/*
* Copyright (c) 2006-2024 RT-Thread Development Team
*
* SPDX-License-Identifier: Apache-2.0
*
* Change Logs:
* Date Author Notes
* 2018-01-26 Bernard Fix pthread_detach issue for a none-joinable
* thread.
* 2019-02-07 Bernard Add _pthread_destroy to release pthread resource.
* 2022-05-10 xiangxistu Modify the recycle logic about resource of pthread.
* 2024-04-15 atwww Modify the recycle logic of TLS in function _pthread_data_destroy,
* make it safe for C++11's thread_local destructors.
*/
#include <rthw.h>
#include <pthread.h>
#include <sched.h>
#include <sys/time.h>
#include "pthread_internal.h"
RT_DEFINE_HW_SPINLOCK(pth_lock);
_pthread_data_t *pth_table[PTHREAD_NUM_MAX] = {NULL};
static int concurrency_level;
_pthread_data_t *_pthread_get_data(pthread_t thread)
{
_pthread_data_t *ptd;
if (thread >= PTHREAD_NUM_MAX) return NULL;
rt_hw_spin_lock(&pth_lock);
ptd = pth_table[thread];
rt_hw_spin_unlock(&pth_lock);
if (ptd && ptd->magic == PTHREAD_MAGIC) return ptd;
return NULL;
}
pthread_t _pthread_data_get_pth(_pthread_data_t *ptd)
{
int index;
rt_hw_spin_lock(&pth_lock);
for (index = 0; index < PTHREAD_NUM_MAX; index ++)
{
if (pth_table[index] == ptd) break;
}
rt_hw_spin_unlock(&pth_lock);
return index;
}
pthread_t _pthread_data_create(void)
{
int index;
_pthread_data_t *ptd = NULL;
ptd = (_pthread_data_t*)rt_malloc(sizeof(_pthread_data_t));
if (!ptd) return PTHREAD_NUM_MAX;
memset(ptd, 0x0, sizeof(_pthread_data_t));
ptd->canceled = 0;
ptd->cancelstate = PTHREAD_CANCEL_DISABLE;
ptd->canceltype = PTHREAD_CANCEL_DEFERRED;
ptd->magic = PTHREAD_MAGIC;
rt_hw_spin_lock(&pth_lock);
for (index = 0; index < PTHREAD_NUM_MAX; index ++)
{
if (pth_table[index] == NULL)
{
pth_table[index] = ptd;
break;
}
}
rt_hw_spin_unlock(&pth_lock);
/* full of pthreads, clean magic and release ptd */
if (index == PTHREAD_NUM_MAX)
{
ptd->magic = 0x0;
rt_free(ptd);
}
return index;
}
static inline void _destroy_item(int index, _pthread_data_t *ptd)
{
extern _pthread_key_data_t _thread_keys[PTHREAD_KEY_MAX];
void *data;
if (_thread_keys[index].is_used)
{
data = ptd->tls[index];
if (data && _thread_keys[index].destructor)
{
_thread_keys[index].destructor(data);
}
}
}
#ifdef RT_USING_CPLUSPLUS11
#define NOT_USE_CXX_TLS -1
#endif
void _pthread_data_destroy(_pthread_data_t *ptd)
{
pthread_t pth;
if (ptd)
{
/* if this thread create the local thread data,
* destruct thread local key
*/
if (ptd->tls != RT_NULL)
{
int index;
#ifdef RT_USING_CPLUSPLUS11
/* If C++11 is enabled and emutls is used,
* destructors of C++ object must be called safely.
*/
extern pthread_key_t emutls_get_pthread_key(void);
pthread_key_t emutls_pthread_key = emutls_get_pthread_key();
if (emutls_pthread_key != NOT_USE_CXX_TLS)
{
/* If execution reaches here, C++ 'thread_local' may be used.
* Destructors of c++ class object must be called before emutls_key_destructor.
*/
int start = ((emutls_pthread_key - 1 + PTHREAD_KEY_MAX) % PTHREAD_KEY_MAX);
int i = 0;
for (index = start; i < PTHREAD_KEY_MAX; index = (index - 1 + PTHREAD_KEY_MAX) % PTHREAD_KEY_MAX, i ++)
{
_destroy_item(index, ptd);
}
}
else
#endif
{
/* If only C TLS is used, that is, POSIX TLS or __Thread_local,
* just iterate the _thread_keys from index 0.
*/
for (index = 0; index < PTHREAD_KEY_MAX; index ++)
{
_destroy_item(index, ptd);
}
}
/* release tls area */
rt_free(ptd->tls);
ptd->tls = RT_NULL;
}
pth = _pthread_data_get_pth(ptd);
/* remove from pthread table */
rt_hw_spin_lock(&pth_lock);
pth_table[pth] = NULL;
rt_hw_spin_unlock(&pth_lock);
/* delete joinable semaphore */
if (ptd->joinable_sem != RT_NULL)
{
rt_sem_delete(ptd->joinable_sem);
ptd->joinable_sem = RT_NULL;
}
/* clean magic */
ptd->magic = 0x0;
/* clear the "ptd->tid->pthread_data" */
ptd->tid->pthread_data = RT_NULL;
/* free ptd */
rt_free(ptd);
}
}
static void _pthread_cleanup(rt_thread_t tid)
{
/* clear cleanup function */
tid->cleanup = RT_NULL;
/* restore tid stack */
rt_free(tid->stack_addr);
/* restore tid control block */
rt_free(tid);
}
static void pthread_entry_stub(void *parameter)
{
void *value;
_pthread_data_t *ptd;
ptd = (_pthread_data_t *)parameter;
/* execute pthread entry */
value = ptd->thread_entry(ptd->thread_parameter);
/* According to "detachstate" to whether or not to recycle resource immediately */
if (ptd->attr.detachstate == PTHREAD_CREATE_JOINABLE)
{
/* set value */
ptd->return_value = value;
rt_sem_release(ptd->joinable_sem);
}
else
{
/* release pthread resource */
_pthread_data_destroy(ptd);
}
}
/**
* @brief Creates a new thread in a POSIX-compliant system.
*
* The `pthread_create` function initializes a new thread in the calling process. The new thread starts execution
* by invoking the function specified by the `start` parameter. The thread runs concurrently with the calling thread.
*
* @param[out] pid
* A pointer to a `pthread_t` object where the ID of the newly created thread will be stored.
* This ID can be used to refer to the thread in subsequent function calls.
*
* @param[in] attr
* A pointer to a `pthread_attr_t` object that specifies attributes for the new thread, or `NULL` for default attributes.
* Default attributes include:
* - Detached state: joinable.
* - Stack size: implementation-defined default.
*
* @param[in] start
* A pointer to the function that the new thread will execute. This function must have the following signature:
* `void *start(void *parameter)`.
*
* @param[in] parameter
* A pointer to data passed as an argument to the `start` function. The meaning and handling of this data is determined
* by the `start` function.
*
* @return
* Returns 0 on success. On failure, a non-zero error code is returned, indicating the error condition:
* - `EAGAIN`: Insufficient resources to create another thread.
* - `EINVAL`: Invalid attributes specified in `attr`.
* - `EPERM`: Insufficient permissions to set the requested attributes.
*
* @note
* It is the caller's responsibility to manage the lifetime of any resources associated with the new thread.
* If the thread is not detached, it must be joined using `pthread_join` to avoid resource leaks.
*
* @see pthread_join, pthread_exit, pthread_attr_init
*/
int pthread_create(pthread_t *pid,
const pthread_attr_t *attr,
void *(*start)(void *), void *parameter)
{
int ret = 0;
void *stack;
char name[RT_NAME_MAX];
static rt_uint16_t pthread_number = 0;
pthread_t pth_id;
_pthread_data_t *ptd;
/* pid shall be provided */
RT_ASSERT(pid != RT_NULL);
/* allocate posix thread data */
pth_id = _pthread_data_create();
if (pth_id == PTHREAD_NUM_MAX)
{
ret = ENOMEM;
goto __exit;
}
/* get pthread data */
ptd = _pthread_get_data(pth_id);
RT_ASSERT(ptd != RT_NULL);
if (attr != RT_NULL)
{
ptd->attr = *attr;
}
else
{
/* use default attribute */
pthread_attr_init(&ptd->attr);
}
if (ptd->attr.stacksize == 0)
{
ret = EINVAL;
goto __exit;
}
rt_snprintf(name, sizeof(name), "pth%02d", pthread_number ++);
/* pthread is a static thread object */
ptd->tid = (rt_thread_t) rt_malloc(sizeof(struct rt_thread));
if (ptd->tid == RT_NULL)
{
ret = ENOMEM;
goto __exit;
}
memset(ptd->tid, 0, sizeof(struct rt_thread));
if (ptd->attr.detachstate == PTHREAD_CREATE_JOINABLE)
{
ptd->joinable_sem = rt_sem_create(name, 0, RT_IPC_FLAG_FIFO);
if (ptd->joinable_sem == RT_NULL)
{
ret = ENOMEM;
goto __exit;
}
}
else
{
ptd->joinable_sem = RT_NULL;
}
/* set parameter */
ptd->thread_entry = start;
ptd->thread_parameter = parameter;
/* stack */
if (ptd->attr.stackaddr == 0)
{
stack = (void *)rt_malloc(ptd->attr.stacksize);
}
else
{
stack = (void *)(ptd->attr.stackaddr);
}
if (stack == RT_NULL)
{
ret = ENOMEM;
goto __exit;
}
/* initial this pthread to system */
if (rt_thread_init(ptd->tid, name, pthread_entry_stub, ptd,
stack, ptd->attr.stacksize,
ptd->attr.schedparam.sched_priority, 20) != RT_EOK)
{
ret = EINVAL;
goto __exit;
}
/* set pthread id */
*pid = pth_id;
/* set pthread cleanup function and ptd data */
ptd->tid->cleanup = _pthread_cleanup;
ptd->tid->pthread_data = (void *)ptd;
/* start thread */
if (rt_thread_startup(ptd->tid) == RT_EOK)
return 0;
/* start thread failed */
rt_thread_detach(ptd->tid);
ret = EINVAL;
__exit:
if (pth_id != PTHREAD_NUM_MAX)
{
_pthread_data_destroy(ptd);
}
return ret;
}
RTM_EXPORT(pthread_create);
/**
* @brief Marks a thread as detached, allowing its resources to be automatically released upon termination.
*
* The `pthread_detach` function separates the specified thread from the calling thread. Once a thread is detached,
* its resources will be automatically reclaimed by the system upon the thread's termination. A detached thread cannot
* be joined using `pthread_join`.
*
* @param[in] thread
* The thread ID of the thread to be detached. This must be a valid thread ID returned by `pthread_create`.
*
* @return
* Returns 0 on success. On failure, an error code is returned:
* - `EINVAL`: The specified thread is not joinable or is already detached.
* - `ESRCH`: No thread with the specified ID could be found.
*
* @note
* - Detaching a thread allows it to run independently. Once detached, the thread's termination status cannot
* be retrieved, and it cannot be joined.
* - Threads can be created in a detached state using attributes set with `pthread_attr_setdetachstate`.
*
* @see pthread_create, pthread_join, pthread_attr_setdetachstate
*/
int pthread_detach(pthread_t thread)
{
int ret = 0;
_pthread_data_t *ptd = _pthread_get_data(thread);
if (ptd == RT_NULL)
{
/* invalid pthread id */
ret = EINVAL;
goto __exit;
}
if (ptd->attr.detachstate == PTHREAD_CREATE_DETACHED)
{
/* The implementation has detected that the value specified by thread does not refer
* to a joinable thread.
*/
ret = EINVAL;
goto __exit;
}
if ((RT_SCHED_CTX(ptd->tid).stat & RT_THREAD_STAT_MASK) == RT_THREAD_CLOSE)
{
/* destroy this pthread */
_pthread_data_destroy(ptd);
goto __exit;
}
else
{
/* change to detach state */
ptd->attr.detachstate = PTHREAD_CREATE_DETACHED;
/* detach joinable semaphore */
if (ptd->joinable_sem)
{
rt_sem_delete(ptd->joinable_sem);
ptd->joinable_sem = RT_NULL;
}
}
__exit:
return ret;
}
RTM_EXPORT(pthread_detach);
/**
* @brief Waits for the specified thread to terminate and retrieves its exit status.
*
* The `pthread_join` function blocks the calling thread until the specified thread terminates.
* If the specified thread has already terminated, it returns immediately. The exit status of
* the terminated thread can optionally be retrieved via the `value_ptr` parameter.
*
* @param[in] thread
* The thread ID of the thread to wait for. This must be a joinable thread created with `pthread_create`.
*
* @param[out] value_ptr
* A pointer to a location where the exit status of the terminated thread will be stored.
* If the thread terminated by calling `pthread_exit`, the value passed to `pthread_exit`
* will be stored at this location. If this parameter is `NULL`, the exit status is ignored.
*
* @return
* Returns 0 on success. On failure, an error code is returned:
* - `ESRCH`: The specified thread does not exist.
* - `EINVAL`: The specified thread is not joinable.
* - `EDEADLK`: A deadlock was detected (e.g., a thread tries to join itself).
*
* @note
* - Threads must not be detached to use `pthread_join`.
* - If `pthread_join` is not called for joinable threads, their resources are not released, leading to resource leaks.
*
* @see pthread_create, pthread_exit, pthread_detach
*/
int pthread_join(pthread_t thread, void **value_ptr)
{
_pthread_data_t *ptd;
rt_err_t result;
ptd = _pthread_get_data(thread);
if (ptd == RT_NULL)
{
return EINVAL; /* invalid pthread id */
}
if (ptd->tid == rt_thread_self())
{
/* join self */
return EDEADLK;
}
if (ptd->attr.detachstate == PTHREAD_CREATE_DETACHED)
{
return EINVAL; /* join on a detached pthread */
}
result = rt_sem_take(ptd->joinable_sem, RT_WAITING_FOREVER);
if (result == RT_EOK)
{
/* get return value */
if (value_ptr != RT_NULL)
*value_ptr = ptd->return_value;
/* destroy this pthread */
_pthread_data_destroy(ptd);
}
else
{
return ESRCH;
}
return 0;
}
RTM_EXPORT(pthread_join);
/**
* @brief Returns the thread ID of the calling thread.
*
* The `pthread_self` function returns the thread ID of the calling thread. The thread ID is unique to the
* thread within a process and can be used to identify the calling thread in the context of multithreading.
*
* The value returned by `pthread_self` can be compared with the thread IDs of other threads to determine
* if two threads are the same.
*
* @return
* The thread ID of the calling thread.
*
* @note
* - The thread ID returned by `pthread_self` is not the same as the operating system's thread ID.
* - This function does not affect the calling thread's state or execution.
* - The thread ID returned by `pthread_self` is only meaningful in the context of the current process.
*
* @see pthread_create, pthread_equal, pthread_join
*/
pthread_t pthread_self (void)
{
rt_thread_t tid;
_pthread_data_t *ptd;
tid = rt_thread_self();
if (tid == NULL) return PTHREAD_NUM_MAX;
/* get pthread data from pthread_data of thread */
ptd = (_pthread_data_t *)rt_thread_self()->pthread_data;
RT_ASSERT(ptd != RT_NULL);
return _pthread_data_get_pth(ptd);
}
RTM_EXPORT(pthread_self);
/**
* @brief Retrieves the clock ID for the specified thread.
*
* The `pthread_getcpuclockid` function retrieves the clock ID associated with the CPU time used
* by the specified thread.
*
* @param[in] thread
* The thread whose CPU clock ID is to be retrieved. If the thread is the calling thread,
* the current thread's ID is used.
*
* @param[out] clock_id
* A pointer to a `clockid_t` variable that will be filled with the clock ID associated
* with the specified thread.
*
* @return
* - `0` on success.
* - `EINVAL` if the `thread` is not a valid thread identifier.
* - `ESRCH` if the specified thread does not exist.
*
* @note
* The clock returned by this function is specific to the thread and is different from the
* system-wide clock. It measures the CPU time consumed by the specified thread, not wall-clock
* time. The thread's CPU time can be obtained using `clock_gettime` with the returned `clock_id`.
*
* @see clock_gettime, pthread_create, pthread_self
*/
int pthread_getcpuclockid(pthread_t thread, clockid_t *clock_id)
{
if(_pthread_get_data(thread) == NULL)
{
return EINVAL;
}
*clock_id = (clockid_t)rt_tick_get();
return 0;
}
RTM_EXPORT(pthread_getcpuclockid);
/**
* @brief Retrieves the current concurrency level of the program.
*
* The `pthread_getconcurrency` function returns the current concurrency level of the program.
* This value represents the number of threads that can run concurrently in the program,
* based on the current settings of the pthreads library. It is used to help tune the behavior
* of thread scheduling in some systems.
*
* @return
* The current concurrency level of the program.
* - The value is an integer representing the number of threads that are permitted to run
* concurrently in the system, based on the library's current configuration.
* - A return value of `0` typically means that the system is using the default concurrency
* level, which may be determined automatically by the system or by thread creation behavior.
*
* @note
* - The behavior and meaning of concurrency levels can be implementation-dependent,
* and it may vary across different systems or environments.
* - The function is typically used for diagnostic purposes, and its behavior may not
* affect thread execution directly.
*
* @see pthread_setconcurrency
*/
int pthread_getconcurrency(void)
{
return concurrency_level;
}
RTM_EXPORT(pthread_getconcurrency);
/**
* @brief Sets the concurrency level of the program.
*
* The `pthread_setconcurrency` function sets the number of threads that are allowed to run concurrently.
* The concurrency level defines the maximum number of threads that can be executed in parallel by the system.
* This is useful for tuning thread behavior and controlling system resource usage, especially in environments
* with limited resources (e.g., CPU cores).
*
* @param[in] new_level
* The new concurrency level to be set. This value represents the number of threads that can execute concurrently.
* - A value of `0` typically means that the system will automatically determine the concurrency level based on
* the system's configuration and available resources.
* - A non-zero value explicitly sets the maximum number of threads that can run concurrently.
*
* @return
* - `0` on success.
* - `EINVAL` if the `new_level` is invalid or if the system does not support this functionality.
*
* @note
* - The behavior of this function is system-dependent. Some systems may ignore the concurrency setting
* and automatically manage the concurrency based on available resources (e.g., CPU cores).
* - This function may not have any effect on systems that do not support concurrency settings at the library level.
* - The concurrency level controls thread scheduling policies and is intended to influence how the thread library
* manages threads, not how the operating system schedules them at the kernel level.
*
* @see pthread_getconcurrency
*/
int pthread_setconcurrency(int new_level)
{
concurrency_level = new_level;
return 0;
}
RTM_EXPORT(pthread_setconcurrency);
/**
* @brief Retrieves the scheduling policy and parameters of a thread.
*
* The `pthread_getschedparam` function retrieves the scheduling policy and the scheduling parameters
* (such as priority) for the specified thread. This allows you to check the scheduling settings of a thread
* and can be useful for thread management and performance tuning in a multithreaded application.
*
* @param[in] thread
* The thread whose scheduling policy and parameters are to be retrieved. This is typically a valid
* `pthread_t` identifier of a thread that has already been created.
*
* @param[out] policy
* A pointer to an integer where the scheduling policy of the specified thread will be stored. The
* value will be one of the following constants defined in `<sched.h>`:
* - `SCHED_FIFO`: First-in, first-out scheduling policy.
* - `SCHED_RR`: Round-robin scheduling policy.
* - `SCHED_OTHER`: Default policy, which is typically used by non-realtime threads.
* - `SCHED_IDLE`: For idle threads (system-level threads that do minimal work).
* - `SCHED_BATCH`: For threads that should be scheduled with lower priority than interactive threads.
* - `SCHED_DEADLINE`: A policy that allows specifying real-time deadlines (on systems that support it).
*
* @param[out] param
* A pointer to a `struct sched_param` where the scheduling parameters (e.g., priority) for the thread
* will be stored. The `sched_param` structure typically contains:
* - `sched_priority`: The priority value associated with the thread's scheduling policy.
*
* @return
* - `0` on success.
* - `ESRCH` if the specified thread does not exist.
* - `EINVAL` if an invalid argument is provided, such as an invalid thread ID or null pointers for the policy or parameters.
*
* @note
* - This function retrieves the current scheduling settings for a thread. These settings can be used
* to monitor or adjust thread behavior.
* - The scheduling policies and priorities may be platform-dependent and subject to system configuration.
*
* @see pthread_setschedparam, sched_getparam
*/
int pthread_getschedparam(pthread_t thread, int *policy, struct sched_param *param)
{
_pthread_data_t *ptd;
ptd = _pthread_get_data(thread);
pthread_attr_getschedpolicy(&ptd->attr, policy);
pthread_attr_getschedparam(&ptd->attr, param);
return 0;
}
RTM_EXPORT(pthread_getschedparam);
/**
* @brief Sets the scheduling policy and parameters for a thread.
*
* The `pthread_setschedparam` function sets the scheduling policy and scheduling parameters (such as priority)
* for the specified thread. This allows you to control how the thread is scheduled by the operating system.
* It is useful for adjusting thread behavior, especially for real-time or performance-sensitive applications.
*
* @param[in] thread
* The thread whose scheduling policy and parameters are to be set. This is a valid `pthread_t` identifier.
*
* @param[in] policy
* The scheduling policy to be set for the thread. This can be one of the following values:
* - `SCHED_FIFO`: First-in, first-out scheduling policy, where threads are scheduled based on their arrival time.
* - `SCHED_RR`: Round-robin scheduling policy, where each thread is allocated a fixed time slice and scheduled cyclically.
* - `SCHED_OTHER`: Default policy for non-realtime threads.
* - `SCHED_IDLE`: For threads intended to run only when no other threads are runnable.
* - `SCHED_BATCH`: For threads that should run with lower priority than interactive threads.
* - `SCHED_DEADLINE`: For real-time threads that have a specified deadline (if supported).
*
* @param[in] param
* A pointer to a `struct sched_param`, which contains the scheduling parameters, typically the thread's priority.
* The `sched_priority` field is the most commonly used parameter, and it controls the thread's priority within
* the specified scheduling policy.
*
* @return
* - `0` on success.
* - `EINVAL` if an invalid policy or parameter is provided.
* - `ESRCH` if the specified thread does not exist.
* - `EPERM` if the caller does not have permission to modify the thread's scheduling attributes.
*
* @note
* - The `sched_param` structure's `sched_priority` field specifies the priority of the thread. The priority
* range depends on the policy used. For example, for `SCHED_FIFO` and `SCHED_RR`, higher priority values
* correspond to higher priority threads, while for `SCHED_OTHER`, priorities are not as strictly enforced.
* - Changing a thread's scheduling parameters may affect its execution behavior, including how it competes with
* other threads for CPU time.
* - The system may not allow you to modify scheduling parameters for all threads, depending on system configuration
* and privileges.
*
* @see pthread_getschedparam
*/
int pthread_setschedparam(pthread_t thread, int policy, const struct sched_param *param)
{
_pthread_data_t *ptd;
ptd = _pthread_get_data(thread);
pthread_attr_setschedpolicy(&ptd->attr, policy);
pthread_attr_setschedparam(&ptd->attr, param);
return 0;
}
RTM_EXPORT(pthread_setschedparam);
/**
* @brief Sets the scheduling priority for a thread.
*
* The `pthread_setschedprio` function adjusts the priority of the specified thread while leaving its
* scheduling policy unchanged. This is useful for fine-tuning thread behavior in multithreaded applications.
*
* @param[in] thread
* The thread whose scheduling priority is to be changed. This must be a valid `pthread_t` identifier.
*
* @param[in] prio
* The new scheduling priority for the thread. The priority must fall within the valid range for the
* thread's current scheduling policy, as defined by `sched_get_priority_min` and `sched_get_priority_max`.
*
* @return
* - `0` on success.
* - `EINVAL` if the specified priority is invalid for the thread's current scheduling policy.
* - `ESRCH` if the specified thread does not exist.
* - `EPERM` if the calling process lacks the necessary privileges to set the thread's priority.
*
* @note
* - Changing a thread's priority may require elevated privileges (e.g., root) on certain systems, especially
* for real-time priorities.
* - The priority range and behavior depend on the thread's current scheduling policy. For example:
* - `SCHED_FIFO` and `SCHED_RR`: Priorities are used for strict scheduling.
* - `SCHED_OTHER`: Priorities may have minimal or no effect.
* - The behavior of this function is platform-dependent and may vary between different operating systems.
*
* @see pthread_setschedparam, pthread_getschedparam
*/
int pthread_setschedprio(pthread_t thread, int prio)
{
_pthread_data_t *ptd;
struct sched_param param;
ptd = _pthread_get_data(thread);
param.sched_priority = prio;
pthread_attr_setschedparam(&ptd->attr, ¶m);
return 0;
}
RTM_EXPORT(pthread_setschedprio);
/**
* @brief Terminates the calling thread and optionally returns a value.
*
* The `pthread_exit` function terminates the calling thread. It can optionally provide an exit status that can be
* retrieved by other threads that join the calling thread using `pthread_join`. If the thread is detached, the
* exit status is ignored and the system automatically reclaims resources once the thread terminates.
*
* @param[in] value
* A pointer to a value that will be returned to any thread that calls `pthread_join` on this thread.
* If `NULL`, no value is returned.
*
* @note
* - This function does not terminate the process. It only terminates the calling thread.
* - If the calling thread is the main thread, `pthread_exit` allows other threads to continue execution.
* - If a thread terminates without calling `pthread_exit`, it returns control to the system when the thread's function ends.
*
* @see pthread_join, pthread_create
*/
void pthread_exit(void *value)
{
_pthread_data_t *ptd;
_pthread_cleanup_t *cleanup;
rt_thread_t tid;
if (rt_thread_self() == RT_NULL)
{
return;
}
/* get pthread data from pthread_data of thread */
ptd = (_pthread_data_t *)rt_thread_self()->pthread_data;
rt_enter_critical();
/* disable cancel */
ptd->cancelstate = PTHREAD_CANCEL_DISABLE;
/* set return value */
ptd->return_value = value;
rt_exit_critical();
/*
* When use pthread_exit to exit.
* invoke pushed cleanup
*/
while (ptd->cleanup != RT_NULL)
{
cleanup = ptd->cleanup;
ptd->cleanup = cleanup->next;
cleanup->cleanup_func(cleanup->parameter);
/* release this cleanup function */
rt_free(cleanup);
}
/* get the info aboult "tid" early */
tid = ptd->tid;
/* According to "detachstate" to whether or not to recycle resource immediately */
if (ptd->attr.detachstate == PTHREAD_CREATE_JOINABLE)
{
/* set value */
rt_sem_release(ptd->joinable_sem);
}
else
{
/* release pthread resource */
_pthread_data_destroy(ptd);
}
/*
* second: detach thread.
* this thread will be removed from scheduler list
* and because there is a cleanup function in the
* thread (pthread_cleanup), it will move to defunct
* thread list and wait for handling in idle thread.
*/
rt_thread_detach(tid);
/* reschedule thread */
rt_schedule();
}
RTM_EXPORT(pthread_exit);
/**
* @brief Executes a routine once in a multithreaded environment.
*
* The `pthread_once` function ensures that the specified initialization routine is executed exactly once,
* even if multiple threads attempt to execute it simultaneously. It is typically used for one-time
* initialization tasks in a multithreaded program.
*
* @param[in] once_control
* A pointer to a `pthread_once_t` control variable. The init_routine can only be excuted
* when (*once_control) is zero.
*
* @param[in] init_routine
* A pointer to the initialization routine to be executed. This routine takes no arguments and
* returns no value. It is guaranteed to be executed exactly once.
*
* @return
* - `0` on success.
*
* @note
* - The `pthread_once` function is thread-safe and guarantees that the `init_routine` is called only once.
* - The `once_control` variable must remain valid and should not be modified by the application after
* initialization.
* - If the initialization routine fails or encounters an error, it is the responsibility of the routine
* to handle it appropriately.
*
* @see pthread_mutex_lock, pthread_mutex_unlock
*/
int pthread_once(pthread_once_t *once_control, void (*init_routine)(void))
{
RT_ASSERT(once_control != RT_NULL);
RT_ASSERT(init_routine != RT_NULL);
rt_enter_critical();
if (!(*once_control))
{
/* call routine once */
*once_control = 1;
rt_exit_critical();
init_routine();
}
rt_exit_critical();
return 0;
}
RTM_EXPORT(pthread_once);
int pthread_atfork(void (*prepare)(void), void (*parent)(void), void (*child)(void))
{
return EOPNOTSUPP;
}
RTM_EXPORT(pthread_atfork);
/**
* @brief Sends a signal to a specific thread.
*
* The `pthread_kill` function sends the specified signal to the target thread. This allows fine-grained
* control over signal handling in multithreaded applications.
*
* @param[in] thread
* The target thread to which the signal is sent. This is a valid `pthread_t` identifier.
*
* @param[in] sig
* The signal to be sent. This can be any valid signal, such as those defined in `<signal.h>`. For example:
* - `SIGTERM`: Request thread termination.
* - `SIGUSR1` or `SIGUSR2`: User-defined signals.
* - `0`: Used to check if the thread is still valid without sending a signal.
*
* @return
* - `0` on success.
* - `ESRCH` if the specified thread does not exist or is invalid.
* - `EINVAL` if the signal number `sig` is invalid.
*
* @note
* - The signal is delivered to the specified thread only if the thread has the appropriate signal handlers
* set up. Unhandled signals might result in the default action for that signal.
* - If `sig` is `0`, no signal is sent, but the function checks if the thread is valid and exists.
* - Signal handling behavior is shared across threads in a process. For example, blocking or ignoring a signal
* in one thread affects the entire process.
*
* @see pthread_sigmask, sigaction
*/
int pthread_kill(pthread_t thread, int sig)
{
#ifdef RT_USING_SIGNALS
_pthread_data_t *ptd;
int ret;
ptd = _pthread_get_data(thread);
if (ptd)
{
ret = rt_thread_kill(ptd->tid, sig);
if (ret == -RT_EINVAL)
{
return EINVAL;
}
return ret;
}
return ESRCH;
#else
return ENOSYS;
#endif
}
RTM_EXPORT(pthread_kill);
#ifdef RT_USING_SIGNALS
/**
* @brief Modifies or retrieves the signal mask of the calling thread.
*
* The `pthread_sigmask` function allows a thread to block, unblock, or examine the signals in its signal mask.
* Signals that are blocked are not delivered to the thread until they are unblocked.
*
* @param[in] how
* Specifies how the signal mask is modified. Possible values:
* - `SIG_BLOCK`: Add the signals in `set` to the current signal mask.
* - `SIG_UNBLOCK`: Remove the signals in `set` from the current signal mask.
* - `SIG_SETMASK`: Replace the current signal mask with the signals in `set`.
*
* @param[in] set
* A pointer to a `sigset_t` containing the signals to be modified in the mask. Can be `NULL` if no change is needed.
*
* @param[out] oset
* A pointer to a `sigset_t` where the previous signal mask will be stored. Can be `NULL` if the previous mask is not required.
*
* @return
* - `0` on success.
*
* @note
* - Signal masks are thread-specific in a multithreaded program.
* - The `pthread_sigmask` function is designed for multithreaded programs, whereas `sigprocmask` should not be used.
* - Blocking a signal prevents it from being delivered to the thread until unblocked.
*
* @see sigprocmask, sigaction, pthread_kill
*/
int pthread_sigmask(int how, const sigset_t *set, sigset_t *oset)
{
return sigprocmask(how, set, oset);
}
#endif
/**
* @brief Unregisters a cleanup handler and optionally executes it.
*
* The `pthread_cleanup_pop` function unregisters a cleanup handler that was previously registered
* using `pthread_cleanup_push`. If the `execute` parameter is non-zero, the cleanup handler is executed
* at the point where the thread terminates or is canceled.
*
* If `execute` is zero, the handler is unregistered without being executed. This allows the handler
* to be removed from the cleanup stack without performing any actions.
*
* @param[in] execute
* If non-zero, the cleanup handler is executed when the thread terminates or is canceled.
* If zero, the handler is simply removed from the stack without executing it.
*
* @note
* - Cleanup handlers are executed in the reverse order of their registration (i.e., last-in, first-out).
* - It is important to use `pthread_cleanup_push` to register cleanup handlers and `pthread_cleanup_pop`
* to ensure they are properly unregistered and executed if needed.
* - This function should be paired with `pthread_cleanup_push` to manage cleanup handlers effectively.
*
* @see pthread_cleanup_push, pthread_exit, pthread_cancel
*/
void pthread_cleanup_pop(int execute)
{
_pthread_data_t *ptd;
_pthread_cleanup_t *cleanup;
if (rt_thread_self() == NULL) return;
/* get pthread data from pthread_data of thread */
ptd = (_pthread_data_t *)rt_thread_self()->pthread_data;
RT_ASSERT(ptd != RT_NULL);
if (execute)
{
rt_enter_critical();
cleanup = ptd->cleanup;
if (cleanup)
ptd->cleanup = cleanup->next;
rt_exit_critical();
if (cleanup)
{
cleanup->cleanup_func(cleanup->parameter);
rt_free(cleanup);
}
}
}
RTM_EXPORT(pthread_cleanup_pop);
/**
* @brief Registers a cleanup handler to be executed when the calling thread terminates.
*
* The `pthread_cleanup_push` function registers a cleanup handler that is executed when the calling thread
* is canceled or exits (either normally or via `pthread_exit`). The cleanup handler will be executed
* in the reverse order of their registration.
*
* The cleanup handler can be used to release resources such as memory or file descriptors when the thread
* is terminated, whether it terminates normally or is canceled.
*
* @param[in] routine
* A pointer to the cleanup handler function. The function must have the following signature:
* `void routine(void* arg);`. It is invoked when the thread terminates or is canceled.
*
* @param[in] arg
* A pointer to the argument that will be passed to the cleanup handler (`routine`).
* This allows the handler to perform actions with the passed argument.
*
* @note
* - The cleanup handler is automatically invoked when a thread terminates or is canceled.
* - The cleanup handlers are executed in the reverse order of their registration, similar to how
* destructors are executed in a stack-based fashion.
* - `pthread_cleanup_pop` must be called to unregister the cleanup handler. It ensures that the handler
* is only invoked during the thread's termination process.
*
* @see pthread_cleanup_pop, pthread_cancel, pthread_exit
*/
void pthread_cleanup_push(void (*routine)(void *), void *arg)
{
_pthread_data_t *ptd;
_pthread_cleanup_t *cleanup;
if (rt_thread_self() == NULL) return;
/* get pthread data from pthread_data of thread */
ptd = (_pthread_data_t *)rt_thread_self()->pthread_data;
RT_ASSERT(ptd != RT_NULL);
cleanup = (_pthread_cleanup_t *)rt_malloc(sizeof(_pthread_cleanup_t));
if (cleanup != RT_NULL)
{
cleanup->cleanup_func = routine;
cleanup->parameter = arg;
rt_enter_critical();
cleanup->next = ptd->cleanup;
ptd->cleanup = cleanup;
rt_exit_critical();
}
}
RTM_EXPORT(pthread_cleanup_push);
/*
* According to IEEE Std 1003.1, 2004 Edition , following pthreads
* interface support cancellation point:
* mq_receive()
* mq_send()
* mq_timedreceive()
* mq_timedsend()
* msgrcv()
* msgsnd()
* msync()
* pthread_cond_timedwait()
* pthread_cond_wait()
* pthread_join()
* pthread_testcancel()
* sem_timedwait()
* sem_wait()
*
* A cancellation point may also occur when a thread is
* executing the following functions:
* pthread_rwlock_rdlock()
* pthread_rwlock_timedrdlock()
* pthread_rwlock_timedwrlock()
* pthread_rwlock_wrlock()
*
* The pthread_cancel(), pthread_setcancelstate(), and pthread_setcanceltype()
* functions are defined to be async-cancel safe.
*/
/**
* @brief Sets the cancelability state of the calling thread.
*
* The `pthread_setcancelstate` function allows a thread to enable or disable its ability to be canceled
* by another thread. Cancelability determines if and when a thread responds to a cancellation request.
*
* @param[in] state
* The new cancelability state for the calling thread. Possible values:
* - `PTHREAD_CANCEL_ENABLE`: The thread can be canceled.
* - `PTHREAD_CANCEL_DISABLE`: The thread cannot be canceled.
*
* @param[out] oldstate
* A pointer to an integer where the previous cancelability state will be stored. Can be `NULL` if
* the previous state is not needed.
*
* @return
* - `0` on success.
* - `EINVAL` if the `state` is not a valid cancelability state.
*
* @note
* - The cancelability state affects how the thread responds to cancellation requests:
* - In the `PTHREAD_CANCEL_DISABLE` state, cancellation requests are held pending until the state is changed to `PTHREAD_CANCEL_ENABLE`.
* - Cancelability is distinct from the cancelability type, which controls the timing of cancellation (deferred or asynchronous).
* - By default, threads are created with `PTHREAD_CANCEL_ENABLE`.
*
* @see pthread_cancel, pthread_setcanceltype
*/
int pthread_setcancelstate(int state, int *oldstate)
{
_pthread_data_t *ptd;
if (rt_thread_self() == NULL) return EINVAL;
/* get pthread data from pthread_data of thread */
ptd = (_pthread_data_t *)rt_thread_self()->pthread_data;
RT_ASSERT(ptd != RT_NULL);
if ((state == PTHREAD_CANCEL_ENABLE) || (state == PTHREAD_CANCEL_DISABLE))
{
if (oldstate)
*oldstate = ptd->cancelstate;
ptd->cancelstate = state;
return 0;
}
return EINVAL;
}
RTM_EXPORT(pthread_setcancelstate);
/**
* @brief Sets the cancellation type of the calling thread.
*
* The `pthread_setcanceltype` function allows a thread to specify when it should respond to
* a cancellation request. The cancellation type can be set to deferred or asynchronous.
*
* @param[in] type
* The new cancellation type for the calling thread. Possible values:
* - `PTHREAD_CANCEL_DEFERRED`: Cancellation occurs at cancellation points (default behavior).
* - `PTHREAD_CANCEL_ASYNCHRONOUS`: Cancellation occurs immediately when a request is received.
*
* @param[out] oldtype
* A pointer to an integer where the previous cancellation type will be stored. Can be `NULL`
* if the previous type is not required.
*
* @return
* - `0` on success.
* - `EINVAL` if the `type` is not a valid cancellation type.
*
* @note
* - The cancellation type determines when a thread processes a cancellation request:
* - **Deferred**: The thread responds to cancellation only at well-defined cancellation points.
* - **Asynchronous**: The thread can be canceled immediately, which may lead to resource inconsistencies.
* - By default, threads use `PTHREAD_CANCEL_DEFERRED`.
* - Asynchronous cancellation should be used cautiously as it can interrupt a thread at any point.
*
* @see pthread_cancel, pthread_setcancelstate, pthread_testcancel
*/
int pthread_setcanceltype(int type, int *oldtype)
{
_pthread_data_t *ptd;
if (rt_thread_self() == NULL) return EINVAL;
/* get pthread data from pthread_data of thread */
ptd = (_pthread_data_t *)rt_thread_self()->pthread_data;
RT_ASSERT(ptd != RT_NULL);
if ((type != PTHREAD_CANCEL_DEFERRED) && (type != PTHREAD_CANCEL_ASYNCHRONOUS))
return EINVAL;
if (oldtype)
*oldtype = ptd->canceltype;
ptd->canceltype = type;
return 0;
}
RTM_EXPORT(pthread_setcanceltype);
/**
* @brief Explicitly checks for pending cancellation requests in the calling thread.
*
* The `pthread_testcancel` function allows a thread to determine if it has a pending
* cancellation request. If a cancellation request is pending and the thread's cancelability
* state is set to `PTHREAD_CANCEL_ENABLE`, the thread will terminate immediately.
*
* @note
* - This function is a cancellation point, meaning it checks for cancellation and responds if applicable.
* - If the thread's cancelability state is `PTHREAD_CANCEL_DISABLE`, the function has no effect.
* - The thread will invoke any cleanup handlers registered with `pthread_cleanup_push` before termination.
*
* @return
* This function does not return if a cancellation is performed. Otherwise, it returns normally.
*
* @see pthread_setcancelstate, pthread_setcanceltype, pthread_cancel
*/
void pthread_testcancel(void)
{
int cancel = 0;
_pthread_data_t *ptd;
if (rt_thread_self() == NULL) return;
/* get pthread data from pthread_data of thread */
ptd = (_pthread_data_t *)rt_thread_self()->pthread_data;
RT_ASSERT(ptd != RT_NULL);
if (ptd->cancelstate == PTHREAD_CANCEL_ENABLE)
cancel = ptd->canceled;
if (cancel)
pthread_exit((void *)PTHREAD_CANCELED);
}
RTM_EXPORT(pthread_testcancel);
/**
* @brief Sends a cancellation request to a specified thread.
*
* The `pthread_cancel` function requests the cancellation of the thread identified by `thread`.
* The actual response to the request depends on the target thread's cancelability state and type.
*
* @param[in] thread
* The identifier of the thread to be canceled.
*
* @return
* - `0` on success.
* - `EINVAL` if the specified thread does not exist.
*
* @note
* - Cancellation is an asynchronous mechanism. The thread may not terminate immediately or at all
* if its cancelability state is set to `PTHREAD_CANCEL_DISABLE`.
* - If the thread is cancelable and terminates, it invokes cleanup handlers registered with
* `pthread_cleanup_push` before termination.
* - The thread's cancellation type determines when it processes the cancellation request:
* - `PTHREAD_CANCEL_DEFERRED` (default): At specific cancellation points.
* - `PTHREAD_CANCEL_ASYNCHRONOUS`: Immediately upon receipt of the request.
*
* @see pthread_setcancelstate, pthread_setcanceltype, pthread_testcancel
*/
int pthread_cancel(pthread_t thread)
{
_pthread_data_t *ptd;
_pthread_cleanup_t *cleanup;
rt_thread_t tid;
/* get posix thread data */
ptd = _pthread_get_data(thread);
if (ptd == RT_NULL)
{
return EINVAL;
}
tid = ptd->tid;
/* cancel self */
if (ptd->tid == rt_thread_self())
return 0;
/* set canceled */
if (ptd->cancelstate == PTHREAD_CANCEL_ENABLE)
{
ptd->canceled = 1;
if (ptd->canceltype == PTHREAD_CANCEL_ASYNCHRONOUS)
{
/*
* When use pthread_cancel to exit.
* invoke pushed cleanup
*/
while (ptd->cleanup != RT_NULL)
{
cleanup = ptd->cleanup;
ptd->cleanup = cleanup->next;
cleanup->cleanup_func(cleanup->parameter);
/* release this cleanup function */
rt_free(cleanup);
}
/* According to "detachstate" to whether or not to recycle resource immediately */
if (ptd->attr.detachstate == PTHREAD_CREATE_JOINABLE)
{
/* set value */
rt_sem_release(ptd->joinable_sem);
}
else
{
/* release pthread resource */
_pthread_data_destroy(ptd);
}
/*
* second: detach thread.
* this thread will be removed from scheduler list
* and because there is a cleanup function in the
* thread (pthread_cleanup), it will move to defunct
* thread list and wait for handling in idle thread.
*/
rt_thread_detach(tid);
}
}
return 0;
}
RTM_EXPORT(pthread_cancel);