Argh, hit 'send' too fast. Here's some mpstat output. Note all xcalls
are registered to just one of the two CPUs.
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys
wt idl
0 0 0 101 643 205 43 0 1 4 0 0 0
1 0 99
1 0 0 0 127 103 37 0 3 4 0 48 0
0 0 100
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys
wt idl
0 0 0 99 631 201 42 0 3 2 0 0 0
2 0 98
1 0 0 0 136 104 39 0 2 1 0 47 0
1 0 99
CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys
wt idl
0 0 0 100 637 203 49 0 3 0 0 0 0
1 0 99
1 0 0 0 138 106 37 0 4 1 0 48 0
1 0 99
On May 8, 2009, at 2:33 AM, Dale Ghent wrote:
While doing some new hme/qfe driver testing on a 2x400MHz ultra2
tonight, I noticed something odd - lots of cross-calls while this
box was sitting idle. Roughly 100/second. A quick 20 second dtrace
revealed the outlaying stack:
SUNW,UltraSPARC-II`send_one_mondo+0x14
SUNW,UltraSPARC-II`send_mondo_set+0x1c
unix`xt_all+0xac
SUNW,UltraSPARC-II`do_scrub_ecache_line+0xa4
genunix`callout_list_expire+0x5c
genunix`callout_expire+0x14
genunix`callout_realtime+0x14
genunix`cyclic_softint+0xb4
unix`cbe_level1+0x8
unix`intr_thread+0x168
2010
Maybe I've just never noticed this before, having not worked on this
hardware since I last powered it down years ago, but is the E$
scrubber supposed to be triggering so many xcalls?
/dale
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org
