wifi: eswifi: fix random crash due to work->handler set to NULL #50153
Conversation
mzajac-avs
changed the title
mzajac-avs
force-pushed
the
l475_fix_random_crash
branch
from
51360aa to
4e14e90
Compare
| @@ -397,7 +397,7 @@ static int eswifi_off_put(struct net_context *context) | |||
| } | |||
|
|
|||
| if (--socket->usage <= 0) { | |||
| memset(socket, 0, sizeof(*socket)); | |||
| socket->context = NULL; | |||
There was a problem hiding this comment.
That certainly works, but wouldn't it be possible to cancel any ongoing/scheduled work instead? I mean once socket is 'released' it's no more assigned to a context, and can be reused at any time, so we don't want anything to access it anymore after this line.
There was a problem hiding this comment.
Well I'm not sure did I try to do this correctly but I added a static struct k_work_sync work_sync to the file and called k_work_cancel_delayable_sync(&socket->read_work, &work_sync); before the memset(). I tired it a few times but it seems that the execution of the application frozen each time (but there was no crash). The stack showed that I'm in arch_cpu_idle () at /home/mzajac/Repo/Internal/zephyr-root/zephyr/arch/arm/core/aarch32/cpu_idle.S:105
There was a problem hiding this comment.
@loicpoulain Is it fine for you then ?
There was a problem hiding this comment.
@erwango yes, though it's more a workaround than a fix. but better than nothing.
|
@mzajac-avs Please check your commit message. It should conform to the following template: For references, I'd advise you to check:
|
This fixes a random crash caused by race condition in the eswifi driver used by the disco L475 iot1 board. Signed-off-by: Michał Zając <m.zajac@avsystem.com>
mzajac-avs
force-pushed
the
l475_fix_random_crash
branch
from
4e14e90 to
172b852
Compare
|
@erwango Fixed the commit message |
|
@loicpoulain, @nandojve can you approve again? Your approval got lost. |
This fixes the random crash on L475 disco iot1 board described here: #49963