Re:Re: [PATCH] net: usb: qmi_wwan: Add support for Cinterion MV32

["Slark Xiao" <slark_xiao@163.com>]

At 2022-08-10 14:55:42, "Bjørn Mork" <bjorn@mork.no> wrote:
>Slark Xiao <slark_xiao@163.com> writes:
>
>> There are 2 models for MV32 serials. MV32-W-A is designed
>> based on Qualcomm SDX62 chip, and MV32-W-B is designed based
>> on Qualcomm SDX65 chip. So we use 2 different PID to separate it.
>>
>> Test evidence as below:
>> T:  Bus=03 Lev=01 Prnt=01 Port=02 Cnt=03 Dev#=  3 Spd=480 MxCh= 0
>> D:  Ver= 2.10 Cls=ef(misc ) Sub=02 Prot=01 MxPS=64 #Cfgs=  1
>> P:  Vendor=1e2d ProdID=00f3 Rev=05.04
>> S:  Manufacturer=Cinterion
>> S:  Product=Cinterion PID 0x00F3 USB Mobile Broadband
>> S:  SerialNumber=d7b4be8d
>> C:  #Ifs= 4 Cfg#= 1 Atr=a0 MxPwr=500mA
>> I:  If#=0x0 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=50 Driver=qmi_wwan
>> I:  If#=0x1 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=40 Driver=option
>> I:  If#=0x2 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=40 Driver=option
>> I:  If#=0x3 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=30 Driver=option
>
>The patch looks nice, but I have a couple of questions since you're one
>of the first pushing one of these SDX6x modems.
>
>Is that protocol pattern fixed on this generation of Qualcomm chips?  It
>looks like an extension of what they started with the SDX55 generation,
>where the DIAG port was identified by ff/ff/30 across multiple vendors.
>
 Seems yes. I checked some different usb_compositions and found that
 diag port is using protocol '30' always.

>Specifically wrt this driver and patch, I wonder if we can/should match
>on ff/ff/50 instead of interface number here?  I note that the interface

I checked all our edited usb_compositions and all QC default usb 
compositions(9025, 90db, 9067,90d5,9084,9091,90ad,90b8,90e5), 
ff/ff/50 is rmnet used only. 

>numbers are allocated sequentionally. Probably in the order these
>function are enabled by the firmware? If so, are we sure this is static?

This needs more time to confirm. I will keep you updated.

>Or could we risk config variants where the RMNET/QMI function have a
>different interface number for the same PIDs?
>
>And another possibility you might consider.  Assuming that ff/ff/50
>uniquely identifies RMNET/QMI functions regardless of PID, would you
>consider a VID+class match to catch all of them?  This would not only
>support both the PIDs of this patch in one go, but also any future PIDs
>without the need for further driver patches.
>
>
>Bjørn

I have a concern, if Cinterion or other Vendors, like Quectel, use other 
chip (such as intel, mediateck and so on), this methods may won't work,
because  they share a same VID. Also this may be changed once Qualcomm 
update the protocol patterns for future chip.