ocfs2: fix the application IO timeout when fstrim is running

Gang He · torvalds · commit 5500ab4ed3b8 · 2019-03-05T21:07:13.000-08:00
The user reported this problem, the upper application IO was timeout when fstrim was running on this ocfs2 partition. the application monitoring resource agent considered that this application did not work, then this node was fenced by the cluster brain (e.g. pacemaker). The root cause is that fstrim thread always holds main_bm meta-file related locks until all the cluster groups are trimmed. This patch will make fstrim thread release main_bm meta-file related locks when each cluster group is trimmed, this will let the current application IO has a chance to claim the clusters from main_bm meta-file. Link: http://lkml.kernel.org/r/20190111090014.31645-1-ghe@suse.com Signed-off-by: Gang He <ghe@suse.com> Reviewed-by: Changwei Ge <ge.changwei@h3c.com> Cc: Mark Fasheh <mfasheh@versity.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Junxiao Bi <junxiao.bi@oracle.com> Cc: Joseph Qi <joseph.qi@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
@@ -7532,18 +7532,18 @@ static int ocfs2_trim_group(struct super_block *sb,
 	return count;
 }
 
-int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range)
+static
+int ocfs2_trim_mainbm(struct super_block *sb, struct fstrim_range *range)
 {
 	struct ocfs2_super *osb = OCFS2_SB(sb);
-	u64 start, len, trimmed, first_group, last_group, group;
+	u64 start, len, trimmed = 0, first_group, last_group = 0, group = 0;
 	int ret, cnt;
 	u32 first_bit, last_bit, minlen;
 	struct buffer_head *main_bm_bh = NULL;
 	struct inode *main_bm_inode = NULL;
 	struct buffer_head *gd_bh = NULL;
 	struct ocfs2_dinode *main_bm;
 	struct ocfs2_group_desc *gd = NULL;
-	struct ocfs2_trim_fs_info info, *pinfo = NULL;
 
 	start = range->start >> osb->s_clustersize_bits;
 	len = range->len >> osb->s_clustersize_bits;
@@ -7552,6 +7552,9 @@ int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range)
 	if (minlen >= osb->bitmap_cpg || range->len < sb->s_blocksize)
 		return -EINVAL;
 
+	trace_ocfs2_trim_mainbm(start, len, minlen);
+
+next_group:
 	main_bm_inode = ocfs2_get_system_file_inode(osb,
 						    GLOBAL_BITMAP_SYSTEM_INODE,
 						    OCFS2_INVALID_SLOT);
@@ -7570,64 +7573,34 @@ int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range)
 	}
 	main_bm = (struct ocfs2_dinode *)main_bm_bh->b_data;
 
-	if (start >= le32_to_cpu(main_bm->i_clusters)) {
-		ret = -EINVAL;
-		goto out_unlock;
-	}
-
-	len = range->len >> osb->s_clustersize_bits;
-	if (start + len > le32_to_cpu(main_bm->i_clusters))
-		len = le32_to_cpu(main_bm->i_clusters) - start;
-
-	trace_ocfs2_trim_fs(start, len, minlen);
-
-	ocfs2_trim_fs_lock_res_init(osb);
-	ret = ocfs2_trim_fs_lock(osb, NULL, 1);
-	if (ret < 0) {
-		if (ret != -EAGAIN) {
-			mlog_errno(ret);
-			ocfs2_trim_fs_lock_res_uninit(osb);
+	/*
+	 * Do some check before trim the first group.
+	 */
+	if (!group) {
+		if (start >= le32_to_cpu(main_bm->i_clusters)) {
+			ret = -EINVAL;
 			goto out_unlock;
 		}
 
-		mlog(ML_NOTICE, "Wait for trim on device (%s) to "
-		     "finish, which is running from another node.\n",
-		     osb->dev_str);
-		ret = ocfs2_trim_fs_lock(osb, &info, 0);
-		if (ret < 0) {
-			mlog_errno(ret);
-			ocfs2_trim_fs_lock_res_uninit(osb);
-			goto out_unlock;
-		}
+		if (start + len > le32_to_cpu(main_bm->i_clusters))
+			len = le32_to_cpu(main_bm->i_clusters) - start;
 
-		if (info.tf_valid && info.tf_success &&
-		    info.tf_start == start && info.tf_len == len &&
-		    info.tf_minlen == minlen) {
-			/* Avoid sending duplicated trim to a shared device */
-			mlog(ML_NOTICE, "The same trim on device (%s) was "
-			     "just done from node (%u), return.\n",
-			     osb->dev_str, info.tf_nodenum);
-			range->len = info.tf_trimlen;
-			goto out_trimunlock;
-		}
+		/*
+		 * Determine first and last group to examine based on
+		 * start and len
+		 */
+		first_group = ocfs2_which_cluster_group(main_bm_inode, start);
+		if (first_group == osb->first_cluster_group_blkno)
+			first_bit = start;
+		else
+			first_bit = start - ocfs2_blocks_to_clusters(sb,
+								first_group);
+		last_group = ocfs2_which_cluster_group(main_bm_inode,
+						       start + len - 1);
+		group = first_group;
 	}
 
-	info.tf_nodenum = osb->node_num;
-	info.tf_start = start;
-	info.tf_len = len;
-	info.tf_minlen = minlen;
-
-	/* Determine first and last group to examine based on start and len */
-	first_group = ocfs2_which_cluster_group(main_bm_inode, start);
-	if (first_group == osb->first_cluster_group_blkno)
-		first_bit = start;
-	else
-		first_bit = start - ocfs2_blocks_to_clusters(sb, first_group);
-	last_group = ocfs2_which_cluster_group(main_bm_inode, start + len - 1);
-	last_bit = osb->bitmap_cpg;
-
-	trimmed = 0;
-	for (group = first_group; group <= last_group;) {
+	do {
 		if (first_bit + len >= osb->bitmap_cpg)
 			last_bit = osb->bitmap_cpg;
 		else
@@ -7659,21 +7632,81 @@ int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range)
 			group = ocfs2_clusters_to_blocks(sb, osb->bitmap_cpg);
 		else
 			group += ocfs2_clusters_to_blocks(sb, osb->bitmap_cpg);
-	}
-	range->len = trimmed * sb->s_blocksize;
+	} while (0);
 
-	info.tf_trimlen = range->len;
-	info.tf_success = (ret ? 0 : 1);
-	pinfo = &info;
-out_trimunlock:
-	ocfs2_trim_fs_unlock(osb, pinfo);
-	ocfs2_trim_fs_lock_res_uninit(osb);
 out_unlock:
 	ocfs2_inode_unlock(main_bm_inode, 0);
 	brelse(main_bm_bh);
+	main_bm_bh = NULL;
 out_mutex:
 	inode_unlock(main_bm_inode);
 	iput(main_bm_inode);
+
+	/*
+	 * If all the groups trim are not done or failed, but we should release
+	 * main_bm related locks for avoiding the current IO starve, then go to
+	 * trim the next group
+	 */
+	if (ret >= 0 && group <= last_group)
+		goto next_group;
 out:
+	range->len = trimmed * sb->s_blocksize;
+	return ret;
+}
+
+int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range)
+{
+	int ret;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+	struct ocfs2_trim_fs_info info, *pinfo = NULL;
+
+	ocfs2_trim_fs_lock_res_init(osb);
+
+	trace_ocfs2_trim_fs(range->start, range->len, range->minlen);
+
+	ret = ocfs2_trim_fs_lock(osb, NULL, 1);
+	if (ret < 0) {
+		if (ret != -EAGAIN) {
+			mlog_errno(ret);
+			ocfs2_trim_fs_lock_res_uninit(osb);
+			return ret;
+		}
+
+		mlog(ML_NOTICE, "Wait for trim on device (%s) to "
+		     "finish, which is running from another node.\n",
+		     osb->dev_str);
+		ret = ocfs2_trim_fs_lock(osb, &info, 0);
+		if (ret < 0) {
+			mlog_errno(ret);
+			ocfs2_trim_fs_lock_res_uninit(osb);
+			return ret;
+		}
+
+		if (info.tf_valid && info.tf_success &&
+		    info.tf_start == range->start &&
+		    info.tf_len == range->len &&
+		    info.tf_minlen == range->minlen) {
+			/* Avoid sending duplicated trim to a shared device */
+			mlog(ML_NOTICE, "The same trim on device (%s) was "
+			     "just done from node (%u), return.\n",
+			     osb->dev_str, info.tf_nodenum);
+			range->len = info.tf_trimlen;
+			goto out;
+		}
+	}
+
+	info.tf_nodenum = osb->node_num;
+	info.tf_start = range->start;
+	info.tf_len = range->len;
+	info.tf_minlen = range->minlen;
+
+	ret = ocfs2_trim_mainbm(sb, range);
+
+	info.tf_trimlen = range->len;
+	info.tf_success = (ret < 0 ? 0 : 1);
+	pinfo = &info;
+out:
+	ocfs2_trim_fs_unlock(osb, pinfo);
+	ocfs2_trim_fs_lock_res_uninit(osb);
 	return ret;
 }
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
@@ -686,6 +686,9 @@ void ocfs2_trim_fs_lock_res_init(struct ocfs2_super *osb)
 {
 	struct ocfs2_lock_res *lockres = &osb->osb_trim_fs_lockres;
 
+	/* Only one trimfs thread are allowed to work at the same time. */
+	mutex_lock(&osb->obs_trim_fs_mutex);
+
 	ocfs2_lock_res_init_once(lockres);
 	ocfs2_build_lock_name(OCFS2_LOCK_TYPE_TRIM_FS, 0, 0, lockres->l_name);
 	ocfs2_lock_res_init_common(osb, lockres, OCFS2_LOCK_TYPE_TRIM_FS,
@@ -698,6 +701,8 @@ void ocfs2_trim_fs_lock_res_uninit(struct ocfs2_super *osb)
 
 	ocfs2_simple_drop_lockres(osb, lockres);
 	ocfs2_lock_res_free(lockres);
+
+	mutex_unlock(&osb->obs_trim_fs_mutex);
 }
 
 static void ocfs2_orphan_scan_lock_res_init(struct ocfs2_lock_res *res,
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
@@ -407,6 +407,7 @@ struct ocfs2_super
 	struct ocfs2_lock_res osb_rename_lockres;
 	struct ocfs2_lock_res osb_nfs_sync_lockres;
 	struct ocfs2_lock_res osb_trim_fs_lockres;
+	struct mutex obs_trim_fs_mutex;
 	struct ocfs2_dlm_debug *osb_dlm_debug;
 
 	struct dentry *osb_debug_root;
diff --git a/fs/ocfs2/ocfs2_trace.h b/fs/ocfs2/ocfs2_trace.h
@@ -712,6 +712,8 @@ TRACE_EVENT(ocfs2_trim_extent,
 
 DEFINE_OCFS2_ULL_UINT_UINT_UINT_EVENT(ocfs2_trim_group);
 
+DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_trim_mainbm);
+
 DEFINE_OCFS2_ULL_ULL_ULL_EVENT(ocfs2_trim_fs);
 
 /* End of trace events for fs/ocfs2/alloc.c. */
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
@@ -1847,6 +1847,8 @@ static int ocfs2_mount_volume(struct super_block *sb)
 	if (ocfs2_is_hard_readonly(osb))
 		goto leave;
 
+	mutex_init(&osb->obs_trim_fs_mutex);
+
 	status = ocfs2_dlm_init(osb);
 	if (status < 0) {
 		mlog_errno(status);
