On 2/9/19 9:41 AM, John Ferlan wrote:
On 2/6/19 2:18 PM, Eric Blake wrote:
Right now, we only use NBD exports during storage migration, where we control the NBD client and therefore don't care about the export name (qemu's default of naming the export after the device is fine). But upcoming patches for exposing backups over NBD wants to use a different export name (matching the libvirt domain XML rather than the qemu device name); so enhance the QMP glue code to allow this flexibility.
Signed-off-by: Eric Blake <eblake@redhat.com> --- src/qemu/qemu_monitor.h | 1 + src/qemu/qemu_monitor_json.h | 1 + src/qemu/qemu_migration.c | 2 +- src/qemu/qemu_monitor.c | 5 +++-- src/qemu/qemu_monitor_json.c | 2 ++ tests/qemumonitorjsontest.c | 2 +- 6 files changed, 9 insertions(+), 4 deletions(-)
I assume/hope some day a test will be added that will show what the @export variable does. I haven't read ahead at all - I'm trying to just go in order and get some feedback sooner so you can start pushing stuff.
Reviewed-by: John Ferlan <jferlan@redhat.com>
Yes, I need to do better at in-tree tests (and not just stuff I've mentioned in the commit messages) for my v5. The export name gets used during pull-style backups, where I purposefully went with naming the exports 'vda' and 'vdb' for a guest with two disks, rather than based on any file name or block node name; 'qemu-nbd --list' connected to the NBD server during a pull backup is an easy way of testing the export names made it through the stack. -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org