zephyr/kernel/poll.c

816 lines
21 KiB
C
Raw Permalink Normal View History

/*
* Copyright (c) 2017 Wind River Systems, Inc.
* Copyright (c) 2023 Arm Limited (or its affiliates). All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*/
/**
* @file
*
* @brief Kernel asynchronous event polling interface.
*
* This polling mechanism allows waiting on multiple events concurrently,
* either events triggered directly, or from kernel objects or other kernel
* constructs.
*/
#include <zephyr/kernel.h>
#include <zephyr/kernel_structs.h>
#include <kernel_internal.h>
#include <wait_q.h>
#include <ksched.h>
#include <zephyr/internal/syscall_handler.h>
#include <zephyr/sys/dlist.h>
#include <zephyr/sys/util.h>
#include <zephyr/sys/__assert.h>
#include <stdbool.h>
/* Single subsystem lock. Locking per-event would be better on highly
* contended SMP systems, but the original locking scheme here is
* subtle (it relies on releasing/reacquiring the lock in areas for
* latency control and it's sometimes hard to see exactly what data is
* "inside" a given critical section). Do the synchronization port
* later as an optimization.
*/
static struct k_spinlock lock;
enum POLL_MODE { MODE_NONE, MODE_POLL, MODE_TRIGGERED };
static int signal_poller(struct k_poll_event *event, uint32_t state);
static int signal_triggered_work(struct k_poll_event *event, uint32_t status);
void k_poll_event_init(struct k_poll_event *event, uint32_t type,
int mode, void *obj)
{
__ASSERT(mode == K_POLL_MODE_NOTIFY_ONLY,
"only NOTIFY_ONLY mode is supported\n");
__ASSERT(type < (BIT(_POLL_NUM_TYPES)), "invalid type\n");
__ASSERT(obj != NULL, "must provide an object\n");
event->poller = NULL;
/* event->tag is left uninitialized: the user will set it if needed */
event->type = type;
event->state = K_POLL_STATE_NOT_READY;
event->mode = mode;
event->unused = 0U;
event->obj = obj;
SYS_PORT_TRACING_FUNC(k_poll_api, event_init, event);
}
/* must be called with interrupts locked */
static inline bool is_condition_met(struct k_poll_event *event, uint32_t *state)
{
switch (event->type) {
case K_POLL_TYPE_SEM_AVAILABLE:
if (k_sem_count_get(event->sem) > 0U) {
*state = K_POLL_STATE_SEM_AVAILABLE;
return true;
}
break;
case K_POLL_TYPE_DATA_AVAILABLE:
if (!k_queue_is_empty(event->queue)) {
*state = K_POLL_STATE_FIFO_DATA_AVAILABLE;
return true;
}
break;
case K_POLL_TYPE_SIGNAL:
if (event->signal->signaled != 0U) {
*state = K_POLL_STATE_SIGNALED;
return true;
}
break;
case K_POLL_TYPE_MSGQ_DATA_AVAILABLE:
if (event->msgq->used_msgs > 0) {
*state = K_POLL_STATE_MSGQ_DATA_AVAILABLE;
return true;
}
break;
#ifdef CONFIG_PIPES
case K_POLL_TYPE_PIPE_DATA_AVAILABLE:
if (k_pipe_read_avail(event->pipe)) {
*state = K_POLL_STATE_PIPE_DATA_AVAILABLE;
return true;
}
#endif /* CONFIG_PIPES */
case K_POLL_TYPE_IGNORE:
break;
default:
__ASSERT(false, "invalid event type (0x%x)\n", event->type);
break;
}
return false;
}
static struct k_thread *poller_thread(struct z_poller *p)
{
return p ? CONTAINER_OF(p, struct k_thread, poller) : NULL;
}
static inline void add_event(sys_dlist_t *events, struct k_poll_event *event,
struct z_poller *poller)
{
struct k_poll_event *pending;
pending = (struct k_poll_event *)sys_dlist_peek_tail(events);
if ((pending == NULL) ||
(z_sched_prio_cmp(poller_thread(pending->poller),
poller_thread(poller)) > 0)) {
sys_dlist_append(events, &event->_node);
return;
}
SYS_DLIST_FOR_EACH_CONTAINER(events, pending, _node) {
if (z_sched_prio_cmp(poller_thread(poller),
poller_thread(pending->poller)) > 0) {
sys_dlist_insert(&pending->_node, &event->_node);
return;
}
}
sys_dlist_append(events, &event->_node);
}
/* must be called with interrupts locked */
static inline void register_event(struct k_poll_event *event,
struct z_poller *poller)
{
switch (event->type) {
case K_POLL_TYPE_SEM_AVAILABLE:
__ASSERT(event->sem != NULL, "invalid semaphore\n");
add_event(&event->sem->poll_events, event, poller);
break;
case K_POLL_TYPE_DATA_AVAILABLE:
__ASSERT(event->queue != NULL, "invalid queue\n");
add_event(&event->queue->poll_events, event, poller);
break;
case K_POLL_TYPE_SIGNAL:
__ASSERT(event->signal != NULL, "invalid poll signal\n");
add_event(&event->signal->poll_events, event, poller);
break;
case K_POLL_TYPE_MSGQ_DATA_AVAILABLE:
__ASSERT(event->msgq != NULL, "invalid message queue\n");
add_event(&event->msgq->poll_events, event, poller);
break;
#ifdef CONFIG_PIPES
case K_POLL_TYPE_PIPE_DATA_AVAILABLE:
__ASSERT(event->pipe != NULL, "invalid pipe\n");
add_event(&event->pipe->poll_events, event, poller);
break;
#endif /* CONFIG_PIPES */
case K_POLL_TYPE_IGNORE:
/* nothing to do */
break;
default:
__ASSERT(false, "invalid event type\n");
break;
}
event->poller = poller;
}
/* must be called with interrupts locked */
static inline void clear_event_registration(struct k_poll_event *event)
{
bool remove_event = false;
event->poller = NULL;
switch (event->type) {
case K_POLL_TYPE_SEM_AVAILABLE:
__ASSERT(event->sem != NULL, "invalid semaphore\n");
remove_event = true;
break;
case K_POLL_TYPE_DATA_AVAILABLE:
__ASSERT(event->queue != NULL, "invalid queue\n");
remove_event = true;
break;
case K_POLL_TYPE_SIGNAL:
__ASSERT(event->signal != NULL, "invalid poll signal\n");
remove_event = true;
break;
case K_POLL_TYPE_MSGQ_DATA_AVAILABLE:
__ASSERT(event->msgq != NULL, "invalid message queue\n");
remove_event = true;
break;
#ifdef CONFIG_PIPES
case K_POLL_TYPE_PIPE_DATA_AVAILABLE:
__ASSERT(event->pipe != NULL, "invalid pipe\n");
remove_event = true;
break;
#endif /* CONFIG_PIPES */
case K_POLL_TYPE_IGNORE:
/* nothing to do */
break;
default:
__ASSERT(false, "invalid event type\n");
break;
}
if (remove_event && sys_dnode_is_linked(&event->_node)) {
sys_dlist_remove(&event->_node);
}
}
/* must be called with interrupts locked */
static inline void clear_event_registrations(struct k_poll_event *events,
int num_events,
k_spinlock_key_t key)
{
while (num_events--) {
clear_event_registration(&events[num_events]);
k_spin_unlock(&lock, key);
key = k_spin_lock(&lock);
}
}
static inline void set_event_ready(struct k_poll_event *event, uint32_t state)
{
event->poller = NULL;
event->state |= state;
}
static inline int register_events(struct k_poll_event *events,
int num_events,
struct z_poller *poller,
bool just_check)
{
int events_registered = 0;
for (int ii = 0; ii < num_events; ii++) {
k_spinlock_key_t key;
uint32_t state;
key = k_spin_lock(&lock);
if (is_condition_met(&events[ii], &state)) {
set_event_ready(&events[ii], state);
poller->is_polling = false;
} else if (!just_check && poller->is_polling) {
register_event(&events[ii], poller);
events_registered += 1;
} else {
/* Event is not one of those identified in is_condition_met()
* catching non-polling events, or is marked for just check,
* or not marked for polling. No action needed.
*/
;
}
k_spin_unlock(&lock, key);
}
return events_registered;
}
static int signal_poller(struct k_poll_event *event, uint32_t state)
{
struct k_thread *thread = poller_thread(event->poller);
__ASSERT(thread != NULL, "poller should have a thread\n");
if (!z_is_thread_pending(thread)) {
return 0;
}
z_unpend_thread(thread);
arch_thread_return_value_set(thread,
state == K_POLL_STATE_CANCELLED ? -EINTR : 0);
if (!z_is_thread_ready(thread)) {
return 0;
}
z_ready_thread(thread);
return 0;
}
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
int z_impl_k_poll(struct k_poll_event *events, int num_events,
k_timeout_t timeout)
{
int events_registered;
k_spinlock_key_t key;
struct z_poller *poller = &_current->poller;
poller->is_polling = true;
poller->mode = MODE_POLL;
__ASSERT(!arch_is_in_isr(), "");
__ASSERT(events != NULL, "NULL events\n");
__ASSERT(num_events >= 0, "<0 events\n");
SYS_PORT_TRACING_FUNC_ENTER(k_poll_api, poll, events);
events_registered = register_events(events, num_events, poller,
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
K_TIMEOUT_EQ(timeout, K_NO_WAIT));
key = k_spin_lock(&lock);
/*
* If we're not polling anymore, it means that at least one event
* condition is met, either when looping through the events here or
* because one of the events registered has had its state changed.
*/
if (!poller->is_polling) {
clear_event_registrations(events, events_registered, key);
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC_EXIT(k_poll_api, poll, events, 0);
return 0;
}
poller->is_polling = false;
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
if (K_TIMEOUT_EQ(timeout, K_NO_WAIT)) {
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC_EXIT(k_poll_api, poll, events, -EAGAIN);
return -EAGAIN;
}
static _wait_q_t wait_q = Z_WAIT_Q_INIT(&wait_q);
int swap_rc = z_pend_curr(&lock, key, &wait_q, timeout);
/*
* Clear all event registrations. If events happen while we're in this
* loop, and we already had one that triggered, that's OK: they will
* end up in the list of events that are ready; if we timed out, and
* events happen while we're in this loop, that is OK as well since
* we've already know the return code (-EAGAIN), and even if they are
* added to the list of events that occurred, the user has to check the
* return code first, which invalidates the whole list of event states.
*/
key = k_spin_lock(&lock);
clear_event_registrations(events, events_registered, key);
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC_EXIT(k_poll_api, poll, events, swap_rc);
return swap_rc;
}
#ifdef CONFIG_USERSPACE
static inline int z_vrfy_k_poll(struct k_poll_event *events,
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
int num_events, k_timeout_t timeout)
{
int ret;
k_spinlock_key_t key;
struct k_poll_event *events_copy = NULL;
uint32_t bounds;
/* Validate the events buffer and make a copy of it in an
* allocated kernel-side buffer.
*/
if (K_SYSCALL_VERIFY(num_events >= 0)) {
ret = -EINVAL;
goto out;
}
if (K_SYSCALL_VERIFY_MSG(!u32_mul_overflow(num_events,
sizeof(struct k_poll_event),
&bounds),
"num_events too large")) {
ret = -EINVAL;
goto out;
}
events_copy = z_thread_malloc(bounds);
if (!events_copy) {
ret = -ENOMEM;
goto out;
}
key = k_spin_lock(&lock);
if (K_SYSCALL_MEMORY_WRITE(events, bounds)) {
k_spin_unlock(&lock, key);
goto oops_free;
}
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
(void)memcpy(events_copy, events, bounds);
k_spin_unlock(&lock, key);
/* Validate what's inside events_copy */
for (int i = 0; i < num_events; i++) {
struct k_poll_event *e = &events_copy[i];
if (K_SYSCALL_VERIFY(e->mode == K_POLL_MODE_NOTIFY_ONLY)) {
ret = -EINVAL;
goto out_free;
}
switch (e->type) {
case K_POLL_TYPE_IGNORE:
break;
case K_POLL_TYPE_SIGNAL:
K_OOPS(K_SYSCALL_OBJ(e->signal, K_OBJ_POLL_SIGNAL));
break;
case K_POLL_TYPE_SEM_AVAILABLE:
K_OOPS(K_SYSCALL_OBJ(e->sem, K_OBJ_SEM));
break;
case K_POLL_TYPE_DATA_AVAILABLE:
K_OOPS(K_SYSCALL_OBJ(e->queue, K_OBJ_QUEUE));
break;
case K_POLL_TYPE_MSGQ_DATA_AVAILABLE:
K_OOPS(K_SYSCALL_OBJ(e->msgq, K_OBJ_MSGQ));
break;
#ifdef CONFIG_PIPES
case K_POLL_TYPE_PIPE_DATA_AVAILABLE:
K_OOPS(K_SYSCALL_OBJ(e->pipe, K_OBJ_PIPE));
break;
#endif /* CONFIG_PIPES */
default:
ret = -EINVAL;
goto out_free;
}
}
ret = k_poll(events_copy, num_events, timeout);
(void)memcpy((void *)events, events_copy, bounds);
out_free:
k_free(events_copy);
out:
return ret;
oops_free:
k_free(events_copy);
K_OOPS(1);
}
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
#include <syscalls/k_poll_mrsh.c>
#endif /* CONFIG_USERSPACE */
/* must be called with interrupts locked */
static int signal_poll_event(struct k_poll_event *event, uint32_t state)
{
struct z_poller *poller = event->poller;
int retcode = 0;
if (poller != NULL) {
if (poller->mode == MODE_POLL) {
retcode = signal_poller(event, state);
} else if (poller->mode == MODE_TRIGGERED) {
retcode = signal_triggered_work(event, state);
} else {
/* Poller is not poll or triggered mode. No action needed.*/
;
}
poller->is_polling = false;
if (retcode < 0) {
return retcode;
}
}
set_event_ready(event, state);
return retcode;
}
void z_handle_obj_poll_events(sys_dlist_t *events, uint32_t state)
{
struct k_poll_event *poll_event;
k_spinlock_key_t key = k_spin_lock(&lock);
poll_event = (struct k_poll_event *)sys_dlist_get(events);
if (poll_event != NULL) {
(void) signal_poll_event(poll_event, state);
}
k_spin_unlock(&lock, key);
}
void z_impl_k_poll_signal_init(struct k_poll_signal *sig)
{
sys_dlist_init(&sig->poll_events);
sig->signaled = 0U;
/* signal->result is left uninitialized */
k_object_init(sig);
SYS_PORT_TRACING_FUNC(k_poll_api, signal_init, sig);
}
#ifdef CONFIG_USERSPACE
static inline void z_vrfy_k_poll_signal_init(struct k_poll_signal *sig)
{
K_OOPS(K_SYSCALL_OBJ_INIT(sig, K_OBJ_POLL_SIGNAL));
z_impl_k_poll_signal_init(sig);
}
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
#include <syscalls/k_poll_signal_init_mrsh.c>
#endif /* CONFIG_USERSPACE */
void z_impl_k_poll_signal_reset(struct k_poll_signal *sig)
{
sig->signaled = 0U;
SYS_PORT_TRACING_FUNC(k_poll_api, signal_reset, sig);
}
void z_impl_k_poll_signal_check(struct k_poll_signal *sig,
unsigned int *signaled, int *result)
{
*signaled = sig->signaled;
*result = sig->result;
SYS_PORT_TRACING_FUNC(k_poll_api, signal_check, sig);
}
#ifdef CONFIG_USERSPACE
void z_vrfy_k_poll_signal_check(struct k_poll_signal *sig,
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
unsigned int *signaled, int *result)
{
K_OOPS(K_SYSCALL_OBJ(sig, K_OBJ_POLL_SIGNAL));
K_OOPS(K_SYSCALL_MEMORY_WRITE(signaled, sizeof(unsigned int)));
K_OOPS(K_SYSCALL_MEMORY_WRITE(result, sizeof(int)));
z_impl_k_poll_signal_check(sig, signaled, result);
}
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
#include <syscalls/k_poll_signal_check_mrsh.c>
#endif /* CONFIG_USERSPACE */
int z_impl_k_poll_signal_raise(struct k_poll_signal *sig, int result)
{
k_spinlock_key_t key = k_spin_lock(&lock);
struct k_poll_event *poll_event;
sig->result = result;
sig->signaled = 1U;
poll_event = (struct k_poll_event *)sys_dlist_get(&sig->poll_events);
if (poll_event == NULL) {
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC(k_poll_api, signal_raise, sig, 0);
return 0;
}
int rc = signal_poll_event(poll_event, K_POLL_STATE_SIGNALED);
SYS_PORT_TRACING_FUNC(k_poll_api, signal_raise, sig, rc);
z_reschedule(&lock, key);
return rc;
}
#ifdef CONFIG_USERSPACE
static inline int z_vrfy_k_poll_signal_raise(struct k_poll_signal *sig,
int result)
{
K_OOPS(K_SYSCALL_OBJ(sig, K_OBJ_POLL_SIGNAL));
return z_impl_k_poll_signal_raise(sig, result);
}
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
#include <syscalls/k_poll_signal_raise_mrsh.c>
static inline void z_vrfy_k_poll_signal_reset(struct k_poll_signal *sig)
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
{
K_OOPS(K_SYSCALL_OBJ(sig, K_OBJ_POLL_SIGNAL));
z_impl_k_poll_signal_reset(sig);
userspace: Support for split 64 bit arguments System call arguments, at the arch layer, are single words. So passing wider values requires splitting them into two registers at call time. This gets even more complicated for values (e.g k_timeout_t) that may have different sizes depending on configuration. This patch adds a feature to gen_syscalls.py to detect functions with wide arguments and automatically generates code to split/unsplit them. Unfortunately the current scheme of Z_SYSCALL_DECLARE_* macros won't work with functions like this, because for N arguments (our current maximum N is 10) there are 2^N possible configurations of argument widths. So this generates the complete functions for each handler and wrapper, effectively doing in python what was originally done in the preprocessor. Another complexity is that traditional the z_hdlr_*() function for a system call has taken the raw list of word arguments, which does not work when some of those arguments must be 64 bit types. So instead of using a single Z_SYSCALL_HANDLER macro, this splits the job of z_hdlr_*() into two steps: An automatically-generated unmarshalling function, z_mrsh_*(), which then calls a user-supplied verification function z_vrfy_*(). The verification function is typesafe, and is a simple C function with exactly the same argument and return signature as the syscall impl function. It is also not responsible for validating the pointers to the extra parameter array or a wide return value, that code gets automatically generated. This commit includes new vrfy/msrh handling for all syscalls invoked during CI runs. Future commits will port the less testable code. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2019-08-06 22:34:31 +02:00
}
#include <syscalls/k_poll_signal_reset_mrsh.c>
#endif /* CONFIG_USERSPACE */
static void triggered_work_handler(struct k_work *work)
{
struct k_work_poll *twork =
CONTAINER_OF(work, struct k_work_poll, work);
/*
* If callback is not set, the k_work_poll_submit_to_queue()
* already cleared event registrations.
*/
if (twork->poller.mode != MODE_NONE) {
k_spinlock_key_t key;
key = k_spin_lock(&lock);
clear_event_registrations(twork->events,
twork->num_events, key);
k_spin_unlock(&lock, key);
}
/* Drop work ownership and execute real handler. */
twork->workq = NULL;
twork->real_handler(work);
}
static void triggered_work_expiration_handler(struct _timeout *timeout)
{
struct k_work_poll *twork =
CONTAINER_OF(timeout, struct k_work_poll, timeout);
twork->poller.is_polling = false;
twork->poll_result = -EAGAIN;
k_work_submit_to_queue(twork->workq, &twork->work);
}
extern int z_work_submit_to_queue(struct k_work_q *queue,
struct k_work *work);
static int signal_triggered_work(struct k_poll_event *event, uint32_t status)
{
struct z_poller *poller = event->poller;
struct k_work_poll *twork =
CONTAINER_OF(poller, struct k_work_poll, poller);
if (poller->is_polling && twork->workq != NULL) {
struct k_work_q *work_q = twork->workq;
z_abort_timeout(&twork->timeout);
twork->poll_result = 0;
z_work_submit_to_queue(work_q, &twork->work);
}
return 0;
}
static int triggered_work_cancel(struct k_work_poll *work,
k_spinlock_key_t key)
{
/* Check if the work waits for event. */
if (work->poller.is_polling && work->poller.mode != MODE_NONE) {
/* Remove timeout associated with the work. */
z_abort_timeout(&work->timeout);
/*
* Prevent work execution if event arrives while we will be
* clearing registrations.
*/
work->poller.mode = MODE_NONE;
/* Clear registrations and work ownership. */
clear_event_registrations(work->events, work->num_events, key);
work->workq = NULL;
return 0;
}
/*
* If we reached here, the work is either being registered in
* the k_work_poll_submit_to_queue(), executed or is pending.
* Only in the last case we have a chance to cancel it, but
* unfortunately there is no public API performing this task.
*/
return -EINVAL;
}
void k_work_poll_init(struct k_work_poll *work,
k_work_handler_t handler)
{
SYS_PORT_TRACING_OBJ_FUNC_ENTER(k_work_poll, init, work);
*work = (struct k_work_poll) {};
k_work_init(&work->work, triggered_work_handler);
work->real_handler = handler;
z_init_timeout(&work->timeout);
SYS_PORT_TRACING_OBJ_FUNC_EXIT(k_work_poll, init, work);
}
int k_work_poll_submit_to_queue(struct k_work_q *work_q,
struct k_work_poll *work,
struct k_poll_event *events,
int num_events,
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
k_timeout_t timeout)
{
int events_registered;
k_spinlock_key_t key;
__ASSERT(work_q != NULL, "NULL work_q\n");
__ASSERT(work != NULL, "NULL work\n");
__ASSERT(events != NULL, "NULL events\n");
__ASSERT(num_events > 0, "zero events\n");
SYS_PORT_TRACING_FUNC_ENTER(k_work_poll, submit_to_queue, work_q, work, timeout);
/* Take ownership of the work if it is possible. */
key = k_spin_lock(&lock);
if (work->workq != NULL) {
if (work->workq == work_q) {
int retval;
retval = triggered_work_cancel(work, key);
if (retval < 0) {
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC_EXIT(k_work_poll, submit_to_queue, work_q,
work, timeout, retval);
return retval;
}
} else {
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC_EXIT(k_work_poll, submit_to_queue, work_q,
work, timeout, -EADDRINUSE);
return -EADDRINUSE;
}
}
work->poller.is_polling = true;
work->workq = work_q;
work->poller.mode = MODE_NONE;
k_spin_unlock(&lock, key);
/* Save list of events. */
work->events = events;
work->num_events = num_events;
/* Clear result */
work->poll_result = -EINPROGRESS;
/* Register events */
events_registered = register_events(events, num_events,
&work->poller, false);
key = k_spin_lock(&lock);
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
if (work->poller.is_polling && !K_TIMEOUT_EQ(timeout, K_NO_WAIT)) {
/*
* Poller is still polling.
* No event is ready and all are watched.
*/
__ASSERT(num_events == events_registered,
"Some events were not registered!\n");
/* Setup timeout if such action is requested */
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
if (!K_TIMEOUT_EQ(timeout, K_FOREVER)) {
z_add_timeout(&work->timeout,
triggered_work_expiration_handler,
kernel/timeout: Make timeout arguments an opaque type Add a k_timeout_t type, and use it everywhere that kernel API functions were accepting a millisecond timeout argument. Instead of forcing milliseconds everywhere (which are often not integrally representable as system ticks), do the conversion to ticks at the point where the timeout is created. This avoids an extra unit conversion in some application code, and allows us to express the timeout in units other than milliseconds to achieve greater precision. The existing K_MSEC() et. al. macros now return initializers for a k_timeout_t. The K_NO_WAIT and K_FOREVER constants have now become k_timeout_t values, which means they cannot be operated on as integers. Applications which have their own APIs that need to inspect these vs. user-provided timeouts can now use a K_TIMEOUT_EQ() predicate to test for equality. Timer drivers, which receive an integer tick count in ther z_clock_set_timeout() functions, now use the integer-valued K_TICKS_FOREVER constant instead of K_FOREVER. For the initial release, to preserve source compatibility, a CONFIG_LEGACY_TIMEOUT_API kconfig is provided. When true, the k_timeout_t will remain a compatible 32 bit value that will work with any legacy Zephyr application. Some subsystems present timeout (or timeout-like) values to their own users as APIs that would re-use the kernel's own constants and conventions. These will require some minor design work to adapt to the new scheme (in most cases just using k_timeout_t directly in their own API), and they have not been changed in this patch, instead selecting CONFIG_LEGACY_TIMEOUT_API via kconfig. These subsystems include: CAN Bus, the Microbit display driver, I2S, LoRa modem drivers, the UART Async API, Video hardware drivers, the console subsystem, and the network buffer abstraction. k_sleep() now takes a k_timeout_t argument, with a k_msleep() variant provided that works identically to the original API. Most of the changes here are just type/configuration management and documentation, but there are logic changes in mempool, where a loop that used a timeout numerically has been reworked using a new z_timeout_end_calc() predicate. Also in queue.c, a (when POLL was enabled) a similar loop was needlessly used to try to retry the k_poll() call after a spurious failure. But k_poll() does not fail spuriously, so the loop was removed. Signed-off-by: Andy Ross <andrew.j.ross@intel.com>
2020-03-06 00:18:14 +01:00
timeout);
}
/* From now, any event will result in submitted work. */
work->poller.mode = MODE_TRIGGERED;
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC_EXIT(k_work_poll, submit_to_queue, work_q, work, timeout, 0);
return 0;
}
/*
* The K_NO_WAIT timeout was specified or at least one event
* was ready at registration time or changed state since
* registration. Hopefully, the poller mode was not set, so
* work was not submitted to workqueue.
*/
/*
* If poller is still polling, no watched event occurred. This means
* we reached here due to K_NO_WAIT timeout "expiration".
*/
if (work->poller.is_polling) {
work->poller.is_polling = false;
work->poll_result = -EAGAIN;
} else {
work->poll_result = 0;
}
/* Clear registrations. */
clear_event_registrations(events, events_registered, key);
k_spin_unlock(&lock, key);
/* Submit work. */
k_work_submit_to_queue(work_q, &work->work);
SYS_PORT_TRACING_FUNC_EXIT(k_work_poll, submit_to_queue, work_q, work, timeout, 0);
return 0;
}
int k_work_poll_submit(struct k_work_poll *work,
struct k_poll_event *events,
int num_events,
k_timeout_t timeout)
{
SYS_PORT_TRACING_FUNC_ENTER(k_work_poll, submit, work, timeout);
int ret = k_work_poll_submit_to_queue(&k_sys_work_q, work,
events, num_events, timeout);
SYS_PORT_TRACING_FUNC_EXIT(k_work_poll, submit, work, timeout, ret);
return ret;
}
int k_work_poll_cancel(struct k_work_poll *work)
{
k_spinlock_key_t key;
int retval;
SYS_PORT_TRACING_FUNC_ENTER(k_work_poll, cancel, work);
/* Check if the work was submitted. */
if (work == NULL || work->workq == NULL) {
SYS_PORT_TRACING_FUNC_EXIT(k_work_poll, cancel, work, -EINVAL);
return -EINVAL;
}
key = k_spin_lock(&lock);
retval = triggered_work_cancel(work, key);
k_spin_unlock(&lock, key);
SYS_PORT_TRACING_FUNC_EXIT(k_work_poll, cancel, work, retval);
return retval;
}