eciadsl works every second time with 2.6.15 kernel, worked with breezy's kernel

[Matteo Z]

First time I launch "eciadsl-start" (USB modem driver see http://eciadsl.flashtux.org/index.php) it fails while it's executing the 4th step. I only see this:

[CODE]
[4/5] Connecting to provider ...
[/CODE]

and nothing more...

and I must terminate it with Ctrl-C and re-launch "eciadsl-start", and now (at the second try) it works correctly.

I've tried to set debug mode, but in the log file there isn't nothing about eciadsl/pppd.

Eciadsl worked fine in breezy with kernel 2.6.12, and works even in dapper if you downgrade to kernel 2.6.12. This bug doesn't permit to use the modem for a server since you have to manually unplug the modem and reexecute the command in order to connect.

Note: the original reporter indicated the bug was in package 'linux-image-2.6.15-10-386'; however, that package was not published in Ubuntu.

[Matteo Z (matteozandi)]

linux-image-2.6.15-20-386 is the right package :)

[Chuck Short (zulcss)]

Then we should assign it to 2.6.15

[bob23450 (bzanetti-deactivatedaccount)]

I experienced the very same behaviour with both Kubuntu 6.06.1 (kernel 2.6.15.26 AMD64 generic) and Kubuntu 6.10 (kernel 2.6.17.10 AMD64 generic), both freshly installed + pppoe package from Ubuntu archives + eciadsl 0.11 usermode (Debian package: http://eciadsl.flashtux.org/download/eciadsl-usermode_0.11-1_amd64.deb).

BTW my ADSL modem is IPM Datacom Dataway USB (Telecom Italia Alice).

I can't find any error related to pppd / pppoe or other networking inside /var/log/messages, but the first time the script simply hangs forever at the beginning of step 4. The second time it works fine, even without unplugging and replugging the modem.

I also tried to disable the integrated Ethernet adapter but it didn't help.

I didn't find any workaround so far so I'm sticking to the double-manual activation.

[bob23450 (bzanetti-deactivatedaccount)]

Well, I did it. I managed to make "eciadsl-start" work the first time, at least under Kubuntu Edgy AMD64 (but probably it may work on other versions / platforms).
The trick was to change the following line of the script /usr/bin/eciadsl-start from:
  nice --20 /usr/sbin/pppd call adsl updetach
to
  nice --6 /usr/sbin/pppd call adsl updetach

With such a change the script doesn't hang any more inside pppd and completes succesfully.
My skill on linux kernel things is very low, but I suppose the niceness requested to the process pppd was too extreme. After many try-and-reboot I found that a value less than 10 is required to work correctly. In fact, pppd will spawn pppoeci (see /etc/ppp/peers/adsl) with -10 more niceness (don't ask me why).
I hope some skilled guy will find a more scientific diagnosis and remedy, in the meantime I'm happy with this.

In short, to make things work try the following:
1. Edit as root the file /usr/bin/eciadsl-start
2. Find the line with the word "nice"
3. Replace "--20" with "--6" (eventually try another number)
4. Save, close, reboot, run eciadsl-start as usual, pray God, say wow!

[Gareth Fitzworthington (mapping-gp-deactivatedaccount)]

This bug has had no activity for a considerable period. This is a check to see if there is still interest in investigating this bug report.
Appears to have been solved with script editing.

[bob23450 (bzanetti-deactivatedaccount)]

Just a few notes.
* I saw this problem also in every later kernel versions
* The script editing above actually has little effect (I posted it with too much precipitation, I apologize)
* The best workaraound I ever found is outlined here: http://eciadsl.flashtux.org/forum/viewtopic.php?t=3152 (it doesn't solve the problem but it works for me most of the times).
* IMHO the problem is most probably related to the driver, not the kernel
So, for me, the report can be closed.
Best regards.

[Gareth Fitzworthington (mapping-gp-deactivatedaccount)]

Closing.