[ros-kernel] Can someone give me a hand in here

[Skywing]

All of the x86 Windows calling conventions (that is,
__cdecl/__fastcall/___stdcall/__thiscall  variants) specify that the only
nonvolatile registers are ecx, edx, and eax.  Furthermore, eax [and possibly
edx for 64-bit results] are used to return values in all x86 Windows calling
conventions.  So, the compiler will not save any other registers before
calling your function.  If you overwrite those registers without restoring
their previous values before returning control, a variety of corruption
problems can result.

-----Original Message-----
From: ros-kernel-bounces at reactos.com [mailto:ros-kernel-bounces at reactos.com]
On Behalf Of Waldo Alvarez
Sent: Thursday, December 18, 2003 11:25 AM
To: ReactOS Kernel List
Subject: RE: [ros-kernel] Can someone give me a hand in here

On Wed, 17 Dec 2003, Skywing wrote:

> You must never read past the terminating null, *especially* in kernel
mode.
> If the string is on a page boundary, that could cause a crash.

Then I suppose I have to check for such a case, using an extra register or 
intentionally putting a 0 at page boundaries saving in the stack that 
final data, and walking through pages in such a way. There will be just a 
very ligth performance degradation.
 
> The only registers that you don't need to save (on x86) are eax, ecx, and
> edx.

Yes I tested it saving ebx and still the same. Thanks for the info. But 
what's the reason? The compiler checks for it before the 
function is called everytime?

> -----Original Message-----
> From: ros-kernel-bounces at reactos.com
[mailto:ros-kernel-bounces at reactos.com]
> On Behalf Of Waldo Alvarez
> Sent: Wednesday, December 17, 2003 1:21 PM
> To: ReactOS Kernel List
> Subject: [ros-kernel] Can someone give me a hand in here
> 
> Hello:
> 
> I was changing this file in ROS to speedup strlen but ReactOS 
> crashes with a page fault very quick since strlen is used almost 
> everywhere. But I'm not used to this kind of assembler format. Can someone

> take a look at it. It could read at most 3 bytes beyond the ending 0 in 
> the string but i don't think it could be a problem since the page size is 
> 4 kb the other reason could be the ebx register I'm not saving. as opposed

> to the previous function.
> 
> Thanks in advance
> Waldo Alvarez
> 
> 
> _______________________________________________
> Ros-kernel mailing list
> Ros-kernel at reactos.com
> http://reactos.com/mailman/listinfo/ros-kernel
> 

_______________________________________________
Ros-kernel mailing list
Ros-kernel at reactos.com
http://reactos.com/mailman/listinfo/ros-kernel