Huawei E3276 problem

[Tommy Cheng]

I will test unbinding tomorrow.  But I did notice that the system hung when I
tried to unload cdc_ncm.ko.

By the way, the platform I tested on is MIPS SoC (Big Endian).  Maybe the
system hung problem has something to do with the platform.  I will apply
your patch and test it tomorrow.  Thanks a lot.




----- Original Message -----
From: Bjørn Mork <bjorn at mork.no>
To: Tommy Cheng <tommy7765 at yahoo.com>
Cc: Olof Ermis <olof.ermis at gmail.com>; "libqmi-devel at lists.freedesktop.org" <libqmi-devel at lists.freedesktop.org>
Sent: Tuesday, November 13, 2012 6:42 PM
Subject: Re: Huawei E3276 problem

Tommy Cheng <tommy7765 at yahoo.com> writes:

> The debug output is as follows: 
>
> # insmod /tmp/cdc_ncm.ko 
> cdc_ncm: 14-Mar-2012
> cdc_ncm 1-1:1.1: XX: ctx->control=81eec080, ctx->data=81eec080, ctx->ether_desc=814b67c8, intf=81eec080
> cdc_ncm 1-1:1.1: XX: failed to claim interface: -16
> usb 1-1: bind() failure
> usbcore: registered new interface driver cdc_ncm

Thanks for testing both of you.  The return code (EBUSY) confirm the
assumption that this fails because we already own the claimed
interface.

I just tested and found that the same patch will apply to both
net-next/master and the current mainline because the context is
unaltered even with the major changes we made to the driver.

So I think I'll try to ask davem if this can be accepted for v3.7 and
stable after all.

Just one more question:  Does unbinding work without any further
changes?  You can test with e.g.
echo '1-1:1.1' >/sys/bus/usb/drivers/cdc_ncm/unbind

I am wondering because it will attempt to release the data interface
when unbinding from the control interface, and this may cause problems
when they are the same.  In which case we need to put a protecting
conditional there too.



Bjørn