calcru: runtime went backwards, RELENG_6, SMP

Dmitry Morozovsky marck at
Tue Jun 12 09:33:35 UTC 2007

On Mon, 11 Jun 2007, Matthew Dillon wrote:

MD> :==============================================================================
MD> :     2 u  354 1024  377    5.305  -66314. 4321.47
MD> :  2 u  365 1024  377    6.913  -66316. 4305.33
MD> :       2 u  358 1024  377    7.939  -66308. 4304.90
MD> :
MD> :Any directions to debug this?
MD> :
MD> :Sincerely,
MD> :D.Marck                                     [DM5020, MCK-RIPE, DM3-RIPN]
MD>     Since you are running on HEAD now, could you also kgdb the live kernel
MD>     and print cpu_ticks?  I believe the sequence is (someone correct me if
MD>     I am wrong):
MD>     kgdb /kernel /dev/mem
MD>     print cpu_ticks

s,/kernel,/boot/kernel/kernel, ;-)

well, strange enough result for me:

(kgdb) print cpu_ticks
$1 = (cpu_tick_f *) 0xffffffff8036cef0 <rdtsc>

Does this mean that kernel uses tsc? sysctl reports

kern.timecounter.choice: TSC(-100) ACPI-fast(1000) i8254(0) dummy(-1000000)
kern.timecounter.hardware: ACPI-fast

MD>     As for further tests... try building a non-SMP kernel (i.e. one that
MD>     only recognizes one cpu) and see if the problem occurs there.  That
MD>     will determine whether there is a basic problem with time keeping or
MD>     whether it is an issue with SMP.

As I'd reported in the first mail, at least RELENG_6/amd64/GENERIC (SMP is on 
by default only on HEAD) behaves similarly.

There may be some hardware issues with this particular mobo though...

D.Marck                                     [DM5020, MCK-RIPE, DM3-RIPN]
*** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck at ***

More information about the freebsd-stable mailing list