Fibocom L850-GL hangs upon connect.

[Bjørn Mork]

Aleksander Morgado <aleksander at aleksander.es> writes:

> Fibocom engineers say there is no way to "switch" to USB after the
> device has been booted in PCIe mode.

Makes some sense.  But there must be some way to select mode in the
bootloader?

> Was told that Intel may be already working on the PCIe support for
> these devices in the kernel,

How hard can that be if you have access to
 - docs,
 - Windows driver source, and
 - modem and host hardware?

Personally I don't have too high hopes.  And while we wait, I believe it
is time to warn people about this problem.  The issue won't be fixed
unless people start voting with their wallets again.  Be careful when
buying new laptops. The included LTE modem might never work in Linux,
and BIOS whitelisting prevents you from ever replacing it.

We've been extremely lucky for many years now, with every LTE modem
vendor at least trying to implement some common "class spec". Although I
am not too impressed by Qualcomms openness wrt docs, they do deserve
credit for publishing enough of their QMI specs to make this happen. And
Microsoft deserves a very big "Thank You" for making MBIM a Windows
certification requirement. Together this has ensured almost complete
Linux LTE driver support for 6 years without much work at all!

Intel does impressive work in the community when it comes to CPUs, GPUs,
NICs, WiFi NICs, and of course also USB core with the xhci host driver
etc.  And probably much more.  But I haven't yet seen anyone from the
Infineon part of Intel doing anything in public.  Did I miss something?
Drivers don't come out of the blue.  When Intel launched their first
"mvm" Wifi chips in 2013, the driver support had already been developed
in the mainline kernel for half a year!  The Windows drivers weren't
publically availaible yet the first time I ran an iwl 7260 with a
mainline Linux driver.

Buth xmm chips? Haven't seen a single Intel commit anywhere.  Why should
that change now?  Infineon is a different Intel than we're used to.

> but while that is not a reality, there seems to be absolutely no way
> to handle the device in the Thinkpad running Linux, unless Lenovo
> decides to add the corresponding USB vid:pid in the BIOS whitelist.

Yes.  And while the rant above is mostly about Intel, I believe it is
important to note that the Lenovo BIOS whitelist is the real problem.
This "feature" is a commercial thing only, preventing you from buying
non Lenovo parts.  The certification excuse is just that - an excuse.
There is obviously no excuse for blacklisting the same hardware running
in a different bus mode.  Or for blacklisting the same hardware with a
non-Lenovo PID.  Or for blacklisting modems which have been certified
for older laptop models. 

Buy laptops without BIOS whitelisting if you can.  Avoid Lenovo and
HP. I believe most other vendors are safe?



Bjørn