[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#362056: More board combinations tested

[sent again, this time with the proper list address]

First of all, please use the BTS instead of responding privately. I
have attached your response, and this is posted to the BTS as well.


On Sat, 15 Apr 2006 15:54:20 -0400 Jack Carroll wrote:

> > It is possible, I believe, to blacklist a module from ever getting
> > automatically loaded by udev.
> > 
> > Also, it is possible to avoid udev altogether, by using the ramdisk
> > generator "yaird" instead of "initramfs-tools", chosen by default.
> 
> 
> 	This sounds interesting.  What would happen, the PCI bus
> wouldn't be scanned automatically at boot time, and the modules
> wouldn't be loaded until the ifconfig command caused them to be
> loaded?

Correct.


>  But how would interface names eth0, eth1, ... be associated
> with drivers and physical devices?  Does the order of the lines
> in /etc/modules take care of that?

Yes, I believe so.

There's also the option of binding device names to MAC addresses using
ifrename or ifupdow-scripts-zg2.


> > > Then, the command ifconfig eth2 192.168.136.55 causes the machine
> > > to freeze instantly.
> > 
> > Even if the problem can be worked around as mentioned above, this
> > still sounds like a kernel bug to me.
> 
> 	Sounds like it to me, too.  I just found that the same thing
> happens even if there aren't any other PCI Ethernet boards plugged
> in.  lsmod shows two modules loaded which look like Tulip drivers.  I
> think I'll try booting the floppy installer, which uses a 2.4.27
> kernel, and see if the Tulip board works at install time.
> 	Aha!  It attempts to load the de4x5 driver and gets an error
> message, but doesn't hang the machine.  But if I bring up another
> console, and command:
> 		modprobe tulip
> it works.  Specifically, we have a bug in the de4x5 module.
> 	Under 2.6.8, the problems are different.  Removing the de2104x
> module and loading the tulip module doesn't hang the machine, but the
> ifconfig command complains that there's no such device as eth2.
> Can't get the board recognized by the CD installer using the 2.6.8
> kernel, or an installed 2.6.8 kernel.

Thanks for the additional testing.

I'll leave it to to others (more clever on the inner workings of the
kernel) to deal with the bug itself...


 - Jonas


-- 
* Jonas Smedegaard - idealist og Internet-arkitekt
* Tlf.: +45 40843136  Website: http://dr.jones.dk/

 - Enden er n_r: http://www.shibumi.org/eoti.htm
--- Begin Message --- 
On Sat, Apr 15, 2006 at 01:39:53PM +0200, Jonas Smedegaard wrote:
> On Sat, 15 Apr 2006 00:45:14 -0400 Jack Carroll wrote:
> 
> > 	But, if there is a PCI Tulip board installed, its driver
> > module is loaded automatically at boot time after 3c59x.
> 
> A driver module loaded automatically sounds like udev is installed.

	It is.  I see it in the boot messages.

 
> It is possible, I believe, to blacklist a module from ever getting
> automatically loaded by udev.
> 
> Also, it is possible to avoid udev altogether, by using the ramdisk
> generator "yaird" instead of "initramfs-tools", chosen by default.


	This sounds interesting.  What would happen, the PCI bus wouldn't be
scanned automatically at boot time, and the modules wouldn't be loaded until
the ifconfig command caused them to be loaded?  But how would interface
names eth0, eth1, ... be associated with drivers and physical devices?  Does
the order of the lines in /etc/modules take care of that?

 
> > Then, the command ifconfig eth2 192.168.136.55 causes the machine to
> > freeze instantly.
> 
> Even if the problem can be worked around as mentioned above, this still
> sounds like a kernel bug to me.

	Sounds like it to me, too.  I just found that the same thing happens
even if there aren't any other PCI Ethernet boards plugged in.  lsmod shows
two modules loaded which look like Tulip drivers.  I think I'll try booting
the floppy installer, which uses a 2.4.27 kernel, and see if the Tulip board
works at install time.
	Aha!  It attempts to load the de4x5 driver and gets an error
message, but doesn't hang the machine.  But if I bring up another console,
and command:
		modprobe tulip
it works.  Specifically, we have a bug in the de4x5 module.
	Under 2.6.8, the problems are different.  Removing the de2104x
module and loading the tulip module doesn't hang the machine, but the
ifconfig command complains that there's no such device as eth2.  Can't get
the board recognized by the CD installer using the 2.6.8 kernel, or an
installed 2.6.8 kernel.

> 
> I just wanted to draw attention to the possible workarounds.
> 
> 
>  - Jonas

	Thanks.  I have one workaround in hand now, which is plug in two PCI
boards that use the same driver (3c59x), and one ISA-PNP board.  That works
under both 2.6.8 and 2.6.15, though the interrupts are assigned differently.

--- End Message ---

Attachment: pgpg8ZaMHDdyZ.pgp
Description: PGP signature

Reply to: