From patchwork Fri Mar 10 22:07:46 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Benjamin Coddington X-Patchwork-Id: 9618059 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id 7909660415 for ; Fri, 10 Mar 2017 22:07:53 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 69875287A2 for ; Fri, 10 Mar 2017 22:07:53 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 5DEE7287AA; Fri, 10 Mar 2017 22:07:53 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.9 required=2.0 tests=BAYES_00,RCVD_IN_DNSWL_HI autolearn=unavailable version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id E968C287A2 for ; Fri, 10 Mar 2017 22:07:52 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932904AbdCJWHu (ORCPT ); Fri, 10 Mar 2017 17:07:50 -0500 Received: from mx1.redhat.com ([209.132.183.28]:34438 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932482AbdCJWHt (ORCPT ); Fri, 10 Mar 2017 17:07:49 -0500 Received: from int-mx13.intmail.prod.int.phx2.redhat.com (int-mx13.intmail.prod.int.phx2.redhat.com [10.5.11.26]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 4547785545; Fri, 10 Mar 2017 22:07:49 +0000 (UTC) Received: from bcodding.csb (vpn-48-189.rdu2.redhat.com [10.10.48.189]) by int-mx13.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id v2AM7mfN022714; Fri, 10 Mar 2017 17:07:48 -0500 Received: by bcodding.csb (Postfix, from userid 24008) id 271FA10C3105; Fri, 10 Mar 2017 17:07:47 -0500 (EST) From: Benjamin Coddington To: Alexander Viro , Trond Myklebust , Anna Schumaker Cc: linux-nfs@vger.kernel.org, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org Subject: [PATCH v2] NFS: switch back to to ->iterate() Date: Fri, 10 Mar 2017 17:07:46 -0500 Message-Id: <4a243778db6f9ccd052364ba2e4dded4390d1cc0.1489183437.git.bcodding@redhat.com> X-Scanned-By: MIMEDefang 2.68 on 10.5.11.26 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.28]); Fri, 10 Mar 2017 22:07:49 +0000 (UTC) Sender: linux-nfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-nfs@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP NFS has some optimizations for readdir to choose between using READDIR or READDIRPLUS based on workload, and which NFS operation to use is determined by subsequent interactions with lookup, d_revalidate, and getattr. Concurrent use of nfs_readdir() via ->iterate_shared() can cause those optimizations to repeatedly invalidate the pagecache used to store directory entries during readdir(), which causes some very bad performance for directories with many entries (more than about 10000). There's a couple ways to fix this in NFS, but no fix would be as simple as going back to ->iterate() to serialize nfs_readdir(), and neither fix I tested performed as well as going back to ->iterate(). The first required taking the directory's i_lock for each entry, with the result of terrible contention. The second way adds another flag to the nfs_inode, and so keeps the optimizations working for large directories. The difference from using ->iterate() here is that much more memory is consumed for a given workload without any performance gain. The workings of nfs_readdir() are such that concurrent users are serialized within read_cache_page() waiting to retrieve pages of entries from the server. By serializing this work in iterate_dir() instead, contention for cache pages is reduced. Waiting processes can have an uncontended pass at the entirety of the directory's pagecache once previous processes have completed filling it. v2 - Keep the bits needed for parallel lookup Signed-off-by: Benjamin Coddington --- fs/nfs/dir.c | 37 ++++++++++++------------------------- 1 file changed, 12 insertions(+), 25 deletions(-) diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index fad81041f5ab..18cdf11a1720 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -57,7 +57,7 @@ static void nfs_readdir_clear_array(struct page*); const struct file_operations nfs_dir_operations = { .llseek = nfs_llseek_dir, .read = generic_read_dir, - .iterate_shared = nfs_readdir, + .iterate = nfs_readdir, .open = nfs_opendir, .release = nfs_closedir, .fsync = nfs_fsync_dir, @@ -145,7 +145,6 @@ struct nfs_cache_array_entry { }; struct nfs_cache_array { - atomic_t refcount; int size; int eof_index; u64 last_cookie; @@ -201,20 +200,11 @@ void nfs_readdir_clear_array(struct page *page) int i; array = kmap_atomic(page); - if (atomic_dec_and_test(&array->refcount)) - for (i = 0; i < array->size; i++) - kfree(array->array[i].string.name); + for (i = 0; i < array->size; i++) + kfree(array->array[i].string.name); kunmap_atomic(array); } -static bool grab_page(struct page *page) -{ - struct nfs_cache_array *array = kmap_atomic(page); - bool res = atomic_inc_not_zero(&array->refcount); - kunmap_atomic(array); - return res; -} - /* * the caller is responsible for freeing qstr.name * when called by nfs_readdir_add_to_array, the strings will be freed in @@ -680,7 +670,6 @@ int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, goto out_label_free; } memset(array, 0, sizeof(struct nfs_cache_array)); - atomic_set(&array->refcount, 1); array->eof_index = -1; status = nfs_readdir_alloc_pages(pages, array_size); @@ -743,7 +732,8 @@ int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page) static void cache_page_release(nfs_readdir_descriptor_t *desc) { - nfs_readdir_clear_array(desc->page); + if (!desc->page->mapping) + nfs_readdir_clear_array(desc->page); put_page(desc->page); desc->page = NULL; } @@ -751,16 +741,8 @@ void cache_page_release(nfs_readdir_descriptor_t *desc) static struct page *get_cache_page(nfs_readdir_descriptor_t *desc) { - struct page *page; - - for (;;) { - page = read_cache_page(desc->file->f_mapping, + return read_cache_page(desc->file->f_mapping, desc->page_index, (filler_t *)nfs_readdir_filler, desc); - if (IS_ERR(page) || grab_page(page)) - break; - put_page(page); - } - return page; } /* @@ -966,11 +948,13 @@ static int nfs_readdir(struct file *file, struct dir_context *ctx) static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int whence) { + struct inode *inode = file_inode(filp); struct nfs_open_dir_context *dir_ctx = filp->private_data; dfprintk(FILE, "NFS: llseek dir(%pD2, %lld, %d)\n", filp, offset, whence); + inode_lock(inode); switch (whence) { case 1: offset += filp->f_pos; @@ -978,13 +962,16 @@ static loff_t nfs_llseek_dir(struct file *filp, loff_t offset, int whence) if (offset >= 0) break; default: - return -EINVAL; + offset = -EINVAL; + goto out; } if (offset != filp->f_pos) { filp->f_pos = offset; dir_ctx->dir_cookie = 0; dir_ctx->duped = 0; } +out: + inode_unlock(inode); return offset; }