-
Notifications
You must be signed in to change notification settings - Fork 63
Vsock kernel support #194
Vsock kernel support #194
Conversation
src/init.c
Outdated
| #else | ||
| ctl_serial = "sh.hyper.channel.0"; | ||
| tty_serial = "sh.hyper.channel.1"; | ||
| if (hyper_cmd("/sbin/modprobe vmw_vsock_virtio_transport") < 0) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nice patch, could you please remove /sbin ?
related with #191
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
sure, I'll drop it. Thanks for pointing it out.
aff23bf to
16258c0
Compare
src/init.c
Outdated
|
|
||
| static int hyper_vsock_msg_accept(struct hyper_event *he, int efd) | ||
| { | ||
| return hyper_vsock_accept(he, efd, &ctl.chan, &hyper_vsock_ttyfd_ops); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ctl.tty?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
indeed, thanks!
|
iperf (patched with vsock support, see https://github.com/bergwolf/iperf/tree/vsock) tests with vsock connection. vsock bandwidth is about 3x of IPv4. Both were tested in the same container (64MB with single core). vsock result: IPv4 iperf results: |
listen control channel on port 2718, and message channel on 2719. Signed-off-by: Peng Tao <bergwolf@gmail.com>
Signed-off-by: Peng Tao <bergwolf@gmail.com>
Set up control channel and message channel if unset yet. Signed-off-by: Peng Tao <bergwolf@gmail.com>
Signed-off-by: Peng Tao <bergwolf@gmail.com>
To have larger buffer size and also handle HUP events so that it is possible to reconnect to hyperstart. Signed-off-by: Peng Tao <bergwolf@gmail.com>
Signed-off-by: Peng Tao <bergwolf@gmail.com>
So that we know when runv closes a connection. Signed-off-by: Peng Tao <bergwolf@gmail.com>
1. when read returns 0, it means remotes has closed socket, for normal TCP connections. 2. when read returns -ENOTCONN for vsock, it means remote has closed the socket. We should be able to just handle case 1) if we patch vsock kernel implementation to match normal TCP connection behavior. Signed-off-by: Peng Tao <bergwolf@gmail.com>
Signed-off-by: Peng Tao <bergwolf@gmail.com>
Signed-off-by: Peng Tao <bergwolf@gmail.com>
|
retest this please |
Uh oh!
There was an error while loading. Please reload this page.