KeRundownThread question

[Hartmut Birr]

Hi,

in KeRundownThread  is an ASSERT statement. What is the reason for that?
ApcDisable is never changed. It is always 1 for mutex objects and always
0 for mutant objects. If a mutant object is on the list, ros does crash.

- Hartmut

Index: ntoskrnl/ke/kthread.c
===================================================================
--- ntoskrnl/ke/kthread.c       (revision 14297)
+++ ntoskrnl/ke/kthread.c       (working copy)
@@ -360,7 +360,7 @@

/* Get the Mutant */
         Mutant = CONTAINING_RECORD(CurrentEntry, KMUTANT, MutantListEntry);
-        ASSERT(Mutant->ApcDisable);
+//        ASSERT(Mutant->ApcDisable);

[Alex Ionescu]

Hartmut Birr wrote:
> Hi,
>
> in KeRundownThread  is an ASSERT statement. What is the reason for that?
> ApcDisable is never changed. It is always 1 for mutex objects and always
> 0 for mutant objects. If a mutant object is on the list, ros does crash.
> Hartmut
> Index: ntoskrnl/ke/kthread.c
>
> --- ntoskrnl/ke/kthread.c       (revision 14297)
> +++ ntoskrnl/ke/kthread.c       (working copy)
> @@ -360,7 +360,7 @@
>     /* Get the Mutant */
>     Mutant = CONTAINING_RECORD(CurrentEntry, KMUTANT, MutantListEntry);
>    ASSERT(Mutant->ApcDisable);
>
> +//        ASSERT(Mutant->ApcDisable);

It should actually be a bugcheck with THREAD_TERMINATE_HELD_MUTEX, but I 
only discovered this weird Bugcode much later after coding this. See 
some information here:

http://groups-beta.google.com/group/comp.os.ms-windows.nt.misc/browse_frm/th...

here:

/
// MessageId: *THREAD_TERMINATE_HELD_MUTEX*
//
// MessageText:
//
//  A kernel thread terminated while holding a mutex
//

Best regards,
Alex Ionescu

[Hartmut Birr]

Alex Ionescu wrote:
> It should actually be a bugcheck with THREAD_TERMINATE_HELD_MUTEX, but
> I only discovered this weird Bugcode much later after coding this. See
> some information here:
>
> http://groups-beta.google.com/group/comp.os.ms-windows.nt.misc/browse_frm/th...
>
> here:
>
> /
> // MessageId: *THREAD_TERMINATE_HELD_MUTEX*
> //
> // MessageText:
> //
> //  A kernel thread terminated while holding a mutex
> //
>
> Best regards,
> Alex Ionescu

Exist there a way to check, if the mutex is only used in user mode? A
user mode thread can create and lock a mutex. An other thread terminates
this thraed and ros does crash.

- Hartmut

[Alex Ionescu]

Hartmut Birr wrote:
> Exist there a way to check, if the mutex is only used in user mode? A
> user mode thread can create and lock a mutex. An other thread terminates
> this thraed and ros does crash.
> Hartmut

Hi,

How can a user-mode thread create a mutex?

Best regards,
Alex Ionescu

[Hartmut Birr]

Alex Ionescu wrote:
> How can a user-mode thread create a mutex?

Simply with CreateMutex. I change the assert statement to:

ASSERT(Mutant->Header.Size == sizeof(KMUTANT)/sizeof(ULONG));

This does ros crash only for mutex objects and not for mutant objects.

- Hartmut

[Alex Ionescu]

Hartmut Birr wrote:
> Alex Ionescu wrote:
> > How can a user-mode thread create a mutex?
>
> Simply with CreateMutex.

CreateMutex creates a Mutant....
> I change the assert statement to:
>
> ASSERT(Mutant->Header.Size == sizeof(KMUTANT)/sizeof(ULONG));
>
> This does ros crash only for mutex objects and not for mutant objects.

That's what the assert should do as well. Mutants have APC Disable = 0. 
Mutexes have it =1. Usermode can only create mutant, so apcdisable will 
always be 0.

I don't understand how user-mode can create a mutex?
> Hartmut

Best regards,
Alex Ionescu

[Hartmut Birr]

Alex Ionescu wrote:
> That's what the assert should do as well. Mutants have APC Disable =
> 0. Mutexes have it =1. Usermode can only create mutant, so apcdisable
> will always be 0.

This means the ASSERT statemant is wrong. The correct statement is:

ASSERT(Mutant->ApcDisable == 0);

- Hartmut

[Alex Ionescu]

Hartmut Birr wrote:
> Alex Ionescu wrote:
> > That's what the assert should do as well. Mutants have APC Disable =
> > 0. Mutexes have it =1. Usermode can only create mutant, so apcdisable
> > will always be 0.
>
> This means the ASSERT statemant is wrong. The correct statement is:
>
> ASSERT(Mutant->ApcDisable == 0);
> Hartmut

Hi,

Right, sorry for not seeing this! Can you please fix it, my kthread.c is 
completely modified due to my local scheduler rewrite.
Thank you!

Best regards,
Alex Ionescu