freebsd stable - Aug 2013 - 9.2-RC3 - suspend/resume causes slow system performance

I opened a ticket about this: kern/181632

Basically, if I do a 'zzz' and then wake the system up, operations like
a
buildworld or portmaster take much longer after the resume - systat shows
very low CPU/disk utilization.  It's 100% repeatable, I don't recall
this
happening on 9.1

I build 9.2-RC3 from source (as I have for years) - this is a fairly
generic SMP system (i5 750 with 4 cpus).

Let me know if I can provide any data - this is a fairly significant
regression for me as I have taken to bringing the system up as needed vs.
leaving it on all of the time.

Wow. Uhm, can you downgrade to 9.1 and see if it still happens?

Would you be able to bisect the kernel source and see if you can find where
along stable/9 it happened? That's the fastest way to determine what broke.

Thanks!



-adrian



On 29 August 2013 06:32, Mike Harding <mvharding at gmail.com> wrote:
> I opened a ticket about this: kern/181632
>
> Basically, if I do a 'zzz' and then wake the system up, operations
like a
> buildworld or portmaster take much longer after the resume - systat shows
> very low CPU/disk utilization.  It's 100% repeatable, I don't
recall this
> happening on 9.1
>
> I build 9.2-RC3 from source (as I have for years) - this is a fairly
> generic SMP system (i5 750 with 4 cpus).
>
> Let me know if I can provide any data - this is a fairly significant
> regression for me as I have taken to bringing the system up as needed vs.
> leaving it on all of the time.
> _______________________________________________
> freebsd-stable at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe at
freebsd.org"
>

On 29/08/2013 15:32, Mike Harding wrote:
> I opened a ticket about this: kern/181632
> 
> Basically, if I do a 'zzz' and then wake the system up, operations
like a
> buildworld or portmaster take much longer after the resume - systat shows
> very low CPU/disk utilization.  It's 100% repeatable, I don't
recall this
> happening on 9.1
> 
> I build 9.2-RC3 from source (as I have for years) - this is a fairly
> generic SMP system (i5 750 with 4 cpus).
> 
> Let me know if I can provide any data - this is a fairly significant
> regression for me as I have taken to bringing the system up as needed vs.
> leaving it on all of the time.
I encountered that yesterday.

Manually stepping down:
sysctl dev.cpu.0.freq=800

And back up again:
sysctl dev.cpu.0.freq=2400

fixed this for me. You can simply run powerd to mask the problem. It
might have been around for ages as far as I know, because I usually
use powerd. I just had it turned of for performance measurements and
forgot to turn it back on before going into suspend. When I resumed
and reran the tests the performance suddenly was abysmal.

-- 
A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?