[libvirt] [PATCH] esx: Explicitly disable unused floppy0 device for GSX

[Matthias Bolte]

2010/8/11 Daniel Veillard <veillard at redhat.com>:
> On Mon, Aug 09, 2010 at 04:44:14PM -0600, Eric Blake wrote:
>> On 08/05/2010 06:45 PM, Matthias Bolte wrote:
>> > floppy0.present defaults to true for GSX. Therefore, it needs to be
>> > explicitly disabled for GSX.
>> > ---
>> >  src/esx/esx_vmx.c |   10 ++++++++++
>> >  1 files changed, 10 insertions(+), 0 deletions(-)
>>
>> If I understand correctly, you want to uniformly make floppy not present
>> except when the user asked for it.  The other ESX clients meet this
>> goal, but for GSX, you have to add an extra argument to make it so.
>> That being the case, I agree with the approach.

That's the intention, but it didn't struck me right away that ESX and
GSX being different here is odd.

I noticed that floppy0.present defaults to true on GSX while
investigating some other GSX related issues in the VMX handling code.

The comments in the esx_vmx.c file indicate that ESX would default to
false in that case, but that's wrong, as I know now. floppy0.present
defaults to true for GSX _and_ ESX, so the assumed default for ESX was
wrong all the time. I probably screwed it when I wrote the initial VMX
handling code. Therefore, this patch is incomplete and we need a v2 to
address the ESX issue too.

>> > @@ -2539,6 +2544,11 @@ esxVMX_FormatConfig(esxVMX_Context *ctx, virCapsPtr caps, virDomainDefPtr def,
>> >          }
>> >      }
>> >
>> > +    if (!hasFloppyDevice && productVersion & esxVI_ProductVersion_GSX) {
>>
>> I'd add an extra () here to make precedence clear:
>>
>> if (!hasFloppyDevice && (productVersion & esxVI_ProductVersion_GSX)) {
>>
>> ACK with that nit fixed.

v2 won't have the product version check anymore.

>
>  Agreed, ACK
>
> Matthias you push it ?
>
> Daniel
>

Well, no, as we need a v2 for this :)

Matthias