/boot/kernel/ums.ko - unsupported file type ; link_elf_obj: symbol evdev_is_grabbed undefined

Status
Not open for further replies.
ziomario

ziomario

Hello to everyone.

I've just upgraded from 13.1-RELEASE-p4 to 13.1-RELEASE-p6 and something is not working correctly anymore. The problem is that when xfce4 is ready to be used,my mouse is frozen. I've changed the USB port where it is attached,but it is still frozen. So,I've thought that the problem might not be the mouse,but something else. The cause of the error is shown on the picture below :

index.jpg


The error is caused by the file ums.ko that replaced the old ums.ko file. The new file may be corrupted or not well patched,because I've swapped it with the older one and everything worked great again. So,now I'm using the p6 but with the ums.ko file which belongs to p4. I want to clarify that the new ums.ko file has been put there by FreeBSD when I have upgraded p4 to p6 issuing the commands : "freebsd-update fetch ; freebsd-update install",not manually by me. The error is not the result of some experiment made by me. Could it be caused by a bug that I should report somewhere ? where ?
 
If you are 100% sure that the problem module came from freebsd-update, then yes, it needs to be reported. Though given that ums is one of the most used drivers, I'd expect a lot of noise already about this if freebsd-update broke it for someone.

Also, "ums.ko is newer than linker.hints file" message suggests otherwise.
 
yuripv79 said:
If you are 100% sure that the problem module came from freebsd-update, then yes, it needs to be reported. Though given that ums is one of the most used drivers, I'd expect a lot of noise already about this if freebsd-update broke it for someone.

Also, "ums.ko is newer than linker.hints file" message suggests otherwise.
Click to expand...

I'm sure yes,because I swapped the new ums.ko file with the older one and now I'm using the system again. The error "ums.ko is newer than linker.hints file" started to appear when I have swapped the files.
 
p6 didn't update or modify ums(4).

Code: 
The following files will be updated as part of updating to
13.1-RELEASE-p6:
/bin/freebsd-version
/boot/kernel/if_ena.ko
/boot/kernel/if_ix.ko
/boot/kernel/kernel
/lib/geom/geom_eli.so
/usr/lib32/geom/geom_eli.so
/usr/share/zoneinfo/America/Bogota
/usr/share/zoneinfo/America/Cambridge_Bay
/usr/share/zoneinfo/America/Godthab
/usr/share/zoneinfo/America/Inuvik
/usr/share/zoneinfo/America/Iqaluit
/usr/share/zoneinfo/America/Nuuk
/usr/share/zoneinfo/America/Ojinaga
/usr/share/zoneinfo/America/Pangnirtung
/usr/share/zoneinfo/America/Rankin_Inlet
/usr/share/zoneinfo/America/Resolute
/usr/share/zoneinfo/America/Whitehorse
/usr/share/zoneinfo/America/Yellowknife
/usr/share/zoneinfo/Asia/Kuala_Lumpur
/usr/share/zoneinfo/Asia/Singapore
/usr/share/zoneinfo/Canada/Yukon
/usr/share/zoneinfo/Singapore
/usr/share/zoneinfo/zone.tab
/usr/share/zoneinfo/zone1970.tab
 
Right, but looks like it modified the /boot/kernel/kernel, and it's where the evdev_is_grabbed symbol comes in 13-STABLE and -CURRENT (device evdev is in GENERIC).

ziomario did you build your kernel with patches, or from -STABLE source, and now the kernel binary was reverted by freebsd-update, while ums.ko was not?
 
Yeah, I'm thinking at some point or another ziomario built a custom kernel and did it from the wrong branch.
 
Can you give a look at the size of the ums.ko file used on the p6 ? In my case is : 39.0 KiB (39936 bytes) ; this file freezes the system. The old ums.ko file that I'm using come from p4 ; it does not freeze the system and it's size is different : 38.9 KiB (39864 bytes)
 
Code: 
# freebsd-version -kru

13.1-RELEASE-p6
13.1-RELEASE-p6
13.1-RELEASE-p6

Nope,I don't think that I've built a custom kernel. The freezing happened after having upgraded from p4 to p6.
 
A guess. It could be you need to rebuild some packages with modules. Then maybe the linker.hints file messages will go away.
If applicable it can be a good idea to rebuild drm-510-kmod and/or nvidia-driver-510.60.02
 
It's hard to do that because the ums.ko file of p6 freezes the mouse. I tried to use a different DM (lxde) and I saw the mouse also frozen. Anyway,it does not freezes the keyboard or any other device. What you suggest me to do ?
 
Code: 
[marietto@marietto /]$ md5 /boot/kernel/ums.ko
MD5 (/boot/kernel/ums.ko) = bf011d5f6bf65b30ffb5444921eb5eee

[marietto@marietto /]$ md5 /boot/kernel/ums.ko.no
MD5 (/boot/kernel/ums.ko.no) = 2e2eac01dc1ff4593af33904cb34e449
 
yuripv79 said:
Right, but looks like it modified the /boot/kernel/kernel, and it's where the evdev_is_grabbed symbol comes in 13-STABLE and -CURRENT (device evdev is in GENERIC).

ziomario did you build your kernel with patches, or from -STABLE source, and now the kernel binary was reverted by freebsd-update, while ums.ko was not?
Click to expand...

Nop.
 
Mario fyi,

- I pulled the git kernel source 13.2-beta1
- build the kernel
- installed the kernel
- rebuild : x11/nvidia-driver
- rebuild : graphics/drm-510-kmod
- rebuild :graphics/drm-kmod
- Installed these 3 packages.

/boot/kernel/ums.ko
Size : 34376 bytes
md5sum : 7d44f3b1100a4bd90edc1a24ce281a87

I have not rebuild world yet so,
freebsd-version -kru
13.2-BETA1
13.2-BETA1
13.1-RELEASE-p5
 
Alain De Vos said:
Mario fyi,

- I pulled the git kernel source 13.2-beta1
- build the kernel
- installed the kernel
- rebuild : x11/nvidia-driver
- rebuild : graphics/drm-510-kmod
- rebuild :graphics/drm-kmod
- Installed these 3 packages.

/boot/kernel/ums.ko
Size : 34376 bytes
md5sum : 7d44f3b1100a4bd90edc1a24ce281a87
Click to expand...

the 13.2-beta 1 ums.ko file is different from the ums.ko file that does not work here. Its MD5 (/boot/kernel/ums.ko.no) = 2e2eac01dc1ff4593af33904cb34e449 ; and anyway I didn't perform any upgrade to 13.2.
 
Alain De Vos said:
A guess. It could be you need to rebuild some packages with modules. Then maybe the linker.hints file messages will go away.
If applicable it can be a good idea to rebuild drm-510-kmod and/or nvidia-driver-510.60.02
Click to expand...

I don't use the nvidia-driver-510.60.02 anymore,but 525.78.01. To learn how to do that I started a new thread here :

forums.freebsd.org

Solved - How to install on FreeBSD 13.1-RELEASE a different nvidia-driver version than one that exists between ports and packages.

Hello to every FreeBSD lover. I'm making an experiment which involves FreeBSD 13.1-RELEASE and Linux/Ubuntu 20.04. It provides that the nvidia drivers of these OSes are of the same version. So,actually Ubuntu supports the following nvidia-driver versions : Ubuntu 20.04 : 510.108.03 ...
forums.freebsd.org forums.freebsd.org

and it seems that it works great :

Code: 
# nvidia-smi
Thu Feb 16 20:31:07 2023        
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.78.01    Driver Version: 525.78.01    CUDA Version: N/A      |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
| 50%   36C    P0    30W / 120W |      0MiB /  3072MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  NVIDIA GeForce ...  Off  | 00000000:02:00.0 Off |                  N/A |
| 31%   57C    P0    74W / 250W |      0MiB / 11264MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+
 
anyway,there is something that does not work. I don't know if it is relevant. When I invoke the nvidia-smi with the shim,it does not show the CUDA version used :

Code: 
# /home/marietto/Desktop/Files/CG/CUDA/libc6-shim/bin/./nv-sglrun nvidia-smi

shim init
Thu Feb 16 20:46:21 2023        
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.78.01    Driver Version: 525.78.01    CUDA Version: N/A      |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
| 50%   34C    P0    29W / 120W |      0MiB /  3072MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
|   1  NVIDIA GeForce ...  Off  | 00000000:02:00.0 Off |                  N/A |
| 29%   44C    P0    44W / 250W |      0MiB / 11264MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

but the version of CUDA is shown if I do the same within the ubuntu linuxulator. Don't know why.
 
yeah,it seems that it works like this. Infact as I said if I run "nvidia-smi" within the ubuntu userland,I see that it is using CUDA 12. But you know that shkhln created the shim to use the CUDA libraries directly in FreeBSD ? So,I can install Blender from the FreeBSD ports and it can use CUDA. I wonder why when I invoke nvidia-smi with the tool sglrun,it does not show the CUDA version used. BUT cycles is able to recognize the CUDA libraries and I can render the 3D animations using my nvidia GPU (outside the linuxulator). Anyway I'm not sure that all of this is relevant with the wrong ums.ko file installed today by freebsd-update. What's the role of the ums.ko file ? what should it do ?
 
Status
Not open for further replies.
Back
Top