I would like to extend my deep thanks for Mateusz Jo{\'n}czyk for the
helpful suggestions in postings to this list on 6-Jan-2025

        https://lists.debian.org/debian-mips/2025/01/msg00000.html
        https://lists.debian.org/debian-mips/2025/01/msg00001.html

in response to my problem report of 5-Dec-2024

        https://lists.debian.org/debian-mips/2024/12/msg00000.html

Two key points raised by Mateusz are the (previously unknown to me)
distinction between the 5KEc and 5KEf CPU variants (the 5KEc lacks
hardware floating-point instructions, while the 5KEf has them), and
the suggestion to try a different vmlinuz kernel version.

I'm pleased to report that the kernel switch was the needed fix to my
installer script: I change it to use vmlinuz-6.1.0-27-5kc-malta.  The
job then completed successfully, and gave me a working virtual
machine. It now uses the vmlinuz-6.1.0-28-5kc-malta kernel that became
available after Debian 12 system updates.

It is, however, still puzzling that, in numerous attempts, the
installer could reach 75% to 79% completion with the
vmlinuz-5.10.0-20-5kc-malta kernel, before going into a hung state
with repeated stuck-CPU complaints.  That is why I didn't think to try
a kernel switch.  That is probably a good lesson for the future when
installers fail!

My Debian 12 build farm now has VMs with these CPUs:

        alpha             hppa              mips32            s390x
        arm64             loong64           mips64            sparc64
        armhf             m68k              ppc64le           x86

They run with several different versions of qemu-system-xxx.

I made experiments with one of the more recent versions of the
qemu-system-mips64el, trying all combinations of -machine and -cpu
options, while requesting 32GB of DRAM.  Alas, the largest memory size
supported by QEMU appears to be just 2GB, addressable with 31 bits,
despite the processor being a 64-bit one.

Our site has had numerous SGI MIPS-based physical systems since 1992,
but the last of them, an SGI O2 R10000-SC, died with a power-supply
failure about a year ago.  My records show that I once had guest
access to a large SGI Origin 2000 with 64 R10000 CPUs and 8GB DRAM,
which is why I expected to find QEMU support for larger memory sizes.

-------------------------------------------------------------------------------
- Nelson H. F. Beebe                    Tel: +1 801 581 5254                  -
- University of Utah                                                          -
- Department of Mathematics, 110 LCB    Internet e-mail: [email protected]  -
- 155 S 1400 E RM 233                       [email protected]  [email protected] -
- Salt Lake City, UT 84112-0090, USA    URL: https://www.math.utah.edu/~beebe -
-------------------------------------------------------------------------------

Reply via email to