From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Snitzer Subject: Re: scsi_error: do not allow IO errors with certain ILLEGAL_REQUEST sense to be retryable Date: Tue, 6 Dec 2011 16:27:04 -0500 Message-ID: <20111206212704.GB30719@redhat.com> References: <1322857889-2623-1-git-send-email-snitzer@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mx1.redhat.com ([209.132.183.28]:45069 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753338Ab1LFV1O (ORCPT ); Tue, 6 Dec 2011 16:27:14 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: "Martin K. Petersen" Cc: linux-scsi@vger.kernel.org, James Bottomley , Hannes Reinecke On Tue, Dec 06 2011 at 4:07pm -0500, Martin K. Petersen wrote: > >>>>> "Mike" == Mike Snitzer writes: > > Mike> Thin provisioned LUNs from multiple array vendors have failed > Mike> WRITE SAME(16) w/ UNMAP bit set with ILLEGAL_REQUEST sense. > > Can you share what these arrays are returning in block limits and the > logical block provisioning VPD? It sounds like they are returning > inconsistent data causing us to issue the wrong command... I don't have that info. But I can ask for it. In the first instance the array had discard alignment constraints (needed to be a multiple of 4k, vendor has since relaxed that). For the second case, the reason for why the array failed the CDB hasn't been shared yet (other than it was an "Invalid field in cdb"). Regardless, shouldn't the SCSI midlayer classify such ILLEGAL_REQUEST sense, with an add. sense I listed in the patch, as a target error?