Promise's PCI DMA ATA interface card support of RH7.2 IA-64

"Murakami Hiroshi" <[email protected]> Fri, 1 Nov 2002 01:47:36 +0900 (JST)
Newsgroups gmane.linux.redhat.ia64.general
Message-ID <[email protected]>
This is revised report of the article 
  "Software RAID function of RH7.2 is not yet reliable".
It is not the software RAID that has trouble, but it is the
device driver(I guess) of the Promise PCI DMA ATA interface card.

Promise Ultra66 PCI DMA ATA interface card, 
Promise Ultra100 PCI DMA ATA interface card can be used 
on my IA-64 machine with 4GB of memory.

The card has two IDE cable sockets and the drive device
correspondence are:
	/dev/hde (Primary Master)
	/dev/hdf (Primary Slave)
	/dev/hdg (Secondary Master)
	/dev/hdh (Secondary Slave)

RedHat7.2 IA-64, I found there is a fatal trouble with the
above Promise cards. First, I have to choose the "good"
BIOS data with these cards. (I was not able to boot the
OS with Promise Ultra 133 TX2 on RH7.2-IA64 using any revision.)

The trouble is, if I connect two IDE disks to both cable as masters
/dev/hde and /dev/hdg, then the OS bangs and freeze
if both drives are accessed simultaneously:
----
dd if=/dev/hde of=/dev/null & dd if=/dev/hdg of=/dev/null 
# wait for while will eventually drop the entire system.
---
If two HDDs are connected to single cable as master and slave
i.e. /dev/hde and /dev/hdf, then:
----
dd if=/dev/hde of=/dev/null & dd if=/dev/hdg of=/dev/null 
# this will not drop the system and finishes normally.
---

On IA-32 RH8.0, I tried the same Promise card and HDDs,
(even Promise Ultra133TX2 can be used ok.) this fatal trouble
never happens.

I guess the device driver of the Promise card for RH7.2 IA64 has
some bug when two IDE bus of the card is used.

H.M.