Re: [Qemu-devel] [PATCH v6 0/8] Vhost and vhost-net support for userspace based backends

[Antonios Motakis]

Hello,

On Tue, Jan 14, 2014 at 12:14 PM, Michael S. Tsirkin <address@hidden> wrote:

On Mon, Jan 13, 2014 at 03:25:11PM +0100, Antonios Motakis wrote:
> In this patch series we would like to introduce our approach for putting a
> virtio-net backend in an external userspace process. Our eventual target is to
> run the network backend in the Snabbswitch ethernet switch, while receiving
> traffic from a guest inside QEMU/KVM which runs an unmodified virtio-net
> implementation.
>
> For this, we are working into extending vhost to allow equivalent functionality
> for userspace. Vhost already passes control of the data plane of virtio-net to
> the host kernel; we want to realize a similar model, but for userspace.
>
> In this patch series the concept of a vhost-backend is introduced.
>
> We define two vhost backend types - vhost-kernel and vhost-user. The former is
> the interface to the current kernel module implementation. Its control plane is
> ioctl based. The data plane is the kernel directly accessing the QEMU allocated,
> guest memory.
>
> In the new vhost-user backend, the control plane is based on communication
> between QEMU and another userspace process using a unix domain socket. This
> allows to implement a virtio backend for a guest running in QEMU, inside the
> other userspace process.
>
> We change -mem-path to QemuOpts and add prealloc, share and unlink as properties
> to it. HugeTLBFS requirements of -mem-path are relaxed, so any valid path can
> be used now.

Wait a second. This does not actually work well: if you mmap
a random file outside HugeTLBFS, kernel won't create huge pages
from this memory so performance of the system as a whole will suffer.

You'll have fix the kernel MM before this scheme can fly.

I'm not sure I completely understand this point. It is up to the user to choose not to use HugeTLBFS. Is there a particular problem with the kernel when not using it?

 

> The new properties allow more fine grained control over the guest
> RAM backing store.
>
> The data path is realized by directly accessing the vrings and the buffer data
> off the guest's memory.
>
> The current user of vhost-user is only vhost-net. We add new netdev backend
> that is intended to initialize vhost-net with vhost-user backend.
>
> Example usage:
>
> qemu -m 1024 -mem-path /hugetlbfs,prealloc=on,share=on \
>      -netdev type=vhost-user,id=net0,path=/path/to/sock,poll_time=2500 \
>      -device virtio-net-pci,netdev=net0
>
> Changes from v5:
>  - Split -mem-path unlink option to a separate patch
>  - Fds are passed only in the ancillary data
>  - Stricter message size checks on receive/send
>  - Netdev vhost-user now includes path and poll_time options
>  - The connection probing interval is configurable
>
> Changes from v4:
>  - Use error_report for errors
>  - VhostUserMsg has new field `size` indicating the following payload length.
>    Field `flags` now has version and reply bits. The structure is packed.
>  - Send data is of variable length (`size` field in message)
>  - Receive in 2 steps, header and payload
>  - Add new message type VHOST_USER_ECHO, to check connection status
>
> Changes from v3:
>  - Convert -mem-path to QemuOpts with prealloc, share and unlink properties
>  - Set 1 sec timeout when read/write to the unix domain socket
>  - Fix file descriptor leak
>
> Changes from v2:
>  - Reconnect when the backend disappears
>
> Changes from v1:
>  - Implementation of vhost-user netdev backend
>  - Code improvements
>
> Antonios Motakis (8):
>   Convert -mem-path to QemuOpts and add prealloc and share properties
>   New -mem-path option - unlink.
>   Decouple vhost from kernel interface
>   Add vhost-user skeleton
>   Add domain socket communication for vhost-user backend
>   Add vhost-user calls implementation
>   Add new vhost-user netdev backend
>   Add vhost-user reconnection
>
>  exec.c                            |  57 +++-
>  hmp-commands.hx                   |   4 +-
>  hw/net/vhost_net.c                | 144 +++++++---
>  hw/net/virtio-net.c               |  42 ++-
>  hw/scsi/vhost-scsi.c              |  13 +-
>  hw/virtio/Makefile.objs           |   2 +-
>  hw/virtio/vhost-backend.c         | 556 ++++++++++++++++++++++++++++++++++++++
>  hw/virtio/vhost.c                 |  46 ++--
>  include/exec/cpu-all.h            |   3 -
>  include/hw/virtio/vhost-backend.h |  40 +++
>  include/hw/virtio/vhost.h         |   4 +-
>  include/net/vhost-user.h          |  17 ++
>  include/net/vhost_net.h           |  15 +-
>  net/Makefile.objs                 |   2 +-
>  net/clients.h                     |   3 +
>  net/hub.c                         |   1 +
>  net/net.c                         |   2 +
>  net/tap.c                         |  16 +-
>  net/vhost-user.c                  | 177 ++++++++++++
>  qapi-schema.json                  |  21 +-
>  qemu-options.hx                   |  24 +-
>  vl.c                              |  41 ++-
>  22 files changed, 1106 insertions(+), 124 deletions(-)
>  create mode 100644 hw/virtio/vhost-backend.c
>  create mode 100644 include/hw/virtio/vhost-backend.h
>  create mode 100644 include/net/vhost-user.h
>  create mode 100644 net/vhost-user.c
>
> --
> 1.8.3.2
>