f2fs: introduce F2FS_UNFAIR_RWSEM to support unfair rwsem

Unfair rwsem should be used when blk-cg is on. Otherwise, there is regression.

FYI, we noticed a -26.7% regression of aim7.jobs-per-min due to commit:

commit: e4544b63a7ee49e7fbebf35ece0a6acd3b9617ae ("f2fs: move f2fs to use reader-unfair rwsems")
https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git master

in testcase: aim7
on test machine: 88 threads 2 sockets Intel(R) Xeon(R) Gold 6238M CPU @ 2.10GHz with 128G memory
with following parameters:

	disk: 4BRD_12G
	md: RAID0
	fs: f2fs
	test: sync_disk_rw
	load: 100
	cpufreq_governor: performance
	ucode: 0x500320a

test-description: AIM7 is a traditional UNIX system level benchmark suite which is used to test and measure the performance of multiuser system.
test-url: https://sourceforge.net/projects/aimbench/files/aim-suite7/

Reported-by: kernel test robot <oliver.sang@intel.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

diff --git a/fs/f2fs/Kconfig b/fs/f2fs/Kconfig
index f46a733..03ef087 100644 (file)
--- a/fs/f2fs/Kconfig
+++ b/fs/f2fs/Kconfig
@@ -143,3 +143,10 @@ config F2FS_IOSTAT
          Support getting IO statistics through sysfs and printing out periodic
          IO statistics tracepoint events. You have to turn on "iostat_enable"
          sysfs node to enable this feature.
+
+config F2FS_UNFAIR_RWSEM
+       bool "F2FS unfair rw_semaphore"
+       depends on F2FS_FS && BLK_CGROUP
+       help
+         Use unfair rw_semaphore, if system configured IO priority by block
+         cgroup.

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index efc4f1f..68d791e 100644 (file)
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -132,7 +132,9 @@ typedef u32 nid_t;
 
 struct f2fs_rwsem {
         struct rw_semaphore internal_rwsem;
+#ifdef CONFIG_F2FS_UNFAIR_RWSEM
         wait_queue_head_t read_waiters;
+#endif
 };
 
 struct f2fs_mount_info {
@@ -2131,7 +2133,9 @@ static inline void __init_f2fs_rwsem(struct f2fs_rwsem *sem,
                const char *sem_name, struct lock_class_key *key)
 {
        __init_rwsem(&sem->internal_rwsem, sem_name, key);
+#ifdef CONFIG_F2FS_UNFAIR_RWSEM
        init_waitqueue_head(&sem->read_waiters);
+#endif
 }
 
 static inline int f2fs_rwsem_is_locked(struct f2fs_rwsem *sem)
@@ -2146,7 +2150,11 @@ static inline int f2fs_rwsem_is_contended(struct f2fs_rwsem *sem)
 
 static inline void f2fs_down_read(struct f2fs_rwsem *sem)
 {
+#ifdef CONFIG_F2FS_UNFAIR_RWSEM
        wait_event(sem->read_waiters, down_read_trylock(&sem->internal_rwsem));
+#else
+       down_read(&sem->internal_rwsem);
+#endif
 }
 
 static inline int f2fs_down_read_trylock(struct f2fs_rwsem *sem)
@@ -2181,7 +2189,9 @@ static inline int f2fs_down_write_trylock(struct f2fs_rwsem *sem)
 static inline void f2fs_up_write(struct f2fs_rwsem *sem)
 {
        up_write(&sem->internal_rwsem);
+#ifdef CONFIG_F2FS_UNFAIR_RWSEM
        wake_up_all(&sem->read_waiters);
+#endif
 }
 
 static inline void f2fs_lock_op(struct f2fs_sb_info *sbi)