diff options
author | Mikulas Patocka <mpatocka@redhat.com> | 2012-10-22 19:37:47 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2012-10-28 13:59:36 -0400 |
commit | 5c1eabe68501d1e1b1586c7f4c46cc531828c4ab (patch) | |
tree | d97aab24d8d0abbd8378e7c851f7a55d266bb851 /include/linux/percpu-rwsem.h | |
parent | e657e078d3dfa9f96976db7a2b5fd7d7c9f1f1a6 (diff) |
percpu-rw-semaphores: use light/heavy barriers
This patch introduces new barrier pair light_mb() and heavy_mb() for
percpu rw semaphores.
This patch fixes a bug in percpu-rw-semaphores where a barrier was
missing in percpu_up_write.
This patch improves performance on the read path of
percpu-rw-semaphores: on non-x86 cpus, there was a smp_mb() in
percpu_up_read. This patch changes it to a compiler barrier and removes
the "#if defined(X86) ..." condition.
From: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Diffstat (limited to 'include/linux/percpu-rwsem.h')
-rw-r--r-- | include/linux/percpu-rwsem.h | 20 |
1 files changed, 7 insertions, 13 deletions
diff --git a/include/linux/percpu-rwsem.h b/include/linux/percpu-rwsem.h index cf80f7e5277f..18f35b54286c 100644 --- a/include/linux/percpu-rwsem.h +++ b/include/linux/percpu-rwsem.h | |||
@@ -12,6 +12,9 @@ struct percpu_rw_semaphore { | |||
12 | struct mutex mtx; | 12 | struct mutex mtx; |
13 | }; | 13 | }; |
14 | 14 | ||
15 | #define light_mb() barrier() | ||
16 | #define heavy_mb() synchronize_sched() | ||
17 | |||
15 | static inline void percpu_down_read(struct percpu_rw_semaphore *p) | 18 | static inline void percpu_down_read(struct percpu_rw_semaphore *p) |
16 | { | 19 | { |
17 | rcu_read_lock(); | 20 | rcu_read_lock(); |
@@ -24,22 +27,12 @@ static inline void percpu_down_read(struct percpu_rw_semaphore *p) | |||
24 | } | 27 | } |
25 | this_cpu_inc(*p->counters); | 28 | this_cpu_inc(*p->counters); |
26 | rcu_read_unlock(); | 29 | rcu_read_unlock(); |
30 | light_mb(); /* A, between read of p->locked and read of data, paired with D */ | ||
27 | } | 31 | } |
28 | 32 | ||
29 | static inline void percpu_up_read(struct percpu_rw_semaphore *p) | 33 | static inline void percpu_up_read(struct percpu_rw_semaphore *p) |
30 | { | 34 | { |
31 | /* | 35 | light_mb(); /* B, between read of the data and write to p->counter, paired with C */ |
32 | * On X86, write operation in this_cpu_dec serves as a memory unlock | ||
33 | * barrier (i.e. memory accesses may be moved before the write, but | ||
34 | * no memory accesses are moved past the write). | ||
35 | * On other architectures this may not be the case, so we need smp_mb() | ||
36 | * there. | ||
37 | */ | ||
38 | #if defined(CONFIG_X86) && (!defined(CONFIG_X86_PPRO_FENCE) && !defined(CONFIG_X86_OOSTORE)) | ||
39 | barrier(); | ||
40 | #else | ||
41 | smp_mb(); | ||
42 | #endif | ||
43 | this_cpu_dec(*p->counters); | 36 | this_cpu_dec(*p->counters); |
44 | } | 37 | } |
45 | 38 | ||
@@ -61,11 +54,12 @@ static inline void percpu_down_write(struct percpu_rw_semaphore *p) | |||
61 | synchronize_rcu(); | 54 | synchronize_rcu(); |
62 | while (__percpu_count(p->counters)) | 55 | while (__percpu_count(p->counters)) |
63 | msleep(1); | 56 | msleep(1); |
64 | smp_rmb(); /* paired with smp_mb() in percpu_sem_up_read() */ | 57 | heavy_mb(); /* C, between read of p->counter and write to data, paired with B */ |
65 | } | 58 | } |
66 | 59 | ||
67 | static inline void percpu_up_write(struct percpu_rw_semaphore *p) | 60 | static inline void percpu_up_write(struct percpu_rw_semaphore *p) |
68 | { | 61 | { |
62 | heavy_mb(); /* D, between write to data and write to p->locked, paired with A */ | ||
69 | p->locked = false; | 63 | p->locked = false; |
70 | mutex_unlock(&p->mtx); | 64 | mutex_unlock(&p->mtx); |
71 | } | 65 | } |