michaelh/zephyr
1
0
Fork 0
Code Releases Activity

kernel/k_sem: Spinlockify

Browse source 
Switch semaphores to use a subsystem spinlock instead of the system
irqlock.

Note that this is only "half way there".  Semaphores will no longer
contend with other irqlock users on SMP systems, but all semaphores
are still sharing the same lock.  Really we want semaphores to be
independently synchronized, but adding 4 bytes to every one (there are
a LOT of these things) for a separate spinlock is too much to pay.

Rather, a proper SMP-aware implementation would spin on the count
variable directly.  But let's not rock that boat quite yet.

Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
This commit is contained in:
Andy Ross 2018-07-24 14:12:36 -07:00 • committed by Anas Nashif
commit da37a53a54
1 changed files with 16 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

22
kernel/sem.c
View file

@ -32,6 +32,15 @@
extern struct k_sem _k_sem_list_start[];
extern struct k_sem _k_sem_list_end[];
/* We use a system-wide lock to synchronize semaphores, which has
* unfortunate performance impact vs. using a per-object lock
* (semaphores are *very* widely used). But per-object locks require
* significant extra RAM. A properly spin-aware semaphore
* implementation would spin on atomic access to the count variable,
* and not a spinlock per se. Useful optimization for the future...
*/
static struct k_spinlock lock;
#ifdef CONFIG_OBJECT_TRACING
struct k_sem *_trace_list_k_sem;
@ -114,12 +123,12 @@ static void do_sem_give(struct k_sem *sem)
void _impl_k_sem_give(struct k_sem *sem)
{
u32_t key = irq_lock();
k_spinlock_key_t key = k_spin_lock(&lock);
sys_trace_void(SYS_TRACE_ID_SEMA_GIVE);
do_sem_give(sem);
sys_trace_end_call(SYS_TRACE_ID_SEMA_GIVE);
_reschedule_irqlock(key);
_reschedule(&lock, key);
}
#ifdef CONFIG_USERSPACE
@ -131,24 +140,25 @@ int _impl_k_sem_take(struct k_sem *sem, s32_t timeout)
__ASSERT(((_is_in_isr() == false) || (timeout == K_NO_WAIT)), "");
sys_trace_void(SYS_TRACE_ID_SEMA_TAKE);
u32_t key = irq_lock();
k_spinlock_key_t key = k_spin_lock(&lock);
if (likely(sem->count > 0U)) {
sem->count--;
irq_unlock(key);
k_spin_unlock(&lock, key);
sys_trace_end_call(SYS_TRACE_ID_SEMA_TAKE);
return 0;
}
if (timeout == K_NO_WAIT) {
irq_unlock(key);
k_spin_unlock(&lock, key);
sys_trace_end_call(SYS_TRACE_ID_SEMA_TAKE);
return -EBUSY;
}
sys_trace_end_call(SYS_TRACE_ID_SEMA_TAKE);
return _pend_curr_irqlock(key, &sem->wait_q, timeout);
int ret = _pend_curr(&lock, key, &sem->wait_q, timeout);
return ret;
}
#ifdef CONFIG_USERSPACE