data race in shell.c

[clamattia]

https://github.com/zephyrproject-rtos/zephyr/blob/main/subsys/shell/shell.c#L1613

If, I am not mistaken. Line 1613 in shell.c can lead to a datarace. Expectation to caller is unclear. For example, code like this is a datarace?:

static char prompt[256];
strcpy(prompt, "Hello");
shell_prompt_change(sh, prompt);

because it boils down to:

char *a;
void shell_thread(void) {
    while(1) {
        // reads a
        if (a != NULL) {
            printf("%s\n", a);
        }
    }
}

void caller_thread(void) {
    static char prompt[256];
    strcpy(prompt, "Hello");
    a = prompt; // equivalent of shell_prompt_change(sh, prompt);
}

Could someone confirm, that this is the case? If so, I can create an issue.
Thanks in advance.

[nordicjm]

@jakub-uC

[clamattia]

Basically:
caller_thread could be concurrently writing to the prompt (optimizer is alowed to swap a=prompt; and strncpy)
shell_thread could be concurrently reading it.
It is not possible to provide a synchronization mechanism.

Proposed solution would be, that shell_prompt_change writes the data to private buffer and cares about the synchronisation with the shell_thread.

[clamattia]

@nordicjm Any news on this? May I create the issue?

[nordicjm]

Feel free to make the issue, Jakub is the shell maintainer so was bringing this to their attention, they'd need to review the bug

[clamattia]

#64972 created issue from this