Re: [PATCH] Replace nvidia timer override quirk with pci id list

[Andi Kleen]

On Fri, Feb 08, 2008 at 04:13:35PM +0100, Prakash Punnoor wrote:
> On the day of Friday 08 February 2008 Andi Kleen hast written:
> > On Thu, Feb 07, 2008 at 10:21:18PM +0100, Prakash Punnoor wrote:
> > > On the day of Thursday 07 February 2008 Andi Kleen hast written:
> > > > Replace the old "for all of nvidia" quirk with a quirk containing pci
> > > > device ID. I goobled this list together from pci.ids and googling and
> > > > it may be incomplete, but so far I haven't had complaints.
> > > >
> > > > +	QBRIDGE(PCI_VENDOR_ID_NVIDIA, 0x02f0, nvidia_timer), /* mcp 51/nf4 ?
> > > > */
> > >
> > > If you want to skip timer override on this board, this is a *NAK* from
> > > me. I told you the last time, it only works reliably here on MCP51 with
> > > timer
> >
> > Hmm, if you told me it got lost somewhere, sorry.
> 
> I at least found this post: http://lkml.org/lkml/2006/8/13/2 though I remeber 
> we had some discussions. ;-)
> 
> >
> > > override working. Even before Asus released a bios which had an option to
> > > enable the hpet, I needed the override or I got irratic behaviour. Since
> > > I got hpet enabled I gave up on arguing as the wrongly triggered quirk
> > > didn't bug me anymore.
> >
> > Ok we can keep the HPET check if that makes you more happy.
> >
> > > IIRC my nforce2 needed the override. I didn't see that in the list.
> >
> > The list only contains IDs where the override should be ignored; so
> > if it has a correct one and it's not there everything is fine.
> 
> Sorry, I meant the opposite. I needed the acpi_skip_timer_override kernel 
> parameter for nforce2, thus no override. So this chipset is missing here. At 
> least I remember that my nforce2 needed the skipping, 

I hope you remember correctly and mean it this time. It would be better
if you could double check.

I'm a little sceptical because we had this patch in OpenSUSE 10.3
and I didn't think there were complaints from NF2 users.
With the changes you're requesting it turns from something 
very well tested into something experimental.

But NF2 should not need a timer override anyways so probably
ignoring it there is ok.

Actually checking CK804 is already an Nforce2, but you might
have NF2S which has a different ID. Do you have full lspci/lspci -n
output?

Ok I'm appending another patch that adds the NF2S too, can
you please test it on that machine?

> 
> > I'm appending a revised patch. Does it work for you?
> 
> I haven't tested it, but it would "work" as it would bail out in my case 

Can you please test it?

> because of the hpet check. The problem I see with this approach - as with the 
> old one -  it simply wants to ignore the override for a whole bunch of 
> chipsets. (The old one is catastrophic as it even doesn't care for chipset 
> revision.)  And checking for hpet is just heuristics (or what is the 
> rationale behind it?) not a real check whether the override should be ignored 

It was a heuristic originally to detect the NF5 which did need
the override. That is why I first removed it because it should
have been obsolete.

> or not. Are you actually sure that so many nforceX boards have broken bioses? 

Yes, it was a problem in the Nvidia reference BIOS that they sent to OEMs
to base their own BIOS on, so pretty much everybody had this problem.

We went over this with Nvidia engineers with a fine comb at this 
point. If you search the mailing list archives you might even
find the discussions.

-Andi7

---

Replace nvidia timer override quirk with pci id list v3

[This patch was originally in the old ff tree and was intended for .24; but
somehow got lost in the arch merge. Has also shipped with OpenSUSE 10.3.
I think it should go into .25] 

This replaces the old NF3/NF4 reference BIOS timer override quirk with a device 
ID list. We need to ignore the timer override on these systems, but not 
ignore it on NF5 based systems. Previously this was distingushed by checking
for HPET, but a lot of BIOS vendors didn't enable HPET in their pre Vista 
BIOSes.

Replace the old "for all of nvidia" quirk with a quirk containing pci device
ID. I goobled this list together from pci.ids and googling and it may be 
incomplete, but so far I haven't had complaints.

I also straightened out the ifdef jungle a bit.

v1->v2: Readd the HPET check to handle a NF4 system of Prakash Punnoor.
	This means with HPET we always assume timer overrides are ok.
v2->v3: Match any IDs, not only bridges again.
	Add NF2S ID

Cc: lenb@xxxxxxxxxx

Signed-off-by: Andi Kleen <ak@xxxxxxx>

---
 arch/x86/kernel/early-quirks.c |   53 ++++++++++++++++++++++-------------------
 1 file changed, 29 insertions(+), 24 deletions(-)

Index: linux/arch/x86/kernel/early-quirks.c
===================================================================
--- linux.orig/arch/x86/kernel/early-quirks.c
+++ linux/arch/x86/kernel/early-quirks.c
@@ -67,37 +67,30 @@ static int __init nvidia_hpet_check(stru
 {
 	return 0;
 }
-#endif /* CONFIG_X86_IO_APIC */
-#endif /* CONFIG_ACPI */
 
-static void __init nvidia_bugs(int num, int slot, int func)
+static void __init nvidia_timer(int num, int slot, int func)
 {
-#ifdef CONFIG_ACPI
-#ifdef CONFIG_X86_IO_APIC
-	/*
-	 * All timer overrides on Nvidia are
-	 * wrong unless HPET is enabled.
-	 * Unfortunately that's not true on many Asus boards.
-	 * We don't know yet how to detect this automatically, but
-	 * at least allow a command line override.
-	 */
 	if (acpi_use_timer_override)
 		return;
 
-	if (acpi_table_parse(ACPI_SIG_HPET, nvidia_hpet_check)) {
-		acpi_skip_timer_override = 1;
-		printk(KERN_INFO "Nvidia board "
-		       "detected. Ignoring ACPI "
-		       "timer override.\n");
-		printk(KERN_INFO "If you got timer trouble "
-			"try acpi_use_timer_override\n");
+	/* When the board has HPET assume timer override is ok */
+	if (!acpi_table_parse(ACPI_SIG_HPET, nvidia_hpet_check)) {
+		printk(KERN_INFO "Pre Nforce5 Nvidia board detected with HPET.\n"
+		       KERN_INFO "Assuming timer override is ok. If you have trouble\n"
+		       KERN_INFO "please try booting with acpi_skip_timer_override\n");
+		return;
 	}
-#endif
-#endif
-	/* RED-PEN skip them on mptables too? */
 
+	acpi_skip_timer_override = 1;
+	printk(KERN_INFO "Pre NForce5 Nvidia board "
+	       "detected. Ignoring ACPI timer override.\n");
+	printk(KERN_INFO "If you got timer trouble "
+		"try acpi_use_timer_override\n");
 }
 
+#endif
+#endif
+
 static void __init ati_bugs(int num, int slot, int func)
 {
 #ifdef CONFIG_X86_IO_APIC
@@ -121,9 +114,22 @@ struct chipset {
 	void (*f)(int num, int slot, int func);
 };
 
+#define QID(vendor, device, func) { \
+	vendor, device, PCI_ANY_ID, PCI_ANY_ID, \
+	QFLAG_APPLY_ONCE, func }
+
 static struct chipset early_qrk[] __initdata = {
-	{ PCI_VENDOR_ID_NVIDIA, PCI_ANY_ID,
-	  PCI_CLASS_BRIDGE_PCI, PCI_ANY_ID, QFLAG_APPLY_ONCE, nvidia_bugs },
+#if defined(CONFIG_ACPI) && defined(CONFIG_X86_IO_APIC)
+	/* This list should cover at least one PCI ID from each NF3 or NF4
+	   mainboard to handle bogus timer overrides in their reference BIOS. */
+	QID(PCI_VENDOR_ID_NVIDIA, 0x00e1, nvidia_timer), /* nforce 3 */
+	QID(PCI_VENDOR_ID_NVIDIA, 0x00ed, nvidia_timer), /* nforce 3 */
+	QID(PCI_VENDOR_ID_NVIDIA, 0x003d, nvidia_timer), /* mcp 04 ?? */
+	QID(PCI_VENDOR_ID_NVIDIA, 0x02f0, nvidia_timer), /* mcp 51/nf4 ? */
+	QID(PCI_VENDOR_ID_NVIDIA, 0x026f, nvidia_timer), /* mcp 51/nf4 ? */
+	QID(PCI_VENDOR_ID_NVIDIA, 0x005c, nvidia_timer), /* ck 804 */
+	QID(PCI_VENDOR_ID_NVIDIA, 0x0084, nvidia_timer), /* nf2s */
+#endif
 	{ PCI_VENDOR_ID_VIA, PCI_ANY_ID,
 	  PCI_CLASS_BRIDGE_PCI, PCI_ANY_ID, QFLAG_APPLY_ONCE, via_bugs },
 	{ PCI_VENDOR_ID_ATI, PCI_ANY_ID,

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/