https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=287872

--- Comment #8 from Eugene Grosbein <eu...@freebsd.org> ---
I tried to use hwpmc(4) to profile the kernel:

# kenv kern.hwpmc.logbuffersize=64
# kenv kern.hwpmc.nbuffers_pcpu=512
# kldload hwpmc

Then removed all vlans, stopped devd, recreated vlans. Then:

# pmcstat -S inst_retired.any_p -O devd.out &
# service devd start
# kill -INT %1
# pmcstat -R devd.out -g

And "gprof -l /boot/kernel/kernel inst_retired.any_p/kernel.gmon" shows the
following:

  %   cumulative   self              self     total
 time   seconds   seconds    calls  ms/call  ms/call  name
 44.5  131231.00 131231.00    28705  4571.71  4571.71  lock_delay [2]
 22.6  197818.00 66587.00    67341   988.80   988.80  DELAY [19]
  3.1  207038.00  9220.00     2875  3206.96  3807.34  pmap_remove_pages <cycle
4> [36]
  1.4  211181.00  4143.00     1134  3653.44  7948.42  pmap_enter_quick_locked
[38]
  1.3  215064.00  3883.00      956  4061.72  4138.80  get_pv_entry [48]
  1.3  218829.00  3765.00     1053  3575.50  4107.77  pmap_enter [45]
  1.0  221869.00  3040.00     1639  1854.79  4229.11  pmap_try_insert_pv_entry
[39]

Above profiling for stock /sbin/ifconfig using netlink. Below for
/sbin/ifconfig rebuilt without netlink and same profiling procedure:

  %   cumulative   self              self     total
 time   seconds   seconds    calls  ms/call  ms/call  name
 11.1     581.00   581.00      422  1376.78  1376.78  counter_u64_fetch [16]
  8.7    1037.00   456.00      226  2017.70  2017.70  DELAY [25]
  6.6    1383.00   346.00      375   922.67  1037.36  pmap_remove_pages [26]
  6.2    1709.00   326.00      187  1743.32  1743.32  strncmp [28]
  3.2    1875.00   166.00       10 16600.00 21475.00  rtsock_msg_buffer [49]
  2.6    2013.00   138.00       86  1604.65  1889.35  pmap_enter [52]
  2.3    2134.00   121.00      512   236.33  2146.13  sysctl_iflist [7]
  2.3    2253.00   119.00      124   959.68  1009.98  get_pv_entry [57]
  2.1    2361.00   108.00      220   490.91  1000.49  pmap_try_insert_pv_entry
[48]

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to