[libvirt] [PATCH v6 00/17] Use more PCIe less legacy PCI

Laine Stump laine at laine.org
Mon Nov 14 16:45:20 UTC 2016


On 11/14/2016 11:20 AM, Andrea Bolognani wrote:
> On Mon, 2016-11-07 at 14:50 -0500, Laine Stump wrote:
>> This is a rebase of v5 (, with the addition of patches 16 & 17, which
>> detect the type of VFIO assigned devices and assign them to either an
>> Express or a legacy PCI slot as appropriate,
> I'm afraid you're going to have to rebase again, hopefully
> it won't be too much work.
>
> I think that, seeing as all patches...
>
> [...]
>>     [RFC] qemu: if pci-bridge is in PCIe config w/o dmi-to-pci-bridge, add
>>       it
> ... except for this one, that we've decided to drop...
>
> [...]

I made a separate patch to fix the test cases this affects. It will have 
to be pushed before 12-15 can be pushed. I just posted it as a reply to 
my last message on that sub-thread.

>>     qemu: propagate virQEMUDriver object to
>>       qemuDomainDeviceCalculatePCIConnectFlags
>>     qemu: assign vfio devices to PCIe addresses when appropriate
> ... and these two, that AIUI will have to be tweaked
> anyway, have been ACKed with only trivial changes needed,
> the bulk of the series (1-15) can be pushed now.
>
> I would not hold up the entire series because of the last
> two patches as they're fairly self-contained, but up to you.
> Do you want me to look in detail at their current form (they
> looked okay from a quick look), or would you rather implement
> Alex's suggestions first and then make a separate mini-series
> out of them?

You can skip them while I try to implement Alex's suggestion. They 
should be independent of the other two series that are pending.

>
> It would also be awesome if you could try and update
> docs/news.html.in with a couple of sentences describing,
> in broad strokes, what this series changes for users,
> following the ideas outlined in last month's "Toward a
> better NEWS file" thread. You can post it as patch 18/17
> and we can iterate over it, the whole thing is still very
> much a work in progress at this point anyway.


Yeah, I'll give that a look in a bit.

Thanks!




More information about the libvir-list mailing list