Re: bug in softint_execute() ?

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

On Fri, Apr 11, 2008 at 08:48:44AM -0700, Tim Rightnour wrote:

> On 11-Apr-2008 Takeshi Nakayama wrote:
> > FYI, I got a similar strange assertion panic on sparc64 SMP box.
> > 
> > panic: kernel diagnostic assertion "curcpu() == ci" failed: file
> > "../../../../kern/vfs_bio.c", line 1541
> 
> Since you brought it up,  I have also seen this one, once:
> 
> panic: kernel diagnostic assertion "ci == curcpu()" failed: file "/usr/src/hac
> kathon/sys/kern/kern_idle.c", line 69
> Stopped in pid 0.10 (system) at netbsd:cpu_Debugger+0x10:       lwz     r0, r1
> , 0
> 
> 0xaa22be10: at panic+0x228
> 0xaa22be60: at __kernassert+0x4c
> 0xaa22be70: at idle_loop+0x224
> 0xaa22bed0: at cpu_spinstart+0x5c
> 0xaa22bf30: at 0xfffffffc
> 0xaa22bf40: at emptyidlespin+0
> 0xaa22bf50: at 0xfffffffc

I think that one is likely to be a bug in MD code. I know of nothing that
would cause the idle thread to migrate, or l_cpu to be updated incorrectly.

Andrew