linux-kernel - Re: [PATCH 2/2] mm/page_alloc: avoid counting event if no successful allocation

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: <20210709122601.GA3840@techsingularity.net>
Date:   Fri, 9 Jul 2021 13:26:01 +0100
From:   Mel Gorman <mgorman@...hsingularity.net>
To:     Yanfei Xu <yanfei.xu@...driver.com>
Cc:     akpm@...ux-foundation.org, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/2] mm/page_alloc: avoid counting event if no successful
 allocation

On Fri, Jul 09, 2021 at 06:28:55PM +0800, Yanfei Xu wrote:
> While the nr_populated is non-zero, however the nr_account might be
> zero if allocating fails. In this case, not to count event can save
> some cycles.
> 

The much more likely path is that nr_account is positive so we avoid a
branch in the common case.

> And this commit extract the check of "page_array" from a while
> statement to avoid unnecessary checks for it.
> 

I'm surprised the compiler does not catch that page_array is invariant
for the loop. Did you check if gcc generates different code is page_array
is explicitly checked once instead of putting it in the loop?

-- 
Mel Gorman
SUSE Labs