On 08/04/2011 09:09 AM, Kevin Wolf wrote:
Am 04.08.2011 15:20, schrieb Anthony Liguori:
On 08/04/2011 02:22 AM, Paolo Bonzini wrote:
On 08/03/2011 11:42 PM, Juan Quintela wrote:
I can certainly limit the change to IDE if we think machine, floppy,
and rtl8139 are safe.
Ok, only IDE is broken, something done if we are not reverting the
others.
Floppy is broken too, and has the problem that the subsection is sent
almost always. On the other hand IDE is only broken if we have
subsections, and that is a rare case.
So perhaps the best course of action is to only fix floppy, either as in
your patch or by reverting 7d905f71. Let's keep IDE as it is broken as
0.14 was, and apply my patch to master.
Hrm, so the idea is to keep IDE broken but not worry because it is so
rare? Why not just force the subsection to be sent?
Wouldn't that only introduce more broken cases in which the destination
thinks it has successfully read the whole VM state when in fact it has
only been confused by a subsection?
Force the subsection to be sent and increment the version number.
Regards,
Anthony Liguori
Kevin
