Skip to contentSkip to conversation

Catching NForce2 lockup with NMI watchdog - found?

DISCUSSION Ian Kumlien 2 participants 5 messages Dec 6, 2003 – Dec 7, 2003

lore
Share & tools
lore
#0 lore raw

Hi, i'm now running this patch and it survived my grep in /usr/src.

It's mainly a correction of the apic patch and the ACPI halt disconnect patch that was originally done for 2.6...

I'll get back to you about uptime, but i think this is it...

Although i would prefer a not so workaroundish approach =)

Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
i386/kernel/mpparse.c renamed from i386/kernel/mpparse.c.orig +2 -1
@@ -1140,7 +1140,8 @@
 	 */
 	for (i = 0; i < mp_irq_entries; i++) {
 		if ((mp_irqs[i].mpc_dstapic == intsrc.mpc_dstapic) 
-			&& (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq)) {
+			&& (mp_irqs[i].mpc_srcbusirq == intsrc.mpc_srcbusirq)
+			&& (mp_irqs[i].mpc_irqtype == intsrc.mpc_irqtype)) {
 			mp_irqs[i] = intsrc;
 			found = 1;
 			break;
i386/kernel/pci-pc.c renamed from i386/kernel/pci-pc.c.orig +13 -0
@@ -1328,6 +1328,18 @@
 		dev->transparent = 1;
 }
 
+/*
+ * Halt Disconnect and Stop Grant Disconnect (bit 4 at offset 0x6F)
+ * must be disabled when APIC is used (or lockups will happen).
+ */
+static void __devinit pci_fixup_nforce2_disconnect(struct pci_dev *d)
+{
+	u8 t;
+
+	pci_read_config_byte(d, 0x6F, &t);
+	pci_write_config_byte(d, 0x6F, (t & 0xef));
+}
+
 struct pci_fixup pcibios_fixups[] = {
 	{ PCI_FIXUP_HEADER,	PCI_VENDOR_ID_INTEL,	PCI_DEVICE_ID_INTEL_82451NX,	pci_fixup_i450nx },
 	{ PCI_FIXUP_HEADER,	PCI_VENDOR_ID_INTEL,	PCI_DEVICE_ID_INTEL_82454GX,	pci_fixup_i450gx },
@@ -1343,6 +1355,7 @@
 	{ PCI_FIXUP_HEADER,	PCI_VENDOR_ID_VIA,	PCI_DEVICE_ID_VIA_8367_0,	pci_fixup_via_northbridge_bug },
 	{ PCI_FIXUP_HEADER,	PCI_VENDOR_ID_NCR,	PCI_DEVICE_ID_NCR_53C810,	pci_fixup_ncr53c810 },
 	{ PCI_FIXUP_HEADER,	PCI_VENDOR_ID_INTEL,	PCI_ANY_ID,			pci_fixup_transparent_bridge },
+	{ PCI_FIXUP_HEADER,	PCI_VENDOR_ID_NVIDIA,	PCI_DEVICE_ID_NVIDIA_NFORCE2,	pci_fixup_nforce2_disconnect },
 	{ 0 }
 };
 
#1 lore raw
Quoted 4 lines · Ian Kumlien

Hi, i'm now running this patch and it survived my grep in /usr/src.

It's mainly a correction of the apic patch and the ACPI halt disconnect patch that was originally done for 2.6...

Correction? how so? code looks the same, although the line numbers are completely different for mpparse.c and at that location there is different code. (Havent checked the disconnect)

Or do u just mean combination of the two patches?

Quoted 2 lines

I'll get back to you about uptime, but i think this is it...

Why do you think the disconnect is also related? (given some are just running the APIC patch and having (less/)no issues?

Quoted 2 lines

Although i would prefer a not so workaroundish approach =)

23 hrs now..

Craig

#2 lore raw
Quoted 9 lines · Craig Bradney
Quoted 4 lines · Ian Kumlien

Hi, i'm now running this patch and it survived my grep in /usr/src.

It's mainly a correction of the apic patch and the ACPI halt disconnect patch that was originally done for 2.6...

Correction? how so? code looks the same, although the line numbers are completely different for mpparse.c and at that location there is different code. (Havent checked the disconnect)

this is for 2.4.23 =)

Quoted 1 line

Or do u just mean combination of the two patches?

Combination + for 2.4.23

Quoted 4 lines
Quoted 1 line

I'll get back to you about uptime, but i think this is it...

Why do you think the disconnect is also related? (given some are just running the APIC patch and having (less/)no issues?

Since i had issues with just the apic patch, as stated in my mail.

Quoted 3 lines
Quoted 1 line

Although i would prefer a not so workaroundish approach =)

23 hrs now..

I'm at 5:06 and thats a record for it running with apic.

Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net
#3 lore raw
Quoted 4 lines · Ian Kumlien
Quoted 3 lines · Craig Bradney
Quoted 2 lines · Ian Kumlien

Hi, i'm now running this patch and it survived my grep in /usr/src.

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

It's mainly a correction of the apic patch and the ACPI halt disconnect patch that was originally done for 2.6...

Correction? how so? code looks the same, although the line numbers are completely different for mpparse.c and at that location there is different code. (Havent checked the disconnect)

this is for 2.4.23 =)

duh.. ok :)

Quoted 10 lines
Quoted 1 line

Or do u just mean combination of the two patches?

Combination + for 2.4.23

Quoted 4 lines
Quoted 1 line

I'll get back to you about uptime, but i think this is it...

Why do you think the disconnect is also related? (given some are just running the APIC patch and having (less/)no issues?

Since i had issues with just the apic patch, as stated in my mail.

oic.. sorry.. must have missed that.

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

Although i would prefer a not so workaroundish approach =)

23 hrs now..

I'm at 5:06 and thats a record for it running with apic.

1d 9h here now..

I remembered I also dont have preempt still, but I am happy with it if it stays like this. The next time I need to reboot (or get a hang), I'll recompile with preempt it and try again.

Craig

#4 lore raw
Quoted 9 lines · Craig Bradney
Quoted 6 lines · Ian Kumlien
Quoted 3 lines · Craig Bradney

Correction? how so? code looks the same, although the line numbers are completely different for mpparse.c and at that location there is different code. (Havent checked the disconnect)

this is for 2.4.23 =)

duh.. ok :)

Since none of my 2.6 running computers can read from the cdrom i have decided to leave this last one on 2.4 =)

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

Why do you think the disconnect is also related? (given some are just running the APIC patch and having (less/)no issues?

Since i had issues with just the apic patch, as stated in my mail.

oic.. sorry.. must have missed that.

Ok, i only got 2h uptime. But now i'm breaking the record with every second that passes...

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines
Quoted 1 line

Although i would prefer a not so workaroundish approach =)

23 hrs now..

I'm at 5:06 and thats a record for it running with apic.

1d 9h here now..

19h 33m

Quoted 3 lines

I remembered I also dont have preempt still, but I am happy with it if it stays like this. The next time I need to reboot (or get a hang), I'll recompile with preempt it and try again.

I'm more interested about getting NMI to work with io-apic right now, if possible...

Signature · 1 line
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net