Re: 2.6.27-rc0 Power Bugs with HP/Compaq Laptops

Re: 2.6.27-rc0 Power Bugs with HP/Compaq Laptops

[Rafael J. Wysocki]

[Adding CCs]

On Thursday, 7 of August 2008, Matt Parnell wrote:
> I have had a long ongoing ordeal dealing with the problems associated
> with the HP Pavilion Laptops, mine being a dv9000 series. It seems
> that these either have a broken HPET table, a missing HPET table, or
> some other ACPI issue, as this thing has always had many issues
> regarding ACPI. First, C1E was broken until I totally disabled it with
> a patch, but since .27, that code has been rewritten, and now for some
> reason HPET isn't working anymore, as I have to set the
> clocksource=tsc. On top of that, the tsc clocksource is unstable, and
> with or without dynticks the .27 kernels on my laptop feel slower, and
> laggy, as even KDE slows down whilest using it. Everything works fine
> if I disable the local apic (lapic), but when I do, only one of my CPU
> cores is detected and used, and the CPU frequency scaling doesn't work
> either. No matter what I do with .27, it just won't work.
> 
> I've tried with the linux next kernel as suggested by Eduard Munteanu,
> to no avail with the same problem. He said this:
> 
> "This is not the right fix. Instead, HPET should be used as a timer
> source for dynticks. This is not in 2.6.26, but I tried the linux-next
> tree (google it) and it looked like dynticks worked fine, without
> disabling C1E. Can't place my finger on the exact commit, but you could
> check."
> 
> Apparently, it's either this or something closely related that is
> causing such calamity for users of these HP/Compaq laptops as myself.
> I say this not just for me, as there are many thousands who have
> issues with these laptops and their crappy bios implementations. Most
> are n00bs using Ubuntu, who can't even voice what's wrong, as they are
> new to linux anyway.
> 
> That said, here are system specifics regarding the motherboard and make.
> 
> HP Pavilion dv9205us
> Motherboard: Quanta 30B7
> Processor: AMD Turion 64 x2 3.2ghz, TL-50
> Bios: Phoenix (this is known to be buggy, but until now I had no
> problems since 2.6.24. Compiled on Windows, this bios is.)
> 
> Other than that, I have no pertinent information, at least until
> someone asks something specific. I wouldn't be contacting you all if I
> wasn't at the end of my own resources, as I am part of the Zen kernel
> community, and they can usually help out with these sorts of things.
> (zen-sources.org).
> 
> Anyway, I appreciate any help/response I can get out of this.
> 
> Thanks!
> Matt Parnell
> --
> 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/
> 
> 

Re: 2.6.27-rc0 Power Bugs with HP/Compaq Laptops

[Ingo Molnar]

> > "This is not the right fix. Instead, HPET should be used as a timer 
> > source for dynticks. This is not in 2.6.26, but I tried the 
> > linux-next tree (google it) and it looked like dynticks worked fine, 
> > without disabling C1E. Can't place my finger on the exact commit, 
> > but you could check."

Would be nice to figure out which exact change solves the problem, so 
that we could fast-route it into v2.6.27. (and for eventual .26 and .25 
backporting, if it's simple enough)

first thing would be to check whether tip/master works fine:

  http://people.redhat.com/mingo/tip.git/README

if that works fine it would be nice to check whether the pending changes 
in tip/timers/hpet solves this particular issue. That can be done by:

  git-checkout -b tmp.test tip/timers/hpet

[ once you have set up tip as a remote repository, as described in the 
  README above. ]

If tip/timers/hpet solves the problem then it's one of these five 
commits that solves it:

 f92a789: hpet: /dev/hpet - fixes and cleanup, fix
 64a76f6: hpet: /dev/hpet - fixes and cleanup
 f356181: x86: add PCI ID for 6300ESB force hpet
 3243b4e: x86: add another PCI ID for ICH6 force-hpet
 70ef6d5: x86: get irq for hpet timer

	Ingo