> Date: Mon, 6 Jul 2020 13:49:26 +0100
> From: Chavdar Ivanov <[email protected]>
>
> On Mon, 6 Jul 2020 at 02:14, Taylor R Campbell <[email protected]> wrote:
> >
> > This and the earlier panic you reported may be fixed by
> >
> > https://mail-index.netbsd.org/source-changes/2020/07/06/msg119081.html
>
> I upgraded this morning. Unfortunately it did not fix it for me,
> though I did not get a panic. I got a deep CPU loop with a fully
> spinning fan and no reaction to the attempt to break into the
> debugger, so I had to hit the power button.
Well, that's unfortunate!
I am trying to reproduce on a wifi network with WPA2, verified by the
`ifconfig -v' counters to be using software CCMP, while simultaneously
doing heavy cgd I/O, and so far I haven't been able to reproduce.
I'll keep investigating.
The attached patch might help with breaking into the debugger, if
you'd like to give it a try -- won't fix any underlying problem, but
may help us to find out what the problem is.
diff -r 622b4cb96b07 sys/arch/x86/x86/fpu.c
--- a/sys/arch/x86/x86/fpu.c Sun Jul 05 14:24:05 2020 +0000
+++ b/sys/arch/x86/x86/fpu.c Mon Jul 06 15:49:34 2020 +0000
@@ -383,9 +383,10 @@ fpu_kern_enter(void)
struct cpu_info *ci;
int s;
- s = splhigh();
+ s = splnet();
ci = curcpu();
+ KASSERTMSG(ci->ci_ilevel <= IPL_NET, "ilevel=%d", ci->ci_ilevel);
KASSERT(ci->ci_kfpu_spl == -1);
ci->ci_kfpu_spl = s;
