From: Matthias Prager Subject: Re: mpt2sas + raid10 goes boom Date: Tue, 09 Apr 2013 12:11:13 +0200 Message-ID: <5163E941.1050907@matthiasprager.de> References: <5162C25C.3090707@matthiasprager.de> <5162F827.7010603@matthiasprager.de> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Cc: linux-raid , linux-ext4@vger.kernel.org, Chris Dunlop , Matthias Prager To: Baruch Even , linux-scsi Return-path: Received: from dd15408.kasserver.com ([85.13.136.168]:46432 "EHLO dd15408.kasserver.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S935101Ab3DIKLU (ORCPT ); Tue, 9 Apr 2013 06:11:20 -0400 In-Reply-To: Sender: linux-ext4-owner@vger.kernel.org List-ID: Thanks for your insights Baruch. The crc count did not increase any further - so this was probably just small oddity (was zero before when the write-same issue already happened). The real issue however does persist. I found a way to reliably trigger the log messages. Using a program called checksum over a photo share (which does a lot of reads and one write per file). With that in place I switched to the 3.4.38 kernel, with which I'm unable to trigger the problem. I will leave the system at that for now, and try to reproduce it on my testing machine to see, whether c8dc9c6 md: raid1,10: Handle REQ_WRITE_SAME flag in write bios works for me. If c8dc9c6 does the trick, it would still be interesting to know why and how this triggered i/o errors, the strange log message from lsi and 'Resets Between Cmd Acceptance and Completion' as one of the drives says. Would that mean the driver/firmware from lsi is issuing or passing on commands to the drive which it does not understand or can't process? --- Matthias