daemon: refactor IPC usage

[duvanan13]

• Prevent new daemon initialization when using FIFO
• Unit test gateway IPC changes to UNIX_SOCKET
• Shared memory is used for multinode testing
• Resolve the TOCTOU issue

[minminlittleshrimp]

Thanks @duvanan13 for joining the pair programming section.
I believe now you have more experience with source control using git.
Thank you for your the rework on this issue, let see how Michael comment about this fix.

Hello @michael-methner , could you kindly review this fix?
Thank you

[michael-methner]

Hello @duvanan13 , hello @minminlittleshrimp ,
this patch prevents are restart after dlt-daemon was not shutdown gracefully (e.g. due to a segmentation fault). Could modify the check so it checks for a running instance of dlt-daemon instead of the files present?

[duvanan13]

Hello @michael-methner ,
In the dlt_daemon_signal_handler function in daemon, we've already implemented the unlink for FIFO by calling dlt_daemon_exit_trigger. Hence, my point is to add SIGSEGV to the dlt_daemon_signal_handler:

dlt-daemon/src/daemon/dlt-daemon.c

Line 2135 in 358ab08

(void)unlink(tmp);

Please kindly check my point.

[michael-methner]

The handler for SIGSEGV should not be added here. This will cause the application to hang in an endless loop in case a segmentation fault occurs as it will re-execute the same instruction again which caused the segmentation fault (and signal is triggered again)

[andreasbank]

SIGSEGV needs to be handled separately. For example, after calling dlt_daemon_exit_trigger() set the SIG_DFL for SIGSEGV and raise() the same signal (or just let it hit SIGSEGV on its own again).

I really don't advise calling dlt_daemon_exit_trigger() since it calls async-unsafe functions in a signal handler, but since you already break that rule I wont complain.

[minminlittleshrimp]

It should be removed