Merge tag 'for-5.1-part1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs updates from David Sterba:
 "This contains usual mix of new features, core changes and fixes; full
  list below. I'm planning second pull request, with a few more fixes
  that arrived recently but too close to merge window, will send it next
  week.

  New features:

   - support zstd compression levels

   - new ioctl to unregister a device from the module (ie. reverse of
     device scan)

   - scrub prints a message to log when it's about to start or finish

  Core changes:

   - qgroups can now skip part of a tree that does not get updated
     during relocation, because this does not affect the quota
     accounting, estimated speedup in run time is about 20%

   - the compression workspace management had to be enhanced due to zstd
     requirements

   - various enospc fixes, when there's high fragmentation the
     over-reservation can cause ENOSPC that might not happen after a
     flush, in such cases try to wait if the situation improves

  Fixes:

   - various ioctls could overwrite previous return value if
     copy_to_user fails, fix this so the original error is reported

   - more reclaim vs GFP_KERNEL fixes

   - other cleanups and refactoring

   - fix a (valid) lockdep warning in a test when device replace is
     destroying worker threads

   - make qgroup async transaction commit more aggressive, this avoids
     some 'quota limit reached' errors if there are not enough data to
     trigger transaction in order to flush

   - fix deadlock between snapshot deletion and quotas when backref
     walking is called from context that already holds the same locks

   - fsync fixes:
      - fix fsync after succession of renames of different files
      - fix fsync after succession of renames and unlink/rmdir"

* tag 'for-5.1-part1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux: (92 commits)
  btrfs: Remove unnecessary casts in btrfs_read_root_item
  Btrfs: remove assertion when searching for a key in a node/leaf
  Btrfs: add missing error handling after doing leaf/node binary search
  btrfs: drop the lock on error in btrfs_dev_replace_cancel
  btrfs: ensure that a DUP or RAID1 block group has exactly two stripes
  btrfs: init csum_list before possible free
  Btrfs: remove no longer needed range length checks for deduplication
  Btrfs: fix fsync after succession of renames and unlink/rmdir
  Btrfs: fix fsync after succession of renames of different files
  btrfs: honor path->skip_locking in backref code
  btrfs: qgroup: Make qgroup async transaction commit more aggressive
  btrfs: qgroup: Move reserved data accounting from btrfs_delayed_ref_head to btrfs_qgroup_extent_record
  btrfs: scrub: remove unused nocow worker pointer
  btrfs: scrub: add assertions for worker pointers
  btrfs: scrub: convert scrub_workers_refcnt to refcount_t
  btrfs: scrub: add scrub_lock lockdep check in scrub_workers_get
  btrfs: scrub: fix circular locking dependency warning
  btrfs: fix comment its device list mutex not volume lock
  btrfs: extent_io: Kill the forward declaration of flush_write_bio
  btrfs: Fix grossly misleading argument names in extent io search
  ...

38 files changed, 1993 insertions, 960 deletions

diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c
index 3b66c957ea6f..5810463dc6d2 100644
--- a/fs/btrfs/acl.c
+++ b/fs/btrfs/acl.c
@@ -9,6 +9,7 @@
 #include <linux/posix_acl_xattr.h>
 #include <linux/posix_acl.h>
 #include <linux/sched.h>
+#include <linux/sched/mm.h>
 #include <linux/slab.h>
 
 #include "ctree.h"
@@ -72,8 +73,16 @@ static int __btrfs_set_acl(struct btrfs_trans_handle *trans,
 	}
 
 	if (acl) {
+		unsigned int nofs_flag;
+
 		size = posix_acl_xattr_size(acl->a_count);
+		/*
+		 * We're holding a transaction handle, so use a NOFS memory
+		 * allocation context to avoid deadlock if reclaim happens.
+		 */
+		nofs_flag = memalloc_nofs_save();
 		value = kmalloc(size, GFP_KERNEL);
+		memalloc_nofs_restore(nofs_flag);
 		if (!value) {
 			ret = -ENOMEM;
 			goto out;
diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c
index d522494698fa..122cb97c7909 100644
--- a/fs/btrfs/async-thread.c
+++ b/fs/btrfs/async-thread.c
@@ -139,13 +139,11 @@ __btrfs_alloc_workqueue(struct btrfs_fs_info *fs_info, const char *name,
 	}
 
 	if (flags & WQ_HIGHPRI)
-		ret->normal_wq = alloc_workqueue("%s-%s-high", flags,
-						 ret->current_active, "btrfs",
-						 name);
+		ret->normal_wq = alloc_workqueue("btrfs-%s-high", flags,
+						 ret->current_active, name);
 	else
-		ret->normal_wq = alloc_workqueue("%s-%s", flags,
-						 ret->current_active, "btrfs",
-						 name);
+		ret->normal_wq = alloc_workqueue("btrfs-%s", flags,
+						 ret->current_active, name);
 	if (!ret->normal_wq) {
 		kfree(ret);
 		return NULL;
diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index 78556447e1d5..11459fe84a29 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -712,7 +712,7 @@ out:
  * read tree blocks and add keys where required.
  */
 static int add_missing_keys(struct btrfs_fs_info *fs_info,
-			    struct preftrees *preftrees)
+			    struct preftrees *preftrees, bool lock)
 {
 	struct prelim_ref *ref;
 	struct extent_buffer *eb;
@@ -737,12 +737,14 @@ static int add_missing_keys(struct btrfs_fs_info *fs_info,
 			free_extent_buffer(eb);
 			return -EIO;
 		}
-		btrfs_tree_read_lock(eb);
+		if (lock)
+			btrfs_tree_read_lock(eb);
 		if (btrfs_header_level(eb) == 0)
 			btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
 		else
 			btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
-		btrfs_tree_read_unlock(eb);
+		if (lock)
+			btrfs_tree_read_unlock(eb);
 		free_extent_buffer(eb);
 		prelim_ref_insert(fs_info, &preftrees->indirect, ref, NULL);
 		cond_resched();
@@ -1227,7 +1229,7 @@ again:
 
 	btrfs_release_path(path);
 
-	ret = add_missing_keys(fs_info, &preftrees);
+	ret = add_missing_keys(fs_info, &preftrees, path->skip_locking == 0);
 	if (ret)
 		goto out;
 
@@ -1288,11 +1290,15 @@ again:
 					ret = -EIO;
 					goto out;
 				}
-				btrfs_tree_read_lock(eb);
-				btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+
+				if (!path->skip_locking) {
+					btrfs_tree_read_lock(eb);
+					btrfs_set_lock_blocking_read(eb);
+				}
 				ret = find_extent_in_eb(eb, bytenr,
 							*extent_item_pos, &eie, ignore_offset);
-				btrfs_tree_read_unlock_blocking(eb);
+				if (!path->skip_locking)
+					btrfs_tree_read_unlock_blocking(eb);
 				free_extent_buffer(eb);
 				if (ret < 0)
 					goto out;
@@ -1650,7 +1656,7 @@ char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
 		/* make sure we can use eb after releasing the path */
 		if (eb != eb_in) {
 			if (!path->skip_locking)
-				btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+				btrfs_set_lock_blocking_read(eb);
 			path->nodes[0] = NULL;
 			path->locks[0] = 0;
 		}
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index 548057630b69..eb8e20b740d6 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -730,6 +730,28 @@ struct heuristic_ws {
 	struct list_head list;
 };
 
+static struct workspace_manager heuristic_wsm;
+
+static void heuristic_init_workspace_manager(void)
+{
+	btrfs_init_workspace_manager(&heuristic_wsm, &btrfs_heuristic_compress);
+}
+
+static void heuristic_cleanup_workspace_manager(void)
+{
+	btrfs_cleanup_workspace_manager(&heuristic_wsm);
+}
+
+static struct list_head *heuristic_get_workspace(unsigned int level)
+{
+	return btrfs_get_workspace(&heuristic_wsm, level);
+}
+
+static void heuristic_put_workspace(struct list_head *ws)
+{
+	btrfs_put_workspace(&heuristic_wsm, ws);
+}
+
 static void free_heuristic_ws(struct list_head *ws)
 {
 	struct heuristic_ws *workspace;
@@ -742,7 +764,7 @@ static void free_heuristic_ws(struct list_head *ws)
 	kfree(workspace);
 }
 
-static struct list_head *alloc_heuristic_ws(void)
+static struct list_head *alloc_heuristic_ws(unsigned int level)
 {
 	struct heuristic_ws *ws;
 
@@ -769,65 +791,59 @@ fail:
 	return ERR_PTR(-ENOMEM);
 }
 
-struct workspaces_list {
-	struct list_head idle_ws;
-	spinlock_t ws_lock;
-	/* Number of free workspaces */
-	int free_ws;
-	/* Total number of allocated workspaces */
-	atomic_t total_ws;
-	/* Waiters for a free workspace */
-	wait_queue_head_t ws_wait;
+const struct btrfs_compress_op btrfs_heuristic_compress = {
+	.init_workspace_manager = heuristic_init_workspace_manager,
+	.cleanup_workspace_manager = heuristic_cleanup_workspace_manager,
+	.get_workspace = heuristic_get_workspace,
+	.put_workspace = heuristic_put_workspace,
+	.alloc_workspace = alloc_heuristic_ws,
+	.free_workspace = free_heuristic_ws,
 };
 
-static struct workspaces_list btrfs_comp_ws[BTRFS_COMPRESS_TYPES];
-
-static struct workspaces_list btrfs_heuristic_ws;
-
 static const struct btrfs_compress_op * const btrfs_compress_op[] = {
+	/* The heuristic is represented as compression type 0 */
+	&btrfs_heuristic_compress,
 	&btrfs_zlib_compress,
 	&btrfs_lzo_compress,
 	&btrfs_zstd_compress,
 };
 
-void __init btrfs_init_compress(void)
+void btrfs_init_workspace_manager(struct workspace_manager *wsm,
+				  const struct btrfs_compress_op *ops)
 {
 	struct list_head *workspace;
-	int i;
 
-	INIT_LIST_HEAD(&btrfs_heuristic_ws.idle_ws);
-	spin_lock_init(&btrfs_heuristic_ws.ws_lock);
-	atomic_set(&btrfs_heuristic_ws.total_ws, 0);
-	init_waitqueue_head(&btrfs_heuristic_ws.ws_wait);
+	wsm->ops = ops;
 
-	workspace = alloc_heuristic_ws();
+	INIT_LIST_HEAD(&wsm->idle_ws);
+	spin_lock_init(&wsm->ws_lock);
+	atomic_set(&wsm->total_ws, 0);
+	init_waitqueue_head(&wsm->ws_wait);
+
+	/*
+	 * Preallocate one workspace for each compression type so we can
+	 * guarantee forward progress in the worst case
+	 */
+	workspace = wsm->ops->alloc_workspace(0);
 	if (IS_ERR(workspace)) {
 		pr_warn(
-	"BTRFS: cannot preallocate heuristic workspace, will try later\n");
+	"BTRFS: cannot preallocate compression workspace, will try later\n");
 	} else {
-		atomic_set(&btrfs_heuristic_ws.total_ws, 1);
-		btrfs_heuristic_ws.free_ws = 1;
-		list_add(workspace, &btrfs_heuristic_ws.idle_ws);
+		atomic_set(&wsm->total_ws, 1);
+		wsm->free_ws = 1;
+		list_add(workspace, &wsm->idle_ws);
 	}
+}
 
-	for (i = 0; i < BTRFS_COMPRESS_TYPES; i++) {
-		INIT_LIST_HEAD(&btrfs_comp_ws[i].idle_ws);
-		spin_lock_init(&btrfs_comp_ws[i].ws_lock);
-		atomic_set(&btrfs_comp_ws[i].total_ws, 0);
-		init_waitqueue_head(&btrfs_comp_ws[i].ws_wait);
+void btrfs_cleanup_workspace_manager(struct workspace_manager *wsman)
+{
+	struct list_head *ws;
 
-		/*
-		 * Preallocate one workspace for each compression type so
-		 * we can guarantee forward progress in the worst case
-		 */
-		workspace = btrfs_compress_op[i]->alloc_workspace();
-		if (IS_ERR(workspace)) {
-			pr_warn("BTRFS: cannot preallocate compression workspace, will try later\n");
-		} else {
-			atomic_set(&btrfs_comp_ws[i].total_ws, 1);
-			btrfs_comp_ws[i].free_ws = 1;
-			list_add(workspace, &btrfs_comp_ws[i].idle_ws);
-		}
+	while (!list_empty(&wsman->idle_ws)) {
+		ws = wsman->idle_ws.next;
+		list_del(ws);
+		wsman->ops->free_workspace(ws);
+		atomic_dec(&wsman->total_ws);
 	}
 }
 
@@ -837,11 +853,11 @@ void __init btrfs_init_compress(void)
  * Preallocation makes a forward progress guarantees and we do not return
  * errors.
  */
-static struct list_head *__find_workspace(int type, bool heuristic)
+struct list_head *btrfs_get_workspace(struct workspace_manager *wsm,
+				      unsigned int level)
 {
 	struct list_head *workspace;
 	int cpus = num_online_cpus();
-	int idx = type - 1;
 	unsigned nofs_flag;
 	struct list_head *idle_ws;
 	spinlock_t *ws_lock;
@@ -849,19 +865,11 @@ static struct list_head *__find_workspace(int type, bool heuristic)
 	wait_queue_head_t *ws_wait;
 	int *free_ws;
 
-	if (heuristic) {
-		idle_ws	 = &btrfs_heuristic_ws.idle_ws;
-		ws_lock	 = &btrfs_heuristic_ws.ws_lock;
-		total_ws = &btrfs_heuristic_ws.total_ws;
-		ws_wait	 = &btrfs_heuristic_ws.ws_wait;
-		free_ws	 = &btrfs_heuristic_ws.free_ws;
-	} else {
-		idle_ws	 = &btrfs_comp_ws[idx].idle_ws;
-		ws_lock	 = &btrfs_comp_ws[idx].ws_lock;
-		total_ws = &btrfs_comp_ws[idx].total_ws;
-		ws_wait	 = &btrfs_comp_ws[idx].ws_wait;
-		free_ws	 = &btrfs_comp_ws[idx].free_ws;
-	}
+	idle_ws	 = &wsm->idle_ws;
+	ws_lock	 = &wsm->ws_lock;
+	total_ws = &wsm->total_ws;
+	ws_wait	 = &wsm->ws_wait;
+	free_ws	 = &wsm->free_ws;
 
 again:
 	spin_lock(ws_lock);
@@ -892,10 +900,7 @@ again:
 	 * context of btrfs_compress_bio/btrfs_compress_pages
 	 */
 	nofs_flag = memalloc_nofs_save();
-	if (heuristic)
-		workspace = alloc_heuristic_ws();
-	else
-		workspace = btrfs_compress_op[idx]->alloc_workspace();
+	workspace = wsm->ops->alloc_workspace(level);
 	memalloc_nofs_restore(nofs_flag);
 
 	if (IS_ERR(workspace)) {
@@ -926,85 +931,47 @@ again:
 	return workspace;
 }
 
-static struct list_head *find_workspace(int type)
+static struct list_head *get_workspace(int type, int level)
 {
-	return __find_workspace(type, false);
+	return btrfs_compress_op[type]->get_workspace(level);
 }
 
 /*
  * put a workspace struct back on the list or free it if we have enough
  * idle ones sitting around
  */
-static void __free_workspace(int type, struct list_head *workspace,
-			     bool heuristic)
+void btrfs_put_workspace(struct workspace_manager *wsm, struct list_head *ws)
 {
-	int idx = type - 1;
 	struct list_head *idle_ws;
 	spinlock_t *ws_lock;
 	atomic_t *total_ws;
 	wait_queue_head_t *ws_wait;
 	int *free_ws;
 
-	if (heuristic) {
-		idle_ws	 = &btrfs_heuristic_ws.idle_ws;
-		ws_lock	 = &btrfs_heuristic_ws.ws_lock;
-		total_ws = &btrfs_heuristic_ws.total_ws;
-		ws_wait	 = &btrfs_heuristic_ws.ws_wait;
-		free_ws	 = &btrfs_heuristic_ws.free_ws;
-	} else {
-		idle_ws	 = &btrfs_comp_ws[idx].idle_ws;
-		ws_lock	 = &btrfs_comp_ws[idx].ws_lock;
-		total_ws = &btrfs_comp_ws[idx].total_ws;
-		ws_wait	 = &btrfs_comp_ws[idx].ws_wait;
-		free_ws	 = &btrfs_comp_ws[idx].free_ws;
-	}
+	idle_ws	 = &wsm->idle_ws;
+	ws_lock	 = &wsm->ws_lock;
+	total_ws = &wsm->total_ws;
+	ws_wait	 = &wsm->ws_wait;
+	free_ws	 = &wsm->free_ws;
 
 	spin_lock(ws_lock);
 	if (*free_ws <= num_online_cpus()) {
-		list_add(workspace, idle_ws);
+		list_add(ws, idle_ws);
 		(*free_ws)++;
 		spin_unlock(ws_lock);
 		goto wake;
 	}
 	spin_unlock(ws_lock);
 
-	if (heuristic)
-		free_heuristic_ws(workspace);
-	else
-		btrfs_compress_op[idx]->free_workspace(workspace);
+	wsm->ops->free_workspace(ws);
 	atomic_dec(total_ws);
 wake:
 	cond_wake_up(ws_wait);
 }
 
-static void free_workspace(int type, struct list_head *ws)
+static void put_workspace(int type, struct list_head *ws)
 {
-	return __free_workspace(type, ws, false);
-}
-
-/*
- * cleanup function for module exit
- */
-static void free_workspaces(void)
-{
-	struct list_head *workspace;
-	int i;
-
-	while (!list_empty(&btrfs_heuristic_ws.idle_ws)) {
-		workspace = btrfs_heuristic_ws.idle_ws.next;
-		list_del(workspace);
-		free_heuristic_ws(workspace);
-		atomic_dec(&btrfs_heuristic_ws.total_ws);
-	}
-
-	for (i = 0; i < BTRFS_COMPRESS_TYPES; i++) {
-		while (!list_empty(&btrfs_comp_ws[i].idle_ws)) {
-			workspace = btrfs_comp_ws[i].idle_ws.next;
-			list_del(workspace);
-			btrfs_compress_op[i]->free_workspace(workspace);
-			atomic_dec(&btrfs_comp_ws[i].total_ws);
-		}
-	}
+	return btrfs_compress_op[type]->put_workspace(ws);
 }
 
 /*
@@ -1036,18 +1003,17 @@ int btrfs_compress_pages(unsigned int type_level, struct address_space *mapping,
 			 unsigned long *total_in,
 			 unsigned long *total_out)
 {
+	int type = btrfs_compress_type(type_level);
+	int level = btrfs_compress_level(type_level);
 	struct list_head *workspace;
 	int ret;
-	int type = type_level & 0xF;
-
-	workspace = find_workspace(type);
 
-	btrfs_compress_op[type - 1]->set_level(workspace, type_level);
-	ret = btrfs_compress_op[type-1]->compress_pages(workspace, mapping,
+	workspace = get_workspace(type, level);
+	ret = btrfs_compress_op[type]->compress_pages(workspace, mapping,
 						      start, pages,
 						      out_pages,
 						      total_in, total_out);
-	free_workspace(type, workspace);
+	put_workspace(type, workspace);
 	return ret;
 }
 
@@ -1071,9 +1037,9 @@ static int btrfs_decompress_bio(struct compressed_bio *cb)
 	int ret;
 	int type = cb->compress_type;
 
-	workspace = find_workspace(type);
-	ret = btrfs_compress_op[type - 1]->decompress_bio(workspace, cb);
-	free_workspace(type, workspace);
+	workspace = get_workspace(type, 0);
+	ret = btrfs_compress_op[type]->decompress_bio(workspace, cb);
+	put_workspace(type, workspace);
 
 	return ret;
 }
@@ -1089,19 +1055,29 @@ int btrfs_decompress(int type, unsigned char *data_in, struct page *dest_page,
 	struct list_head *workspace;
 	int ret;
 
-	workspace = find_workspace(type);
-
-	ret = btrfs_compress_op[type-1]->decompress(workspace, data_in,
+	workspace = get_workspace(type, 0);
+	ret = btrfs_compress_op[type]->decompress(workspace, data_in,
 						  dest_page, start_byte,
 						  srclen, destlen);
+	put_workspace(type, workspace);
 
-	free_workspace(type, workspace);
 	return ret;
 }
 
+void __init btrfs_init_compress(void)
+{
+	int i;
+
+	for (i = 0; i < BTRFS_NR_WORKSPACE_MANAGERS; i++)
+		btrfs_compress_op[i]->init_workspace_manager();
+}
+
 void __cold btrfs_exit_compress(void)
 {
-	free_workspaces();
+	int i;
+
+	for (i = 0; i < BTRFS_NR_WORKSPACE_MANAGERS; i++)
+		btrfs_compress_op[i]->cleanup_workspace_manager();
 }
 
 /*
@@ -1512,7 +1488,7 @@ static void heuristic_collect_sample(struct inode *inode, u64 start, u64 end,
  */
 int btrfs_compress_heuristic(struct inode *inode, u64 start, u64 end)
 {
-	struct list_head *ws_list = __find_workspace(0, true);
+	struct list_head *ws_list = get_workspace(0, 0);
 	struct heuristic_ws *ws;
 	u32 i;
 	u8 byte;
@@ -1581,18 +1557,29 @@ int btrfs_compress_heuristic(struct inode *inode, u64 start, u64 end)
 	}
 
 out:
-	__free_workspace(0, ws_list, true);
+	put_workspace(0, ws_list);
 	return ret;
 }
 
-unsigned int btrfs_compress_str2level(const char *str)
+/*
+ * Convert the compression suffix (eg. after "zlib" starting with ":") to
+ * level, unrecognized string will set the default level
+ */
+unsigned int btrfs_compress_str2level(unsigned int type, const char *str)
 {
-	if (strncmp(str, "zlib", 4) != 0)
+	unsigned int level = 0;
+	int ret;
+
+	if (!type)
 		return 0;
 
-	/* Accepted form: zlib:1 up to zlib:9 and nothing left after the number */
-	if (str[4] == ':' && '1' <= str[5] && str[5] <= '9' && str[6] == 0)
-		return str[5] - '0';
+	if (str[0] == ':') {
+		ret = kstrtouint(str + 1, 10, &level);
+		if (ret)
+			level = 0;
+	}
+
+	level = btrfs_compress_op[type]->set_level(level);
 
-	return BTRFS_ZLIB_DEFAULT_LEVEL;
+	return level;
 }
diff --git a/fs/btrfs/compression.h b/fs/btrfs/compression.h
index ddda9b80bf20..9976fe0f7526 100644
--- a/fs/btrfs/compression.h
+++ b/fs/btrfs/compression.h
@@ -64,6 +64,16 @@ struct compressed_bio {
 	u32 sums;
 };
 
+static inline unsigned int btrfs_compress_type(unsigned int type_level)
+{
+	return (type_level & 0xF);
+}
+
+static inline unsigned int btrfs_compress_level(unsigned int type_level)
+{
+	return ((type_level & 0xF0) >> 4);
+}
+
 void __init btrfs_init_compress(void);
 void __cold btrfs_exit_compress(void);
 
@@ -87,7 +97,7 @@ blk_status_t btrfs_submit_compressed_write(struct inode *inode, u64 start,
 blk_status_t btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 				 int mirror_num, unsigned long bio_flags);
 
-unsigned btrfs_compress_str2level(const char *str);
+unsigned int btrfs_compress_str2level(unsigned int type, const char *str);
 
 enum btrfs_compression_type {
 	BTRFS_COMPRESS_NONE  = 0,
@@ -97,8 +107,35 @@ enum btrfs_compression_type {
 	BTRFS_COMPRESS_TYPES = 3,
 };
 
+struct workspace_manager {
+	const struct btrfs_compress_op *ops;
+	struct list_head idle_ws;
+	spinlock_t ws_lock;
+	/* Number of free workspaces */
+	int free_ws;
+	/* Total number of allocated workspaces */
+	atomic_t total_ws;
+	/* Waiters for a free workspace */
+	wait_queue_head_t ws_wait;
+};
+
+void btrfs_init_workspace_manager(struct workspace_manager *wsm,
+				  const struct btrfs_compress_op *ops);
+struct list_head *btrfs_get_workspace(struct workspace_manager *wsm,
+				      unsigned int level);
+void btrfs_put_workspace(struct workspace_manager *wsm, struct list_head *ws);
+void btrfs_cleanup_workspace_manager(struct workspace_manager *wsm);
+
 struct btrfs_compress_op {
-	struct list_head *(*alloc_workspace)(void);
+	void (*init_workspace_manager)(void);
+
+	void (*cleanup_workspace_manager)(void);
+
+	struct list_head *(*get_workspace)(unsigned int level);
+
+	void (*put_workspace)(struct list_head *ws);
+
+	struct list_head *(*alloc_workspace)(unsigned int level);
 
 	void (*free_workspace)(struct list_head *workspace);
 
@@ -119,9 +156,18 @@ struct btrfs_compress_op {
 			  unsigned long start_byte,
 			  size_t srclen, size_t destlen);
 
-	void (*set_level)(struct list_head *ws, unsigned int type);
+	/*
+	 * This bounds the level set by the user to be within range of a
+	 * particular compression type.  It returns the level that will be used
+	 * if the level is out of bounds or the default if 0 is passed in.
+	 */
+	unsigned int (*set_level)(unsigned int level);
 };
 
+/* The heuristic workspaces are managed via the 0th workspace manager */
+#define BTRFS_NR_WORKSPACE_MANAGERS	(BTRFS_COMPRESS_TYPES + 1)
+
+extern const struct btrfs_compress_op btrfs_heuristic_compress;
 extern const struct btrfs_compress_op btrfs_zlib_compress;
 extern const struct btrfs_compress_op btrfs_lzo_compress;
 extern const struct btrfs_compress_op btrfs_zstd_compress;
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 5a6c39b44c84..324df36d28bf 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -13,6 +13,7 @@
 #include "print-tree.h"
 #include "locking.h"
 #include "volumes.h"
+#include "qgroup.h"
 
 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, int level);
@@ -45,11 +46,18 @@ noinline void btrfs_set_path_blocking(struct btrfs_path *p)
 	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
 		if (!p->nodes[i] || !p->locks[i])
 			continue;
-		btrfs_set_lock_blocking_rw(p->nodes[i], p->locks[i]);
-		if (p->locks[i] == BTRFS_READ_LOCK)
+		/*
+		 * If we currently have a spinning reader or writer lock this
+		 * will bump the count of blocking holders and drop the
+		 * spinlock.
+		 */
+		if (p->locks[i] == BTRFS_READ_LOCK) {
+			btrfs_set_lock_blocking_read(p->nodes[i]);
 			p->locks[i] = BTRFS_READ_LOCK_BLOCKING;
-		else if (p->locks[i] == BTRFS_WRITE_LOCK)
+		} else if (p->locks[i] == BTRFS_WRITE_LOCK) {
+			btrfs_set_lock_blocking_write(p->nodes[i]);
 			p->locks[i] = BTRFS_WRITE_LOCK_BLOCKING;
+		}
 	}
 }
 
@@ -1288,7 +1296,7 @@ tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct btrfs_path *path,
 		return eb;
 
 	btrfs_set_path_blocking(path);
-	btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+	btrfs_set_lock_blocking_read(eb);
 
 	if (tm->op == MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
 		BUG_ON(tm->slot != 0);
@@ -1378,7 +1386,7 @@ get_old_root(struct btrfs_root *root, u64 time_seq)
 		free_extent_buffer(eb_root);
 		eb = alloc_dummy_extent_buffer(fs_info, logical);
 	} else {
-		btrfs_set_lock_blocking_rw(eb_root, BTRFS_READ_LOCK);
+		btrfs_set_lock_blocking_read(eb_root);
 		eb = btrfs_clone_extent_buffer(eb_root);
 		btrfs_tree_read_unlock_blocking(eb_root);
 		free_extent_buffer(eb_root);
@@ -1486,9 +1494,16 @@ noinline int btrfs_cow_block(struct btrfs_trans_handle *trans,
 	search_start = buf->start & ~((u64)SZ_1G - 1);
 
 	if (parent)
-		btrfs_set_lock_blocking(parent);
-	btrfs_set_lock_blocking(buf);
+		btrfs_set_lock_blocking_write(parent);
+	btrfs_set_lock_blocking_write(buf);
 
+	/*
+	 * Before CoWing this block for later modification, check if it's
+	 * the subtree root and do the delayed subtree trace if needed.
+	 *
+	 * Also We don't care about the error, as it's handled internally.
+	 */
+	btrfs_qgroup_trace_subtree_after_cow(trans, root, buf);
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
 
@@ -1582,7 +1597,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	if (parent_nritems <= 1)
 		return 0;
 
-	btrfs_set_lock_blocking(parent);
+	btrfs_set_lock_blocking_write(parent);
 
 	for (i = start_slot; i <= end_slot; i++) {
 		struct btrfs_key first_key;
@@ -1641,7 +1656,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 			search_start = last_block;
 
 		btrfs_tree_lock(cur);
-		btrfs_set_lock_blocking(cur);
+		btrfs_set_lock_blocking_write(cur);
 		err = __btrfs_cow_block(trans, root, cur, parent, i,
 					&cur, search_start,
 					min(16 * blocksize,
@@ -1856,7 +1871,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		}
 
 		btrfs_tree_lock(child);
-		btrfs_set_lock_blocking(child);
+		btrfs_set_lock_blocking_write(child);
 		ret = btrfs_cow_block(trans, root, child, mid, 0, &child);
 		if (ret) {
 			btrfs_tree_unlock(child);
@@ -1894,7 +1909,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 
 	if (left) {
 		btrfs_tree_lock(left);
-		btrfs_set_lock_blocking(left);
+		btrfs_set_lock_blocking_write(left);
 		wret = btrfs_cow_block(trans, root, left,
 				       parent, pslot - 1, &left);
 		if (wret) {
@@ -1909,7 +1924,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 
 	if (right) {
 		btrfs_tree_lock(right);
-		btrfs_set_lock_blocking(right);
+		btrfs_set_lock_blocking_write(right);
 		wret = btrfs_cow_block(trans, root, right,
 				       parent, pslot + 1, &right);
 		if (wret) {
@@ -2072,7 +2087,7 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 		u32 left_nr;
 
 		btrfs_tree_lock(left);
-		btrfs_set_lock_blocking(left);
+		btrfs_set_lock_blocking_write(left);
 
 		left_nr = btrfs_header_nritems(left);
 		if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
@@ -2127,7 +2142,7 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 		u32 right_nr;
 
 		btrfs_tree_lock(right);
-		btrfs_set_lock_blocking(right);
+		btrfs_set_lock_blocking_write(right);
 
 		right_nr = btrfs_header_nritems(right);
 		if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
@@ -2529,26 +2544,6 @@ done:
 	return ret;
 }
 
-static void key_search_validate(struct extent_buffer *b,
-				const struct btrfs_key *key,
-				int level)
-{
-#ifdef CONFIG_BTRFS_ASSERT
-	struct btrfs_disk_key disk_key;
-
-	btrfs_cpu_key_to_disk(&disk_key, key);
-
-	if (level == 0)
-