[PATCH 0/5] Make real_root used only in ref-verify

[PATCH 0/5] Make real_root used only in ref-verify

[Nikolay Borisov]

Here's a small series that refactors the way btrfs_ref::real_root and
btrfs_ref::skip_qgroup are used. Currently the former is used in ref-verify but
also in order to perform the is_fstree() check on it in delayed-ref core. Given
the complexity and amount of information that the delayed ref machinery hauls
around it becomes really non-evident that delrefs really don't care about
real_root itself but rather only if qgroup processing should happen or not.

Instead of having the check burried in the core this series changes the data
flow in such a way that real_root will only be used for ref-verify's operation
and 'skip_qgroup' will contains the final condition of whether qgroup processing
should take place for a given delref.


Nikolay Borisov (5):
  btrfs: Rename root fields in delayed refs structs
  btrfs: Rely on owning_root field in btrfs_add_delayed_tree_ref to
    detect CHUNK_ROOT
  btrfs: Add additional parameters to
    btrfs_init_tree_ref/btrfs_init_data_ref
  btrfs: pull up qgroup checks from delayed-ref core to init time
  btrfs: make real_root optional

 fs/btrfs/delayed-ref.c | 19 +++++++++---------
 fs/btrfs/delayed-ref.h | 44 +++++++++++++++++++++---------------------
 fs/btrfs/extent-tree.c | 32 +++++++++++++++---------------
 fs/btrfs/file.c        | 13 ++++++++-----
 fs/btrfs/inode.c       |  4 ++--
 fs/btrfs/ref-verify.c  |  4 ++--
 fs/btrfs/relocation.c  | 28 +++++++++++++--------------
 fs/btrfs/tree-log.c    |  2 +-
 8 files changed, 74 insertions(+), 72 deletions(-)

--
2.25.1

[PATCH 1/5] btrfs: Rename root fields in delayed refs structs

[Nikolay Borisov]

Both data and metadata delayed ref structures have fields named
root/ref_root respectively. Those are somewhat cryptic and don't really
convey the real meaning. In fact those roots are really the original
owners of the respective block (i.e in case of a snapshot a data delref
will contain the original root that owns the given block). Rename those
fields accordingly and adjust comments.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/delayed-ref.c | 13 +++++++------
 fs/btrfs/delayed-ref.h | 12 ++++++------
 fs/btrfs/extent-tree.c | 10 +++++-----
 fs/btrfs/ref-verify.c  |  4 ++--
 4 files changed, 20 insertions(+), 19 deletions(-)

diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index ca848b183474..53a80163c1d7 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -922,7 +922,7 @@ int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 
 	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
 	    is_fstree(generic_ref->real_root) &&
-	    is_fstree(generic_ref->tree_ref.root) &&
+	    is_fstree(generic_ref->tree_ref.owning_root) &&
 	    !generic_ref->skip_qgroup) {
 		record = kzalloc(sizeof(*record), GFP_NOFS);
 		if (!record) {
@@ -938,14 +938,15 @@ int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 		ref_type = BTRFS_TREE_BLOCK_REF_KEY;
 
 	init_delayed_ref_common(fs_info, &ref->node, bytenr, num_bytes,
-				generic_ref->tree_ref.root, action, ref_type);
-	ref->root = generic_ref->tree_ref.root;
+				generic_ref->tree_ref.owning_root, action,
+				ref_type);
+	ref->root = generic_ref->tree_ref.owning_root;
 	ref->parent = parent;
 	ref->level = level;
 
 	init_delayed_ref_head(head_ref, record, bytenr, num_bytes,
-			      generic_ref->tree_ref.root, 0, action, false,
-			      is_system);
+			      generic_ref->tree_ref.owning_root, 0, action,
+			      false, is_system);
 	head_ref->extent_op = extent_op;
 
 	delayed_refs = &trans->transaction->delayed_refs;
@@ -997,7 +998,7 @@ int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
 	u64 bytenr = generic_ref->bytenr;
 	u64 num_bytes = generic_ref->len;
 	u64 parent = generic_ref->parent;
-	u64 ref_root = generic_ref->data_ref.ref_root;
+	u64 ref_root = generic_ref->data_ref.owning_root;
 	u64 owner = generic_ref->data_ref.ino;
 	u64 offset = generic_ref->data_ref.offset;
 	u8 ref_type;
diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index e22fba272e4f..8826f3e2b809 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -186,8 +186,8 @@ enum btrfs_ref_type {
 struct btrfs_data_ref {
 	/* For EXTENT_DATA_REF */
 
-	/* Root which refers to this data extent */
-	u64 ref_root;
+	/* Original root this data extent belongs to */
+	u64 owning_root;
 
 	/* Inode which refers to this data extent */
 	u64 ino;
@@ -210,11 +210,11 @@ struct btrfs_tree_ref {
 	int level;
 
 	/*
-	 * Root which refers to this tree block.
+	 * Root which owns this tree block.
 	 *
 	 * For TREE_BLOCK_REF (skinny metadata, either inline or keyed)
 	 */
-	u64 root;
+	u64 owning_root;
 
 	/* For non-skinny metadata, no special member needed */
 };
@@ -277,7 +277,7 @@ static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
 	if (!generic_ref->real_root)
 		generic_ref->real_root = root;
 	generic_ref->tree_ref.level = level;
-	generic_ref->tree_ref.root = root;
+	generic_ref->tree_ref.owning_root = root;
 	generic_ref->type = BTRFS_REF_METADATA;
 }
 
@@ -287,7 +287,7 @@ static inline void btrfs_init_data_ref(struct btrfs_ref *generic_ref,
 	/* If @real_root not set, use @root as fallback */
 	if (!generic_ref->real_root)
 		generic_ref->real_root = ref_root;
-	generic_ref->data_ref.ref_root = ref_root;
+	generic_ref->data_ref.owning_root = ref_root;
 	generic_ref->data_ref.ino = ino;
 	generic_ref->data_ref.offset = offset;
 	generic_ref->type = BTRFS_REF_DATA;
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index ec5de19f0acd..de099de3829e 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1396,7 +1396,7 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 	ASSERT(generic_ref->type != BTRFS_REF_NOT_SET &&
 	       generic_ref->action);
 	BUG_ON(generic_ref->type == BTRFS_REF_METADATA &&
-	       generic_ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID);
+	       generic_ref->tree_ref.owning_root == BTRFS_TREE_LOG_OBJECTID);
 
 	if (generic_ref->type == BTRFS_REF_METADATA)
 		ret = btrfs_add_delayed_tree_ref(trans, generic_ref, NULL);
@@ -3372,9 +3372,9 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
 	 * tree, just update pinning info and exit early.
 	 */
 	if ((ref->type == BTRFS_REF_METADATA &&
-	     ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
+	     ref->tree_ref.owning_root == BTRFS_TREE_LOG_OBJECTID) ||
 	    (ref->type == BTRFS_REF_DATA &&
-	     ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)) {
+	     ref->data_ref.owning_root == BTRFS_TREE_LOG_OBJECTID)) {
 		/* unlocks the pinned mutex */
 		btrfs_pin_extent(trans, ref->bytenr, ref->len, 1);
 		ret = 0;
@@ -3385,9 +3385,9 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
 	}
 
 	if (!((ref->type == BTRFS_REF_METADATA &&
-	       ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
+	       ref->tree_ref.owning_root == BTRFS_TREE_LOG_OBJECTID) ||
 	      (ref->type == BTRFS_REF_DATA &&
-	       ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)))
+	       ref->data_ref.owning_root == BTRFS_TREE_LOG_OBJECTID)))
 		btrfs_ref_tree_mod(fs_info, ref);
 
 	return ret;
diff --git a/fs/btrfs/ref-verify.c b/fs/btrfs/ref-verify.c
index d2062d5f71dd..e2b9f8616501 100644
--- a/fs/btrfs/ref-verify.c
+++ b/fs/btrfs/ref-verify.c
@@ -678,10 +678,10 @@ int btrfs_ref_tree_mod(struct btrfs_fs_info *fs_info,
 
 	if (generic_ref->type == BTRFS_REF_METADATA) {
 		if (!parent)
-			ref_root = generic_ref->tree_ref.root;
+			ref_root = generic_ref->tree_ref.owning_root;
 		owner = generic_ref->tree_ref.level;
 	} else if (!parent) {
-		ref_root = generic_ref->data_ref.ref_root;
+		ref_root = generic_ref->data_ref.owning_root;
 		owner = generic_ref->data_ref.ino;
 		offset = generic_ref->data_ref.offset;
 	}
-- 
2.25.1

[PATCH 2/5] btrfs: Rely on owning_root field in btrfs_add_delayed_tree_ref to detect CHUNK_ROOT

[Nikolay Borisov]

real_root field is going to be used only by ref-verify tool so
limit its use outside of it. Blocks belonging to the chunk root will
always have it as an owner so the check is equivalent.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/delayed-ref.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index 53a80163c1d7..7c725379b8ca 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -906,7 +906,9 @@ int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 	u64 parent = generic_ref->parent;
 	u8 ref_type;
 
-	is_system = (generic_ref->real_root == BTRFS_CHUNK_TREE_OBJECTID);
+	is_system = (generic_ref->tree_ref.owning_root
+		     == BTRFS_CHUNK_TREE_OBJECTID);
+
 
 	ASSERT(generic_ref->type == BTRFS_REF_METADATA && generic_ref->action);
 	BUG_ON(extent_op && extent_op->is_data);
-- 
2.25.1

[PATCH 3/5] btrfs: Add additional parameters to btrfs_init_tree_ref/btrfs_init_data_ref

[Nikolay Borisov]

In order to make 'real_root' used only in ref-verify it's required to
have the necessary context to perform the same checks that this member
is used for. So add 'mod_root' which will contain the root on behalf of
which a delayed ref was created and a 'skip_group' parameter which
will contain callsite-specific override of skip_qgroup.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/delayed-ref.h |  5 +++--
 fs/btrfs/extent-tree.c | 17 +++++++++++------
 fs/btrfs/file.c        | 13 ++++++++-----
 fs/btrfs/inode.c       |  3 ++-
 fs/btrfs/relocation.c  | 21 ++++++++++++++-------
 fs/btrfs/tree-log.c    |  2 +-
 6 files changed, 39 insertions(+), 22 deletions(-)

diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index 8826f3e2b809..8ab79def8e98 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -271,7 +271,7 @@ static inline void btrfs_init_generic_ref(struct btrfs_ref *generic_ref,
 }
 
 static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
-				int level, u64 root)
+				int level, u64 root, u64 mod_root, bool skip_qgroup)
 {
 	/* If @real_root not set, use @root as fallback */
 	if (!generic_ref->real_root)
@@ -282,7 +282,8 @@ static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
 }
 
 static inline void btrfs_init_data_ref(struct btrfs_ref *generic_ref,
-				u64 ref_root, u64 ino, u64 offset)
+				u64 ref_root, u64 ino, u64 offset, u64 mod_root,
+				bool skip_qgroup)
 {
 	/* If @real_root not set, use @root as fallback */
 	if (!generic_ref->real_root)
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index de099de3829e..a4fbe4b1df80 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2439,7 +2439,8 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
 					       num_bytes, parent);
 			generic_ref.real_root = root->root_key.objectid;
 			btrfs_init_data_ref(&generic_ref, ref_root, key.objectid,
-					    key.offset);
+					    key.offset, root->root_key.objectid,
+					    for_reloc);
 			generic_ref.skip_qgroup = for_reloc;
 			if (inc)
 				ret = btrfs_inc_extent_ref(trans, &generic_ref);
@@ -2453,7 +2454,8 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
 			btrfs_init_generic_ref(&generic_ref, action, bytenr,
 					       num_bytes, parent);
 			generic_ref.real_root = root->root_key.objectid;
-			btrfs_init_tree_ref(&generic_ref, level - 1, ref_root);
+			btrfs_init_tree_ref(&generic_ref, level - 1, ref_root,
+					    root->root_key.objectid, for_reloc);
 			generic_ref.skip_qgroup = for_reloc;
 			if (inc)
 				ret = btrfs_inc_extent_ref(trans, &generic_ref);
@@ -3288,7 +3290,7 @@ void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
 	btrfs_init_generic_ref(&generic_ref, BTRFS_DROP_DELAYED_REF,
 			       buf->start, buf->len, parent);
 	btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf),
-			    root->root_key.objectid);
+			    root->root_key.objectid, 0, false);
 
 	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
 		btrfs_ref_tree_mod(fs_info, &generic_ref);
@@ -4741,7 +4743,8 @@ int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
 
 	btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
 			       ins->objectid, ins->offset, 0);
-	btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner, offset);
+	btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner,
+			    offset, 0, false);
 	btrfs_ref_tree_mod(root->fs_info, &generic_ref);
 
 	return btrfs_add_delayed_data_ref(trans, &generic_ref, ram_bytes);
@@ -4934,7 +4937,8 @@ struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
 		btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
 				       ins.objectid, ins.offset, parent);
 		generic_ref.real_root = root->root_key.objectid;
-		btrfs_init_tree_ref(&generic_ref, level, root_objectid);
+		btrfs_init_tree_ref(&generic_ref, level, root_objectid,
+				    root->root_key.objectid, false);
 		btrfs_ref_tree_mod(fs_info, &generic_ref);
 		ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, extent_op);
 		if (ret)
@@ -5351,7 +5355,8 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 
 		btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
 				       fs_info->nodesize, parent);
-		btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid);
+		btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid,
+				    0, false);
 		ret = btrfs_free_extent(trans, &ref);
 		if (ret)
 			goto out_unlock;
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 9785ea32d39c..9a3db1365ae8 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -876,7 +876,8 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 				btrfs_init_data_ref(&ref,
 						root->root_key.objectid,
 						new_key.objectid,
-						args->start - extent_offset);
+						args->start - extent_offset,
+						0, false);
 				ret = btrfs_inc_extent_ref(trans, &ref);
 				BUG_ON(ret); /* -ENOMEM */
 			}
@@ -962,7 +963,8 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 				btrfs_init_data_ref(&ref,
 						root->root_key.objectid,
 						key.objectid,
-						key.offset - extent_offset);
+						key.offset - extent_offset, 0,
+						false);
 				ret = btrfs_free_extent(trans, &ref);
 				BUG_ON(ret); /* -ENOMEM */
 				args->bytes_found += extent_end - key.offset;
@@ -1238,7 +1240,7 @@ int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, bytenr,
 				       num_bytes, 0);
 		btrfs_init_data_ref(&ref, root->root_key.objectid, ino,
-				    orig_offset);
+				    orig_offset, 0, false);
 		ret = btrfs_inc_extent_ref(trans, &ref);
 		if (ret) {
 			btrfs_abort_transaction(trans, ret);
@@ -1263,7 +1265,8 @@ int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
 	other_end = 0;
 	btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
 			       num_bytes, 0);
-	btrfs_init_data_ref(&ref, root->root_key.objectid, ino, orig_offset);
+	btrfs_init_data_ref(&ref, root->root_key.objectid, ino, orig_offset,
+			    0, false);
 	if (extent_mergeable(leaf, path->slots[0] + 1,
 			     ino, bytenr, orig_offset,
 			     &other_start, &other_end)) {
@@ -2626,7 +2629,7 @@ static int btrfs_insert_replace_extent(struct btrfs_trans_handle *trans,
 				       extent_info->disk_len, 0);
 		ref_offset = extent_info->file_offset - extent_info->data_offset;
 		btrfs_init_data_ref(&ref, root->root_key.objectid,
-				    btrfs_ino(inode), ref_offset);
+				    btrfs_ino(inode), ref_offset, 0, false);
 		ret = btrfs_inc_extent_ref(trans, &ref);
 	}
 
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index f5faa6bedbd4..4fa0792be22d 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4921,7 +4921,8 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 					extent_start, extent_num_bytes, 0);
 			ref.real_root = root->root_key.objectid;
 			btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
-					ino, extent_offset);
+					ino, extent_offset,
+					root->root_key.objectid, false);
 			ret = btrfs_free_extent(trans, &ref);
 			if (ret) {
 				btrfs_abort_transaction(trans, ret);
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 5e5066ee03e6..ce3631760ef4 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -1148,7 +1148,8 @@ int replace_file_extents(struct btrfs_trans_handle *trans,
 				       num_bytes, parent);
 		ref.real_root = root->root_key.objectid;
 		btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
-				    key.objectid, key.offset);
+				    key.objectid, key.offset,
+				    root->root_key.objectid, false);
 		ret = btrfs_inc_extent_ref(trans, &ref);
 		if (ret) {
 			btrfs_abort_transaction(trans, ret);
@@ -1159,7 +1160,8 @@ int replace_file_extents(struct btrfs_trans_handle *trans,
 				       num_bytes, parent);
 		ref.real_root = root->root_key.objectid;
 		btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
-				    key.objectid, key.offset);
+				    key.objectid, key.offset,
+				    root->root_key.objectid, false);
 		ret = btrfs_free_extent(trans, &ref);
 		if (ret) {
 			btrfs_abort_transaction(trans, ret);
@@ -1369,7 +1371,8 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, old_bytenr,
 				       blocksize, path->nodes[level]->start);
 		ref.skip_qgroup = true;
-		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid);
+		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
+				    0, true);
 		ret = btrfs_inc_extent_ref(trans, &ref);
 		if (ret) {
 			btrfs_abort_transaction(trans, ret);
@@ -1378,7 +1381,8 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, new_bytenr,
 				       blocksize, 0);
 		ref.skip_qgroup = true;
-		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid);
+		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid, 0,
+				    true);
 		ret = btrfs_inc_extent_ref(trans, &ref);
 		if (ret) {
 			btrfs_abort_transaction(trans, ret);
@@ -1387,7 +1391,8 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 
 		btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, new_bytenr,
 				       blocksize, path->nodes[level]->start);
-		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid);
+		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
+				    0, true);
 		ref.skip_qgroup = true;
 		ret = btrfs_free_extent(trans, &ref);
 		if (ret) {
@@ -1397,7 +1402,8 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 
 		btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, old_bytenr,
 				       blocksize, 0);
-		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid);
+		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid,
+				    0, true);
 		ref.skip_qgroup = true;
 		ret = btrfs_free_extent(trans, &ref);
 		if (ret) {
@@ -2476,7 +2482,8 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 					       upper->eb->start);
 			ref.real_root = root->root_key.objectid;
 			btrfs_init_tree_ref(&ref, node->level,
-					    btrfs_header_owner(upper->eb));
+					    btrfs_header_owner(upper->eb),
+					    root->root_key.objectid, false);
 			ret = btrfs_inc_extent_ref(trans, &ref);
 			if (!ret)
 				ret = btrfs_drop_subtree(trans, root, eb,
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 71b3ddb0333d..cda7bb2e9a1e 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -789,7 +789,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 						ins.objectid, ins.offset, 0);
 				btrfs_init_data_ref(&ref,
 						root->root_key.objectid,
-						key->objectid, offset);
+						key->objectid, offset, 0, false);
 				ret = btrfs_inc_extent_ref(trans, &ref);
 				if (ret)
 					goto out;
-- 
2.25.1

[PATCH 4/5] btrfs: pull up qgroup checks from delayed-ref core to init time

[Nikolay Borisov]

Instead of checking whether qgroup processing for a dealyed ref has to
happen in the core of delayed ref, simply pull the check at init time of
respective delayed ref structures. This eliminates the final use of
real_root in delayed-ref core paving the way to making this member
optional.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/delayed-ref.c | 4 ----
 fs/btrfs/delayed-ref.h | 4 ++++
 fs/btrfs/extent-tree.c | 5 -----
 fs/btrfs/inode.c       | 1 -
 fs/btrfs/relocation.c  | 7 -------
 5 files changed, 4 insertions(+), 17 deletions(-)

diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index 7c725379b8ca..405e93ebddae 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -923,8 +923,6 @@ int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 	}
 
 	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
-	    is_fstree(generic_ref->real_root) &&
-	    is_fstree(generic_ref->tree_ref.owning_root) &&
 	    !generic_ref->skip_qgroup) {
 		record = kzalloc(sizeof(*record), GFP_NOFS);
 		if (!record) {
@@ -1029,8 +1027,6 @@ int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
 	}
 
 	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
-	    is_fstree(ref_root) &&
-	    is_fstree(generic_ref->real_root) &&
 	    !generic_ref->skip_qgroup) {
 		record = kzalloc(sizeof(*record), GFP_NOFS);
 		if (!record) {
diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index 8ab79def8e98..6bb299c66e1e 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -279,6 +279,8 @@ static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
 	generic_ref->tree_ref.level = level;
 	generic_ref->tree_ref.owning_root = root;
 	generic_ref->type = BTRFS_REF_METADATA;
+	generic_ref->skip_qgroup = skip_qgroup || !(is_fstree(root) &&
+				    (!mod_root || is_fstree(mod_root)));
 }
 
 static inline void btrfs_init_data_ref(struct btrfs_ref *generic_ref,
@@ -292,6 +294,8 @@ static inline void btrfs_init_data_ref(struct btrfs_ref *generic_ref,
 	generic_ref->data_ref.ino = ino;
 	generic_ref->data_ref.offset = offset;
 	generic_ref->type = BTRFS_REF_DATA;
+	generic_ref->skip_qgroup = skip_qgroup || !(is_fstree(ref_root) &&
+				    (!mod_root || is_fstree(mod_root)));
 }
 
 static inline struct btrfs_delayed_extent_op *
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index a4fbe4b1df80..5a07386bd6d9 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -2437,11 +2437,9 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
 			key.offset -= btrfs_file_extent_offset(buf, fi);
 			btrfs_init_generic_ref(&generic_ref, action, bytenr,
 					       num_bytes, parent);
-			generic_ref.real_root = root->root_key.objectid;
 			btrfs_init_data_ref(&generic_ref, ref_root, key.objectid,
 					    key.offset, root->root_key.objectid,
 					    for_reloc);
-			generic_ref.skip_qgroup = for_reloc;
 			if (inc)
 				ret = btrfs_inc_extent_ref(trans, &generic_ref);
 			else
@@ -2453,10 +2451,8 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
 			num_bytes = fs_info->nodesize;
 			btrfs_init_generic_ref(&generic_ref, action, bytenr,
 					       num_bytes, parent);
-			generic_ref.real_root = root->root_key.objectid;
 			btrfs_init_tree_ref(&generic_ref, level - 1, ref_root,
 					    root->root_key.objectid, for_reloc);
-			generic_ref.skip_qgroup = for_reloc;
 			if (inc)
 				ret = btrfs_inc_extent_ref(trans, &generic_ref);
 			else
@@ -4936,7 +4932,6 @@ struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
 
 		btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
 				       ins.objectid, ins.offset, parent);
-		generic_ref.real_root = root->root_key.objectid;
 		btrfs_init_tree_ref(&generic_ref, level, root_objectid,
 				    root->root_key.objectid, false);
 		btrfs_ref_tree_mod(fs_info, &generic_ref);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 4fa0792be22d..a980bf37efdc 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4919,7 +4919,6 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 
 			btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF,
 					extent_start, extent_num_bytes, 0);
-			ref.real_root = root->root_key.objectid;
 			btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
 					ino, extent_offset,
 					root->root_key.objectid, false);
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index ce3631760ef4..fed823596248 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -1146,7 +1146,6 @@ int replace_file_extents(struct btrfs_trans_handle *trans,
 		key.offset -= btrfs_file_extent_offset(leaf, fi);
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, new_bytenr,
 				       num_bytes, parent);
-		ref.real_root = root->root_key.objectid;
 		btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
 				    key.objectid, key.offset,
 				    root->root_key.objectid, false);
@@ -1158,7 +1157,6 @@ int replace_file_extents(struct btrfs_trans_handle *trans,
 
 		btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
 				       num_bytes, parent);
-		ref.real_root = root->root_key.objectid;
 		btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
 				    key.objectid, key.offset,
 				    root->root_key.objectid, false);
@@ -1370,7 +1368,6 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, old_bytenr,
 				       blocksize, path->nodes[level]->start);
-		ref.skip_qgroup = true;
 		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
 				    0, true);
 		ret = btrfs_inc_extent_ref(trans, &ref);
@@ -1380,7 +1377,6 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 		}
 		btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, new_bytenr,
 				       blocksize, 0);
-		ref.skip_qgroup = true;
 		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid, 0,
 				    true);
 		ret = btrfs_inc_extent_ref(trans, &ref);
@@ -1393,7 +1389,6 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 				       blocksize, path->nodes[level]->start);
 		btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
 				    0, true);
-		ref.skip_qgroup = true;
 		ret = btrfs_free_extent(trans, &ref);
 		if (ret) {
 			btrfs_abort_transaction(trans, ret);
@@ -1404,7 +1399,6 @@ int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
 				       blocksize, 0);
 		btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid,
 				    0, true);
-		ref.skip_qgroup = true;
 		ret = btrfs_free_extent(trans, &ref);
 		if (ret) {
 			btrfs_abort_transaction(trans, ret);
@@ -2480,7 +2474,6 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 			btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF,
 					       node->eb->start, blocksize,
 					       upper->eb->start);
-			ref.real_root = root->root_key.objectid;
 			btrfs_init_tree_ref(&ref, node->level,
 					    btrfs_header_owner(upper->eb),
 					    root->root_key.objectid, false);
-- 
2.25.1

[PATCH 5/5] btrfs: make real_root optional

[Nikolay Borisov]

Now that real_root is only used in ref-verify core gate it behind
CONFIG_BTRFS_FS_REF_VERIFY ifdef. This shrinks the size of pending
delayed refs by 8 bytes per ref, of which we can have many at any one
time depending on intensity of the workload. Also change the comment
about the member as it no longer deals with qgroups.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/delayed-ref.h | 23 +++++++++--------------
 1 file changed, 9 insertions(+), 14 deletions(-)

diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index 6bb299c66e1e..ddc82caf7b82 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -231,17 +231,10 @@ struct btrfs_ref {
 	 */
 	bool skip_qgroup;
 
-	/*
-	 * Optional. For which root is this modification.
-	 * Mostly used for qgroup optimization.
-	 *
-	 * When unset, data/tree ref init code will populate it.
-	 * In certain cases, we're modifying reference for a different root.
-	 * E.g. COW fs tree blocks for balance.
-	 * In that case, tree_ref::root will be fs tree, but we're doing this
-	 * for reloc tree, then we should set @real_root to reloc tree.
-	 */
+#ifdef CONFIG_BTRFS_FS_REF_VERIFY
+	/* Through which root is this modification. */
 	u64 real_root;
+#endif
 	u64 bytenr;
 	u64 len;
 
@@ -273,9 +266,10 @@ static inline void btrfs_init_generic_ref(struct btrfs_ref *generic_ref,
 static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
 				int level, u64 root, u64 mod_root, bool skip_qgroup)
 {
+#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 	/* If @real_root not set, use @root as fallback */
-	if (!generic_ref->real_root)
-		generic_ref->real_root = root;
+	generic_ref->real_root = mod_root ? mod_root : root;
+#endif
 	generic_ref->tree_ref.level = level;
 	generic_ref->tree_ref.owning_root = root;
 	generic_ref->type = BTRFS_REF_METADATA;
@@ -287,9 +281,10 @@ static inline void btrfs_init_data_ref(struct btrfs_ref *generic_ref,
 				u64 ref_root, u64 ino, u64 offset, u64 mod_root,
 				bool skip_qgroup)
 {
+#ifdef CONFIG_BTRFS_FS_REF_VERIFY
 	/* If @real_root not set, use @root as fallback */
-	if (!generic_ref->real_root)
-		generic_ref->real_root = ref_root;
+	generic_ref->real_root = mod_root ? mod_root : ref_root;
+#endif
 	generic_ref->data_ref.owning_root = ref_root;
 	generic_ref->data_ref.ino = ino;
 	generic_ref->data_ref.offset = offset;
-- 
2.25.1

Re: [PATCH 1/5] btrfs: Rename root fields in delayed refs structs

[David Sterba]

On Mon, Oct 11, 2021 at 01:10:15PM +0300, Nikolay Borisov wrote:
> Both data and metadata delayed ref structures have fields named
> root/ref_root respectively. Those are somewhat cryptic and don't really
> convey the real meaning. In fact those roots are really the original
> owners of the respective block (i.e in case of a snapshot a data delref
> will contain the original root that owns the given block). Rename those
> fields accordingly and adjust comments.

Subject: Re: [PATCH 1/5] btrfs: Rename root fields in delayed refs structs
                                ^

Please use lower case first letter after the "btrfs:" prefix, thanks.

Re: [PATCH 4/5] btrfs: pull up qgroup checks from delayed-ref core to init time

[David Sterba]

On Mon, Oct 11, 2021 at 01:10:18PM +0300, Nikolay Borisov wrote:
> @@ -279,6 +279,8 @@ static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
>  	generic_ref->tree_ref.level = level;
>  	generic_ref->tree_ref.owning_root = root;
>  	generic_ref->type = BTRFS_REF_METADATA;
> +	generic_ref->skip_qgroup = skip_qgroup || !(is_fstree(root) &&
> +				    (!mod_root || is_fstree(mod_root)));

Please avoid overly long conditions when assigning them as expression.
So far we've been doing only simple == or ?:, anything else should be
an if-else.

Re: [PATCH 5/5] btrfs: make real_root optional

[David Sterba]

On Mon, Oct 11, 2021 at 01:10:19PM +0300, Nikolay Borisov wrote:
> @@ -273,9 +266,10 @@ static inline void btrfs_init_generic_ref(struct btrfs_ref *generic_ref,
>  static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
>  				int level, u64 root, u64 mod_root, bool skip_qgroup)
>  {
> +#ifdef CONFIG_BTRFS_FS_REF_VERIFY
>  	/* If @real_root not set, use @root as fallback */
> -	if (!generic_ref->real_root)
> -		generic_ref->real_root = root;
> +	generic_ref->real_root = mod_root ? mod_root : root;

	generic_ref->real_root = mod_root ?: root;

Ie. the short form where the true branch only repeats the condition.

Re: [PATCH 0/5] Make real_root used only in ref-verify

[David Sterba]

On Mon, Oct 11, 2021 at 01:10:14PM +0300, Nikolay Borisov wrote:
> Here's a small series that refactors the way btrfs_ref::real_root and
> btrfs_ref::skip_qgroup are used. Currently the former is used in ref-verify but
> also in order to perform the is_fstree() check on it in delayed-ref core. Given
> the complexity and amount of information that the delayed ref machinery hauls
> around it becomes really non-evident that delrefs really don't care about
> real_root itself but rather only if qgroup processing should happen or not.
> 
> Instead of having the check burried in the core this series changes the data
> flow in such a way that real_root will only be used for ref-verify's operation
> and 'skip_qgroup' will contains the final condition of whether qgroup processing
> should take place for a given delref.
> 
> 
> Nikolay Borisov (5):
>   btrfs: Rename root fields in delayed refs structs
>   btrfs: Rely on owning_root field in btrfs_add_delayed_tree_ref to
>     detect CHUNK_ROOT
>   btrfs: Add additional parameters to
>     btrfs_init_tree_ref/btrfs_init_data_ref
>   btrfs: pull up qgroup checks from delayed-ref core to init time
>   btrfs: make real_root optional

I'll add the branch as-is to for-next so we have some testing coverage,
the suggested changes should not change functionality.

Re: [PATCH 5/5] btrfs: make real_root optional

[Nikolay Borisov]

On 11.10.21 г. 18:05, David Sterba wrote:
> On Mon, Oct 11, 2021 at 01:10:19PM +0300, Nikolay Borisov wrote:
>> @@ -273,9 +266,10 @@ static inline void btrfs_init_generic_ref(struct btrfs_ref *generic_ref,
>>  static inline void btrfs_init_tree_ref(struct btrfs_ref *generic_ref,
>>  				int level, u64 root, u64 mod_root, bool skip_qgroup)
>>  {
>> +#ifdef CONFIG_BTRFS_FS_REF_VERIFY
>>  	/* If @real_root not set, use @root as fallback */
>> -	if (!generic_ref->real_root)
>> -		generic_ref->real_root = root;
>> +	generic_ref->real_root = mod_root ? mod_root : root;
> 
> 	generic_ref->real_root = mod_root ?: root;
> 
> Ie. the short form where the true branch only repeats the condition.

Ah, this is a GNU extension which I had to go an read up on explicitly,
but will keep it in mind :)

>