Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755622AbaF3L45 (ORCPT ); Mon, 30 Jun 2014 07:56:57 -0400 Received: from ip4-83-240-18-248.cust.nbox.cz ([83.240.18.248]:52130 "EHLO ip4-83-240-18-248.cust.nbox.cz" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754035AbaF3Lx0 (ORCPT ); Mon, 30 Jun 2014 07:53:26 -0400 From: Jiri Slaby To: stable@vger.kernel.org Cc: linux-kernel@vger.kernel.org, Jan Kara , Joel Becker , Andrew Morton , Linus Torvalds , Jiri Slaby Subject: [PATCH 3.12 033/181] ocfs2: implement delayed dropping of last dquot reference Date: Mon, 30 Jun 2014 13:50:54 +0200 Message-Id: X-Mailer: git-send-email 2.0.0 In-Reply-To: <61844d8e25eb8899b0836afa9796fa239db80f1f.1404128997.git.jslaby@suse.cz> References: <61844d8e25eb8899b0836afa9796fa239db80f1f.1404128997.git.jslaby@suse.cz> In-Reply-To: References: Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Jan Kara 3.12-stable review patch. If anyone has any objections, please let me know. =============== commit e3a767b60fd8a9f5e133f42f4970cff77ec43173 upstream. We cannot drop last dquot reference from downconvert thread as that creates the following deadlock: NODE 1 NODE2 holds dentry lock for 'foo' holds inode lock for GLOBAL_BITMAP_SYSTEM_INODE dquot_initialize(bar) ocfs2_dquot_acquire() ocfs2_inode_lock(USER_QUOTA_SYSTEM_INODE) ... downconvert thread (triggered from another node or a different process from NODE2) ocfs2_dentry_post_unlock() ... iput(foo) ocfs2_evict_inode(foo) ocfs2_clear_inode(foo) dquot_drop(inode) ... ocfs2_dquot_release() ocfs2_inode_lock(USER_QUOTA_SYSTEM_INODE) - blocks finds we need more space in quota file ... ocfs2_extend_no_holes() ocfs2_inode_lock(GLOBAL_BITMAP_SYSTEM_INODE) - deadlocks waiting for downconvert thread We solve the problem by postponing dropping of the last dquot reference to a workqueue if it happens from the downconvert thread. Signed-off-by: Jan Kara Reviewed-by: Mark Fasheh Reviewed-by: Srinivas Eeda Cc: Joel Becker Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds Signed-off-by: Jiri Slaby --- fs/ocfs2/ocfs2.h | 5 +++++ fs/ocfs2/quota.h | 2 ++ fs/ocfs2/quota_global.c | 35 +++++++++++++++++++++++++++++++++++ fs/ocfs2/super.c | 8 ++++++++ 4 files changed, 50 insertions(+) diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h index 3a903470c794..1f023ae39379 100644 --- a/fs/ocfs2/ocfs2.h +++ b/fs/ocfs2/ocfs2.h @@ -30,6 +30,7 @@ #include #include #include +#include #include #include #include @@ -418,6 +419,10 @@ struct ocfs2_super struct ocfs2_dentry_lock *dentry_lock_list; struct work_struct dentry_lock_work; + /* List of dquot structures to drop last reference to */ + struct llist_head dquot_drop_list; + struct work_struct dquot_drop_work; + wait_queue_head_t osb_mount_event; /* Truncate log info */ diff --git a/fs/ocfs2/quota.h b/fs/ocfs2/quota.h index d5ab56cbe5c5..f266d67df3c6 100644 --- a/fs/ocfs2/quota.h +++ b/fs/ocfs2/quota.h @@ -28,6 +28,7 @@ struct ocfs2_dquot { unsigned int dq_use_count; /* Number of nodes having reference to this entry in global quota file */ s64 dq_origspace; /* Last globally synced space usage */ s64 dq_originodes; /* Last globally synced inode usage */ + struct llist_node list; /* Member of list of dquots to drop */ }; /* Description of one chunk to recover in memory */ @@ -110,6 +111,7 @@ int ocfs2_read_quota_phys_block(struct inode *inode, u64 p_block, int ocfs2_create_local_dquot(struct dquot *dquot); int ocfs2_local_release_dquot(handle_t *handle, struct dquot *dquot); int ocfs2_local_write_dquot(struct dquot *dquot); +void ocfs2_drop_dquot_refs(struct work_struct *work); extern const struct dquot_operations ocfs2_quota_operations; extern struct quota_format_type ocfs2_quota_format; diff --git a/fs/ocfs2/quota_global.c b/fs/ocfs2/quota_global.c index d7b5108789e2..b990a62cff50 100644 --- a/fs/ocfs2/quota_global.c +++ b/fs/ocfs2/quota_global.c @@ -10,6 +10,7 @@ #include #include #include +#include #include @@ -679,6 +680,27 @@ static int ocfs2_calc_qdel_credits(struct super_block *sb, int type) OCFS2_INODE_UPDATE_CREDITS; } +void ocfs2_drop_dquot_refs(struct work_struct *work) +{ + struct ocfs2_super *osb = container_of(work, struct ocfs2_super, + dquot_drop_work); + struct llist_node *list; + struct ocfs2_dquot *odquot, *next_odquot; + + list = llist_del_all(&osb->dquot_drop_list); + llist_for_each_entry_safe(odquot, next_odquot, list, list) { + /* Drop the reference we acquired in ocfs2_dquot_release() */ + dqput(&odquot->dq_dquot); + } +} + +/* + * Called when the last reference to dquot is dropped. If we are called from + * downconvert thread, we cannot do all the handling here because grabbing + * quota lock could deadlock (the node holding the quota lock could need some + * other cluster lock to proceed but with blocked downconvert thread we cannot + * release any lock). + */ static int ocfs2_release_dquot(struct dquot *dquot) { handle_t *handle; @@ -694,6 +716,19 @@ static int ocfs2_release_dquot(struct dquot *dquot) /* Check whether we are not racing with some other dqget() */ if (atomic_read(&dquot->dq_count) > 1) goto out; + /* Running from downconvert thread? Postpone quota processing to wq */ + if (current == osb->dc_task) { + /* + * Grab our own reference to dquot and queue it for delayed + * dropping. Quota code rechecks after calling + * ->release_dquot() and won't free dquot structure. + */ + dqgrab(dquot); + /* First entry on list -> queue work */ + if (llist_add(&OCFS2_DQUOT(dquot)->list, &osb->dquot_drop_list)) + queue_work(ocfs2_wq, &osb->dquot_drop_work); + goto out; + } status = ocfs2_lock_global_qf(oinfo, 1); if (status < 0) goto out; diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index d4e81e4a9b04..d6adb5b2b3a5 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -1945,6 +1945,11 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) ocfs2_disable_quotas(osb); + /* All dquots should be freed by now */ + WARN_ON(!llist_empty(&osb->dquot_drop_list)); + /* Wait for worker to be done with the work structure in osb */ + cancel_work_sync(&osb->dquot_drop_work); + ocfs2_shutdown_local_alloc(osb); ocfs2_truncate_log_shutdown(osb); @@ -2275,6 +2280,9 @@ static int ocfs2_initialize_super(struct super_block *sb, INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes); osb->dentry_lock_list = NULL; + INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs); + init_llist_head(&osb->dquot_drop_list); + /* get some pseudo constants for clustersize bits */ osb->s_clustersize_bits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); -- 2.0.0 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/