From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S265245AbUHHJle (ORCPT ); Sun, 8 Aug 2004 05:41:34 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S265247AbUHHJle (ORCPT ); Sun, 8 Aug 2004 05:41:34 -0400 Received: from vsmtp14.tin.it ([212.216.176.118]:62662 "EHLO vsmtp14.tin.it") by vger.kernel.org with ESMTP id S265245AbUHHJlc (ORCPT ); Sun, 8 Aug 2004 05:41:32 -0400 Date: Sun, 8 Aug 2004 11:39:11 +0200 From: Giuliano Pochini To: "Randy.Dunlap" Cc: linux-kernel@vger.kernel.org Subject: Re: SCSI removable devices problem Message-Id: <20040808113911.6185276c.pochini@shiny.it> In-Reply-To: <20040801092421.3f138fac.rddunlap@osdl.org> References: <20040801141931.6e026422.pochini@shiny.it> <20040801092421.3f138fac.rddunlap@osdl.org> X-Mailer: Sylpheed version 0.9.6 (GTK+ 1.2.10; powerpc-unknown-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 1 Aug 2004 09:24:21 -0700 "Randy.Dunlap" wrote: > On Sun, 1 Aug 2004 14:19:31 +0200 Giuliano Pochini wrote: > > | > | > | It seems that 2.6.7 reads the partition table only once: when the scsi unit > | is added to the list during boot or when I send the "add-single-device" > | command. This is a problem with removable devices because if the drive > | hasn't a disk inserted at boot time, attempting to mount a partitioned disk > | always results in: > | > | mount: /dev/sdb1 is not a valid block device > | > | Also, bad things may happen (not tested) if a disk was in during boot and I > | replace it with another one with a different partitioning. > | > | As a workaround I have to send the "remove-single-device" command after > | having unmounted a volume and "add-single-device" after I have inserted a > | new one. I don't know when this problem was introduced, sorry. > > I think that it's been this way for some time now... > > Does using > blockdev --rereadpt /dev/sdb1 > help? Yes, that works. But IIRC I didn't need that with linux 2.4. The kernel should reread the table automatically when it receaves an UNIT_ATTENTION message from the scsi bus. -- Giuliano.