From patchwork Tue Apr 22 21:29:20 2014 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Weston Andros Adamson X-Patchwork-Id: 4036141 Return-Path: X-Original-To: patchwork-linux-nfs@patchwork.kernel.org Delivered-To: patchwork-parsemail@patchwork1.web.kernel.org Received: from mail.kernel.org (mail.kernel.org [198.145.19.201]) by patchwork1.web.kernel.org (Postfix) with ESMTP id 43C959F1F4 for ; Tue, 22 Apr 2014 21:30:04 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id 3E155201EC for ; Tue, 22 Apr 2014 21:30:00 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 14C8620136 for ; Tue, 22 Apr 2014 21:29:56 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758643AbaDVV3W (ORCPT ); Tue, 22 Apr 2014 17:29:22 -0400 Received: from mail-ig0-f169.google.com ([209.85.213.169]:37421 "EHLO mail-ig0-f169.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757883AbaDVV3Q (ORCPT ); Tue, 22 Apr 2014 17:29:16 -0400 Received: by mail-ig0-f169.google.com with SMTP id h18so3388324igc.0 for ; Tue, 22 Apr 2014 14:29:15 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20130820; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=uc0w5zBZ7LO0kbUaaxODjvXWKOgIcLeX06nu3Zt8Hgw=; b=PzjakaB+JnXPHH+Mn8ywBlkwEc2gpgc5XwUuJQkCE7EYEnB/oQ3dXE9+0+5An0oZ+N y5VTu1Hfq1mZvq6HsULL5DWxqh5WTHMS1NkTcdHGFSI6YUelA+JeUKfRUQO3x8gjA5gh xngSx0N9kZ8655rNCjKLEAcff+brTY2M7KZaNXWTenBj5maW+CmhVipa21nUAIXWuDiF VOR9GQiGHl6jbbeF0ymfZe+sIT70oXj9/ksbHUEzqkaH8BkFhAHoqBVHN1FIWZHNeBZB KNmD31YE12cdxutllS/RXyR2ldo0hbQULDehDH4GZPbYDCx2ShWstoUd88UAQjxKAYLX W3bA== X-Gm-Message-State: ALoCoQksPWgdb+w7WPzTYC69DUuKEseSmCHPmYxrkFXog74VzxPbu27aTdUlsIkbyWWQKcizu+BF X-Received: by 10.50.30.6 with SMTP id o6mr253474igh.43.1398202155366; Tue, 22 Apr 2014 14:29:15 -0700 (PDT) Received: from gavrio-wifi.robotsandstuff.fake (c-98-209-19-144.hsd1.mi.comcast.net. [98.209.19.144]) by mx.google.com with ESMTPSA id j9sm31115582igu.10.2014.04.22.14.29.14 for (version=TLSv1 cipher=RC4-SHA bits=128/128); Tue, 22 Apr 2014 14:29:14 -0700 (PDT) From: Weston Andros Adamson To: trond.myklebust@primarydata.com Cc: linux-nfs@vger.kernel.org, Weston Andros Adamson Subject: [PATCH 12/17] nfs: use > 1 request to handle bsize < PAGE_SIZE Date: Tue, 22 Apr 2014 17:29:20 -0400 Message-Id: <1398202165-78897-13-git-send-email-dros@primarydata.com> X-Mailer: git-send-email 1.8.5.2 (Apple Git-48) In-Reply-To: <1398202165-78897-1-git-send-email-dros@primarydata.com> References: <1398202165-78897-1-git-send-email-dros@primarydata.com> Sender: linux-nfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-nfs@vger.kernel.org X-Spam-Status: No, score=-7.5 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_HI, RP_MATCHES_RCVD, UNPARSEABLE_RELAY autolearn=ham version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on mail.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP Use the newly added support for multiple requests per page for rsize/wsize < PAGE_SIZE, instead of having multiple read / write data structures per pageio header. This allows us to get rid of nfs_flush_multi and nfs_pagein_multi. Signed-off-by: Weston Andros Adamson --- fs/nfs/pagelist.c | 22 +++++++------------ fs/nfs/read.c | 63 ++++++------------------------------------------------ fs/nfs/write.c | 64 ++++++------------------------------------------------- 3 files changed, 22 insertions(+), 127 deletions(-) diff --git a/fs/nfs/pagelist.c b/fs/nfs/pagelist.c index b5c4c13..e819b1b 100644 --- a/fs/nfs/pagelist.c +++ b/fs/nfs/pagelist.c @@ -434,21 +434,13 @@ nfs_wait_on_request(struct nfs_page *req) size_t nfs_generic_pg_test(struct nfs_pageio_descriptor *desc, struct nfs_page *prev, struct nfs_page *req) { - if (!prev) - return req->wb_bytes; - /* - * FIXME: ideally we should be able to coalesce all requests - * that are not block boundary aligned, but currently this - * is problematic for the case of bsize < PAGE_CACHE_SIZE, - * since nfs_flush_multi and nfs_pagein_multi assume you - * can have only one struct nfs_page. - */ - if (desc->pg_bsize < PAGE_SIZE) + if (desc->pg_count > desc->pg_bsize) { + /* should never happen */ + WARN_ON_ONCE(1); return 0; + } - if (desc->pg_count + req->wb_bytes <= desc->pg_bsize) - return req->wb_bytes; - return 0; + return min(desc->pg_bsize - desc->pg_count, (size_t)req->wb_bytes); } EXPORT_SYMBOL_GPL(nfs_generic_pg_test); @@ -526,7 +518,9 @@ static bool nfs_can_coalesce_requests(struct nfs_page *prev, return false; } size = pgio->pg_ops->pg_test(pgio, prev, req); - WARN_ON_ONCE(size && size != req->wb_bytes); + WARN_ON_ONCE(size > req->wb_bytes); + if (size && size < req->wb_bytes) + req->wb_bytes = size; return size > 0; } diff --git a/fs/nfs/read.c b/fs/nfs/read.c index c774810..daeff0c 100644 --- a/fs/nfs/read.c +++ b/fs/nfs/read.c @@ -336,56 +336,12 @@ static void nfs_pagein_error(struct nfs_pageio_descriptor *desc, desc->pg_completion_ops->error_cleanup(&desc->pg_list); } -/* - * Generate multiple requests to fill a single page. - * - * We optimize to reduce the number of read operations on the wire. If we - * detect that we're reading a page, or an area of a page, that is past the - * end of file, we do not generate NFS read operations but just clear the - * parts of the page that would have come back zero from the server anyway. - * - * We rely on the cached value of i_size to make this determination; another - * client can fill pages on the server past our cached end-of-file, but we - * won't see the new data until our attribute cache is updated. This is more - * or less conventional NFS client behavior. - */ -static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc, - struct nfs_pgio_header *hdr) -{ - struct nfs_page *req = hdr->req; - struct page *page = req->wb_page; - struct nfs_read_data *data; - size_t rsize = desc->pg_bsize, nbytes; - unsigned int offset; - - offset = 0; - nbytes = desc->pg_count; - do { - size_t len = min(nbytes,rsize); - - data = nfs_readdata_alloc(hdr, 1); - if (!data) { - nfs_pagein_error(desc, hdr); - return -ENOMEM; - } - data->pages.pagevec[0] = page; - nfs_read_rpcsetup(data, len, offset); - list_add(&data->list, &hdr->rpc_list); - nbytes -= len; - offset += len; - } while (nbytes != 0); - - nfs_list_remove_request(req); - nfs_list_add_request(req, &hdr->pages); - desc->pg_rpc_callops = &nfs_read_common_ops; - return 0; -} - -static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, - struct nfs_pgio_header *hdr) +int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { struct nfs_page *req; struct page **pages; + struct page *last_page; struct nfs_read_data *data; struct list_head *head = &desc->pg_list; @@ -397,11 +353,14 @@ static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, } pages = data->pages.pagevec; + last_page = NULL; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); nfs_list_add_request(req, &hdr->pages); - *pages++ = req->wb_page; + if (last_page != req->wb_page) + *pages++ = req->wb_page; + last_page = req->wb_page; } nfs_read_rpcsetup(data, desc->pg_count, 0); @@ -409,14 +368,6 @@ static int nfs_pagein_one(struct nfs_pageio_descriptor *desc, desc->pg_rpc_callops = &nfs_read_common_ops; return 0; } - -int nfs_generic_pagein(struct nfs_pageio_descriptor *desc, - struct nfs_pgio_header *hdr) -{ - if (desc->pg_bsize < PAGE_CACHE_SIZE) - return nfs_pagein_multi(desc, hdr); - return nfs_pagein_one(desc, hdr); -} EXPORT_SYMBOL_GPL(nfs_generic_pagein); static int nfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc) diff --git a/fs/nfs/write.c b/fs/nfs/write.c index 43892e0..f40db93 100644 --- a/fs/nfs/write.c +++ b/fs/nfs/write.c @@ -1233,52 +1233,6 @@ static void nfs_flush_error(struct nfs_pageio_descriptor *desc, } /* - * Generate multiple small requests to write out a single - * contiguous dirty area on one page. - */ -static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, - struct nfs_pgio_header *hdr) -{ - struct nfs_page *req = hdr->req; - struct page *page = req->wb_page; - struct nfs_write_data *data; - size_t wsize = desc->pg_bsize, nbytes; - unsigned int offset; - int requests = 0; - struct nfs_commit_info cinfo; - - nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq); - - if ((desc->pg_ioflags & FLUSH_COND_STABLE) && - (desc->pg_moreio || nfs_reqs_to_commit(&cinfo) || - desc->pg_count > wsize)) - desc->pg_ioflags &= ~FLUSH_COND_STABLE; - - - offset = 0; - nbytes = desc->pg_count; - do { - size_t len = min(nbytes, wsize); - - data = nfs_writedata_alloc(hdr, 1); - if (!data) { - nfs_flush_error(desc, hdr); - return -ENOMEM; - } - data->pages.pagevec[0] = page; - nfs_write_rpcsetup(data, len, offset, desc->pg_ioflags, &cinfo); - list_add(&data->list, &hdr->rpc_list); - requests++; - nbytes -= len; - offset += len; - } while (nbytes != 0); - nfs_list_remove_request(req); - nfs_list_add_request(req, &hdr->pages); - desc->pg_rpc_callops = &nfs_write_common_ops; - return 0; -} - -/* * Create an RPC task for the given write request and kick it. * The page must have been locked by the caller. * @@ -1286,11 +1240,12 @@ static int nfs_flush_multi(struct nfs_pageio_descriptor *desc, * This is the case if nfs_updatepage detects a conflicting request * that has been written but not committed. */ -static int nfs_flush_one(struct nfs_pageio_descriptor *desc, - struct nfs_pgio_header *hdr) +int nfs_generic_flush(struct nfs_pageio_descriptor *desc, + struct nfs_pgio_header *hdr) { struct nfs_page *req; struct page **pages; + struct page *last_page; struct nfs_write_data *data; struct list_head *head = &desc->pg_list; struct nfs_commit_info cinfo; @@ -1304,11 +1259,14 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc, nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq); pages = data->pages.pagevec; + last_page = NULL; while (!list_empty(head)) { req = nfs_list_entry(head->next); nfs_list_remove_request(req); nfs_list_add_request(req, &hdr->pages); - *pages++ = req->wb_page; + if (last_page != req->wb_page) + *pages++ = req->wb_page; + last_page = req->wb_page; } if ((desc->pg_ioflags & FLUSH_COND_STABLE) && @@ -1321,14 +1279,6 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc, desc->pg_rpc_callops = &nfs_write_common_ops; return 0; } - -int nfs_generic_flush(struct nfs_pageio_descriptor *desc, - struct nfs_pgio_header *hdr) -{ - if (desc->pg_bsize < PAGE_CACHE_SIZE) - return nfs_flush_multi(desc, hdr); - return nfs_flush_one(desc, hdr); -} EXPORT_SYMBOL_GPL(nfs_generic_flush); static int nfs_generic_pg_writepages(struct nfs_pageio_descriptor *desc)