From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Martin K. Petersen" Subject: Re: SG does not ignore dxferp (direct io + mmap) Date: Thu, 01 Dec 2016 08:40:31 -0500 Message-ID: References: <1479752642.19792.43.camel@localhost.localdomain> <20161122083759.xeifuex3xxfimuwz@linux-x5ow.site> <1479839407.28416.21.camel@localhost.localdomain> <2146476957.2165908.1479927335303.JavaMail.zimbra@redhat.com> <1479932524.28416.43.camel@localhost.localdomain> <20161125080758.5bh5jkcgvhw3xuvb@linux-x5ow.site> <1194718949.74785.1480096576577.JavaMail.zimbra@redhat.com> <1480523188.28416.94.camel@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain Return-path: Received: from userp1040.oracle.com ([156.151.31.81]:38617 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754600AbcLANlQ (ORCPT ); Thu, 1 Dec 2016 08:41:16 -0500 In-Reply-To: <1480523188.28416.94.camel@localhost.localdomain> (Ewan D. Milne's message of "Wed, 30 Nov 2016 11:26:28 -0500") Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Christoph Hellwig Cc: "Ewan D. Milne" , Johannes Thumshirn , Laurence Oberman , Eyal Ben David , dgilbert@interlog.com, linux-scsi@vger.kernel.org >>>>> "Ewan" == Ewan D Milne writes: >> I think what we need to understand is what caused the regression in >> the first place, I probably should have been bisecting the original >> failure rather than trying to find where it started working. Ewan> Bisecting leads to this commit: commit 37f19e57a0de3c4a3417aa13ff4d04f1e0dee4b3 Author: Christoph Hellwig Date: Sun Jan 18 16:16:33 2015 +0100 block: merge __bio_map_user_iov into bio_map_user_iov And also remove the unused bdev argument. Signed-off-by: Christoph Hellwig Reviewed-by: Ming Lei Signed-off-by: Jens Axboe Specifically, the problem appears to be caused by the removal of the setting of bio->bi_bdev, which would previously be set to NULL. If I add: diff --git a/block/bio.c b/block/bio.c index 0723d4c..ecac37b 100644 --- a/block/bio.c +++ b/block/bio.c @@ -1351,6 +1351,7 @@ struct bio *bio_map_user_iov(struct request_queue *q, if (iter->type & WRITE) bio->bi_rw |= REQ_WRITE; + bio->bi_bdev = NULL; bio->bi_flags |= (1 << BIO_USER_MAPPED); /* Ewan> The test passes (no zero byte corruption). Ewan> Setting dxferp would cause map_data.null_mapped to be set before Ewan> it is passed to blk_rq_map_user(_iov) which would cause a Ewan> difference in behavior. Christoph? -- Martin K. Petersen Oracle Linux Engineering