Hi, All:
I have built the linux kernel and the glibc for my R3000 CPU. Because
the /sbin/init could not be lunched, I wrote the following program to
replace the init program:
liu.c:
#include <stdio.h>
#include <string.h>
main()
{
char s[20];
strcpy(s,"RUNNING");//failed
}
After built it, I got the following by using objdump:
liu: file format elf32-bigmips
Disassembly of section .text:
00000000004000c0 <main>:
4000c0: 3c1c0fc2 lui $gp,0xfc2
4000c4: 279c8090 addiu $gp,$gp,-32624
4000c8: 0399e021 addu $gp,$gp,$t9
4000cc: 27bdffc0 addiu $sp,$sp,-64
4000d0: afbc0010 sw $gp,16($sp)
4000d4: afbf0038 sw $ra,56($sp)
4000d8: afbe0034 sw $s8,52($sp)
4000dc: afbc0030 sw $gp,48($sp)
4000e0: 03a0f021 move $s8,$sp
4000e4: 27c40018 addiu $a0,$s8,24
4000e8: 8f858018 lw $a1,-32744($gp) <========
failed here !
4000ec: 00000000 nop
4000f0: 24a500b0 addiu $a1,$a1,176
4000f4: 8f99802c lw $t9,-32724($gp)
4000f8: 00000000 nop
4000fc: 0320f809 jalr $t9
400100: 00000000 nop
400104: 8fdc0010 lw $gp,16($s8)
400108: 03c0e821 move $sp,$s8
40010c: 8fbf0038 lw $ra,56($sp)
400110: 8fbe0034 lw $s8,52($sp)
400114: 03e00008 jr $ra
400118: 27bd0040 addiu $sp,$sp,64
40011c: 00000000 nop
0000000000400120 <strcpy>:
400120: 3c1c0fc2 lui $gp,0xfc2
400124: 279c8030 addiu $gp,$gp,-32720
400128: 0399e021 addu $gp,$gp,$t9
40012c: 00851023 subu $v0,$a0,$a1
400130: 2446ffff addiu $a2,$v0,-1
400134: 90a30000 lbu $v1,0($a1)
400138: 24a50001 addiu $a1,$a1,1
40013c: 00a61021 addu $v0,$a1,$a2
400140: a0430000 sb $v1,0($v0)
400144: 00031e00 sll $v1,$v1,0x18
400148: 1460fffa bnez $v1,400134 <strcpy+14>
40014c: 00801021 move $v0,$a0
400150: 03e00008 jr $ra
...
Here is the screen output:
Here 15: retval=0
[init:1:004000c0:0:004000c0:00001000]
in handle_mm_fault()
In do_no_page: calling nopage()
In filemap_nopage()
In do_no_page: called nopage()
calling: update_mmu_cache()
[HIT]
called: update_mmu_cache()
[init:1:0fc100a8:0:004000e8:0003f040]
Unable to handle kernel paging request at virtual address 0fc100a8, epc
== 004000e8
Oops: 0000
$0 : 00000000 1000fc00 00000000 00000000 7fffff68 00000000 00000000
00000000
$8 : 0000fc00 00000010 00000000 00001fe7 80257970 00000001 1000fc01
00000060
$16: 80241598 8020d7a4 80241570 801081c0 00000000 00000000 00000000
00000000
$24: 0000000a 00000000 0fc18090 7fffff50 7fffff50
00000000
epc : 004000e8
Status: 0000fc00
Cause : 00000008
Process init (pid: 1, stackpage=83ff8000)
Stack: 00000fd4 00000fd5 00000fd6 00000fd7 0fc18090 00000fd9 00000fda
00000fdb
00000fdc 00000fdd 00000fde 00000fdf 0fc18090 800938c0 00000000 00000fe3
00000001 7fffffda 00000000 7fffffdd 7fffffde 00000000 00000010 00000000
00000000 00000000 00000fee 00000fef 00000ff0 00000ff1 00000ff2 00000ff3
00000ff4 00000ff5 00000ff6 00000ff7 00000ff8 00000ff9 00000ffa 00000ffb
00000ffc ...
CaCode: afbc0030 03a0f021 27c40018 <8f858018> 00000000 24a500b0
8f99802c 00000000
As we know, $gp is managed by the OS. My question is who assign value to
register t9 (i.e. $25) and where? Why $gp was given 0xfc2 and
then added by -32624 ? Because $gp and $t9 gave a wrong address
0fc100a8, the CPU give a page fault and the OS said that the address is
not
GROWDOWN and faild to continue.
I think it related to my glibc but I do not know the exact place.
If you had this kind problem before, please share your knowlage with me.
Thanks,
Steven Liu
