[Bug 40311] New: Processing of 'DeviceIoControl' calls doesn't corresponds to specification

https://bugs.winehq.org/show_bug.cgi?id=40311

Bug ID: 40311
           Summary: Processing of 'DeviceIoControl' calls doesn't
                    corresponds to specification
           Product: Wine
           Version: unspecified
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: ntoskrnl
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

As described in specification 'Buffer Descriptions for I/O Control Codes' for
'METHOD_IN_DIRECT' or 'METHOD_OUT_DIRECT' transfer types the data buffer, to
which points the pointer passed as 'lpOutputBuffer' argument into
'DeviceIoControl', must be accessible from driver. In other words, when in
control code is specified 'METHOD_IN_DIRECT' tranfer type, the 'lpOutputBuffer'
buffer may be used as the additional buffer of data that driver may transfer to
device. This feature is used in software and drivers for Hantek Oscilloscopes.
The pointer 'lpOutputBuffer' and argument 'nOutBufferSize', which are passed to
'DeviceIoControl' function, must are used as corresponding values for fields
'StartVa', 'ByteOffset' and 'ByteCount' in 'irp->MdlAddress' structure.

But Wine allocates new output buffer that are passed to
'IoBuildDeviceIoControlRequest' (see function 'dispatch_ioctl' in source file
'dlls/ntoskrnl.exe/ntoskrnl.c'). Thus data stored in output buffer aren't
available for driver.

Specification:
https://msdn.microsoft.com/en-us/library/windows/hardware/ff540663(v=vs.85)....

[Bug 40311] Processing of 'DeviceIoControl' calls doesn't corresponds to specification

https://bugs.winehq.org/show_bug.cgi?id=40311

[e-mail] changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 40311] Processing of 'DeviceIoControl' calls doesn' t corresponds to specification

https://bugs.winehq.org/show_bug.cgi?id=40311

Sebastian Lackner  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #1 from Sebastian Lackner  ---
I have sent patches for this issue:

http://source.winehq.org/patches/data/127128
http://source.winehq.org/patches/data/127129
http://source.winehq.org/patches/data/127130
http://source.winehq.org/patches/data/127131

[Bug 40311] Processing of 'DeviceIoControl' calls doesn' t corresponds to specification

https://bugs.winehq.org/show_bug.cgi?id=40311

--- Comment #2 from [e-mail] ---
(In reply to Sebastian Lackner from comment #1)
> I have sent patches for this issue:
>
> http://source.winehq.org/patches/data/127128
> http://source.winehq.org/patches/data/127129
> http://source.winehq.org/patches/data/127130
> http://source.winehq.org/patches/data/127131

At the moment I temporarily solved my problem. But my solution is very dirty
hack (patch
https://github.com/AlexUg/hantekdso.sys/blob/master/hantekdso.sys/resources/...).

Thank you very much for your work. I hope your patches will be included in the
next release.

[Bug 40311] Processing of 'DeviceIoControl' calls doesn' t corresponds to specification

https://bugs.winehq.org/show_bug.cgi?id=40311

--- Comment #3 from Sebastian Lackner  ---
I wasn't aware of this patch, but actually it looks like your solution is not
really that much different from my one. My proposed patch also transfers both
the input and output buffer as a big chunk of data. In my opinion, all other
alternative (for example transferring only the addresses, and then implement
mapping on top of ReadProcessMemory) would require a much bigger overhead. Wine
is also planning to use the driver framework for input devices, so multiple
wineserver calls back and forth would be bad.

Regarding your project, please note that there is actually ongoing effort to
implement Plug&Play support, HID support, and later also USB support. I'm not
sure if your code is too much specialized for one specific USB Oscilloscope,
but probably some parts (for example libusb handling) could also be useful for
Wine. ;)

[Bug 40311] Processing of 'DeviceIoControl' calls doesn't corresponds to specification

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40311

Anastasius Focht  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
            Version|unspecified                 |1.9.5
             Status|UNCONFIRMED                 |NEEDINFO
                 CC|                            |[e-mail]

--- Comment #4 from Anastasius Focht  ---
Hello folks,

what's the status here? What happened to Sebastian's patches?

Regards

[Bug 40311] Processing of 'DeviceIoControl' calls doesn't corresponds to specification

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40311

--- Comment #5 from [e-mail] ---
The bug is currently fixed. If I'm not mistaken, since version 3.4 (I don't
remember exactly).

[Bug 40311] Processing of 'DeviceIoControl' calls doesn't corresponds to specification

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40311

Anastasius Focht  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEEDINFO                    |RESOLVED
      Fixed by SHA1|                            |b466e893116af5007a1302aef56
                   |                            |030f150884b75
         Resolution|---                         |FIXED

--- Comment #6 from Anastasius Focht  ---
Hello,

--- quote ---
The bug is currently fixed. If I'm not mistaken, since version 3.4 (I don't
remember exactly).
--- quote ---

thanks. Indeed, I should have checked the git history from the time Sebastian
commented here plus some extra months ;-)

Here we go.

--- snip ---
$ git log --pretty=format:"%h %an %ad %s" --since "2016-10-16" --until
"2016-12-01" --author "Sebastian" -- dlls/ntoskrnl.exe

b466e89311 Sebastian Lackner Sun Oct 16 08:57:21 2016 +0200 ntoskrnl.exe:
Explicitly set MappedSystemVa for MDLs.
9912ef778f Sebastian Lackner Sun Oct 16 08:53:29 2016 +0200 ntoskrnl.exe: Use
MmInitializeMdl in IoAllocateMdl.
eb7ac554d1 Sebastian Lackner Sun Oct 16 08:51:06 2016 +0200 ntoskrnl.exe: Add
support for METHOD_IN_DIRECT/METHOD_OUT_DIRECT ioctls.
5789b94f19 Sebastian Lackner Sun Oct 16 08:48:55 2016 +0200 ntoskrnl.exe: Defer
deallocation of in_buff in dispatch_ioctl.
--- snip ---

Using commit
https://source.winehq.org/git/wine.git/commitdiff/b466e893116af5007a1302aef5...
("ntoskrnl.exe: Explicitly set MappedSystemVa for MDLs.") as resolution. Part
of Wine 1.9.22 release.

Thanks Sebastian

Regards

[Bug 40311] Processing of 'DeviceIoControl' calls doesn't corresponds to specification

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40311

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #7 from Alexandre Julliard  ---
Closing bugs fixed in 5.8.