Non-maskable interrupt

Non-maskable interrupt

Post by rogerc.. » Tue, 11 May 1999 04:00:00



Does the PowerPC 750 have a non-maskable interrupt pin?  Can anyone
attest to being completely non-maskable?

Same question for the PowerQuic 860.

Thanks.
R.

--== Sent via Deja.com http://www.deja.com/ ==--
---Share what you know. Learn what you don't.---

 
 
 

Non-maskable interrupt

Post by Richard Hendrick » Tue, 11 May 1999 04:00:00



> Does the PowerPC 750 have a non-maskable interrupt pin?  Can anyone
> attest to being completely non-maskable?

Dunno.

Quote:> Same question for the PowerQuic 860.

The 8xx family has an NMI pin, erroneously referred to as
IRQ0~ (I really want to slap the person who put that in our manuals,
btw).  IRQ0~ is totally non-maskable, and it will always cause the core
to jump to the system reset interrupt.  Since it is totally
non-maskable, it is not safe to use for a "fast" interrupt, since it
could occur during a NRI time.
Quote:> Thanks.
> R.

> --== Sent via Deja.com http://www.deja.com/ ==--
> ---Share what you know. Learn what you don't.---


 
 
 

Non-maskable interrupt

Post by Andrew Klossn » Tue, 11 May 1999 04:00:00


Quote:> Does the PowerPC 750 have a non-maskable interrupt pin?  Can anyone
> attest to being completely non-maskable?

NMIs and RISC machines don't get along well together.  If you're deep
in the context saving routine of a normal interrupt and an NMI
happens, your context is overlayed and you can't return from the
original interrupt.  CISC machines save interrupt context in hardware,
with the disadvantage of eliminating very lightweight interrupt handlers.

The 750 NMI is the SRESET pin.  When using it, consider HID0[NHR] and
MSR[RI].


 
 
 

Non-maskable interrupt

Post by Tim Ols » Tue, 11 May 1999 04:00:00



| >
| > Does the PowerPC 750 have a non-maskable interrupt pin?  Can anyone
| > attest to being completely non-maskable?

No, the 750 does not have a non-maskable interrupt.  Few RISC processors
do, because interrupt state save/restore is usually non-atomic --
interrupt state is squirreled away in registers which must be saved in
memory via interrupt handler code before another interrupt can be handled
in a recoverable manner.

If you are worried about interrupts being disabled for long periods of
time, you  should investigate a scheme for lightweight interrupt handling;
using the "first level interrupt handler" only to queue and acknowlege the
hardware interrupt, then re-enable interrupts and let a second-level
interrupt handler do the real work.

--

     -- Tim Olson

 
 
 

Non-maskable interrupt

Post by Ola Liljedah » Wed, 12 May 1999 04:00:00



> > Does the PowerPC 750 have a non-maskable interrupt pin?  Can anyone
> > attest to being completely non-maskable?

> NMIs and RISC machines don't get along well together.  If you're deep
> in the context saving routine of a normal interrupt and an NMI
> happens, your context is overlayed and you can't return from the
> original interrupt.  CISC machines save interrupt context in hardware,
> with the disadvantage of eliminating very lightweight interrupt handlers.

68K only saves PC and SR on the (interrupt) stack when handling an interrupt.
Pretty lightweight except for the memory accesses.

In the PowerPC the only context saved is the PC and MSR. For normal
exceptions/interrupts they are saved to SRR0/1 and have to be restored
to those registers before returning from exception.

The PowerPC 403 (and probably 401 and 405 also) have a critical interrupt
that saves the PC and MSR in SRR2/3. Thus the critical interrupt can nest
with the normal interrupts. There is a special return from critical interrupt.

Critical interrupts can however be masked in the MSR. It is not a non-maskable
interrupt but it can be used as a non-maskable interrupt with the proper
design rules.
--
        Ola Liljedahl

 
 
 

1. : Knowledge about non-maskable interrupts (NMI)

Hi folks,

I have got a Compaq Presario 1215 Laptop and Linux Redhat 6.0 installed.
All works beautifully except when I hit one of the special access buttons
like Volume of the speakers or play/stop of the CD player. Then I get error
messages on the screen like:

Uhhuh, NMI received for unknown reason 28
Dazed and confused, but trying to continue
Do you have a strange power saving mode enabled?

These messages pop up every second and the harddisk is being accessed at
the same time. Sometimes it is also an "unknown reason 38" instead of "28".

Is there any way to "mask" these hardware interrupt by brute force?

Thanks in advance - David.

2. Bug or feature in regcmp(3g)?

3. Non Maskable Interrupts

4. c_runtime_startup problem

5. Is it possible to register Interrupts in device-driver's in non-init entry point

6. Keeping a child process alive when parent dies

7. patchadd fails on non-global non-sparse non-running zones

8. Madge TokenRing Support?

9. X Behavior - non interrupt driven

10. netif_rx from non interrupt context

11. Interrupt registration in non-init context

12. How to run NON-ATAPI Mitsumi CDROM without interrupts ?

13. NEx000 and 1.2.9: Interrupted while interrupts are masked!