From patchwork Mon Aug 5 06:10:30 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Yan, Zheng" X-Patchwork-Id: 2838536 Return-Path: X-Original-To: patchwork-ceph-devel@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 061339F479 for ; Mon, 5 Aug 2013 06:10:51 +0000 (UTC) Received: from mail.kernel.org (localhost [127.0.0.1]) by mail.kernel.org (Postfix) with ESMTP id C4B382011C for ; Mon, 5 Aug 2013 06:10:49 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 2AC792017E for ; Mon, 5 Aug 2013 06:10:48 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754714Ab3HEGKp (ORCPT ); Mon, 5 Aug 2013 02:10:45 -0400 Received: from mga14.intel.com ([143.182.124.37]:63899 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754703Ab3HEGKn (ORCPT ); Mon, 5 Aug 2013 02:10:43 -0400 Received: from azsmga001.ch.intel.com ([10.2.17.19]) by azsmga102.ch.intel.com with ESMTP; 04 Aug 2013 23:10:43 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.89,816,1367996400"; d="scan'208";a="341696574" Received: from zyan5-mobl.sh.intel.com ([10.239.13.8]) by azsmga001.ch.intel.com with ESMTP; 04 Aug 2013 23:10:42 -0700 From: "Yan, Zheng" To: ceph-devel@vger.kernel.org Cc: sage@inktank.com, "Yan, Zheng" Subject: [PATCH 3/3] ceph: rework trim caps code Date: Mon, 5 Aug 2013 14:10:30 +0800 Message-Id: <1375683030-28305-10-git-send-email-zheng.z.yan@intel.com> X-Mailer: git-send-email 1.8.1.4 In-Reply-To: <1375683030-28305-1-git-send-email-zheng.z.yan@intel.com> References: <1375683030-28305-1-git-send-email-zheng.z.yan@intel.com> Sender: ceph-devel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: ceph-devel@vger.kernel.org X-Spam-Status: No, score=-6.9 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_HI, RP_MATCHES_RCVD, UNPARSEABLE_RELAY autolearn=unavailable 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 From: "Yan, Zheng" The trim caps code that handles SESSION_RECALL_STATE message has two issues. First, it uses d_prune_aliases() to prune dentries. This confuses our 'dir complete' check, because d_prune_aliases() unhashes dentries before calling our d_prune() callback. Second, it only prune dentries, inodes with zero reference are still in the icache unless there is memory pressure. The fix is adding d_delete() and drop_inode() VFS callback. VFS calls our d_delete()/drop_inode() callback when a dentry/inode's last reference is dropped. If true is returned, VFS drop the dentry/inode from the cache. When handling SESSION_RECALL_STATE message, we mark inode with CEPH_I_TRIMCAPS flag. Our callbacks check the flag and session->s_trim_caps to decide if VFS should drop a dentry/inode. To trim an inode, we just need to reference dentries associated with the inode, then release them. Signed-off-by: Yan, Zheng --- fs/ceph/caps.c | 30 ++++++++++++++++++++++++++++++ fs/ceph/dir.c | 12 ++++++++++++ fs/ceph/inode.c | 11 +++++++++++ fs/ceph/mds_client.c | 37 ++++++++++++++++++++++++++++++------- fs/ceph/mds_client.h | 4 ++-- fs/ceph/super.c | 1 + fs/ceph/super.h | 3 +++ 7 files changed, 89 insertions(+), 9 deletions(-) diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 49590a0..332dc8d 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -882,6 +882,36 @@ int __ceph_caps_mds_wanted(struct ceph_inode_info *ci) } /* + * Used by ceph_d_delete()/ceph_drop_inode(). check if we should drop + * the inode. + */ +int ceph_trim_caps(struct ceph_inode_info *ci, int drop_inode) +{ + struct ceph_cap *cap; + struct rb_node *p; + int ret = 0; + + if (!spin_trylock(&ci->i_ceph_lock)) + return 0; + + if (ci->i_ceph_flags & CEPH_I_TRIMCAPS) { + for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) { + cap = rb_entry(p, struct ceph_cap, ci_node); + if (atomic_read(&cap->session->s_trim_caps) <= 0) + continue; + if (drop_inode) + atomic_dec(&cap->session->s_trim_caps); + ret = 1; + } + if (!ret) + ci->i_ceph_flags &= ~CEPH_I_TRIMCAPS; + } + + spin_unlock(&ci->i_ceph_lock); + return ret; +} + +/* * called under i_ceph_lock */ static int __ceph_is_any_caps(struct ceph_inode_info *ci) diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c index 0e4da4a..9741f34 100644 --- a/fs/ceph/dir.c +++ b/fs/ceph/dir.c @@ -1043,6 +1043,17 @@ static int ceph_d_revalidate(struct dentry *dentry, unsigned int flags) return valid; } +static int ceph_d_delete(const struct dentry *dentry) +{ + if (dentry->d_inode) { + struct ceph_inode_info *ci = ceph_inode(dentry->d_inode); + /* need trim caps? */ + if (ci->i_ceph_flags & CEPH_I_TRIMCAPS) + return ceph_trim_caps(ci, false); + } + return 0; +} + /* * Release our ceph_dentry_info. */ @@ -1300,6 +1311,7 @@ const struct inode_operations ceph_dir_iops = { const struct dentry_operations ceph_dentry_ops = { .d_revalidate = ceph_d_revalidate, + .d_delete = ceph_d_delete, .d_release = ceph_d_release, .d_prune = ceph_d_prune, }; diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 01e4db1..371590a 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -431,6 +431,17 @@ void ceph_destroy_inode(struct inode *inode) call_rcu(&inode->i_rcu, ceph_i_callback); } +int ceph_drop_inode(struct inode *inode) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + int ret = generic_drop_inode(inode); + + /* need trim caps? */ + if (!ret && (ci->i_ceph_flags & CEPH_I_TRIMCAPS)) + ret = ceph_trim_caps(ci, true); + + return ret ; +} /* * Helpers to fill in size, ctime, mtime, and atime. We have to be diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c index 2a42b5f..7c2b8d6 100644 --- a/fs/ceph/mds_client.c +++ b/fs/ceph/mds_client.c @@ -435,7 +435,7 @@ static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc, s->s_renew_seq = 0; INIT_LIST_HEAD(&s->s_caps); s->s_nr_caps = 0; - s->s_trim_caps = 0; + atomic_set(&s->s_trim_caps, 0); atomic_set(&s->s_ref, 1); INIT_LIST_HEAD(&s->s_waiting); INIT_LIST_HEAD(&s->s_unsafe); @@ -1207,9 +1207,10 @@ static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg) { struct ceph_mds_session *session = arg; struct ceph_inode_info *ci = ceph_inode(inode); + struct dentry *dentry; int used, oissued, mine; - if (session->s_trim_caps <= 0) + if (atomic_read(&session->s_trim_caps) <= 0) return -1; spin_lock(&ci->i_ceph_lock); @@ -1225,16 +1226,39 @@ static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg) if ((used & ~oissued) & mine) goto out; /* we need these caps */ - session->s_trim_caps--; if (oissued) { /* we aren't the only cap.. just remove us */ __queue_cap_release(session, ceph_ino(inode), cap->cap_id, cap->mseq, cap->issue_seq); __ceph_remove_cap(cap); + atomic_dec(&session->s_trim_caps); } else { /* try to drop referring dentries */ + ci->i_ceph_flags |= CEPH_I_TRIMCAPS; spin_unlock(&ci->i_ceph_lock); - d_prune_aliases(inode); + /* + * can't use d_prune_aliases(), because it unhashes dentries + * before calling ceph_d_prune(). + */ +restart: + spin_lock(&inode->i_lock); + hlist_for_each_entry(dentry, &inode->i_dentry, d_alias) { + spin_lock(&dentry->d_lock); + if (!dentry->d_count) { + dget_dlock(dentry); + spin_unlock(&dentry->d_lock); + spin_unlock(&inode->i_lock); + /* + * our d_delete callback returns true when + * CEPH_I_TRIMCAPS is set. This makes VFS + * drop the dentry. + */ + dput(dentry); + goto restart; + } + spin_unlock(&dentry->d_lock); + } + spin_unlock(&inode->i_lock); dout("trim_caps_cb %p cap %p pruned, count now %d\n", inode, cap, atomic_read(&inode->i_count)); return 0; @@ -1257,12 +1281,11 @@ static int trim_caps(struct ceph_mds_client *mdsc, dout("trim_caps mds%d start: %d / %d, trim %d\n", session->s_mds, session->s_nr_caps, max_caps, trim_caps); if (trim_caps > 0) { - session->s_trim_caps = trim_caps; + atomic_set(&session->s_trim_caps, trim_caps); iterate_session_caps(session, trim_caps_cb, session); dout("trim_caps mds%d done: %d / %d, trimmed %d\n", session->s_mds, session->s_nr_caps, max_caps, - trim_caps - session->s_trim_caps); - session->s_trim_caps = 0; + trim_caps - atomic_read(&session->s_trim_caps)); } return 0; } diff --git a/fs/ceph/mds_client.h b/fs/ceph/mds_client.h index c2a19fb..d0e6566 100644 --- a/fs/ceph/mds_client.h +++ b/fs/ceph/mds_client.h @@ -130,8 +130,8 @@ struct ceph_mds_session { /* protected by s_cap_lock */ spinlock_t s_cap_lock; struct list_head s_caps; /* all caps issued by this session */ - int s_nr_caps, s_trim_caps; - int s_num_cap_releases; + int s_nr_caps, s_num_cap_releases; + atomic_t s_trim_caps; struct list_head s_cap_releases; /* waiting cap_release messages */ struct list_head s_cap_releases_done; /* ready to send */ struct ceph_cap *s_cap_iterator; diff --git a/fs/ceph/super.c b/fs/ceph/super.c index 6627b26..bcf26bf 100644 --- a/fs/ceph/super.c +++ b/fs/ceph/super.c @@ -655,6 +655,7 @@ static const struct super_operations ceph_super_ops = { .alloc_inode = ceph_alloc_inode, .destroy_inode = ceph_destroy_inode, .write_inode = ceph_write_inode, + .drop_inode = ceph_drop_inode, .sync_fs = ceph_sync_fs, .put_super = ceph_put_super, .show_options = ceph_show_options, diff --git a/fs/ceph/super.h b/fs/ceph/super.h index e81c0b6..b87691a 100644 --- a/fs/ceph/super.h +++ b/fs/ceph/super.h @@ -421,6 +421,7 @@ static inline struct inode *ceph_find_inode(struct super_block *sb, /* * Ceph inode. */ +#define CEPH_I_TRIMCAPS 1 /* trim caps when possible */ #define CEPH_I_NODELAY 4 /* do not delay cap release */ #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ #define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */ @@ -487,6 +488,7 @@ extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, struct ceph_cap *cap); +extern int ceph_trim_caps(struct ceph_inode_info *ci, int drop_inode); static inline int ceph_caps_issued(struct ceph_inode_info *ci) { @@ -675,6 +677,7 @@ extern const struct inode_operations ceph_file_iops; extern struct inode *ceph_alloc_inode(struct super_block *sb); extern void ceph_destroy_inode(struct inode *inode); +extern int ceph_drop_inode(struct inode *inode); extern struct inode *ceph_get_inode(struct super_block *sb, struct ceph_vino vino);