[GIT PULL for v6.11] vfs inode

From: Christian Brauner
Date: Fri Jul 12 2024 - 09:58:23 EST


Hey Linus,

/* Summary */
This contains smaller performance improvements to inodes and dentries:

# inode:

- Add rcu based inode lookup variants. They avoid one inode hash lock acquire
in the common case thereby significantly reducing contention. We already
support RCU-based operations but didn't take advantage of them during inode
insertion.

Callers of iget_locked() get the improvement without any code changes.
Callers that need a custom callback can switch to iget5_locked_rcu() as e.g.,
did btrfs.

With 20 threads each walking a dedicated 1000 dirs * 1000 files directory
tree to stat(2) on a 32 core + 24GB ram vm:

before: 3.54s user 892.30s system 1966% cpu 45.549 total
after: 3.28s user 738.66s system 1955% cpu 37.932 total (-16.7%)

Long-term we should pick up the effort to introduce more fine-grained locking
and possibly improve on the currently used hash implementation.

- Start zeroing i_state in inode_init_always() instead of doing it in
individual filesystems. This allows us to remove an unneeded lock acquire in
new_inode() and not burden individual filesystems with this.

dcache:

- Move d_lockref out of the area used by RCU lookup to avoid cacheline ping
poing because the embedded name is sharing a cacheline with d_lockref.

- Fix dentry size on 32bit with CONFIG_SMP=y so it does actually end up with
128 bytes in total.

/* Testing */
clang: Debian clang version 16.0.6 (26)
gcc: (Debian 13.2.0-24)

All patches are based on v6.10-rc1 and have been sitting in linux-next.
No build failures or warnings were observed.

/* Conflicts */

Merge conflicts with other trees
================================

[1]: linux-next: manual merge of the vfs-brauner tree with the btrfs tree
https://lore.kernel.org/linux-next/ZnFyeNLLrEcX5_g0@xxxxxxxxxxxxx

Merge conflicts with mainline
=============================

This causes a merge conflict with fs/bcachefs/fs.c. The gist is that
inode->i_state initialization in __bch2_new_inode() isn't needed anymore after
this is merged and that line needs to be deleted from that function.

The following changes since commit 1613e604df0cd359cf2a7fbd9be7a0bcfacfabd0:

Linux 6.10-rc1 (2024-05-26 15:20:12 -0700)

are available in the Git repository at:

git@xxxxxxxxxxxxxxxxxxx:pub/scm/linux/kernel/git/vfs/vfs tags/vfs-6.11.inode

for you to fetch changes up to dc99c0ff53f588bb210b1e8b3314c7581cde68a2:

fs: fix dentry size (2024-07-03 10:34:11 +0200)

Please consider pulling these changes from the signed vfs-6.11.inode tag.

Thanks!
Christian

----------------------------------------------------------------
vfs-6.11.inode

----------------------------------------------------------------
Christian Brauner (1):
fs: fix dentry size

Mateusz Guzik (7):
vfs: add rcu-based find_inode variants for iget ops
btrfs: use iget5_locked_rcu
xfs: preserve i_state around inode_init_always in xfs_reinit_inode
vfs: partially sanitize i_state zeroing on inode creation
xfs: remove now spurious i_state initialization in xfs_inode_alloc
bcachefs: remove now spurious i_state initialization
vfs: move d_lockref out of the area used by RCU lookup

fs/bcachefs/fs.c | 1 -
fs/btrfs/inode.c | 2 +-
fs/inode.c | 108 +++++++++++++++++++++++++++++++++++++------------
fs/xfs/xfs_icache.c | 5 ++-
include/linux/dcache.h | 9 ++++-
include/linux/fs.h | 7 +++-
6 files changed, 99 insertions(+), 33 deletions(-)