Re: [Xenomai-help] Detecting switch to secondary mode

[Jan Kiszka <jan.kiszka@domain.hid>]

[-- Attachment #1: Type: text/plain, Size: 1824 bytes --]

J. Niehaus wrote:
> Hello,
> 
> I have written a xenomai application (posix-skin) with several tasks. All of them should normally use primary mode (SCHED_FIFO). While some are allowed to switch to secondary mode for stuff like I/O and memory allocation others must not switch due to hard realtime constraints. 
> 
> Is there a way to check if and how often a thread has switched to secondary mode and maybe even what system call caused the switch? While a test from within the program itself would be great some information from /proc/xenomai would also do.
> 

Yes, see this example:
http://www.rts.uni-hannover.de/xenomai/lxr/source/ksrc/skins/native/snippets/sigxcpu.c

We have the following code snippet in the signal handler of our RACK
framework:

void signal_handler(int sig)
{
[...]
    switch(sig)
    {
        case SIGXCPU:

            if (signal_flags & HANDLING_TERM) // ignore while
cleaning 		                                              // up
                return;

            printf("\n");
            printf("---=== SIGNAL HANDLER of %s ===---\n", classname);
            printf(" -> SIGXCPU (%02d)\n", sig);
            printf(" !!! WARNING unexpected switch to secondary mode "
                   "!!!\n");
            nentries = backtrace (array, 10);
            strings = backtrace_symbols (array, nentries);
            for (i = 0; i < nentries; i++)
            {
                printf ("  %s\n", strings[i]);
            }
            free (strings);
            printf("---=== END OF %s SIGNAL HANDLER ===---\n",
                   classname);
            return;
[...]

I.e. it does a back-trace to show what happened in user-space (debug
symbols are required). It's VERY useful when letting beginners hack
their first RT applications.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]