10-02-2024, 10:44 PM
(This post was last modified: 10-02-2024, 11:06 PM by mifritscher.)
While the kernel can work with a 586, it seems that init does need at least a 586 :-( The kernel cries with "Attempting to kill init" - the backtrace shows a exception regarding an illegal exception.
Tested with
qemu-system-x86_64 -m 256 -cpu 'pentium2-v1' /dev/sdd
.
The culprit: /lib/runit/runit-init uses CMOV (2 times in the whole binary), and this is supported only on Pentium Pro onwards.
Would it be possible to recompile the binaries with march=586 (or pentium)? Then I could probably run it on an IBM Thinkpad 760 (P1-133 and 64 MB RAM - for the RAM see the other thread) :-)
Btw, even the K6... K6-3 are 586 class CPUs.
Another way to fix this would by to apply the kernel patch found under https://lore.kernel.org/lkml/20210626130...@orca.pet/ .
This way, the user mode binaries do not to be changed :-)
Tested with
qemu-system-x86_64 -m 256 -cpu 'pentium2-v1' /dev/sdd
.
The culprit: /lib/runit/runit-init uses CMOV (2 times in the whole binary), and this is supported only on Pentium Pro onwards.
Would it be possible to recompile the binaries with march=586 (or pentium)? Then I could probably run it on an IBM Thinkpad 760 (P1-133 and 64 MB RAM - for the RAM see the other thread) :-)
Btw, even the K6... K6-3 are 586 class CPUs.
Another way to fix this would by to apply the kernel patch found under https://lore.kernel.org/lkml/20210626130...@orca.pet/ .
This way, the user mode binaries do not to be changed :-)