Handle SIGINT and SIGTERM signals for graceful shutdown of sink service #287
Conversation
sgunes-wirepas
force-pushed
the
fix/sink_sigint
branch
from
f3d7bba to
3149db7
Compare
sink_service/source/main.c
Outdated
| @@ -149,6 +150,30 @@ static int open_and_check_connection(unsigned long baudrate, char * port_name) | |||
| return 0; | |||
| } | |||
|
|
|||
| static volatile sig_atomic_t stop_requested = 0; | |||
There was a problem hiding this comment.
could you prefix it with m_ as it is a global variable of the module.
sink_service/source/main.c
Outdated
| @@ -308,7 +338,7 @@ int main(int argc, char * argv[]) | |||
| goto finish; | |||
| } | |||
|
|
|||
| for (;;) | |||
| while (!stop_requested) | |||
There was a problem hiding this comment.
Did I miss something here, but how does sd_bus_wait exit in order to evaluate this condition?
There was a problem hiding this comment.
Would their be a way to force it from stop_signal_handler ?
There was a problem hiding this comment.
Very good point, thanks. At least when I ran it, sd_bus_wait was returning quite fast (almost immediately) but maybe it is different when there is an unresponsive sink? Do you think calling sd_bus_wait with a timeout value of some seconds be safer?
There was a problem hiding this comment.
Looks like sd_bus_wait is using ppoll internally, and that is returning also when any signal is caught.
sd_bus_wait seems to treat the EINTR error from ppoll as a success:
https://github.com/systemd/systemd/blob/9ef559a0363f6baea17a7dfdc15ff52ba43bc4f6/src/libsystemd/sd-bus/sd-bus.c#L3426
This makes it easier to test the sink service for memory issues, and speeds up the stopping of the docker container (docker sends SIGTERM when stopping).
sgunes-wirepas
force-pushed
the
fix/sink_sigint
branch
from
3149db7 to
42fa87e
Compare
This makes it easier to test the sink service for memory issues (like in #286) using valgrind or sanitizers
Also speeds up the stopping of the docker container (docker sends SIGTERM when stopping).