On Wed, Apr 27, 2016 at 11:25:23PM +0800, Huang Lei wrote:
> From: Huang Lei <lhua...@ebay.com>
>
> In some environments, an interface may have multiple IP addresses
> in same subnet, if TCP client socket doesn't call bind() explicitly,
> OS will chooses an local IP and port for it, usually the primary IP of
> the subnet will be chosen. With this patch, a secondary IP of the
> subnet can be specified as local address of TCP connection.
>
> Signed-off-by: Huang Lei <lhua...@ebay.com>
Thanks for working on this.
I see that you updated some of the documentation to mention the new
support. However, the updates missed a number of locations, at least
the following. Please update them also:
lib/vconn-active.man:2:.IQ "\fBtcp:\fIip\fR[\fB:\fIport\fR]"
lib/vconn-passive.man:2:.IQ "\fBptcp:\fR[\fIport\fR][\fB:\fIip\fR]"
ovsdb/remote-active.man:9:.IP "\fBtcp:\fIip\fB:\fIport\fR"
ovsdb/remote-passive.man:12:.IP "\fBptcp:\fIport\fR[\fB:\fIip\fR]"
I think that this new feature bears mentioning in NEWS, so please add an
item there also.
Please update the comment on inet_parse_active() to explain the new
parameter.
In inet_parse_active(), the treatment of logical_ss and foreign_ss is
inconsistent in the face of failure. If parsing fails in one spot, only
foreign_ss is zeroed; if it fails in another, only local_ss.
Presumably, both should always be zeroed, if supplied.
It would probably be more user-friendly to report an error from
inet_parse_active() if the string includes a local address but local_ss
is NULL.
It looks like local_ss defaults to an IPv4 address even if an IPv6
address is supplied for the remote address. That seems wrong.
I don't see any IPv6 tests. I suggest adding a few.
Thanks,
Ben.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
