Skip to contentSkip to conversation

System hang up with Speedtouch USB hotplug

bug report Duncan Sands 3 participants 6 messages Nov 19, 2001 – Nov 20, 2001

lore
Share & tools
lore
#0 lore raw

Did you try without the preempt patch?

Duncan.

PS: You say "I'm not using binary driver from Alcatel but the free user-mode implementation...". In fact the Alcatel kernel module is open source (GPL)! Here's the situation as I understand it:

Johan Verrept's driver has two parts: a kernel module and a user space management utility. The kernel module is open source (GPL). It has problems with SMP and preempt, but otherwise works well. The user space management daemon is closed source. I think the reason it is closed source is that it uploads some firmware to the modem, and Alcatel wants this firmware to be closed source.

Benoit Papillault's driver is completely user space (it talks to the modem through usbdevfs). However it still uses the Alcatel closed source binary management code, because it needs the firmware for the modem. But I suppose it probably doesn't run any closed source code on your computer. I don't know the details of this project well, so take this with a pinch of salt (actually, you should take my description of the other project with a pinch of salt too).

#1 lore raw
Quoted 4 lines · Duncan Sands

Johan Verrept's driver has two parts: a kernel module and a user space management utility. The kernel module is open source (GPL). It has problems with SMP and preempt, but otherwise works well.

Anything that has problems with SMP is going to have problems with preempt. Further, most any module will need to be compiled as a preemptible kernel version, just like you need SMP versions.

Binary modules just are always going to be a problem unless we standardize more, and I'm not saying that is a good idea ...

Robert Love

#2 lore raw
In reply to Duncan Sands
Quoted 9 lines · Robert Love
Quoted 4 lines · Duncan Sands

Johan Verrept's driver has two parts: a kernel module and a user space management utility. The kernel module is open source (GPL). It has problems with SMP and preempt, but otherwise works well.

Anything that has problems with SMP is going to have problems with preempt. Further, most any module will need to be compiled as a preemptible kernel version, just like you need SMP versions.

Yes, it has problems with preempt because it has problems with SMP. But it seemed clearer to say "with SMP and preempt".

Quoted 2 lines

Binary modules just are always going to be a problem unless we standardize more, and I'm not saying that is a good idea ...

Lucky this isn't a binary module we're talking about then!

Duncan.

#3 lore raw
In reply to Duncan Sands
Quoted 1 line · Duncan Sands

Did you try without the preempt patch?

As I said before, I wasn't able to reproduce the hang, even with the preempt patch, so I don't know.

Quoted 4 lines

Duncan.

PS: You say "I'm not using binary driver from Alcatel but the free user-mode implementation...". In fact the Alcatel kernel module is open source (GPL)!

I didn't looked close at the Alcatel driver... so, maybe the kernel module is gpl.

Quoted 1 line

Here's the situation as I understand it:

[...]
Yes, it's how I understand it too

Signature · 6 lines
 ** Gael Le Mignot, Ing3 EPITA, Coder of The Kilobug Team **
Home Mail : [email protected]          Work Mail : [email protected]
GSM       : 06.71.47.18.22 (in France)   ICQ UIN   : 7299959
Web       : http://kilobug.freesurf.fr or http://drizzt.dyndns.org

"Software is like sex it's better when it's free.", Linus Torvalds
#4 lore raw
In reply to Duncan Sands

By the way, I find the preemptible kernel patch very helpful for debugging SMP problems on a uniprocessor machine.

I'm trying to debug the speedtouch kernel module crashes with SMP using it: the module oopses nicely with preempt too!

Duncan.

#5 lore raw
Quoted 7 lines · Duncan Sands

By the way, I find the preemptible kernel patch very helpful for debugging SMP problems on a uniprocessor machine.

I'm trying to debug the speedtouch kernel module crashes with SMP using it: the module oopses nicely with preempt too!

This is a nice unintended benefit of the preemptible kernel patch, eh? It also helps in your SMP debugging because you can see the lock depth (how many locks are held) via current->preempt_count. The preempt patch will also prevent you from infinite looping in the kernel, assuming the system is able to preempt itself.

Robert Love