This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


Re: [Xen-devel] issues getting more than 16M ram to be used without oops

To: Keir Fraser <Keir.Fraser@xxxxxxxxxxxx>
Subject: Re: [Xen-devel] issues getting more than 16M ram to be used without oopsing. 1.2 and 1.3-unstable
From: Keir Fraser <Keir.Fraser@xxxxxxxxxxxx>
Date: Tue, 06 Apr 2004 10:56:06 +0100
Cc: "Brian Wolfe" <brianw@xxxxxxxxxxxx>, xen-devel@xxxxxxxxxxxxxxxxxxxxx
Delivery-date: Tue, 06 Apr 2004 11:01:25 +0100
Envelope-to: steven.hand@xxxxxxxxxxxx
In-reply-to: Your message of "Tue, 06 Apr 2004 08:11:13 BST." <E1BAkjp-0000QF-00@xxxxxxxxxxxxxxxxx>
List-archive: <http://sourceforge.net/mailarchive/forum.php?forum=xen-devel>
List-help: <mailto:xen-devel-request@lists.sourceforge.net?subject=help>
List-id: List for Xen developers <xen-devel.lists.sourceforge.net>
List-post: <mailto:xen-devel@lists.sourceforge.net>
List-subscribe: <https://lists.sourceforge.net/lists/listinfo/xen-devel>, <mailto:xen-devel-request@lists.sourceforge.net?subject=subscribe>
List-unsubscribe: <https://lists.sourceforge.net/lists/listinfo/xen-devel>, <mailto:xen-devel-request@lists.sourceforge.net?subject=unsubscribe>
Sender: xen-devel-admin@xxxxxxxxxxxxxxxxxxxxx
> > Heh, I know what you mean on the VIA boards. I almost always have an issue
> > booting > 3 month old kernels on them. 8-P I spose I could look at via as
> > a "enforced upgrade" system. ;)
> > 
> > *nod* I'd tried the noacpi, ignorebiostables, etc. All of the usual tricks
> > on the xenolinux.gz kernel image as well to no avail pror to harassing
> > yall. :) I think my friend Adam was feeling a bit abused as a Xen suport
> > route as well. *grin*
> > 
> > Things are getting a LITTLE bit futher on the bootup. :) Any chance the
> > IDE , USB, and/or ethernet interfaces are doign something funky?
> Looks like the Ethernet interface might be doing bad things. Both
> crashes are at the same point in its interrupt handler. I'm not sure
> whether this is a cause or merely a symptom though.
> I'll take a look. Meanwhile, can you put a different card in that
> machine? For example, we know that 3com 3c905's are good. 3c595 is a
> relic. :-)

Okay, I think I've found and fixed the problem. I've made pre-built
images available at:

The problem is that the 3c595 is not using DMA, but is instead using
PIO (polled I/O). This is doubly bad in Xen:
 1. The main CPU is responsibel for transferring all data to/from slow
 on-card memory.
 2. Xen needs to temporarily map the buffer into its address space to
 execute the transfer.

It was the latter which was not being done -- Xen has a one-to-one
mapping of onyl the first 40MB of physical memory. If the buffer
location is any higher than that then we ended up copying received
packets to a random location!

So, if you're looking for decent performance then you want to get
yourself a good network card -- eg. 3c905.

I've checked the probable fix into both 1.2 and 1.3 trees.

 -- Keir

This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
Xen-devel mailing list

<Prev in Thread] Current Thread [Next in Thread>