Subject: saupcpl leak with i386 netbsd-4 and PTHREAD_CONCURRENCY>1?
To: None <current-users@NetBSD.org>
From: Paul Ripke <firstname.lastname@example.org>
Date: 10/16/2006 21:12:54
I realize that PTHREAD_CONCURRENCY>1 isn't recommended for general
consumption as yet, but I decided to give it a go with firefox
(native build on i386) on an oldish dual P-III box with 128 MB RAM
(long story, but the magic blue smoke escaped from my half-decent
After a bit of use, a couple of coredumps from firefox (appears
normal, no matter what concurrency), and one hang requiring
"kill -9", I found the poor thing thrashing badly,
and went looking where the RAM had gone, and discovered:
ksh$ vmstat -m | egrep '^[A-Z]|^saup'
vmstat: Kmem statistics are not being gathered by the kernel.
Memory resource pool statistics
Name Size Requests Fail Releases Pgreq Pgrel Npage Hiwat Minpg Maxpg Idle
saupcpl 1612 119234493 0 119198379 18103 44 18059 18061 0 inf 1
In use 64413K, total allocated 83316K; utilization 77.3%
Shutting down all pthread apps didn't improve matters, the only
option was reboot. This is a plain GENERIC.MP from the netbsd-4 branch
about a month back.
Is this a known issue? Anything I can look at here?
I love deadlines. I like the whooshing sound they make as they fly by.
-- Douglas Adams