FP exception reporting not working on NetBSD host

[Paul Goyette]

I recognize that NetBSD is not one of the officially supported host OS. However, qemu 0.13.0 is available in the NetBSD pkgsrc collection, and works quite well. Well, with one exception (pun intended): It seems that Floating Point exceptions don't get reported properly.

The following code-snippet demonstrates the problem:

volatile int flt_signal = 0;

static sigjmp_buf sigfpe_flt_env;
static void
sigfpe_flt_action(int signo, siginfo_t *info, void *ptr)
{
        flt_signal++;
}

void trigger(void)
{
        struct sigaction sa;
        double d = strtod("0", NULL);

        if (sigsetjmp(sigfpe_flt_env, 0) == 0) {
                sa.sa_flags = SA_SIGINFO;
                sa.sa_sigaction = sigfpe_flt_action;
                sigemptyset(&sa.sa_mask);
                sigaction(SIGFPE, &sa, NULL);
                fpsetmask(FP_X_INV|FP_X_DZ|FP_X_OFL|FP_X_UFL|FP_X_IMP);
                printf("%g\n", 1 / d);
        }
        printf("FPE signal handler invoked %d times.\n");
}

[Paulo Cezar A Junior (pcacjr)]

On Mon, Jan 3, 2011 at 12:14 PM, Paul Goyette <email address hidden>wrote:

> Public bug reported:
>
> I recognize that NetBSD is not one of the officially supported host OS.
> However, qemu 0.13.0 is available in the NetBSD pkgsrc collection, and
> works quite well. Well, with one exception (pun intended): It seems
> that Floating Point exceptions don't get reported properly.
>
> The following code-snippet demonstrates the problem:
>
>
> volatile int flt_signal = 0;
>
> static sigjmp_buf sigfpe_flt_env;
> static void
> sigfpe_flt_action(int signo, siginfo_t *info, void *ptr)
> {
> flt_signal++;
> }
>
> void trigger(void)
> {
> struct sigaction sa;
> double d = strtod("0", NULL);
>
> if (sigsetjmp(sigfpe_flt_env, 0) == 0) {
> sa.sa_flags = SA_SIGINFO;
> sa.sa_sigaction = sigfpe_flt_action;
> sigemptyset(&sa.sa_mask);
> sigaction(SIGFPE, &sa, NULL);
> fpsetmask(FP_X_INV|FP_X_DZ|FP_X_OFL|FP_X_UFL|FP_X_IMP);
> printf("%g\n", 1 / d);
> }
> printf("FPE signal handler invoked %d times.\n");
>

this printf() does miss an argument :-)

> }
>
> ** Affects: qemu
> Importance: Undecided
> Status: New
>
> --
> You received this bug notification because you are a member of qemu-
> devel-ml, which is subscribed to QEMU.
> https://bugs.launchpad.net/bugs/696834
>
> Title:
> FP exception reporting not working on NetBSD host
>
> Status in QEMU:
> New
>
> Bug description:
> I recognize that NetBSD is not one of the officially supported host OS.
> However, qemu 0.13.0 is available in the NetBSD pkgsrc collection, and
> works quite well. Well, with one exception (pun intended): It seems that
> Floating Point exceptions don't get reported properly.
>
> The following code-snippet demonstrates the problem:
>
>
> volatile int flt_signal = 0;
>
> static sigjmp_buf sigfpe_flt_env;
> static void
> sigfpe_flt_action(int signo, siginfo_t *info, void *ptr)
> {
> flt_signal++;
> }
>
> void trigger(void)
> {
> struct sigaction sa;
> double d = strtod("0", NULL);
>
> if (sigsetjmp(sigfpe_flt_env, 0) == 0) {
> sa.sa_flags = SA_SIGINFO;
> sa.sa_sigaction = sigfpe_flt_action;
> sigemptyset(&sa.sa_mask);
> sigaction(SIGFPE, &sa, NULL);
> fpsetmask(FP_X_INV|FP_X_DZ|FP_X_OFL|FP_X_UFL|FP_X_IMP);
> printf("%g\n", 1 / d);
> }
> printf("FPE signal handler invoked %d times.\n");
>

this printf() does miss an argument :-)

> }
>
>
>
>

[Paul Goyette (paul-whooppee)]

On Mon, 3 Jan 2011, Paulo Cezar A Junior wrote:

<snip>

>> printf("FPE signal handler invoked %d times.\n");
>
> this printf() does miss an argument :-)

Yes, it does. The signal handler is also missing a line:

  siglongjmp(sigfpe_flt_env, 1);

That's what I get for extracting bits&pieces of the larger test program.
:)

The following is a complete, standalone test program:

  #include <ieeefp.h>
  #include <setjmp.h>
  #include <signal.h>
  #include <stdio.h>

  volatile int flt_signal = 0;

  static sigjmp_buf sigfpe_flt_env;
  static void
  sigfpe_flt_action(int signo, siginfo_t *info, void *ptr)
  {

   flt_signal++;
   siglongjmp(sigfpe_flt_env, 1);
  }
  int main(int argc, void *argv[])
  {
   struct sigaction sa;
   double d;

   printf("Start\n");
   if (sigsetjmp(sigfpe_flt_env, 0) == 0) {
    sa.sa_flags = SA_SIGINFO;
    sa.sa_sigaction = sigfpe_flt_action;
    sigemptyset(&sa.sa_mask);
    sigaction(SIGFPE, &sa, NULL);
    fpsetmask(FP_X_INV|FP_X_DZ|FP_X_OFL|FP_X_UFL|FP_X_IMP);
    d = 1.0 / strtod("0", NULL);
   }
   printf("FPE signal handler invoked %d times.\n");
  }

Executing the program gives the following results:

  # cc -o test test.c
  # ./test
  Start
  FPE signal handler invoked 0 times.
  #

On "real" hardware,

  {225} cc -o test test.c
  {226} ./test
  Start
  FPE signal handler invoked 1 times.
  {227}

-------------------------------------------------------------------------
| Paul Goyette | PGP Key fingerprint: | E-mail addresses: |
| Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer | | pgoyette at netbsd.org |
-------------------------------------------------------------------------

[Paul Goyette (paul-whooppee)]

On Mon, 3 Jan 2011, Paul Goyette wrote:

> The following is a complete, standalone test program:
>
> #include <ieeefp.h>
> #include <setjmp.h>
> #include <signal.h>
> #include <stdio.h>
>
> volatile int flt_signal = 0;
>
> static sigjmp_buf sigfpe_flt_env;
> static void
> sigfpe_flt_action(int signo, siginfo_t *info, void *ptr)
> {
>
> flt_signal++;
> siglongjmp(sigfpe_flt_env, 1);
> }
> int main(int argc, void *argv[])
> {
> struct sigaction sa;
> double d;
>
> printf("Start\n");
> if (sigsetjmp(sigfpe_flt_env, 0) == 0) {
> sa.sa_flags = SA_SIGINFO;
> sa.sa_sigaction = sigfpe_flt_action;
> sigemptyset(&sa.sa_mask);
> sigaction(SIGFPE, &sa, NULL);
> fpsetmask(FP_X_INV|FP_X_DZ|FP_X_OFL|FP_X_UFL|FP_X_IMP);
> d = 1.0 / strtod("0", NULL);
> }
> printf("FPE signal handler invoked %d times.\n");

And, of course, I still missed the extra agument:

                 printf("FPE signal handler invoked %d times.\n", flt_signal);

> }
>
> Executing the program gives the following results:
>
> # cc -o test test.c
> # ./test
> Start
> FPE signal handler invoked 0 times.
> #
>
> On "real" hardware,
>
> {225} cc -o test test.c
> {226} ./test
> Start
> FPE signal handler invoked 1 times.
> {227}
>
>
> -------------------------------------------------------------------------
> | Paul Goyette | PGP Key fingerprint: | E-mail addresses: |
> | Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com |
> | Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
> | Kernel Developer | | pgoyette at netbsd.org |
> -------------------------------------------------------------------------
>
> --
> You received this bug notification because you are a direct subscriber
> of the bug.
> https://bugs.launchpad.net/bugs/696834
>
> Title:
> FP exception reporting not working on NetBSD host
>
> Status in QEMU:
> New
>
> Bug description:
> I recognize that NetBSD is not one of the officially supported host OS. However, qemu 0.13.0 is available in the NetBSD pkgsrc collection, and works quite well. Well, with one exception (pun intended): It seems that Floating Point exceptions don't get reported properly.
>
> The following code-snippet demonstrates the problem:
>
>
> volatile int flt_signal = 0;
>
> static sigjmp_buf sigfpe_flt_env;
> static void
> sigfpe_flt_action(int signo, siginfo_t *info, void *ptr)
> {
> flt_signal++;
> }
>
> void trigger(void)
> {
> struct sigaction sa;
> double d = strtod("0", NULL);
>
> if (sigsetjmp(sigfpe_flt_env, 0) == 0) {
> sa.sa_flags = SA_SIGINFO;
> sa.sa_sigaction = sigfpe_flt_action;
> sigemptyset(&sa.sa_mask);
> sigaction(SIGFPE, &sa, NULL);
> fpsetmask(FP_X_INV|FP_X_DZ|FP_X_OFL|FP_X_UFL|FP_X_IMP);
> printf("%g\n", 1 / d);
> }
> printf("FPE signal handler invoked %d times.\n");
> }
>
> To unsubscribe from this bug, go to:
> https://bugs.launchpad.net/qemu/+bug/696834/+subscribe
>
> !DSPAM:4d21f0752341470756574!
>
>
>

-----------------------------------------------------------...

[Paulo Cezar A Junior (pcacjr)]

LinkedIn
------------

Bug,

I'd like to add you to my professional network on LinkedIn.

- Paulo

Paulo Cezar
Platform Researcher and Developer at INdT
São Paulo Area, Brazil

Confirm that you know Paulo Cezar
https://www.linkedin.com/e/-g11m92-go1e59vx-23/isd/2968180228/v_gmfnNW/

--
(c) 2011, LinkedIn Corporation

[Thomas Huth (th-huth)]

QEMU 0.13 is completely outdated nowadays - can you still reproduce this problem with the latest version of QEMU (currently version 2.8) ?

[Paul Goyette (paul-whooppee)]

On Tue, 10 Jan 2017, Thomas Huth wrote:

> QEMU 0.13 is completely outdated nowadays - can you still reproduce this
> problem with the latest version of QEMU (currently version 2.8) ?

The test program running on qemu 2.8.0 now produces the expected results
- 1 FP exception

You can close this bug.

+------------------+--------------------------+------------------------+
| Paul Goyette | PGP Key fingerprint: | E-mail addresses: |
| (Retired) | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com |
| Kernel Developer | 0786 F758 55DE 53BA 7731 | pgoyette at netbsd.org |
+------------------+--------------------------+------------------------+

[Thomas Huth (th-huth)]

Thanks for verifying!