lkml.org 
[lkml]   [2022]   [Aug]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH V9 04/15] asm-generic: ticket-lock: Keep ticket-lock the same semantic with qspinlock
Date
From: Guo Ren <guoren@linux.alibaba.com>

Define smp_mb__after_spinlock by smp_mb as default behavior to give RCsc
synchronization point for all architectures. Keep the same semantic with
qspinlock, a acquire (RCpc) synchronization point. More detail, see
include/linux/spinlock.h.

Some architectures could give more robust semantics than smp_mb, eg.
riscv. Some architectures needn't smp_mb__after_spinlock because their
spinlocks have contained an RCsc.

Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
---
include/asm-generic/spinlock.h | 5 +++++
include/asm-generic/ticket_spinlock.h | 18 ++++--------------
2 files changed, 9 insertions(+), 14 deletions(-)

diff --git a/include/asm-generic/spinlock.h b/include/asm-generic/spinlock.h
index 970590baf61b..6f5a1b838ca2 100644
--- a/include/asm-generic/spinlock.h
+++ b/include/asm-generic/spinlock.h
@@ -6,4 +6,9 @@
#include <asm-generic/ticket_spinlock.h>
#include <asm/qrwlock.h>

+/* See include/linux/spinlock.h */
+#ifndef smp_mb__after_spinlock
+#define smp_mb__after_spinlock() smp_mb()
+#endif
+
#endif /* __ASM_GENERIC_SPINLOCK_H */
diff --git a/include/asm-generic/ticket_spinlock.h b/include/asm-generic/ticket_spinlock.h
index cfcff22b37b3..d8e6ec82f096 100644
--- a/include/asm-generic/ticket_spinlock.h
+++ b/include/asm-generic/ticket_spinlock.h
@@ -14,9 +14,8 @@
* a test-and-set.
*
* It further assumes atomic_*_release() + atomic_*_acquire() is RCpc and hence
- * uses atomic_fetch_add() which is RCsc to create an RCsc hot path, along with
- * a full fence after the spin to upgrade the otherwise-RCpc
- * atomic_cond_read_acquire().
+ * uses smp_mb__after_spinlock which is RCsc to create an RCsc hot path, See
+ * include/linux/spinlock.h
*
* The implementation uses smp_cond_load_acquire() to spin, so if the
* architecture has WFE like instructions to sleep instead of poll for word
@@ -32,22 +31,13 @@

static __always_inline void ticket_spin_lock(arch_spinlock_t *lock)
{
- u32 val = atomic_fetch_add(1<<16, &lock->val);
+ u32 val = atomic_fetch_add_acquire(1<<16, &lock->val);
u16 ticket = val >> 16;

if (ticket == (u16)val)
return;

- /*
- * atomic_cond_read_acquire() is RCpc, but rather than defining a
- * custom cond_read_rcsc() here we just emit a full fence. We only
- * need the prior reads before subsequent writes ordering from
- * smb_mb(), but as atomic_cond_read_acquire() just emits reads and we
- * have no outstanding writes due to the atomic_fetch_add() the extra
- * orderings are free.
- */
atomic_cond_read_acquire(&lock->val, ticket == (u16)VAL);
- smp_mb();
}

static __always_inline bool ticket_spin_trylock(arch_spinlock_t *lock)
@@ -57,7 +47,7 @@ static __always_inline bool ticket_spin_trylock(arch_spinlock_t *lock)
if ((old >> 16) != (old & 0xffff))
return false;

- return atomic_try_cmpxchg(&lock->val, &old, old + (1<<16)); /* SC, for RCsc */
+ return atomic_try_cmpxchg_acquire(&lock->val, &old, old + (1<<16));
}

static __always_inline void ticket_spin_unlock(arch_spinlock_t *lock)
--
2.36.1
\
 
 \ /
  Last update: 2022-08-08 09:16    [W:0.067 / U:0.072 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site