Solved X stopped working after upgrading to 14.3.

I checked dmesg -a and discovered a couple of things which I am having some difficulty "translating". I'll list them in order.

atrtc0 can't map interrupt
hdac1 CODEC at address 2 not responding
KLD nvidia-modeset.ko depends on kernel not available or version mismatch
kldload: can't load nvidia_modeset: module already loaded or in kernel
/etc/rc WARNING Unable to load kernel module nvidia-modeset
/etc/rc WARNING unable to load kernel module usbhid
Invalid TXQ id


So, it looks to me as though nvidia-modeset is not being loaded at all.

Ken Gordon
 
1. Show us what "uname -a" says. Make sure you have 14.3. or else your upgrade to 14.3 failed.
2. Show us the first "url:" line in /etc/pkg/FreeBSD.conf -- it should be url: "pkg+https://pkg.FreeBSD.org/${ABI}/quarterly",
3. pkg install -f nvidia-kmod -- to make sure you have the correct bits.
4. Reboot. After that see from dmesg if nvidia-modeset loaded. If not, show us what the nvidia related lines are. 1. & 2. above ensure you have matching kernel and .ko files.
 
So, it looks to me as though nvidia-modeset is not being loaded at all.
No. It's loaded as 4th line of your dmesg shows. Just not working as of mis-match with kernel.

So, request 1 by bakul in post #27 is important here.
If you're building ports locally, running (or to be run after next reboot) kernel and kernel sources (under /usr/src/sys/) are needed to be in sync. Otherwise, the error you encountered could happen (not always, though).
 
1. Show us what "uname -a" says. Make sure you have 14.3. or else your upgrade to 14.3 failed.

uname -ashows 14.3-RELEASE-p5. I remember this from doing it a few days ago. I'll re-run that the next time I boot into FreeBSD (I am in Win10 just now) and will list the exact output from the command.

2. Show us the first "url:" line in /etc/pkg/FreeBSD.conf -- it should be url: "pkg+https://pkg.FreeBSD.org/${ABI}/quarterly",

That URL line is https://pkg.FreeBSDS.org/$(AB1)/quarterly. I had to remove the pkg+ from those lines in order for the upgrades to work. I believe, now, that the reason was that the {SRV} line was missing in that set of commands. I'll revisit those commands, reinsert the pkg+ part. and add the line containing {SRV} then try to correct matters.

3. pkg install -f nvidia-kmod -- to make sure you have the correct bits.

OK.

4. Reboot. After that see from dmesg if nvidia-modeset loaded. If not, show us what the nvidia related lines are. 1. & 2. above ensure you have matching kernel and .ko files.

Will do. And thanks again.

Ken Gordon
 
That URL line is https://pkg.FreeBSDS.org/$(AB1)/quarterly.
There are mutiple typo's in your line (no use of curly braces and no capital I, the I of Isaac ), check your actual reference:
Rich (BB code):
    you have:     https://pkg.FreeBSDS.org/$(AB1)/quarterly
it should be: pkg+https://pkg.FreeBSD.org/${ABI}/quarterly",
The addition of "pkg+" should work but we can deal with that later.

Just to be sure about your settings, please post (cut & paste) the output of
pkg -vv | sed -nE -e '/^OSVERSION|^ABI/ p' -e '/^Repositories:/,$ p'
 
Yeah, there's lots of times when a typo is all that stands between me and a successful setup of just about anything. That really taught me to be incredibly anal about them, and to never be in a rush typing stuff out. That's actually easier than trying to learn a sed/awk command to do the same thing and trying to debug THAT later 😂
 
I happened to have 14.3-RELEASE bhyve VM so I fired it up, freebsd-updated it to -p5, rebooted, installed nvidia-kmod and kldloaded nvidia-modeset. kldstat & dmesg showed nvidia-modeset.ko and nvidia.ko loaded fine. I am not passing the GPU to this VM but this shouldn't matter for now. So not sure what is going on with Ken's machine....
 
There are mutiple typo's in your line (no use of curly braces and no capital I, the I of Isaac ), check your actual reference:
Rich (BB code):
    you have:     https://pkg.FreeBSDS.org/$(AB1)/quarterly
it should be: pkg+https://pkg.FreeBSD.org/${ABI}/quarterly",

Yes, there are multiple typos in that post because I am not copying from my FreeBSD system, but am writing it all down on paper, then typing it in here. I am 83 years old and my eyesight is no longer very good.

So, here are two photos of what I meant to show: the first one is they way it was at first and the second photo is after I fixed it. FYI, adding the
Code:
pkg+
required me to also add the line
Code:
mirror_type: "srv"
. Without that line, including
Code:
pkg+
caused everything to fail. Here are the photos:
FreeBSD_conf.jpg


and after editing it:
FreeBSD_conf2.jpg


The addition of "pkg+" should work but we can deal with that later.

Well, only if the line mirror_type: "srv" is included.

Just to be sure about your settings, please post (cut & paste) the output of
pkg -vv | sed -nE -e '/^OSVERSION|^ABI/ p' -e '/^Repositories:/,$ p'

OK. Will do.

Ken Gordon
 
The following are photos of the results of certain commands requested by some who are trying to help me come to a conclusion on these issues. The first one shows the results of inputting uname -a as root:

uname-a.jpg


Next, shows the result of some of the latest dmesg -a showing that at least one of the nivida files is 1)not loaded, and 2) that there is a version mismatch somewhere.
Dmesg-a_Part.jpg


The last one shows the result of inputting startx as a normal user. This worked fine under 14.2-RELEASE-p3 but no longer works under 14.3-RELEASE-p5.
StartxResult.jpg


At this point, I am wondering if using freebsd-update rollback will at least put my system back to 14.2, but I doubt that will work, and at this point, I am less than enthusiastic about attempting it..

Ken Gordon
 
At this point, I am wondering if using freebsd-update rollback will at least put my system back to 14.2, but I doubt that will work, and at this point, I am less than enthusiastic about attempting it..
I'm also less enthousiastic about that, most importantly because packages for 14.2-RELEASE are no longer available!

Yes, there are multiple typos in that post because I am not copying from my FreeBSD system, but am writing it all down on paper, then typing it in here. I am 83 years old and my eyesight is no longer very good.
No worries. I'm guessing and hoping that you are not responding by means of typing things on a mobile phone of sorts but on a system on hand running perhaps MS Windows.

I mention this as an option to perhaps consider trying later (we can deal with photo's under these exceptional circumstances); for example from Windows you can set up a terminal connection by means of PuTTY and then given a working network connections of course, you can easily cut and paste any non-graphical stuff.
 
A couple more things to try to see what is going on. Report what kldxref -d /boot/modules/nvidia-modeset.ko returns. I get
# kldxref -d /boot/modules/nvidia-modeset.ko
/boot/modules/nvidia-modeset.ko
module nvidia_modeset
depends on kernel.1403000 (1403000,1499999)
interface nvidia_modeset.1
depends on nvidia.1 (1,1)
depends on linux.1 (1,1)
depends on linux_common.1 (1,1)

That kernel.1403000 line indicates the kernel version(s) for which this .ko will supposedly work.

Second report what md5 /boot/modules/nvidia.ko /boot/modules/nvidia-modeset.ko returns. I get
# md5 /boot/modules/nvidia.ko /boot/modules/nvidia-modeset.ko /boot/kernel/kernel
MD5 (/boot/modules/nvidia.ko) = 6189c33384d10ad6ff34e73312eda90d
MD5 (/boot/modules/nvidia-modeset.ko) = cd99eb9a6c7bc03be69826caff9f2d31
MD5 (/boot/kernel/kernel) = fc523e4840a1bc75f11c4c46b3e5b2b0


Third see if /boot/kernel/linker.hints and /boot/modules/linker.hints exist. They are created by kldxref.
 
Thank you all for the suggestions. I will follow up on those asap. Dinner time here now, and we have guests.

I must admit to getting darned tired of dealing with this issue. I MAY decide to wait until Ver. 15-RELEASE is available, and then do a complete start-over.

As I have mentioned here several times, my main reason for moving to FreeBSD in the first place was to get completely rid of Billy Gates' garbage. And I had come very close to achieving what I wanted before running into this problem.

Rats!

Ken Gordon
 
I think it helps to have a backup plan for when your graphics go out. This is one of the FreeBSD quirks you must navigate.
Generally I don't touch my packages when we are between major versions switch. It can be a minefield.
SCFB is not that bad.
True that the resolution is limited to GOP sizes and it is not accelerated but for a short stint in a position like this it should be considered.
There is no way I would dump my desktop for a graphics problem.

example 4
 
Also note this from above manual:
To configure multiple graphics processors, the BusID can be added.A list of graphics processor bus IDs can be displayed by executing:

If you have dual graphics like Optimus laptop you need to add Intel CPU graphics BUSID to scfb configuration file.
 
It appears that the nvidia-modeset module is included in any Nvidia driver. I recently upgraded to the latest Nvidia driver, which is version 580. Attempting to find nvidia-modeset on my system returns absolutely nothing. I am now wondering if the module was somehow left out of the latest driver-set.

I will attempt to drop back to a later version of the Nvidia driver-set, which I believe is 507, to see if this makes any difference in my present issue.

Later,

Ken Gordon
 
It appears that the nvidia-modeset module is included in any Nvidia driver. I recently upgraded to the latest Nvidia driver, which is version 580. Attempting to find nvidia-modeset on my system returns absolutely nothing. I am now wondering if the module was somehow left out of the latest driver-set.

I will attempt to drop back to a later version of the Nvidia driver-set, which I believe is 507, to see if this makes any difference in my present issue.

Later,

Ken Gordon
Kernel modules for nvidia GPUs other than nvidia-drm.ko is installed by x11/nvidia-kmod and its variants (-304, -340, -390, -470 or -devel) now.

These were included in corresponding x11/nvidia-driver ports before, but splitted out to allow kmod repo builder to support nvidia GPU driver kmods recently at commit f4e907a49258. This is done before 2025Q4 branches.
 
OK. I'm going to begin again:
It seems obvious to me that the nvidia-modeset.ko module, which is required by X, is not being loaded.

Inputting pkg info nvidia* returns the following:

nvidia-driver-580.95.05
nvidia-kmod-580.95.05.1403506
nvidia-settings-580.95.05
nvidia-xconfig-580-95.05


Therefore, all the necessary nvidia files SHOULD be installed OK.

Next, inputting kldstat returns a long list of .ko files, NONE of which are nvidia-modeset.ko. The list is quite long, so I hesitate to type them all here at this point.

Next, inputting kldload nvidia-modeset returns the following:
KLD nvidia-modeset.ko: depends on kernel - not available or version mismatch
kldload: can't load nvidia-modeset: module already loaded or in kernel


Except, it isn't.

This indicates to me that the error message itself is either wrong, or needs "translation".

J think my next step is to UNinstall all the present nvidia files, and install earlier ones, if that is possible. I am suspecting that the latest version, 580, is either missing the nvidia-modeset module or it has been misnamed, or that there is something within FreeBSD-14.3-RELEASE which is preventing it from being loaded or recognized.

Ken Gordon
 
OK. Success. X is now working. Here is what I did:

pkg delete nvidia*

The system them deleted all of the nvidia-580 files, 4 of them.

I then input the following:

pkg search nvidia

The result was a long list of available nvidia apps. The most recent prior to version 580 was 470.

I then input the command: pkg install nividia-driver-470-470.256.02_2

The system then installed that app, and the app nvidia-kmod-470-470.256.02.14030000

I then input kldload nvidia-modest.ko. which finished correctly.

I then exited root, cd-ed to home/Ken and input startx

X then started correctly, loaded my desktop (MATE) and everything appears correct.

I then started a terminal, and input uname -a, returned was:

FreeBSD KensOffice 14.3-RELEASE-p5 FreeBSD 14.3-RELEASE-p5 GENERIC amd64

So, it would appear that the problem is that the latest version of the nividia drivers are either missing the nvidia-modeset.ko module, or it is misnamed, or otherwise screwed up, and that MY problem was NOT based in FreeBSD-14.3-RELEASE at all.

Ken Gordon
 
So, it would appear that the problem is that the latest version of the nividia drivers are either missing the nvidia-modeset.ko module, or it is misnamed, or otherwise screwed up, and that MY problem was NOT based in FreeBSD-14.3-RELEASE at all.
It cannot happen. nvidia.ko and nvidia-modeset.ko is built in single port at once.
Code:
# cd /usr/ports/x11/nvidia-kmod
# make
  (snip)
===>  Staging for nvidia-kmod-580.95.05.1500501
===>   Generating temporary packing list
===> src (install)
===> src/nvidia (install)
install -T release -o root -g wheel -m 444   nvidia.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
===> src/nvidia-modeset (install)
install -T release -o root -g wheel -m 444   nvidia-modeset.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
===> firmware (install)
===> firmware/gsp_tu10x (install)
install -T release -o root -g wheel -m 444   nvidia_gsp_tu10x_fw.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
===> firmware/gsp_ga10x (install)
install -T release -o root -g wheel -m 444   nvidia_gsp_ga10x_fw.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
====> Compressing man pages (compress-man)
# # ls -l work/stage/boot/modules/
total 167886
-r--r--r--  1 root wheel   2106704 Oct 30 07:24 nvidia-modeset.ko
-r--r--r--  1 root wheel 112565392 Oct 30 07:24 nvidia.ko
-r--r--r--  1 root wheel  74848192 Oct 30 07:24 nvidia_gsp_ga10x_fw.ko
-r--r--r--  1 root wheel  30323752 Oct 30 07:24 nvidia_gsp_tu10x_fw.ko
# make package
===>  Building packages for nvidia-kmod-580.95.05.1500501
===>   Building nvidia-kmod-580.95.05.1500501
# tar tvf work/pkg/nvidia-kmod-580.95.05.1500501.pkg
-rw-r--r--  0 root   wheel    2343 Jan  1  1970 +COMPACT_MANIFEST
-rw-r--r--  0 root   wheel    3948 Jan  1  1970 +MANIFEST
-rw-r--r--  0 root   wheel     199 Oct 30 07:24 /usr/local/share/licenses/nvidia-kmod-580.95.05.1500501/catalog.mk
-rw-r--r--  0 root   wheel      89 Oct 30 07:24 /usr/local/share/licenses/nvidia-kmod-580.95.05.1500501/LICENSE
-rw-r--r--  0 root   wheel   22653 Oct 30 07:24 /usr/local/share/licenses/nvidia-kmod-580.95.05.1500501/NVIDIA
-r--r--r--  0 root   wheel 112565392 Oct 30 07:24 /boot/modules/nvidia.ko
-r--r--r--  0 root   wheel   2106704 Oct 30 07:24 /boot/modules/nvidia-modeset.ko
-r--r--r--  0 root   wheel  74848192 Oct 30 07:24 /boot/modules/nvidia_gsp_ga10x_fw.ko
-r--r--r--  0 root   wheel  30323752 Oct 30 07:24 /boot/modules/nvidia_gsp_tu10x_fw.ko
#

If any step fails, pkg simply dissapears from the repo, or fully succeeded.

Have you read my comment #16 in this thread?
 
It cannot happen. nvidia.ko and nvidia-modeset.ko is built in single port at once.
Code:
# cd /usr/ports/x11/nvidia-kmod
# make
  (snip)
===>  Staging for nvidia-kmod-580.95.05.1500501
===>   Generating temporary packing list
===> src (install)
===> src/nvidia (install)
install -T release -o root -g wheel -m 444   nvidia.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
===> src/nvidia-modeset (install)
install -T release -o root -g wheel -m 444   nvidia-modeset.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
===> firmware (install)
===> firmware/gsp_tu10x (install)
install -T release -o root -g wheel -m 444   nvidia_gsp_tu10x_fw.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
===> firmware/gsp_ga10x (install)
install -T release -o root -g wheel -m 444   nvidia_gsp_ga10x_fw.ko /usr/ports/x11/nvidia-kmod/work/stage/boot/modules/
====> Compressing man pages (compress-man)
# # ls -l work/stage/boot/modules/
total 167886
-r--r--r--  1 root wheel   2106704 Oct 30 07:24 nvidia-modeset.ko
-r--r--r--  1 root wheel 112565392 Oct 30 07:24 nvidia.ko
-r--r--r--  1 root wheel  74848192 Oct 30 07:24 nvidia_gsp_ga10x_fw.ko
-r--r--r--  1 root wheel  30323752 Oct 30 07:24 nvidia_gsp_tu10x_fw.ko
# make package
===>  Building packages for nvidia-kmod-580.95.05.1500501
===>   Building nvidia-kmod-580.95.05.1500501
# tar tvf work/pkg/nvidia-kmod-580.95.05.1500501.pkg
-rw-r--r--  0 root   wheel    2343 Jan  1  1970 +COMPACT_MANIFEST
-rw-r--r--  0 root   wheel    3948 Jan  1  1970 +MANIFEST
-rw-r--r--  0 root   wheel     199 Oct 30 07:24 /usr/local/share/licenses/nvidia-kmod-580.95.05.1500501/catalog.mk
-rw-r--r--  0 root   wheel      89 Oct 30 07:24 /usr/local/share/licenses/nvidia-kmod-580.95.05.1500501/LICENSE
-rw-r--r--  0 root   wheel   22653 Oct 30 07:24 /usr/local/share/licenses/nvidia-kmod-580.95.05.1500501/NVIDIA
-r--r--r--  0 root   wheel 112565392 Oct 30 07:24 /boot/modules/nvidia.ko
-r--r--r--  0 root   wheel   2106704 Oct 30 07:24 /boot/modules/nvidia-modeset.ko
-r--r--r--  0 root   wheel  74848192 Oct 30 07:24 /boot/modules/nvidia_gsp_ga10x_fw.ko
-r--r--r--  0 root   wheel  30323752 Oct 30 07:24 /boot/modules/nvidia_gsp_tu10x_fw.ko
#

If any step fails, pkg simply dissapears from the repo, or fully succeeded.

Have you read my comment #16 in this thread?

Well, I read your comment at messages #17, thoroughly. Have you read my reply at message # 21?

I suppose there is something else which could cause version 580 of the Nvidia drivers to not work correctly, but if everything in version 580 is exactly correct, why did changing my system to Nvidia drivers version 470 fix my problem, and why was I getting the "...not available or mismatch" error when using version 580?

I have just now read your comment #16, and found this:

20250930:<br> AFFECTS: users of x11/nvidia-driver*<br> AUTHOR: junchoon@dec.sakura.ne.jp<br><br> Because of splitting out kernel modules (kmod) part from x11/nvidia-driver*<br> into newly introduced x11/nvidia-kmod*, you need to deinstall currently<br> installed x11/nvidia-driver* and freshly install it again with new version.<br> This would pull in corresponding x11/nvidia-kmod* because x11/nvidia-driver*<br> explicitly depends on corresponding x11/nvidia-kmod*.

According to that, I would have to, first delete all nvidia apps, then re-install version 580 in order to pull in nvidia-kmod files. OK. I'll try that.

Standby.

Ken Gordon
 
OK. I first pkg delete nvidia*

Then did pkg install nvidia-driver. Still no nvidia-modeset module loaded.

I get the same error messages as previously.

Rebooted, and did all the above again.

STILL get the same error messages and nvidia-modeset is NEVER loaded.

Going back to my previous method of installing version 470.

Ken Gordon
 
20250930:&lt;br&gt; AFFECTS: users of x11/nvidia-driver*&lt;br&gt; AUTHOR: junchoon@dec.sakura.ne.jp&lt;br&gt;&lt;br&gt; Because of splitting out kernel modules (kmod) part from x11/nvidia-driver*&lt;br&gt; into newly introduced x11/nvidia-kmod*, you need to deinstall currently&lt;br&gt; installed x11/nvidia-driver* and freshly install it again with new version.&lt;br&gt; This would pull in corresponding x11/nvidia-kmod* because x11/nvidia-driver*&lt;br&gt; explicitly depends on corresponding x11/nvidia-kmod*.

The above does NOT work for me.

Dropping back to version 470 DOES work for me.

Ken Gordon
 
Back
Top