zephyr/kernel/sys_clock.c
Andy Ross 9098a45c84 kernel: New timeslicing implementation
Instead of checking every time we hit the low-level context switch
path to see if the new thread has a "partner" with which it needs to
share time, just run the slice timer always and reset it from the
scheduler at the points where it has already decided a switch needs to
happen.  In TICKLESS_KERNEL situations, we pay the cost of extra timer
interrupts at ~10Hz or whatever, which is low (note also that this
kind of regular wakeup architecture is required on SMP anyway so the
scheduler can "notice" threads scheduled by other CPUs).  Advantages:

1. Much simpler logic.  Significantly smaller code.  No variance or
   dependence on tickless modes or timer driver (beyond setting a
   simple timeout).

2. No arch-specific assembly integration with _Swap() needed

3. Better performance on many workloads, as the accounting now happens
   at most once per timer interrupt (~5 Hz) and true rescheduling and
   not on every unrelated context switch and interrupt return.

4. It's SMP-safe.  The previous scheme kept the slice ticks as a
   global variable, which was an unnoticed bug.

Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2018-10-16 15:03:10 -04:00

585 lines
14 KiB
C

/* system clock support */
/*
* Copyright (c) 1997-2015 Wind River Systems, Inc.
*
* SPDX-License-Identifier: Apache-2.0
*/
#include <kernel_structs.h>
#include <toolchain.h>
#include <linker/sections.h>
#include <wait_q.h>
#include <drivers/system_timer.h>
#include <syscall_handler.h>
#ifdef CONFIG_SYS_CLOCK_EXISTS
#ifdef _NON_OPTIMIZED_TICKS_PER_SEC
#warning "non-optimized system clock frequency chosen: performance may suffer"
#endif
#endif
#ifdef CONFIG_SYS_CLOCK_EXISTS
static void _handle_expired_timeouts(sys_dlist_t *expired);
#if defined(CONFIG_TIMER_READS_ITS_FREQUENCY_AT_RUNTIME)
int z_clock_hw_cycles_per_sec = CONFIG_SYS_CLOCK_HW_CYCLES_PER_SEC;
#endif
#else
#if defined(CONFIG_TIMER_READS_ITS_FREQUENCY_AT_RUNTIME)
int z_clock_hw_cycles_per_sec;
#endif
#endif
/* Note that this value is 64 bits, and thus non-atomic on almost all
* Zephyr archtictures. And of course it's routinely updated inside
* timer interrupts. Access to it must be locked.
*/
static volatile u64_t tick_count;
u64_t z_last_tick_announced;
#ifdef CONFIG_TICKLESS_KERNEL
/*
* If this flag is set, system clock will run continuously even if
* there are no timer events programmed. This allows using the
* system clock to track passage of time without interruption.
* To save power, this should be turned on only when required.
*/
int _sys_clock_always_on = 1;
static u32_t next_ts;
#endif
u32_t z_tick_get_32(void)
{
#ifdef CONFIG_TICKLESS_KERNEL
return (u32_t)z_clock_uptime();
#else
return (u32_t)tick_count;
#endif
}
u32_t _impl_k_uptime_get_32(void)
{
#ifdef CONFIG_TICKLESS_KERNEL
__ASSERT(_sys_clock_always_on,
"Call k_enable_sys_clock_always_on to use clock API");
#endif
return __ticks_to_ms(z_tick_get_32());
}
#ifdef CONFIG_USERSPACE
Z_SYSCALL_HANDLER(k_uptime_get_32)
{
#ifdef CONFIG_TICKLESS_KERNEL
Z_OOPS(Z_SYSCALL_VERIFY(_sys_clock_always_on));
#endif
return _impl_k_uptime_get_32();
}
#endif
s64_t z_tick_get(void)
{
#ifdef CONFIG_TICKLESS_KERNEL
return z_clock_uptime();
#else
unsigned int key = irq_lock();
s64_t ret = tick_count;
irq_unlock(key);
return ret;
#endif
}
void z_tick_set(s64_t val)
{
unsigned int key = irq_lock();
__ASSERT_NO_MSG(val > tick_count);
__ASSERT_NO_MSG(val > z_last_tick_announced);
tick_count = val;
irq_unlock(key);
}
s64_t _impl_k_uptime_get(void)
{
#ifdef CONFIG_TICKLESS_KERNEL
__ASSERT(_sys_clock_always_on,
"Call k_enable_sys_clock_always_on to use clock API");
#endif
return __ticks_to_ms(z_tick_get());
}
#ifdef CONFIG_USERSPACE
Z_SYSCALL_HANDLER(k_uptime_get, ret_p)
{
u64_t *ret = (u64_t *)ret_p;
Z_OOPS(Z_SYSCALL_MEMORY_WRITE(ret, sizeof(*ret)));
*ret = _impl_k_uptime_get();
return 0;
}
#endif
s64_t k_uptime_delta(s64_t *reftime)
{
s64_t uptime, delta;
uptime = k_uptime_get();
delta = uptime - *reftime;
*reftime = uptime;
return delta;
}
u32_t k_uptime_delta_32(s64_t *reftime)
{
return (u32_t)k_uptime_delta(reftime);
}
/* handle the expired timeouts in the nano timeout queue */
#ifdef CONFIG_SYS_CLOCK_EXISTS
/*
* Handle timeouts by dequeuing the expired ones from _timeout_q and queue
* them on a local one, then doing the real handling from that queue. This
* allows going through the second queue without needing to have the
* interrupts locked since it is a local queue. Each expired timeout is marked
* as _EXPIRED so that an ISR preempting us and releasing an object on which
* a thread was timing out and expired will not give the object to that thread.
*
* Always called from interrupt level, and always only from the system clock
* interrupt.
*/
static inline void handle_timeouts(s32_t ticks)
{
sys_dlist_t expired;
unsigned int key;
/* init before locking interrupts */
sys_dlist_init(&expired);
key = irq_lock();
sys_dnode_t *next = sys_dlist_peek_head(&_timeout_q);
struct _timeout *timeout = (struct _timeout *)next;
K_DEBUG("head: %p, delta: %d\n",
timeout, timeout ? timeout->delta_ticks_from_prev : -2112);
if (next == NULL) {
irq_unlock(key);
return;
}
/*
* Dequeue all expired timeouts from _timeout_q, relieving irq lock
* pressure between each of them, allowing handling of higher priority
* interrupts. We know that no new timeout will be prepended in front
* of a timeout which delta is 0, since timeouts of 0 ticks are
* prohibited.
*/
while (next != NULL) {
/*
* In the case where ticks number is greater than the first
* timeout delta of the list, the lag produced by this initial
* difference must also be applied to others timeouts in list
* until it was entirely consumed.
*/
s32_t tmp = timeout->delta_ticks_from_prev;
if (timeout->delta_ticks_from_prev < ticks) {
timeout->delta_ticks_from_prev = 0;
} else {
timeout->delta_ticks_from_prev -= ticks;
}
ticks -= tmp;
next = sys_dlist_peek_next(&_timeout_q, next);
if (timeout->delta_ticks_from_prev == 0) {
sys_dnode_t *node = &timeout->node;
sys_dlist_remove(node);
/*
* Reverse the order that that were queued in the
* timeout_q: timeouts expiring on the same ticks are
* queued in the reverse order, time-wise, that they are
* added to shorten the amount of time with interrupts
* locked while walking the timeout_q. By reversing the
* order _again_ when building the expired queue, they
* end up being processed in the same order they were
* added, time-wise.
*/
sys_dlist_prepend(&expired, node);
timeout->delta_ticks_from_prev = _EXPIRED;
} else {
break;
}
irq_unlock(key);
key = irq_lock();
timeout = (struct _timeout *)next;
}
irq_unlock(key);
_handle_expired_timeouts(&expired);
}
#else
#define handle_timeouts(ticks) do { } while (false)
#endif
/**
*
* @brief Announce ticks to the kernel
*
* This function is only to be called by the system clock timer driver when a
* tick is to be announced to the kernel. It takes care of dequeuing the
* timers that have expired and wake up the threads pending on them.
*
* @return N/A
*/
void z_clock_announce(s32_t ticks)
{
z_last_tick_announced += ticks;
__ASSERT_NO_MSG(z_last_tick_announced >= tick_count);
#ifdef CONFIG_SMP
/* sys_clock timekeeping happens only on the main CPU */
if (_arch_curr_cpu()->id) {
return;
}
#endif
#ifndef CONFIG_TICKLESS_KERNEL
unsigned int key;
K_DEBUG("ticks: %d\n", ticks);
key = irq_lock();
tick_count += ticks;
irq_unlock(key);
#endif
handle_timeouts(ticks);
#ifdef CONFIG_TIMESLICING
z_time_slice(ticks);
#endif
#ifdef CONFIG_TICKLESS_KERNEL
u32_t next_to = _get_next_timeout_expiry();
next_to = next_to == K_FOREVER ? 0 : next_to;
next_to = !next_to || (next_ts
&& next_to) > next_ts ? next_ts : next_to;
if (next_to) {
/* Clears current program if next_to = 0 and remaining > 0 */
int dt = next_to ? next_to : (_sys_clock_always_on ? INT_MAX : K_FOREVER);
z_clock_set_timeout(dt, false);
}
#endif
}
int k_enable_sys_clock_always_on(void)
{
#ifdef CONFIG_TICKLESS_KERNEL
int prev_status = _sys_clock_always_on;
_sys_clock_always_on = 1;
_enable_sys_clock();
return prev_status;
#else
return -ENOTSUP;
#endif
}
void k_disable_sys_clock_always_on(void)
{
#ifdef CONFIG_TICKLESS_KERNEL
_sys_clock_always_on = 0;
#endif
}
#ifdef CONFIG_SYS_CLOCK_EXISTS
extern u64_t z_last_tick_announced;
/* initialize the timeouts part of k_thread when enabled in the kernel */
void _init_timeout(struct _timeout *t, _timeout_func_t func)
{
/*
* Must be initialized here and when dequeueing a timeout so that code
* not dealing with timeouts does not have to handle this, such as when
* waiting forever on a semaphore.
*/
t->delta_ticks_from_prev = _INACTIVE;
/*
* Must be initialized here so that k_wakeup can
* verify the thread is not on a wait queue before aborting a timeout.
*/
t->wait_q = NULL;
/*
* Must be initialized here, so the _handle_one_timeout()
* routine can check if there is a thread waiting on this timeout
*/
t->thread = NULL;
/*
* Function must be initialized before being potentially called.
*/
t->func = func;
/*
* These are initialized when enqueing on the timeout queue:
*
* thread->timeout.node.next
* thread->timeout.node.prev
*/
}
void _init_thread_timeout(struct _thread_base *thread_base)
{
_init_timeout(&thread_base->timeout, NULL);
}
/* remove a thread timing out from kernel object's wait queue */
static inline void _unpend_thread_timing_out(struct k_thread *thread,
struct _timeout *timeout_obj)
{
if (timeout_obj->wait_q) {
_unpend_thread_no_timeout(thread);
thread->base.timeout.wait_q = NULL;
}
}
/*
* Handle one timeout from the expired timeout queue. Removes it from the wait
* queue it is on if waiting for an object; in this case, the return value is
* kept as -EAGAIN, set previously in _Swap().
*/
static inline void _handle_one_expired_timeout(struct _timeout *timeout)
{
struct k_thread *thread = timeout->thread;
unsigned int key = irq_lock();
timeout->delta_ticks_from_prev = _INACTIVE;
K_DEBUG("timeout %p\n", timeout);
if (thread) {
_unpend_thread_timing_out(thread, timeout);
_mark_thread_as_started(thread);
_ready_thread(thread);
irq_unlock(key);
} else {
irq_unlock(key);
if (timeout->func) {
timeout->func(timeout);
}
}
}
/*
* Loop over all expired timeouts and handle them one by one. Should be called
* with interrupts unlocked: interrupts will be locked on each interation only
* for the amount of time necessary.
*/
static void _handle_expired_timeouts(sys_dlist_t *expired)
{
struct _timeout *timeout, *next;
SYS_DLIST_FOR_EACH_CONTAINER_SAFE(expired, timeout, next, node) {
_handle_one_expired_timeout(timeout);
}
}
/* returns _INACTIVE if the timer is not active */
int _abort_timeout(struct _timeout *timeout)
{
if (timeout->delta_ticks_from_prev == _INACTIVE) {
return _INACTIVE;
}
if (!sys_dlist_is_tail(&_timeout_q, &timeout->node)) {
sys_dnode_t *next_node =
sys_dlist_peek_next(&_timeout_q, &timeout->node);
struct _timeout *next = (struct _timeout *)next_node;
next->delta_ticks_from_prev += timeout->delta_ticks_from_prev;
}
sys_dlist_remove(&timeout->node);
timeout->delta_ticks_from_prev = _INACTIVE;
return 0;
}
/* returns _INACTIVE if the timer has already expired */
int _abort_thread_timeout(struct k_thread *thread)
{
return _abort_timeout(&thread->base.timeout);
}
static inline void _dump_timeout(struct _timeout *timeout, int extra_tab)
{
#ifdef CONFIG_KERNEL_DEBUG
char *tab = extra_tab ? "\t" : "";
K_DEBUG("%stimeout %p, prev: %p, next: %p\n"
"%s\tthread: %p, wait_q: %p\n"
"%s\tticks remaining: %d\n"
"%s\tfunction: %p\n",
tab, timeout, timeout->node.prev, timeout->node.next,
tab, timeout->thread, timeout->wait_q,
tab, timeout->delta_ticks_from_prev,
tab, timeout->func);
#endif
}
static inline void _dump_timeout_q(void)
{
#ifdef CONFIG_KERNEL_DEBUG
struct _timeout *timeout;
K_DEBUG("_timeout_q: %p, head: %p, tail: %p\n",
&_timeout_q, _timeout_q.head, _timeout_q.tail);
SYS_DLIST_FOR_EACH_CONTAINER(&_timeout_q, timeout, node) {
_dump_timeout(timeout, 1);
}
#endif
}
/* find the closest deadline in the timeout queue */
s32_t _get_next_timeout_expiry(void)
{
struct _timeout *t = (struct _timeout *)
sys_dlist_peek_head(&_timeout_q);
return t ? t->delta_ticks_from_prev : K_FOREVER;
}
/*
* Add timeout to timeout queue. Record waiting thread and wait queue if any.
*
* Cannot handle timeout == 0 and timeout == K_FOREVER.
*
* If the new timeout is expiring on the same system clock tick as other
* timeouts already present in the _timeout_q, it is be _prepended_ to these
* timeouts. This allows exiting the loop sooner, which is good, since
* interrupts are locked while trying to find the insert point. Note that the
* timeouts are then processed in the _reverse order_ if they expire on the
* same tick.
*
* This should not cause problems to applications, unless they really expect
* two timeouts queued very close to one another to expire in the same order
* they were queued. This could be changed at the cost of potential longer
* interrupt latency.
*
* Must be called with interrupts locked.
*/
void _add_timeout(struct k_thread *thread, struct _timeout *timeout,
_wait_q_t *wait_q, s32_t timeout_in_ticks)
{
__ASSERT(timeout_in_ticks >= 0, "");
timeout->delta_ticks_from_prev = timeout_in_ticks;
timeout->thread = thread;
timeout->wait_q = (sys_dlist_t *)wait_q;
K_DEBUG("before adding timeout %p\n", timeout);
_dump_timeout(timeout, 0);
_dump_timeout_q();
/* If timer is submitted to expire ASAP with
* timeout_in_ticks (duration) as zero value,
* then handle timeout immedately without going
* through timeout queue.
*/
if (!timeout_in_ticks) {
_handle_one_expired_timeout(timeout);
return;
}
s32_t *delta = &timeout->delta_ticks_from_prev;
struct _timeout *in_q;
#ifdef CONFIG_TICKLESS_KERNEL
/*
* If some time has already passed since timer was last
* programmed, then that time needs to be accounted when
* inserting the new timeout. We account for this
* by adding the already elapsed time to the new timeout.
* This is like adding this timout back in history.
*/
u32_t adjusted_timeout;
*delta += (int)(z_clock_uptime() - z_last_tick_announced);
adjusted_timeout = *delta;
#endif
SYS_DLIST_FOR_EACH_CONTAINER(&_timeout_q, in_q, node) {
if (*delta <= in_q->delta_ticks_from_prev) {
in_q->delta_ticks_from_prev -= *delta;
sys_dlist_insert_before(&_timeout_q, &in_q->node,
&timeout->node);
goto inserted;
}
*delta -= in_q->delta_ticks_from_prev;
}
sys_dlist_append(&_timeout_q, &timeout->node);
inserted:
K_DEBUG("after adding timeout %p\n", timeout);
_dump_timeout(timeout, 0);
_dump_timeout_q();
#ifdef CONFIG_TICKLESS_KERNEL
if (adjusted_timeout < _get_next_timeout_expiry()) {
z_clock_set_timeout(adjusted_timeout, false);
}
#endif
}
/*
* Put thread on timeout queue. Record wait queue if any.
*
* Cannot handle timeout == 0 and timeout == K_FOREVER.
*
* Must be called with interrupts locked.
*/
void _add_thread_timeout(struct k_thread *thread,
_wait_q_t *wait_q,
s32_t timeout_in_ticks)
{
_add_timeout(thread, &thread->base.timeout, wait_q, timeout_in_ticks);
}
#endif /* CONFIG_SYS_CLOCK_EXISTS */