Closed yagarea closed 6 months ago
none of these are related to sway. though if they don't reproduce on another wayland compositor that would be interesting, but i bet it would be the same
arch linux
do you have the mesa
package? that is the one with the regular opengl drivers (check for /usr/lib/dri/radeonsi_dri.so
). from the chromium logs it looks like it only tries vulkan, which it normally doesn't do
glxinfo
and eglinfo
should list radeonsi
(the driver) in the output somewhere, and glxgears
should work to show you some gears (from mesa-demos
/mesa-utils
)
ah right, you posted eglinfo already. yeah the driver is present there..
Confirming the same issue, tested on Asus PN50 with a Ryzen 4700U and integrated graphics. Computer completely freezes.
Does reverting 7e69a7076fc8a4eb788e0229b1c99dd0b7b04bb7 fix it for you? It does for me.
Edit: Don't know how I missed #7897
ah, i forgot that chromium is still most likely depending on that. that explains all of those cases at least
someone would have to tell the chromium developers about it
For me it isn't really chromium-based applications (Discord on wayland works with HW acceleration) but more like an issue with xwayland - can't launch Cyberpunk, window would flash for a second and then disappear. The error I got is:
vulkan: No DRI3 support detected - required for presentation
Note: you can probably enable DRI3 in your Xorg config
Does reverting 7e69a70 fix it for you? It does for me.
I've observed this on 1.8.1, so that commit wasn't there in the first place. Seeing your answer above, I'm wondering if we're talking about two different issues. In fact, our issue might be different than OP's.
In my case, the whole system freezes on AMD. I've just switched to the AMD system and observed it for the first time in sway. Will put in some more testing to see if other DEs freeze on this machine and kernel, and do a memtest since I could be running into a hardware issue as well.
The crash of electron apps you're observing is present on Intel as well, and I found it's correlated with high dpi screens where you use scaling, in combination of using Ozone. Running those apps on Xwayland doesn't crash them but makes them blurry.
I was having freezes akin to this, I switched back to linux 6.6.10 and they went away
Which kernel did you observe them on? I haven't seen one in a couple of days now, but can't remember if I was on a fresher kernel before. Right now on 6.1.72 from nixos/nixpkgs#a68bc4feaf4bbf4b626226ff8f0f8110588d4ebc
the 6.7 series
Since commit https://github.com/swaywm/sway/commit/7e69a7076fc8a4eb788e0229b1c99dd0b7b04bb7, all Vulkan application running through XWayland crash for me, saying that DRI3 is not available.
Edit: -Dlegacy-wl-drm
is the way to go. I missed that while scrolling down!
Yeah, please try -Dlegacy-wl-drm
if you have issues.
-Dlegacy-wl-drm
How can I try it ?
Assuming that your version has commit https://github.com/swaywm/sway/commit/08a06a7b6bbb324e9fc6e49e96379340404135b4, you just need to add that flag to your sway
call. If you are using some kind of display manager, you might need to read into how to do that there.
this is occurring with the current archlinux release of sway, which is pre-that commit for me, although I think it's a kernel issue since downgrading to 6.6.10 fixes it for me
I discovered that creating swap stops the freezes. Apps still crash but system will not freeze. They just die quicker...
I still call it progress
~@emersion perhaps we should split the "No DRI3 support detected" problem into a separate issue? It doesn't sound like it's the same as @yagarea's and I'd like to still have it tracked so that we don't have to use -Dlegacy-wl-drm
to run Vulkan apps on Xwayland.~
Fixed in xwayland git so won't bother tracking.
Update: I've replaced ram modules and crashes went away on that machine. memtest86 with the old memory wasn't completing, so I'll blame it on ram.
Update: I've replaced ram modules and crashes went away on that machine. memtest86 with the old memory wasn't completing, so I'll blame it on ram.
Well, It turned out that these issues ware caused by faulty RAM even in my case. I can not prove it, but moving to another intel based laptop made these problems disappear.
If someone can prove and decide if this was caused entirely by defective RAM feel free to close this issue.
@emersion I guess we can close it and reopen if someone else comes back with the same issue. Both @yagarea and I are back on Intel for DE machine, so it's going to be hard to drill this down further.
I moved my install from intel i7 + intel iris to AMD ryzen 5 + AMD radeon, by swaping SSDs. I experience these issues:
Symptoms
AMDGPU securedisplay: generic failrure
in log and this is in journalctl:journalctl
:Chromium browser straight up crashes immidietly on lauch:
More information
Specs:
This is list of installed packages containing
amd
:This is list of packages containing
radeon
:This is output of
lspci -nn
:Output of
lspci -k | grep -A 3 -E "(VGA|3D)"
:I have found this in
dmesg
log:...
this is output of
eglinfo -B
:These are my kernel parameters:
What I tried:
xf86-video-amdgpu
fwupdmgr
xf86-video-amdgpu
)amdvlk
drivers and installed radeon driversCONFIG_DRM_AMDGPU_CIK=Y
to kernel parameters. This results inUnknown kernel command line parameters "CONFIG_DRM_AMDGPU_CIK=Y", will be passed to user space.
in dmesg.radeon.dpm=0
andradeon.dpm=1
with no effect~/.cache/mesa
directoryNothing seems to resolve this issue.
Question
What should I do to get rid of this freezes and crashes ? Do you have any tips how to diagnose these kind of issues ?
Thank you for help