From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S267489AbUHPJan (ORCPT ); Mon, 16 Aug 2004 05:30:43 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S267494AbUHPJan (ORCPT ); Mon, 16 Aug 2004 05:30:43 -0400 Received: from eurogra4543-2.clients.easynet.fr ([212.180.52.86]:23696 "HELO server5.heliogroup.fr") by vger.kernel.org with SMTP id S267489AbUHPJal (ORCPT ); Mon, 16 Aug 2004 05:30:41 -0400 From: Hubert Tonneau To: linux-kernel@vger.kernel.org Subject: USB storage crash report in 2.6 Date: Mon, 16 Aug 2004 09:11:48 GMT Message-ID: <04BQXJO12@server5.heliogroup.fr> X-Mailer: Pliant 92 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org When I try to copy large amount of datas (more than 100 GB) between USB attached disks, I get a crash with Linux 2.6 Here are the extra details I can provide: At crash time, the machine completely freezes, so I can only report what I can see on the screen, mainly the stack trace report: scan_async ehci_watchlog ehci_work ehci_watchlog run_timer_softirq __do_softirq do_softirq smp_apic_timer_interrupt apic_timer_interrupt default_idle call_console_driver printk The machine is running a Dell sx270 (P4 hyperthreading) running Linux 2.6.8.1 What I do is try to copy roughly 60 GB from disk sdc to sda (EXT3 to EXT3), and then another 60 GB from sdc to sdb The crash appends after copying roughly 100 GB. It can be reproduced. I had no problem when I first copied from sda to sdc and sdb to sdc (it was XFS to EXT3). I tried the following, without success: . downgrade to 2.6.7 . desable kernel preempting . change target filesystem from EXT3 to EXT2 then to XFS What did success is downgrade to 2.4.27 Anyway, I got no crash on all our various machines running 2.6.6 and 2.6.7, including some busy ones, unless some USB device is attached (once reading a deffective DVD, and this time copying between USB disks). Just in case, I also remember that I also had a server repetingly freezing in early 2.6, and it was related to a bad cable generating tiny error on the SCSI bus (MTP fusion controler) and it was properly handled by 2.4, not by 2.6 so the problem might be related to handling tiny problems in the SCSI layer as well (assuming it was not an MTP fusion driver problem).