Merge tag 'for-4.17-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "This contains a few fixups to the qgroup patches that were merged this
  dev cycle, unaligned access fix, blockgroup removal corner case fix
  and a small debugging output tweak"

* tag 'for-4.17-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: print-tree: debugging output enhancement
  btrfs: Fix race condition between delayed refs and blockgroup removal
  btrfs: fix unaligned access in readdir
  btrfs: Fix wrong btrfs_delalloc_release_extents parameter
  btrfs: delayed-inode: Remove wrong qgroup meta reservation calls
  btrfs: qgroup: Use independent and accurate per inode qgroup rsv
  btrfs: qgroup: Commit transaction in advance to reduce early EDQUOT

fs/btrfs/ctree.h

@@ -459,6 +459,25 @@ struct btrfs_block_rsv {
 	unsigned short full;
 	unsigned short type;
 	unsigned short failfast;
+
+	/*
+	 * Qgroup equivalent for @size @reserved
+	 *
+	 * Unlike normal @size/@reserved for inode rsv, qgroup doesn't care
+	 * about things like csum size nor how many tree blocks it will need to
+	 * reserve.
+	 *
+	 * Qgroup cares more about net change of the extent usage.
+	 *
+	 * So for one newly inserted file extent, in worst case it will cause
+	 * leaf split and level increase, nodesize for each file extent is
+	 * already too much.
+	 *
+	 * In short, qgroup_size/reserved is the upper limit of possible needed
+	 * qgroup metadata reservation.
+	 */
+	u64 qgroup_rsv_size;
+	u64 qgroup_rsv_reserved;
 };
 
 /*
@@ -714,6 +733,12 @@ struct btrfs_delayed_root;
  */
 #define BTRFS_FS_EXCL_OP			16
 
+/*
+ * To info transaction_kthread we need an immediate commit so it doesn't
+ * need to wait for commit_interval
+ */
+#define BTRFS_FS_NEED_ASYNC_COMMIT		17
+
 struct btrfs_fs_info {
 	u8 fsid[BTRFS_FSID_SIZE];
 	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];

fs/btrfs/delayed-inode.c

@@ -556,6 +556,12 @@ static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
 	dst_rsv = &fs_info->delayed_block_rsv;
 
 	num_bytes = btrfs_calc_trans_metadata_size(fs_info, 1);
+
+	/*
+	 * Here we migrate space rsv from transaction rsv, since have already
+	 * reserved space when starting a transaction.  So no need to reserve
+	 * qgroup space here.
+	 */
 	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes, 1);
 	if (!ret) {
 		trace_btrfs_space_reservation(fs_info, "delayed_item",
@@ -577,7 +583,10 @@ static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
 		return;
 
 	rsv = &fs_info->delayed_block_rsv;
-	btrfs_qgroup_convert_reserved_meta(root, item->bytes_reserved);
+	/*
+	 * Check btrfs_delayed_item_reserve_metadata() to see why we don't need
+	 * to release/reserve qgroup space.
+	 */
 	trace_btrfs_space_reservation(fs_info, "delayed_item",
 				      item->key.objectid, item->bytes_reserved,
 				      0);
@@ -602,9 +611,6 @@ static int btrfs_delayed_inode_reserve_metadata(
 
 	num_bytes = btrfs_calc_trans_metadata_size(fs_info, 1);
 
-	ret = btrfs_qgroup_reserve_meta_prealloc(root, num_bytes, true);
-	if (ret < 0)
-		return ret;
 	/*
 	 * btrfs_dirty_inode will update the inode under btrfs_join_transaction
 	 * which doesn't reserve space for speed.  This is a problem since we
@@ -616,6 +622,10 @@ static int btrfs_delayed_inode_reserve_metadata(
 	 */
 	if (!src_rsv || (!trans->bytes_reserved &&
 			 src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC)) {
+		ret = btrfs_qgroup_reserve_meta_prealloc(root,
+				fs_info->nodesize, true);
+		if (ret < 0)
+			return ret;
 		ret = btrfs_block_rsv_add(root, dst_rsv, num_bytes,
 					  BTRFS_RESERVE_NO_FLUSH);
 		/*
@@ -634,6 +644,8 @@ static int btrfs_delayed_inode_reserve_metadata(
 						      "delayed_inode",
 						      btrfs_ino(inode),
 						      num_bytes, 1);
+		} else {
+			btrfs_qgroup_free_meta_prealloc(root, fs_info->nodesize);
 		}
 		return ret;
 	}

fs/btrfs/delayed-ref.c

@@ -540,8 +540,10 @@ add_delayed_ref_head(struct btrfs_fs_info *fs_info,
 		     struct btrfs_delayed_ref_head *head_ref,
 		     struct btrfs_qgroup_extent_record *qrecord,
 		     u64 bytenr, u64 num_bytes, u64 ref_root, u64 reserved,
-		     int action, int is_data, int *qrecord_inserted_ret,
+		     int action, int is_data, int is_system,
+		     int *qrecord_inserted_ret,
 		     int *old_ref_mod, int *new_ref_mod)
+
 {
 	struct btrfs_delayed_ref_head *existing;
 	struct btrfs_delayed_ref_root *delayed_refs;
@@ -585,6 +587,7 @@ add_delayed_ref_head(struct btrfs_fs_info *fs_info,
 	head_ref->ref_mod = count_mod;
 	head_ref->must_insert_reserved = must_insert_reserved;
 	head_ref->is_data = is_data;
+	head_ref->is_system = is_system;
 	head_ref->ref_tree = RB_ROOT;
 	INIT_LIST_HEAD(&head_ref->ref_add_list);
 	RB_CLEAR_NODE(&head_ref->href_node);
@@ -772,6 +775,7 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
 	struct btrfs_delayed_ref_root *delayed_refs;
 	struct btrfs_qgroup_extent_record *record = NULL;
 	int qrecord_inserted;
+	int is_system = (ref_root == BTRFS_CHUNK_TREE_OBJECTID);
 
 	BUG_ON(extent_op && extent_op->is_data);
 	ref = kmem_cache_alloc(btrfs_delayed_tree_ref_cachep, GFP_NOFS);
@@ -800,8 +804,8 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
 	 */
 	head_ref = add_delayed_ref_head(fs_info, trans, head_ref, record,
 					bytenr, num_bytes, 0, 0, action, 0,
-					&qrecord_inserted, old_ref_mod,
-					new_ref_mod);
+					is_system, &qrecord_inserted,
+					old_ref_mod, new_ref_mod);
 
 	add_delayed_tree_ref(fs_info, trans, head_ref, &ref->node, bytenr,
 			     num_bytes, parent, ref_root, level, action);
@@ -868,7 +872,7 @@ int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
 	 */
 	head_ref = add_delayed_ref_head(fs_info, trans, head_ref, record,
 					bytenr, num_bytes, ref_root, reserved,
-					action, 1, &qrecord_inserted,
+					action, 1, 0, &qrecord_inserted,
 					old_ref_mod, new_ref_mod);
 
 	add_delayed_data_ref(fs_info, trans, head_ref, &ref->node, bytenr,
@@ -898,9 +902,14 @@ int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
 	delayed_refs = &trans->transaction->delayed_refs;
 	spin_lock(&delayed_refs->lock);
 
+	/*
+	 * extent_ops just modify the flags of an extent and they don't result
+	 * in ref count changes, hence it's safe to pass false/0 for is_system
+	 * argument
+	 */
 	add_delayed_ref_head(fs_info, trans, head_ref, NULL, bytenr,
 			     num_bytes, 0, 0, BTRFS_UPDATE_DELAYED_HEAD,
-			     extent_op->is_data, NULL, NULL, NULL);
+			     extent_op->is_data, 0, NULL, NULL, NULL);
 
 	spin_unlock(&delayed_refs->lock);
 	return 0;

fs/btrfs/delayed-ref.h

@@ -127,6 +127,7 @@ struct btrfs_delayed_ref_head {
 	 */
 	unsigned int must_insert_reserved:1;
 	unsigned int is_data:1;
+	unsigned int is_system:1;
 	unsigned int processing:1;
 };
 

fs/btrfs/disk-io.c

@@ -1824,6 +1824,7 @@ static int transaction_kthread(void *arg)
 
 		now = get_seconds();
 		if (cur->state < TRANS_STATE_BLOCKED &&
+		    !test_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags) &&
 		    (now < cur->start_time ||
 		     now - cur->start_time < fs_info->commit_interval)) {
 			spin_unlock(&fs_info->trans_lock);

fs/btrfs/extent-tree.c

@@ -2601,13 +2601,19 @@ static int cleanup_ref_head(struct btrfs_trans_handle *trans,
 	trace_run_delayed_ref_head(fs_info, head, 0);
 
 	if (head->total_ref_mod < 0) {
-		struct btrfs_block_group_cache *cache;
+		struct btrfs_space_info *space_info;
+		u64 flags;
 
-		cache = btrfs_lookup_block_group(fs_info, head->bytenr);
-		ASSERT(cache);
-		percpu_counter_add(&cache->space_info->total_bytes_pinned,
+		if (head->is_data)
+			flags = BTRFS_BLOCK_GROUP_DATA;
+		else if (head->is_system)
+			flags = BTRFS_BLOCK_GROUP_SYSTEM;
+		else
+			flags = BTRFS_BLOCK_GROUP_METADATA;
+		space_info = __find_space_info(fs_info, flags);
+		ASSERT(space_info);
+		percpu_counter_add(&space_info->total_bytes_pinned,
 				   -head->num_bytes);
-		btrfs_put_block_group(cache);
 
 		if (head->is_data) {
 			spin_lock(&delayed_refs->lock);
@@ -5559,14 +5565,18 @@ static void space_info_add_new_bytes(struct btrfs_fs_info *fs_info,
 
 static u64 block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
 				    struct btrfs_block_rsv *block_rsv,
-				    struct btrfs_block_rsv *dest, u64 num_bytes)
+				    struct btrfs_block_rsv *dest, u64 num_bytes,
+				    u64 *qgroup_to_release_ret)
 {
 	struct btrfs_space_info *space_info = block_rsv->space_info;
+	u64 qgroup_to_release = 0;
 	u64 ret;
 
 	spin_lock(&block_rsv->lock);
-	if (num_bytes == (u64)-1)
+	if (num_bytes == (u64)-1) {
 		num_bytes = block_rsv->size;
+		qgroup_to_release = block_rsv->qgroup_rsv_size;
+	}
 	block_rsv->size -= num_bytes;
 	if (block_rsv->reserved >= block_rsv->size) {
 		num_bytes = block_rsv->reserved - block_rsv->size;
@@ -5575,6 +5585,13 @@ static u64 block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
 	} else {
 		num_bytes = 0;
 	}
+	if (block_rsv->qgroup_rsv_reserved >= block_rsv->qgroup_rsv_size) {
+		qgroup_to_release = block_rsv->qgroup_rsv_reserved -
+				    block_rsv->qgroup_rsv_size;
+		block_rsv->qgroup_rsv_reserved = block_rsv->qgroup_rsv_size;
+	} else {
+		qgroup_to_release = 0;
+	}
 	spin_unlock(&block_rsv->lock);
 
 	ret = num_bytes;
@@ -5597,6 +5614,8 @@ static u64 block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
 			space_info_add_old_bytes(fs_info, space_info,
 						 num_bytes);
 	}
+	if (qgroup_to_release_ret)
+		*qgroup_to_release_ret = qgroup_to_release;
 	return ret;
 }
 
@@ -5738,17 +5757,21 @@ static int btrfs_inode_rsv_refill(struct btrfs_inode *inode,
 	struct btrfs_root *root = inode->root;
 	struct btrfs_block_rsv *block_rsv = &inode->block_rsv;
 	u64 num_bytes = 0;
+	u64 qgroup_num_bytes = 0;
 	int ret = -ENOSPC;
 
 	spin_lock(&block_rsv->lock);
 	if (block_rsv->reserved < block_rsv->size)
 		num_bytes = block_rsv->size - block_rsv->reserved;
+	if (block_rsv->qgroup_rsv_reserved < block_rsv->qgroup_rsv_size)
+		qgroup_num_bytes = block_rsv->qgroup_rsv_size -
+				   block_rsv->qgroup_rsv_reserved;
 	spin_unlock(&block_rsv->lock);
 
 	if (num_bytes == 0)
 		return 0;
 
-	ret = btrfs_qgroup_reserve_meta_prealloc(root, num_bytes, true);
+	ret = btrfs_qgroup_reserve_meta_prealloc(root, qgroup_num_bytes, true);
 	if (ret)
 		return ret;
 	ret = reserve_metadata_bytes(root, block_rsv, num_bytes, flush);
@@ -5756,7 +5779,13 @@ static int btrfs_inode_rsv_refill(struct btrfs_inode *inode,
 		block_rsv_add_bytes(block_rsv, num_bytes, 0);
 		trace_btrfs_space_reservation(root->fs_info, "delalloc",
 					      btrfs_ino(inode), num_bytes, 1);
-	}
+
+		/* Don't forget to increase qgroup_rsv_reserved */
+		spin_lock(&block_rsv->lock);
+		block_rsv->qgroup_rsv_reserved += qgroup_num_bytes;
+		spin_unlock(&block_rsv->lock);
+	} else
+		btrfs_qgroup_free_meta_prealloc(root, qgroup_num_bytes);
 	return ret;
 }
 
@@ -5777,20 +5806,23 @@ static void btrfs_inode_rsv_release(struct btrfs_inode *inode, bool qgroup_free)
 	struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
 	struct btrfs_block_rsv *block_rsv = &inode->block_rsv;
 	u64 released = 0;
+	u64 qgroup_to_release = 0;
 
 	/*
 	 * Since we statically set the block_rsv->size we just want to say we
 	 * are releasing 0 bytes, and then we'll just get the reservation over
 	 * the size free'd.
 	 */
-	released = block_rsv_release_bytes(fs_info, block_rsv, global_rsv, 0);
+	released = block_rsv_release_bytes(fs_info, block_rsv, global_rsv, 0,
+					   &qgroup_to_release);
 	if (released > 0)
 		trace_btrfs_space_reservation(fs_info, "delalloc",
 					      btrfs_ino(inode), released, 0);
 	if (qgroup_free)
-		btrfs_qgroup_free_meta_prealloc(inode->root, released);
+		btrfs_qgroup_free_meta_prealloc(inode->root, qgroup_to_release);
 	else
-		btrfs_qgroup_convert_reserved_meta(inode->root, released);
+		btrfs_qgroup_convert_reserved_meta(inode->root,
+						   qgroup_to_release);
 }
 
 void btrfs_block_rsv_release(struct btrfs_fs_info *fs_info,
@@ -5802,7 +5834,7 @@ void btrfs_block_rsv_release(struct btrfs_fs_info *fs_info,
 	if (global_rsv == block_rsv ||
 	    block_rsv->space_info != global_rsv->space_info)
 		global_rsv = NULL;
-	block_rsv_release_bytes(fs_info, block_rsv, global_rsv, num_bytes);
+	block_rsv_release_bytes(fs_info, block_rsv, global_rsv, num_bytes, NULL);
 }
 
 static void update_global_block_rsv(struct btrfs_fs_info *fs_info)
@@ -5882,7 +5914,7 @@ static void init_global_block_rsv(struct btrfs_fs_info *fs_info)
 static void release_global_block_rsv(struct btrfs_fs_info *fs_info)
 {
 	block_rsv_release_bytes(fs_info, &fs_info->global_block_rsv, NULL,
-				(u64)-1);
+				(u64)-1, NULL);
 	WARN_ON(fs_info->trans_block_rsv.size > 0);
 	WARN_ON(fs_info->trans_block_rsv.reserved > 0);
 	WARN_ON(fs_info->chunk_block_rsv.size > 0);
@@ -5906,7 +5938,7 @@ void btrfs_trans_release_chunk_metadata(struct btrfs_trans_handle *trans)
 	WARN_ON_ONCE(!list_empty(&trans->new_bgs));
 
 	block_rsv_release_bytes(fs_info, &fs_info->chunk_block_rsv, NULL,
-				trans->chunk_bytes_reserved);
+				trans->chunk_bytes_reserved, NULL);
 	trans->chunk_bytes_reserved = 0;
 }
 
@@ -6011,6 +6043,7 @@ static void btrfs_calculate_inode_block_rsv_size(struct btrfs_fs_info *fs_info,
 {
 	struct btrfs_block_rsv *block_rsv = &inode->block_rsv;
 	u64 reserve_size = 0;
+	u64 qgroup_rsv_size = 0;
 	u64 csum_leaves;
 	unsigned outstanding_extents;
 
@@ -6023,9 +6056,17 @@ static void btrfs_calculate_inode_block_rsv_size(struct btrfs_fs_info *fs_info,
 						 inode->csum_bytes);
 	reserve_size += btrfs_calc_trans_metadata_size(fs_info,
 						       csum_leaves);
+	/*
+	 * For qgroup rsv, the calculation is very simple:
+	 * account one nodesize for each outstanding extent
+	 *
+	 * This is overestimating in most cases.
+	 */
+	qgroup_rsv_size = outstanding_extents * fs_info->nodesize;
 
 	spin_lock(&block_rsv->lock);
 	block_rsv->size = reserve_size;
+	block_rsv->qgroup_rsv_size = qgroup_rsv_size;
 	spin_unlock(&block_rsv->lock);
 }
 
@@ -8403,7 +8444,7 @@ static void unuse_block_rsv(struct btrfs_fs_info *fs_info,
 			    struct btrfs_block_rsv *block_rsv, u32 blocksize)
 {
 	block_rsv_add_bytes(block_rsv, blocksize, 0);
-	block_rsv_release_bytes(fs_info, block_rsv, NULL, 0);
+	block_rsv_release_bytes(fs_info, block_rsv, NULL, 0, NULL);
 }
 
 /*

fs/btrfs/file.c

@@ -1748,7 +1748,7 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
 			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
 					     lockstart, lockend, &cached_state);
 		btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes,
-					       (ret != 0));
+					       true);
 		if (ret) {
 			btrfs_drop_pages(pages, num_pages);
 			break;

fs/btrfs/inode.c

@@ -31,6 +31,7 @@
 #include <linux/uio.h>
 #include <linux/magic.h>
 #include <linux/iversion.h>
+#include <asm/unaligned.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -5905,11 +5906,13 @@ static int btrfs_filldir(void *addr, int entries, struct dir_context *ctx)
 		struct dir_entry *entry = addr;
 		char *name = (char *)(entry + 1);
 
-		ctx->pos = entry->offset;
-		if (!dir_emit(ctx, name, entry->name_len, entry->ino,
-			      entry->type))
+		ctx->pos = get_unaligned(&entry->offset);
+		if (!dir_emit(ctx, name, get_unaligned(&entry->name_len),
+					 get_unaligned(&entry->ino),
+					 get_unaligned(&entry->type)))
 			return 1;
-		addr += sizeof(struct dir_entry) + entry->name_len;
+		addr += sizeof(struct dir_entry) +
+			get_unaligned(&entry->name_len);
 		ctx->pos++;
 	}
 	return 0;
@@ -5999,14 +6002,15 @@ static int btrfs_real_readdir(struct file *file, struct dir_context *ctx)
 		}
 
 		entry = addr;
-		entry->name_len = name_len;
+		put_unaligned(name_len, &entry->name_len);
 		name_ptr = (char *)(entry + 1);
 		read_extent_buffer(leaf, name_ptr, (unsigned long)(di + 1),
 				   name_len);
-		entry->type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
+		put_unaligned(btrfs_filetype_table[btrfs_dir_type(leaf, di)],
+				&entry->type);
 		btrfs_dir_item_key_to_cpu(leaf, di, &location);
-		entry->ino = location.objectid;
-		entry->offset = found_key.offset;
+		put_unaligned(location.objectid, &entry->ino);
+		put_unaligned(found_key.offset, &entry->offset);
 		entries++;
 		addr += sizeof(struct dir_entry) + name_len;
 		total_len += sizeof(struct dir_entry) + name_len;

fs/btrfs/print-tree.c

@@ -189,9 +189,10 @@ void btrfs_print_leaf(struct extent_buffer *l)
 	fs_info = l->fs_info;
 	nr = btrfs_header_nritems(l);
 
-	btrfs_info(fs_info, "leaf %llu total ptrs %d free space %d",
-		   btrfs_header_bytenr(l), nr,
-		   btrfs_leaf_free_space(fs_info, l));
+	btrfs_info(fs_info,
+		   "leaf %llu gen %llu total ptrs %d free space %d owner %llu",
+		   btrfs_header_bytenr(l), btrfs_header_generation(l), nr,
+		   btrfs_leaf_free_space(fs_info, l), btrfs_header_owner(l));
 	for (i = 0 ; i < nr ; i++) {
 		item = btrfs_item_nr(i);
 		btrfs_item_key_to_cpu(l, &key, i);
@@ -325,7 +326,7 @@ void btrfs_print_leaf(struct extent_buffer *l)
 	}
 }
 
-void btrfs_print_tree(struct extent_buffer *c)
+void btrfs_print_tree(struct extent_buffer *c, bool follow)
 {
 	struct btrfs_fs_info *fs_info;
 	int i; u32 nr;
@@ -342,15 +343,19 @@ void btrfs_print_tree(struct extent_buffer *c)
 		return;
 	}
 	btrfs_info(fs_info,
-		   "node %llu level %d total ptrs %d free spc %u",
-		   btrfs_header_bytenr(c), level, nr,
-		   (u32)BTRFS_NODEPTRS_PER_BLOCK(fs_info) - nr);
+		   "node %llu level %d gen %llu total ptrs %d free spc %u owner %llu",
+		   btrfs_header_bytenr(c), level, btrfs_header_generation(c),
+		   nr, (u32)BTRFS_NODEPTRS_PER_BLOCK(fs_info) - nr,
+		   btrfs_header_owner(c));
 	for (i = 0; i < nr; i++) {
 		btrfs_node_key_to_cpu(c, &key, i);
-		pr_info("\tkey %d (%llu %u %llu) block %llu\n",
+		pr_info("\tkey %d (%llu %u %llu) block %llu gen %llu\n",
 		       i, key.objectid, key.type, key.offset,
-		       btrfs_node_blockptr(c, i));
+		       btrfs_node_blockptr(c, i),
+		       btrfs_node_ptr_generation(c, i));
 	}
+	if (!follow)
+		return;
 	for (i = 0; i < nr; i++) {
 		struct btrfs_key first_key;
 		struct extent_buffer *next;
@@ -372,7 +377,7 @@ void btrfs_print_tree(struct extent_buffer *c)
 		if (btrfs_header_level(next) !=
 		       level - 1)
 			BUG();
-		btrfs_print_tree(next);
+		btrfs_print_tree(next, follow);
 		free_extent_buffer(next);
 	}
 }

fs/btrfs/print-tree.h

@@ -7,6 +7,6 @@
 #define BTRFS_PRINT_TREE_H
 
 void btrfs_print_leaf(struct extent_buffer *l);
-void btrfs_print_tree(struct extent_buffer *c);
+void btrfs_print_tree(struct extent_buffer *c, bool follow);
 
 #endif

fs/btrfs/qgroup.c

@@ -11,6 +11,7 @@
 #include <linux/slab.h>
 #include <linux/workqueue.h>
 #include <linux/btrfs.h>
+#include <linux/sizes.h>
 
 #include "ctree.h"
 #include "transaction.h"
@@ -2375,8 +2376,21 @@ int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-static bool qgroup_check_limits(const struct btrfs_qgroup *qg, u64 num_bytes)
+/*
+ * Two limits to commit transaction in advance.
+ *
+ * For RATIO, it will be 1/RATIO of the remaining limit
+ * (excluding data and prealloc meta) as threshold.
+ * For SIZE, it will be in byte unit as threshold.
+ */
+#define QGROUP_PERTRANS_RATIO		32
+#define QGROUP_PERTRANS_SIZE		SZ_32M
+static bool qgroup_check_limits(struct btrfs_fs_info *fs_info,
+				const struct btrfs_qgroup *qg, u64 num_bytes)
 {
+	u64 limit;
+	u64 threshold;
+
 	if ((qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_RFER) &&
 	    qgroup_rsv_total(qg) + (s64)qg->rfer + num_bytes > qg->max_rfer)
 		return false;
@@ -2385,6 +2399,31 @@ static bool qgroup_check_limits(const struct btrfs_qgroup *qg, u64 num_bytes)
 	    qgroup_rsv_total(qg) + (s64)qg->excl + num_bytes > qg->max_excl)
 		return false;
 
+	/*
+	 * Even if we passed the check, it's better to check if reservation
+	 * for meta_pertrans is pushing us near limit.
+	 * If there is too much pertrans reservation or it's near the limit,
+	 * let's try commit transaction to free some, using transaction_kthread
+	 */
+	if ((qg->lim_flags & (BTRFS_QGROUP_LIMIT_MAX_RFER |
+			      BTRFS_QGROUP_LIMIT_MAX_EXCL))) {
+		if (qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_EXCL)
+			limit = qg->max_excl;
+		else
+			limit = qg->max_rfer;
+		threshold = (limit - qg->rsv.values[BTRFS_QGROUP_RSV_DATA] -
+			    qg->rsv.values[BTRFS_QGROUP_RSV_META_PREALLOC]) /
+			    QGROUP_PERTRANS_RATIO;
+		threshold = min_t(u64, threshold, QGROUP_PERTRANS_SIZE);
+
+		/*
+		 * Use transaction_kthread to commit transaction, so we no
+		 * longer need to bother nested transaction nor lock context.
+		 */
+		if (qg->rsv.values[BTRFS_QGROUP_RSV_META_PERTRANS] > threshold)
+			btrfs_commit_transaction_locksafe(fs_info);
+	}
+
 	return true;
 }
 
@@ -2434,7 +2473,7 @@ static int qgroup_reserve(struct btrfs_root *root, u64 num_bytes, bool enforce,
 
 		qg = unode_aux_to_qgroup(unode);
 
-		if (enforce && !qgroup_check_limits(qg, num_bytes)) {
+		if (enforce && !qgroup_check_limits(fs_info, qg, num_bytes)) {
 			ret = -EDQUOT;
 			goto out;
 		}

fs/btrfs/transaction.c

@@ -2267,6 +2267,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans)
 	 */
 	cur_trans->state = TRANS_STATE_COMPLETED;
 	wake_up(&cur_trans->commit_wait);
+	clear_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags);
 
 	spin_lock(&fs_info->trans_lock);
 	list_del_init(&cur_trans->list);

fs/btrfs/transaction.h

@@ -199,6 +199,20 @@ int btrfs_clean_one_deleted_snapshot(struct btrfs_root *root);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans);
 int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
 				   int wait_for_unblock);
+
+/*
+ * Try to commit transaction asynchronously, so this is safe to call
+ * even holding a spinlock.
+ *
+ * It's done by informing transaction_kthread to commit transaction without
+ * waiting for commit interval.
+ */
+static inline void btrfs_commit_transaction_locksafe(
+		struct btrfs_fs_info *fs_info)
+{
+	set_bit(BTRFS_FS_NEED_ASYNC_COMMIT, &fs_info->flags);
+	wake_up_process(fs_info->transaction_kthread);
+}
 int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans);
 int btrfs_should_end_transaction(struct btrfs_trans_handle *trans);
 void btrfs_throttle(struct btrfs_fs_info *fs_info);