Hi,
I've updated the dom0 pvops patches against 2.6.24, and they now have a
new git home:
http://git.et.redhat.com/?p=linux-2.6-dom0-pvops.git
git://et.redhat.com/linux-2.6-dom0-pvops.git
I'm maintaining the tree in this repo in two separate forms:
* Development branch:
The master branch is an ongoing, shareable development branch,
currently including a fair amount of debug code and history that
we will not ever want to push upstream
* Clean patch queues:
I'm also maintaining the same tree as a completely different set
of patches/git history --- the tag "dom0-2.6.24-A0" is a clean
patch queue applying directly against upstream v2.6.24, and the
tag "dom0-2.6.24-A0-dev" is the corresponding tag on the
master
branch which results in _exactly_ the same tree, but which has
the full development history behind it. (git-diff-tree will
show they are identical for verification.)
So for patch review and upstream pushing, the dom0-2.6.24-A0 tag is what
you want to be looking at; for development and git merging, the master
branch is where the activity will be.
New in this release:
--------------------
dom0-2.6.24-A0 adds:
Initial APIC support:
This is a minimal brute-force merge of the 2.6.18-xen APIC code,
enough to get interrupts being handled and to allow other dom0
work to proceed; this will still need to be worked into a pvops-
friendly format.
PCI DMA support:
Adds an x86_64-style dma_mapping indirection, tested on
baremetal, and adds Xen pfn-to-mfn translations to allow DMA to
function under pvops. This is fully pvops-friendly as it
stands, but does not yet support swiotlb (so for now it's for
development and testing on 32-bit <4GB boxes only.)
Various minor bugfixes.
Status
------
Boots through to initrd/module loading, but modular drivers are still
not working right, failing on init with a variety of symptoms including
fatal controller errors or init timeouts. We currently suspect problems
in the caching/writethrough settings on the PCI accesses, but that's
just a guess right now.
The full patch will NOT boot on baremetal due to the heavy-handed APIC
merge; but the dma-mapping patches applied alone DO boot on baremetal.
Still includes lots of debug printks for irq and dma mapping, don't
expect a quiet boot.
--Stephen
