From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756594Ab1BRIRE (ORCPT ); Fri, 18 Feb 2011 03:17:04 -0500 Received: from moutng.kundenserver.de ([212.227.126.186]:58509 "EHLO moutng.kundenserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753608Ab1BRIRC (ORCPT ); Fri, 18 Feb 2011 03:17:02 -0500 From: Arnd Bergmann To: Andrew Morton Subject: Re: [Bugme-new] [Bug 29172] New: releasing loop on top of other loop leads to deadlock Date: Fri, 18 Feb 2011 09:16:42 +0100 User-Agent: KMail/1.12.2 (Linux/2.6.37; KDE/4.3.2; x86_64; ; ) Cc: "Rafael J. Wysocki" , Maciej Rutecki , linux-kernel@vger.kernel.org References: <20110217131244.bb64711a.akpm@linux-foundation.org> In-Reply-To: <20110217131244.bb64711a.akpm@linux-foundation.org> MIME-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <201102180916.42494.arnd@arndb.de> X-Provags-ID: V02:K0:HgBPZ1N0hc4q2ns73cjMnpfkX1H9TRxREMfLvP+rU9p c/qEG23RzEOe4V7WZnQv/H7U7VsapfBOF3dR2BdQUblcNscHkD 7xXw4DeAlAY4RjQO9xbl79B5zFgOdAuUMhmmzCWWtA5NrWykNk sxyRbh13A9b+R5l6XKtAMIzIAK5Z4TmeD72/H0l7DieKUq0tBg lOHxYRWek//gMha3yA9JA== Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thursday 17 February 2011, Andrew Morton wrote: > On Tue, 15 Feb 2011 10:00:57 GMT > bugzilla-daemon@bugzilla.kernel.org wrote: > > > https://bugzilla.kernel.org/show_bug.cgi?id=29172 > > post-2.6.37 deadlock in loop. I'd agree with the reporter that Arnd might be involved ;) Yes, sounds right, we should fix it and get the fix into stable. This comes from an automated conversion that was done under the assumption that it cannot break from making the locking stricter. A related bug happened in the nbd driver, and I think the solution here should also be to remove the new mutex. Upon closer review of the code, I cannot see any reason why it would actually be needed: there is no global data to be protected and all device specific data is protected by another mutex. Arnd