In a little more detail: edit /etc/apt/sources.list, add contrib non-free at the end of lines ending with main, then run sudo apt install apt-file, sudo apt update, and then look for packages providing the firmware files you need with apt-file search
so here's a question. My system had an nvidia driver installed and working fine 24 hours ago. I've since completely messed up the driver situation. Is it possible to "boot to a previous kernel" from the command line such that I'll restore the old kernel with the proper driver?
@duhaime The main issue I see with that approach is that you won't be able to access the remote system at all if you make any mistake that makes it unable to boot.
@duhaime /var/log/apt/history.log. And with integers appended.
@duhaime I mean, it is a high possibility no matter what you do. I once talked someone through a remote downgrade of his Debian system (or possibly Ubuntu). So it's possible. But it was definitely a bit tense.
@duhaime No, upgrading the whole system would. (I'm not on Ubuntu (nor Debian), so I'm not sure about the actual commands; basically, apt update and then apt upgrade. Did you run something like this?).
does stack overflow support the show/hide details syntax like github markdown does? so I could keep the gory details for the insatiable but keep them tidy and out of sight?
that gives: ``` ii libnvidia-cfg1-440:amd64 440.64.00-0ubuntu1 amd64 NVIDIA binary OpenGL/GLX configuration library ii libnvidia-common-440 440.64.00-0ubuntu1 all Shared files used by the NVIDIA libraries rc libnvidia-compute-390:i386 390.116-0ubuntu0.18.04.1 i386 NVIDIA libcompute package ii libnvidia-compute-440:amd64 440.64.00-0ubuntu1 amd64 NVIDIA libcompute package
@duhaime What do you think you did to mess up the driver situation? Your logs show you have installed/uninstalled several things. What step did start it all?
I'm wondering if I should try removing all nvidia drivers and installing afresh, as the nvidia error log plainly informs me that the driver install failed: gist.githubusercontent.com/duhaime/…
@duhaime Yes, that sounds reasonable. Unless it makes your system crash. But you should be clear what you are trying to install. And also confirm it works with your card
@duhaime That log is relevant, yes. It suggests to look at /var/lib/dkms/nvidia/410.48/build/make.log and /var/log/nvidia-installer.log. What is the content of those files?
> WARNING: Unable to determine the path to install the libglvnd EGL vendor library config files. Check that you have pkg-config and the libglvnd development libraries installed, or specify a path with --glvnd-egl-config-path. Will install libEGL vendor library config file to /usr/share/glvnd/egl_vendor.d
Though with Debian binary files, dependencies should be taken care of. That's one of the many reasons to use them.
What does dpkg -l | grep libglvnd show?
Oh, and one other file to look at is the Xorg log file.
hmm, there is no /var/lib/dkms/nvidia/410.48/build/make.log. The installer log is the file linked above (gist.githubusercontent.com/duhaime/…) -- it references iself
dpkg -l | grep libglvnd returns:
ii libglvnd-core-dev:amd64 1.0.0-2ubuntu2.3 amd64 Vendor neutral GL dispatch library -- core development files
ii libglvnd-dev:amd64 1.0.0-2ubuntu2.3 amd64 Vendor neutral GL dispatch library -- development files
ii libglvnd0:amd64 1.0.0-2ubuntu2.3 amd64 Vendor neutral GL dispatch library
@duhaime Ok. Could you please check what is on the "Kernel modules" line in the blocks following "VGA compatible controller: NVIDIA..." in the output of lspci -k?
@duhaime Considering you've probably done a local NVDIA installation, I think it would be best if you got rid of everything nvidia related, including any local installations, and start over.
You could also clean up your kernels a bit while you're about it.
@duhaime A few things you may find useful: according to the doc on download.nvidia.com, nvidia-installer should have a --uninstall option meant to restore the system to its pre-install state. Though I'm afraid it likely won't work as expected if you also installed nvidia-related stuff using aptafter having run nvidia-installer.
When in doubt, to know where a loaded module comes from, you can use modinfo (searching for the filename line, e.g. modinfo | grep '^filename') and then use dpkg -S /path/to/file to search for the package which owns that file.
@fra-san That's a good point. I'd run the uninstall option to get rid of the local files. It should be fine if the plan is just to remove all NVIDIA related stuff.
@FaheemMitha Thanks. I've been using Ubuntu for a while a few years ago (never Debian), but sadly I'm not familiar with even the most basic commands anymore.
@duhaime Some basics: There are two drivers for nvidia cards: nouveau and nvidia. Both mutually incompatible. If one is loaded the other most of the time doesn't work.
@duhaime If you want nvidia: The installed kernel defines the number of the Nvidia driver to use (if it exists). Here is the search page from Nvidia: nvidia.com/Download/index.aspx?lang=en-us
@duhaime DO NOT DOWNLOAD THOSE, I am linking only to find out what is available from NVIDIA.
@duhaime says that the driver number should be 440.82 (you had installed 435.21 before borking it (your words)).
@duhaime The exact path to install any of those drivers is usually complex, convoluted and prone to error. Sometimes I think that NVIDIA wants it that way for Linux. You need expert help, I can not stress that enough. Find someone that knows and follow his (and only his) advice.
@duhaime However, I have the impression that using a card for computing (anything) makes it unable to provide video (I might be very wrong, as I am very lost on this area). So, maybe, on installing CUDA you will have everything you actually need.
@duhaime I still believe that you need a full re-install, you have tried everything that is out there making the problem even more complex that it needed to be and creating additional problems.
@duhaime Is there anyone that could physically access the computer and insert an Ubuntu live pen-drive? If it works on a live distro it will work on the installed system. Just an idea .... :-)
@duhaime In any case keep the problem at askubuntu.com, not here, Please. People versed on such issues usually live there, not here.
@duhaime And, good luck. :-)
@Kusalananda Yes, the intr option has been deprecated since kernel 2.6. But, did you know that soft has this warning on the manual ?? : NB: A so-called "soft" timeout can cause silent data corruption in certain cases.
@Kusalananda Ah, yes, thanks: *Also, @Isaac, good modification of your answer there. +1 * It becomes easy to correct a mistake when one is told which one is it.
@Isaac OpenBSD uses another NFS implementation, but I would expect there to be data corruption if the a softmounted NFS mount goes away for an extended period (triggering the timeout). It's akin to pulling a disk out of a running system.
@Kusalananda There is no more detail of which corruption would occur. A truncated file because it got forcefully stopped in the middle of a copy doesn't sound like "silent corruption" to me. But I really don't remember all the details. The alternative is hard but that blocks for quite long (as long as the server takes to recover).
@Kusalananda [This answer](https://serverfault.com/a/9506/427450) To this question: What are the advantages/disadvantages of hard versus soft mounts in UNIX? Seems to apply to UNIX, doesn't it ?
@Kusalananda I suppose that that is a very fine issue and that that is not very much relevant to a personal way to backup data. But worth knowing in any case.
@Kusalananda change fine to small, probably conveys better my meaning.
@StephenKitt Stephen, do you mean that Debian Stable will not upgrade an application package, e.g. not upgrading Wine 3.0 to Wine 3.1 or higher? Will not upgrade Linux kernel from hypothetically 1.1 to 1.2?
The “stable” releases don’t get much in the way of updates; only fixes for security issues, shipped continuously, and fixes for serious bugs, which are shipped periodically in point releases. Fixes are normally backported to the version of the package which shipped initially in any given release; there are exceptions, e.g. for Chromium and Firefox ESR.
I have always heard that Debian Stable and Ubuntu LTS are comparable. I am using Ubuntu LTS, and by apt update and apt upgrade I can always get higher versions of Linux kernel and application packages. So how are Debian Stable and Ubuntu LTS comparable?
@Tim In stable It depends. If some security update or serious bug require the use of a new version to correct the problem, then, any package will get upgraded (updated to the new version). In the specific case of Wine, yes, wine will most of the time will be retained at the version that shiped with the original release of Debian. The kernel, for sure will remain static (in version number, not in the binary file) for the whole life of stable.
@Tim However, testing will get upgraded packages quite quickly since their release from the developer. In any case, many times (within reason) a newer package could be installed from backports or even from sid (be careful, very careful in this case).
