GAS/GCC Assembler Code Modification
Hi folks,
I have a question regarding GAS and GCC.
Does the GAS modify the Assembler Code? If I try to assemble (for example) the following line:
popl %esi
the gas changes that to
pop %esi
Ok I guess this is ok, because the ESI Register is 'only' 32 Bit. So movw (= mov) would be ok.
Next example:
If I try to assemble
movl $0xb,%eax
the gas changes that to
mov $0xb,%eax
How can I tell gas that I dont want him to optimize the code?
Best regards,
Martin
PS: Sry for my poor english.
Quoted 8 lines · Martin
popl %esi the gas changes that to
pop %esi Ok I guess this is ok, because the ESI Register is 'only' 32 Bit. So movw (= mov) would be ok.Next example:
If I try to assemble
movl $0xb,%eax the gas changes that to
mov $0xb,%eax
There is no optimisation happening here. it's only that gas is using only mov as mnemonic, irrespective of whether you write movw or movl. I suggest you have a look at the generated instruction bytecodes.
while on linux, the default it assumes is 32 bit mode. so, though your movw also it will show as mov, but the generated instruction bytecode will also be prefixed with a mode byte, compared to had same movw been used in 16 bit mode.
Quoted 1 line
How can I tell gas that I dont want him to optimize the code?
Do you mean anything other than above as 'optimize the code' ?
I don't know if I cleared some things for you or increased the mess.
--
sandeep
--------------------------------------------------------------------------
I predict that today will be remembered until tomorrow!
--------------------------------------------------------------------------
Am Samstag, 15. Mai 2004 10:28 schrieb Martin:
Quoted 11 lines
Hi folks,
I have a question regarding GAS and GCC.
Does the GAS modify the Assembler Code? If I try to assemble (for example) the following line:popl %esi
the gas changes that to
pop %esi
though gas 'optimizes', which cannot be configured(!), in your example gas doesn't, re listing by gas, itself! (my version ex binutils 2.14.90) - apparently, your disassembler translates differently. compare the code...
the only 'optimization' by gas, which I'm aware of, is branches distance coding and selection of alignment fill code. branches coding can be forced to always 'long' (4 bytes) by declaring the destn label '.global'.
while assembling gas tries to deduce the oprands' formats by mnemonics and register nameing and, tries some correction if the proper size specs were missing, plus emanating a concerning warning (not an error!) message.
best,
hp
Signature · 1 line
mail to 'hp': lx at lxhp : in-berlin : de
Quoted 11 lines · peter willy krause
Am Samstag, 15. Mai 2004 10:28 schrieb Martin:
Quoted 9 lines
I have a question regarding GAS and GCC.
Does the GAS modify the Assembler Code? If I try to assemble (for example) the following line:popl %esi
the gas changes that to
pop %esi
Quoted 7 lines
though gas 'optimizes', which cannot be configured(!), in your example gas doesn't, re listing by gas, itself! (my version ex binutils 2.14.90) - apparently, your disassembler translates differently. compare the code...
the only 'optimization' by gas, which I'm aware of, is branches distance coding and selection of alignment fill code. branches coding can be forced to always 'long' (4 bytes) by declaring the destn label '.global'.
Alright.
Quoted 3 lines
while assembling gas tries to deduce the oprands' formats by mnemonics and register nameing and, tries some correction if the proper size specs were missing, plus emanating a concerning warning (not an error!) message.
Ok. I guess here is my error in reasoning. I just wondered because in AlephOnes Paper about Stacksmashing he printed the following code:
[...]
popl %esi
movl %esi,0x8(%esi)
movb $0x0,0x7(%esi)
movl $0x0,0xc(%esi)
movl $0xb,%eax
movl %esi,%ebx
leal 0x8(%esi),%ecx
leal 0xc(%esi),%edx
int $0x80
movl $0x1, %eax
movl $0x0, %ebx[...]
After assembling with gas and disassembling with gdb I got the following code:
[...]
pop %esi
mov %esi,0x8(%esi)
movb $0x0,0x7(%esi)
movl $0x0,0xc(%esi)
mov $0xb,%eax
mov %esi,%ebx
lea 0x8(%esi),%ecx
lea 0xc(%esi),%edx
int $0x80
mov $0x1, %eax
mov $0x0, %ebx [...]
But if you are right, the code above should be the same.
But I still have one question. Wouldn't be the code from AlephOne bigger (bytesize) than the code I got?
Best regards,
thanks for your help.
Martin
Am Samstag, 15. Mai 2004 21:08 schrieb Martin:
Quoted 2 lines
But I still have one question. Wouldn't be the code from AlephOne
^^^^^^^^^^
I've no idea of what that is. got a link?
best
hp
Signature · 1 line
mail to 'hp': lx at lxhp : in-berlin : de
Quoted 5 lines · peter willy krause
Am Samstag, 15. Mai 2004 21:08 schrieb Martin:
Quoted 2 lines
But I still have one question. Wouldn't be the code from AlephOne
^^^^^^^^^^I've no idea of what that is. got a link?
..my fault. sry.
http://www.phrack.org/show.php?p=49&a=14
(etwa in der Mitte des Textes befindet sich der Code)