sched/hrtimer: Update the comments.

This commit updated the comments.

Signed-off-by: ouyangxiangzhen <ouyangxiangzhen@xiaomi.com>
This commit is contained in:
ouyangxiangzhen
2026-01-28 14:43:19 +08:00
committed by Donny(董九柱)
parent 3eedf5f22b
commit dac48484ce
6 changed files with 62 additions and 53 deletions
+30 -23
View File
@@ -113,17 +113,14 @@ extern "C"
* Name: hrtimer_init
*
* Description:
* Initialize a high-resolution timer instance. This function sets the
* expiration callback and its argument. The timer is initialized in the
* inactive state and is not armed until hrtimer_start() is called.
* Initialize a high-resolution timer instance.
*
* Input Parameters:
* hrtimer - Pointer to the hrtimer instance to be initialized
* func - Expiration callback function
* arg - Argument passed to the callback
*
* Returned Value:
* None
*
****************************************************************************/
#define hrtimer_init(hrtimer) memset(hrtimer, 0, sizeof(hrtimer_t))
@@ -132,14 +129,23 @@ extern "C"
* Name: hrtimer_cancel
*
* Description:
* Cancel a high-resolution timer asynchronously. This function set the
* Cancel a high-resolution timer asynchronously.
*
* If the timer is currently pending, it will be removed from the
* hrtimer queue and will not be executed.
*
* If the timer callback is currently executing. This function set the
* timer to the cancelled state. The caller will acquire the limited
* ownership of the hrtimer, which allow the caller restart the hrtimer,
* but the callback function may still be executing on another CPU, which
* prevent the caller from freeing the hrtimer. The caller must call
* `hrtimer_cancel` to wait for the callback to be finished. Please use
* the function with care. Concurrency errors are prone to occur in this
* use case.
* but the callback function may still be executing on another CPU,
* which prevent the caller from freeing the hrtimer.
* The caller must call `hrtimer_cancel_sync` to wait for the callback
* to be finished. Please use the function with care.
* Concurrency errors are prone to occur in this use case.
*
* If the canceled timer was the earliest expired timer in the queue,
* the expiration of the underlying hardware timer will be updated to the
* expiration time of the next earliest timer
*
* This function is non-blocking and does not wait for a running callback
* to finish.
@@ -151,11 +157,8 @@ extern "C"
* OK (0) on success; a negated errno value on failure.
* > 0 on if the timer callback is running.
*
* Assumptions/Notes:
* - This function acquires the global hrtimer spinlock to protect both
* the red-black tree and the timer state.
* - The caller must ensure that the timer structure is not freed until
* it is guaranteed that any running callback has returned.
* Assumptions:
* - The hrtimer is not NULL.
*
****************************************************************************/
@@ -165,13 +168,16 @@ int hrtimer_cancel(FAR hrtimer_t *hrtimer);
* Name: hrtimer_cancel_sync
*
* Description:
* Cancel a high-resolution timer and wait until it becomes inactive.
*
* Cancel a high-resolution timer and synchronously wait the callback to
* be finished. This function set the timer to the cancelled state and wait
* for all references to be released. The caller will then acquire full
* ownership of the hrtimer. After the function returns, the caller can
* safely deallocate the hrtimer.
* be finished.
*
* If the timer callback is running, this function set the timer to the
* cancelled state and wait for all all references to be released.
* The caller will then acquire full ownership of the hrtimer. After the
* function returns, the caller can safely deallocate the hrtimer.
* - If sleeping is allowed (normal task context), yields CPU briefly
* to avoid busy-waiting.
* - Otherwise (interrupt or idle task context), spins until completion.
*
* Input Parameters:
* hrtimer - Pointer to the high-resolution timer instance to cancel.
@@ -193,13 +199,14 @@ int hrtimer_cancel_sync(FAR hrtimer_t *hrtimer);
* a relative timeout, depending on the selected mode.
*
* Input Parameters:
* hrtimer - Timer instance to start
* hrtimer - Pointer to high-resolution timer.
* func - Expiration callback function
* expired - Expiration time in nanoseconds
* mode - HRTIMER_MODE_ABS or HRTIMER_MODE_REL
*
* Returned Value:
* OK on success; a negated errno value on failure.
*
****************************************************************************/
int hrtimer_start_absolute(FAR hrtimer_t *hrtimer, hrtimer_entry_t func,
+14 -5
View File
@@ -116,7 +116,8 @@ extern uintptr_t g_hrtimer_running[CONFIG_SMP_NCPUS];
* now - The current time (nsecs).
*
* Returned Value:
* None
* None.
*
****************************************************************************/
void hrtimer_process(uint64_t now);
@@ -141,6 +142,7 @@ void hrtimer_process(uint64_t now);
*
* Assumptions:
* The underlying timer start function returns 0 on success.
*
****************************************************************************/
static inline_function void hrtimer_reprogram(uint64_t next_expired)
@@ -174,15 +176,21 @@ static inline_function void hrtimer_reprogram(uint64_t next_expired)
*
* Description:
* Compare two high-resolution timer nodes to determine their ordering
* in the container. Used internally by the RB-tree macros.
* in the hrtimer queue. Used internally by the RB-tree macros.
*
* Note that RB-tree can not guarantee that timers with the same expired
* time will be processed in a FIFO order. However, this violation
* of the sorted queue invariant is acceptable and can not affect the
* functional correctness for the hrtimer.
*
* Input Parameters:
* a - Pointer to the first hrtimer node.
* b - Pointer to the second hrtimer node.
*
* Returned Value:
* <0 if a expires before b
* >=0 if a expires after b
* <0 if a expires before b.
* >=0 if a expires after b.
*
****************************************************************************/
#ifdef CONFIG_HRTIMER_TREE
@@ -304,6 +312,7 @@ static inline_function bool hrtimer_insert(FAR hrtimer_t *hrtimer)
*
* Returned Value:
* Pointer to the earliest timer, or NULL if none are pending.
*
****************************************************************************/
static inline_function FAR hrtimer_t *hrtimer_get_first(void)
@@ -324,7 +333,7 @@ static inline_function FAR hrtimer_t *hrtimer_get_first(void)
* of the value you are reading.
*
* Input Parameters:
* ptr - The pointer to be read.
* ptr - The pointer to be read.
*
* Returned Value:
* The value in the queue.
+2 -5
View File
@@ -67,11 +67,8 @@
* OK (0) on success; a negated errno value on failure.
* > 0 on if the timer callback is running.
*
* Assumptions/Notes:
* - This function acquires the global hrtimer spinlock to protect
* the container.
* - The caller must ensure that the timer structure is not freed until
* it is guaranteed that any running callback has returned.
* Assumptions:
* - The hrtimer is not NULL.
*
****************************************************************************/
+7 -7
View File
@@ -42,19 +42,19 @@ uintptr_t g_hrtimer_running[CONFIG_SMP_NCPUS];
/* Global spinlock protecting the high-resolution timer subsystem.
*
* This spinlock serializes access to the hrtimer container and
* This spinlock serializes access to the hrtimer queue and
* protects timer state transitions. It must be held whenever the
* timer container is modified.
* timer queue is modified.
*/
seqcount_t g_hrtimer_lock = SEQLOCK_INITIALIZER;
/* Container for all active high-resolution timers.
/* HRTimer queue for all active high-resolution timers.
*
* When CONFIG_HRTIMER_TREE is enabled, timers are stored in a container.
* When CONFIG_HRTIMER_TREE is enabled, timers are stored in a queue.
* When disabled, timers are stored in a linked list.
*
* The container is ordered by absolute expiration time in
* The queue is ordered by absolute expiration time in
* both configurations.
*/
@@ -90,11 +90,11 @@ struct list_node g_hrtimer_list =
* Name: RB_GENERATE
*
* Description:
* Instantiate the container helper functions for the hrtimer
* Instantiate the queue helper functions for the hrtimer
* subsystem.
*
* This macro generates the static inline functions required to
* manipulate the hrtimer container, including insertion,
* manipulate the hrtimer queue, including insertion,
* removal, and lookup operations.
*
* Note: This is only compiled when CONFIG_HRTIMER_TREE is enabled.
+3 -3
View File
@@ -40,9 +40,9 @@
*
* Description:
* Process all expired high-resolution timers. This function repeatedly
* retrieves the earliest timer from the active timer container, checks
* retrieves the earliest timer from the active timer queue, checks
* if it has expired relative to the current time, removes it from the
* container, and invokes its callback function. Processing continues
* queue, and invokes its callback function. Processing continues
* until:
*
* 1. No additional timers have expired, or
@@ -62,7 +62,7 @@
* None.
*
* Assumptions/Notes:
* - This function acquires a spinlock to protect the timer container.
* - This function acquires a spinlock to protect the timer queue.
* - Timer callbacks are invoked with interrupts enabled
* to avoid deadlocks.
* - DEBUGASSERT ensures that timer callbacks are valid.
+6 -10
View File
@@ -43,20 +43,16 @@
* in nanoseconds.
*
* Input Parameters:
* hrtimer - Pointer to the hrtimer structure.
* hrtimer - Pointer to the hrtimer.
* func - Expiration callback function.
* expired - Expiration time in nanoseconds. Interpretation
* depends on mode.
* expired - Expiration time in nanoseconds.
*
* Returned Value:
* OK (0) on success, or a negated errno value on failure.
*
* Assumptions/Notes:
* - This function disables interrupts briefly via spinlock to safely
* insert the timer into the container.
* - Absolute mode sets the timer to expire at the given absolute time.
* - Relative mode sets the timer to expire after 'ns'
* nanoseconds from the current time.
* Assumptions:
* - hrtimer is not NULL and func is not NULL.
*
****************************************************************************/
int hrtimer_start_absolute(FAR hrtimer_t *hrtimer, hrtimer_entry_t func,
@@ -72,7 +68,7 @@ int hrtimer_start_absolute(FAR hrtimer_t *hrtimer, hrtimer_entry_t func,
flags = write_seqlock_irqsave(&g_hrtimer_lock);
/* Ensure no core can write the hrtimer. */
/* Ensure no running core can write the hrtimer. */
hrtimer_cancel_running(hrtimer);