[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [PATCH V3 0/6] virtio 1.0 pci optimizations and fixes
From: |
Jason Wang |
Subject: |
[Qemu-devel] [PATCH V3 0/6] virtio 1.0 pci optimizations and fixes |
Date: |
Fri, 6 Nov 2015 16:02:43 +0800 |
Hi all:
This series tries to fix the following issues:
- qemu abort when trying to adjust endianness for zero length eventfd,
this prevent fast mmio eventfd from being used in ppc. Fixing by
skip the endianness adjustment for zero length eventfd.
- 1.0 mmio is slow since it was using datamatch eventfd. Fixing this
by usinng zero length mmio eventfd, then we could get speed up through
kernel fast mmio bus on ept capable machine.
- 1.0 mmio is slow compared to pio (at least on some
archs/setups). Fixing this by re-introducing pio notification
capability. This will be useful for the arch/setups that fast mmio
does not work.
- Some virtio pci 1.0 fields were not migrated. This will cause
unexpected behaviour if migrate during driver initialization. Fixing
this by introduce a transport specific callback and get/put
transport specific fields for virtio 1.0.
- queue_enable read was broken. Fixing by set the queue_enable to true
during guest write and clear it during reset.
Please review.
Thanks
Changes from V2:
- rebase to master
- conditionally use zero length io eventfd based on new kvm cap
- replace "modern_state" with "extra_state" in patch 1
- check for the existence of k->load_extra_state in patch 1
Changes from V1:
- skip zero length eventfd endianness adjustment
- don't use pci specific name ("modern") in virtio core, using "extra"
instead and in virtio pci callback, using subsections which could
allow us to extend the future improvement without changing the core.
- don't check virtio_virtqueue_needed() in virtio_extra_state_needed()
- drop the ppc 2.5 machine type patch
- squash Eduardo's 2.5 machine type patches into this series
Jason Wang (6):
virtio-pci: fix 1.0 virtqueue migration
memory: don't try to adjust endianness for zero length eventfd
KVM: add support for any length io eventfd
virtio-pci: use zero length mmio eventfd for 1.0 notification cap when
possible
virtio-pci: introduce pio notification capability for modern device
virtio-pci: unbreak queue_enable read
hw/virtio/virtio-pci.c | 277 +++++++++++++++++++++++++++++++++++++----
hw/virtio/virtio-pci.h | 28 +++--
hw/virtio/virtio.c | 58 +++++++++
include/hw/compat.h | 4 +
include/hw/virtio/virtio-bus.h | 3 +
include/sysemu/kvm.h | 8 ++
kvm-all.c | 4 +
kvm-stub.c | 1 +
memory.c | 8 +-
9 files changed, 361 insertions(+), 30 deletions(-)
--
2.1.4
- [Qemu-devel] [PATCH V3 0/6] virtio 1.0 pci optimizations and fixes,
Jason Wang <=
- [Qemu-devel] [PATCH V3 1/6] virtio-pci: fix 1.0 virtqueue migration, Jason Wang, 2015/11/06
- [Qemu-devel] [PATCH V3 2/6] memory: don't try to adjust endianness for zero length eventfd, Jason Wang, 2015/11/06
- [Qemu-devel] [PATCH V3 3/6] KVM: add support for any length io eventfd, Jason Wang, 2015/11/06
- [Qemu-devel] [PATCH V3 4/6] virtio-pci: use zero length mmio eventfd for 1.0 notification cap when possible, Jason Wang, 2015/11/06
- [Qemu-devel] [PATCH V3 5/6] virtio-pci: introduce pio notification capability for modern device, Jason Wang, 2015/11/06
- [Qemu-devel] [PATCH V3 6/6] virtio-pci: unbreak queue_enable read, Jason Wang, 2015/11/06