Hi,
I'm currently running on -CURRENT from November, though I think the message
for amdgpu.ko ("kernel: linker_load_file: /boot/modules/amdgpu.ko -
unsupported file type") started to appear as early as last August. Since
this isn't my daily driver (yet!), I didn't notice for quite a while, and
scfb is adequate for what little I currently do on FreeBSD.
When I initially installed this machine in Aug 2020, my GPU (Vega10 Renoir)
wasn't supported by DRM, so over time I made several builds of the
graphics/drm-devel-kmod port, which worked for a while last year.
In the last weeks, in the time I managed to set aside for this work, I made
several attempts to bring this machine up-to-date, using "pkg upgrade",
building my own from /usr/src (always up-to-date using 'git pull'), and
beinstall.sh. (both separately and in combination). I always work with
separate build environments for these attempts.
All my upgrade attempts have resulted in a system that boots, but neither
starts X nor lets me log in on a console. There are "linker_load_file ... "
message for several .ko files, all preceded by a line like this:
"KLD hconf.ko: depends on kernel - not available or version mismatchmodule
name"
here are the paths I get this message about:
/boot/kernel/cuse.ko
/boot/kernel/hconf.ko
/boot/kernel/hms.ko
/boot/kernel/hmt.ko
/boot/kernel/intpm.ko
/boot/kernel/lindebugfs.ko
/boot/kernel/linux.ko
/boot/kernel/linux64.ko
/boot/modules/amdgpu.ko
/boot/modules/drm.ko
The information I found on the Internet about this error message seem to
indicate that I need to re-build the .ko being reported by the message with
the current kernel, which is .. hard if I can't even log in ;-). I would
also not expect a version mismatch after a pkg upgrade (but no, I didn't go
through all the output that generated).
Does anyone have an idea what I can do to get past this obstacle? I could
of course always re-install, but that feels like giving in :-)
TIA
Michael
--
Michael Schuster
http://recursiveramblings.wordpress.com/
recursion, n: see 'recursion'
