linux-btrfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Damien Le Moal <dlemoal@kernel.org>
To: Johannes Thumshirn <jth@kernel.org>, linux-btrfs@vger.kernel.org
Cc: Naohiro Aota <naohiro.aota@wdc.com>,
	Johannes Thumshirn <johannes.thumshirn@wdc.com>
Subject: Re: [PATCH 2/3] btrfs: zoned: get rid of treelog_bg_lock
Date: Tue, 24 Jun 2025 10:33:52 +0900	[thread overview]
Message-ID: <bd2825ee-6210-4fce-9b1c-16fed908f3a1@kernel.org> (raw)
In-Reply-To: <20250623165629.316213-3-jth@kernel.org>

On 6/24/25 1:56 AM, Johannes Thumshirn wrote:
> From: Johannes Thumshirn <johannes.thumshirn@wdc.com>
> 
> Lockstat analysis of benchmark workloads shows a very high contention of
> the treelog_bg_lock. But the treelog_bg_lock only protects a single
> field in 'struct btrfs_fs_info', namely 'u64 treelog_bg'.
> 
> Use READ_ONCE()/WRITE_ONCE() to access 'btrfs_fs_info::treelog_bg'.
> 
> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
> ---
>  fs/btrfs/disk-io.c     |  1 -
>  fs/btrfs/extent-tree.c | 45 +++++++++++-------------------------------
>  fs/btrfs/fs.h          |  1 -
>  fs/btrfs/zoned.c       |  2 +-
>  fs/btrfs/zoned.h       |  7 +++----
>  5 files changed, 15 insertions(+), 41 deletions(-)
> 
> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
> index 24896322376d..a54218717cb4 100644
> --- a/fs/btrfs/disk-io.c
> +++ b/fs/btrfs/disk-io.c
> @@ -2789,7 +2789,6 @@ void btrfs_init_fs_info(struct btrfs_fs_info *fs_info)
>  	spin_lock_init(&fs_info->defrag_inodes_lock);
>  	spin_lock_init(&fs_info->super_lock);
>  	spin_lock_init(&fs_info->unused_bgs_lock);
> -	spin_lock_init(&fs_info->treelog_bg_lock);
>  	spin_lock_init(&fs_info->zone_active_bgs_lock);
>  	rwlock_init(&fs_info->tree_mod_log_lock);
>  	rwlock_init(&fs_info->global_root_lock);
> diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
> index a9bda68a1883..46358a555f78 100644
> --- a/fs/btrfs/extent-tree.c
> +++ b/fs/btrfs/extent-tree.c
> @@ -3809,22 +3809,6 @@ static int do_allocation_clustered(struct btrfs_block_group *block_group,
>  	return find_free_extent_unclustered(block_group, ffe_ctl);
>  }
>  
> -/*
> - * Tree-log block group locking
> - * ============================
> - *
> - * fs_info::treelog_bg_lock protects the fs_info::treelog_bg which
> - * indicates the starting address of a block group, which is reserved only
> - * for tree-log metadata.
> - *
> - * Lock nesting
> - * ============
> - *
> - * space_info::lock
> - *   block_group::lock
> - *     fs_info::treelog_bg_lock
> - */
> -
>  /*
>   * Simple allocator for sequential-only block group. It only allows sequential
>   * allocation. No need to play with trees. This function also reserves the
> @@ -3844,7 +3828,6 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
>  	u64 log_bytenr;
>  	u64 data_reloc_bytenr;
>  	int ret = 0;
> -	bool skip = false;
>  
>  	ASSERT(btrfs_is_zoned(block_group->fs_info));
>  
> @@ -3852,13 +3835,9 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
>  	 * Do not allow non-tree-log blocks in the dedicated tree-log block
>  	 * group, and vice versa.
>  	 */
> -	spin_lock(&fs_info->treelog_bg_lock);
> -	log_bytenr = fs_info->treelog_bg;
> +	log_bytenr = READ_ONCE(fs_info->treelog_bg);
>  	if (log_bytenr && ((ffe_ctl->for_treelog && bytenr != log_bytenr) ||
>  			   (!ffe_ctl->for_treelog && bytenr == log_bytenr)))
> -		skip = true;
> -	spin_unlock(&fs_info->treelog_bg_lock);
> -	if (skip)
>  		return 1;
>  
>  	/*
> @@ -3894,14 +3873,13 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
>  
>  	spin_lock(&space_info->lock);
>  	spin_lock(&block_group->lock);
> -	spin_lock(&fs_info->treelog_bg_lock);
>  
>  	if (ret)
>  		goto out;
>  
>  	ASSERT(!ffe_ctl->for_treelog ||
> -	       block_group->start == fs_info->treelog_bg ||
> -	       fs_info->treelog_bg == 0);
> +	       block_group->start == log_bytenr ||
> +	       log_bytenr == 0);
>  	ASSERT(!ffe_ctl->for_data_reloc ||
>  	       block_group->start == data_reloc_bytenr ||
>  	       data_reloc_bytenr == 0);
> @@ -3917,7 +3895,7 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
>  	 * Do not allow currently using block group to be tree-log dedicated
>  	 * block group.
>  	 */
> -	if (ffe_ctl->for_treelog && !fs_info->treelog_bg &&
> +	if (ffe_ctl->for_treelog && log_bytenr == 0 &&

Similar comment as previous patch. It feels that all the log_bytenr uses should
really be replaced with READ_ONCE(fs_info->treelog_bg). But I am not sure about
the concurrency with that field changing...

>  	    (block_group->used || block_group->reserved)) {
>  		ret = 1;
>  		goto out;
> @@ -3948,8 +3926,8 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
>  		goto out;
>  	}
>  
> -	if (ffe_ctl->for_treelog && !fs_info->treelog_bg)
> -		fs_info->treelog_bg = block_group->start;
> +	if (ffe_ctl->for_treelog && READ_ONCE(fs_info->treelog_bg) == 0)
> +		WRITE_ONCE(fs_info->treelog_bg, block_group->start);
>  
>  	if (ffe_ctl->for_data_reloc) {
>  		if (READ_ONCE(fs_info->data_reloc_bg) == 0)

[...]

> diff --git a/fs/btrfs/zoned.h b/fs/btrfs/zoned.h
> index 6e11533b8e14..c1b3a5c3a799 100644
> --- a/fs/btrfs/zoned.h
> +++ b/fs/btrfs/zoned.h
> @@ -383,14 +383,13 @@ static inline void btrfs_zoned_meta_io_unlock(struct btrfs_fs_info *fs_info)
>  static inline void btrfs_clear_treelog_bg(struct btrfs_block_group *bg)
>  {
>  	struct btrfs_fs_info *fs_info = bg->fs_info;
> +	u64 treelog_bg = READ_ONCE(fs_info->treelog_bg);
>  
>  	if (!btrfs_is_zoned(fs_info))
>  		return;
>  
> -	spin_lock(&fs_info->treelog_bg_lock);
> -	if (fs_info->treelog_bg == bg->start)
> -		fs_info->treelog_bg = 0;
> -	spin_unlock(&fs_info->treelog_bg_lock);
> +	if (treelog_bg == bg->start)

Variable treelog_bg is not needed.

> +		WRITE_ONCE(fs_info->treelog_bg, 0);
>  }
>  
>  static inline void btrfs_zoned_data_reloc_lock(struct btrfs_inode *inode)


-- 
Damien Le Moal
Western Digital Research

  reply	other threads:[~2025-06-24  1:35 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-06-23 16:56 [PATCH 0/3] btrfs: zoned: reduce lock contention in zoned extent allocator Johannes Thumshirn
2025-06-23 16:56 ` [PATCH 1/3] btrfs: zoned: get rid of relocation_bg_lock Johannes Thumshirn
2025-06-24  1:29   ` Damien Le Moal
2025-06-23 16:56 ` [PATCH 2/3] btrfs: zoned: get rid of treelog_bg_lock Johannes Thumshirn
2025-06-24  1:33   ` Damien Le Moal [this message]
2025-06-23 16:56 ` [PATCH 3/3] btrfs: zoned: don't hold space_info lock on zoned allocation Johannes Thumshirn

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=bd2825ee-6210-4fce-9b1c-16fed908f3a1@kernel.org \
    --to=dlemoal@kernel.org \
    --cc=johannes.thumshirn@wdc.com \
    --cc=jth@kernel.org \
    --cc=linux-btrfs@vger.kernel.org \
    --cc=naohiro.aota@wdc.com \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).