reactos-kernel

[ros-kernel] Problems with the last changes (mdl.c..)

[Hartmut B. <har...@te...>]

Hi,

I've some problems with the last changes. Some weeks (month ?) ago, I've 
changed the parameter in subsys/smss/init.c line 374 from FALSE to TRUE. 
This means, the registry files are build, if they do not exist. With the 
last changes and existing registry files, ros does crash. Without the 
registry files, ros builds the registry files and does boot without any 
problems. I've remove the last changes (mdl.c, aspace.c ..) and there is no 
problem. What's wrong between the registry and the last changes ?

- Hartmut



  

Re: [ros-kernel] Problems with the last changes (mdl.c..)

[<we...@cw...>]

Quoting Hartmut Birr <har...@te...>:
> I've some problems with the last changes. Some weeks (month ?) ago, I've
> changed the parameter in subsys/smss/init.c line 374 from FALSE to TRUE.
> This means, the registry files are build, if they do not exist. With the
> last changes and existing registry files, ros does crash. Without the 
> registry files, ros builds the registry files and does boot without any
> problems. I've remove the last changes (mdl.c, aspace.c ..) and there is
> no problem. What's wrong between the registry and the last changes ?
> 
I don't know. The crash report you posted shows the cr3 value as zero so
most likely either the thread (or its parent process) being switched to
is corrupt. The mdl changes set up a large (256MB) region in the kernel address 
space so it may be causing some problems just by shuffling pointer values 
around. The parent process seems to be c0238a0a so if the system you
debugging on can set memory breakpoints it would be interesting to see what
happens to the PageDirectoryBase member; if this thread has been run before
then a pointer to it will be on the stack otherwise Ki386ContextSwitch
could be changed (e.g. load the process pointer into edi rather than ebx) so it 
is possible to read it off from an exception report.