faultive address reported on SIGSEGV was simply junk

Hi,

A while back I had the 2.6 kernel running on my Mandrake sparc box, see: 
http://archives.mandrakelinux.com/cooker/2004-02/msg00092.php

Mandrake's glibc maintainer (Gwenole Beauchesne) took a look at it and 
had the following to say: 
http://archives.mandrakelinux.com/cooker/2004-02/msg00298.php

====

On Sun, 1 Feb 2004, Thomas Backlund wrote:

> Now for the next step...
> I'll log into the buildhost to get the latest changes to the kernels,
> and will add them to my 2.4 series as well....

Does it work correctly now wrt. signal handlers? Last time I checked on
Stefan's machine, the faultive address reported on SIGSEGV was simply
junk.

> Gwenole,
> as you seems to be the maintainer of dietlibc-devel,
> would it be possible to update it to 0.24 ?

It is, but I don't think I will update stage1 dietlibc as it works, unless
gcc wants it to be updated too.

> It has many fixes for the sparc/sparc64 branch...,
> and many other interesting fixes...

All interesting AMD64 fixes are already in our dietlibc. ;-)

Bye,
Gwenole.
====

I'm wondering if this also takes place on aurora with the 2.6 kernel. And how it's been fixed.

The following peice of code can be used to test it:


====
#include <stdio.h>
#include <signal.h>
#include <errno.h>
#include <unistd.h>
#include <sys/mman.h>
#include <stdlib.h>

static volatile char *page;

static void sigsegv_handler(int sig, siginfo_t *sip, void *scp)
{
  printf("Caught SIGSEGV to %08x\n", sip->si_addr);
  if (sip->si_addr != page)
          abort();
  exit(0);
}

int main(void)
{
  struct sigaction sigsegv_sa;
  sigemptyset(&sigsegv_sa.sa_mask);
  sigsegv_sa.sa_sigaction = sigsegv_handler;
  sigsegv_sa.sa_flags = SA_SIGINFO;
  if (sigaction(SIGSEGV, &sigsegv_sa, 0) < 0) {
    perror("sigaction");
    return 1;
  }
 
  {
    int page_size = getpagesize();
    volatile unsigned char *p = (volatile unsigned char *)mmap(0, 
page_size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0);
    if (p == MAP_FAILED) {
      perror("mmap");
      return 2;
    }

    if (mprotect(p, page_size, PROT_READ) < 0) {
      perror("mprotect");
      return 3;
    }

    printf("p = %p\n", p);
    page = p;
    *p = 0;

    munmap(p, page_size);
  }
  return 0;
}
====

On i586 I saw this behaviour:

./test.i586
p = 0x40018000
Caught SIGSEGV to 40018000

While on sparc this:

./test.sparc32
p = 0x70018000
Caught SIGSEGV to 70018377
Aborted

regards,

Stefan

_______________________________________________
Aurora-sparc-devel mailing list
[email protected]
http://lists.auroralinux.org/mailman/listinfo/aurora-sparc-devel
Aurora FAQ: http://www.ecs.soton.ac.uk/~mas01r/aurorafaq.html