Current lockup on 'sysctl lock'

About 3-5 time per day recent current locksup on my noteboot.

Symptoms: 
system responsive, not locked processes works, I can  switch X to sc0
and back.

processes locked on 'sysctl lock' (see below)
allways there is one process sleeping on allproc

usually it is related with high activity (this case evolution gets bunch
of night mail and pass it through spam-assassin)

I do not know why proc 1037 have such strange backtrace but suspect that
evolution was forked before exec spamc (or like).


proc 658 was created after lock (I switched on free console and type
login) - login was executed but even does not ask password.

May be it is same issue as reported there about lockup of top and ps.

Unfortunately there was no comconsole, so I have retype most important
info:

db> ps
1037 ... 0000000 new [INACTIVE] evolution-1.5
...
 790 ... 008c001  stop(threaded) evolution-1.5
     thread ... [SUSP]
     thread ... [SLPQ spltwt ...][SLP]
     thread ... [SUSP]
     thread ... [SLPQ select ...][SLP][SUSP]
     thread ... [SUSP]
 781 ... 0004001 [SLPQ sysctl lock ...][SLP] wireless-applet
...
 731 ... 0004000 [SLPQ sysctl lock ...][SLP] metacity
...
 477 ... 0000000 [SLPQ allproc ...][SLP] cron
...
 658 ... 0004002 [SLPQ sysctl lock ...][SLP] login
...
db> tr 1037
sched_switch(0,...)
(null)(1,..) at 0
_end(1,...)
_end(1,...)
... a lot of _end(), different args on stack
_end(1,...)
_end() at ...
db> tr 781
sched_switch()
mi_switch()
sleepq_switch()
sleepq_wait()
cv_wait()
_sx_xlock()
userland_sysctl()
__syscall()
...
db> tr 731
sched_switch()
mi_switch()
sleepq_switch()
sleepq_wait()
cv_wait()
_sx_xlock()
userland_sysctl()
__syscall()
...
db> tr 477
sched_switch()
mi_switch()
sleepq_switch()
sleepq_wait()
cv_wait()
_sx_xlock()
fork1()
fork()
syscall()
...
db> tr 658
sched_switch()
mi_switch()
sleepq_switch()
sleepq_wait()
cv_wait()
_sx_xlock()
userland_sysctl()
__syscall()
...


-- 
Vladimir B. Grebenschikov
SWsoft Inc.