Re: [PATCH v9 16/17] libxl: convert libxl_device_vkb_add to an async operation

[Roger Pau Monne <roger.pau@citrix.com>]

Ian Jackson wrote:
> Roger Pau Monne writes ("Re: [PATCH v9 16/17] libxl: convert libxl_device_vkb_add to an async operation"):
>> Ian Jackson wrote:
> ...
>>> This is rather unfriendly - doesn't it amount to aborting on an
>>> invalid domain configuration.  And I'm not sure I understand the
>>> logic, which is new in your patch.
>> This was part of the previous code, but this was done inside of a switch
>> instead of an if. Since it's not possible to specify vkb devices in the
>> config file (at least xl.cfg(5) doesn't mention anything about vkbs),
>> I'm quite sure d_config->vkbs == NULL and d_config->num_vkbs == 0 always
>> or something has gone extremely wrong, just added the abort to be sure.
> 
> People who write other code which calls libxl might try to pass vkbds,
> but then the code would crash.  And if they didn't test it with HVM
> domains, they wouldn't notice that their code had a crash bug.
> 
> I think it should cause the creation to fail with ERROR_INVAL.

I've been thinking about this and I think that maybe we should only
perform this piece of code if num_vkbs is 0, so we make sure there's at
least one vkb device, but not fail if num_vkbs > 0.