[Bug libfortran/25377] New: libgfortran/io/transfer.c:2159 unlock_unit causes segfault

[Bug libfortran/25377] New: libgfortran/io/transfer.c:2159 unlock_unit causes segfault

[jb at gcc dot gnu dot org]

I tried to use gfortran together with a library that uses pthreads. This caused
a segfault. Note that my own code didn't directly call pthreads nor use openmp
(this is trunk, not gomp branch).

~/src/benchmark/matmul-bench% gfortran -g -O0 matmul-bench.f90 -o mbgfc-goto
-lgoto -lpthread

~/src/benchmark/matmul-bench% gdb ./mbgfc-goto
GNU gdb 6.3-debian
Copyright 2004 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "i486-linux-gnu"...Using host libthread_db library
"/lib/tls/i686/cmov/libthread_db.so.1".

(gdb) b transfer.c:2159
No source file named transfer.c.
Make breakpoint pending on future shared library load? (y or [n]) y

Breakpoint 1 (transfer.c:2159) pending.
(gdb) r
Starting program: /home/joib/src/benchmark/matmul-bench/mbgfc-goto
[Thread debugging using libthread_db enabled]
[New Thread 1076887104 (LWP 30593)]
Breakpoint 2 at 0x40188b98: file ../../../trunk/libgfortran/io/transfer.c, line
2159.
Pending breakpoint "transfer.c:2159" resolved
 Single precision matrix multiplication test
[Switching to Thread 1076887104 (LWP 30593)]

Breakpoint 2, *_gfortran_st_write_done (dtp=0xbfa1665c)
    at ../../../trunk/libgfortran/io/transfer.c:2159
2159        unlock_unit (dtp->u.p.current_unit);
(gdb) step
*_gfortrani_unlock_unit (u=0x804f118) at
../../../trunk/libgfortran/io/unit.c:555
555     {
(gdb)
538     gthr-posix.h: No such file or directory.
        in gthr-posix.h
(gdb)
539     in gthr-posix.h
(gdb)
557     }
(gdb)
*_gfortrani_unlock_unit (u=0xbfa1665c) at gthr-posix.h:539
539     gthr-posix.h: No such file or directory.
        in gthr-posix.h
(gdb)
0x402ed223 in in6addr_any () from /lib/tls/i686/cmov/libc.so.6
(gdb)
Single stepping until exit from function in6addr_any,
which has no line number information.

Program received signal SIGSEGV, Segmentation fault.
0x402ed231 in in6addr_any () from /lib/tls/i686/cmov/libc.so.6


-- 
           Summary: libgfortran/io/transfer.c:2159 unlock_unit causes
                    segfault
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jb at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug libfortran/25377] libgfortran/io/transfer.c:2159 unlock_unit causes segfault

[jakub at gcc dot gnu dot org]

------- Comment #1 from jakub at gcc dot gnu dot org  2005-12-13 13:10 -------
Can you attach a testcase?
The gdb session doesn't really provide any useful info.


-- 

jakub at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug libfortran/25377] libgfortran/io/transfer.c:2159 unlock_unit causes segfault

[jb at gcc dot gnu dot org]

------- Comment #2 from jb at gcc dot gnu dot org  2005-12-13 16:13 -------
Unfortunately the goto blas library where I noticed this error is not free
software. But here is a simple testcase that shows the same problem.

File hello.c:

#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#define NUM_THREADS     2

void *PrintHello(void *threadid)
{
   printf("\n%d: Hello World!\n", threadid);
   pthread_exit(NULL);
}

void fhellomt_ ()
{
   pthread_t threads[NUM_THREADS];
   pthread_attr_t attr;
   int rc, t, status;
   pthread_attr_init (&attr);
   pthread_attr_setdetachstate (&attr, PTHREAD_CREATE_JOINABLE);
   for(t=0; t<NUM_THREADS; t++){
      printf("Creating thread %d\n", t);
      rc = pthread_create(&threads[t], &attr, PrintHello, (void *)t);
      if (rc){
         printf("ERROR; return code from pthread_create() is %d\n", rc);
         exit(-1);
      }
   }
   pthread_attr_destroy (&attr);
   for(t=0; t<NUM_THREADS; t++)
   {
      rc = pthread_join(threads[t], (void **)&status);
      if (rc)
      {
         printf("ERROR; return code from pthread_join() is %d\n", rc);
         exit(-1);
      }
      printf("Completed join with thread %d status= %d\n",t, status);
   }
}


File fpthello.f90:

! Pthreads hello world
program fpthello
  implicit none
  call fhellomt ()
  print *, 'done'
end program fpthello

Compiled with:

gcc -c hello.c
gfortran fpthello.f90 hello.o -lpthread 

Running it produces:

./a.out
Creating thread 0
Creating thread 1

0: Hello World!
Completed join with thread 0 status= 0

1: Hello World!
Completed join with thread 1 status= 0
 done
zsh: segmentation fault  ./a.out


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug middle-end/25377] [4.2 Regression] weak_ref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW
          Component|libfortran                  |middle-end
     Ever Confirmed|0                           |1
           Keywords|                            |wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2006-01-29 17:42:14
               date|                            |
            Summary|libgfortran/io/transfer.c:21|[4.2 Regression] weak_ref
                   |59 unlock_unit causes       |sibcalled with -fPIC
                   |segfault                    |
   Target Milestone|---                         |4.2.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

------- Comment #3 from pinskia at gcc dot gnu dot org  2006-01-31 16:24 -------
Reduced testcase:
int f(void);
static typeof(f) g __attribute__((weakref("f")));
int h(void)
{
  return g();
}

-----
Actually is obviously why x86 goes wrong:
in ix86_function_ok_for_sibcall:
  /* If we are generating position-independent code, we cannot sibcall
     optimize any indirect call, or a direct call to a global function,
     as the PLT requires %ebx be live.  */
  if (!TARGET_64BIT && flag_pic && (!decl || TREE_PUBLIC (decl)))
    return false; 

TREE_PUBLIC Is the wrong check here. 

s390 fixed it a different correct way:
http://gcc.gnu.org/ml/gcc-patches/2005-12/msg00646.html

See the thread:
http://gcc.gnu.org/ml/gcc/2005-12/msg00718.html

Again checking TREE_PUBLIC is not the correct check really.
In fact reading tree.h, it seems so too as it is just commenting about the
name:
   nonzero means name is to be accessible from outside this module.

Which has nothing to do with the function being external.


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |target
  GCC build triplet|i686-pc-linux-gnu           |
   GCC host triplet|i686-pc-linux-gnu           |
            Summary|[4.2 Regression] weak_ref   |[4.2 Regression] weakref
                   |sibcalled with -fPIC        |sibcalled with -fPIC


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

------- Comment #4 from pinskia at gcc dot gnu dot org  2006-01-31 16:26 -------
Testing the obvious fix like what s390 did.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

------- Comment #5 from pinskia at gcc dot gnu dot org  2006-01-31 16:42 -------
And yes that fixed my reduced testcase.
The patch is:
Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c  (revision 110396)
+++ config/i386/i386.c  (working copy)
@@ -2211,7 +2211,7 @@ ix86_function_ok_for_sibcall (tree decl,
   /* If we are generating position-independent code, we cannot sibcall
      optimize any indirect call, or a direct call to a global function,
      as the PLT requires %ebx be live.  */
-  if (!TARGET_64BIT && flag_pic && (!decl || TREE_PUBLIC (decl)))
+  if (!TARGET_64BIT && flag_pic && (!decl || !targetm.binds_local_p (decl)))
     return false;

   if (decl)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

------- Comment #6 from pinskia at gcc dot gnu dot org  2006-02-01 01:17 -------
Patch posted:
http://gcc.gnu.org/ml/gcc-patches/2006-02/msg00005.html


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |http://gcc.gnu.org/ml/gcc-
                   |                            |patches/2006-
                   |                            |02/msg00005.html
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[rth at gcc dot gnu dot org]

------- Comment #7 from rth at gcc dot gnu dot org  2006-02-03 00:10 -------
*** Bug 25527 has been marked as a duplicate of this bug. ***


-- 

rth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dev at stuffit dot at


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

------- Comment #9 from pinskia at gcc dot gnu dot org  2006-02-03 03:57 -------
Fixed.


-- 

pinskia at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

------- Comment #8 from pinskia at gcc dot gnu dot org  2006-02-03 03:56 -------
Subject: Bug 25377

Author: pinskia
Date: Fri Feb  3 03:56:55 2006
New Revision: 110537

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=110537
Log:
2006-02-02  Andrew Pinski  <pinskia@physics.uc.edu>

        PR target/25377
        * config/i386/i386.c (ix86_function_ok_for_sibcall):
        Use targetm.binds_local_p instead of TREE_PUBLIC.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/i386/i386.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377

[Bug target/25377] [4.2 Regression] weakref sibcalled with -fPIC

[pinskia at gcc dot gnu dot org]

------- Comment #10 from pinskia at gcc dot gnu dot org  2006-02-05 15:01 -------
*** Bug 26104 has been marked as a duplicate of this bug. ***


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |magnus_os at yahoo dot se


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25377