From: Zygo Blaxell <ce3g8jdj@umail.furryterror.org>
To: Joanne Koong <joannelkoong@gmail.com>
Cc: dsterba@suse.com, boris@bur.io, loemra.dev@gmail.com,
fdmanana@suse.com, wqu@suse.com, linux-btrfs@vger.kernel.org
Subject: Re: [PATCH v1] btrfs: add LOGICAL_INO_V2 flag for avoiding the tree modification log
Date: Tue, 22 Sep 2026 22:28:53 -0400 [thread overview]
Message-ID: <arM5ZbekdkYsCfke@hungrycats.org> (raw)
In-Reply-To: <20260922201258.874070-1-joannelkoong@gmail.com>
On Tue, Sep 22, 2026 at 01:12:57PM -0700, Joanne Koong wrote:
> At Meta, we use the LOGICAL_INO_V2 ioctl to estimate sharing-aware space
> usage. We sample the logical address space and resolve each sample back
> to the inodes which reference it. The measurement is statistical, so
> with enough samples it gives good bounds on real usage.
>
> Currently in the LOGICAL_INO_V2 ioctl path, iterate_extent_inodes()
> (which can walk backrefs either against the current trees or against the
> commit roots) acquires a tree modification log sequence number when
> walking the current trees, since those trees are being modified while
> the walk is in progress. The log is then enabled filesystem wide for as
> long as any such walk is running, and is only disabled again once the
> last walk has finished.
>
> While it is enabled the cost impacts the whole filesystem, not only the
> caller. Modifications of interior nodes of the extent tree and of
> subvolume trees have to be recorded, serialized on a single filesystem
> wide lock. Delayed refs with a sequence number at or above the oldest
> active one can not be run (the run path bails out with -EAGAIN in
> btrfs_run_delayed_refs_for_head()), and metadata refs above it can not
> be merged, so they accumulate and that work is deferred to transaction
> commit.
>
> Currently, userspace has no way to ask for anything cheaper. Whenever a
> transaction is running, the walk uses the tree modification log, even if
> the caller would tolerate results that could be up to one commit interval
> stale.
>
> Give userspace this option by adding BTRFS_LOGICAL_INO_ARGS_COMMIT_ROOT,
> which makes the backref walk use the commit roots and therefore never
> enables the tree modification log. Behavior is unchanged unless the flag
> gets explicitly passed by userspace. iterate_extent_inodes() already
> takes a search_commit_root argument for this, which scrub, send and the
> data relocation warning path all pass as true. The flag simply plumbs it
> through to userspace.
I was about to send the same patch after tracking down some performance
issues. This is the same use case as btdu, and bees goes a lot faster
without the tree mod log as well. Much appreciated!
> Assisted-by: LLM
> Signed-off-by: Joanne Koong <joannelkoong@gmail.com>
> ---
> fs/btrfs/backref.c | 6 ++++--
> fs/btrfs/backref.h | 3 ++-
> fs/btrfs/ioctl.c | 9 +++++++--
> include/uapi/linux/btrfs.h | 10 ++++++++++
> 4 files changed, 23 insertions(+), 5 deletions(-)
>
> diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
> index 1be632c742bd..364ec3ad7d8a 100644
> --- a/fs/btrfs/backref.c
> +++ b/fs/btrfs/backref.c
> @@ -2549,7 +2549,8 @@ static int build_ino_list(u64 inum, u64 offset, u64 num_bytes, u64 root, void *c
> }
>
> int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
> - void *ctx, bool ignore_offset)
> + void *ctx, bool ignore_offset,
> + bool search_commit_root)
> {
> struct btrfs_backref_walk_ctx walk_ctx = { 0 };
> int ret;
> @@ -2575,7 +2576,8 @@ int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
> walk_ctx.extent_item_pos = logical - found_key.objectid;
> walk_ctx.fs_info = fs_info;
>
> - return iterate_extent_inodes(&walk_ctx, false, build_ino_list, ctx);
> + return iterate_extent_inodes(&walk_ctx, search_commit_root,
> + build_ino_list, ctx);
> }
>
> static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
> diff --git a/fs/btrfs/backref.h b/fs/btrfs/backref.h
> index 179791de6b19..f8dc8a3723cf 100644
> --- a/fs/btrfs/backref.h
> +++ b/fs/btrfs/backref.h
> @@ -226,7 +226,8 @@ int iterate_extent_inodes(struct btrfs_backref_walk_ctx *ctx,
> iterate_extent_inodes_t *iterate, void *user_ctx);
>
> int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
> - void *ctx, bool ignore_offset);
> + void *ctx, bool ignore_offset,
> + bool search_commit_root);
>
> int paths_from_inode(u64 inum, struct inode_fs_paths *ipath);
>
> diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
> index 52aab510aea0..63257f8fb897 100644
> --- a/fs/btrfs/ioctl.c
> +++ b/fs/btrfs/ioctl.c
> @@ -3286,6 +3286,7 @@ static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
> struct btrfs_ioctl_logical_ino_args AUTO_KFREE(loi);
> struct btrfs_data_container AUTO_KVFREE(inodes);
> bool ignore_offset;
> + bool search_commit_root;
>
> if (!capable(CAP_SYS_ADMIN))
> return -EPERM;
> @@ -3296,6 +3297,7 @@ static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
>
> if (version == 1) {
> ignore_offset = false;
> + search_commit_root = false;
> size = min_t(u32, loi->size, SZ_64K);
> } else {
> /* All reserved bits must be 0 for now */
> @@ -3303,10 +3305,12 @@ static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
> return -EINVAL;
>
> /* Only accept flags we have defined so far */
> - if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET))
> + if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET |
> + BTRFS_LOGICAL_INO_ARGS_COMMIT_ROOT))
> return -EINVAL;
>
> ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
> + search_commit_root = loi->flags & BTRFS_LOGICAL_INO_ARGS_COMMIT_ROOT;
> size = min_t(u32, loi->size, SZ_16M);
> }
>
> @@ -3314,7 +3318,8 @@ static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
> if (IS_ERR(inodes))
> return PTR_ERR(inodes);
>
> - ret = iterate_inodes_from_logical(loi->logical, fs_info, inodes, ignore_offset);
> + ret = iterate_inodes_from_logical(loi->logical, fs_info, inodes,
> + ignore_offset, search_commit_root);
> if (ret == -EINVAL)
> return -ENOENT;
> if (ret < 0)
> diff --git a/include/uapi/linux/btrfs.h b/include/uapi/linux/btrfs.h
> index 0a13baf3d8d1..b6f196bfa600 100644
> --- a/include/uapi/linux/btrfs.h
> +++ b/include/uapi/linux/btrfs.h
> @@ -732,6 +732,16 @@ struct btrfs_ioctl_logical_ino_args {
> */
> #define BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET (1ULL << 0)
>
> +/*
> + * Resolve backrefs against the commit roots instead of the current trees.
> + *
> + * Resolving against the current trees is more expensive for the filesystem as
> + * a whole, not only for the caller. Callers that do not need to observe the
> + * currently running transaction should set this. The result can be up to one
> + * commit interval stale.
> + */
> +#define BTRFS_LOGICAL_INO_ARGS_COMMIT_ROOT (1ULL << 1)
> +
> enum btrfs_dev_stat_values {
> /* disk I/O failure stats */
> BTRFS_DEV_STAT_WRITE_ERRS, /* EIO or EREMOTEIO from lower layers */
> --
> 2.52.0
>
>
prev parent reply other threads:[~2026-09-23 2:36 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-22 20:12 [PATCH v1] btrfs: add LOGICAL_INO_V2 flag for avoiding the tree modification log Joanne Koong
2026-09-22 21:41 ` Qu Wenruo
2026-09-23 2:28 ` Zygo Blaxell [this message]
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=arM5ZbekdkYsCfke@hungrycats.org \
--to=ce3g8jdj@umail.furryterror.org \
--cc=boris@bur.io \
--cc=dsterba@suse.com \
--cc=fdmanana@suse.com \
--cc=joannelkoong@gmail.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=loemra.dev@gmail.com \
--cc=wqu@suse.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