All of lore.kernel.org
 help / color / mirror / Atom feed
From: Mark Harmstone <mark@harmstone.com>
To: linux-btrfs@vger.kernel.org
Cc: Mark Harmstone <mark@harmstone.com>
Subject: [PATCH v7 16/16] btrfs: populate fully_remapped_bgs_list on mount
Date: Mon, 24 Nov 2025 18:53:08 +0000	[thread overview]
Message-ID: <20251124185335.16556-17-mark@harmstone.com> (raw)
In-Reply-To: <20251124185335.16556-1-mark@harmstone.com>

Add a function btrfs_populate_fully_remapped_bgs_list() which gets
called on mount, which looks for fully remapped block groups
(i.e. identity_remap_count == 0) which haven't yet had their chunk
stripes and device extents removed.

This happens when a filesystem is unmounted while async discard has not
yet finished, as otherwise the data range occupied by the chunk stripes
would be permanently unusable.

Signed-off-by: Mark Harmstone <mark@harmstone.com>
---
 fs/btrfs/block-group.c      | 79 +++++++++++++++++++++++++++++++++++++
 fs/btrfs/block-group.h      |  2 +
 fs/btrfs/disk-io.c          |  9 +++++
 fs/btrfs/free-space-cache.c | 18 +++++++++
 fs/btrfs/relocation.c       |  4 ++
 5 files changed, 112 insertions(+)

diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c
index a6005a54273f..e82a8d8618be 100644
--- a/fs/btrfs/block-group.c
+++ b/fs/btrfs/block-group.c
@@ -4823,6 +4823,11 @@ void btrfs_mark_bg_fully_remapped(struct btrfs_block_group *bg,
 	struct btrfs_fs_info *fs_info = trans->fs_info;
 
 	if (btrfs_test_opt(fs_info, DISCARD_ASYNC)) {
+		spin_lock(&bg->lock);
+		set_bit(BLOCK_GROUP_FLAG_STRIPE_REMOVAL_PENDING,
+			&bg->runtime_flags);
+		spin_unlock(&bg->lock);
+
 		btrfs_discard_queue_work(&fs_info->discard_ctl, bg);
 	} else {
 		spin_lock(&fs_info->unused_bgs_lock);
@@ -4842,3 +4847,77 @@ void btrfs_mark_bg_fully_remapped(struct btrfs_block_group *bg,
 		spin_unlock(&fs_info->unused_bgs_lock);
 	}
 }
+
+/*
+ * Compare the block group and chunk trees, and find any fully-remapped block
+ * groups which haven't yet had their chunk stripes and device extents removed,
+ * and put them on the fully_remapped_bgs list so this gets done.
+ *
+ * This happens when a block group becomes fully remapped, i.e. its last
+ * identity mapping is removed, and the volume is unmounted before async
+ * discard has finished. It's important this gets done as until it is the
+ * chunk's stripes are dead space.
+ */
+int btrfs_populate_fully_remapped_bgs_list(struct btrfs_fs_info *fs_info)
+{
+	struct rb_node *node_bg, *node_chunk;
+
+	node_bg = rb_first_cached(&fs_info->block_group_cache_tree);
+	node_chunk = rb_first_cached(&fs_info->mapping_tree);
+
+	while (node_bg && node_chunk) {
+		struct btrfs_block_group *bg;
+		struct btrfs_chunk_map *map;
+
+		bg = rb_entry(node_bg, struct btrfs_block_group, cache_node);
+		map = rb_entry(node_chunk, struct btrfs_chunk_map, rb_node);
+
+		ASSERT(bg->start == map->start);
+
+		if (!(bg->flags & BTRFS_BLOCK_GROUP_REMAPPED))
+			goto next;
+
+		if (bg->identity_remap_count != 0)
+			goto next;
+
+		if (map->num_stripes == 0)
+			goto next;
+
+		spin_lock(&fs_info->unused_bgs_lock);
+
+		if (list_empty(&bg->bg_list)) {
+			btrfs_get_block_group(bg);
+			list_add_tail(&bg->bg_list,
+				      &fs_info->fully_remapped_bgs);
+		} else {
+			list_move_tail(&bg->bg_list,
+				       &fs_info->fully_remapped_bgs);
+		}
+
+		spin_unlock(&fs_info->unused_bgs_lock);
+
+		/*
+		 * Ideally we'd want to call btrfs_discard_queue_work() here,
+		 * but it'd do nothing as the discard worker hasn't been
+		 * started yet.
+		 *
+		 * The block group will get added to the discard list when
+		 * btrfs_handle_fully_remapped_bgs() gets called, when we
+		 * commit the first transaction.
+		 */
+		if (btrfs_test_opt(fs_info, DISCARD_ASYNC)) {
+			spin_lock(&bg->lock);
+			set_bit(BLOCK_GROUP_FLAG_STRIPE_REMOVAL_PENDING,
+				&bg->runtime_flags);
+			spin_unlock(&bg->lock);
+		}
+
+next:
+		node_bg = rb_next(node_bg);
+		node_chunk = rb_next(node_chunk);
+	}
+
+	ASSERT(!node_bg && !node_chunk);
+
+	return 0;
+}
diff --git a/fs/btrfs/block-group.h b/fs/btrfs/block-group.h
index 743702f4e5d1..bdb75bf1c3f5 100644
--- a/fs/btrfs/block-group.h
+++ b/fs/btrfs/block-group.h
@@ -93,6 +93,7 @@ enum btrfs_block_group_flags {
 	 * transaction.
 	 */
 	BLOCK_GROUP_FLAG_NEW,
+	BLOCK_GROUP_FLAG_STRIPE_REMOVAL_PENDING,
 };
 
 enum btrfs_caching_type {
@@ -416,5 +417,6 @@ int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
 bool btrfs_block_group_should_use_size_class(const struct btrfs_block_group *bg);
 void btrfs_mark_bg_fully_remapped(struct btrfs_block_group *bg,
 				  struct btrfs_trans_handle *trans);
+int btrfs_populate_fully_remapped_bgs_list(struct btrfs_fs_info *fs_info);
 
 #endif /* BTRFS_BLOCK_GROUP_H */
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index e29a6d5ecb2d..7e0e5f6f3223 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -3637,6 +3637,15 @@ int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_device
 		goto fail_sysfs;
 	}
 
+	if (btrfs_fs_incompat(fs_info, REMAP_TREE)) {
+		ret = btrfs_populate_fully_remapped_bgs_list(fs_info);
+		if (ret) {
+			btrfs_err(fs_info,
+			"failed to populate fully_remapped_bgs list: %d", ret);
+			goto fail_sysfs;
+		}
+	}
+
 	btrfs_zoned_reserve_data_reloc_bg(fs_info);
 	btrfs_free_zone_cache(fs_info);
 
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index e15fa8567f7c..7f7744a78de2 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -3068,6 +3068,7 @@ bool btrfs_is_free_space_trimmed(struct btrfs_block_group *block_group)
 	bool ret = true;
 
 	if (block_group->flags & BTRFS_BLOCK_GROUP_REMAPPED &&
+	    !test_bit(BLOCK_GROUP_FLAG_STRIPE_REMOVAL_PENDING, &block_group->runtime_flags) &&
 	    block_group->identity_remap_count == 0) {
 		return true;
 	}
@@ -3849,6 +3850,23 @@ void btrfs_trim_fully_remapped_block_group(struct btrfs_block_group *bg)
 	const u64 max_discard_size = READ_ONCE(discard_ctl->max_discard_size);
 	u64 end = btrfs_block_group_end(bg);
 
+	if (!test_bit(BLOCK_GROUP_FLAG_STRIPE_REMOVAL_PENDING, &bg->runtime_flags)) {
+		bg->discard_cursor = end;
+
+		if (bg->used == 0) {
+			spin_lock(&fs_info->unused_bgs_lock);
+			if (!list_empty(&bg->bg_list)) {
+				list_del_init(&bg->bg_list);
+				btrfs_put_block_group(bg);
+			}
+			spin_unlock(&fs_info->unused_bgs_lock);
+
+			btrfs_mark_bg_unused(bg);
+		}
+
+		return;
+	}
+
 	bytes = end - bg->discard_cursor;
 
 	if (max_discard_size &&
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 2e1a87ab0509..25766b418a6b 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -4792,6 +4792,10 @@ int btrfs_last_identity_remap_gone(struct btrfs_chunk_map *chunk,
 
 	btrfs_remove_bg_from_sinfo(bg);
 
+	spin_lock(&bg->lock);
+	clear_bit(BLOCK_GROUP_FLAG_STRIPE_REMOVAL_PENDING, &bg->runtime_flags);
+	spin_unlock(&bg->lock);
+
 	ret = remove_chunk_stripes(trans, chunk, path);
 	if (ret) {
 		btrfs_abort_transaction(trans, ret);
-- 
2.51.0


  parent reply	other threads:[~2025-11-24 18:53 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-11-24 18:52 [PATCH v7 00/16] Remap tree Mark Harmstone
2025-11-24 18:52 ` [PATCH v7 01/16] btrfs: add definitions and constants for remap-tree Mark Harmstone
2025-12-19  1:21   ` David Sterba
2025-11-24 18:52 ` [PATCH v7 02/16] btrfs: add REMAP chunk type Mark Harmstone
2025-12-19  1:27   ` David Sterba
2025-11-24 18:52 ` [PATCH v7 03/16] btrfs: allow remapped chunks to have zero stripes Mark Harmstone
2025-11-24 18:52 ` [PATCH v7 04/16] btrfs: remove remapped block groups from the free-space tree Mark Harmstone
2025-11-24 18:52 ` [PATCH v7 05/16] btrfs: don't add metadata items for the remap tree to the extent tree Mark Harmstone
2025-11-24 18:52 ` [PATCH v7 06/16] btrfs: add extended version of struct block_group_item Mark Harmstone
2025-11-24 18:52 ` [PATCH v7 07/16] btrfs: allow mounting filesystems with remap-tree incompat flag Mark Harmstone
2025-11-24 18:53 ` [PATCH v7 08/16] btrfs: redirect I/O for remapped block groups Mark Harmstone
2025-11-24 18:53 ` [PATCH v7 09/16] btrfs: handle deletions from remapped block group Mark Harmstone
2025-11-25 15:59   ` Boris Burkov
2025-11-25 16:06   ` Boris Burkov
2025-12-19  1:45   ` David Sterba
2025-11-24 18:53 ` [PATCH v7 10/16] btrfs: handle setting up relocation of block group with remap-tree Mark Harmstone
2025-12-19  1:56   ` David Sterba
2026-01-06 12:20     ` Mark Harmstone
2025-11-24 18:53 ` [PATCH v7 11/16] btrfs: move existing remaps before relocating block group Mark Harmstone
2025-12-19  2:07   ` David Sterba
2026-01-06 18:28     ` Mark Harmstone
2025-11-24 18:53 ` [PATCH v7 12/16] btrfs: replace identity remaps with actual remaps when doing relocations Mark Harmstone
2025-11-24 18:53 ` [PATCH v7 13/16] btrfs: add do_remap param to btrfs_discard_extent() Mark Harmstone
2025-11-24 18:53 ` [PATCH v7 14/16] btrfs: allow balancing remap tree Mark Harmstone
2025-12-19  2:12   ` David Sterba
2025-11-24 18:53 ` [PATCH v7 15/16] btrfs: handle discarding fully-remapped block groups Mark Harmstone
2025-11-25 16:02   ` Boris Burkov
2025-11-24 18:53 ` Mark Harmstone [this message]
2025-11-27  8:37   ` [PATCH v7 16/16] btrfs: populate fully_remapped_bgs_list on mount Boris Burkov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20251124185335.16556-17-mark@harmstone.com \
    --to=mark@harmstone.com \
    --cc=linux-btrfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.