Bug#507994: Please include version 1.6 of hso.c

[Moritz Muehlenhoff <jmm@inutil.org>]

On Sun, Dec 14, 2008 at 02:22:10PM +0100, Torsten Jerzembeck wrote:
> Hi,
> 
> I've tested the hso drivers, especially under less-than-optimal
> reception conditions (travelling by train from Stuttgart to Karlsruhe ->
> frequent changes between UMTS, GPRS an no reception). My results so far
> are:
> 
> - The 1.2 driver included with the current kernel sources is next to
>   useless. Really poor reception, and very frequent lock-ups of the
>   system. It seems that this version doesn't handle handovers and
>   network outages gracefully at all and has severe problems when
>   disconnecting the USB modem.
> 
> - The 1.6 driver from the Pharscape forum that aba is talking about
>   gives _much_ better reception and is _much_ more stable than the 1.2.
>   However, even this causes system lockups. I've managed to nail down
>   the cause, it seems that the code has problems with rapid changes
>   between "no reception" and GPRS/UMTS when under network load.
> 
> - I tried to backport the hso.c module from linux-next (kindly provided
>   by Florian Weimer; my version is attached to this mail). It uses an
>   additional element in the tty_struct that is not present in earlier
>   versions of the kernel. After changing the code not to use
>   tty_kref_put and tty_kref_get, it compiles and can be inserted, but
>   the system freezes on the first access of the hardware. I suppose
>   there is some kind of locking missing, but I'm not fluent enough in C
>   to debug this.
> 
> My next step will be using a vanilla 2.6.27.9 kernel with the 1.6 hso.c
> from the Pharscape forum, as this is reported to be stable and cause no
> freezes.

As of Linux 2.6.31-rc4, the version 1.6 of hso still isn't merged into
mainline. It has seen a couple of bugfixes, so it might be possible that
fixes have been cherrypicked from 1.6 without bumping the version number.

Does the 2.6.30 driver from unstable work correctly? Otherwise you should
prod the author of the 1.6 version to submit his changes to the mainline
kernel.

Cheers,
        Moritz