From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Martin K. Petersen" Subject: Re: [RFC] Simlify dif_verify routines and fixup fileio protection information code. Date: Wed, 15 Apr 2015 11:05:33 -0400 Message-ID: References: <1428945575-30839-1-git-send-email-sagig@mellanox.com> <552D4C7B.9000402@dev.mellanox.co.il> <552E384F.3030203@dev.mellanox.co.il> <552E76D1.7040204@dev.mellanox.co.il> Mime-Version: 1.0 Content-Type: text/plain Return-path: In-Reply-To: <552E76D1.7040204@dev.mellanox.co.il> (Sagi Grimberg's message of "Wed, 15 Apr 2015 17:33:53 +0300") Sender: target-devel-owner@vger.kernel.org To: Sagi Grimberg Cc: Akinobu Mita , "Nicholas A. Bellinger" , target-devel@vger.kernel.org, "linux-scsi@vger.kernel.org" , "Martin K. Petersen" , Christoph Hellwig List-Id: linux-scsi@vger.kernel.org >>>>> "Sagi" == Sagi Grimberg writes: Sagi> I don't think this is sufficient. With this we actually write Sagi> unprotected data for WRITE_SAME (i.e. write data blocks but not Sagi> storing the corresponding PI information). When this data will be Sagi> read back you will see PI errors (you currently don't see those Sagi> because your backend drive contains escape values I assume). Sagi> I'd say the correct fix is to calc PI for the block Indeed! Sagi> and "write_same" it... Well, the ref tag needs to be incremented for each block (for Type 1). -- Martin K. Petersen Oracle Linux Engineering