x86: Fix HPET for Fujitsu u9200

x86: Fix HPET for Fujitsu u9200

[Janne Kulmala]

HPET timer is listed in the ACPI table, but needs a quirk entry in order to
work. Unfortunately, the quirk code runs after first HPET hpet_enable() which
has already determined that the timer doesn't work (reads 0xFFFFFFFF). This
patch allows hpet_enable() to be called again after running the quirk code.

Signed-off-by: Janne Kulmala <janne.t.kulmala@tut.fi>

--- linux-2.6.28/arch/x86/kernel/quirks.c~	2008-12-16 12:58:48.000000000 +0200
+++ linux-2.6.28/arch/x86/kernel/quirks.c	2008-12-16 12:59:14.000000000 +0200
@@ -168,6 +168,8 @@
  			 ich_force_enable_hpet);
  DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH8_1,
  			 ich_force_enable_hpet);
+DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH8_4,
+			 ich_force_enable_hpet);
  DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH9_7,
  			 ich_force_enable_hpet);

--- linux-2.6.28/arch/x86/kernel/hpet.c~	2008-12-16 13:08:06.000000000 +0200
+++ linux-2.6.28/arch/x86/kernel/hpet.c	2008-12-16 13:10:29.000000000 +0200
@@ -811,7 +811,7 @@

  out_nohpet:
  	hpet_clear_mapping();
-	boot_hpet_disable = 1;
+	hpet_address = 0;
  	return 0;
  }

Re: x86: Fix HPET for Fujitsu u9200

[Ingo Molnar]

* Janne Kulmala <janne.t.kulmala@tut.fi> wrote:

> HPET timer is listed in the ACPI table, but needs a quirk entry in order 
> to work. Unfortunately, the quirk code runs after first HPET 
> hpet_enable() which has already determined that the timer doesn't work 
> (reads 0xFFFFFFFF). This patch allows hpet_enable() to be called again 
> after running the quirk code.
>
> Signed-off-by: Janne Kulmala <janne.t.kulmala@tut.fi>

applied to tip/timers/hpet, thanks Janne!

Venki, is this new hpet force-enable quirk entry generally OK:

 DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH8_1,
                         ich_force_enable_hpet);
+DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_ICH8_4,
+                        ich_force_enable_hpet);

i guess it's OK since we enable hpet on ICH8_1 already. Should we perhaps 
convert these quirks somehow to some ">= threshold_version" condition, 
instead of having to enumerate everything?

	Ingo