mbox series

[v2,0/3] fs/file.c: optimize the critical section of file_lock in

Message ID 20240622154904.3774273-1-yu.ma@intel.com (mailing list archive)
Headers show
Series fs/file.c: optimize the critical section of file_lock in | expand

Message

Ma, Yu June 22, 2024, 3:49 p.m. UTC
pts/blogbench-1.1.0 is a benchmark designed to replicate the
load of a real-world busy file server by multiple threads of
random reads, writes, and rewrites. When running default configuration
with multiple parallel threads, hot spin lock contention is observed
from alloc_fd(), file_closed_fd() and put_unused_fd() around file_lock.

These 3 patches are created to reduce the critical section of file_lock
in alloc_fd() and close_fd(). As a result, pts/blogbench-1.1.0 has been
improved by 32% for read and 17% for write with over 30% kernel cycles
reduced on ICX 160 cores configuration with v6.10-rc4.

v1 -> v2:
1. Rebased the patch set to latest v6.10-rc4 and updated the performance
results.
2. Updated patch 1 to resolve the bug identified by Mateusz Guzik with rlimit
check and ensure allocated fd is greater than start.
3. Updated patch 3 to remove sanity_check directly per the alignment with
maintainer

Yu Ma (3):
  fs/file.c: add fast path in alloc_fd()
  fs/file.c: conditionally clear full_fds
  fs/file.c: remove sanity_check from alloc_fd()

 fs/file.c | 46 ++++++++++++++++++++++++----------------------
 1 file changed, 24 insertions(+), 22 deletions(-)


base-commit: 6ba59ff4227927d3a8530fc2973b80e94b54d58f