On 20/02/24 11:34 am, Peter Xu wrote:
> On Fri, Feb 16, 2024 at 09:06:23AM +0000, Het Gala wrote:
>> migration QAPIs can now work with either 'channels' or 'uri' as their
>> argument.
>>
>> Signed-off-by: Het Gala <het.gala@nutanix.com>
>> ---
>> tests/qtest/migration-test.c | 7 +++++++
>> 1 file changed, 7 insertions(+)
>>
>> diff --git a/tests/qtest/migration-test.c b/tests/qtest/migration-test.c
>> index e7f2719dcf..0bc69b1943 100644
>> --- a/tests/qtest/migration-test.c
>> +++ b/tests/qtest/migration-test.c
>> @@ -704,6 +704,13 @@ typedef struct {
>> */
>> const char *connect_uri;
>>
>> + /*
>> + * Optional: list of migration stream channels, each connected
>> + * to a dst QEMU. It can be used instead of URI to carry out
>> + * the same task as listen_uri or connect_uri.
>> + */
>> + MigrationChannelList *connect_channels;
>> +
>> /* Optional: callback to run at start to set migration parameters */
>> TestMigrateStartHook start_hook;
>> /* Optional: callback to run at finish to cleanup */
> Please squash this patch into the follow up patch that uses it. Thanks,
Yes sure.
I am also planning to convert this field into a bool (just say whether
connect_channels would be present or not). It would prove useful for
positive cases actually where only channel is being used, because if I
convert them before hand itself port is 0 but kernel converts port 0 and
gives a random port number for migration. And positive tests fail there.
Will be more clear when I post the v2 patchset. Let me know, if it does
not sound right then.
Regards,
Het Gala