Re: [PATCH v2] USB: serial: qcserial: Support for SDX55 based Sierra Wireless 5G modules

From: Bjørn Mork
Date: Fri Jul 02 2021 - 03:33:16 EST


Stefan Brüns <stefan.bruens@xxxxxxxxxxxxxx> writes:

> On Donnerstag, 24. Juni 2021 13:15:07 CEST Bjørn Mork wrote:
>> Johan Hovold <johan@xxxxxxxxxx> writes:
>> > Could you please also post the output of usb-devices (or lsusb -v) for
>> > this device in MBIM mode?
>>
>> Yes, this would be nice to have.
>
> See below.
>
>> I suspect that this device is like other SDX55 devices we've seen, using
>> class/subclass/function to map the vendor specific functions
>> too. Dropping static interface numbers. If correct, then the patch is
>> bogus and the interface numbers might change based on firmware version
>> and configuration.
>
> Do you really expect Sierra do to something sensible? According to their
> documentation functions are matched by interface numbers.

Well... I expect them to be as sensible as any other vendor. And I
expect docs to be intendend as guidance only ;-)

> They still use broken interface descriptors with holes in interface numbering
> (i.e. interface number 2 does not exist, which violates the USB standard).

Right. Wrt the violation, I think that train left a decade ago.

> Interface Descriptor:
> bLength 9
> bDescriptorType 4
> bInterfaceNumber 3
> bAlternateSetting 0
> bNumEndpoints 3
> bInterfaceClass 255 Vendor Specific Class
> bInterfaceSubClass 0
> bInterfaceProtocol 0
[..]
> Interface Descriptor:
> bLength 9
> bDescriptorType 4
> bInterfaceNumber 4
> bAlternateSetting 0
> bNumEndpoints 2
> bInterfaceClass 255 Vendor Specific Class
> bInterfaceSubClass 255 Vendor Specific Subclass
> bInterfaceProtocol 48


So Sierra do follow the same pattern we've seen on other X55 devices:

ff/00/00 - AT
ff/ff/30 - QCDM

See commits

accf227de4d2 ("USB: serial: option: Add support for Quectel RM500Q")
d6c1ddd938d8 ("USB: serial: option: add Quectel EM160R-GL")

for other examples.

This obviously doesn't make any difference if your configuration is the
only one. But I believe that is unlikely. There are probably ways the
layout can be changed, even if currenly not documented. The advantage
of class/subclass/protocol matching to function is that it works
regardless of the number of functions and their interface number.


Bjørn