Mirrors/esp-idf
1
0
Fork 0
mirror of https://github.com/espressif/esp-idf.git synced 2026-10-02 03:00:34 +03:00
Code Issues Packages Projects Releases Wiki Activity

fix(ble/bluedroid): allow osi_event re-post during POSTING window

Browse Source 
Do not reject osi_thread_post_event() when only POSTING is set.
QUEUED already prevents double-queueing; rejecting POSTING caused
HCI downstream lost wakeup. Add generic osi_event and hci downstream
diagnostics for post failures.


(cherry picked from commit 011138ffd9)

Co-authored-by: zhanghaipeng <zhanghaipeng@espressif.com>
This commit is contained in:
Zhang Hai Peng
2026-08-25 19:55:18 +08:00
parent 6f3d6977fe
commit c149a80c15
2 changed files with 45 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
components/bt/common/osi/thread.c
+27 -4
View File
@@ -534,13 +534,30 @@ static bool osi_event_can_bind_locked(const struct osi_event *event, osi_thread_
static bool osi_event_can_post_locked(const struct osi_event *event)
{
if (event->thread == NULL || event->queue_idx >= event->thread->work_queue_num) {
OSI_TRACE_EVENT("%s deny ev=%p flags=0x%x qidx=%u",
__func__, event, event ? event->flags : 0,
event ? event->queue_idx : 0);
return false;
}
return !OSI_EVENT_HAS_FLAG(event, OSI_EVENT_FLAG_DELETING) &&
event->item.func != NULL &&
!OSI_EVENT_HAS_FLAG(event, OSI_EVENT_FLAG_QUEUED) &&
!OSI_EVENT_HAS_FLAG(event, OSI_EVENT_FLAG_POSTING);
if (OSI_EVENT_HAS_FLAG(event, OSI_EVENT_FLAG_DELETING) ||
event->item.func == NULL ||
OSI_EVENT_HAS_FLAG(event, OSI_EVENT_FLAG_QUEUED)) {
OSI_TRACE_EVENT("%s deny ev=%p flags=0x%x qidx=%u wq_len=%d",
__func__, event, event->flags, event->queue_idx,
osi_thread_queue_wait_size(event->thread, event->queue_idx));
return false;
}
/* Do NOT gate on OSI_EVENT_FLAG_POSTING here. POSTING marks the window in
* osi_thread_post_event() between osi_thread_post() (enqueue) and the
* poster clearing the flag. During that window the generic event handler
* may already have run and cleared QUEUED. A concurrent post that arrives
* after QUEUED is cleared is a legitimate re-post (new work arrived while
* the handler was draining) and must be accepted; rejecting it causes a
* lost wakeup. QUEUED alone prevents genuine double-queueing. POSTING is
* retained only for osi_event_is_idle()/osi_event_should_free(). */
return true;
}
static bool osi_event_is_alive_locked(const struct osi_event *event)
@@ -714,6 +731,7 @@ static void osi_thread_generic_event_handler(void *context)
OSI_EVENT_SET_FLAG(event, OSI_EVENT_FLAG_RUNNING);
func = event->item.func;
func_context = event->item.context;
OSI_TRACE_DEBUG("%s enter ev=%p flags=0x%x", __func__, event, event->flags);
osi_mutex_unlock(&event->lock);
if (func != NULL) {
@@ -722,6 +740,7 @@ static void osi_thread_generic_event_handler(void *context)
osi_mutex_lock(&event->lock, OSI_MUTEX_MAX_TIMEOUT);
OSI_EVENT_CLEAR_FLAG(event, OSI_EVENT_FLAG_RUNNING);
OSI_TRACE_DEBUG("%s exit ev=%p flags=0x%x", __func__, event, event->flags);
osi_mutex_unlock(&event->lock);
osi_event_release(event);
@@ -755,6 +774,7 @@ bool osi_thread_post_event(struct osi_event *event, uint32_t timeout)
uint8_t queue_idx = 0;
if (!osi_event_acquire(event)) {
OSI_TRACE_EVENT("%s acquire fail ev=%p", __func__, event);
return false;
}
@@ -782,6 +802,9 @@ bool osi_thread_post_event(struct osi_event *event, uint32_t timeout)
osi_mutex_unlock(&event->lock);
if (!ret) {
OSI_TRACE_EVENT("%s enqueue fail ev=%p qidx=%u wq_len=%d",
__func__, event, queue_idx,
osi_thread_queue_wait_size(thread, queue_idx));
osi_event_release(event);
}
osi_event_release(event);