[dpdk-dev] [PATCH 00/12] qemu vhost-user support

Tetsuya Mukawa mukawa at igel.co.jp
Fri Jan 30 11:02:48 CET 2015


On 2015/01/30 15:36, Huawei Xie wrote:
> vhost-user supports passing vring information to a seperate vhost enabled
> process, normally a user space vSwitch, through unix domain socket.
>
> In previous DPDK version, we implement a user space character device driver
> vhost-cuse in user space DPDK process. vring informations are passed to the
> driver through ioctl call, including eventfds for interrupt injection and
> host notification. We need to develop a kernel module to copy that fd from
> qemu into our process. We also need some trick to map guest memory.
> (TODO: kickfd/callfd is reversed which causes confusion)
>
> known issue in vhost-user implementation in QEMU, reported by haifeng.lin at huawei.com
> * QEMU doesn't send correct memory region information with multiple numa node configuration
>  	http://lists.gnu.org/archive/html/qemu-devel/2014-12/msg01454.html
>
> Thanks Tetsuya for reporting the issue that "FD_ISSET would crash when receive -1
> as fd on Ubuntu 14.04".
>
> Huawei Xie (12):
>   enable VIRTIO_NET_F_CTRL_RX
>   create vhost_cuse directory and move vhost-net-cdev.c into vhost_cuse
>   rename vhost-net-cdev.h to vhost-net.h
>   move fd copying(from qemu process into vhost process) to eventfd_copy.c
>   copy host_memory_map from virtio-net.c to a new file virtio-net-cdev.c
>   make host_memory_map more generic
>   split set_memory_table into two parts
>   add select based event driven processing
>   free memory when receive new set_memory_table message
>   vhost user support
>   support dev->ifname in vhost-user
>   support calling rte_vhost_driver_register after rte_vhost_driver_session_start
>
>  lib/librte_vhost/Makefile                     |   8 +-
>  lib/librte_vhost/rte_virtio_net.h             |   5 +-
>  lib/librte_vhost/vhost-net-cdev.c             | 389 ---------------------
>  lib/librte_vhost/vhost-net-cdev.h             | 113 -------
>  lib/librte_vhost/vhost-net.h                  | 121 +++++++
>  lib/librte_vhost/vhost_cuse/eventfd_copy.c    |  89 +++++
>  lib/librte_vhost/vhost_cuse/eventfd_copy.h    |  40 +++
>  lib/librte_vhost/vhost_cuse/vhost-net-cdev.c  | 417 +++++++++++++++++++++++
>  lib/librte_vhost/vhost_cuse/virtio-net-cdev.c | 401 ++++++++++++++++++++++
>  lib/librte_vhost/vhost_cuse/virtio-net-cdev.h |  48 +++
>  lib/librte_vhost/vhost_rxtx.c                 |   2 +-
>  lib/librte_vhost/vhost_user/fd_man.c          | 234 +++++++++++++
>  lib/librte_vhost/vhost_user/fd_man.h          |  66 ++++
>  lib/librte_vhost/vhost_user/vhost-net-user.c  | 469 ++++++++++++++++++++++++++
>  lib/librte_vhost/vhost_user/vhost-net-user.h  | 106 ++++++
>  lib/librte_vhost/vhost_user/virtio-net-user.c | 322 ++++++++++++++++++
>  lib/librte_vhost/vhost_user/virtio-net-user.h |  49 +++
>  lib/librte_vhost/virtio-net.c                 | 455 +++----------------------
>  lib/librte_vhost/virtio-net.h                 |  43 +++
>  19 files changed, 2460 insertions(+), 917 deletions(-)
>  delete mode 100644 lib/librte_vhost/vhost-net-cdev.c
>  delete mode 100644 lib/librte_vhost/vhost-net-cdev.h
>  create mode 100644 lib/librte_vhost/vhost-net.h
>  create mode 100644 lib/librte_vhost/vhost_cuse/eventfd_copy.c
>  create mode 100644 lib/librte_vhost/vhost_cuse/eventfd_copy.h
>  create mode 100644 lib/librte_vhost/vhost_cuse/vhost-net-cdev.c
>  create mode 100644 lib/librte_vhost/vhost_cuse/virtio-net-cdev.c
>  create mode 100644 lib/librte_vhost/vhost_cuse/virtio-net-cdev.h
>  create mode 100644 lib/librte_vhost/vhost_user/fd_man.c
>  create mode 100644 lib/librte_vhost/vhost_user/fd_man.h
>  create mode 100644 lib/librte_vhost/vhost_user/vhost-net-user.c
>  create mode 100644 lib/librte_vhost/vhost_user/vhost-net-user.h
>  create mode 100644 lib/librte_vhost/vhost_user/virtio-net-user.c
>  create mode 100644 lib/librte_vhost/vhost_user/virtio-net-user.h
>  create mode 100644 lib/librte_vhost/virtio-net.h
>

Hi Xie,

it seems "checkpath.pl" reports warnings in some patches.
I guess some of them can be fixed.

Thanks,
Tetsuya



More information about the dev mailing list