btrfs: zoned: activate new BG only from extent allocation context
cover btrfs: zoned: activate new BG only from extent allocation context
In btrfs_make_block_group(), we activate the allocated block group, expecting that the block group is soon used for the extent allocation.
With a lot of IOs, btrfs_async_reclaim_data_space() tries to prepare for them by pre-allocating data block groups. That preallocation can consume all the active zone counting. It is OK if they are soon written and filled. However, that's not the case. As a result, an allocation of non-data block groups fails due to the lack of an active zone resource.
This series fixes the issue by activating a new block group only when it's called from find_free_extent(). This series introduces CHUNK_ALLOC_FORCE_FOR_EXTENT in btrfs_chunk_alloc_enum to distinguish the context.
Naohiro Aota (2):
btrfs: return allocated block group from do_chunk_alloc()
btrfs: zoned: activate block group only for extent allocation
fs/btrfs/block-group.c | 36 +++++++++++++++++++++++++++--------- fs/btrfs/block-group.h | 1 + fs/btrfs/extent-tree.c | 2 +- 3 files changed, 29 insertions(+), 10 deletions(-)
Signature · 1 line
2.35.1
PATCH 1/2 btrfs: return allocated block group from do_chunk_alloc()
Return the allocated block group from do_chunk_alloc(). This is a preparation patch for the next patch.
--- fs/btrfs/block-group.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-)
fs/btrfs/block-group.c +13 -3
@@ -3433,7 +3433,7 @@ int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type)return btrfs_chunk_alloc (trans, alloc_flags,CHUNK_ALLOC_FORCE ); } -static int do_chunk_alloc (struct btrfs_trans_handle *trans,u64 flags) +static struct btrfs_block_group *do_chunk_alloc (struct btrfs_trans_handle *trans,u64 flags) {struct btrfs_block_group *bg;int ret;
@@ -3520,7 +3520,11 @@ static int do_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags) out:btrfs_trans_release_chunk_metadata (trans); -return ret; +if (ret) +return ERR_PTR (ret); + +btrfs_get_block_group (bg); +return bg; }/*
@@ -3635,6 +3639,7 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, {struct btrfs_fs_info *fs_info = trans->fs_info;struct btrfs_space_info *space_info; +struct btrfs_block_group *ret_bg;bool wait_for_alloc = false;bool should_alloc = false;int ret =0 ;
@@ -3728,9 +3733,14 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,force_metadata_allocation (fs_info); } - ret =do_chunk_alloc (trans, flags); + ret_bg =do_chunk_alloc (trans, flags); trans->allocating_chunk = false; +if (IS_ERR (ret_bg)) + ret =PTR_ERR (ret_bg); +else +btrfs_put_block_group (ret_bg); +spin_lock (&space_info->lock);if (ret <0 ) {if (ret == -ENOSPC )
Signature · 1 line
2.35.1
PATCH 2/2 btrfs: zoned: activate block group only for extent allocation
In reply to Naohiro AotaIn btrfs_make_block_group(), we activate the allocated block group, expecting that the block group is soon used for allocation. However, the chunk allocation from flush_space() context broke the assumption. There can be a large time gap between the chunk allocation time and the extent allocation time from the chunk.
Activating the empty block groups pre-allocated from flush_space() context can exhaust the active zone counter of a device. Once we use all the active zone counts for empty pre-allocated BGs, we cannot activate new BG for the other things: metadata, tree-log, or data relocation BG. That failure results in a fake -ENOSPC.
This patch introduces CHUNK_ALLOC_FORCE_FOR_EXTENT to distinguish the chunk allocation from find_free_extent(). Now, the new block group is activated only in that context.
--- fs/btrfs/block-group.c | 24 ++++++++++++++++-------- fs/btrfs/block-group.h | 1 + fs/btrfs/extent-tree.c | 2 +- 3 files changed, 18 insertions(+), 9 deletions(-)
fs/btrfs/block-group.c +16 -8
@@ -2481,12 +2481,6 @@ struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *tranreturn ERR_PTR (ret); } -/* -* New block group is likely to be used soon. Try to activate it now. -* Failure is OK for now. -*/ -btrfs_zone_activate (cache); - ret =exclude_super_stripes (cache);if (ret) {/* We may have excluded something, so call this just in case */
@@ -3642,8 +3636,14 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,struct btrfs_block_group *ret_bg;bool wait_for_alloc = false;bool should_alloc = false; +bool from_extent_allocation = false;int ret =0 ; +if (force ==CHUNK_ALLOC_FORCE_FOR_EXTENT ) { + from_extent_allocation = true; + force =CHUNK_ALLOC_FORCE ; + } +/* Don't re-enter if we're already allocating a chunk */ if (trans->allocating_chunk)return -ENOSPC ;
@@ -3736,9 +3736,17 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, ret_bg =do_chunk_alloc (trans, flags); trans->allocating_chunk = false; -if (IS_ERR (ret_bg)) +if (IS_ERR (ret_bg)) { ret =PTR_ERR (ret_bg); -else + }else if (!from_extent_allocation) { +/* +* New block group is likely to be used soon. Try to activate +* it now. Failure is OK for now. +*/ +btrfs_zone_activate (ret_bg); + } + +if (!ret)btrfs_put_block_group (ret_bg);spin_lock (&space_info->lock);
fs/btrfs/block-group.h +1 -0
@@ -40,6 +40,7 @@ enum btrfs_chunk_alloc_enum {CHUNK_ALLOC_NO_FORCE ,CHUNK_ALLOC_LIMITED ,CHUNK_ALLOC_FORCE , + CHUNK_ALLOC_FORCE_FOR_EXTENT, }; struct btrfs_caching_control {
fs/btrfs/extent-tree.c +1 -1
@@ -4082,7 +4082,7 @@ static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info, } ret =btrfs_chunk_alloc (trans, ffe_ctl->flags, -CHUNK_ALLOC_FORCE ); +CHUNK_ALLOC_FORCE_FOR_EXTENT );/* Do not bail out on ENOSPC since we can do more. */ if (ret == -ENOSPC )
Signature · 1 line
2.35.1
Re: [PATCH 2/2] btrfs: zoned: activate block group only for extent allocation
Quoted 59 lines · Naohiro Aota
In
btrfs_make_block_group(), we activate the allocated block group, expecting that the block group is soon used for allocation. However, the chunk allocation fromflush_space()context broke the assumption. There can be a large time gap between the chunk allocation time and the extent allocation time from the chunk.Activating the empty block groups pre-allocated from
flush_space()context can exhaust the active zone counter of a device. Once we use all the active zone counts for empty pre-allocated BGs, we cannot activate new BG for the other things: metadata, tree-log, or data relocation BG. That failure results in a fake-ENOSPC.This patch introduces
CHUNK_ALLOC_FORCE_FOR_EXTENTto distinguish the chunk allocation fromfind_free_extent(). Now, the new block group is activated only in that context.Fixes eb66a010d518 ("btrfs: zoned: activate new block group")Cc [email protected] # 5.16+: c7d1b9109dd0: btrfs: return allocated block group from do_chunk_alloc()Cc [email protected] # 5.16+Signed-off-by Naohiro Aota <[email protected]>--- fs/btrfs/block-group.c | 24 ++++++++++++++++-------- fs/btrfs/block-group.h | 1 + fs/btrfs/extent-tree.c | 2 +- 3 files changed, 18 insertions(+), 9 deletions(-)fs/btrfs/block-group.c +6
-6@@ -2481,12 +2481,6 @@ struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *tranreturn ERR_PTR (ret); } -/* -* New block group is likely to be used soon. Try to activate it now. -* Failure is OK for now. -*/ -btrfs_zone_activate (cache); - ret =exclude_super_stripes (cache);if (ret) {/* We may have excluded something, so call this just in case */ @@ -3642,8 +3636,14 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,struct btrfs_block_group *ret_bg;bool wait_for_alloc = false;bool should_alloc = false; +bool from_extent_allocation = false;int ret =0 ; +if (force ==CHUNK_ALLOC_FORCE_FOR_EXTENT ) { + from_extent_allocation = true; + force =CHUNK_ALLOC_FORCE ; + } +/* Don't re-enter if we're already allocating a chunk */ if (trans->allocating_chunk)return -ENOSPC ;
fs/btrfs/block-group.c int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, · line 3741
Quoted 8 lines
fs/btrfs/block-group.c +2
-2@@ -3736,9 +3736,17 @@ int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, ret_bg =do_chunk_alloc (trans, flags); trans->allocating_chunk = false; -if (IS_ERR (ret_bg)) +if (IS_ERR (ret_bg)) { ret =PTR_ERR (ret_bg); -else + }else if (!from_extent_allocation) {
Please ignore this series as this condition is flipped.
Quoted 39 lines
fs/btrfs/block-group.c +8
-0+/* +* New block group is likely to be used soon. Try to activate +* it now. Failure is OK for now. +*/ +btrfs_zone_activate (ret_bg); + } + +if (!ret)btrfs_put_block_group (ret_bg);spin_lock (&space_info->lock);fs/btrfs/block-group.h +1
-0@@ -40,6 +40,7 @@ enum btrfs_chunk_alloc_enum {CHUNK_ALLOC_NO_FORCE ,CHUNK_ALLOC_LIMITED ,CHUNK_ALLOC_FORCE , + CHUNK_ALLOC_FORCE_FOR_EXTENT, }; struct btrfs_caching_control {fs/btrfs/extent-tree.c +1
-1@@ -4082,7 +4082,7 @@ static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info, } ret =btrfs_chunk_alloc (trans, ffe_ctl->flags, -CHUNK_ALLOC_FORCE ); +CHUNK_ALLOC_FORCE_FOR_EXTENT );/* Do not bail out on ENOSPC since we can do more. */ if (ret == -ENOSPC )Signature · 1 line
2.35.1