This patchset allows QEMU to poll and check the device used buffer
after sending all SVQ control commands, instead of polling and checking
immediately after sending each SVQ control command, so that QEMU can
send all the SVQ control commands in parallel, which have better
performance improvement.
I use vdpa_sim_net to simulate vdpa device, refactor
vhost_vdpa_net_load() to call vhost_vdpa_net_load_mac() 30 times,
to build a test environment for sending multiple SVQ control commands.
The monotonic time to finish vhost_vdpa_net_load() is as follows:
QEMU microseconds
--------------------------------------------------
not patched 85.092
--------------------------------------------------
patched 79.222
So this is a save of (85.092 - 79.222)/30 = 0.2 ms per command.
This patchset resolves the GitLab issue at
https://gitlab.com/qemu-project/qemu/-/issues/1578.
v2:
- recover accidentally deleted rows
- remove extra newline
- refactor `need_poll_len` to `cmds_in_flight`
- return -EINVAL when vhost_svq_poll() return 0 or check
on buffers written by device fails
- change the type of `in_cursor`, and refactor the
code for updating cursor
- return directly when vhost_vdpa_net_load_{mac,mq}()
returns a failure in vhost_vdpa_net_load()
v1: https://lists.nongnu.org/archive/html/qemu-devel/2023-04/msg02668.html
Hawkins Jiawei (2):
vdpa: rename vhost_vdpa_net_cvq_add()
vdpa: send CVQ state load commands in parallel
net/vhost-vdpa.c | 165 +++++++++++++++++++++++++++++++++++++----------
1 file changed, 130 insertions(+), 35 deletions(-)
--
2.25.1
