I’ve seen v6.5.8 is out on the rpi github repo : https://github.com/raspberrypi/linux/commit/8b2a6b131d5b3c0fdac5e1582ca0993d06a6d6a4
Yes it is along with the new -rc. I have both compiled and tested. I am waiting to see what they do with upgrading 6.1. They seemed to have stopped upgrading their branches for some reason. I am afraid they dropped everything trying to chase down @tartanpion memory leak and I will have to re-compile all kernels over again.
1 Like
Well they never upgraded the 6.1.y kernel tree today but all of the other kernels are current now with upstream. The 6.1.y tree did have several bug fixes. I will be glad when the bug fixes start slowing down from their kernel overhaul. 
No memory leak fix yet.
RPi’s latest trees pushed to the unstable branch when the mirrors sync.
linux-rpi4 6.1.58-4
linux-rpi4-headers 6.1.58-4
linux-rpi4-mainline 6.5.8-1
linux-rpi4-mainline-headers 6.5.8-1
linux-rpi4-rc 6.6.rc7-1
linux-rpi4-rc-headers 6.6.rc7-1
linux-rpi5 6.1.58-4
linux-rpi5 headers-6.1.58-4
rpi-overlays 6.1.58-4
what is raspberrypi-utils package, raspi-config?
raspberrypi-utils is RPi’s new git that replaced their depreciated raspberrypi-userland git (our old raspberrypi-userland-git package).
Explanation:
https://github.com/raspberrypi/userland/blob/master/README.md
The latest round of pi kernel fixups has been pushed to the unstable branch when the mirrors sync.
@tartanpion Check and see if this kernel fixes your memory leak issue.
linux-rpi4 6.1.58-5
linux-rpi4-headers 6.1.58-5
linux-rpi5 6.1.58-5
linux-rpi5-headers 6.1.58-5
rpi-overlays 6.1.58-5
The possible fix is only on a PR https://github.com/raspberrypi/linux/pull/5684. So unless you make a patch like in pios.
If anyone wants to test, then
sudo rpi-update pulls/5684
I was going to patch the latest kernel but saw most of the PR was there except the debug code.
I am not an expert but I see nothing in the commit but everything concerning 5684 is in draft /open.
Since I’m on 4 hard reboots since this morning because I use the raspberry I was monitoring this PR…
Looks like you are right. I guess I clicked on the wrong link. I will redo after breakfast.
Added:
@tartanpion Here is the test kernel. I built it from their work branch using their last commit adding the debug. It lacks a couple of commits as with their main tree and it has not been tested for the pi5 yet so I do not want to push it to our repo.
https://github.com/raspberrypi/linux/commit/36ed2f9d67a40900ce6b7b719fc8b6d6f662360c
If you would run it through your tests and see if it is ok. If not capture the debug info so we can provide what they want here.
https://github.com/raspberrypi/linux/issues/5674
Download the kernel packages and unpack artifacts.zip and install. Click the Download button here.
https://gitlab.manjaro.org/manjaro-arm/packages/core/linux-rpi4/-/jobs/13169
I don’t have time to test more. But already i continue to have that when playing hevc hardware accelerated :
oct. 29 19:07:54 bloodmoon-pc kernel: rpivid feb10000.codec: rpivid_h265_start: (3840x2160)
oct. 29 19:07:54 bloodmoon-pc kernel: rpivid feb10000.codec: SPS changed
oct. 29 19:07:54 bloodmoon-pc kernel: rpivid feb10000.codec: PPS changed
oct. 29 19:07:54 bloodmoon-pc kernel: rpivid feb10000.codec: Missing DPB ent 0, timestamp=0
oct. 29 19:07:54 bloodmoon-pc kernel: rpivid feb10000.codec: Missing DPB ent 1, timestamp=0
oct. 29 19:07:54 bloodmoon-pc kernel: rpivid feb10000.codec: Missing DPB ent 2, timestamp=0
oct. 29 19:07:54 bloodmoon-pc kernel: rpivid feb10000.codec: Missing DPB ent 3, timestamp=0
oct. 29 19:07:55 bloodmoon-pc kernel: vc4-drm gpu: swiotlb buffer is full (sz: 6352896 bytes), total 32768 (slots), used 199 (slots)
oct. 29 19:07:55 bloodmoon-pc kernel: vc4-drm gpu: swiotlb buffer is full (sz: 5558272 bytes), total 32768 (slots), used 587 (slots)
oct. 29 19:07:56 bloodmoon-pc kernel: vc4-drm gpu: swiotlb buffer is full (sz: 6496256 bytes), total 32768 (slots), used 129 (slots)
oct. 29 19:07:59 bloodmoon-pc kernel: rpivid feb10000.codec: PPS changed
oct. 29 19:08:02 bloodmoon-pc kernel: rpivid feb10000.codec: PPS changed
oct. 29 19:08:02 bloodmoon-pc kernel: rpivid feb10000.codec: Missing DPB ent 0, timestamp=0
sudo cat /sys/kernel/debug/dri/1/hvs_dlist_allocs
Allocated nodes:
node [00000020 + 0000000b]
node [0000002b + 00000009]
node [0000003d + 00000009]
Stale nodes:
node [0000002b + 00000009] channel 0 frcnt 35
How are you playing that video so when they start asking questions. I do not see the same output here playing HEVC with mpv. journalctl does not report anything when playing.
[ray@jellyfin AKA.2023]$ mpv nasa.ts
(+) Video --vid=1 (hevc 3840x2160 59.940fps)
(+) Audio --aid=1 (mp2 2ch 48000Hz)
Audio --aid=2 (aac 2ch 48000Hz)
Using hardware decoding (drm).
(+) Video --vid=1 (hevc 3840x2160 59.940fps)
(+) Audio --aid=1 (mp2 2ch 48000Hz)
Audio --aid=2 (aac 2ch 48000Hz)
Subs --sid=1 (*) (eia_608)
AO: [pipewire] 48000Hz stereo 2ch s16p
VO: [gpu] 3840x2160 drm_prime[rpi4_8]
AV: 00:04:23 / 00:04:24 (100%) A-V: 0.000 Dropped: 8285
Exiting... (End of file)
With mpv rpi-ffmpeg, nothing special.
I have news errors
cma: cma_alloc: linux,cma: alloc failed, req-size: 4050 pages, ret: -12
v3d fec00000.v3d: MMU error from client CLE (4) at 0x1e8a1000, pte invalid
Perhaps you should try a sdcard with gnome wayland and see if you encounter the same errors and the same freeze because it’s not easy to switch in a TTY after to list the debug.
Edit : I forgot to put cma-512 back
I will try playing on a TTY and see what it does. I am not reaally understanding the rpivid errors. I thought that had to do with camera. The cma error is weird also as you have 512 you posted above in your config.txt.
I mean i do not play in a tty. But when the freeze occurs, even the keyboard does not work anymore to switch in a tty to read the debug.
And
vc4-drm gpu: swiotlb buffer is full
is here with or without cma-512.
I have try to remove that because they no more mention that in their new web page.
linux-rpi4-mainline kernel packages has been promoted to the 6.6.y tree. The linux-rpi4-rc packages are now obsolete in the unstable branch until the 6.7.y tree is added to the RPi github. New linux-rpi4-mainline kernel packages pushed to the unstable branch when the mirrors sync.
This will most likely be the next LTS kernel and if so RPi will move to it from the 6.1.y tree…
https://www.techradar.com/pro/theres-a-new-version-of-linux-but-dont-go-getting-too-excited
linux-rpi4-mainline 6.6.0-1
linux-rpi4-mainline-headers 6.6.0-1
6.6-mainline updated, thanks.
edit,
since kernel new 2712 build, Pi4/sway/wlroots with mesa-22.1.8 & 23.2.1 can not start, egl/dri error .
so, build mesa-23.3-dev with Pi5-VII supported, solved.
does new kernel matter?
Maybe mesa fixed a bug and added pi5 support.
but with mesa-23.3-git lost vulkan device.
Using video driver wayland
Vulkan device: V3D 4.2.14
Vulkan device type: integrated gpu
Vulkan version: 1.2.267 (api) 23.99.99 (driver)
Max. texture size: 4096
Max. uniform buffer range: 1073741824
Min. uniform buffer offset alignment: 32
Resolution: 1280 x 720
v3dv_drm_handle_device: drmGetMagic failedKernel 6.1.58-2 just doesn’t boot on my Raspberry Pi 4.
Complains about some bluetooth HW issue, then spits errors that a CPU is stuck. Plus SD access LED stays steady lit, then nothing.
Too early in the boot process to take screenshot or copy / paste, but well, it doesn’t boot.
Reverting to 6.1.54-1 brings my pi 4 back to life.