Last week or the week before I mentioned somewhere that I was having
font issues in the console of my 9.0 sysv build on the new machine
(ryzen 5 3400G i.e. a Picasso APU). Today I ended up spending a lot
longer than I had expected looking at the bootscript! I was going
to mention this on -support as a "probably not needed for general
use" fix, but now that I've booted 5.3.1 I'm not sure if it is
needed at all (except for 5.2 kernels).
The first problem was that in my kernel config I started from 'make
defconfig' and didn't notice that fonts are in 'library routines'
(Select compiled-in fonts). And specifically for using 12x22, the
Sparc console 12x22 font. Clearly, without that there was no way
that I could load my preferred 12x22.
But using setfont for an 8x16 font in the console init script was
still failing although I could run it on each tty after logging in.
There was also an error message from that script, although it said
[ok] afterwards:
*ERROR* construct Invalid Connector Object ID for Adapter Service for connector
index:2! type 0 expected 3
Instrumenting the script, and suspecting a race, I eventually
settled on the following change (and put the error as a comment):
diff -Naur a/lfs/init.d/console b/lfs/init.d/console
--- a/lfs/init.d/console 2013-01-24 22:25:14.000000000 +0000
+++ b/lfs/init.d/console 2019-09-23 16:50:09.579343530 +0100
@@ -80,7 +80,11 @@
done
# Set the font (if not already set above) and the keymap
- [ "${use_fb}" == "1" ] || [ -z "${FONT}" ] || setfont $FONT || failed=1
+ # Without the sleep, setfont can fail on amdgpu framebuffer
+ # and even with it there can be an error message on 5.2 kernels
+ # *ERROR* construct Invalid Connector Object ID for Adapter Service
+ # for connector index:2! type 0 expected 3
+ [ "${use_fb}" == "1" ] || [ -z "${FONT}" ] || sleep 1 ; setfont $FONT ||
failed=1
[ -z "${KEYMAP}" ] ||
loadkeys ${KEYMAP} >/dev/null 2>&1 ||
Using that in 5.2.17 now successfully loads my specified font but
still produces the ERROR message (I tried adding 2>/dev/null but
that did not suppress it). So, I was going to use this on future
builds.
But when booting 5.3.1 I no-longer get the error message so whenever
I make my next build (probably not this week) I'll try without the
patch. I suspect the problem might be specific to Picasso APUs and
5.2 kernels (noted with 5.2.11 and 5.2.17, no earlier kernels were
used).
Fun, isn't it ?
ĸen
--
thread 'main' panicked at 'giraffe',
/tmp/rustc-1.32.0-src/src/test/run-fail/while-panic.rs:17:13
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
