[Bug target/52205] New: SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[Bug target/52205] New: SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[ian at airs dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

             Bug #: 52205
           Summary: SPARC Solaris 2.11 unwind through signal handler fails
                    with -fnon-call-exceptions
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: ian@airs.com
                CC: ro@gcc.gnu.org


Compile this C++ program with -fnon-call-exceptions:

#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

void
die(const char* msg)
{
  perror(msg);
  exit(EXIT_FAILURE);
}

void
handler(int signo, siginfo_t* info, void *context)
{
  printf("in handler signal %d\n", signo);
  throw signo;
}

int
main(int, const char**)
{
  struct sigaction act;
  memset(&act, 0, sizeof act);
  act.sa_sigaction = handler;
  sigfillset(&act.sa_mask);
  act.sa_flags = SA_SIGINFO;
  if (sigaction(SIGSEGV, &act, NULL) != 0)
    die("sigaction");

  try
    {
      *reinterpret_cast<char*>(0) = 1;
    }
  catch (int signo)
    {
      printf("caught signal %d\n", signo);
      exit(EXIT_SUCCESS);
    }

  printf("did not catch\n");
  exit(EXIT_FAILURE);
}

When this program is run, it should print something like

in handler signal 11
caught signal 11

This works fine on GNU/Linux and on x86 and x86_64 Solaris 2.11.  When run on
SPARC Solaris 2.11, however, it prints

in handler signal 11
Segmentation Fault

I see this in both 32-bit and 64-bit mode.  If I tweak
libgcc/config/sparc/sol2-unwind.h so that sparc_is_sighandler sets *nframes to
3 rather than 2, then the test passes (I only tried this in 32-bit mode, not in
64-bit mode).  The cuh_pattern test in sparc_is_sighandler does not match, so
presumably it needs to be adjusted for Solaris 2.11.  However, I'm not sure how
to properly and safely correct it.

[Bug target/52205] SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |DUPLICATE

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-02-10 23:34:28 UTC ---
This is a dup of bug 51921.

*** This bug has been marked as a duplicate of bug 51921 ***

[Bug target/52205] SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ebotcazou at gcc dot
                   |                            |gnu.org

--- Comment #2 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-02-11 10:47:46 UTC ---
> I see this in both 32-bit and 64-bit mode.  If I tweak
> libgcc/config/sparc/sol2-unwind.h so that sparc_is_sighandler sets *nframes to
> 3 rather than 2, then the test passes (I only tried this in 32-bit mode, not in
> 64-bit mode).  The cuh_pattern test in sparc_is_sighandler does not match, so
> presumably it needs to be adjusted for Solaris 2.11.  However, I'm not sure how
> to properly and safely correct it.

OK, let's at least try, first in 32-bit mode.  What code path is taken exactly
in sparc_is_sighandler on Solaris 11?

[Bug target/52205] SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[ian at airs dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

Ian Lance Taylor <ian at airs dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
   Last reconfirmed|                            |2012-02-12
         Resolution|DUPLICATE                   |
     Ever Confirmed|0                           |1

--- Comment #3 from Ian Lance Taylor <ian at airs dot com> 2012-02-12 04:12:07 UTC ---
I'm using a system Rainer gave me access to.  It's sparc-sun-solaris2.11. 
uname -a reports

SunOS mayon 5.11 11.0 sun4v sparc SUNW,SPARC-Enterprise-T5220 Solaris

readelf -V /lib/libc.so.1 shows that the highest version is SUNW_1.22.7.

Using mainline as of a couple of February 8, 2012 or so.  Looking at
libgcc/config/sparc/sol2-unwind.h.

This case matches:

  if(/* Solaris 8+ - multi-threaded
       ----------------------------
       <__sighndlr>:    save  %sp, -96, %sp
       <__sighndlr+4>:    mov  %i0, %o0
       <__sighndlr+8>:    mov  %i1, %o1
       <__sighndlr+12>:    call  %i3
       <__sighndlr+16>:    mov  %i2, %o2
       <__sighndlr+20>:    ret         <--- PC
       <__sighndlr+24>:    restore  */
        pc[-5] == 0x9de3bfa0
     && pc[-4] == 0x90100018
     && pc[-3] == 0x92100019
     && pc[-2] == 0x9fc6c000
     && pc[-1] == 0x9410001a
     && pc[ 0] == 0x81c7e008
     && pc[ 1] == 0x81e80000)

In that condition, cuh_pattern is set to 0x92100019.  This doesn't match any of
the choices in the code, so it returns 1 with *nframes = 2.  In order to work
correctly, it needs to return with *nframes = 3.

cuh_pattern is an instruction loaded from some code.  That code looks like
this:

   0xff298f48 <call_user_handler+876>:  mov  %i1, %o1
   0xff298f4c <call_user_handler+880>:  call  0xff2a552c <__sighndlr>
   0xff298f50 <call_user_handler+884>:  mov  %i5, %o2
   0xff298f54 <call_user_handler+888>:  ld  [ %fp + 0x4c ], %i5
   0xff298f58 <call_user_handler+892>:  ld  [ %fp + 0x44 ], %g5

[Bug target/52205] SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |ASSIGNED
         AssignedTo|unassigned at gcc dot       |ebotcazou at gcc dot
                   |gnu.org                     |gnu.org

--- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-02-12 10:06:07 UTC ---
Fixing.

[Bug target/52205] SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #5 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-02-12 15:33:17 UTC ---
Created attachment 26643
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26643
Tentative fix

Would you mind giving it a try on the Solaris 11 machine?  TIA.

[Bug target/52205] SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[ian at airs dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #6 from Ian Lance Taylor <ian at airs dot com> 2012-02-12 19:52:02 UTC ---
The patch fixes the test case and also passes some relevant Go tests.

Rainer, if OK, I'd like to leave it to you to comment on the patch and do a
full testsuite run.

[Bug target/52205] SPARC Solaris 2.11 unwind through signal handler fails with -fnon-call-exceptions

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #7 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-02-12 20:24:56 UTC ---
> The patch fixes the test case and also passes some relevant Go tests.

Great.  For the records, it was also tested on 5 different versions of Solaris
8, 9 and 10, covering all the cases in the code, both 32-bit and 64-bit.

[Bug target/52205] unwinding through signal handler fails

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #8 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2012-02-14 18:39:32 UTC ---
> --- Comment #6 from Ian Lance Taylor <ian at airs dot com> 2012-02-12 19:52:02 UTC ---
> The patch fixes the test case and also passes some relevant Go tests.
>
> Rainer, if OK, I'd like to leave it to you to comment on the patch and do a
> full testsuite run.

I'm currently doing so, and preliminary results look good.  I've got a
couple of comments on the patch as-is which I find harder than necessary
to read/understand:

* It uses far too magic numbers unnecessarily and/or without
  explanation.  E.g. looking for the caller frame with hardcoded offsets
  (cuh_patter, sah_address) is unnecessary: just use struct frame in
  <sys/frame.h> for that, where fr_savfp is the saved frame pointer at
  that exact offset.

* All the patterns should be acompanied by the disassembly to make it
  easier to compare with that you see e.g. in a debugger.

* Talking of Solaris 8 in Solaris Containers is unnecessary and
  incomplete: this is just the Solaris 8 alternate thread library (in
  /usr/lib/lwp) which is copied over /usr/lib/libthread.so.1 when
  booting the branded zone, but is also used by gcc on Solaris 8 by
  default.  This alternate thread library became the default in Solaris
  9, thus they are the same.

    Rainer

[Bug target/52205] unwinding through signal handler fails

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #9 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-02-14 22:04:35 UTC ---
> I'm currently doing so, and preliminary results look good.  I've got a
> couple of comments on the patch as-is which I find harder than necessary
> to read/understand:
> 
> * It uses far too magic numbers unnecessarily and/or without
>   explanation.  E.g. looking for the caller frame with hardcoded offsets
>   (cuh_patter, sah_address) is unnecessary: just use struct frame in
>   <sys/frame.h> for that, where fr_savfp is the saved frame pointer at
>   that exact offset.
> 
> * All the patterns should be acompanied by the disassembly to make it
>   easier to compare with that you see e.g. in a debugger.
> 
> * Talking of Solaris 8 in Solaris Containers is unnecessary and
>   incomplete: this is just the Solaris 8 alternate thread library (in
>   /usr/lib/lwp) which is copied over /usr/lib/libthread.so.1 when
>   booting the branded zone, but is also used by gcc on Solaris 8 by
>   default.  This alternate thread library became the default in Solaris
>   9, thus they are the same.

Fine with me (I won't make any of these changes myself though).

[Bug target/52205] unwinding through signal handler fails

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |NEW
         AssignedTo|ebotcazou at gcc dot        |unassigned at gcc dot
                   |gnu.org                     |gnu.org

[Bug target/52205] unwinding through signal handler fails

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #11 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-02-15 08:13:29 UTC ---
Author: ebotcazou
Date: Wed Feb 15 08:13:22 2012
New Revision: 184256

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=184256
Log:
    PR target/51921
    PR target/52205
    * config/sparc/sol2-unwind.h (sparc64_is_sighandler): Add support for
    Solaris 11 and slightly reformat.
    (sparc_is_sighandler): Likewise.

Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/config/sparc/sol2-unwind.h

[Bug target/52205] unwinding through signal handler fails

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #10 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-02-15 08:13:20 UTC ---
Author: ebotcazou
Date: Wed Feb 15 08:13:09 2012
New Revision: 184255

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=184255
Log:
    PR target/51921
    PR target/52205
    * config/sparc/sol2-unwind.h (sparc64_is_sighandler): Add support for
    Solaris 11 and slightly reformat.
    (sparc_is_sighandler): Likewise.

Modified:
    trunk/libgcc/ChangeLog
    trunk/libgcc/config/sparc/sol2-unwind.h

[Bug target/52205] unwinding through signal handler fails

[ro at CeBiTec dot Uni-Bielefeld.DE]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

--- Comment #12 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> 2012-02-16 19:46:02 UTC ---
> Fine with me (I won't make any of these changes myself though).

I'll probably give it a whirl, but only after 4.7 has branched.  For
4.8, there might be considerably simplifications possible since the old
MxN libthread is gone then.

    Rainer

[Bug target/52205] unwinding through signal handler fails

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

--- Comment #13 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-05-26 13:49:41 UTC ---
.

[Bug target/52205] unwinding through signal handler fails

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52205

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #14 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-05-26 14:02:12 UTC ---
.