64 bit time_t oddness
Garance A Drosihn
drosih at rpi.edu
Sun Feb 15 16:05:05 PST 2004
At 5:13 PM -0600 2/15/04, matt wrote:
>Anyhow, here is the strangeness. After compiling/installing the
>64 bit time_t world/kernel, I'm getting some serious lag on this
>machine. When logged in via ssh, the shell freezes every few
>seconds for a brief period of time. Then I get this with top -S:
>last pid: 81822; load averages: 1.18, 1.06, 0.98 ...
>46 processes: 2 running, 28 sleeping, 16 waiting
>CPU states: 8.9% user, 0.0% nice, 72.3% system,
> 0.0% interrupt, 18.8% idle
>Mem: 11M Active, 5368K Inact, 19M Wired, 32K Cache,
> 8848K Buf, 954M Free
>Swap: 2022M Total, 2022M Free
> PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
> 11 root -16 0 0K 40K RUN 31:42 38.96% 38.96% idle:
> 8 root 171 52 0K 40K pgzero 1:12 3.96% 3.96% pagezero
> 473 root 130 0 3624K 1952K select 0:47 0.00% 0.00% dhclient
>At any given time, I only seem to have ~40% (or less) cpu
>available for anything. The clean make buildworld/buildkernel
>with 32 bit time_t didn't have this problem, the sources are
>identical (no cvsup's between builds).
I am not seeing this behavior. My 64-bit time_t system is sitting
at 99% to 100% idle. One thing I notice is that you're running
dhclient there, and I am not. (I did do some testing with a DHCP
setup, but not much). Is your ethernet connection configured
correctly? Did you remember to remove /var/db/dhclient.leases ?
I wonder if there's anything else that needs to be done for people
who are using DHCP.
Do you have any interesting options turned on (or off) in your
kernel? My kernel == the GENERIC kernel, except that I have
commented out the WITNESS and WITNESS_SKIPSPIN options. I am
running the ULE scheduler (SCHED_ULE), which was a recent
change to the GENERIC kernel. Are you using the older scheduler?
Garance Alistair Drosehn = gad at gilead.netel.rpi.edu
Senior Systems Programmer or gad at freebsd.org
Rensselaer Polytechnic Institute or drosih at rpi.edu
More information about the freebsd-sparc64