Re: When you failed to boot with your 2GB+ memory system from ASC 29160

On Mon, 23 Sep 2002, Murakami Hiroshi wrote:

>Date: Mon, 23 Sep 2002 11:27:22 +0900 (JST)
>From: Murakami Hiroshi <[email protected]>
>To: [email protected]
>Content-Type: text/plain; charset=US-ASCII
>List-Id: Discussion of Linux on the Intel IA-64 processor
>    <ia64-list.redhat.com>
>Subject: When you failed to boot with your 2GB+ memory system from ASC
>    29160
>
>
>  I have got solved a trouble with the Itanium system.
>The trouble is, when the memory is up to 2GB
>(512KB sdram in 4 slots simultaneously), the
>any Linux IA-64 OS boots fine.
>When the physical memory is increased beyond 2GB,
>for example 2.5GB or 4GB, then suddenly the
>the boot partition on the SCSI HDD is not found
>is happen and no way to boot from HDD by any mean.
>EFI SHELL's map info tells as if the vfat partition
>is not present. If the memory is reduced to 2GB or 512MB
>in total, then the map information tells there is vfat
>partition. I have been puzzled with this trouble for 3 weeks.
>When the total memory size is upto 2GB it boots fine and
>2.5GB or 4GB, it lose the vfat boot partion.
>#
>Finally I got the true solution. The PCI SCSI adaptor card
>the Adaptec ASC 29160, the card's BIOS revision was newest
>3.10.0, which had been the the source of this trouble.
>  I flushed-up the card's BIOS to revision 2.57.2, then this
>bad behavior completely went away. And it boots happily for ever...
>
>PS.
>Adaptec 2940UW PCI card, with my rom revision 1.32 did not
>work fine with my Itanium box to cause the red screen with
>the register dump before it tries to boot, fixed with the
>newest rom revision 2.20.0 after flushed up and it can boot
>the SCSI HDD also. It is only Ultra SCSI (40MB/S BUS) 
>compared with 29160's Ultra 160 SCSI (160MB/S SCSI BUS).

Sounds like some weirdo bug related to memory holes.  There have 
been various XFree86 bug reports of machines MCA'ing when memory 
is added or removed from the machines.  So far they've turned out 
to be weirdo bugs due ot memory holes.


-- 
Mike A. Harris		ftp://people.redhat.com/mharris
OS Systems Engineer
XFree86 maintainer
Red Hat Inc.