2.4.19-rc3-ac2 I2O (Promise SX6000) funkiness

2.4.19-rc3-ac2 I2O (Promise SX6000) funkiness

[Stuffed Crust]

[-- Attachment #1: Type: text/plain, Size: 3621 bytes --]

I'm trying to use a Promise SX6000 board.  This one has a 128 meg stick
on it, and five 120 gig drives in a RAID5 configuration.

The array is created and initialized fine (save the 12 hours or so of
incessant beeping by the card..)

I fire up the i2o_pci and i2o_block modules and the card and its array
are detected:

	I2O Core - (C) Copyright 1999 Red Hat Software
	I2O: Event thread created as pid 9
	I2O configuration manager v 0.04.
	  (C) Copyright 1999 Red Hat Software
	Linux I2O PCI support (c) 1999 Red Hat Software.
	i2o: Checking for PCI I2O controllers...
	PCI: Found IRQ 10 for device 00:11.1
	PCI: Sharing IRQ 10 with 00:07.2
	i2o: I2O controller on bus 0 at 137.
	i2o: PCI I2O controller at 0xDA000000 size=4194304
	I2O: Promise workarounds activated.
	I2O: MTRR workaround for Intel i960 processor
	i2o/iop0: Installed at IRQ10
	i2o: 1 I2O controller found and installed.
	Activating I2O controllers...
	This may take a few minutes if there are many devices
	i2o/iop0: Reset rejected, trying to clear
	i2o/iop0: LCT has 9 entries.
	i2o/iop0: Configuration dialog desired.
	I2O Block Storage OSM v0.9
	   (c) Copyright 1999-2001 Red Hat Software.
	i2o_block: registered device at major 80   
	i2o_block: Checking for Boot device...
	i2o_block: Checking for I2O Block devices...
	i2ob: Installing tid 11 device at unit 0   
	i2o/hda: Max segments 28, queue depth 8, byte limit 49152.
	i2o/hda: Type 68: 468987MB, 512 byte sectors.
	i2o/hda: Maximum sectors/read set to 96.   
	 i2o/hda: i2o/hda1

So now it's time to create a filesystem.

	mke2fs -j -m -0 /dev/i2o/hda1

It starts out fine, then the console starts spewing these messages after
about 30/2664 inode tables.

	i2o/iop0: No handler for event (0x00000020)
	i2o/iop0: No handler for event (0x00000020)
	i2o/iop0: No handler for event (0x00000020)
	i2o/iop0: No handler for event (0x00000020)
	i2o/iop0: No handler for event (0x00000020)
	...

After I break out of mke2fs, the errors continue for a while, then
change to:

	i2o/iop0: Hardware Failure: Unknown Error  
	i2o/iop0: Hardware Failure: Unknown Error  
	i2o/iop0: Hardware Failure: Unknown Error  
	...
	i2o_post_wait event completed after timeout.
	i2o_post_wait event completed after timeout.
	i2o_post_wait event completed after timeout.
	...
	i2ob_release: controller rejected unclaim.

This happens with 2.4.19rc2aa1 SMP, 2.4.19rc3aa1 SMP.
It also happens with 2.4.19rc3ac2 SMP (with noapic+nosmp, as I have the
endless APIC Error bug with that kernel) 
And finally, it happens on the "stock" RedHat 2.4.18-5 Uniprocessor kernel.

So, can anyone tell me what's going on here?  Is the controller on the
fritz?  Is the host PC all screwed up?  (memtest86 says the RAM is
likely fine; though the SX6000 provides no means for me to verify it 
on the card itself) 

FWIW, this is a newer SX6000 with PDC20276 chips on it, and it's sharing
IRQ10 with the usb controller.  Disabling it makes no difference.

Further details can be provided upon request...

I can move the controller to another machine this weekend for further
testing.  But in the mean time, Just what is event 0x00000020?

I suppose I can call up Promise, but as their driver won't even compile,
I don't have much faith in their ability to do anything but shave a
couple of hours off of my life.

 - Pizza
-- 
Solomon Peachy                                   pizza@f*cktheusers.org
I'm not broke, but I'm badly bent.                         ICQ #1318344
Patience comes to those who wait.                         Melbourne, FL

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: 2.4.19-rc3-ac2 I2O (Promise SX6000) funkiness

[Alan Cox]

On Thu, 2002-07-25 at 04:43, Stuffed Crust wrote:
> It starts out fine, then the console starts spewing these messages after
> about 30/2664 inode tables.
> 
> 	i2o/iop0: No handler for event (0x00000020)
> 	i2o/iop0: No handler for event (0x00000020)
> 	i2o/iop0: No handler for event (0x00000020)
> 	i2o/iop0: No handler for event (0x00000020)
> 	i2o/iop0: No handler for event (0x00000020)

Its spewing events we didnt ask for then it dies

> 
> 	i2o/iop0: Hardware Failure: Unknown Error  
> 	i2o/iop0: Hardware Failure: Unknown Error  
> 	i2o/iop0: Hardware Failure: Unknown Error  

That says it all. Thats a message from the controller on the SX6000
saying "I broke".

> FWIW, this is a newer SX6000 with PDC20276 chips on it, and it's sharing
> IRQ10 with the usb controller.  Disabling it makes no difference.

You will need at list 2.4.19ac from a cold boot for this card to work.
Also check your boot logs to ensure it skipped over the PDC20276
controllers correctly. I think the code is right but I am not yet sure.

If we don't skip the 20276 chips properly then we reconfigure them and
the raid card has kittens (rightly so)

> Further details can be provided upon request...
> 
> I can move the controller to another machine this weekend for further
> testing.  But in the mean time, Just what is event 0x00000020?

Its an invalid I2O event code. Promise cards emit them before crashing
when you ask them to do anything they take offence too.

> I suppose I can call up Promise, but as their driver won't even compile,
> I don't have much faith in their ability to do anything but shave a
> couple of hours off of my life.

Actually the promise provided driver for the supertrak cards seems to
work fine. I actually used it to debug the generic i2o_block support for
their chips. It won't help if we are eating the ide controllers however