Return-Path: Received: from natasha.panasas.com ([67.152.220.90]:52137 "EHLO natasha.panasas.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755827Ab1JDKcu (ORCPT ); Tue, 4 Oct 2011 06:32:50 -0400 From: Boaz Harrosh To: Trond Myklebust , Benny Halevy , Brent Welch , NFS list , open-osd Subject: [PATCH 09/19] ore: Support for short read/writes Date: Tue, 4 Oct 2011 12:32:41 +0200 Message-ID: <1317724361-27575-1-git-send-email-bharrosh@panasas.com> In-Reply-To: <4E8ADEDA.4050709@panasas.com> References: <4E8ADEDA.4050709@panasas.com> Content-Type: text/plain Sender: linux-nfs-owner@vger.kernel.org List-ID: MIME-Version: 1.0 Memory conditions and max_bio constraints might cause us to not comply to the full length of the requested IO. Instead of failing the complete IO we can issue a shorter read/write and report how much was actually executed in the ios->length member. All users must check ios->length at IO_done or upon return of ore_read/write and re-issue the reminder of the bytes. Because other wise there is no error returned like before. This is part of the effort to support the pnfs-obj layout driver. Signed-off-by: Boaz Harrosh --- fs/exofs/ore.c | 30 +++++++++++++++++++++++------- 1 files changed, 23 insertions(+), 7 deletions(-) diff --git a/fs/exofs/ore.c b/fs/exofs/ore.c index 2f39f23..0b992e1 100644 --- a/fs/exofs/ore.c +++ b/fs/exofs/ore.c @@ -377,8 +377,8 @@ static int _add_stripe_unit(struct ore_io_state *ios, unsigned *cur_pg, unsigned pg = *cur_pg; struct request_queue *q = osd_request_queue(_ios_od(ios, per_dev->dev)); - - per_dev->length += cur_len; + unsigned len = cur_len; + int ret; if (per_dev->bio == NULL) { unsigned pages_in_stripe = ios->layout->group_width * @@ -390,7 +390,8 @@ static int _add_stripe_unit(struct ore_io_state *ios, unsigned *cur_pg, if (unlikely(!per_dev->bio)) { ORE_DBGMSG("Failed to allocate BIO size=%u\n", bio_size); - return -ENOMEM; + ret = -ENOMEM; + goto out; } } @@ -403,15 +404,24 @@ static int _add_stripe_unit(struct ore_io_state *ios, unsigned *cur_pg, added_len = bio_add_pc_page(q, per_dev->bio, ios->pages[pg], pglen, pgbase); - if (unlikely(pglen != added_len)) - return -ENOMEM; + if (unlikely(pglen != added_len)) { + ret = -ENOMEM; + goto out; + } pgbase = 0; ++pg; } BUG_ON(cur_len); + per_dev->length += len; *cur_pg = pg; - return 0; + ret = 0; +out: /* we fail the complete unit on an error eg don't advance + * per_dev->length and cur_pg. This means that we might have a bigger + * bio than the CDB requested length (per_dev->length). That's fine + * only the oposite is fatal. + */ + return ret; } static int _prepare_for_striping(struct ore_io_state *ios) @@ -476,7 +486,13 @@ static int _prepare_for_striping(struct ore_io_state *ios) out: ios->numdevs = max_comp + mirrors_p1; ios->pages_consumed = cur_pg; - return ret; + if (unlikely(ret)) { + if (length == ios->length) + return ret; + else + ios->length -= length; + } + return 0; } int ore_create(struct ore_io_state *ios) -- 1.7.2.3