When xTimerDelete is called using a dymanic timer handle, the timer
handle should immediately be considered unusable for any operation;
including checking if the timer is still running. Under high system
loads, that memory region may see fast reuse while furi_timer_free
is sleeping between timer active checks. That reuse could result in
memory at that pointer causing the timer active check to return true.
Rework the furi_timer_delete process (in the case of dynamically
allocated callback memory) to stop the timer, wait for it to stop,
free the memory, and then delete the timer. Timers without dynamically
allocated callback memory are just sent a delete command; no need
to stop it first.
Fixes: ff33bc6aea ("Furi: wait for timer wind down in destructor (#1716)")
Signed-off-by: Kris Bahnsen <Kris@KBEmbedded.com>
* FuriHal: retry gauge/charger initialization
* FuriHal: lower logging level for flash known errata
* FuriHal: graceful fail if subghz chip is not working
* Furi: issue stop command even if timer is not active, document timer behavior
* Furi: remove direct FreeRTOS timers use
* Furi: eliminate FreeRTOS headers leak. What did it cost? Everything...
* SubGhz: proper public api for protocols. Format Sources.
* Furi: slightly less redundant declarations
* Desktop: proper types in printf
* Sync API Symbols
* Furi: add timer reset and fix dolphin service, fix unit tests
* Furi: proper timer restart method naming and correct behavior in timer stopped state.
---------
Co-authored-by: hedger <hedger@nanode.su>
* Furi: properly handle thread free before TCB scrapping, add furi_free - more invasive version of free to memmgr. FuriHal: add DWT comparator api to cortex. Updater, RPC: refactor various thread shanenigans. Code cleanup.
* Rollback free macros and related changes
* Furi: make `furi_is_irq_context` public
* Furi: proper name and documentation for furi_kernel_is_irq_or_masked.
* Target: bump symbol table version
* Furi: proper doxygen context for warnings
Co-authored-by: Aleksandr Kutuzov <alleteam@gmail.com>
