux-kernel - Re: [PATCH v2 3/9] mm, page_owner: convert page_owner

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]

Date:	Wed, 25 Nov 2015 16:08:11 +0100
From:	Vlastimil Babka <vbabka@...e.cz>
To:	Michal Hocko <mhocko@...nel.org>
Cc:	linux-mm@...ck.org, linux-kernel@...r.kernel.org,
	Andrew Morton <akpm@...ux-foundation.org>,
	Joonsoo Kim <iamjoonsoo.kim@....com>,
	Minchan Kim <minchan@...nel.org>,
	Sasha Levin <sasha.levin@...cle.com>,
	"Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
	Mel Gorman <mgorman@...e.de>,
	Peter Zijlstra <peterz@...radead.org>
Subject: Re: [PATCH v2 3/9] mm, page_owner: convert page_owner_inited to
 static key

[+CC PeterZ]

On 11/25/2015 03:52 PM, Michal Hocko wrote:
> On Tue 24-11-15 13:36:15, Vlastimil Babka wrote:
>> CONFIG_PAGE_OWNER attempts to impose negligible runtime overhead when enabled
>> during compilation, but not actually enabled during runtime by boot param
>> page_owner=on. This overhead can be further reduced using the static key
>> mechanism, which this patch does.
> 
> Is this really worth doing?

Well, I assume that jump labels exist for a reason, and allocation hot paths are
sufficiently sensitive to be worth it? It's not an extra maintenance burden for
us anyway. Just a bit different content of the if () line.

> If we do not have jump labels then the check
> will be atomic rather than a simple access, so it would be more costly,
> no? Or am I missing something?

Well, atomic read is a simple READ_ONCE on x86_64. That excludes some compiler
optimizations, but it's not expensive for the CPU. The optimization would be
caching the value of the flag to a register, which would only potentially affect
multiple checks from the same function (and its inlines). Which doesn't happen
AFAIK, as it's just once in the allocation and once in the free path?

Now I admit I have no idea if there are architectures that don't support jump
labels *and* have an expensive atomic read, and whether we care?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/