Re: SPL NOT ZERO ON SYSCALL ENTRY

[Andrew Doran <ad%NetBSD.org@localhost>]

On Mon, Apr 28, 2008 at 05:54:03PM +0200, Matthias Drochner wrote:

> ad%NetBSD.org@localhost said:
> > I just made some changes to i386/lock_stubs.S that prevent the same
> > errors when running with a preemptive kernel. There is a good chance
> > that it also fixes this problem.. 
> 
> I just managed to trigger such a problem again.
> CPU1 hit the breakpoint I planted
> after the "SPL NOT LOWERED ON SYSCALL EXIT" in Xsyscall().
> CPU0 was in Xtrap0e (which might be inaccurate).
> And again:
> (gdb) print cpu_info_primary.ci_next->ci_istate
> $2 = {ipending = 0x40000000, ilevel = 0x7}

ci_mtx_oldspl, ci_mtx_count?

> I've got a coredump and a kernel with debug info. Is there
> anything I could do to help?

I'm out of ideas as to what could cause this. You could instrument
arch/x86 and arch/i386 to record the last place that the IPL was raised
to IPL_HIGH before the panic. A bit hit and miss, but perhaps it could
be useful.

Andrew