Re: fwd: Mylex dac960 not SMP-safe?

From: Olli Lounela (olli@mpoli.fi)
Date: Mon Feb 12 2001 - 18:55:01 EST


On Mon, Feb 12, 2001 at 09:02:00AM -0800, Leonard N. Zubkoff wrote:
>
> I seem to recall that the Intel Providence motherboard has been
> problematic in many configurations. Have you contacted Mylex Technical
> Support to inquire about compatibility issues between the AcceleRAID 250
> and the PR440FX motherboard? They are far more expert than I when
> hardware compatibility is in question.

I have done that now, but don't really expect response before tomorrow at
earliest.

Just to mention what I earlier forgot, I've made darn sure the cooling is
good enough. The raid pack is in external box, so power quality should not
affect the issue.

> Busmastering jumper? The only jumper on the AcceleRAID 250 I recall is the one
> that controls whether the SISL support for onboard Symbios SCSI chips is
> activated. Having this jumper in the wrong position could lead to interrupt
> problems, I expect.

OK. The documentation about the meaning of SISL was not exactly the best in
the World, so to say. The onboard controller is aic7880, I didn't think it
uses Symbios chips... ohwell.

> If it hangs right after the driver banner is printed, but before the
> configuration of the controller is reported, this almost certainly means that
> the driver is not receiving interrupts from the DAC960. This generally

This is the case. With nosmp, it gets recognized and initialized all right,
the filesystems in the disk set pass e2fsck without problems, but the boot
still eventually hangs hard at eepro100 initialization once init changes to
runlevel 3 (this is RedHat, alright).

There is a difference here, from Mylex init hang I can do soft reboot, eepro
init hangs hard. And yes, I've waited half an hour to see if it recovers.
No such luck.

With 2.2-UP (stock RH 2.2.16-3) I seem to be able to run the machine with no
problems, though SMP variant of the same still hangs hard, apparently at
network traffic.

> indicates compatibility problems between the motherboard, its BIOS, and the
> DAC960 hardware/firmware, a buggy motherboard, or it can indicate that the

I've now also upgraded the flsh ROMs to the latest I found from Mylex' web
site (Firmware 4.08-0-35, BIOS 4.10-41).

> Linux kernel is not dealing properly with the motherboard APIC configuration.
> I thought that the latter type of issue is no longer likely.

This is what it sounds like, to me, if the driver works fine. Especially
since 2.2 is different from 2.4, and also since the motherboard has worked
before. Perhaps the kernel mis-initializes APIC, and interrupts go awry.

I have now tested several kernels, going all the way back to 2.4.0-test10
(I thought there might be something in common with Jan-Benedict Glaw's
report about PCI problems with DAC1164). No difference at all with any of
the kernels.

However, I still have no real idea how to start debugging the case. I sure
hope it isn't undebuggable.

> eepro100, so I am initially much more suspicious of the PR440FX motherboard
> combination.

My feelings exactly. I suspect SMP PPro is no longer all that common, and
Mylex there might be much less so. eepro100/B with 82557 should not be
uncommon, however. Strange, however, that it's the dac960 that exposes the
problem.

> I notice that the AcceleRAID 250 and eepro100 are sharing IRQ 10. WHile in
> theory this should work fine, and does in other systems, is it possible to
> place the AcceleRAID 250 in a different slot or assign it another IRQ?

Forcing these to different IRQ's made no difference, nor did shuffling
around the PCI-cards (just two, S3 Virge and Mylex).

Finally got 2.4.2pre3 nosmp boot dmesg, attached. Also current lspci -vvvxx

-- 
    Olli               ...and he thought I'm serious! Hahahaha...



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



This archive was generated by hypermail 2b29 : Thu Feb 15 2001 - 21:00:20 EST