New Raspberry Pi Kernels & Related Packages

ARM
524

It seems it is auto-selected if you select ACPI.

After messing with this, while I like the use of grub and look forward to it, however the UEFI firmware is just not there yet. Too much does not work, or maybe it does but I don’t know how to make it work.

For me:
WiFi runs slowly, at ~1.5Mb/s
fake_hardware clock and ntp fail to set the system clock
no gpio
no sound devices
no video hardware acceleration

I will see about adding u-boot to the mix, but first I will take a look at how the latest tumbleweed boots.

525

Reading the uefi git it does have a lot of stuff that does not work and some stuff will work but needs patching which they said “It is beyound the scope of this git”. I believe from reading the README my take away is it is in it’s infancy stage and really needs more support with commits accepted and implemented to the upstream kernel.

Although it is one thing of many that does not work but did you install crda and set the our country’s regulatory domain for US. I know it has helped others here improving the wifi speed.

526

Agreed that it is too early, I believe the issue is more with the kernel than the UEFI firmware. I think this is due to ACPI. Tumbleweed does not have sound and they don’t configure the gpiomem device. Interestingly, not even as “is not set”. Yet they compile most everything for GPIO and it has /dev/gpiochip devices, which are also missing for me. I am going to investigate this further.

And WiFi runs as fast as my router, so I need to figure out that too.

527

I discovered today that overlay_map.dtb went missing in the linux-rpi4 kernel install again somehow. I have pushed new kernels to the testing and the unstable branches when the mirrors sync. This affects using the vc4-kms-v3d overlay getting the right things loaded.

linux-rpi4 5.10.25-3
linux-rpi4-headers 5.10.25-3
528

I tried taking the Tumbleweed kernel config and adding the config for gpiomem, and after a build that took forever, no device. Oddly, no /dev/gpiochip devices either, but which were created on the Tumbleweed image.

I also tried building it as a module with your kernel config. The module loads without any messages and yet no /dev/gpiomem. I believe that indicates that the device is not found.

It is possible this is a 5.12 issue and not related to either UEFI or ACPI.

I wonder if I need to load one of the gpio overlays?

529

@Darksky @0n0w1c
Here is a link to Jeff Geerling’s artikel and my question about booting rpi4 from SSDs with manjaro rpi4 kernel. Is this ready now?

530

Yes sir! The modules got put in the kernels and you will have to flash the eeprom with the image in the beta directory.

https://github.com/raspberrypi/rpi-eeprom/commit/a1e86f13504dee2022cf16138949b8c3c9eeed76#diff-6b00460f7e3086a1f5f94c819ce44d98a0e06d2b4aabdd5c01d4b520ee3584c9

Added: @0n0w1c

Looks like I need to upgrade the eeprom package. Will do it after a while. I’m in the process of compiling the 5.12 kernel. I have not seen any comments on the net about the wifi issue so I reverted the config to get rid of the efi junk to see if it affects the wifi.

1 Like
531

The latest rpi-eeprom package has been pushed to the unstable branch when the mirrors sync.

rpi-eeprom 20210325-1

@0n0w1c While testing the new package it appears that the NVMe fix is also in the stable folder:

https://github.com/raspberrypi/rpi-eeprom/commit/ff27ccf69403b01e337fc4ee6e7ae75244028cce#diff-f82cf0fa3d9e1f6b15ee62a421fe938454c0a1744a7c1ed94f3c05a136785574

532

No doubt this is due to my current configuration with UEFI test kernel.
Just weird things in different areas with EFI.

$ sudo rpi-eeprom-update -a

VCHI initialization failed
This tool only works with a Raspberry Pi 4

I’ll now run it next on a normal install.

Edit: And of course it ran as expected and successfully updated the eeprom.

533

Wow! If it was me I would bail on uefi and wait for something stable.

534

@Darksky @0n0w1c

Thank you both for everything! :penguin:

Is there a Wiki-Howto around?
May be something like “Howto dd you Manjaro from SD to SSD”

535

Yes, either UEFI is not ready, I expect too much or I have no idea what I am doing… or all of the above.

536

Ah, I found a significant issue, no /proc/device-tree directory. In each case, when I boot either via UEFI shell or grub2, no device tree is passed to the kernel. However, in the case of Tumbleweed, a device tree is passed to the kernel. Hmm.

Edit: As usual, I am my own worst enemy. I now have the device tree passed to the kernel when using grub2. I have sound and a /dev/gpiomem device… although my video colors are messed up. But progress! :slight_smile:

Edit 2: And now I have v3d working which fixes the wonky colors and rpi-eeprom-update works.

$ lsmod

Module                  Size  Used by
ghash_generic          16384  0
ghash_ce               24576  0
gf128mul               16384  2 ghash_generic,ghash_ce
gcm                    24576  0
ccm                    20480  0
algif_aead             20480  0
ctr                    16384  0
algif_skcipher         20480  0
cmac                   16384  0
md5                    16384  0
md4                    16384  0
algif_hash             20480  0
af_alg                 32768  3 algif_hash,algif_skcipher,algif_aead
joydev                 24576  0
brcmfmac              323584  0
brcmutil               24576  1 brcmfmac
sha256_generic         16384  0
raspberrypi_hwmon      16384  0
bcm2835_codec          49152  0
v3d                    81920  5
bcm2835_v4l2           45056  0
bcm2835_isp            32768  0
v4l2_mem2mem           45056  1 bcm2835_codec
sha256_arm64           28672  0
cfg80211              868352  1 brcmfmac
gpu_sched              40960  1 v3d
videobuf2_dma_contig    24576  2 bcm2835_codec,bcm2835_isp
i2c_brcmstb            16384  0
rfkill                 36864  5 cfg80211
i2c_bcm2835            16384  0
bcm2835_mmal_vchiq     32768  3 bcm2835_codec,bcm2835_v4l2,bcm2835_isp
videobuf2_vmalloc      20480  1 bcm2835_v4l2
videobuf2_memops       20480  2 videobuf2_vmalloc,videobuf2_dma_contig
videobuf2_v4l2         32768  4 bcm2835_codec,bcm2835_v4l2,v4l2_mem2mem,bcm2835_isp
videobuf2_common       65536  8 bcm2835_codec,videobuf2_vmalloc,videobuf2_dma_contig,videobuf2_v4l2,bcm2835_v4l2,v4l2_mem2mem,videobuf2_memops,bcm2835_isp
vc4                   274432  18
bcm2835_gpiomem        16384  0
vc_sm_cma              36864  2 bcm2835_mmal_vchiq,bcm2835_isp
videodev              278528  6 bcm2835_codec,videobuf2_v4l2,bcm2835_v4l2,videobuf2_common,v4l2_mem2mem,bcm2835_isp
cec                    53248  1 vc4
mc                     61440  6 videodev,bcm2835_codec,videobuf2_v4l2,videobuf2_common,v4l2_mem2mem,bcm2835_isp
snd_bcm2835            28672  1
drm_kms_helper        270336  3 vc4
snd_soc_core          245760  1 vc4
snd_compress           20480  1 snd_soc_core
snd_pcm_dmaengine      20480  1 snd_soc_core
snd_pcm               126976  5 vc4,snd_bcm2835,snd_compress,snd_soc_core,snd_pcm_dmaengine
snd_timer              36864  1 snd_pcm
snd                    98304  9 snd_bcm2835,snd_timer,snd_compress,snd_soc_core,snd_pcm
rpivid_mem             16384  0
syscopyarea            16384  1 drm_kms_helper
sysfillrect            16384  1 drm_kms_helper
sysimgblt              16384  1 drm_kms_helper
nvmem_rmem             16384  0
fb_sys_fops            16384  1 drm_kms_helper
uio_pdrv_genirq        16384  0
uio                    24576  1 uio_pdrv_genirq
sch_fq_codel           20480  7
fuse                  126976  1
drm                   565248  14 gpu_sched,drm_kms_helper,v3d,vc4
backlight              20480  1 drm
ip_tables              32768  0
x_tables               40960  1 ip_tables
crct10dif_ce           20480  1
sr_mod                 28672  0
cdrom                  45056  1 sr_mod
hid_apple              20480  0
btrfs                1409024  1
blake2b_generic        20480  0
xor                    20480  1 btrfs
xor_neon               16384  1 xor
raid6_pq              110592  1 btrfs
overlay               122880  0
crypto_user            16384  0
efivarfs               20480  1
ipv6                  536576  50

Edit 3: WiFi speed seems to have returned to normal. However, I have a remaining issue with my argonone case fan software… seems I am still missing /dev/gpiochip0 and /dev/gpiochip1. I need to make sure CONFIG_GPIO_SYSFS=m/y.

537

unstable update, youtube cursor circle awhile and no sound, don’t know which cause.

538

If possible, would you pass along the kernel configuration changes made for it to be efi bootable? I am of the mind to continue with UEFI, for my personal projects at least.

539

Well the revert on 5.12 was a waste of time concerning the wifi issue that prevented a shutdown so I might stay with the config to enable uefi for 5.12. It does not seem to bother anything booting regularly. In the logs I see it looks for uefi files and since there are none it proceeds to boot normally.

Both 5.11 and 5.12 upgraded today so I am in the process of compiling them.

1 Like
540

New linux-rpi4-mainline and linux-rpi4-rc packages has been pushed to the unstable branch when the mirrors sync.

linux-rpi4-rc still has the wpa_supplicant issue with not closing out and preventing a reboot or shutdown. The fix is to have NetworkManager to use iwd instead as posted in post #510 above.

https://forum.manjaro.org/t/new-raspberry-pi-kernels-related-packages/4721/510

The linux-rpi4-rc kernel has the uefi modules enabled if one wants to experiment.

linux-rpi4-mainline 5.11.10-1
linux-rpi4-mainlineheaders 5.11.10-1
linux-rpi4-rc 5.12.rc5-1
linux-rpi4-rc-headers 5.12.rc5-1
1 Like
541

I updated the rc kernel on my uefi system and things seem to be in good shape, except no /dev/dri/card0 device, so no gui.

Edit: And of course, it works fine on a normal install.

Edit 2: I installed fbturbo for now. It would seem drm is not finding the vc4. This is a new development for me. Odd, as now all other hardware issues have been resolved. I have just one remaning issue, systemd-timesyncd timeouts.

Edit 3:
$ cat /proc/device-tree/gpu/status

disabled
542

I am still messing with the kernel build. Not sure yet what the issue is. It is first noticed by the 4 raspberries… they are purple, not red (and subsequently llvmpipe is discolored). I think this is an issue related to the framebuffer. I do not know if it has the same cause as the failure to find the vc4.

In dmesg, on a working system I see messages concerning vc4-drm. No such messages with the rc4 or rc5 kernel builds. The only UEFI build that has worked is the tumbleweed kernel build… but it is huge and has other issues.

Edit: Adding this link for my reference.

543

The 5.12 wifi issue supposedly just got fixed today. I am recompiling it to test.

https://github.com/raspberrypi/linux/commit/b97a629902bc062791b8d24d5de28b9556802ea1

1 Like