Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1750711AbWEQQsG (ORCPT ); Wed, 17 May 2006 12:48:06 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750716AbWEQQsF (ORCPT ); Wed, 17 May 2006 12:48:05 -0400 Received: from osa.unixfolk.com ([209.204.179.118]:25300 "EHLO osa.unixfolk.com") by vger.kernel.org with ESMTP id S1750711AbWEQQsD (ORCPT ); Wed, 17 May 2006 12:48:03 -0400 Date: Wed, 17 May 2006 09:47:59 -0700 (PDT) From: Dave Olson To: Roland Dreier Cc: "Bryan O'Sullivan" , openib-general@openib.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH 35 of 53] ipath - some interrelated stability and cleanliness fixes In-Reply-To: Message-ID: References: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1101 Lines: 29 On Mon, 15 May 2006, Roland Dreier wrote: | This looks like a pastiche of several patches. Why can't it be split | up into logical pieces? | | > Call dma_free_coherent without ipath_mutex held. | | Why? Doesn't freeing work with the mutex held? Sure, that's the way the previous code worked. We are seeing a bug (with both our driver native MPI processes and mthca mvapic), where when 8 processes using "simultaneously exit", we get watchdogs and/or hangs in the close routines. Moving the freeing outside the mutex was an attempt to see if we were running into some VM issues by doing lots of page unlocking and freeing with the mutex held. It seemed to help somewhat, but not to solve the problem. It also allows other processes to open and close in a somewhat more timely fashion. Dave Olson olson@unixfolk.com http://www.unixfolk.com/dave - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/