[Ksummit-2012-discuss] PCI breakout session

Benjamin Herrenschmidt benh at kernel.crashing.org
Mon Jun 18 20:48:32 UTC 2012


On Mon, 2012-06-18 at 14:11 +0300, Michael S. Tsirkin wrote:

> > Now if you're thinking of virtio, well, I argue that having used IO
> > space as the basic for virtio communication is a HUGE DESIGN BUG :-) I
> > know it's marginally faster to emulate that MMIO but that isn't an
> > excuse to perpetuate that historical abortion.
> 
> Unfortunately it's significantly faster on x86 :(.

How "significantly" ? How relevant in the overall performance of
virtio ? Can the ring handling be tuned to limit the need here ? Having
ring head/tails in shared memory and only using IO/MMIO when not
NAPI'ing for example ?

> > If not already (I haven't checked recently), virtio drivers should start
> > accepting an MMIO BAR (using iomap makes it trivial to not care about
> > the BAR type).
> 
> Yes this works with a Linux guest but breaks non-Linux ones.

Why would that ? Those other guests have virtio driver that can be
updated as well.

There is no excuse for sticking with IO space, really. However, again,
this mailing list isn't the right forum to discuss this :-)

Cheers,
Ben.




More information about the Ksummit-2012-discuss mailing list