From: Grzegorz Kulewski <kangur@polcom.net>
To: Alistair John Strachan <s0348365@sms.ed.ac.uk>
Cc: Giampaolo Tomassoni <g.tomassoni@libero.it>,
linux-kernel@vger.kernel.org,
linux-atm-general@lists.sourceforge.net
Subject: Re: [ATMSAR] Request for review - update #1
Date: Sun, 4 Sep 2005 19:22:15 +0200 (CEST) [thread overview]
Message-ID: <Pine.LNX.4.63.0509041904390.29195@alpha.polcom.net> (raw)
In-Reply-To: <200509041754.54995.s0348365@sms.ed.ac.uk>
On Sun, 4 Sep 2005, Alistair John Strachan wrote:
> On Sunday 04 September 2005 17:41, Grzegorz Kulewski wrote:
>> On Sun, 4 Sep 2005, Alistair John Strachan wrote:
>>> On Sunday 04 September 2005 12:05, Giampaolo Tomassoni wrote:
>>>> Dears,
>>>>
>>>> thanks to Jiri Slaby who found a bug in the AAL0 handling of the ATMSAR
>>>> module.
>>>>
>>>> I attach a fixed version of the atmsar patch as a diff against the
>>>> 2.6.13 kernel tree.
>>>
>>> [snip]
>>>
>>> Just out of curiosity, is there ANY reason why this has to be done in the
>>> kernel? The PPPoATM module for pppd implements (via linux-atm) a
>>> completely userspace ATM decoder.. if anything, now redundant ATM stack
>>> code should be REMOVED from Linux!
>>>
>>> Most distributions (to my knowledge) supporting the speedtouch modem do
>>> so using the method prescribed on speedtouch.sf.net; an entirely
>>> userspace procedure. pppd does all the ATM magic.
>>>
>>> Does this have real-world applications beyond the Speedtouch DSL modems?
>>> If not, I propose adding this code to linux-atm, not the kernel, since
>>> most users of USB speedtouch DSL modems will not be using the kernel's
>>> ATM.
>>
>> I am using SpeedTouch 330 modem with kernel driver (on Gentoo).
>>
>> The instalation is currently (with firmware loader instead of modem_run)
>> very simple: USE="atm" emerge ppp, download firmware and place it in
>> /lib/firmware, compile the kernel with speedtch support.
>
> Compared to "place the firmware in /lib/firmware" on many other distros, this
> sounds like a lot of work! The kernel speedtch provides no advantages to its
> userspace alternative.
Are you saying that you do not need to install ppp and compile (or install
binary) kernel on other distros???
>> I tried to use userspace driver some time ago but it wasn't working for me
>> so I gave up. I was using modem_run with kernel driver for long time to
>> load the firmware but there were many problems with it too (nearly every
>> kernel or modem_run upgrade was breaking something, modem_run was hanging
>> in D state in most unapropriate moments and so on).
>
> This is not the case any longer.
Maybe. But there were many bugs in modem_run, usbfs, usb drivers in
kernel, ACPI, IRQ routing and so on. They caused modem_run to hang or do
something stupid without even telling user what is broken. In my
experience when one bug was fixed other appeared in the next kernel. So
even if now everything works ok you have no guarantee that next release
won't break something... :)
>> Now I am using pure kernel driver and firmware loader and it works 100%
>> ok. There were no problems with it for long time. And I don't even want to
>> look at this userspace driver again.
>
> Conversely people (including myself) found the kernel implementation to be
> buggy, and when userspace breaks, you can simply restart it.. when the kernel
> breaks, you have to reboot.
1. But you still use the kernel even with userspace driver. So it still
can break forcing you to reboot.
2. I have no problems with kernel driver. All problems that I saw in the
past were problems in other subsystems (usbfs, usb drivers, ACPI, IRQ, ...).
3. Restarting modem_run was never enought for me. I always at least had to
unplug the modem from USB port. Or more often reboot. So I see no gain
here.
>> Since Linux newer was (or is going to be) userspace-driver OS, maybe we
>> should leave it that way...
>
> No.
>
> What can be done in userspace, within valid performance constraints, usually
> should be. This has always been the Linux kernel way.
But not device drivers. I do not think that Linux has good infrastructure
for drivers in userspace (yes I know that there are some). Also are you
sure that performance and latencies are ok with userspace driver even if
system is under load?
> The speedtouch modem is a USB device, and there are many existing userspace
> "driver" implementations for USB devices. Including speedtouch.
>
> I'm not necessarily saying this code shouldn't be in the kernel, I'd just be
> interested to know why it has to be.
Well, as you are saing it hasn't... :) But since it is there and works
for me I am interested in not changing this state without really good
reason.
Grzegorz Kulewski
next prev parent reply other threads:[~2005-09-04 17:22 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-09-04 11:05 Giampaolo Tomassoni
2005-09-04 12:00 ` [Linux-ATM-General] " Francois Romieu
2005-09-04 13:13 ` R: " Giampaolo Tomassoni
2005-09-04 15:32 ` Francois Romieu
2005-09-04 18:44 ` R: " Giampaolo Tomassoni
2005-09-04 19:11 ` matthieu castet
2005-09-04 19:33 ` R: " Giampaolo Tomassoni
2005-09-04 19:42 ` Giampaolo Tomassoni
2005-09-05 6:05 ` Zoran Stojsavljevic
2005-09-05 6:21 ` R: " Giampaolo Tomassoni
2005-09-04 16:20 ` Alistair John Strachan
2005-09-04 16:41 ` Grzegorz Kulewski
2005-09-04 16:54 ` Alistair John Strachan
2005-09-04 17:22 ` Grzegorz Kulewski [this message]
2005-09-05 15:04 ` Duncan Sands
2005-09-04 18:36 ` R: " Giampaolo Tomassoni
2005-09-04 18:51 ` Alistair John Strachan
2005-09-05 9:36 ` David Woodhouse
2005-09-05 13:52 ` Alistair John Strachan
2005-09-05 13:56 ` David Woodhouse
2005-09-05 14:18 ` Alistair John Strachan
2005-09-05 14:32 ` David Woodhouse
2005-09-05 14:46 ` Duncan Sands
2005-09-04 16:55 ` Jiri Slaby
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Pine.LNX.4.63.0509041904390.29195@alpha.polcom.net \
--to=kangur@polcom.net \
--cc=g.tomassoni@libero.it \
--cc=linux-atm-general@lists.sourceforge.net \
--cc=linux-kernel@vger.kernel.org \
--cc=s0348365@sms.ed.ac.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®