LinuxQuestions.org - forcing e1000e driver on new device (old kernel... old everything)

- Linux - Hardware (https://www.linuxquestions.org/questions/linux-hardware-18/)

- - forcing e1000e driver on new device (old kernel... old everything) (https://www.linuxquestions.org/questions/linux-hardware-18/forcing-e1000e-driver-on-new-device-old-kernel-old-everything-4175428961/)

forcing e1000e driver on new device (old kernel... old everything)

Hi!

I've got this box that (for whatever reason, can't really tell why) got a ubuntu 10.04 installed on it. The problem is that it's got an intel ethernet card that's not got registered to use e1000e driver (and I check in my updated kubuntu box and it uses e1000e by modules.pcimap).

The card has PCI id 8086:1503.

Now, we hacked modules.pcimap and added a line like the one I have in my computer, then modprobed e1000e, the module got loaded but no new device.

Then rebooted, just in case, and the module didn't get loaded... so I bet there's something missing.

So, how can I force e1000e to use this device? I checked with modinfo and I don't see a parameter to tell e1000e which PCI id to use. Thanks in advance.

Yes e1000e is the driver for this card and it *should* be automatically loaded.

Maybe it's conflicting with another one, loaded first.

To find out, please send output of following command:

Code:

lspci -k | grep -A3 net

And attach to your next post the file lsmod.txt that you will write this way:

Code:

lsmod | sort > lsmod.txt

PS And please give us the output of

Code:

uname -r

as well so we know how old is your kernel. That could help determine how the modules are supposed to be selected then loaded.

It's an old kernel.... but I don't have the computer in front of me. Just asked the guys in front of the computer to do it... hold on a second. Just keep in mind that the device is not (was not, anyways) in modules.pcimap.... So there has to be a way to force the module to try to work on the device even if it thinks it shouldn't.

First command:

Code:

00:19.0 Ethernet controller: Intel Corporation Device 1503 (rev 05)

        Kernel modules: e1000e

00:1a.0 USB Controller: Intel Corporation Cougar Point USB Enhanced Host Controller #2 (rev 05)

        Kernel driver in use: ehci_hcd

Second

Code:

ahci                  32232  1 

bitblit                4707  1 fbcon

fbcon                  35102  71 

font                    7557  1 fbcon

hid                    67032  1 usbhid

lp                      7028  0 

Module                  Size  Used by

parport                32635  3 ppdev,parport_pc,lp

parport_pc            26250  1 

ppdev                  5259  0 

psmouse                63245  0 

serio_raw              3978  0 

shpchp                28884  0 

snd                    54148  6 snd_hda_codec_realtek,snd_hda_intel,snd_hda_codec,snd_hwdep,snd_pcm,snd_timer

snd_hda_codec          74201  2 snd_hda_codec_realtek,snd_hda_intel

snd_hda_codec_realtek  203312  0 

snd_hda_intel          22037  0 

snd_hwdep              5412  1 snd_hda_codec

snd_page_alloc          7172  2 snd_hda_intel,snd_pcm

snd_pcm                70886  2 snd_hda_intel,snd_hda_codec

snd_timer              19098  1 snd_pcm

softcursor              1189  1 bitblit

soundcore              6620  1 snd

tileblit                2031  1 fbcon

usbhid                36174  0 

usb_storage            39457  0 

vga16fb                11385  1 

vgastate                8961  1 vga16fb

uname:

Code:

2.6.32-24-generic-pae

I guess in the first command, the module that shows up is because of the line we hacked in pcimap.

What puzzles me is ther be no driver in use. My Ethernet card needs the same driver and this is what I get as output of the first command:

Code:

00:19.0 Ethernet controller: Intel Corporation 82566MM Gigabit Network Connection (rev 03)

        Subsystem: Lenovo ThinkPad T61

        Kernel driver in use: e1000e

        Kernel modules: e1000e

So:
(1) Does e1000e shows in lsmod after you issue the"modprobe e1000e" command?
(2) Type as root "tail -f /var/log/messages in a terminal", then:
(a) unplug the Ethernet wire then
(b) plug it in
What do you see in the log?
(3) same as (2) but type as root
(a) "ifconfig eth0 up" then
(b) "ifconfig eth0 down"

Anyway, 2.6.32 is not that old and your card is supposed to be supported.

I just checked in vanilla 2.6.32 and it's not supported from what I can see.

In 3.6-rc5
http://git.kernel.org/?p=linux/kerne...1a24195dd08217 Line 412:

Code:

#define E1000_DEV_ID_PCH2_LV_V 0x1503

Not so in 2.6.32. Take a look: http://git.kernel.org/?p=linux/kerne...d34d935308c8d7

That's right.

So I suggest you simply upgrade your e1000e driver to the most recent version that you will find here.

In the tarball, named e100e-2.1.4.tar.gz you will find a README including instructions to compile and install it. Of course your card is supported in it, I just checked ;)

The guys in maracaibo are checking if it works with the ubuntu server (I just tested compiling it on my updated laptop).

Now, let's suppose it works.... I dare making the same question: Is it possible to force a module to use a device? Perhaps with something generic like

Code:

modprobe e1000e pciid=8086:1503

Thanks in advance.

Your command won't bind the kernel module to the device. To do that "modprobe e1000e" is sufficient, provided the device be not yet claimed by another driver.

But don't bother to do manually what udev is supposed to do automatically for you.

If I understand correctly:
(1) After installing a module, run as root the command "modprobe -a". This will update the modules dependencies (needed by modprobe to work properly) and in addition update the file /lib/modules<kernel version>/modules.alias. To do that the command "modinfo <kernel module>" is automatically run for all installed modules and the lines beginning with "alias:" of the command's output are written in the aforementioned file, followed by the module's name.
(2) an udev rule like this one:

Code:

ACTION=="add", SUBSYSTEM=="pci", MODALIAS=="*", RUN+="/sbin/modprobe $modalias"

is usually triggered by a "udevadm trigger <options>" command, included in a startup script, which request device events from the kernel and apply the rule to theses events.

In our example the "modinfo e1000e" command's output includes, among others, following line:

Code:

alias: pci:v00008086d00001503sv*sd*bc*sc*i*

which says "this is an alias for pci id 8086:1503".
So after running "depmod -a" we will see in /lib/modules<kernel version>/modules.alias following line:

Code:

alias pci:v00008086d00001503sv*sd*bc*sc*i* e1000e

When the "udevadm trigger <options>" command will be run at startup, the kernel will say "Eh, I've found the pci device 8086:1503", then the udev rule will be applied, which after reading the relevant line of modules.alias, will trigger the "/sbin/modprobe e1000e" command for you.

Of course, all that process takes much less time to be executed than it took me to write it down ;)

To know more, see here.