From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932825Ab0I0IgO (ORCPT ); Mon, 27 Sep 2010 04:36:14 -0400 Received: from mga02.intel.com ([134.134.136.20]:46433 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932793Ab0I0IgN (ORCPT ); Mon, 27 Sep 2010 04:36:13 -0400 Message-Id: <8u3s8d$jmp9g4@orsmga001.jf.intel.com> X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.57,242,1283756400"; d="scan'208";a="661431812" Date: Mon, 27 Sep 2010 09:36:09 +0100 To: Hugh Dickins Subject: Re: How best to pin pages in physical memory? Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, Rik van Riel , Christoph Hellwig References: <8u3s8d$jmkug0@orsmga001.jf.intel.com> From: Chris Wilson In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 26 Sep 2010 16:29:23 -0700 (PDT), Hugh Dickins wrote: > If i915_get_object_get_pages() isn't doing its job, I think you need to > wonder why not - maybe somewhere doing a put_page or page_cache_release, > freeing one or all pages too soon? Thanks, that's a very useful bit of review. I surmised that after seeing corruption in a batch buffer after a swap storm that the pages were not as safe required. The fact that swapping was involved could have been a coincidence, except for the growing number of reports from other users reporting crashes in conjunction with swapping. So back to seeing whether i915_gem_object_[get/put]_pages is sufficient for our uses. -Chris -- Chris Wilson, Intel Open Source Technology Centre