Re: [dhcwg] [v6ops] DHCPv6 Prefix Delegation (DHCPv6 PD) on cellular links - works ok

[Alexandre Petrescu <[e-mail]>]

Le 30/11/2017 à 20:03, Bjørn Mork a écrit :
>> On Nov 30, 2017, at 11:18 AM, Alexandre Petrescu
>> <[e-mail]<mailto:[e-mail]>>
>> wrote:
>> 
>> v6opsers,
>> 
>> We succeeded in making DHCPv6 PD work natively on a cellular link. 
>> I would claim it to be a first[*]. Packet dump available upon
>> request.
>> 
>> The UE device is a Huawei E392u-12 USB dongle produced in 2011.
>> It hosts a baseband processor Qualcomm model MDM9600 (or model
>> MDM9200, or an Intel - not sure).
> 
> MDM9200.  

Noted.

I would like to ask you how do you know it's an MDM9200 in Huawei 
E392u-12 USB dongle?  Did you open the USB dongle and saw MDM9200 
printed on a IC packaging?  We could only open one cover and cant see 
the IC; further intrusion seems to risk to break it mechanically.

> Not that it matters much...

Well it does matter, because it may mean progress or counter-progress.

I am saying it's MDM9600 because Huawei announced the E392 to include a 
Qualcomm MDM9600, in year 2011:
> “We are pleased to collaborate with Qualcomm to announce the world’s
> first multi-mode LTE TDD Dongle E392 based on Qualcomm’s MDM9600,
> which is anticipated to be commercially available in the third
> quarter of 2011,” said [...] of Huawei Devices India.

On the same MDM920x product line, the MDM9207 is more recent, probably 
year 2016, whereas MDM9200 is surely year 2011.  MDM9207 does block 
DHCPv6 whereas MDM9200 does not block DHCPv6.

It would be a counter-progress if newer chipsets blocked DHCPv6 whereas 
older chipsets allowed it.

That's why I suppose the E392 to be hosting a different Qualcomm product 
line - the MDM9600.  That MDM9600 product line would supposedly support 
DHCPv6 throughout all its subsequent generations.

But I really dont know whether E392 holds a MDM9200 or a MDM9600; so I 
have to go again through this painful Qualcomm 'business case' via 
Sierra and Orange, to ask them again to unblock multicast/547.

If I knew precisely that E392 holds MDM9200 I could tell them: why 
MDM9200 allows DHCPv6 whereas MDM9207 blocks it?

Until then I can only tell them: why MDM9200-or-MDM9600 allows DHCPv6 
whereas MDM9207 blocks it?

(add to that the fact that some xda-developpers say that E392 is 
probably hosting an Intel modem, rather than Qualcomm)

>> It lets all standard DHCPv6 messaging through. The DHCPv6 client 
>> 'odhcpv6' is running on linux laptop.
> 
> Can't help being a bit curious about the Linux driver used for this 
> test? PPP?  qmi_wwan?  Something else?

For this test we used ModemManager with mmcli command line.  We did not 
use pppd.  The kernel modules were: qmi_wwan, cdc_wdm, option and 
usb_wwan.  It was first necessary to switch the dongle from "USB 
storage" mode to "modem" mode.  To achieve that it was necessary to use 
a proprietary string.  That proprietary string is found by first 
sniffing the serial line on a Windows machine.  This whole paragraph is 
publicly documented by programmers in forums.

> (crossing fingers for qmi_wwan :-)

Yes :-)

Alex

> 
>> DHCPv6-PD is blocked on other baseband processors, including
>> Qualcomm and Balong versions; they are hosted by many smartphones,
>> USB cellular dongles, mobile personal WiFi hotspots and IoT
>> devices.  A sample list is available upon request.
> 
> Yes, the "smarter" these things get, the harder it is to make them
> work properly.
> 
> 
> Bjørn
> 
> _______________________________________________ dhcwg mailing list 
> [e-mail] https://www.ietf.org/mailman/listinfo/dhcwg
>