QMI IPv6 crashes

[Bjørn Mork]

Dan Williams <dcbw at redhat.com> writes:

> So I tried this with the MC7750 and a Verizon IPv6-only connection.  I
> set NOARP, and set router_solicitations,
> mldv2_unsolicited_report_interval, mldv1_unsolicited_report_interval,
> and ndisc_notify to "0" to disable them.  I also verified that with QMI
> static IP configuration NetworkManager was *not* sending router
> solicitations.  I then ran tcpdump on the WWAN interface to capture any
> packets.  No packets were captured in two tries, and the device still
> crashed with:
>
> Src: FatalError
> Dat: 00000DFE
> Str: ps_icmp6_nd.c
> Fmt: Can't find the ND control block
> 00000000 00000000 00000000 00000000
> Task: DS
>
> Not sure where to proceed from here...

At least we know that it's not caused by anything we send to the modem.
Could still be something we DON'T send of course...

But if the firmeware reallly is so fragile that we have to imitate the
Windows event sequence just to make it survive, then I am not sure it's
worth it.  Better let it die.

I found it quite entertaining at first that Sierra now distributes a
version of the Qualcomm GobiNet driver with a firmware bug workaround
taken from qmi_wwan. But it's really tragic if we step back and think
about it.  Sierra obviously acknowledge the firmware bug, but they seem
incapable of fixing the firmware. That's a bit terryfying, even if we
accept that the bug comes from Qualcomm.  Sierra should really be in a
position to have the bug fixed either way.  They shouldn't need to add
workarounds to drivers.  That's just stupid if you control the firmware.


Bjørn