Re: [ros-dev] [NTDLL/LDR] - Fix a few bugs (wrong variable usage, wrong variable initialization) which led to incorrect snapping of import address table.

[Minas Abrahamyan]

Hi Alexey,

Since you are still working on these LdrXXX functions then you maybe
find possibility
to add there some code to prevent them trapping into infinite cycle (bug 5881:
http://www.reactos.org/bugzilla/show_bug.cgi?id=5881  )

While booted up with RAM 25 or 24 MB ROS kernel falls into infinite
cycle there,
but being booted up with 15 Mbs ROS stalls and and hangs up.

To test it is enough to run it on any VM with 24Mb of RAM.

What occurs there is infinite recursive calls, in form of
LdrpLoadModule ->calls LdrFixupImports -> LdrpGetOrLoadModule ->
LdrpLoadModule and so forth.
(detailed bt is in bug 5881 description).
Maybe ROS kernel doesn't check memory availability, something like
XXXMalloc returning NULL or so.

Thanks!
-M.A.

On Wed, Mar 23, 2011 at 4:25 PM,  [e-mail] wrote:
> Author: fireball
> Date: Wed Mar 23 12:25:53 2011
> New Revision: 51123
>
> URL: http://svn.reactos.org/svn/reactos?rev=51123&view=rev
> Log:
> [NTDLL/LDR]
> Fix a few bugs (wrong variable usage, wrong variable initialization) which led to incorrect snapping of import address table.
> Wrap LdrpSnapThunk() invocations into SEH.
>
> Modified:
>    trunk/reactos/dll/ntdll/ldr/ldrpe.c
>
> Modified: trunk/reactos/dll/ntdll/ldr/ldrpe.c
> URL: http://svn.reactos.org/svn/reactos/trunk/reactos/dll/ntdll/ldr/ldrpe.c?rev=5...

[Aleksey Bragin]

Hi Minas,
thanks for watching these commits!

I have a rewrite of all these functions, which would fix this issue,  
however I can't commit it before I fix other code first.

WBR,
Aleksey Bragin.

On Mar 29, 2011, at 5:26 PM, Minas Abrahamyan wrote:
> Hi Alexey,
>
> Since you are still working on these LdrXXX functions then you maybe
> find possibility
> to add there some code to prevent them trapping into infinite cycle  
> (bug 5881:
> http://www.reactos.org/bugzilla/show_bug.cgi?id=5881  )
>
> While booted up with RAM 25 or 24 MB ROS kernel falls into infinite
> cycle there,
> but being booted up with 15 Mbs ROS stalls and and hangs up.
>
> To test it is enough to run it on any VM with 24Mb of RAM.
>
> What occurs there is infinite recursive calls, in form of
> LdrpLoadModule ->calls LdrFixupImports -> LdrpGetOrLoadModule ->
> LdrpLoadModule and so forth.
> (detailed bt is in bug 5881 description).
> Maybe ROS kernel doesn't check memory availability, something like
> XXXMalloc returning NULL or so.
>
> Thanks!
> -M.A.
>
> On Wed, Mar 23, 2011 at 4:25 PM,  [e-mail] wrote:
> > Author: fireball
> > Date: Wed Mar 23 12:25:53 2011
> > New Revision: 51123
> >
> > URL: http://svn.reactos.org/svn/reactos?rev=51123&view=rev
> > Log:
> > [NTDLL/LDR]
> > Fix a few bugs (wrong variable usage, wrong variable
> >
> > initialization) which led to incorrect snapping of import address  
> > table.
> > Wrap LdrpSnapThunk() invocations into SEH.
> >
> > Modified:
> >    trunk/reactos/dll/ntdll/ldr/ldrpe.c
> >
> > Modified: trunk/reactos/dll/ntdll/ldr/ldrpe.c
> > URL: http://svn.reactos.org/svn/reactos/trunk/reactos/dll/ntdll/ 
> > ldr/ldrpe.c?rev=51123&r1=51122&r2=51123&view=diff
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev

[Minas Abrahamyan]

Hi Alexey,
> I have a rewrite of all these functions, which would fix this issue,

So one may hope that this problem will be eventually solved? Sounds great! :)

Then if bug 6031 (
http://www.reactos.org/bugzilla/show_bug.cgi?id=6031 ) would be
solved, I will try,
then together this will eliminate very the root of the strange evil bug! :)

Best Regards,
-M.A.

On Wed, Mar 30, 2011 at 1:29 AM, Aleksey Bragin [e-mail] wrote:
> Hi Minas,
> thanks for watching these commits!
>
> I have a rewrite of all these functions, which would fix this issue, however
> I can't commit it before I fix other code first.
>
> WBR,
> Aleksey Bragin.
>
> On Mar 29, 2011, at 5:26 PM, Minas Abrahamyan wrote:
> > Hi Alexey,
> >
> > Since you are still working on these LdrXXX functions then you maybe
> > find possibility
> > to add there some code to prevent them trapping into infinite cycle (bug
> > 5881:
> > http://www.reactos.org/bugzilla/show_bug.cgi?id=5881  )
> >
> > While booted up with RAM 25 or 24 MB ROS kernel falls into infinite
> > cycle there,
> > but being booted up with 15 Mbs ROS stalls and and hangs up.
> >
> > To test it is enough to run it on any VM with 24Mb of RAM.
> >
> > What occurs there is infinite recursive calls, in form of
> > LdrpLoadModule ->calls LdrFixupImports -> LdrpGetOrLoadModule ->
> > LdrpLoadModule and so forth.
> > (detailed bt is in bug 5881 description).
> > Maybe ROS kernel doesn't check memory availability, something like
> > XXXMalloc returning NULL or so.
> >
> > Thanks!
> > -M.A.
> >
> > On Wed, Mar 23, 2011 at 4:25 PM,  [e-mail] wrote:
> > > Author: fireball
> > > Date: Wed Mar 23 12:25:53 2011
> > > New Revision: 51123
> > >
> > > URL: http://svn.reactos.org/svn/reactos?rev=51123&view=rev
> > > Log:
> > > [NTDLL/LDR]
> > > Fix a few bugs (wrong variable usage, wrong variable initialization)
> > >
> > > which led to incorrect snapping of import address table.
> > > Wrap LdrpSnapThunk() invocations into SEH.
> > >
> > > Modified:
> > >   trunk/reactos/dll/ntdll/ldr/ldrpe.c
> > >
> > > Modified: trunk/reactos/dll/ntdll/ldr/ldrpe.c
> > > URL:
> > > http://svn.reactos.org/svn/reactos/trunk/reactos/dll/ntdll/ldr/ldrpe.c?rev=5...
> >
> > Ros-dev mailing list
> > [e-mail]
> > http://www.reactos.org/mailman/listinfo/ros-dev
>
> Ros-dev mailing list
> [e-mail]
> http://www.reactos.org/mailman/listinfo/ros-dev