This is the mail archive of the mailing list for the binutils project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

o32 links ...

I just tried to link ``main(){}'' under Linux and it somehow runs in an
endless loop:

Program received signal SIGINT, Interrupt.
0x4018c0 in __frame_state_for (pc_target=0x7fffff77, state_in=0x2acc1c08)
(gdb) info registers 
          zero       at       v0       v1       a0       a1       a2       a3
 R0   00000000 0fb82ea0 00000000 00000000 00000000 00000000 00000000 00000000 
            t0       t1       t2       t3       t4       t5       t6       t7
 R8   ffffffff 00000000 10000000 f7ffffff 08000000 04849145 7ffffa10 7ffffa0c 
            s0       s1       s2       s3       s4       s5       s6       s7
 R16  0fb8b380 004006e0 7ffff9c0 7fffff77 2acc1c08 7fffff77 00000000 7fffe9a0 
            t8       jp       k0       k1       gp       sp       fp       ra
 R24  7ffffa10 004018bc 7ffffe4c 00000000 100082f0 7ffffd20 7ffff700 004018c8 
            pc    cause badvaddr       lo       hi      fsr      fir
      004018c0 00000400 8df8b938 cccccccd 00000000 00000000 00002310 

  401898:       afbf01f0        sw      $ra,496($sp)
  40189c:       afbc01ec        sw      $gp,492($sp)
  4018a0:       afb201e0        sw      $s2,480($sp)
  4018a4:       afb101dc        sw      $s1,476($sp)
  4018a8:       afb001d8        sw      $s0,472($sp)
  4018ac:       00a0a021        move    $s4,$a1
  4018b0:       8f998070        lw      $t9,-32656($gp)
  4018b4:       00000000        nop
  4018b8:       27390e40        addiu   $t9,$t9,3648
  4018bc:       00000000        nop
This is where $25 points to.
  4018c0:       0320f809        jalr    $t9
  4018c4:       00000000        nop
  4018c8:       8fbc0010        lw      $gp,16($sp)
  4018cc:       00409021        move    $s2,$v0
  4018d0:       1240000c        beqz    $s2,401904 <__frame_state_for+0x8c>
  4018d4:       02402021        move    $a0,$s2
  4018d8:       27a50018        addiu   $a1,$sp,24
  4018dc:       8f998070        lw      $t9,-32656($gp)
  4018e0:       00000000        nop

Here disassembly & relocations from libgcc.a, function __frame_state_for:

     eb0:       8f990000        lw      $t9,0($gp)
                        eb0: R_MIPS_GOT16       .text
     eb4:       00000000        nop
     eb8:       27390440        addiu   $t9,$t9,1088
                        eb8: R_MIPS_LO16        .text
     ebc:       00000000        nop
     ec0:       0320f809        jalr    $t9

The values inserted for the GOT16 / LO16 relocations in the final link
don't make any sense at all to me and are completly different from the
2.8.1 linker output.

The function is __frame_state_for() from egcs 1.0.3a frame.c as it's
trying to call the static function find_fde.

Is there some facility which allows dumping of the got's contents in a
reasonably informative format?

Hmm...  The linker used to default to IRIX5-style shared libs which is
what Linux uses right now, just if this makes a difference.

This is today's CVS binutils + your elflink.h patch.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]