linux-kernel - Re: [PATCH] epoll: restrict EPOLLEXCLUSIVE to POLLIN and POLLOUT

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] [day] [month] [year] [list]

Message-ID: <994cbd84457a4986e03ad0c4369600fa@silodev.com>
Date:	Fri, 05 Feb 2016 11:49:47 +0200
From:	Madars Vitolins <m@...odev.com>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	Jason Baron <jbaron@...mai.com>, torvalds@...ux-foundation.org,
	mtk.manpages@...il.com, mingo@...nel.org, peterz@...radead.org,
	viro@....linux.org.uk, normalperson@...t.net, corbet@....net,
	luto@...capital.net, hagen@...u.net, linux-kernel@...r.kernel.org,
	linux-fsdevel@...r.kernel.org, linux-api@...r.kernel.org
Subject: Re: [PATCH] epoll: restrict EPOLLEXCLUSIVE to POLLIN and POLLOUT

Andrew Morton @ 2016-02-05 00:59 rakstīja:
> On Thu, 04 Feb 2016 23:44:05 +0200 Madars Vitolins <m@...odev.com> 
> wrote:
> 
>> Hi Jason,
>> 
>> 
>> Just run off the original tests with this patch (eventpoll.c from
>> 4.5-rc2 + patch bellow). Got the same good results, no regression.
>> 
>> $ time ./bankcl
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 1359856158.04 USD
>> Account balance is: 101528948.40 USD
>> 
>> real	0m41.826s
>> user	0m29.513s
>> sys	0m6.490s
>> 
>> 
>> Test case:
>> https://mvitolin.wordpress.com/2015/12/05/endurox-testing-epollexclusive-flag/
>> 
>> PS,
>> 
>> Original cases 0m24.953s vs 0m41.826s now probably is related with my 
>> pc
>> setup. As I just now re-run test with original patch, got the same ~41
>> sec.
>> 
>> So I am fine with this patch!
>> 
> 
> Thanks, I shall add your Tested-by:
> 
> One thing we're sorely missing is an epoll test suite, in
> tools/testing/selftests.  If anyone has anything which we can use to
> kick things off, please hand it over ;)

Not bad idea, probably we need a "tools/testing/selftests/eventpoll" 
folder under which we should have test cases for various epoll scenarios 
with common "run.sh". In spare time I can try to build a case for 
EXCLUSIVE flag (with queues & multiple processes :) ).

Thanks,
Madars