From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from sj-iport-5.cisco.com (sj-iport-5.cisco.com [171.68.10.87]) by ozlabs.org (Postfix) with ESMTP id 019B267C33 for ; Thu, 23 Nov 2006 10:04:05 +1100 (EST) To: Hoang-Nam Nguyen Subject: Re: [PATCH 2.6.19] ehca: bug fix: use wqe offset instead wqe address to determine pending work requests References: <200611202354.13030.hnguyen@linux.vnet.ibm.com> <200611221029.10077.hnguyen@linux.vnet.ibm.com> From: Roland Dreier Date: Wed, 22 Nov 2006 15:04:01 -0800 In-Reply-To: <200611221029.10077.hnguyen@linux.vnet.ibm.com> (Hoang-Nam Nguyen's message of "Wed, 22 Nov 2006 10:29:09 +0100") Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linuxppc-dev@ozlabs.org, linux-kernel@vger.kernel.org, openib-general@openib.org, raisch@de.ibm.com List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , > We found this bug actually through a code review by random. Since > (un)fortunately the queue pages were layouted in order, we've not > seen it earlier. It's certainly a bug and can cause kernel panic > if above observation is not met, probably in stress situation > of system. That means the "former" code accesses next page that > it has not allocated. OK. After thinking about this, I'm going to queue it for 2.6.20 -- we're _way_ too close to the 2.6.19 final release to put in patches that aren't either small and obvious, or fix a problem someone hit in real life. - R.