Re: [Qemu-devel] [PATCH 6/6] migration: Pass Error ** argument to {save, load}_vmstate

[Juan Quintela]

Laurent Vivier <address@hidden> wrote:
> On 25/04/2017 19:00, Juan Quintela wrote:
>> Laurent Vivier <address@hidden> wrote:
>>> On 25/04/2017 12:24, Juan Quintela wrote:

>>>>  {
>>>> +    Error *err = NULL;
>>>> +
>>>>      if (replay_snapshot) {
>>>>          if (replay_mode == REPLAY_MODE_RECORD) {
>>>> -            if (save_vmstate(replay_snapshot) != 0) {
>>>> +            if (save_vmstate(replay_snapshot, &err) != 0) {
>>>>                  error_report("Could not create snapshot for icount 
>>>> record");
>>>>                  exit(1);
>>>>              }
>>>>          } else if (replay_mode == REPLAY_MODE_PLAY) {
>>>> -            if (load_vmstate(replay_snapshot) != 0) {
>>>> +            if (load_vmstate(replay_snapshot, &err) != 0) {
>>>>                  error_report("Could not load snapshot for icount replay");
>>>>                  exit(1);
>>>>              }
>>>
>>> You can use "&error_fatal" in these cases.
>> 
>> I was very happy with your suggestion.  But then I realized that if I
>> use error_fatal, I would lost the error_report() messages, right?
>
> The "Could not create snapshot for icount record" and "Could not load
> snapshot for icount replay", yes.
>
> But you keep the one from the inside load_vmstate().

    if (!bdrv_all_can_snapshot(&bs)) {
        error_setg(errp, "Device '%s' is writable but does not support "
                   "snapshots", bdrv_get_device_name(bs));
        return ret;
    }

This is the 1st error on save_vmstate.  My understanding is that now I
get something like (two messages):

Device 'foo' is writable but does not support snapshots
Could not create snapshot for icount record

If I change to error_fatal, only the 1st message will appear.  And I am
not replay maintainer to decide the change O:-)


> Isn't it enough?

Dunno.

Later, Juan.