[PATCH v2 4/6] hidclass.sys: Send rawinput messages with HID report.

[Rémi Bernon]

On 3/5/21 6:10 PM, Alexandre Julliard wrote:
> Rémi Bernon <rbernon at codeweavers.com> writes:
> 
>> On 3/5/21 6:03 PM, Alexandre Julliard wrote:
>>> Rémi Bernon <rbernon at codeweavers.com> writes:
>>>
>>>> @@ -237,6 +238,34 @@ static NTSTATUS copy_packet_into_buffer(HID_XFER_PACKET *packet, BYTE* buffer, U
>>>>            return STATUS_BUFFER_OVERFLOW;
>>>>    }
>>>>    +static void HID_Device_sendRawInput(DEVICE_OBJECT *device,
>>>> HID_XFER_PACKET *packet)
>>>> +{
>>>> +    BASE_DEVICE_EXTENSION *ext = device->DeviceExtension;
>>>> +    UCHAR report_id;
>>>> +
>>>> +    SERVER_START_REQ(send_hardware_message)
>>>> +    {
>>>> +        req->win                  = 0;
>>>> +        req->flags                = 0;
>>>> +        req->input.type           = HW_INPUT_HID;
>>>> +        req->input.hid.usage_page = ext->preparseData->caps.UsagePage;
>>>> +        req->input.hid.usage      = ext->preparseData->caps.Usage;
>>>> +        req->input.hid.length     = 0;
>>>> +
>>>> +        if (!(report_id = ext->preparseData->reports[0].reportID))
>>>> +        {
>>>> +            wine_server_add_data(req, &report_id, sizeof(report_id));
>>>> +            req->input.hid.length++;
>>>> +        }
>>>> +
>>>> +        wine_server_add_data(req, packet->reportBuffer, packet->reportBufferLen);
>>>> +        req->input.hid.length += packet->reportBufferLen;
>>>> +
>>>> +        wine_server_call(req);
>>>> +    }
>>>> +    SERVER_END_REQ;
>>>> +}
>>> It would be better to find a way to do this using exported APIs. I
>>> don't
>>> think we want to have even more modules make direct server calls (the
>>> goal is for all server calls to be done on the Unix side of the syscall
>>> interface).
>>
>> Alright, does it include __wine_send_input?
> 
> At least for now that would be better, yes.
> 

About that, is there any known use of INPUT_HARDWARE / HARDWAREINPUT 
type? I tried a few SendInput combination on Windows and it doesn't seem 
to have any effect.

On Wine it looks like it's plugged and able send fake Wine hardware 
messages, but it doesn't seem to be tested anywhere, or really used 
anywhere.

Would it be appropriate to use this structure to send HID rawinput 
message then? Maybe disable its support from SendInput and only allow it 
from __wine_send_input?

The INPUT structure doesn't have any explicit variable size member, but 
SendInput has a size parameter, so I guess we can keep any additional 
data internally.
-- 
Rémi Bernon <rbernon at codeweavers.com>