From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758236Ab3LXETV (ORCPT ); Mon, 23 Dec 2013 23:19:21 -0500 Received: from mail.linux-iscsi.org ([67.23.28.174]:42506 "EHLO linux-iscsi.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752176Ab3LXETT (ORCPT ); Mon, 23 Dec 2013 23:19:19 -0500 Message-ID: <1387858826.5567.236.camel@haakon3.risingtidesystems.com> Subject: bio_integrity_verify() bug causing READ verify to be silently skipped From: "Nicholas A. Bellinger" To: "Martin K. Petersen" Cc: linux-scsi , LKML , Jens Axboe , Christoph Hellwig , Hannes Reinecke , James Bottomley Date: Mon, 23 Dec 2013 20:20:26 -0800 Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.4.4-1 Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Martin & Co, So after playing with the mainline DIF client against an initial WIP target DIF support patch, I've started hitting a bug in bio_integrity_verify() that causes READ verify logic to be silently skipped for both WIP target and existing scsi_debug DIF code. The issue is with the scsi_end_request() -> blk_end_request() completion path, where eventually blk_update_request() -> req_bio_endio() -> bio_advance() is called to increment bio->bi_idx to a non-zero value. Given that bio_integrity_verify() is using bio_for_each_segment(), the loop starts from the updated bio->bi_idx, and not a zero value, which ends up skipping individual bio segment calls to bi->verify_fn(). The following patch changes bio_integrity_verify() to use bio_for_each_segment_all() instead of bio_for_each_segment() to ensure that the segment walk always starts from zero, regardless of the current bio->bi_idx value after bio_advance(). Note this bug has been observed with v3.13-rc3 code, and I haven't yet looked back to figure out when this bug was first introduced.. Any ideas..? Interestingly enough, the scsi-mq alpha code does not suffer from this bug, as blk_end_request() is never called from scsi_mq_end_request() -> blk_mq_end_io() completion path code. Thank you, --nab >>From 32242942edca095e8dd126cb1408f2842340773e Mon Sep 17 00:00:00 2001 From: Nicholas Bellinger Date: Tue, 24 Dec 2013 04:00:24 +0000 Subject: [PATCH] bio-integrity: Fix bio_integrity_verify segment start bug This patch addresses a bug in bio_integrity_verify() code that has been causing DIF READ verify operations to be silently skipped. The issue is that bio->bi_idx will have been incremented within bio_advance() code in the normal blk_update_request() -> req_bio_endio() completion path, and bio_integrity_verify() is using bio_for_each_segment() which starts the bio segment walk at the current bio->bi_idx. So instead use bio_for_each_segment_all() to always start the bio segment walk from zero, regardless of the current bio->bi_idx value after bio_advance() has been called. Cc: Martin K. Petersen Cc: Jens Axboe Cc: Christoph Hellwig Signed-off-by: Nicholas Bellinger --- fs/bio-integrity.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/bio-integrity.c b/fs/bio-integrity.c index fc60b31..f1d5cde 100644 --- a/fs/bio-integrity.c +++ b/fs/bio-integrity.c @@ -450,7 +450,7 @@ static int bio_integrity_verify(struct bio *bio) bix.disk_name = bio->bi_bdev->bd_disk->disk_name; bix.sector_size = bi->sector_size; - bio_for_each_segment(bv, bio, i) { + bio_for_each_segment_all(bv, bio, i) { void *kaddr = kmap_atomic(bv->bv_page); bix.data_buf = kaddr + bv->bv_offset; bix.data_size = bv->bv_len; -- 1.7.10.4