From mboxrd@z Thu Jan 1 00:00:00 1970 From: Manfred Spraul Subject: Re: ipc-msg broken again on 3.11-rc7? Date: Tue, 03 Sep 2013 09:49:48 +0200 Message-ID: <5225949C.9030201@colorfullife.com> References: <521DE5D7.4040305@synopsys.com>

<52205597.3090609@synopsys.com>

<5224BCF6.2080401@c olorfullife.com> Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="------------020005030001010506050103" Return-path: In-Reply-To: Sender: owner-linux-mm@kvack.org To: Vineet Gupta Cc: "sedat.dilek@gmail.com" , Linus Torvalds , Davidlohr Bueso , Davidlohr Bueso , linux-next , LKML , Stephen Rothwell , Andrew Morton , linux-mm , Andi Kleen , Rik van Riel , Jonathan Gonzalez List-Id: linux-next.vger.kernel.org This is a multi-part message in MIME format. --------------020005030001010506050103 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Hi Vineet, On 09/03/2013 09:34 AM, Vineet Gupta wrote: > However assuming we are going ahead with debugging this - can you > please confirm whether you see the issue on x86 as well as I have not > tested that ? I vaguely remember one of your earlier posts suggested > you did Thx, -Vineet I'm unable to reproduce the issue so far. I ran - something like 4000 stock msgctl08 with 4 cores on x86. - a few runs with modified msgctl08 with either slowed down reader or slowed down writer threads. (i.e.: force queue full or queue empty waits). - a few runs (modified&unmodified) with all but one core taken offline. I have not yet tested with PREEMPT enabled. A few more ideas: - what is the output of ipcs -q? Are the queues empty or full? - what is the output of then WCHAN field with ps when it hangs? Something like / #ps/ -o pid,f,stat,pcpu,pmem,psr,/wchan/=/WIDE/-/WCHAN/ -o comm,args -- Manfred --------------020005030001010506050103 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit

Hi Vineet,

On 09/03/2013 09:34 AM, Vineet Gupta wrote:

However assuming we are going ahead with debugging this - can you please confirm whether you see the issue on x86 as well as I have not tested that ? I vaguely remember one of your earlier posts suggested you did Thx, -Vineet

I'm unable to reproduce the issue so far. I ran
- something like 4000 stock msgctl08 with 4 cores on x86.
- a few runs with modified msgctl08 with either slowed down reader or slowed down writer threads. (i.e.: force queue full or queue empty waits).
- a few runs (modified&unmodified) with all but one core taken offline.

I have not yet tested with PREEMPT enabled.

A few more ideas:
- what is the output of ipcs -q? Are the queues empty or full?
- what is the output of then WCHAN field with ps when it hangs?
Something like
#ps -o pid,f,stat,pcpu,pmem,psr,wchan=WIDE-WCHAN -o comm,args

--
Manfred
--------------020005030001010506050103-- -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org