Il 04/07/2014 09:23, Ming Lei ha scritto:
I think it is good and better to go to 2.1, and it should save lots of
write syscall.
Also should regression be caused, the per thread trick may be
resorted to, which should be simple.
If we have the "right" solution (which we do, unlike the plug/unplug
case), and the benefit is there but limited, I don't see a reason to
rush another patch in 2.1.
Some reasonable level of performance degradation or increased host CPU
utilization was expected in 2.1; of course 40% is not reasonable.
With multi virtqueue's coming for virtio-blk, it should save more, and I
also plan to use the improved qemu bh to help merge requests from
multi queue, without introducing extra notifier.
But virtio-blk multiqueue is 2.2 material, and so is coalescing of irqfd
writes. I think Kevin or Stefan should queue this patch (with the
smp_mb optimization, IMHO) for block-next.
Paolo
