btrfs: defer adding raid type kobject until after chunk relocation

Any time the first block group of a new type is created, we add a new
kobject to sysfs to hold the attributes for that type.  Kobject-internal
allocations always use GFP_KERNEL, making them prone to fs-reclaim races.
While it appears as if this can occur any time a block group is created,
the only times the first block group of a new type can be created in
memory is at mount and when we create the first new block group during
raid conversion.

This patch adds a new list to track pending kobject additions and then
handles them after we do chunk relocation.  Between relocating the
target chunk (or forcing allocation of a new chunk in the case of data)
and removing the old chunk, we're in a safe place for fs-reclaim to
occur.  We're holding the volume mutex, which is already held across
page faults, and the delete_unused_bgs_mutex, which will only stall
the cleaner thread.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

5 files changed, 62 insertions, 20 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index ffa72ca5755d..8d3aa56b928b 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -385,8 +385,9 @@ struct btrfs_dev_replace {
 
 /* For raid type sysfs entries */
 struct raid_kobject {
-	int raid_type;
+	u64 flags;
 	struct kobject kobj;
+	struct list_head list;
 };
 
 struct btrfs_space_info {
@@ -940,6 +941,8 @@ struct btrfs_fs_info {
 	u32 thread_pool_size;
 
 	struct kobject *space_info_kobj;
+	struct list_head pending_raid_kobjs;
+	spinlock_t pending_raid_kobjs_lock; /* uncontended */
 
 	u64 total_pinned;
 
@@ -2700,6 +2703,7 @@ int btrfs_can_relocate(struct btrfs_fs_info *fs_info, u64 bytenr);
 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 			   struct btrfs_fs_info *fs_info, u64 bytes_used,
 			   u64 type, u64 chunk_offset, u64 size);
+void btrfs_add_raid_kobjects(struct btrfs_fs_info *fs_info);
 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
 				struct btrfs_fs_info *fs_info,
 				const u64 chunk_offset);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 1657d6aa4fa6..38b387ae78f8 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -2431,6 +2431,8 @@ int open_ctree(struct super_block *sb,
 	INIT_LIST_HEAD(&fs_info->delayed_iputs);
 	INIT_LIST_HEAD(&fs_info->delalloc_roots);
 	INIT_LIST_HEAD(&fs_info->caching_block_groups);
+	INIT_LIST_HEAD(&fs_info->pending_raid_kobjs);
+	spin_lock_init(&fs_info->pending_raid_kobjs_lock);
 	spin_lock_init(&fs_info->delalloc_root_lock);
 	spin_lock_init(&fs_info->trans_lock);
 	spin_lock_init(&fs_info->fs_roots_radix_lock);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index e10f74dac493..0b1f01dd02de 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -9918,9 +9918,39 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
 	return 0;
 }
 
+/* link_block_group will queue up kobjects to add when we're reclaim-safe */
+void btrfs_add_raid_kobjects(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_space_info *space_info;
+	struct raid_kobject *rkobj;
+	LIST_HEAD(list);
+	int index;
+	int ret = 0;
+
+	spin_lock(&fs_info->pending_raid_kobjs_lock);
+	list_splice_init(&fs_info->pending_raid_kobjs, &list);
+	spin_unlock(&fs_info->pending_raid_kobjs_lock);
+
+	list_for_each_entry(rkobj, &list, list) {
+		space_info = __find_space_info(fs_info, rkobj->flags);
+		index = btrfs_bg_flags_to_raid_index(rkobj->flags);
+
+		ret = kobject_add(&rkobj->kobj, &space_info->kobj,
+				  "%s", get_raid_name(index));
+		if (ret) {
+			kobject_put(&rkobj->kobj);
+			break;
+		}
+	}
+	if (ret)
+		btrfs_warn(fs_info,
+			   "failed to add kobject for block cache, ignoring");
+}
+
 static void link_block_group(struct btrfs_block_group_cache *cache)
 {
 	struct btrfs_space_info *space_info = cache->space_info;
+	struct btrfs_fs_info *fs_info = cache->fs_info;
 	int index = btrfs_bg_flags_to_raid_index(cache->flags);
 	bool first = false;
 
@@ -9931,27 +9961,20 @@ static void link_block_group(struct btrfs_block_group_cache *cache)
 	up_write(&space_info->groups_sem);
 
 	if (first) {
-		struct raid_kobject *rkobj;
-		int ret;
-
-		rkobj = kzalloc(sizeof(*rkobj), GFP_NOFS);
-		if (!rkobj)
-			goto out_err;
-		rkobj->raid_type = index;
-		kobject_init(&rkobj->kobj, &btrfs_raid_ktype);
-		ret = kobject_add(&rkobj->kobj, &space_info->kobj,
-				  "%s", get_raid_name(index));
-		if (ret) {
-			kobject_put(&rkobj->kobj);
-			goto out_err;
+		struct raid_kobject *rkobj = kzalloc(sizeof(*rkobj), GFP_NOFS);
+		if (!rkobj) {
+			btrfs_warn(cache->fs_info,
+				"couldn't alloc memory for raid level kobject");
+			return;
 		}
+		rkobj->flags = cache->flags;
+		kobject_init(&rkobj->kobj, &btrfs_raid_ktype);
+
+		spin_lock(&fs_info->pending_raid_kobjs_lock);
+		list_add_tail(&rkobj->list, &fs_info->pending_raid_kobjs);
+		spin_unlock(&fs_info->pending_raid_kobjs_lock);
 		space_info->block_group_kobjs[index] = &rkobj->kobj;
 	}
-
-	return;
-out_err:
-	btrfs_warn(cache->fs_info,
-		   "failed to add kobject for block cache, ignoring");
 }
 
 static struct btrfs_block_group_cache *
@@ -10167,6 +10190,7 @@ int btrfs_read_block_groups(struct btrfs_fs_info *info)
 			inc_block_group_ro(cache, 1);
 	}
 
+	btrfs_add_raid_kobjects(info);
 	init_global_block_rsv(info);
 	ret = 0;
 error:
diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index 6af7b58e1a90..ca067471cd46 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -272,7 +272,7 @@ static ssize_t raid_bytes_show(struct kobject *kobj,
 {
 	struct btrfs_space_info *sinfo = to_space_info(kobj->parent);
 	struct btrfs_block_group_cache *block_group;
-	int index = to_raid_kobj(kobj)->raid_type;
+	int index = btrfs_bg_flags_to_raid_index(to_raid_kobj(kobj)->flags);
 	u64 val = 0;
 
 	down_read(&sinfo->groups_sem);
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 73de042158f1..4fc6acf65220 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -3003,6 +3003,16 @@ static int btrfs_relocate_chunk(struct btrfs_fs_info *fs_info, u64 chunk_offset)
 	if (ret)
 		return ret;
 
+	/*
+	 * We add the kobjects here (and after forcing data chunk creation)
+	 * since relocation is the only place we'll create chunks of a new
+	 * type at runtime.  The only place where we'll remove the last
+	 * chunk of a type is the call immediately below this one.  Even
+	 * so, we're protected against races with the cleaner thread since
+	 * we're covered by the delete_unused_bgs_mutex.
+	 */
+	btrfs_add_raid_kobjects(fs_info);
+
 	trans = btrfs_start_trans_remove_block_group(root->fs_info,
 						     chunk_offset);
 	if (IS_ERR(trans)) {
@@ -3130,6 +3140,8 @@ static int btrfs_may_alloc_data_chunk(struct btrfs_fs_info *fs_info,
 			if (ret < 0)
 				return ret;
 
+			btrfs_add_raid_kobjects(fs_info);
+
 			return 1;
 		}
 	}