Sparc slowdown - problem identified...

[Tillman]

On Fri, Aug 15, 2003 at 10:14:19PM -0700, Marcel Moolenaar wrote:
> On Fri, Aug 15, 2003 at 10:31:59PM -0600, Tillman wrote:
> > 
> > hme0: <Sun HME 10/100 Ethernet> mem 0xe0000000-0xe0007fff at device 1.1 on pci1
> > hme0: Ethernet address: 08:00:20:c6:7f:c7
> > 
> > hme1: <Sun HME 10/100 Ethernet> mem 0x2800000-0x2807fff at device 0.1 on pci3
> > hme1: Ethernet address: 08:00:20:c6:7f:c7
> > 
> > arp: 192.168.23.3 is on hme0 but got reply from 00:10:4b:69:2a:86 on hme1
> > arp: 192.168.23.3 is on hme0 but got reply from 00:10:4b:69:2a:86 on hme1
> > arp: 192.168.23.3 is on hme0 but got reply from 00:10:4b:69:2a:86 on hme1
> > 
> > Any ideas?
> 
> Both hme0 and hme1 have the same MAC address. Suspicious...
> 
> > Is it possible that the hme interfaces are numbered in a
> > different order with the new kernel, similar to how the disk devices
> > could have been renumbered (but that wasn't an issue for me)?
> 
> Yes, definitely. When you enable OFW_NEWPCI, compare the old dmesg(8)
> with the new one to see what has changed and correct your setup
> accordingly.

Results:

 Old buildkernel time:

real    477m30.067s
user    400m42.592s
sys     24m45.618s

 New buildkernel time:

real    151m42.832s
user    125m20.230s
sys     13m10.643s

For folks attempting this, note that my 4-port ethernet card is now
detected /after/ the onboard interface. Thus, I have hme0 & hme[1-4]
instead of hme[0-3] & hme4. Rather than reconfigure, it was easier to
just change cables around :-)

-T


-- 
Yield to temptation; it may not pass your way again.
	- Robert Heinlein