From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from zcars04f.nortelnetworks.com (zcars04f.nortelnetworks.com [47.129.242.57]) by ozlabs.org (Postfix) with ESMTP id E29332BF12 for ; Sat, 4 Dec 2004 09:08:22 +1100 (EST) Message-ID: <41B0E18C.2060801@nortelnetworks.com> Date: Fri, 03 Dec 2004 15:58:36 -0600 From: Chris Friesen MIME-Version: 1.0 To: Linux kernel , linuxppc-dev@ozlabs.org Content-Type: text/plain; charset=us-ascii; format=flowed Subject: looking for help, attempting to debug high latency in 2.4 kernel List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Hi guys, I'm running 2.4.22 on ppc32. Embedded system, just 74xx cpu, dual tulip network links and fiberchannel card. Card has 1.5GB of memory, rootfs is tmpfs, system has some NFS mounts. We're seeing userspace delayed for almost a second every so often. No idea what's causing it. We've got scheduler instrumentation showing a usespace task starting to run, then a bunch of tasks being put on the runqueue, and finally we wake up and start running normally again almost a full second later. The timing is not exactly the same in each case, but it's usually close to a second of delay. I'm looking for some tips on how to go about tracking this down. Where are my delays likely to be coming from? Long-running code paths? Long periods with interrupts off? I dunno where to start. Would porting lockmeter be a good idea? Is there anyone who's already done this for ppc? I appreciate any help. Thanks, Chris