Re: [Qemu-devel] [RFC][PATCH] ide: Break migration by splitting error status from status register

[Paolo Bonzini]

On 05/31/2011 12:09 PM, Kevin Wolf wrote:
> When adding the werror=stop mode, some flags were added to s->status
> which are used to determine what kind of operation should be restarted
> when the VM is continued.
>
> Unfortunately, it turns out that s->status is in fact a device register
> and as such is visible to the guest (some of the abused bits are even
> writable for the guest).
>
> Splitting the internal status and the status register into two different
> variables is easy enough, but this will break migration: We must have a
> way to detect what s->status really means. Is it only the status register
> (as used by new versions) or do we have to extract internal error status
> flags?
>
> Here we seem to be lacking some kind of optional subsection that would
> be simply ignored by older versions, but can contain information for new
> versions. Is there any precedence on how to solve this?

You need to stop writing either status field to the migration stream; 
instead you recreate the "wrong" status field before saving, and set the 
"right" status fields from the saved data after loading.

On top of this, you use a subsection to save bits 3-7 of the "real" IDE 
status registers.  These had been hijacked, so there is no room for them 
in the migration stream.  Of course, the subsection is needed only if 
any of those bits is set.

Paolo