From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: Re: Core scsi layer crashes in 2.6.8.1 Date: 29 Sep 2004 10:11:59 -0400 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <1096467125.2028.11.camel@mulgrave> References: <1096401785.13936.5.camel@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from stat16.steeleye.com ([209.192.50.48]:27338 "EHLO hancock.sc.steeleye.com") by vger.kernel.org with ESMTP id S268447AbUI2OMH (ORCPT ); Wed, 29 Sep 2004 10:12:07 -0400 In-Reply-To: <1096401785.13936.5.camel@localhost.localdomain> List-Id: linux-scsi@vger.kernel.org To: Alan Cox Cc: Linux Kernel Mailing List , SCSI Mailing List On Tue, 2004-09-28 at 16:03, Alan Cox wrote: > Second problem is with the scsi handling logic for errors. If you rmmod > a scsi driver while it is error handling you get a chain of errors > starting with > > Illegal transition Cancel->Offline > Badness is scsi_device_set_state > path: > scsi_device_set_state > scsi_unjam_host > scsi_error_handler > These state transition warnings are currently expected in this code (they're basically verbose warnings). What was the oops? I have a theory that we should be taking a device reference before waking up the error handler, otherwise host removal can race with error handling. James