core/msg: prevent thread from stalling indefinitely

[HendrikVE]

Contribution description

Before, when calling msg_send_receive() with an invalid target_pid the thread trying to send the message would have been put into blocked state. But when sending the message is failing, this thread will then remain blocked forever, since it can't expect to receive a response from a thread that never got its message (because this thread simply does not exist).

Testing procedure

All you need to do is to call msg_send_receive with target_pid=0.

Declaration of AI-Tools / LLMs usage:

AI-Tools / LLMs that were used are:

• none

[riot-ci]

Murdock results

❌ FAILED

78965c6 tests/core/msg_send_receive: exclude atmega8

Build failures (1)

Application	Target	Toolchain	Runtime (s)	Worker
examples/networking/dtls/dtls-wolfssl	cc1350-launchpad	gnu	7.79	flatserv

Artifacts

[benpicco]

Looks good to me

[HendrikVE]

@maribu Would you mind adding the second ACK? :)

[crasbe]

I wonder if it would be worth adding a test case for this?

[HendrikVE]

I wonder if it would be worth adding a test case for this?

Makes sense to me. I'd try to integrate it into the existing test at tests/core/msg_send_receive/main.c

[HendrikVE]

I added some checks to the existing code. The test now also checks for the return value of msg_send_receive and it makes sure that the main thread is still scheduled by calling thread_yield. If it was blocked due to the bug the final puts("Test successful."); line would not get executed.

[maribu]

The CI failure is not a glitch, sadly

[crasbe]

The CI failure is not a glitch, sadly

No, it's the classic "changed a test, now it doesn't fit into the memory of $smöl_microcontroller anymore".

I ran make generate-Makefile.ci on skyleaf and this is the result:

buechse@skyleaf:~/RIOTstuff/riot-vanillaice/RIOT/tests/core/msg_send_receive$ git diff Makefile.ci
diff --git a/tests/core/msg_send_receive/Makefile.ci b/tests/core/msg_send_receive/Makefile.ci
index d28f5b77f9..5f2032215b 100644
--- a/tests/core/msg_send_receive/Makefile.ci
+++ b/tests/core/msg_send_receive/Makefile.ci
@@ -4,6 +4,7 @@ BOARD_INSUFFICIENT_MEMORY := \
     arduino-uno \
     atmega328p \
     atmega328p-xplained-mini \
+    atmega8 \
     nucleo-f031k6 \
     nucleo-l011k4 \
     samd10-xmini \

[HendrikVE]

Do I get this right that the solution is to add the atmega8 to the list? ^^

[crasbe]

Yes.

[crasbe]

Fascinating. I ran make generate-Makefile.ci locally on skyleaf and it did not add cc1350-launchpad to the examples/networking/dtls/dtls-wolfssl/Makefile.ci...

Considering the build failed twice because of it, it might be a good idea to do it nevertheless :D

But the good news is that no other application has to be modified.