Re: MBIM device refusing to be enabled

[Dan Williams <dcbw@redhat.com>]

On Wed, 2013-10-09 at 13:22 +0200, Oliver Neukum wrote:
> On Wed, 2013-10-09 at 13:09 +0200, Bjørn Mork wrote:
> > Bjørn Mork <bjorn@mork.no> writes:
> 
> Hi,
> 
> > 
> > > Sorry, that was nonsense.  rfkill would only affect the HwRadioState.
> > 
> > Turns out I wasn't completely off anyway... by pure luck ;-)
> > 
> > > Looking at the commands following this message, it appears that MM
> > > correctly attempts to enable the SwRadioState but fails.  I don't know
> > > why.
> 
> Cool.
> 
> > Is the device by any chance a Sierra Wireless device?
> 
> Yes. Rebranded by HP but it is Sierra.
> HP lt4111 LTE/EV-DO/HSPA+ Gobi 4G Module
> 
> > I was able to recreate the behaviour you see after experimenting a bit
> > with my MC7710.  This device (and I assume most other MBIM capable
> > Sierra Wireless minicards) can be configured to enter low power mode on
> > rfkill (W_DISABLE asserted), instead of powering off.  Sony for example
> > are known to configure the built-in Sierra devices in this mode.
> > 
> > When low power mode is forced by rfkill this way, the firmware
> > erroneously[1] claims
> > 
> >  HwRadioState: on
> >  SwRadioState: off
> > 
> > Any attempt to change this software state using MBIM will fail with
> > MBIM_STATUS_FAILURE.  Exactly like your log shows.  The modem must be
> > enabled using rfkill before MM can use it. Changing the firmware
> > behaviour will not do any good - it will just cause the modem to power
> > off and disappear instead.
> > 
> > So I would start looking at rfkill after all.  There are often problems
> > with these platform drivers and newer laptops, due to the lack of
> > documentation from the vendors.  Here's one (now fixed) example:
> > https://bugzilla.kernel.org/show_bug.cgi?id=47751
> 
> Thanks. A quick look shows that there's something wrong. There is no
> rfkill entry for the device. I am investigating.

PCIe minicard pinouts have a W_DISABLE# pin (row 1 pin 20, active low)
that kills the card.  If your machine grounds that PIN or mishandles it
somehow, and the card cares about that pin, then the card will show as
rfkilled.

I have cards that care: Samsung Y3300/Y3400, and most Ericsson devices.
The devices themselves show as rfkilled (either through QMI, AT
commands, etc) but you can't do anything about it unless you have a way
to float W_DISABLE.

So depending on your laptop's slot pinout and BIOS behavior, and whether
or not the rfkill driver correctly supports your laptop, you might be
out of luck.

Dan