On Friday 21 February 2003 03:01, Steve Parker wrote:
> I see that the Alcatel kernel-level driver for the Alcatel SpeedTouch USB
> Modem is now included in the 2.5 kernel.
Hi Steve, that is so, and I am maintaining it.
> This seems strange, since there seem to have been (in the past, at least)
> a lot of (panic) problems reported with it, and the speedtouch.sf.net (and
> complimentary speedtouchconf.sf.net) are fully capable of running this
> modem in userspace.
> Have these problems been resolved? Is the kernel driver as stable as the
> userspace one? Are there demonstrable perfomance benefits in the kernel
> driver?
These problems are being resolved. Most of them have already been resolved.
The cvs version for 2.4, which you can find at
http://www.linux-usb.org/SpeedTouch/
is quite stable. In theory it can still crash (due to various micro races), but in
practice it does not. In any case, these micro races will be fixed soon. The
2.5 version, which is essentially identical to 2.4 cvs, doesn't work very well
in the current 2.5 kernel. I don't know why. I am working on it.
I have nothing against the user space version, which I used for many moons.
The kernel version is certainly much lighter weight - less CPU, less memory.
Whether this matters for you depends on your machine/needs. My machine
is slow, and I need all the CPU time I can get!
> I've certainly been using my modem for well over a year with the userspace
> driver (speedtouch.sf.net) with - as at 2.4.18 - an unpatched kernel, and
> no troubles whatsoever. Is there any need for the Alcatel code in the
> kernel when the n_hdlc and ppp configs already cater for this modem, once
> the microcode is loaded?
The main disadvantages of the kernel mode driver were:
(1) unstable, and very unstable on SMP/preempt boxes
(2) required running the closed source speedmgmt program
(3) required compiling your own kernel
The driver is in 2.5, and is heading for inclusion in 2.4, so I expect that in the
future most distributions will ship with the speedtch module compiled. Thus
(3) is going away.
The cvs version of the user space driver contains a patch for modem_run
which enables it to be used with the kernel driver in place of speedmgmt
(use the -k flag). Thus (2) has already gone away.
As I mentioned, (1) is (almost) dealt with.
> What is the status of this driver WRT the Alcatel microcode? Last I heard,
> the Alcatel microcode was required by both the Alcatel kernel-level and
> the speedtouch.sf.net drivers.
That is correct, but you no longer need to upload it using speedmgmt.
There are efforts underway to reverse engineer the microcode, with
some success. The goal of writing our own microcode (ARM processor
by the way) is still far, far away though.
All the best,
Duncan.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Feb 23 2003 - 22:00:32 EST