On Mon, Feb 03, 2003 at 04:40:08PM +0100, Pavel Machek wrote:
> should apt-get install
not packaged yet I'm afraid. Get
http://movementarian.org/oprofile-0.5.tar.gz, ./configure \
--with-kernel-support && make && make install
> --ctr0-event=CPU_CLK_UNHALTED && opcontrol --start
If it's an Intel CPU (not pentium 4) this should be OK
For Athlon, use INST_RETIRED
> do the trick? If that works does it mean oprofile is okay?
Run a few kernel compiles and stress it a bit. Run "op_time" to check
you get output...
> > I don't pretend to understand the PM layer at all, but it looks like
> > that both nmi.c's and oprofile's resume functions will get called. This
> > won't work: if oprofile has control of the perfctr's/nmi stuff, you
> > can't let the NMI watchdog's resume() be called, as it may conflict with
> > what oprofile is trying to resume.
>
> oprofile() should already have checks to prevent that, and I added one
I haven't seen where you added these checks. They did not exist before:
oprofile took responsibility for the NMI/perfctr handling off the NMI
watchdog entirely, then handed it back when oprofile finished.
Basically you have to ensure that only the right pm callback is called,
not both.
> [ if (nmi_watchdog == NMI_LOCAL_APIC)
I don't see an added line like this ?
regards
john
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Feb 07 2003 - 22:00:12 EST