Re: [microblaze-uclinux] problem with signals (SIGIO) in the petalinux-2.6.2 kernel

[Islam Samir <islambio@xxxxxxxxx>]

Versions: PetaLinux dist v0.2-rc3 ---   Kernel: 2.4.x

Just a follow-up :) I tried the same code on the petalinux 2.4 kernel. The first time the signal is received every thing is ok. The second time the signal is received the following error pops:

kernel BUG at sched.c: 562 !

and the system hangs.

I changed the signal from SIGIO to SIGCHLD on the same kernel 2.4. The results: 
-First signal is received correctly. Second signal is not received and terminates the process (intst)... but there's no kernel BUG stuff and there's no system hang-up.

I need signals in my application to be able to handle interrupts in the user space. I have a data acquisition system (DAQ) attached to the microblaze that sends interrupts whenever a new sample is ready. I want my user space application to read the new sample through XGPIO. I used asynchronous notifications (FASYNC) flag for that purpose after some modifications to the XGPIO
 driver.

If signals are not suitable, what is the alternative ?? please any ideas... ?? !!

My project deadline is really soon, and the whole project depends on the DAQ. If I couldn't handle the new sample from the DAQ, then there's no project :((

Regards,
Islam

Islam Samir <islambio@xxxxxxxxx> wrote:

Hi !
I'm trying to implement a signal handler for the SIGIO signal in the user space.. The  singal is caught successfully with the signal handler. The problem is when the singal handler reaches the "return" point, it completely crashes the whole system !!

I traced the program using microblaze-uclinux-gdb and followed step by step what happens after the call to "return" in the signal handler, simply it goes into the address area of the process *stack* ... it then continues exceution from there !! just executing some "garbage" instructions in the stack untill a total system crash happens.

Is it normal for the signal handler to return into the stack ??
Please note that the very same attached code works very well on my desktop linux (i386) with NO problems!!

---

Get your own web address.
Have a HUGE year through Yahoo! Small Business.#include
#include

/*for fcntl()*/
#include
/*for fcntl()*/
#include

/*for getpid()*/
#include
/*for getpid()*/
#include

/*for pthreads*/
#include

/*for memset()*/
#include

/*mutex is short for MUTual EXclusion*/
pthread_mutex_t dum_mutex = PTHREAD_MUTEX_INITIALIZER;

typedef void (*sighandler_t)(int);

void sighnd(int sig)
{
struct sigaction sa;
memset(&sa, 0, sizeof(sa));
sa.sa_handler = &sighnd;
int ret = sigaction(SIGIO, &sa, NULL);

/* try to lock the dummy mutex to prevent any other threads of this handler to run before the current one finishes */
int i = pthread_mutex_lock (&dum_mutex);

if(!ret)
printf("\n\nRe-registered the signal handler just in case linux resets it, registeration result: %d", ret);
printf("\n\nHandler caught interrupt signal SIGIO");
fflush(stdout);
/*unlock the dummy mutex*/
pthread_mutex_unlock (&dum_mutex);
return;
}

int main(void)
{

struct sigaction sa;
memset(&sa, 0, sizeof(sa));
sa.sa_handler = &sighnd;
int ret = sigaction(SIGIO, &sa, NULL);

/*sighandler_t ret = signal(SIGIO, &sighnd);*/

if(!ret)
printf("\n\nRegistered a handler to signal!\nWaiting for interrupts...");

//wait for signals from the kernel
for(;;)
pause();//this function returns only when the signal handler returns

return 0;

}
#include
#include
#include


int main(void)
{
int i;
//the signal chld doesn't kill other processes
i = kill(50, SIGIO);
i = kill(51, SIGIO);
i = kill(52, SIGIO);
i = kill(53, SIGIO);

//printf("\n\n%i\n\n", i);

}

---

Yahoo! oneSearch: Finally,  mobile search 
that gives answers, not web links.