linux-kernel - Re: 'autogroup' sched code KILLING responsiveness

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: <20110123145429.GA2523@zhy>
Date:	Sun, 23 Jan 2011 22:54:29 +0800
From:	Yong Zhang <yong.zhang0@...il.com>
To:	Christian Kujau <christian@...dbynature.de>
Cc:	Michael Witten <mfwitten@...il.com>, linux-kernel@...r.kernel.org,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	Mike Galbraith <efault@....de>, Ingo Molnar <mingo@...e.hu>
Subject: Re: 'autogroup' sched code KILLING responsiveness

On Sun, Jan 23, 2011 at 02:50:08AM -0800, Christian Kujau wrote:
> On Fri, 21 Jan 2011 at 10:20, Michael Witten wrote:
> > With that code in place, a resource-intensive activity (such as
> > compiling the Linux kernel) causes my computer to become
> > unresponsive for many seconds at a time; the entire screen
> > does not refresh, typed keys are dropped or are handled very
> > late, etc (even in Linux's plain virtual consoles).
> 
> Unfortunately, I'd like to add a "me too". 2.6.38-rc1 behaves fine, but 
> with CONFIG_SCHED_AUTOGROUP=y and doing I/O and CPU intensive work (git 
> prune/git repack on a Linux git tree), system load goes up to ~13 and 
> becomes unresponse for some time too. This even happens when I start the 
> jobs with nice -n10.
> 
> Without CONFIG_SCHED_AUTOGROUP enabled and doing the same work,
> systemload goes up to 1 or maybe 2.
> 
> I'm on UP as well (PowerPC G4), disabling CONFIG_SCHED_AUTOGROUP helps 
> here too.

I think below patch will fix it.

---
From: Yong Zhang <yong.zhang0@...il.com>
Subject: [PATCH] sched: tg->se->load should be initialised to tg->shares

Michael reported that when enable autogroup on UP, system
responsiveness becomes very bad.
Because in init_tg_cfs_entry() we initialise se->load
to 0 instead of tg->shares, in the end we have 0-weight
sched entity on rq, then lead to misbehavior.

Reported-by: Michael Witten <mfwitten@...il.com>
Reported-by: Christian Kujau <christian@...dbynature.de>
Signed-off-by: Yong Zhang <yong.zhang0@...il.com>
---
 kernel/sched.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 18d38e4..3ec2c6c 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -7857,7 +7857,7 @@ static void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq,
 		se->cfs_rq = parent->my_q;
 
 	se->my_q = cfs_rq;
-	update_load_set(&se->load, 0);
+	update_load_set(&se->load, tg->shares);
 	se->parent = parent;
 }
 #endif
-- 
1.7.1


--
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/