Re: crash running tests (x86_fpregs_xmm_read) under UTM(QEMU+maOS-Hypervisor) with today's -current

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Thu, Mar 20, 2025 at 09:37:12PM -0700, Greg A. Woods wrote:
> Running tests on ~yesterday's -current (amd64 running under UTM on
> macOS), and this is the last thing I see:
> 
>     x86_fpregs_xmm_core: [0.076360s] Passed.
>     x86_fpregs_xmm_read: Connection to nbt.local closed by remote host.
> 
> The console said:
> 
> [ 1753.0907783] panic: kernel diagnostic assertion "!(l->l_md.md_flags & MDL_FPU_IN_CPU)" failed: file "/Volumes/work/woods/g-NetBSD-src/sys/arch/x86/x86/fpu.c", line 149
> [ 1753.0907783] cpu1: Begin traceback...
> [ 1753.0907783] vpanic() at netbsd:vpanic+0x171
> [ 1753.0907783] kern_assert() at netbsd:kern_assert+0x4b
> [ 1753.0907783] process_read_fpregs_xmm() at netbsd:process_read_fpregs_xmm+0x122
> [ 1753.0907783] process_read_fpregs() at netbsd:process_read_fpregs+0x9
> [ 1753.0907783] proc_regio() at netbsd:proc_regio+0x69
> [ 1753.1016891] do_ptrace() at netbsd:do_ptrace+0xc4f
> [ 1753.1016891] sys_ptrace() at netbsd:sys_ptrace+0x2b
> [ 1753.1016891] syscall() at netbsd:syscall+0x112
> [ 1753.1016891] --- syscall (number 26) ---
> [ 1753.1016891] netbsd:syscall+0x112:
> [ 1753.1016891] cpu1: End traceback...
> 
> [ 1753.1016891] dumping to dev 168,2 (offset=1951, size=2096140):
> [ 1753.1016891] dump fatal page fault in supervisor mode
> [ 1754.1092530] trap type 6 code 0 rip 0xffffffff80f774a5 cs 0x8 rflags 0x10246 cr2 0xdeadbf17 ilevel 0x6 rsp 0xffff8f0279cccce8
> [ 1754.1092530] curlwp 0xffff881219751c00 pid 1059.1059 lowest kstack 0xffff8f0279cc82c0

I've seen similar panics on the Xen testbed with PVH or PVHVM,
also while running fpregs tests.
I've not seen it for some time now, but it has always been once in a while

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--