Anthony Liguori wrote:
Avi Kivity wrote:
On 06/15/2009 04:23 PM, Anthony Liguori wrote:
How would qemu know which slots to optimize for?
In practice, I don't see that as a real problem. We should (a) add
an ioapic and four more pci links (b) recommend that slots be
assigned in ascending order, and everything works.
I don't see your concern about libvirt allocating slots. If a human
can plug a card into a slot, so can libvirt. Doing an interactive
back-and-forth (equivalent to plugging a card while blindfolded, then
looking to see which slot we hit) is certainly more difficult.
Let's take a concrete example because I think you missed my point.
For the r2d board, if you have 1 on-board NIC, it has to go in slot
2. Additional NICs can go in any slot, but the primary on-board NIC
is expected to live in slot 2. It's possible to not have that
on-board NIC.
Libvirt does not support r2d. I hope it won't start to support it.
We can have default values for these types of devices or something like
pci_addr=auto.
If you let QEMU allocate which PCI slot a device goes in, we can hide
this detail from libvirt. If you have libvirt do PCI slot allocation
by default, it has to know about this restriction in the r2d board
unless you have a clever way to express this sort of information.
Once QEMU has allocated a device to a slot, libvirt can do a good job
maintaining that relationship.
The end user should have a mechanism to control device slot positioning.
For example, if you have several pci devices, some
get high rate of interrupts and some not, if you want to optimize you
guest you should isolate the high rate 'interesting' devices.
This is something the user will need to do. I agree that the default
behavior might be 'auto'
Also, while moving from one qemu version to another, you'll need to
represent the older behavior. -qemu-0.10 is not good enough
since there will be multiple versions in the future with multiple
distributions setting their defaults.
Regards,
Anthony Liguori
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
