On 12/19/2011 06:15 PM, Anthony Liguori wrote:
On 12/19/2011 04:55 PM, Corey Bryant wrote:
diff --git a/net.c b/net.c
index f7bebf8..9296224 100644
--- a/net.c
+++ b/net.c
@@ -952,6 +952,14 @@ static const struct {
.type = QEMU_OPT_STRING,
.help = "script to shut down the interface",
}, {
+ .name = "br",
+ .type = QEMU_OPT_STRING,
+ .help = "bridge name",
+ }, {
I don't think passing br= makes a whole of sense for -net tap. I think
it would make more sense to make sure that helper could take a shell
string so you could do:
-netdev tap,helper="/usr/libexec/qemu-bridge-helper --br=br0"
Regards,
Anthony Liguori
Ok but do you think the -net bridge options should remain as-is? It
seems like
execution of the helper should be consistent. Here are the current
options for
-net bridge:
-net bridge,helper=/usr/local/libexec/qemu-bridge-helper,br=br0
Yes. -net bridge is syntactic sugar for -net tap with specific knowledge
of the qemu-bridge-helper.
If someone wrote a 'qemu-openvswitch-helper' then you could imagine a
'-net openvswitch' option that passed a bunch of openvswitch specific
arguments.
Regards,
Anthony Liguori
It seems like the helper should accept the following arguments:
--vnet-hdr --br=<bridge name> --fd=<unix fd>
(It already accept these, but the --br= and --fd= syntax aren't required at the
moment.)
Then QEMU would only allow the following to be specified for -netdev tap:
-netdev tap,helper="/usr/libexec/qemu-bridge-helper"
or
-netdev tap,helper="/usr/libexec/qemu-bridge-helper --br=bridge"
and would ignore or reject --vnet-hdr and --fd=. --vnet-hdr and --fd= would
always be specified internally when the helper is exec'd.