Sudden loss of video using Chrome 100 in Mint

While using Chrome 100 in Mint 19.3 (kernel 5.4.0.107), the screen suddenly went black, the monitor reported no signal, then the DVI message appeared and the last operating window of Chrome reappeared. BUT the keyboard did nothing, clicking the mouse did nothing, and I had to hit the power button to invoke a hard reboot.

This has now happened 3 times. It only happened when using Chrome, in one case ONLY chrome was open, and only using the newest kernel version of Mint 19.3. It did not happen (yet) using new Chrome 100 in Ubuntu Mate, same kernel (but it might).

There are so many variables, I do not know where to go to diagnose, let alone repair. It is easy for me to use Chromium and Firefox instead, and it is easy to boot to the older kernel, but that avoids knowing what happened.

What additional information would help y’all figure this out?

I’m just poking sticks into the dark, hoping I accidentally hit something.

I would guess, the reason for that happening is that Chrome is trying to use some type of codec or whatever special technology, which requires low level driver usage in the OS, most likely related to the hardware.

If such a low level driver would crash due to whatever reason, then this could easily explain your issue.

Please, try to reproduce the issue and precisely log what you did moments before it happened. Then we can find the trigger causing the symptoms you are experiencing.

Here is one out of three.
I downloaded some podcasts to a folder on the hard drive.
I opened them in VLC to edit the MP3 tags.
While VLC was open, I opened Chrome to search for a title of one of the podcasts.
Before I could hit enter in the search window, the screen went dark.
Only those two programs were open.
I can try the exact same procedure and see if it reproduces the event.

When it happens, could you have a look at your log files to see if there are any relevant messages.

Neville

I need a bit of guidance here.
I have systemd and the normal Mint.
Which logs should I access if I am doing a hard reboot (i.e., holding down the power button for 5 seconds) after the fact?
Should I open them without re-opening Chrome?

You can stream them into files and then trigger the issue. Once reboot finishes, you can look into the files for hints.

@cliffsloane
It does not matter when you read logfiles. They cumulate, newest stuff is at the bottom. They will not erase if you boot.

Just use an editor to view them, go to the bottom
Neville

I managed to invoke a crash, and as @Akito suggested, it is definitely about the graphics card (Radeon). The whole section of the log that I copied has 543 lines, so I present it as a Google drive link.
https://drive.google.com/file/d/1dx3ve9zPEIvTudH6LzhtK_94s9-FwOxr/view?usp=sharing

Great, we are progressing! :smiley:

And now, the plot thickens!
In Ubuntu Mate, I enter the following:

modinfo -V radeon

and get this:

kmod version 27
+XZ -ZLIB +LIBCRYPTO -EXPERIMENTAL

But in the troubled Mint, it gives this:

kmod version 24
-XZ -ZLIB -EXPERIMENTAL

Could this have something to do with it? It was version 24 all along, and I only just noticed the difference between Ubuntu and Mint.

Sounds reasonable. Though, I have no idea if that’s the case. We need to research a bit more.

Couldn’t find any relevant information on this, yet.

@cliffsloane
My Debian 11 says

# modinfo -V radeon
kmod version 28
-ZSTD +XZ -ZLIB +LIBCRYPTO -EXPERIMENTAL

Different again
Neville

Sorry - glossed over most of the posts here - one thing I’ve had to do in the past occasionally, in Chrome (google chrome - and - Chromium) is chrome://settings and find advanced, and uncheck “hardware acceleration” - have you tried that?

They seem to have obfuscated it now - I had to search for it :

Would it be worth my effort to uninstall Chrome 100 and install (in this case, from .deb) an older version? I found this site:
http://170.210.201.179/linux/chrome/deb/pool/main/g/google-chrome-stable/
The only reason I hesitate is that I have three other browsers working just fine in Mint (FF, Chromium 99 and Edge).

I would just go with one browser that works, as long as it does all you want.
Browsers are huge… they make for big update downloads. You dont want too many of them. Firefox is something like 100Mb every time you download an update.

Neville

well, it happened again! This time, it was Chromium that crashed (100.0.4896.60), and in Ubuntu Mate.
I uploaded 3,500 lines of syslog here:
https://drive.google.com/file/d/1srPu1V33q-Ef2_P_A-98WWTNu8J9eVdJ/view?usp=sharing
I started with a fully initialized kernel at 19:42, but the good stuff starts far later, at 19:47. This is where all the failures with radeon get noted.

Also, there was an update for Chrome in Mint. I installed it (100.0.4896.75) and now it crashes as soon as I open Chrome.
Looks like I have to use other browsers and/or install an older version of Chrome.

Here is a manageable number of lines from journalctl:

Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_008010_GRBM_STATUS = 0xA0003030
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_008014_GRBM_STATUS2 = 0x00000003
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_000E50_SRBM_STATUS = 0x200290C0
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_008674_CP_STALLED_STAT1 = 0x00000000
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_008678_CP_STALLED_STAT2 = 0x00000000
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_00867C_CP_BUSY_STAT = 0x00000000
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_008680_CP_STAT = 0x80100000
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: R_00D034_DMA_STATUS_REG = 0x44C83D57
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: GPU reset succeeded, trying to resume
Apr 04 19:46:51 cliff-desktop kernel: [drm] PCIE gen 2 link speeds already enabled
Apr 04 19:46:51 cliff-desktop kernel: [drm] PCIE GART of 512M enabled (table at 0x0000000000142000).
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: WB enabled
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: fence driver on ring 0 use gpu addr 0x0000000010000c00 and cpu addr 0x00000000c5595b5d
Apr 04 19:46:51 cliff-desktop kernel: radeon 0000:01:00.0: fence driver on ring 5 use gpu addr 0x00000000000521d0 and cpu addr 0x0000000085d25a25
Apr 04 19:46:51 cliff-desktop kernel: [drm] ring test on 0 succeeded in 1 usecs
Apr 04 19:46:51 cliff-desktop kernel: [drm] ring test on 5 succeeded in 1 usecs
Apr 04 19:46:51 cliff-desktop kernel: [drm] UVD initialized successfully.
Apr 04 19:47:02 cliff-desktop kernel: radeon 0000:01:00.0: ring 0 stalled for more than 10184msec
Apr 04 19:47:02 cliff-desktop kernel: radeon 0000:01:00.0: GPU lockup (current fence id 0x0000000000003004 last fence id 0x000000000000302c on ring 0)
Apr 04 19:47:02 cliff-desktop kernel: [drm:r600_ib_test [radeon]] ERROR radeon: fence wait failed (-35).
Apr 04 19:47:02 cliff-desktop kernel: [drm:radeon_ib_ring_tests [radeon]] ERROR radeon: failed testing IB on GFX ring (-35).
Apr 04 19:48:13 cliff-desktop systemd-shutdown[1]: Syncing filesystems and block devices.
Apr 04 19:48:13 cliff-desktop systemd-shutdown[1]: Sending SIGTERM to remaining processes…

So, it happens in Ubuntu and Mint
it happens in more than ine browser
the log messages are about Radeon

If it was Debian , I would be looking at
/etc/apt/sources .list
to see if non-free is present
this allows it to install proprietary packages

Then I would be searching thru the package repository looking for relevant packages that have not been installed. In cases like this I have always managed to find a package that fixed the issue. Look for stuff that mentions drivers or radeon .

I just dont have Ubuntu or Mint experience and I dont know your hardware, so I cant be specific

Neville

It happens in Chrome version 100+ and Chromium version 100+. Not version 99 or less.
It never happened before this week, and never happens in Firefox, Edge, Brave.
It also does not happen in Windows 10, any browser.

This Radeon thing is not a separate video card, but part of the mainboard.

BTW, when I say “never”, it is more accurate to say “not so far.”

Never say never with computers

It happens in Chrome version 100+ and Chromium version 100+. Not version 99 or less.
It never happened before this week, and never happens in Firefox, Edge, Brave.
It also does not happen in Windows 10, any browser.

OK that changes things.
It seems like a browser version issue, doesnt it.
You are not going to fix that yourself … need a workaround, like use an earlier version or use Firefox which works.
Or there might be some browser setting you can fiddle to suppress it.

These intermittant things are not easy. The log doesnt tell us much other than that the graphics board/chip is unhappy.

Neville