[content deleted]

Index > Windows > [content deleted]

Author

Thread

asmcoder

Joined: 02 Jun 2008
Posts: 784

asmcoder 27 Oct 2008, 16:28

[content deleted]

Last edited by asmcoder on 14 Aug 2009, 14:55; edited 1 time in total

27 Oct 2008, 16:28

LocoDelAssembly
Your code has a bug

Joined: 06 May 2005
Posts: 4624
Location: Argentina

LocoDelAssembly 27 Oct 2008, 18:52

http://board.flatassembler.net/topic.php?p=83613#83613

27 Oct 2008, 18:52

Hrstka

Joined: 05 May 2008
Posts: 63
Location: Czech republic

Hrstka 29 Oct 2008, 19:36

Perhaps you mean something like spinlock?

29 Oct 2008, 19:36

asmcoder

Joined: 02 Jun 2008
Posts: 784

asmcoder 16 Nov 2008, 09:47

[content deleted]

Last edited by asmcoder on 14 Aug 2009, 14:55; edited 1 time in total

16 Nov 2008, 09:47

asmfan

Joined: 11 Aug 2006
Posts: 392
Location: Russian

asmfan 16 Nov 2008, 13:26

xchg doesn't need lock prefix at all. this instruction is special about lock.

16 Nov 2008, 13:26

Hrstka

Joined: 05 May 2008
Posts: 63
Location: Czech republic

Hrstka 17 Nov 2008, 01:38

Have you read http://en.wikipedia.org/wiki/Spinlock#Example_implementation properly?

Code:

lock:                       ; The lock variable. 1 = locked, 0 = unlocked.
    dd      0

The lock variable is defined somewhere in YOUR program, not in the kernel. You also have to change the name of the variable, otherwise fasm will complain. Then you do this:

Code:

call spin_lock
; now you can manipulate your structure without any harm
; ...
call spin_unlock

and that's all.

Quote:

xchg doesn't need lock prefix at all.

Right, it's unnecessary in this case, if the lock prefix is there the CPU just ignores it.

17 Nov 2008, 01:38

bitRAKE

Joined: 21 Jul 2003
Posts: 4224
Location: vpcmpistri

bitRAKE 17 Nov 2008, 04:23

Code:

lock:    ; 0 = locked, address = unlocked
    dd struct_address

Here is an alternate method...

Code:

    ; here is the spin...
    xor eax,eax
@@: xchg eax,[lock]
    test eax,eax
    je @B


; now use address in EAX


    ; should put a serializing instruction here for safety
    mov [lock],eax ; put it back when done (unlock)

...unlock doesn't need to be atomic, but all use of the structure should be complete. Serialization forces all instructions to retire prior to lock update. This should be faster, imho. (note: lock needs to be aligned to prevent spliting the write.)

17 Nov 2008, 04:23

asmfan

Joined: 11 Aug 2006
Posts: 392
Location: Russian

asmfan 17 Nov 2008, 09:47

bitRAKE, it's wrong method of acquiring spinlock. it should be done by lock + cmpxchg(8/16b). test eax,eax can be already invalid.
also i doublt that xchg need memory aligned. if it'is aligned even mov is atomic write instruction (on latest processors).

17 Nov 2008, 09:47

Feryno

Joined: 23 Mar 2005
Posts: 515
Location: Czech republic, Slovak republic

Feryno 17 Nov 2008, 11:49

I like this way (simple, no registers destroyed, the PAUSE optimalization for newer CPUs)

Code:

call acquire_spinlock
; ... do the job
call release_spinlock
ret

; subprocedures:

acquire_spinlock_pause:
 pause
align 10h
acquire_spinlock:
 lock bts dword [spinlock],0
 jc      acquire_spinlock_pause
      ret

align 10h
release_spinlock:
       lock btr dword [spinlock],0
 ret

; data:

spinlock dd 0

17 Nov 2008, 11:49

asmcoder

Joined: 02 Jun 2008
Posts: 784

asmcoder 17 Nov 2008, 13:47

[content deleted]

Last edited by asmcoder on 14 Aug 2009, 14:55; edited 1 time in total

17 Nov 2008, 13:47

asmfan

Joined: 11 Aug 2006
Posts: 392
Location: Russian

asmfan 17 Nov 2008, 15:34

yes i was wrong. xchg is good for this.
cmpxchg is for more complicated situations.
pause - it's good point for spin-lock loops.

17 Nov 2008, 15:34

asmcoder

Joined: 02 Jun 2008
Posts: 784

asmcoder 17 Nov 2008, 16:04

[content deleted]

Last edited by asmcoder on 14 Aug 2009, 14:55; edited 1 time in total

17 Nov 2008, 16:04

asmfan

Joined: 11 Aug 2006
Posts: 392
Location: Russian

asmfan 17 Nov 2008, 16:34

Suppose it was to me, then you can "safe" UTFG & RTFM lil script-kiddie malware writer. Your posts here on this forum are so pitiful that sometimes i think that you're not a programmer.
if not, please use google and read something.

17 Nov 2008, 16:34

< Last Thread | Next Thread >

Forum Rules:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum