Re: [Libusb-devel] Error sending control message: No such file or directory

[Tim R. <ti...@pr...>]

Blaine Booher wrote:
> On Friday 19 December 2008, Alan Stern wrote:
>   
>> On Fri, 19 Dec 2008, Blaine Booher wrote:
>>     
>> To make this work, you need to edit the source file 
>> drivers/usb/core/devio.c.  Change the function check_ctrlrecip() to 
>> simply return 0 without doing any of its normal tests.
>>
>> Alan Stern
>>
>>
>>     
>
> Hey everyone,
>   I wanted to open a disucssion (if one is necessary) about this issue.  The issue being that the kernel (in check_ctrlrecip()) blocks control messages with request type 0x22 and request 0x09.  The 0x09 is SET_REPORT in the HID API, which is valid.

Two corrections to that.  First, it doesn't block ALL control messages
with bmRequestType=0x22 and bRequest=0x09; it only blocks those where
the endpoint number in the wIndex field is not within an interface
claimed by the caller.  Second, all of the HID requests use
bmRequestType=0x21, targeting an interface.  0x22 targets the endpoint. 
The 0x22/0x09 message is NOT valid in the HID API.

>   So... since there are devices available for purchase on the market (that may or may not be Windows-only) that do in fact use this request type and request, should the kernel be updated to not block these requests?
>   

This is a complicated question, and although I have my own opinions,
this needs to be decided by the Linux kernel USB team.

>   I believe that this requsttype has to do with specifics regarding HID devices.  Is this actually a rare usage of an HID specification, or is it straight up non-compliant?
>   

It is straight up non-compliant.  Section 7.2 (page 50) of the HID spec
is quite clear on this.

-- 
Tim Roberts, ti...@pr...
Providenza & Boekelheide, Inc.