Re: number of CPUs and IPI panic

On Mon, 2004-11-15 at 22:10, Doug White wrote:
> On Fri, 12 Nov 2004, Stephan Uphoff wrote:
> 
> > On Fri, 2004-11-12 at 09:19, Nik Azim Azam wrote:
> > > Win2K Advanced Server works just fine here.
> >
> > The root cause seems to be a misunderstanding of the interrupt delivery
> > and buffering mechanism. The fix may take a few days since it probably
> > requires extensive changes to the IPI layer.
> 
> [getting in late on this]
> 
> Tomorrow I will have access to a Dell PowerEdge 6500 (4x500MHz) I can use
> to capture info or test patches. Or see if the bug reproduces there in the
> first place :)
> 
> rwatson mumbled something about wanting to determine what triggered the
> prior IPI that the panic later trips over.  I'll have to find a sutiable
> place to hide that information...

Hopefully we can skip the debugging phase and go directly to the testing
phase. However it would be helpful to know if the PowerEdge 6500
encounters any problems without the patch.

The feedback I have received so far looks promising:

Andy Farkas wrote:
>> You can download the patch at:
>>       http://people.freebsd.org/~ups/ipi4_patch
>>
>> Any feedback is appreciated.
>>
>>       Stephan
>>
> I have been hammering the box for the last 4 hours without
> a panic.
> 
> Without your patch, I could easily get a panic by boot -s then
> fsck. I did this twice (without panic) using your patch.
> 
> Without your patch, I would usually get a panic during make
> buildworld -j4. So far with your patch its already done a
> buildworld -j8, and its about to finish a buildworld -j4.
> (2 buildworld in a row)
> 
> This is good indeed.
> 
> -andyf
> 
> ps. ok to post to list.