On Sun, 4 Sep 2022, Stephen Walsh wrote: > > What happens if you boot the known-good initrd (v5.15) with the > > crashy vmlinux binary (v5.18)? > > > The combination won't boot normally but should fail normally (wrong > > modules) and not fail with an oops. > > It still crashes, but I can see the process. >
OK. I assume that it was the initrd change that prevented the oops and not the addition of ignore_loglevel. Your results suggest to me that the problem was triggered by the growth of the initrd between v5.15 and v5.18. My guess is that the bootloader made use of the 8 MB memory chunk which was later discarded by the kernel. [ 0.000000] Ignoring memory chunk at 0x7800000:0x800000 before the first chunk This may actually be a bootloader bug, i.e. memory chunks not in the correct sequence (?). Anyway, I agree with Ingo, this problem has to do with memory layout and/or bootloader behaviour and the way to avoid it is probably a memfile. As I have no experience with booting Linux on Amigas I can't be of much more help unfortunately. So I'll stop guessing and leave this to the Amiga experts. Regards, Finn
