Peter Maydell <peter.mayd...@linaro.org> wrote: > On 11 October 2018 at 11:45, Philippe Mathieu-Daudé <phi...@redhat.com> wrote: >> On 11/10/2018 11:00, Mao Zhongyi wrote: >>> Cc: Jan Kiszka <jan.kis...@web.de> >>> Cc: Peter Maydell <peter.mayd...@linaro.org> >>> Cc: Gerd Hoffmann <kra...@redhat.com> >>> To: qemu-...@nongnu.org >>> >>> Signed-off-by: Mao Zhongyi <maozhon...@cmss.chinamobile.com> >>> --- >>> hw/arm/musicpal.c | 16 ++++++++-------- >>> hw/audio/marvell_88w8618.c | 3 +-- >>> include/hw/audio/wm8750.h | 1 + >>> 3 files changed, 10 insertions(+), 10 deletions(-) >>> >>> diff --git a/hw/arm/musicpal.c b/hw/arm/musicpal.c >>> index ac266f9253..6425f1d50f 100644 >>> --- a/hw/arm/musicpal.c >>> +++ b/hw/arm/musicpal.c >>> @@ -406,7 +406,7 @@ static void mv88w8618_eth_realize(DeviceState >>> *dev, Error **errp) >>> } >>> >>> static const VMStateDescription mv88w8618_eth_vmsd = { >>> - .name = "mv88w8618_eth", >>> + .name = TYPE_MV88W8618_ETH, >> >> I understand TYPE_device name might changed, but once used, >> VMStateDescription shouldn't, else this would break migration. >> Thus I wouldn't recommend using TYPE_name in VMStateDescription.name. >> >> Since I'm not sure I cc'ed the migration maintainers. > > Yes, that's the usual approach -- the vmstate struct names > aren't inherently the same as the QOM type names, and so > we keep them decoupled to avoid accidentally breaking migration.
Yeap, I agree. Later, Juan.