Bug#941336: nvidia-legacy-340xx-kernel-dkms: Kernel module nvidia-legacy-340xx can't be loaded
Paul Vojta
vojta at math.berkeley.edu
Wed May 20 01:48:26 BST 2020
On Fri, Apr 24, 2020 at 09:44:39PM +0200, Andreas Beckmann wrote:
> On 22/01/2020 01.26, Paul Vojta wrote:
> > I noticed similarities with Bug #941427 (relating to non-legacy nvidia drivers),
> > and I found that the procedure in Message 10 in that bug report worked for me.
>
> That procedure at some point caused the kernel module to be rebuilt in a
> "clean" environment.
>
> > I suspect it's a problem with some package's "Depends" or "Pre-depends"
> > declarations.
> >
> > (This solves the problem reported in this bug report, but the machine still
> > has problems with suspend/resume. That's a problem for another bug report.)
>
> I'm closing that bug now, since it is unclear how to reproduce the
> broken module build.
>
>
> Andreas
Dear Andreas,
Thank you for your recent email.
Attached are my detailed notes on how to reproduce the bug.
Tl;DR:
Hardware is Apple MacMini4.1, purchased new Sept. 2010,
Core 2 Duo 2.4GHz, 2G ram.
Boot is via rEFInd.
Installed nvidia-legacy driver suite via:
apt-get -y install dkms xserver-xorg-core libgtk2.0-0
followed by
apt-get -y install xserver-xorg-video-nvidia-legacy-340xx
Things worked fine upon rebooting after first "apt-get".
Upon rebooting after the second one, the bug appeared, with dmesg
showing a kernel oops.
The notes were from February 3 (and earlier). I'd been meaning to send them
earlier, but I've been really busy with other stuff.
Sincerely,
Paul Vojta
-------------- next part --------------
Got bullseye netinst image:
wget https://cdimage.debian.org/cdimage/bullseye_di_alpha1/amd64/iso-cd/debian-bullseye-DI-alpha1-amd64-netinst.iso
It is dated Nov. 30 2019, and its md5sum is:
d31305eb576f61b0af43e91a2400ff1f debian-bullseye-DI-alpha1-amd64-netinst.iso
Burned it to a CD.
1. Boot for initial install.
Start up computer. Hit some (arrow) keys at the rEFInd screen to stop its
countdown. Insert CD, and reboot from within rEFInd.
Then boot from within rEFInd using the CD icon. (Or, repeatedly hit the `c'
key immediately after the startup sound to get it to boot the CD directly.)
Select "Graphical install" [wait about 30 seconds]
Choose language, location, keyboard, and network (choose wired).
For some reason it sometimes asks for the firmware file brcm/bcm43xx-0.fw,
but when I insert removable media containing it, it just repeats the
request.
Set up users, clock
Partitioning:
First click "Go Back", scroll down, and select "Execute a shell".
mount -t ext4 /dev/sda5 /mnt
cd /mnt
mkdir z3
mv bin boot dev etc home opt proc root run sbin usr var z3
rm i* v* l*
mkdir -p usr/lib
cp -pr z3/usr/lib/firmware usr/lib
cd /
umount /mnt
exit
Select "Partition disks"
Choose "Manual partitioning"
For the ext4 partition:
select the partition, click "Continue",
Select "Use as", click "Continue", choose ext4, click "Continue"...
Select "Format the partition", click "Continue", choose, "no, keep ...",
click "Continue"
Select "Mount point", click "Continue", choose "/", click "Continue"
Select "Done setting up the partition", click "Continue"
For the swap partition (this is probably unnecessary):
Set "Use as" to "Swap area" (same procedure as above)
Select "Done setting up the partition", click "Continue"
Scroll down and select "Finish partitioning and write the changes to disk".
Installation... [about 7 minutes]
Scan extra installation media: No
For mirrors, select "United States", and mirrors.ocf.berkeley.edu
Proxy: No
[about 4 minutes]
Popularity-contest: No
Select software to install: only ssh server and standard system utilities
[about 5 minutes]
Install boot loader to master boot record (of /dev/sda)
Rebooted. Chose "Boot from FAT volume" in rEFInd (choosing OS didn't work).
cp -p /z3/etc/ssh/ssh_host* /etc/ssh
[I can ssh to the machine now]
vi /etc/apt/sources.list
[added contrib and non-free to main two entries]
apt-get update
ln /z3/var/cache/apt/archives/*.deb /var/cache/apt/archives
apt-get install tcsh nvi firmware-brcm80211 firmware-misc-nonfree
chsh -s /usr/bin/tcsh
chsh -s /usr/bin/tcsh vojta
vi /etc/default/keyboard
[add ctrl:swapcaps to XKBOPTIONS]
dpkg-reconfigure debconf
[Choose readline interface]
apt-get -y install dkms xserver-xorg-core libgtk2.0-0
Rebooted. Chose "Boot vmlinuz-5.4.0-3-amd64" in rEFInd.
Boot went OK. I was able to log in (via getty, which was expected).
apt-get -y install xserver-xorg-video-nvidia-legacy-340xx
Rebooted, same method
Screen went black during the reboot sequence. I was able to log in
remotely via ssh, and from the console (as a getty login, as before),
but the latter wasn't very useful since there was no video showing.
dmesg output showed a kernel (or driver?) oops.
To burn it onto a USB stick:
Boot the MacOS.
Insert the USB stick.
Follow the directions from the web site:
https://linuxnewbieguide.org/how-to-put-a-linux-iso-onto-a-usb-stick-and-make-it-bootable-on-a-mac/
It says open up Applications --> Utilities --> Disk Utility
(it popped that up for me when I inserted the USB drive, which had the
iso copied to it via dd).
Select the USB stick and click "erase"
In a command window now:
diskutil list
Locate the USB drive (it's /dev/disk1 in this case), and type:
diskutil unmountDisk /dev/disk1
hdiutil convert -format UDRW -o netinst2.img netinst.iso
sudo dd if=netinst2.img.dmg of=/dev/disk1
[up to 14 minutes]
[A popup will appear. Important: ignore it.]
diskutil eject /dev/disk1
Booting from the stick:
To boot from it, reboot and hold down the alt key while it's coming up.
This only seems to have messed up rEFInd.
Fixing rEFInd:
I reinstalled refind (booted up MacOS, found install.sh in
~/Downloads/refind*, and ran it).
Now the bug is back.
However, the bug only occurs if I select "linux-5.4.0-3" from the
rEFInd screen. If I boot from the disk's boot sector, then the boot
process goes to a grub screen and things work. Note that the grub
screen is 80x24 (or maybe 80x25), whereas when booting via rEFInd
the screen has a much higher resolution.
More information about the pkg-nvidia-devel
mailing list