Re: [Xen-devel] [PATCH] hvmloader / qemu-xen: Getting rid of resource conflict for OpRegion.

[Ross Philipson <Ross.Philipson@xxxxxxxxxx>]

> -----Original Message-----
> From: firemeteor.guo@xxxxxxxxx [mailto:firemeteor.guo@xxxxxxxxx] On
> Behalf Of G.R.
> Sent: Thursday, January 10, 2013 11:29 AM
> To: Ross Philipson
> Cc: Ian Campbell; xen-devel; Keir (Xen.org); Jean.guyader@xxxxxxxxx;
> Stefano Stabellini
> Subject: Re: [Xen-devel] [PATCH] hvmloader / qemu-xen: Getting rid of
> resource conflict for OpRegion.
> 
> >> >> My impression from Ian's feedback is that the patch cannot be
> >> accepted
> >> >> before the concern being resolved.
> >> >> (Let alone the existing code in the tree has already open a hole
> >> >> (first 18 bytes in my example))
> >> >> And you mentioned that such info can be obtained from my host
> bios.
> >> >> I've dumped the ACPI table from my host system and had it
> >> disassembled
> >> >> by iasl.
> >> >> But I lack of the knowledge to interpret the content.
> >> >> Could you lend me a hand in case that's a trivial task for you?
> >> >> You can find the DSDT.dsl file attached in my previous mail
> (@2012-
> >> 12-
> >> >> 23).
> >> >
> >> > Sure I can take a look. Can you send me a dump of your e820 map on
> the
> >> > system in question (and an lspci dump while you are at it)? Also I
> >> think
> >> > you may have mentioned it but what base address is the ASLS
> register
> >> > reporting?
> >>
> >> First of all, are you asking for the guest or the host? Host, I
> guess?
> >> I'm not sure how to dump the e820 map.
> >> Previously I used acpidump to get acpitables and use iasl to
> disassemble
> >> them.
> >> Are you talking about the same work flow?
> >
> > Host. The easiest way to get it is if you can get serial output from
> Xen
> > which should report it early on. Or your kernel may report it; mine
> does
> > and I can get it via dmesg (traced as "Xen-provided memory map").
> >
> > For PCI I just wanted the output from lspci (e.g. lspci -v and lspci -
> xxx).
> >
> 
> This shoud be the OpRegion base: cd996018 is the host address.
> igd_write_opregion: Map OpRegion: cd996018 -> feff4018
> 
> Other info are available in the attachment, including pci info && the
> xl dmesg log (for e820 map).

Well unfortunately not much else in that ACPI NVS region is defined. There
is the GNVS region that is defined but it is tiny and at the tail end of
it. So it is not clear what else is in that region or around the IGD
region (frankly it could be unused or contain other bits not related the
graphics at all). Some of the values in GNVS could be pointers to other
regions in that NVS too.

I guess the only truly 100% secure way to deal with it is to trap to qemu
(as was suggested) and block access to the area around IGD. It didn't
sound like this was a popular option. Anyway sorry I could not be
more helpful but it looks like your patch to map the extra page made
it in unstable.

> 
> Thanks,
> Timothy

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel