Re: [Qemu-devel] [PULL 01/17] migration: remove register_savevm()

[Juan Quintela]

Peter Maydell <address@hidden> wrote:
> On 6 June 2017 at 03:51, David Gibson <address@hidden> wrote:
>> From: Laurent Vivier <address@hidden>
>>
>> We can replace the four remaining calls of register_savevm() by
>> calls to register_savevm_live(). So we can remove the function and
>> as we don't allocate anymore the ops pointer with g_new0()
>> we don't have to free it then.
>>
>> Signed-off-by: Laurent Vivier <address@hidden>
>> Reviewed-by: Juan Quintela <address@hidden>
>> Signed-off-by: David Gibson <address@hidden>
>> ---
>>  hw/net/vmxnet3.c            |  8 ++++++--
>>  hw/s390x/s390-skeys.c       |  9 +++++++--
>>  hw/s390x/s390-virtio-ccw.c  |  8 ++++++--
>>  include/migration/vmstate.h |  8 --------
>>  migration/savevm.c          | 16 ----------------
>>  slirp/slirp.c               |  8 ++++++--
>>  6 files changed, 25 insertions(+), 32 deletions(-)
>
> Great to see register_savevm() finally disappearing.
>
> Any chance of an update to docs/migration.txt, which still
> mentions register_savevm(), but on the other hand doesn't
> say anything about register_savevm_live() and unregister_savevm().
> (Doc comments in the .h file for those functions would be
> nice too...)

Ok, will take a look.

> Things that would be interesting to explain/document:
>  * what is special about vmxnet3 that makes it the only pci device
>    that needs to use this rather than having a vmstate struct?

Will take a look.  vmxnet3 used to be a mess (in relation to migration).

>  * why does s390-skeys call the register function with a NULL
>    pointer but the unregister pointer with a device pointer?

No clue, will left that 

> (Could we replace the uses of these which pass a dev pointer
> with vmstate structs and then drop the dev parameter?)

Not sure, have to take a look.

Thanks, Juan.