On Sun, 26 March 2000, Edward Schernau wrote:
>
> I had the box get twitchy on me with the NIC and the
> HPT controller sharing an interrupt. ...
> I've got everything on its own interrupt now, and
> its stable. ...
> I use ONLY the HPT, and disable the ATA33 IDE
> controller in BIOS.
>
> CPU0 CPU1
> 0: 11176 39389 IO-APIC-edge timer
> 1: 598 2698 IO-APIC-edge keyboard
> 2: 0 0 XT-PIC cascade
> 4: 3 0 IO-APIC-edge serial
> 5: 0 0 IO-APIC-edge soundblaster
> 8: 2 0 IO-APIC-edge rtc
> 10: 6476 10406 IO-APIC-edge serial
> 11: 36 133 IO-APIC-level eth0
> 12: 1839 3615 IO-APIC-edge PS/2 Mouse
> 13: 1 0 XT-PIC fpu
> 15: 3519 3555 IO-APIC-level ide0
>
Ed (and anyone else who's tried):
Have you ever used BOTH HPT366 channels? On my system they come up sharing one IRQ.
This seems to be an HPT366 hardware limitation that can't be altered?
It looks like you're NOT using "noapic"? I have to use it or I don't even get through
boot without disk I/O errors. Did you happen to try "noapic" on your twitchy config
before un-sharing IRQs?
Depending on answers to the above, we may be able to explain the different stability
experiences different folks are having and define one principle to making the BP6
stable: "If you have active shared IRQs (including the HPT366's dual ide channels),
you have use the 'noapic' kernel option."
Also, even thought I'm only running ATA-33 (on IBM Deskstar 34GXP ATA-66-capable
drives) and using noapic, I lose DMA interrupts under simultaneous heavy load on both
HPT366 IDE channels unless I enable the HPT366 fast interrupt handling that's provided
by Andre Hedrick's 3/17 IDE patches for 2.2.15-15. (An experimental kernel option -- I
don't have the precise name handy. This patch was apparently superceded yesterday by
Andre's 3/25 patches for 2.2.15-14.)
I lose DMA interrupts when running the same test on a non-HPT ide channel and on only
one HPT channel. I think this supports your observation re. shared IRQs. (Don't have
a UDMA drive on ide0, so can't easily repeat the test on the PIIX ide channels, which
don't share an IRQ.)
With this config (noapic, latest ide patches on 2.2.15-15 kernel, and experimental
HPT366 fast interrupt support) I seem to get stable disk operation with the Bonnie
disk benchmark running simultaneously on all three of my UDMA/ATA-33 channels
(ide1/2/3).
Anybody had similar or clearly conflicting experiences?
-- Bill Clay
... still tweakin' after all these weeks ...
Bill Clay
Charlotte NC USA
[EMAIL PROTECTED]
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
