Control: tag -1 - unreproducible + confirmed upstream
On 03.08.2025 09:27, Michael Tokarev wrote:
qemu-system-i386 -enable-kvm -smp 6 -m 4G -cdrom NetBSD-10.1-i386.iso
Choose an option; RETURN for default; SPACE to stop countdown. Option 1 will be chosen in 0 seconds. 20389236+589100+746196 [919406+939408+1024752]=0x17787ac WARNING: couldn't open /var/db/entropy-file Loading /stand/i386/10.1/modules/cd9660/cd9660.kmod WARNING: 1 module failed to load [reboot] .. This module loading failure looks rather suspicious. And it doesn't happen with 10.1.0-rc1. It does not look right at all. Maybe your failure is actually similar to mine, the prob is that the netbsd kernel is unable to correctly read the cd-rom?
Ok. git bisection shows that this commit, introduced between 9.0 and 9.1, is the one which broke netbsd.iso like above (not like in the original report, which I still can't reproduce): commit 88dd4ca06c8392155289e5462cd26af3762a1b04 Author: Zhao Liu <[email protected]> Date: Wed Apr 24 23:49:14 2024 +0800 i386/cpu: Use APIC ID info to encode cache topo in CPUID[4] Refer to the fixes of cache_info_passthrough ([1], [2]) and SDM, the CPUID.04H:EAX[bits 25:14] and CPUID.04H:EAX[bits 31:26] should use the nearest power-of-2 integer. .... (I had to apply a fix target/i386/cpu.h on top of it, or else qemu does not build at all due to missing ArchCPU.l1_cache_per_core). Next, the following commit (part of 10.1.0-rc0) fixes the issue for me: commit f985a1195ba2d9c6f6f33e83fe2e419a7e8acb60 Author: Chuang Xu <[email protected]> Date: Mon Jul 14 16:08:56 2025 +0800 i386/cpu: Fix number of addressable IDs field for CPUID.01H.EBX[23:16] When QEMU is started with: -cpu host,migratable=on,host-cache-info=on,l3-cache=off -smp 180,sockets=2,dies=1,cores=45,threads=2 On Intel platform: CPUID.01H.EBX[23:16] is defined as "max number of addressable IDs for logical processors in the physical package". .... Since the latter commit which fixed the issue in 10.1 isn't directly back-portable to 10.0, we'll think about other possibilities in this area. BTW, the same issue is present when using qemu-system-x86_64. And it only affects intel-based CPUs, it does not happen on AMD. That's why I weren't able to repro it before. Let's mark it as confirmed+upstream, hopefully it's the same issue as reported. /mjt
