In my personal branch (which hopefully I'll be submitting soon), I
changed hwthread_update_threads() to preserve current_threadid, which fixed
a problem I ran into.
Does this patch also work for you?
--- a/src/rtos/hwthread.c
+++ b/src/rtos/hwthread.c
@@ -87,13 +92,10 @@ static int hwthread_update_threads(struct rtos *rtos)
enum target_debug_reason current_reason = DBG_REASON_UNDEFINED;
if (rtos == NULL)
- return -1;
+ return ERROR_FAIL;
target = rtos->target;
- /* wipe out previous thread details if any */
- rtos_free_threadlist(rtos);
-
/* determine the number of "threads" */
if (target->smp) {
for (head = target->head; head != NULL; head = head->next) {
@@ -107,6 +109,11 @@ static int hwthread_update_threads(struct rtos *rtos)
} else
thread_list_size = 1;
+ /* Wipe out previous thread details if any, but preserve threadid.
*/
+ int64_t current_threadid = rtos->current_threadid;
+ rtos_free_threadlist(rtos);
+ rtos->current_threadid = current_threadid;
+
/* create space for new thread details */
rtos->thread_details = malloc(sizeof(struct thread_detail) *
thread_list_size);
Tim
On Mon, Apr 8, 2019 at 9:36 AM Daniel Goehring OS via OpenOCD-devel <
[email protected]> wrote:
> Hello,
>
> For the OpenOCD project, I’d like to commit a change to remove the call
> to “rtos_update_threads()” in function gdb_handle_vcont_packet() (file
> gdb_server.c, line 2775), but wanted to get input from the community first.
>
> To provide some environment background, OpenOCD is configured to
> communicate to a GDB client using the GDB remote serial protocol. The
> chip being debugged is an Ampere Computing eMAG 8180 and both
> OpenOCD/GDB are configured SMP multi-core with HW RTOS support enabled.
> The HW RTOS feature allows CPU target selection using the GDB threads
> command.
>
> It’s not clear to me why “rtos_update_threads()” is called here because
> rtos state should already be up-to-date. In addition, this routine calls
> “rtos_free_threadlist()” which wipes out previous thread details such as
> “current_threadid” and “current_thread”. In a GDB multi-cpu (SMP) debug
> environment, resetting current_threadid and current_thread breaks
> functionality on an Ampere Computing eMAG 8180 development platform.
>
> Removing the call fixes the issue I was seeing (current_threadid and
> current_thread values remain intact throughout the call to
> gdb_handle_vcont_packet()).
>
> Specifics below…
>
> File <path>/src/server/gdb_server.c, line 2775
> Lines in question:
> -start-
> /* FIXME: why is this necessary? rtos state should be up-to-date here
> already! */
> rtos_update_threads(target);
> -end-
>
> Thanks,
> Dan Goehring
>
> _______________________________________________
> OpenOCD-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/openocd-devel
>
_______________________________________________
OpenOCD-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/openocd-devel
