From mboxrd@z Thu Jan 1 00:00:00 1970 From: Carsten Emde Subject: Re: Memory leak in 3.4.x-rt ? Date: Tue, 25 Sep 2012 18:18:31 +0200 Message-ID: <5061D957.4060305@osadl.org> References: <503775B3.4080504@cantastic.org> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: =?UTF-8?B?UmFsZiBSw7ZzY2g=?= , linux-rt-users@vger.kernel.org To: Jean-Michel Hautbois Return-path: Received: from toro.web-alm.net ([62.245.132.31]:58490 "EHLO toro.web-alm.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758222Ab2IYQUK (ORCPT ); Tue, 25 Sep 2012 12:20:10 -0400 In-Reply-To: Sender: linux-rt-users-owner@vger.kernel.org List-ID: On 09/25/2012 05:39 PM, Jean-Michel Hautbois wrote: > 2012/8/24 Ralf R=C3=B6sch: >> Hi, >> >> we work on our TX4938 (Toshiba Mips, little endian mode) based embed= ded >> system (motion control) >> running successfully a 3.2 RT patched kernel for month now. >> >> I just give a try to a 3.4 RT patched kernel and detected a memory p= roblem >> as you can see from the logs below: >> >> 3.2 with RT FULL PREEMPT is o.k >> 3.4 without RT FULL PREEMPT is .ok. >> 3.4 with RT FULL PREEMPT is not o.k. >> [..] >> uname -a >> Linux MCU 3.4.9-rt17-11244-gb12622a #23 PREEMPT RT Fri Aug 24 12:15:= 49 CEST >> 2012 mips GNU/Linux >> [root@MCU:/opt]# watch -n 1 cat /proc/meminfo | grep MemFree >> MemFree: 204304 kB >> MemFree: 204208 kB >> MemFree: 204112 kB >> MemFree: 204064 kB >> MemFree: 203968 kB >> MemFree: 203920 kB >> MemFree: 203824 kB >> MemFree: 203776 kB >[..] > You have a big leak (around 100kB/sec), yes, but can you give us more= details ? > What kind of application ? Network ? video encoder ? > Which drivers are you using ? > Did you try using some advanced tools in order to check that your > application cannot have a memory leak in a specific use case (race > condition ?) ? > You will need to be more precise on the memory leak detection, becaus= e > with what you gave, I (maybe others can) cannot help you more... In a first step, I would propose to run "cat /proc/meminfo" repeatedly=20 and no longer only check MemFree but check all other variables. One of=20 them will increase in the same amount as MemFree decreases. Which one i= s=20 it? Slab? This information will then lead us where to search further. -Carsten. -- To unsubscribe from this list: send the line "unsubscribe linux-rt-user= s" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html