[ros-dev] Re:[ros-svn] [ion] 12748: Fix incorrect LPC Object export...we export pointers not the object itself! Also remove more TIME stuf

[Hartmut Birr]

Filip Navara schrieb:

> Filip Navara wrote:
>
>> Hartmut Birr wrote:
>>
>>> I think that Alex's changes are correct. If Alex's changes are 
>>> wrong, all other type object initialisations are also wrong. There 
>>> is only one little bug, the EXPORTED/IMPORTED definition is missing.
>>
>>
>> Hmm, you might be right...
>
>
> I have searched the net a bit and looked into my books.
>
> Windows NT Filesystem Internals:
> "...at system initialization time, the NT I/O Manager registers all 
> the different I/O Manager objects (including the file object 
> structure) with the NT Object Manager. The ObCreateObjectType() Object 
> Manager routine is used for this purpose. Although this routine is not 
> exposed by the NT Executive, it serves to make the NT Object Manager 
> aware of a new object type. When invoking this routine, the I/O 
> Manager also supplies the functions that must be invoked by the Object 
> Manager to manipulate the object being defined. For file object 
> structures, the I/O Manager supplies an internal routine called 
> IopCloseFile() to be invoked whenever any handle associated with the 
> file object has been closed."
>
> The ObCreateObjectType function has prototype similar to this:
>
> and it's used like
>
>    RtlZeroMemory(&ObjectTypeInitializer, 
> sizeof(OBJECT_TYPE_INITIALIZER));
>    ObjectTypeInitializer.Length = sizeof(OBJECT_TYPE_INITIALIZER);
>    ...
>    RtlInitUnicodeString(&ObjectTypeName, L"Port");
>    ObCreateObjectType(&ObjectTypeName, &ObjectTypeInitializer, 0, 
> &LpcPortObjectType);
>
> so yes, the exported *ObjectType variables are really of type 
> POBJECT_TYPE.
>
> My apologies to Alex.
>
> Regards,
> Filip
>
>
> _______________________________________________
> Ros-dev mailing list
> Ros-dev at reactos.com
> http://reactos.com:8080/mailman/listinfo/ros-dev
>
>
Hi,

I've add some code to my test driver:

  Status = ZwCreateEvent(&hEvent, EVENT_ALL_ACCESS, NULL, 
NotificationEvent, FALSE);
  DPRINT1("%x\n", Status);
  if (NT_SUCCESS(Status))
  {
     Status = ObReferenceObjectByHandle(hEvent, EVENT_ALL_ACCESS, 
(POBJECT_TYPE)ExEventObjectType, KernelMode, (PVOID*)&Event, NULL);
     DPRINT1("%x\n", Status);
     if (NT_SUCCESS(Status))
     {
        ObDereferenceObject((PVOID)Event);
     }
     Status = ObReferenceObjectByHandle(hEvent, EVENT_ALL_ACCESS, 
(POBJECT_TYPE)&ExEventObjectType, KernelMode, (PVOID*)&Event, NULL);
     DPRINT1("%x\n", Status);
     if (NT_SUCCESS(Status))
     {
        ObDereferenceObject((PVOID)Event);
     }
     ZwClose(hEvent);
  }

The result on W2K is:

(memtest.c:185) 0
(memtest.c:189) 0
(memtest.c: 195) c0000024 -> STATUS_OBJECT_TYPE_MISMATCH

This means Alex's changes and the other object type initialisations are 
correct.

- Hartmut