I'm studying relocation of ELF for MIPS arch.
But I'm a little confused at the calculation of `S`, which is refered in
"SYSTEM V APPLICATION BINARAY INTERFACE MIPS RISC Processor Supplement
3rd Edition".
For example, I get a relocation ELF executable "a.ex_".
mips-elf-objdump -r a.ex_
----------------------------
RELOCATION RECORDS FOR [.text]:
OFFSET TYPE VALUE
00000010 R_MIPS_HI16 .rodata
00000014 R_MIPS_LO16 .rodata
...
mips-elf-objdump -dS a.ex_
----------------------------
static int enc(const int c)
{
0: 27bdfff8 addiu sp,sp,-8
4: afbe0000 sw s8,0(sp)
8: 03a0f021 move s8,sp
c: afc40008 sw a0,8(s8)
return BASE64_ARRAY[c];
10: 3c020000 lui v0,0x0
14: 2443001c addiu v1,v0,28
...
Here, relocation should be applied to location 0x10 and 0x14.
The loader places the segment '.rodata' in 0x00435000 in run time.
Intuitively, relocation will be done like this:
Original sh_addr Final sh_addr
---------------------------------
0 0x00435000
The location 0x1c will be relocated to 0x0043501c. So the instructions
at 0x10 and 0x14 will be fixed as:
10: 3c020043 lui v0,0x0043
14: 2443501c addiu v1,v0,0x501c
The document says
S Represents the value of the symbol whose index resides in the relocation
entry, unless the the symbol is STB_LOCAL and is of type
STT_SECTION in which case S represents the original sh_addr minus
the final sh_addr.
Calculation looks like:
S = 0 - 0x00435000 = 0xffbcb000
AHL = 0<<16 + 0x1c = 0x0000001c
R_MIPS_HI16=((S+AHL) - (short)(S+AHL)) >> 16 = 0xffbd
R_MIPS_LO16=(S+AHL) = 0xb01c
The results seem incorrect. The point is, whether S is equal to
final_sh_addr-original_sh_addr or S is equal to original_sh_addr-final_sh_addr.
Best Regards.
Pan Ruochen
2007/8/28
