Raspberry Pi 4 bcm2711 / Raspberry Pi 3 bcm2837 (aarch64)

business_kid · 11-18-2022, 11:11 AM

I can't use u-boot - period. U-Boot reads uEnv.txt, the Pi itself reads config.txt. I'm booting with these modifications to /boot/config.txt

Code:

#kernel=u-boot.bin
...
disable_overscan=0
overscan_top=32
overscan_bottom=32
overscan_left=32
overscan_right=32
...
kernel=vmlinuz-6.0.9
initrd=uInitrd-6.0.9 followkernel

Without these mods the screen overscans, I can't access the menu bar and am basically flying blind as all sorts of stuff is at the screen edge. We've discussed this before, and fingered my tv as the culprit.

business_kid · 11-20-2022, 07:09 AM

Right, I actually got back and tested that all from the Pi. Just as feedback, he overscan parameters don't seem to be working, and the firefox on your current MMC image is rearing up in alarm if I try to download ungoogled chromium in Firefox. I'm not unduly worried by either error as it boots and I have a terminal, which is all I want really.

The image filesystem landed in poor shape, so I'll check my download before I do much. It's the 6.2.x version I'll want to see behaving itself.

BTW, I'm intrigued by what those rpi-eeprom-update & rpi-userland scripts do, but they do plenty. Is there a document somewhere? They have been changes since 15.0

sndwvs · 11-20-2022, 07:33 AM

here are the sources rpi-eeprom, rpi-userland

business_kid · 11-20-2022, 11:59 AM

Right, thanks for those. They appear to be the Raspbian utilities which accounts for the bloat and lack of comments.

I have tried to get the screen correctly sized without success. The first page of text output on boot seems to obey the "overscan" directives but page 2 of the boot is oversized and off screen. Here's some things from dmesg that seem odd:

Code:

[    0.000000] Kernel command line: coherent_pool=1M 8250.nr_uarts=1 snd_bcm2835.enable_headphones=0 snd_bcm2835.enable_hdmi=0 snd_bcm2835.enable_hdmi=0  smsc95xx.macaddr=DC:A6:32:73:7D:DF vc_mem.mem_base=0x3ec00000 vc_mem.mem_size=0x40000000  root=/dev/mmcblk0p2 ro rootwait nofont console=tty1 selinux=0 plymouth.enable=0 smsc95xx.turbo_mode=N dwc_otg.lpm_enable=0 elevator=noop snd-bcm2835.enable_compat_alsa=0
[    0.000000] Kernel parameter elevator= does not have any effect anymore.
               Please use sysfs to set IO scheduler for individual devices.
[    0.000000] Unknown kernel command line parameters "nofont selinux=0", will be passed to user space.

[   52.332555] hdmi-audio-codec hdmi-audio-codec.0.auto: Not able to map channels to speakers (-22)
[   52.332580] hdmi-audio-codec hdmi-audio-codec.0.auto: ASoC: error at snd_soc_pcm_dai_prepare on i2s-hifi: -22
[   52.332587]  MAI: ASoC: __soc_pcm_prepare() failed (-22)
<<Repeated 20+ times>

[   52.587848]  MAI: __soc_pcm_open() failed (-19)
[   52.588993] hdmi-audio-codec hdmi-audio-codec.1.auto: ASoC: error at snd_soc_dai_startup on i2s-hifi: -19
[   52.589112]  MAI: __soc_pcm_open() failed (-19)
[   52.590313] hdmi-audio-codec hdmi-audio-codec.1.auto: ASoC: error at snd_soc_dai_startup on i2s-hifi: -19
[   52.590380]  MAI: __soc_pcm_open() failed (-19)
[   52.608150] hdmi-audio-codec hdmi-audio-codec.1.auto: ASoC: error at snd_soc_dai_startup on i2s-hifi: -19
[   52.608249]  MAI: __soc_pcm_open() failed (-19)
[   52.608992] hdmi-audio-codec hdmi-audio-codec.1.auto: ASoC: error at snd_soc_dai_startup on i2s-hifi: -19
[   52.609079]  MAI: __soc_pcm_open() failed (-19)
[   52.611349] hdmi-audio-codec hdmi-audio-codec.1.auto: ASoC: error at snd_soc_dai_startup on i2s-hifi: -19
[   52.611534]  MAI: __soc_pcm_open() failed (-19)
[   52.922620] Bluetooth: RFCOMM TTY layer initialized
[   52.922678] Bluetooth: RFCOMM socket layer initialized
[   52.922730] Bluetooth: RFCOMM ver 1.11

Despite all those errors, sound works

There's also nothing about video drivers, or the GPU in dmesg. No module loaded? Every mention of video or hdmi is in connection with the bcm 2835, which is the audio chip unless I'm mistaken. The only thing I got was this

Code:

bash-5.2# dmesg |grep framebuffer
[    0.632152] simple-framebuffer 3e4e0000.framebuffer: framebuffer at 0x3e4e0000, 0x6e7000 bytes
[    0.632167] simple-framebuffer 3e4e0000.framebuffer: format=a8r8g8b8, mode=1824x984x32, linelength=7296
[    0.638342] simple-framebuffer 3e4e0000.framebuffer: fb0: simplefb registered!

and that's apparently incapable of following "overscan" orders.

Looking at /var/log/Xorg.0.log, it has modelines, and maximum dotclock. As I go back to the days of Analogue monitors, I can cobble myself up something in /etc/X11/xorg.conf.d/50-video.conf to sort it, but I won't bother unless the new 6.2.x drivers need to be bent straight. It does appear that the endless list of parameters in /boot/cmdline.txt need a once-over as some are deprecated. I'm currently using that.

sndwvs · 11-20-2022, 12:20 PM

Quote:

Originally Posted by business_kid

Despite all those errors, sound works

There's also nothing about video drivers, or the GPU in dmesg. No module loaded? Every mention of video or hdmi is in connection with the bcm 2835, which is the audio chip unless I'm mistaken.

more information bcm2835

business_kid · 11-20-2022, 01:22 PM

Ok so the BCM2835 has the GPU in some corner of it. I somehow thought it was in the BCM2711.

Interestingly, I've a backup of 15.0 here, with kernel 5.16.7, and dmesg includes this (note the bold section)

Code:

Kernel command line: coherent_pool=1M 8250.nr_uarts=1 snd_bcm2835.enable_compat_alsa=0 snd_bcm2835.enable_hdmi=1 video=HDMI-A-1:1920x1080M@60,margin_left=48,margin_right=48,margin_top=48,margin_bottom=48 smsc95xx.macaddr=DC:A6:32:73:7D:DF vc_mem.mem_base=0x3eb00000 vc_mem.mem_size=0x3ff00000  root=/dev/sda2 ro rootwait nofont console=tty1 selinux=0 plymouth.enable=0 smsc95xx.turbo_mode=N dwc_otg.lpm_enable=0 elevator=noop snd-bcm2835.enable_compat_alsa=0

At a semi-educated guess, that 15.0 command line looks like the o/p of config.txt being passed as kernel parameters. The current kernel command line (2 posts down) isn't half that elaborate, as if config.txt is being ignored.

Again, this isn't a priority. I'll be making noise if I can't stop the 6.2.x kernels from overscanning.

business_kid · 11-21-2022, 11:19 AM

So the penny finally dropped here. I moved cmdline.txt to cmdline.orig, and put the kernel command line as logged in the dmesg of 15.0 in as cmdline.txt on your current image with the 6.0.9 kernel, and it booted with no extra errors and solved my overscan problem.

I know you recommend u-boot. But that pays no heed to the parameters in config.txt. I have been using the standard RazPi boot sequence of config.txt & cmdline.txt. But config.txt is obviously not being processed on this iteration of current.

I'm usually ok - just my quirky monitor, a usb keyboard & mouse, a single hdmi monitor and a usb wifi adapter. I'm not using I2C, funny baud rates or GPIO. But guys doing clever things with this Pi may want a very peculiar setup and if the setup is not being processed, that's an issue. I've attached the new working cmdline.txt. I'm pretty sure the overscan setting in config.txt was 32, not 48 and have no idea where the difference came from.

sndwvs · 11-22-2022, 01:38 PM

Quote:

Originally Posted by business_kid

So the penny finally dropped here. I moved cmdline.txt to cmdline.orig, and put the kernel command line as logged in the dmesg of 15.0 in as cmdline.txt on your current image with the 6.0.9 kernel, and it booted with no extra errors and solved my overscan problem.

I know you recommend u-boot. But that pays no heed to the parameters in config.txt. I have been using the standard RazPi boot sequence of config.txt & cmdline.txt. But config.txt is obviously not being processed on this iteration of current.

I'm usually ok - just my quirky monitor, a usb keyboard & mouse, a single hdmi monitor and a usb wifi adapter. I'm not using I2C, funny baud rates or GPIO. But guys doing clever things with this Pi may want a very peculiar setup and if the setup is not being processed, that's an issue. I've attached the new working cmdline.txt. I'm pretty sure the overscan setting in config.txt was 32, not 48 and have no idea where the difference came from.

When using the kernel=Image parameter, the loading line is taken from cmdline.txt. Under kernel=u-boot.bin, the loading line is taken from /boot/uEnv.txt extraargs supplements the bootargs parameter from boot.scr

business_kid · 11-22-2022, 02:23 PM

I'm looking here at the /var/log/messages that I have built up with about half a dozen boots on this MMC card with the 6.0.9 image. My principal changes to config.txt was overscan arguments, and kernel/initrd arguments. Just focussing on the kernel command line,the arguments passed are interesting. Here's the config files

/boot/uEnv.txt

Code:

rootdev=UUID=32539e08-a5c0-4084-9f32-0932043d95e3
extraargs=coherent_pool=1M 8250.nr_uarts=1 snd_bcm2835.enable_compat_alsa=1 snd_bcm2835.enable_hdmi=1 snd_bcm2835.enable_headphones=1 smsc95xx.macaddr=00:90:5e:df:8c:81

/boot/cmdline.txt(ORIGINAL)

Code:

root=/dev/mmcblk0p2 ro rootwait nofont console=tty1 selinux=0 plymouth.enable=0 smsc95xx.turbo_mode=N dwc_otg.lpm_enable=0 elevator=noop snd-bcm2835.enable_compat_alsa=0

And the logged kernel command line

Quote:

root=UUID=32539e08-a5c0-4084-9f32-0932043d95e3 ro rootwait rootfstype=ext4 init=/sbin/init console=ttyS0,115200n8 earlyprintk console=tty1 loglevel=4 usb-
storage.quirks= coherent_pool=1M 8250.nr_uarts=1 snd_bcm2835.enable_compat_alsa=1 snd_bcm2835.enable_hdmi=1 snd_bcm2835.enable_headphones=1 smsc95xx.macaddr=00:90:5e:df:8c:81

which seems to be the 2 files one after the other. So any configuration with u-boot has to be in uEnv.txt translated into kernel arguments.

Working with config.txt & cmdline.txt, but before I fixed overscanning, only this bit of the total command line was coming from config.txt, or elsewhere

Code:

coherent_pool=1M 8250.nr_uarts=1 snd_bcm2835.enable_headphones=0 snd_bcm2835.enable_hdmi=0 snd_bcm2835.enable_hdmi=0  smsc95xx.macaddr=DC:A6:32:73:7D:DF vc_mem.mem_base=0x3ec00000 vc_mem.mem_size=0x40000000

and because of the hdmi=0, the thing was booting on 1280x720. All those "snd_bcm2835_enable..." options should be set to 1 and not 0.

To fix it, I suggest you find why config.txt settings are not being passed, or expand cmdline.txt. I'm just passing this on, as I had the logs to hand. Maybe I'm behind on the eeprom updates or something.

sndwvs · 11-25-2022, 01:40 AM

Quote:

Originally Posted by business_kid

To fix it, I suggest you find why config.txt settings are not being passed, or expand cmdline.txt. I'm just passing this on, as I had the logs to hand. Maybe I'm behind on the eeprom updates or something.

config.txt is only processed when the rpi loader is running (kernel=Image)

business_kid · 11-25-2022, 04:26 AM

Well, I am using that the "kernel=vmlinuz-6.0.9" setup, because I'm not able to run u-boot with my monitor. I'm actually using the copies with version numbers to avoid confusion, but they're identical. My setup on your recent image is identical to my 15.0 setup. 15.0 works, your recent image doesn't, because it isn't passing extra options from config.txt to the kernel. That's why I'm flagging it.

I worked around the issue with an expanded cmdline.txt. But my settings are strictly "Works For Me" and won't necessarily work for others. Come to think of it, u-boot has no provision for setup except the kernel options in uEnv.txt & cmdline.txt. How is u-boot ever going to load an overlay?

sndwvs · 11-25-2022, 04:52 AM

Quote:

Originally Posted by business_kid

Well, I am using that the "kernel=vmlinuz-6.0.9" setup, because I'm not able to run u-boot with my monitor. I'm actually using the copies with version numbers to avoid confusion, but they're identical. My setup on your recent image is identical to my 15.0 setup. 15.0 works, your recent image doesn't, because it isn't passing extra options from config.txt to the kernel. That's why I'm flagging it.

I worked around the issue with an expanded cmdline.txt. But my settings are strictly "Works For Me" and won't necessarily work for others. Come to think of it, u-boot has no provision for setup except the kernel options in uEnv.txt & cmdline.txt. How is u-boot ever going to load an overlay?

missing parameters just need to be added/replaced in /boot/uEnv.txt parameter extraargs=.

business_kid · 11-25-2022, 06:42 AM

That's crazy. I can imagine that working for 'switch' type arguments (e.g. turbo=0). The Pi boot system loads the overlays before it boots the kernel. There's no kernel firmware for the bcm2835. How do you specify, for example, your choice of video overlays or the cpu/gpu frequency with kernel arguments? The kernel only loads kernel firmware, not boot firmware.

sndwvs · 11-25-2022, 06:52 AM

when using u-boot, this is a chain from the rpi loader (where config.txt is used, loading overlays) and then passes to u-boot where uEnv.txt is already used.

business_kid · 11-25-2022, 09:53 AM

Ah

. Pity is doesn't set my monitor.