You could try adding bootcode_delay in config.txt to delay for a few seconds before the bootloader attempts to loader start.elf
The other way of getting logs is NETCONSOLE which could be enabled on-demand via a GPIO filter
I tested first with hdmi_enable_4kp60=1 and then with core_freq=550 both cold boots and reboots and it booted fine.timg236 wrote: ↑Wed Jun 03, 2020 4:35 pmThanks for testing.lingon wrote: ↑Wed Jun 03, 2020 4:22 pmThe download problem was due to too strict cookie settings. I installed that release and the Raspberry Pi 4 cold boots and reboots fine with it from the eMMC module.
Jun 3 2020 13:08:21
Copyright (c) 2012 Broadcom
version b5de8c32f4f45a12a1fdfe107254df82965f9d56 (clean) (release) (start)
Is there anything else I could test?
Would it be possible to get the eMMC speed up to DDR50 from 22.98 MB/sec reported by hdparm?
It would useful if you could test hdmi_enable_4kp60=1 or core_freq=550 in config.txt because that's slightly more slightly to trigger a race condition.
Bootloader / start.elf performance is unlikely to get any better until the drivers are written to use the new SDHCI controller on 2711 (instead of legacy Aransan EMMC). On Linux the driver should be running at full speed but IIRC the DMA only works in the bottom 1 gigabyte of memory so you could be hitting bounce buffers in that test.
I tried booting from the eMMC module with that release also on my older Raspberry Pi computers. I have also tried out the latest release version on the same computers. Unfortunately there is a regression concerning eMMC booting in both of these firmware versions.timg236 wrote: ↑Wed Jun 03, 2020 12:29 pm
Please could you try this potential firmware fix? It's a from a dev branch and is otherwise untested so please backup the card first!
https://drive.google.com/drive/folders/ ... sp=sharing