Skip to contentSkip to conversation

btrfs: zoned: reserve data_reloc block group on mount

PATCH In review Johannes Thumshirn 3 participants 3 messages Jun 4, 2025 – Jun 5, 2025 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH v5 1/1 btrfs: zoned: reserve data_reloc block group on mount

From: Johannes Thumshirn <[email protected]>

Create a block group dedicated for data relocation on mount of a zoned filesystem.

If there is already more than one empty DATA block group on mount, this one is picked for the data relocation block group, instead of a newly created one.

This is done to ensure, there is always space for performing garbage collection and the filesystem is not hitting ENOSPC under heavy overwrite workloads.

---
Changes to v4:
- Use btrfs_commit_transaction() instead of btrfs_end_transaction

 fs/btrfs/disk-io.c |  1 +
 fs/btrfs/zoned.c   | 61 ++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/zoned.h   |  3 +++
 3 files changed, 65 insertions(+)
fs/btrfs/disk-io.c +1 -0
index 3def93016963..b211dc8cdb86 100644
@@ -3562,6 +3562,7 @@ int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_device
 		goto fail_sysfs;
 	}
 
+	btrfs_zoned_reserve_data_reloc_bg(fs_info);
 	btrfs_free_zone_cache(fs_info);
 
 	btrfs_check_active_zone_reservation(fs_info);
fs/btrfs/zoned.c +61 -0

73 lines not on this page · raw message

fs/btrfs/zoned.h +3 -0
index 9672bf4c3335..6e11533b8e14 100644
@@ -88,6 +88,7 @@ void btrfs_zone_finish_endio(struct btrfs_fs_info *fs_info, u64 logical,
 void btrfs_schedule_zone_finish_bg(struct btrfs_block_group *bg,
 				   struct extent_buffer *eb);
 void btrfs_clear_data_reloc_bg(struct btrfs_block_group *bg);
+void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info);
 void btrfs_free_zone_cache(struct btrfs_fs_info *fs_info);
 bool btrfs_zoned_should_reclaim(const struct btrfs_fs_info *fs_info);
 void btrfs_zoned_release_data_reloc_bg(struct btrfs_fs_info *fs_info, u64 logical,
@@ -241,6 +242,8 @@ static inline void btrfs_schedule_zone_finish_bg(struct btrfs_block_group *bg,
 
 static inline void btrfs_clear_data_reloc_bg(struct btrfs_block_group *bg) { }
 
+static inline void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info) { }
+
 static inline void btrfs_free_zone_cache(struct btrfs_fs_info *fs_info) { }
 
 static inline bool btrfs_zoned_should_reclaim(const struct btrfs_fs_info *fs_info)
Signature · 1 line
2.49.0
#1 lore raw
Quoted 95 lines · Johannes Thumshirn

From: Johannes Thumshirn <[email protected]>

Create a block group dedicated for data relocation on mount of a zoned filesystem.

If there is already more than one empty DATA block group on mount, this one is picked for the data relocation block group, instead of a newly created one.

This is done to ensure, there is always space for performing garbage collection and the filesystem is not hitting ENOSPC under heavy overwrite workloads.

---
Changes to v4:
- Use btrfs_commit_transaction() instead of btrfs_end_transaction

 fs/btrfs/disk-io.c |  1 +
 fs/btrfs/zoned.c   | 61 ++++++++++++++++++++++++++++++++++++++++++++++
 fs/btrfs/zoned.h   |  3 +++
 3 files changed, 65 insertions(+)
fs/btrfs/disk-io.c +1 -0
index 3def93016963..b211dc8cdb86 100644
@@ -3562,6 +3562,7 @@ int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_device
                goto fail_sysfs;
        }
 
+       btrfs_zoned_reserve_data_reloc_bg(fs_info);
        btrfs_free_zone_cache(fs_info);
 
        btrfs_check_active_zone_reservation(fs_info);
fs/btrfs/zoned.c +44 -0
index 19710634d63f..4e122d6c19c0 100644
@@ -17,6 +17,7 @@
 #include "fs.h"
 #include "accessors.h"
 #include "bio.h"
+#include "transaction.h"
 
 /* Maximum number of zones to report per blkdev_report_zones() call */
 #define BTRFS_REPORT_NR_ZONES   4096
@@ -2443,6 +2444,66 @@ void btrfs_clear_data_reloc_bg(struct btrfs_block_group *bg)
        spin_unlock(&fs_info->relocation_bg_lock);
 }
 
+void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
+{
+       struct btrfs_space_info *data_sinfo = fs_info->data_sinfo;
+       struct btrfs_space_info *space_info = data_sinfo->sub_group[0];
+       struct btrfs_trans_handle *trans;
+       struct btrfs_block_group *bg;
+       struct list_head *bg_list;
+       u64 alloc_flags;
+       bool initial = false;
+       bool did_chunk_alloc = false;
+       int index;
+       int ret;
+
+       if (!btrfs_is_zoned(fs_info))
+               return;
+
+       if (fs_info->data_reloc_bg)
+               return;
+
+       if (sb_rdonly(fs_info->sb))
+               return;
+
+       ASSERT(space_info->subgroup_id == BTRFS_SUB_GROUP_DATA_RELOC);
+       alloc_flags = btrfs_get_alloc_profile(fs_info, space_info->flags);
+       index = btrfs_bg_flags_to_raid_index(alloc_flags);
+
+       bg_list = &data_sinfo->block_groups[index];
+again:
+       list_for_each_entry(bg, bg_list, list) {
+               if (bg->used > 0)
+                       continue;
+
+               if (!initial) {
+                       initial = true;
+                       continue;
+               }
+
+               fs_info->data_reloc_bg = bg->start;
+               set_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC, &bg->runtime_flags);
+               btrfs_zone_activate(bg);
+
+               return;
+       }

fs/btrfs/zoned.c void btrfs_clear_data_reloc_bg(struct btrfs_block_group *bg) · line 2499

Quoted 10 lines
fs/btrfs/zoned.c +10 -0
+
+       if (did_chunk_alloc)
+               return;
+
+       trans = btrfs_join_transaction(fs_info->tree_root);
+       if (IS_ERR(trans))
+               return;
+
+       ret = btrfs_chunk_alloc(trans, space_info, alloc_flags, CHUNK_ALLOC_FORCE);
+       btrfs_commit_transaction(trans);

Ok so the commit makes a difference and I suppose it fixes the zoned specific corruption you mentioned before.

Can we please get a comment here that explains why it's needed? Because normally we don't need to do it, it's enough to call btrfs_end_transaction() and anyone is able to use the new block group.

Thanks.

Quoted 35 lines
fs/btrfs/zoned.c +7 -0
+       if (ret == 1) {
+               did_chunk_alloc = true;
+               bg_list = &space_info->block_groups[index];
+               goto again;
+       }
+}
+
 void btrfs_free_zone_cache(struct btrfs_fs_info *fs_info)
 {
        struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
fs/btrfs/zoned.h +3 -1
index 9672bf4c3335..6e11533b8e14 100644
@@ -88,6 +88,7 @@ void btrfs_zone_finish_endio(struct btrfs_fs_info *fs_info, u64 logical,
 void btrfs_schedule_zone_finish_bg(struct btrfs_block_group *bg,
                                   struct extent_buffer *eb);
 void btrfs_clear_data_reloc_bg(struct btrfs_block_group *bg);
+void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info);
 void btrfs_free_zone_cache(struct btrfs_fs_info *fs_info);
 bool btrfs_zoned_should_reclaim(const struct btrfs_fs_info *fs_info);
 void btrfs_zoned_release_data_reloc_bg(struct btrfs_fs_info *fs_info, u64 logical,
@@ -241,6 +242,8 @@ static inline void btrfs_schedule_zone_finish_bg(struct btrfs_block_group *bg,
 
 static inline void btrfs_clear_data_reloc_bg(struct btrfs_block_group *bg) { }
 
+static inline void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info) { }
+
 static inline void btrfs_free_zone_cache(struct btrfs_fs_info *fs_info) { }
 
 static inline bool btrfs_zoned_should_reclaim(const struct btrfs_fs_info *fs_info)
--

2.49.0

#2 lore raw
Quoted 9 lines · Filipe Manana

fs/btrfs/zoned.c void btrfs_clear_data_reloc_bg(struct btrfs_block_group *bg) · line 2499

Quoted 2 lines
fs/btrfs/zoned.c +2 -0
+       ret = btrfs_chunk_alloc(trans, space_info, alloc_flags, CHUNK_ALLOC_FORCE);
+       btrfs_commit_transaction(trans);

Ok so the commit makes a difference and I suppose it fixes the zoned specific corruption you mentioned before.

Can we please get a comment here that explains why it's needed? Because normally we don't need to do it, it's enough to call btrfs_end_transaction() and anyone is able to use the new block group.

With patch

https://lore.kernel.org/linux-btrfs/[email protected]/

applied even v4 works (with a clean test environment). So I think I can go back to v4 but need to have the above applied first (after I've fixed the kbuild error).