Sounds like I may have been misunderstood. Sorry about that. Using rpmfusion now, nothing against them at all. Pointed out the differences between our systems and the sources of packages related between those systems.
I clearly replied to:
Then it has to be a difference in installed packages. My system isn’t too old.
I use Fedora 37 every single day, including rpmfusion. My goal was to identify the cause, which appeared to be related. Package sources matter as it helps to identify differences and what I have working and what they once had an issue with. Not because a package source is bad or broken, rather, because it’s not default and we can look into it as a contributing factor with a package that may not be cooperating for a specific user’s config.
Hopefully that’s clearer on the subject.
Tltr, I lack the aforementioned package the customer has. Hence the difference. Could be relevant to their past experience. Since rpmfusion isn’t default, adding from them is like adding from AUR, etc to a lesser degree. Not bad at all, just not vanilla and worth looking into.
Yep, never, ever suggested otherwise. Rather, was indicating source of a specific package. A package not on my own system with the same repos enabled. Litterly trying to answer their question.
There was a bit of back and forth, so I can see where you may have believed I indicated that there was something wrong with rpmfusion. Not at all. Not even saying the package sources are bad. I suggested the source of the issue may be related to that particular package at that particular time.
Good rule of thumb. I won’t ever blame a repo as bad. I may blame a package release as being a contributor to an issue at times. Seems reasonable to me as part of a troubleshooting process. Hopefully that clears up any confusion on intent.
@Matt_Hartley no problem, just wanted to get clarification. Intel-media-driver is directly referenced in the doucmentation for getting hardware acceleration on Fedora for Firefox working. It is pretty clear they are working on the issue and have made tons of improvements. The more corner cases we find the better it should get.
Preliminary results: I had no hard freeze yet. But I still got rendering issues, like flickering top few display lines directly after wakeup from s2idle. But I’d say these occur less often than with intel-media-driver installed and are less impacting. i915.enable_psr=1 and intel-media-driver installed gave me the most stable results so far.
I will continue running the current setup (no intel-media-driver and no kernel arguments) to watch out for any freezes.
I don’t have this issue any more and I am running my Framework with intel-media-driver installed and also PSR version 2 (which is the default setting on fedora) now. What seems to cause some trouble is gnome night light, but I have not examined that and just deactivated night light for now.
Currently running on Kernel 6.3.8
Hm, out of sudden, I see this short freezing and flickering again with the LTS kernel on Arch Linux (linux-lts-6.1.51). This is a bit mysterious at the moment. Downgrading to the previous LTS kernel doesn’t help, and there seem to be no other recent suspicious package upgrades. What seems to help is the good old i915.enable_psr=0.
I tried switching to the “normal” kernel (linux-6.4.12), and this seems to help for now…
Does anyone else see this on recent (LTS) kernels?
I have a slight suspicion that this is related to this (reverting) commit, which first appeared in 6.1.45 [PATCH 6.1 127/127] Revert "drm/i915: Disable DC states for all commits" - Greg Kroah-Hartman . The reverted commit is related to PSR, and it’s not present in the mainline kernel, so this would also explain why I only see this in the “longterm” kernel. (But I’m not going to investigate further as long as 6.4.12 works for me.)
Hi Matt and everyone, I’ve already gone through the fix here so I’m not in need of any immediate assistance, but I’m posting the below just in case it’s helpful for future troubleshooting.
For me, the flickering/freezing started happening after I went through the fingerprint reader fix (here). My info:
13th Gen Intel Core i5-1340P × 16
Intel Raptor Lake-P [Iris Xe Graphics] (rev 04)
Fedora Linux 38 (Workstation Edition) (64-bit)
Windowing System: Wayland
GNOME version: 44.5
Kernel version: Linux 6.2.9-300.fc38.x86_64 (Kernel driver in use: i915)
A freeze happened earlier today while browsing on Firefox (Flathub, so I can play YouTube). Here are the logs:
journalctl --since "2023-10-18 17:53:00" --until "2023-10-18 18:01:00"
Oct 18 17:56:17 fedora PackageKit[52285]: daemon quit
Oct 18 17:56:17 fedora systemd[1]: packagekit.service: Deactivated successfully.
Oct 18 17:56:17 fedora systemd[1]: packagekit.service: Consumed 1.948s CPU time.
Oct 18 17:56:17 fedora audit[1]: SERVICE_STOP pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=packagekit comm="systemd" exe="/usr/lib/systemd/syst>
Oct 18 17:57:00 fedora org.mozilla.firefox.desktop[52023]: [Child 3261, MediaDecoderStateMachine #2] WARNING: Decoder=7fa5be36c900 state=DECODING_METADATA Decode metadata failed, shutt>
Oct 18 17:57:00 fedora org.mozilla.firefox.desktop[52023]: [Child 3261, MediaDecoderStateMachine #2] WARNING: Decoder=7fa5be36c900 Decode error: NS_ERROR_DOM_MEDIA_METADATA_ERR (0x806e>
Oct 18 17:57:09 fedora org.mozilla.firefox.desktop[52023]: [Child 3261, MediaDecoderStateMachine #2] WARNING: Decoder=7fa5c21ac400 state=DECODING_METADATA Decode metadata failed, shutt>
Oct 18 17:57:09 fedora org.mozilla.firefox.desktop[52023]: [Child 3261, MediaDecoderStateMachine #2] WARNING: Decoder=7fa5c21ac400 Decode error: NS_ERROR_DOM_MEDIA_METADATA_ERR (0x806e>
Oct 18 17:57:21 fedora org.mozilla.firefox.desktop[52023]: [Child 3261, MediaDecoderStateMachine #2] WARNING: Decoder=7fa5c283b800 state=DECODING_METADATA Decode metadata failed, shutt>
Oct 18 17:57:21 fedora org.mozilla.firefox.desktop[52023]: [Child 3261, MediaDecoderStateMachine #2] WARNING: Decoder=7fa5c283b800 Decode error: NS_ERROR_DOM_MEDIA_METADATA_ERR (0x806e>
Oct 18 17:58:08 fedora org.mozilla.firefox.desktop[50473]: libva info: VA-API version 1.18.0
Oct 18 17:58:08 fedora org.mozilla.firefox.desktop[50473]: libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/iHD_drv_video.so
Oct 18 17:58:08 fedora org.mozilla.firefox.desktop[50473]: libva info: Trying to open /usr/lib/x86_64-linux-gnu/dri/intel-vaapi-driver/iHD_drv_video.so
Oct 18 17:58:08 fedora org.mozilla.firefox.desktop[50473]: libva info: Found init function __vaDriverInit_1_18
Oct 18 17:58:08 fedora org.mozilla.firefox.desktop[50473]: libva info: va_openDriver() returns 0
Oct 18 17:58:08 fedora org.mozilla.firefox.desktop[50473]: ATTENTION: default value of option mesa_glthread overridden by environment.
Oct 18 18:00:22 fedora kernel: Lockdown: systemd-logind: hibernation is restricted; see man kernel_lockdown.7
Oct 18 18:00:24 fedora systemd-logind[928]: Lid closed.
A small suggestion: a previous post indicated that a fix for this Panel Self Refresh issue was featured in the Fedora installation guide, but I noticed this was not the case for the current Fedora 38 Installation on the Framework Laptop 13 guide (or here, though I’m not sure if this issue is too specific). Perhaps this is a good addition to Step 6?
I think all of your issues stem from using this kernel. In my instance I have not needed PSR=0 since January of this year. The current kernel is 6.5.6-200.fc38.x86_64.
Why? The regular Firefox rpm can play youtube with hardware acceleration without issue as well.
It is probably not included since it has been resolved for quite some time. Please make sure you are fully updated and running the latest kernel. 99% of the issues you ever run into on a newly released processor are eventually resolved by updating through them. As fixes and improvements hit the kernel your performance and stability improve.
I am also curious if you have the following package installed: xorg-x11-drv-intel. If you do I highly encourage you to nuke it, space it, drop it in the ocean, bury it, or simply dnf remove. It is old, out of date, and actually contributes to the whole issue. Also I recommend installing the latest intel-media-driver from rpmfusion. This driver is where a number of improvements for the graphics reside, and the rpmfusion one tends to be very up to date.
After doing all of the above steps, remove that psr=0 kernel parameter. Admittedly I am on a 12th gen Intel, but the recommendations I made should apply. The biggest one is the kernel. I suspect either your system is not fully updated, or at some point grub was changed to select that kernel explicitly and is why that outdated kernel is even still on your system, or it has been somehow versionlocked. If you can run rpm -qa or dnf list installed kernel, it should show all installed kernels.
Thank you very much for your insights, Matt and @nadb!
Short version: I’ve rectified the kernel issue, burned xorg-x11-drv-intel with fire, and removed the i915.enable_psr=0 parameter as advised. (I haven’t gone through the rpmfusion step yet, because I’d like to test this for a few days to see if it’s sufficient.)
Long version: I’m not entirely sure how my kernel ended up this way, because I used a freshly downloaded ISO file for my install last Saturday, but I had incorrectly understood that it was the latest version because GNOME Software indicated that my system was up to date. (So I assume this only indicates installation, and not active usage.)
In any case, I remembered seeing a list of something that looked like a list of kernel versions while rebooting the other day, so I rebooted, pressed F8, and saw that I could simply select the correct one (later made persistent):
Fedora Linux (6.5.7-200.fc38.x86_64) 38 (Workstation Edition)
Fedora Linux (6.5.6-200.fc38.x86_64) 38 (Workstation Edition)
*Fedora Linux (6.2.9-300.fc38.x86_64) 38 (Workstation Edition)
Fedora Linux (0-rescue-7db6c6aba4d84f0e939088d60f407bf8) 38 (Workstation Edition)
UEFI Firmware Settings
Also, to answer this question for later reference:
Apparently, the kernel being old and such had made it so that the regular Firefox could not play anything on YouTube. Therefore, I had dug up this solution. However, now that the root causes of the issue are resolved, I’m happy to report that the regular Firefox can indeed play YouTube videos without a problem.
@Agnes happy to help. Fedora does not usually update the download isos through the lifespan of a Fedora version. So that image is a near release date iso. Also packagekit and the Software Application are slow to the draw sometimes. Updating in a console or terminal is often more reliable or rather in sync with the current state, particularly if you use dnf update --refresh.
