Re: [MM] grabbing QMI/WDM ports

Marius Bjørnstad Kotsbak <marius kotsbak gmail com> writes:

> I might also get time to test this as i just received my ZTE 820D
> modem. But I just wonder what the point of doing this is? Is it giving
> better performance 

In my testing with a Huawei E392 (which uses the same Qualcomm chipset
as the ZTE 820D) I've never been able to exceed 8 Mbits/s upstream when
using PPP and the option driver.  I've seen upstream bandwidths as high
as 30+ Mbits/s when using the network interface and the qmi_wwan driver.
So better performance is certainly a main argument for dropping PPP for

I've not done any actual analysis of why the PPP upstream bandwith is
limited. I'm guessing that modem packet processing may be the primary
limiting factor, but I don't know this. Comments and test results are
appreciated.  What data rates are you getting?

Note that the downstream bandwidth isn't affected (to the same degree at
least).  But that is reasonable, as adding a PPP header to the received
packets is no work at all...

> or more possibilities (checking signal quality
> while connected?)? The modem works great in Ubuntu 12.04 using the 3.2
> kernel without the qmi_wwan module (using just option module).

I believe QMI exposes a lot more than what's available through the AT
command interface, but why care as long as the AT interaface provides
more than enough for normal usage?  So yes, the modem works just fine
with PPP and the option driver.  You don't really need another driver.
But what's the fun in that :-)

Another issue, which I just noticed yesterday, and which may be purely a
Huawei firmware issue: I am completely unable to do IPv6 connections
using PPP.  The PDP context is reset to "IP" (from "IPV6" or "IPV4V6")
whenever I issue an "ATDT*99#" command.  I'm not claiming that IPv6
works flawlessly using the qmi_wwan driver either (in fact it doesn't
work at all without additional hacks), but at least it is possible to
make it work.  Further research is required before this goes anywhere
near mainstream, though.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]