unable to start GUI on new Thinkpad

[sacerdos_daemonis]

My computer died and I bought a new one. The last one had a Radeon driver that worked well with OpenBSD so I bought another Radeon figuring it would have a good chance of being compatible. OpenBSD is installed and works fine in TTY but cannot run a GUI. I assume the machine is too new for OpenBSD and I need to bite the bullet and install a Linux system until OpenBSD catches up, but I am checking here in a desperate hope that I can get it working. One of the few suggestions on wiki.x.org is using a different driver.

Lenovo Thinkpad E595
Radeon driver
Ryzen 3 processor

I do not use a DM.
I copied .xinitrc from /etc to /home. An error message in the log says to make sure /etc/sysctl.conf is set to machdep.allowaperture=1. I created an /etc/sysctl.conf file, but do not remember if that must be followed by a sysctl shell command. If so I missed a step.

Running startx results in

Server terminated with error (1).
xinit: giving up
xinit: unable to connect to X server: Connection refused
xinit: server error

I took photographs of the log file so I can add more information if necessary. The parts that got my attention are:

 
check DevMem: failed to open /dev/xf86 and /dev/mem
 (operation not permitted)
linear framebuffer access unavailable
using wscons driver on /dev/tty64

Using system config directory "/usr/X11R6/share/X11/xorg.conf
No Layout section. Using the first screen section.
No section available. Using defaults.

A list of defaults is followed by

 
Matched ati as autoconfigured driver 0.
Matched vesa as autoconfigured driver 1.

Followed by a long list of modules including ati, Radeon and vesa.

 
VESA(0): initiating int10
(EE) VESA(0): Cannot read int vect
(II) UnloadModule: "vesa"
(II) UnloadSubModule "int10"
(II) Unloading int10
(II) UnloadSubModule "vbe"
(II) Unloading vbe
(EE) Screen(s) found, but none have a usable configuration.
Fatal server error:
no screens found

There are no other EE entries, no WW warnings, no Not Implemented entires.
I do not know if any of the entries are only related to video acceleration.

[TronDD]

I forget the details of it, but not every system allows users to run X with startx.

Try enabling xenodm.

[jggimi]

From the fragments posted, it appears your X video is being driven through the vesa(4) driver, which requires use of the xf86(4) aperture driver on architectures such as amd64 and i386. xf86(4) must be enabled via setting machdep.allowaperture in /etc/sysctl.conf. The default is disabled, and that default only works with supported DRM hardware. If I recall correctly, vesa(4) requires machdep.allowaperture=2.

[fvgit]

Originally Posted by sacerdos_daemonis

Lenovo Thinkpad E595
Radeon driver
Ryzen 3 processor

Your CPU should have the built-in Radeon Vega 3 graphics card. Doesn't it show up in the dmesg as amdgpu or sth.? There are already a few sparse posts on the mailing lists mentioning working Ryzen 3, 5 or 7 systems incl. their Vega GPUs, I think. So in theory it should already work although it can be hit and miss according to some posts in this subreddit:

AMDGPU on Ryzen still leads unrecoverable system lockups at unpredictable times. I think anyone considering getting one of those should be aware that this might be an issue. Seems to be less frequent with newer Ryzen 3xxx integrated GPUs.

[Mako_Elite]

Can you post dmesg output ???

Checking specs of your Lenovo Thinkpad E595..it uses

Integrated AMD Radeon™ Vega 3
Integrated AMD Radeon™ Vega 8
Integrated AMD Radeon™ RX Vega 10

lookking in OpenBSD manual "radeon"..I don't see these listed in supported GPU. Perhaps not supported?

Best way is to start with full dmesg output and see what we got and go from there.

[sacerdos_daemonis]

@Mako Elite
I do not know how to record a dmesg from a TTY and transfer it to a computer with a browser.

@jggimi
Yes it is defaulting to vesa. machdep.allowaperture=1 was the instruction given but I shall keep =2 in mind if further testing is possible.

I may have started the thread prematurely. It seems to be a case of unsupported hardware. Yesterday the system worked fine with the terminal. Today it will not boot. After listing details of the cpus the first three "fail to become ready, fail to identify" followed by a reboot.

I am baffled why the system functioned after installation (which required a reboot) then cannot even boot after a shutdown. Looks like I must install another system and wait for OpenBSD's hardware support to catch up. Unless I try something silly like reinstalling and not running fw_update, but that would probably make things worse.

[Mako_Elite]

Originally Posted by sacerdos_daemonis

@Mako Elite
I do not know how to record a dmesg from a TTY and transfer it to a computer with a browser.

@jggimi
Yes it is defaulting to vesa. machdep.allowaperture=1 was the instruction given but I shall keep =2 in mind if further testing is possible.

I may have started the thread prematurely. It seems to be a case of unsupported hardware. Yesterday the system worked fine with the terminal. Today it will not boot. After listing details of the cpus the first three "fail to become ready, fail to identify" followed by a reboot.

I am baffled why the system functioned after installation (which required a reboot) then cannot even boot after a shutdown. Looks like I must install another system and wait for OpenBSD's hardware support to catch up. Unless I try something silly like reinstalling and not running fw_update, but that would probably make things worse.

You can get dmesg to the file like this..type:

dmesg > dmesg.txt

This will create file named dmesg.txt. Yo can call it whatever you want. Then you can transfer the file to USB stick and use in other machine.
It seems now that you can't even boot now. Did you modified any files?
I would try fresh install and not changing anything after install. If you have access to internet directly with LAN cable (not wifi), then make sure it is connected to the laptop during install. This way if there are any firmware available OpenBSD will download them and install them for you automatically.

I have Lenovo Thinkpad T530 with Intel HD 4000 Graphics and it works like a champ.
My desktop is Lenovo ThinkStation P500 with Radeon HD 7950.

Don't give up yet, OpenBSD is really good with Lenovo hardware. Your machine might be too new for now.

I ditched my fairly new iMac , Mac Pro and older MacBook Pro for OpenBSD and I never looked back.

Let me know if you need more help.

Ludovit

[fvgit]

Originally Posted by Mako_Elite

Don't give up yet, OpenBSD is really good with Lenovo hardware.

For what it's worth, nycbug.org's dmesgd list contains 4 recent (2020) entries regarding Ryzen on OpenBSD.

• Lenovo T495, cpu0: AMD Ryzen 7 PRO 3700U w/ Radeon Vega Mobile Gfx, 2296.07 MHz, 17-18-01
• Lenovo Ideapad 530S-14ARR, cpu0: AMD Ryzen 3 2200U with Radeon Vega Mobile Gfx, 2495.74 MHz, 17-11-00
• Motile M142, cpu0: AMD Ryzen 5 3500U with Radeon Vega Mobile Gfx, 2096.35 MHz, 17-18-01

If the system already booted it probably can boot again. Maybe it's a UEFI/BIOS setting that's causing trouble?

[Head_on_a_Stick]

Which version of OpenBSD is this?

I have a ThinkPad E485 with a second generation Ryzen CPU and it's fully supported and perfectly stable under -current.

To show us the dmesg(8) output use this to upload it to a pastebin and then post the returned URL:

curl -F 'f:1=<-' ix.io < /var/run/dmesg.boot

That Vega iGPU should be using radeondrm(4) rather than xf86(4) but I'm pretty sure you'll need -current and the amdgpu-firmware-20200619 images.

[sacerdos_daemonis]

Originally Posted by Mako_Elite

You can get dmesg to the file like this..type:

dmesg > dmesg.txt

Thanks. It is easier than I expected. I shall add that tidbit to my collection.

I have Lenovo Thinkpad T530 with Intel HD 4000 Graphics and it works like a champ.

That is why I bought the machine. My last two laptops were also Lenovo. I had no worries buying another one other than the possibility of unsupported new hardware. Which seems to be the case.

[sacerdos_daemonis]

Originally Posted by fvgit

If the system already booted it probably can boot again. Maybe it's a UEFI/BIOS setting that's causing trouble?

I set the bios to both UEFI and Legacy. The boot problem occurs when trying to start the CPUs, so I assume the boot loader is not an issue. Although my knowledge is severely limited.

[sacerdos_daemonis]

Originally Posted by Head_on_a_Stick

Which version of OpenBSD is this?

6.7-release

That Vega iGPU should be using radeondrm(4) rather than xf86(4) but I'm pretty sure you'll need -current and the amdgpu-firmware-20200619 images.

I shall look into that possibility when I have an opportunity.

[sacerdos_daemonis]

I reinstalled 6.7, copied the snapshot bsd.rd to / and upgraded to -current.
During the reboot "cpu failed to start" but the system still booted. ??? strange
I am attempting to include a dmesg.txt file with this post.

the good
The system is running, including fvwm.

the bad
I am not able to install packages. I have tried Iridium and leafpad, but neither could be installed because many of the components are unavailable in the snapshot repository. I have an idea to install packages from 6.7-release, but do not know if it is safe to mix -release and -current. Or if it is even possible. EXPORT PKG_PATH to mirror /pub/OpenBSD/6.7/packages/amd64? Or would the best course be to reinstall 6.7, install all the packages I want, then upgrade to -current?

the unknown
Because of the "cpu failed to start" message and my recent experience with the 6.7 installation, I do not know if the system will boot again. I shall know that tomorrow morning.

Since a GUI can be used with -current, I am not sure if a new thread should be started for the package installation issue or continue with this one.

[jggimi]

Your dmesg shows that only cpu0 is functional. cpu1-cpu3 "failed to become ready" and "failed to identify".


You cannot install -release/-stable packages on -current. Since you upgraded, you may have some prior packages which interfere with your attempted package installations. I am using the same snapshot (September 6) and can install leafpad from the -current package repository, updated September 10.

[sacerdos_daemonis]

The only package installed is Midnight Commander and I did not upgrade it after moving to -current. I shall hope it was a problem with the mirror and either wait a little while and try again or try another mirror.

[jggimi]

This may seem like an onerous request. If you post your pkg_add console log, someone may be able to give you specific help.

[jggimi]

Again, providing advice in the blind.... It may help to know that when you install midnight commander, pkg_add(1) installs 17 additional dependent packages, including several libraries. You might try first performing a # pkg_add -u to update any of the previously installed packages prior to trying to install a new package.

[sacerdos_daemonis]

I would have simply posted the entire output of the pkg_add message, but the only method I have to record it is pen and paper. Lots of "because of libraries", "minor too small", "major too small" and "cannot resolve" with any package I tried. Since I did not know if the system would work MC is the only package I installed previously.

There is also a bigger problem. When shutting down "System shutdown has arrived" is displayed and the system freezes. It does not shut down and there is no "synching discs" message. The computer must be shut down with the power button. That is obviously not a feasible situation. At this point I think the best course is try to install another system (if the hardware is not too new) and wait for OpenBSD 6.9 and try again.

[jggimi]

Package installation failures

The script(1) tool can be used to log a console or terminal session. The output file can be cleaned of control characters with col(1) -b, and the resulting output stored in a FAT filesystem on a USB stick for transport to the workstation you're using to communicate here. Or, you can take one or more photographs of the console, store them on a website, and link to the images here.

It takes an effort to capture the information needed for someone to be able to help you, as nobody can peer over your shoulder at your system and see the errors you've received. And partial descriptions of the errors are insufficient.

I can think of some possible errors you might have introduced which cause this problem, but there isn't enough information here to point to any of them. I don't even know if you've tried my # pkg_add -u suggestion.

If you'd like to give OpenBSD one more try before abandoning it, I'd suggest a clean installation instead of an upgrade, as your upgraded system appears to be troubled. I have already noted I'm using the same snapshot and (if you're not using PKG_PATH) the same snapshot package repository, and cannot replicate your problem.

System shutdown issues

Again, it's tough to know what is going on with your system. The halt(8) or shutdown(8) command can force a dump for diagnostics, but that is unlikely to be helpful to you, even if you are able to produce a dump in this situation.

[sacerdos_daemonis]

Originally Posted by jggimi

I'd suggest a clean installation instead of an upgrade, as your upgraded system appears to be troubled.

It was a fresh installation. Other than the addition of one package. At this point I do not have the patience to go through another installation and upgrade just to see if the third time will be the charm. Maybe in a few days but right now.

If you'd like to give OpenBSD one more try before abandoning it,

It is not a case of abandoning OpenBSD. It would be a hiatus until hardware support catches up. I started this thread for ideas but was never optimistic. To be honest I am not confident in being able to use another BSD or Linux. This hardware may too new for them as well.