[9fans] thousands of interrupts per second

[9fans] thousands of interrupts per second

[David Leimbach]

I figured maybe something is going on with venti and an initial
snapshot after an install, but it's been doing this now for > 6hours.

I've got gnot showing about 15000 and higher interrupts per second...

Re: [9fans] thousands of interrupts per second

[erik quanstrom]

On Tue Nov 20 17:38:05 EST 2007, leimy2k@gmail.com wrote:
> I figured maybe something is going on with venti and an initial
> snapshot after an install, but it's been doing this now for > 6hours.
> 
> I've got gnot showing about 15000 and higher interrupts per second...

could be timesync misconfiguration.  i did this to myself once by
starting several timesyncs.  do you have any clues as to what the
machine is doing?

- erik

Re: [9fans] thousands of interrupts per second

[andrey mirtchovski]

the best way to catch timesync-related issues is to use a stop watch
to measure the duration of 'sleep 60'. seriously :)

On Nov 20, 2007 5:08 PM, erik quanstrom <quanstro@quanstro.net> wrote:
>
> On Tue Nov 20 17:38:05 EST 2007, leimy2k@gmail.com wrote:
> > I figured maybe something is going on with venti and an initial
> > snapshot after an install, but it's been doing this now for > 6hours.
> >
> > I've got gnot showing about 15000 and higher interrupts per second...
>
> could be timesync misconfiguration.  i did this to myself once by
> starting several timesyncs.  do you have any clues as to what the
> machine is doing?
>
> - erik
>

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 20, 2007 4:21 PM, andrey mirtchovski <mirtchovski@gmail.com> wrote:
> the best way to catch timesync-related issues is to use a stop watch
> to measure the duration of 'sleep 60'. seriously :)
>
>
> On Nov 20, 2007 5:08 PM, erik quanstrom <quanstro@quanstro.net> wrote:
> >
> > On Tue Nov 20 17:38:05 EST 2007, leimy2k@gmail.com wrote:
> > > I figured maybe something is going on with venti and an initial
> > > snapshot after an install, but it's been doing this now for > 6hours.
> > >
> > > I've got gnot showing about 15000 and higher interrupts per second...
> >
> > could be timesync misconfiguration.  i did this to myself once by
> > starting several timesyncs.  do you have any clues as to what the
> > machine is doing?

Oh timesync isn't what I thought it was I guess.

> >
> > - erik
> >
>

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 21, 2007 6:28 AM, David Leimbach <leimy2k@gmail.com> wrote:
>
> On Nov 20, 2007 4:21 PM, andrey mirtchovski <mirtchovski@gmail.com> wrote:
> > the best way to catch timesync-related issues is to use a stop watch
> > to measure the duration of 'sleep 60'. seriously :)
> >
> >
> > On Nov 20, 2007 5:08 PM, erik quanstrom <quanstro@quanstro.net> wrote:
> > >
> > > On Tue Nov 20 17:38:05 EST 2007, leimy2k@gmail.com wrote:
> > > > I figured maybe something is going on with venti and an initial
> > > > snapshot after an install, but it's been doing this now for > 6hours.
> > > >
> > > > I've got gnot showing about 15000 and higher interrupts per second...
> > >
> > > could be timesync misconfiguration.  i did this to myself once by
> > > starting several timesyncs.  do you have any clues as to what the
> > > machine is doing?
>
> Oh timesync isn't what I thought it was I guess.

Would timesync cause venti or something else disk I/O related to go
completely fubar?  It looks like my plan 9 box is now hosed again and
will not boot.  My hard disk activity LED was almost solid on during
this interrupt storm that was going on almost all day yesterday.

Not sure if timesync can cause that kind of behavior, and I've no idea
why a fresh install with no post-configuration should behave that way.

Dave

Re: [9fans] thousands of interrupts per second

[andrey mirtchovski]

timesync can cause the kernel to think that the cpu has a much
different frequency than it really does, which can cause time to speed
up, timers to fire much sooner than usual and so on. see what
/dev/cputype says...

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 21, 2007 6:34 AM, andrey mirtchovski <mirtchovski@gmail.com> wrote:
> timesync can cause the kernel to think that the cpu has a much
> different frequency than it really does, which can cause time to speed
> up, timers to fire much sooner than usual and so on. see what
> /dev/cputype says...
>

At this point I'm booted from the Plan 9 CD and it says I've an
AMD-Athlon 2080.... it is, in fact, an Athlon XP 2800+

Re: [9fans] thousands of interrupts per second

[erik quanstrom]

what chipset & ethernet controller.

- erik

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 21, 2007 6:47 AM, erik quanstrom <quanstro@quanstro.net> wrote:
> what chipset & ethernet controller.
>
> - erik
>

It's an nForce 2 I believe (haven't really turned this thing on for
like a year or so).

3com Ethernet controller 3c59x seems correct :-)

It's got an nvidia ethernet controller too, but I never expect that to
work anywhere but linux and windows.

Dave

Re: [9fans] thousands of interrupts per second

[andrey mirtchovski]

that is probably correct. doesn't amd only "suggest" the clock
frequency to be what they think the processor speed is when compared
to the original 1GHz Athlon?

what is the current time reported by the system? try running timesync
and monitor the cpufreq for sudden increase/decreases. of course that
may have nothing to do with your real problem, but i have experience
high interrupt counts after timesync mangled my cpufreq :)

> At this point I'm booted from the Plan 9 CD and it says I've an
> AMD-Athlon 2080.... it is, in fact, an Athlon XP 2800+
>

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 21, 2007 6:51 AM, andrey mirtchovski <mirtchovski@gmail.com> wrote:
> that is probably correct. doesn't amd only "suggest" the clock
> frequency to be what they think the processor speed is when compared
> to the original 1GHz Athlon?

Yeah their product numbers are not really the clock speed.  My
understanding is they were saying  an Athlon 2800 was clocked slower
but still about as fast as an Intel at 2800... but that might just be
my own bad memory.

>
> what is the current time reported by the system? try running timesync
> and monitor the cpufreq for sudden increase/decreases. of course that
> may have nothing to do with your real problem, but i have experience
> high interrupt counts after timesync mangled my cpufreq :)
>

Starting timesync from the live cd does cause a lot of interrupts yes.
 But my network wasn't configured either, which is seemingly the same
thing going on with my plan 9 installation since I didn't configure
any network.

perhaps timesync should not be on in the default installed
distribution, and users should turn that on themselves once they have
their network configured properly?

>
> > At this point I'm booted from the Plan 9 CD and it says I've an
> > AMD-Athlon 2080.... it is, in fact, an Athlon XP 2800+
> >
>

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 21, 2007 7:27 AM, David Leimbach <leimy2k@gmail.com> wrote:
> On Nov 21, 2007 6:51 AM, andrey mirtchovski <mirtchovski@gmail.com> wrote:
> > that is probably correct. doesn't amd only "suggest" the clock
> > frequency to be what they think the processor speed is when compared
> > to the original 1GHz Athlon?
>
> Yeah their product numbers are not really the clock speed.  My
> understanding is they were saying  an Athlon 2800 was clocked slower
> but still about as fast as an Intel at 2800... but that might just be
> my own bad memory.
>
> >
> > what is the current time reported by the system? try running timesync
> > and monitor the cpufreq for sudden increase/decreases. of course that
> > may have nothing to do with your real problem, but i have experience
> > high interrupt counts after timesync mangled my cpufreq :)
> >
>
> Starting timesync from the live cd does cause a lot of interrupts yes.
>  But my network wasn't configured either, which is seemingly the same
> thing going on with my plan 9 installation since I didn't configure
> any network.
>
> perhaps timesync should not be on in the default installed
> distribution, and users should turn that on themselves once they have
> their network configured properly?
>
>
> >
> > > At this point I'm booted from the Plan 9 CD and it says I've an
> > > AMD-Athlon 2080.... it is, in fact, an Athlon XP 2800+
> > >
> >
>

Ok another fresh install, I killed timesync, but am still getting
thousands of interrupts.  And I'm noticing I'm getting about 18 fossil
instances and 14 venti instances.

not sure if that's normal or not.

Re: [9fans] thousands of interrupts per second

[Russ Cox]

> Ok another fresh install, I killed timesync, but am still getting
> thousands of interrupts.  And I'm noticing I'm getting about 18 fossil
> instances and 14 venti instances.
> 
> not sure if that's normal or not.

Are you sure you have a recent CD?
What you're describing sounds like the venti sync problem
I fixed a month ago.

Russ

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 21, 2007 8:06 AM, Russ Cox <rsc@swtch.com> wrote:
> > Ok another fresh install, I killed timesync, but am still getting
> > thousands of interrupts.  And I'm noticing I'm getting about 18 fossil
> > instances and 14 venti instances.
> >
> > not sure if that's normal or not.
>
> Are you sure you have a recent CD?
> What you're describing sounds like the venti sync problem
> I fixed a month ago.
>
> Russ
>
>
Hmmm I deleted the original plan 9 iso file I burned this from.  It
could be pretty old, but I'm pretty sure only by a few weeks.

I'll re-download.

Dave

Re: [9fans] thousands of interrupts per second

[Gabriel Diaz]

Hello

I'm getting the same with a new kernel (pulled minutes ago), i tried two
kernel configurations (for pcf kernel), one configured with a default pcf
and other with settings to build just what i think i need.

With both kernels i get high intr. rates (stats shows ~280000), with and
without timesync.

The disk has dma on and fossil is not flushing to venti.

cpu is a core 2 duo 6600 at 2.4Ghz and sleep 60s seems to last 60 seconds :)

chipset is intel 945p and eth is a realtek 8111b gigabit (rtl8169 driver).

i'm still removing what i can from pcf to see if there is something
offending there (first i though about usbaudio, but i tried with no usb
devices connected and get the same).

I tried other kernel, an older cpu one i have around, and intr rate is
normal.

gabi


On Nov 21, 2007 5:14 PM, David Leimbach <leimy2k@gmail.com> wrote:
>
> On Nov 21, 2007 8:06 AM, Russ Cox <rsc@swtch.com> wrote:
> > > Ok another fresh install, I killed timesync, but am still getting
> > > thousands of interrupts.  And I'm noticing I'm getting about 18 fossil
> > > instances and 14 venti instances.
> > >
> > > not sure if that's normal or not.
> >
> > Are you sure you have a recent CD?
> > What you're describing sounds like the venti sync problem
> > I fixed a month ago.
> >
> > Russ
> >
> >
> Hmmm I deleted the original plan 9 iso file I burned this from.  It
> could be pretty old, but I'm pretty sure only by a few weeks.
>
> I'll re-download.
>
> Dave
>

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 21, 2007 10:28 AM, Gabriel Diaz <gabidiaz@gmail.com> wrote:
> Hello
>
> I'm getting the same with a new kernel (pulled minutes ago), i tried two
> kernel configurations (for pcf kernel), one configured with a default pcf
> and other with settings to build just what i think i need.
>
> With both kernels i get high intr. rates (stats shows ~280000), with and
> without timesync.
>
> The disk has dma on and fossil is not flushing to venti.
>
> cpu is a core 2 duo 6600 at 2.4Ghz and sleep 60s seems to last 60 seconds :)
>
> chipset is intel 945p and eth is a realtek 8111b gigabit (rtl8169 driver).
>
> i'm still removing what i can from pcf to see if there is something
> offending there (first i though about usbaudio, but i tried with no usb
> devices connected and get the same).
>
> I tried other kernel, an older cpu one i have around, and intr rate is
> normal.
>
> gabi
>
>
>
Well at least it's not my hardware, and not that I burned an ancient image :-)

Dave

Re: [9fans] thousands of interrupts per second

[erik quanstrom]

i'd guess the most reciently changed devices should head the list of
leading suspects:

--rw-rw-r-- M 72 glenda   sys   1591 Oct  4 01:40 pcmciamodem.c
--rw-rw-r-- M 72 glenda   sys  30411 Oct  8 19:31 ether82557.c
--rw-rw-r-- M 72 geoff    sys  37571 Oct  8 20:30 sdiahci.c
--rw-rw-r-- M 72 geoff    sys  34689 Oct 10 16:11 ether82563.c
--rw-rw-r-- M 72 jmk      sys  22388 Nov  8 16:20 ethervt6102.c
--rw-rw-r-- M 72 jmk      sys  45327 Nov 11 10:58 etherigbe.c

- erik

Re: [9fans] thousands of interrupts per second

[David Leimbach]

On Nov 20, 2007 4:08 PM, erik quanstrom <quanstro@quanstro.net> wrote:
>
> On Tue Nov 20 17:38:05 EST 2007, leimy2k@gmail.com wrote:
> > I figured maybe something is going on with venti and an initial
> > snapshot after an install, but it's been doing this now for > 6hours.
> >
> > I've got gnot showing about 15000 and higher interrupts per second...
>
> could be timesync misconfiguration.  i did this to myself once by
> starting several timesyncs.  do you have any clues as to what the
> machine is doing?
>

This is a straight fresh install, with 0 post install configuration
being done by myself.  It does say how it can't find my IP address,
which would make me think you're right on about the timesync issue,
but that it could be getting screwed up by just one?


> - erik
>