Skip to contentSkip to conversation

eal: add rte_exit() main lcore limitation description

PATCH In review Suanming Mou 2 participants 2 messages Jul 14, 2021 – Jun 30, 2023

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 eal: add rte_exit() main lcore limitation description

Currently, rte_eal_mp_wait_lcore() function will be called by rte_exit() as the routine below:
rte_exit() -> rte_eal_cleanup() -> rte_service_finalize() -> rte_eal_mp_wait_lcore()

As rte_eal_mp_wait_lcore() is announced can only be called from main lcore, rte_exit() gets the limitation implicitly as well. Or once rte_exit() is called from a worker core, the rte_exit() procedure will get stuck in the rte_eal_mp_wait_lcore() function as the core status is still running.

This commit adds the limitation to rte_exit() to make things clear.

---
 lib/eal/include/rte_common.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
lib/eal/include/rte_common.h +1 -1
index d5a32c66a5..ebe2869cc3 100644
@@ -857,7 +857,7 @@ rte_str_to_size(const char *str)
  * Function to terminate the application immediately, printing an error
  * message and returning the exit_code back to the shell.
  *
- * This function never returns
+ * To be executed on the MAIN lcore only. This function never returns.
  *
  * @param exit_code
  *     The exit code to be returned by the application
Signature · 1 line
2.25.1
#1 lore raw
Quoted 5 lines · On Wed, 14 Jul 2021 17:13:00 +0300 Suanming Mou <[email protected]> wrote:

Currently, rte_eal_mp_wait_lcore() function will be called by rte_exit() as the routine below:
rte_exit() -> rte_eal_cleanup() -> rte_service_finalize() -> rte_eal_mp_wait_lcore()

Quoted 8 lines

As rte_eal_mp_wait_lcore() is announced can only be called from main lcore, rte_exit() gets the limitation implicitly as well. Or once rte_exit() is called from a worker core, the rte_exit() procedure will get stuck in the rte_eal_mp_wait_lcore() function as the core status is still running.

This commit adds the limitation to rte_exit() to make things clear.

This patch has had no response in almost two years.
The root cause of the problem is a the assumption in the service library. That is a bug and should be filed against rte_service, not by having a caveat in documentation that no one will read anyway.