Default 2.6.29 IRQ 16/nvidia: IRQF_DISABLED is not guaranteed on shared IRQs

[interzoneuk]

Hi.

I have noticed this since using 2.6.29.1 kernel (amd64)

[ 15.617627] IRQ 16/nvidia: IRQF_DISABLED is not guaranteed on shared IRQs

here is what is using irqs:-

mrgentoo morgan # cat /proc/interrupts

CPU0 CPU1
0: 125 0 IO-APIC-edge timer
1: 2 633 IO-APIC-edge i8042
4: 0 2 IO-APIC-edge
7: 1 0 IO-APIC-edge
8: 0 0 IO-APIC-edge rtc0
9: 0 0 IO-APIC-fasteoi acpi
14: 0 0 IO-APIC-edge pata_amd
15: 0 0 IO-APIC-edge pata_amd
16: 2 3153 IO-APIC-fasteoi nvidia
18: 0 3 IO-APIC-fasteoi ohci1394
20: 94 33981 IO-APIC-fasteoi ohci_hcd:usb1
22: 37 27539 IO-APIC-fasteoi sata_nv, HDA Intel
23: 56 37136 IO-APIC-fasteoi sata_nv, ehci_hcd:usb2
25: 213 266212 PCI-MSI-edge eth0
NMI: 0 0 Non-maskable interrupts
LOC: 329291 282776 Local timer interrupts
RES: 53901 51045 Rescheduling interrupts
CAL: 1174 118 Function call interrupts
TLB: 2374 1850 TLB shootdowns
SPU: 0 0 Spurious interrupts
ERR: 1
MIS: 0

Anyone know how to prevent the message ?

[thefirstm]

Is it causing you any problems? If it is not, then just ignore it. I get that message with both 2.6.29 and 2.6.30-rc, but it doesn't cause any problems on either.

[kernelOfTruth]

if you need a "fix" for this:

ask google

(3rd hit provides one)

Re: PATCH: 2.6.26-rc8: Fix IRQF_DISABLED for shared interrupts

[StringCheesian]

Originally Posted by kernelOfTruth

if you need a "fix" for this:

ask google

(3rd hit provides one)

Re: PATCH: 2.6.26-rc8: Fix IRQF_DISABLED for shared interrupts

But that patch doesn't look like a fix. In fact, it looks like it's probably the patch that introduced the warning in the first place. Which is good, it exposed a preexisting problem between the kernel and the nvidia driver. But it's definitely not a fix.

[zander]

I don't believe the warning is indicative of a real problem in case of the NVIDIA Linux graphics driver. I'll take a closer look, though.

[kernelOfTruth]

Originally Posted by StringCheesian

But that patch doesn't look like a fix. In fact, it looks like it's probably the patch that introduced the warning in the first place. Which is good, it exposed a preexisting problem between the kernel and the nvidia driver. But it's definitely not a fix.

argh !

sorry, I of course meant the 4th

http://zen-sources.org/content/irqfs...ed-fix-2629-rc

-----Start-Copy-------

---
kernel/irq/manage.c | 12 ++++++++++++
1 file changed, 12 insertions(+)
--- a/kernel/irq/manage.c 2008-06-30 12:28:58.000000000 -0700
+++ b/kernel/irq/manage.c 2008-06-30 12:46:54.000000000 -0700
@@ -539,6 +539,18 @@ int request_irq(unsigned int irq, irq_ha
struct irqaction *action;
int retval;

+ /*
+ * handle_IRQ_event() always ignores IRQF_DISABLED except for
+ * the _first_ irqaction (sigh). That can cause oopsing, but
+ * the behavior is classified as "will not fix" so we need to
+ * start nudging drivers away from using that idiom.
+ */
+ if ((irqflags & (IRQF_SHARED|IRQF_DISABLED))
+ == (IRQF_SHARED|IRQF_DISABLED))
+ pr_warning("IRQ %d/%s: IRQF_DISABLED is not "
+ "guaranteed on shared IRQs\n",
+ irq, devname);
+
#ifdef CONFIG_LOCKDEP
/*
* Lockdep wants atomic interrupt handlers:

------END-------

[StringCheesian]

Originally Posted by zander

I don't believe the warning is indicative of a real problem in case of the NVIDIA Linux graphics driver. I'll take a closer look, though.

I just assumed that the warning was something serious. Sorry, I didn't mean to exaggerate.

[interzoneuk]

Thanks for taking a look Zander. I am getting the segfault with some 3d apps (rarely)..

I did search on google and found that exact patch but was under the impression that it was already in the kernel and that it just warned you..

I have tried the following (in case that helped)
Kernel settings:-

] Support for Interrupt Remapping (EXPERIMENTAL)
] Support for DMA Remapping Devices (EXPERIMENTAL)

- I have also installed a beta bios for my MB that enables dma remapping

Still the same

I have also tried putting it in MSI mode - NVreg_EnableMSI=1

And get similar warning although a different irq is now shared

26: 0 581 PCI-MSI-edge nvidia

cheers

[zander]

@interzoneuk: It is unlikely that an NVIDIA GPU interrupt delivery problem would manifest itself in user-space segmentation faults.

[botio]

Hi, my 6600GT AGP shares an IRQ with an USB bus.
I also get this message:
IRQ 16/nvidia: IRQF_DISABLED is not guaranteed on shared IRQs

16: 5545144 0 IO-APIC-fasteoi uhci_hcd:usb2, uhci_hcd:usb5, nvidia

My system freezed every time I used an bluetooth dongle plugged on this port, I changed the port and now it's ok. I didn't checked if it still responded to ping though

I guess there are actually issues sharing IRQs with the graphic card.