Re: [Qemu-devel] [PATCH] Migration compatibility for serial

[Michael S. Tsirkin]

On Wed, Jun 17, 2015 at 11:59:05AM +0100, Dr. David Alan Gilbert wrote:
> * Michael S. Tsirkin (address@hidden) wrote:
> > On Wed, Jun 17, 2015 at 11:51:57AM +0100, Dr. David Alan Gilbert wrote:
> > > * Juan Quintela (address@hidden) wrote:
> > > > "Michael S. Tsirkin" <address@hidden> wrote:
> > > > > On Wed, Jun 17, 2015 at 10:11:48AM +0200, Paolo Bonzini wrote:
> > > > >> 
> > > > >> 
> > > > >> On 17/06/2015 09:52, Michael S. Tsirkin wrote:
> > > > >> > > No, please.  Upstream QEMU doesn't want to get into judgement 
> > > > >> > > about when
> > > > >> > > migration quality might be "good enough" that you can drop 
> > > > >> > > subsections.
> > > > >> > >  It's one thing to perfect the .needed functions to make the 
> > > > >> > > appearance
> > > > >> > > of subsections as unlikely as possible, but adding flags is not
> > > > >> > > something we've done so far---and not something at least *I* 
> > > > >> > > want to do.
> > > > >> > 
> > > > >> > Not like this, sure.  But e.g. patches that force specific fields 
> > > > >> > to
> > > > >> > behave in a way consistent with QEMU 2.2, with appropriate
> > > > >> > doducmentation would be ok I think.
> > > > >> 
> > > > >> That's not what 2.2 means in "pc-i440fx-2.2".  It means "same 
> > > > >> hardware
> > > > >> as 2.2", not "bug-compatible with 2.2".
> > > > >> 
> > > > >> Refining the .needed functions (e.g. see commit bfa7362889) is just
> > > > >> that: describing when a subsection is needed.  Forcing specific 
> > > > >> fields
> > > > >> to behave in a way consistent with QEMU 2.2 is bug compatibility.
> > > > >> 
> > > > >> Paolo
> > > > >
> > > > > We do bug-compatible if it's not a big pain, too.
> > > > 
> > > > In this case, there is disagreement about what is better:
> > > > - correct solution
> > > > - bug compatible
> > > > 
> > > > We can't have both in this case :-(
> > > > 
> > > > Notice that if "both" are 2.2 <improved>, i.e. 2.3 with -M
> > > > pc-i440fx-2.2, we also got the correct behaviour.  So the matrix is
> > > > something like:
> > > > 
> > > > Source: 2.2  Destination: 2.2 -> bug compatible 2.2
> > > > Source: 2.3  Destination: 2.2 -> breaks if serial is being used, works 
> > > > otherwise
> > > > Source: 2.3  Destination: 2.3 with -M pc-i440fx-2.2: works always
> > > 
> > > To be fair the 2.3->2.2 is more subtle; opening it is unlikely
> > > to generate the subsections; it needs a bit more than that (certainly on 
> > > Linux)
> > > figuring out exactly what triggers each subsection is trickier.
> > > 
> > > Dave
> > 
> > And more importantly, what is the result of skipping them,
> > like you proposed. E.g. if guests crash that's no
> > better than failing migration.
> 
> I believe it's the same behaviour as qemu serial migration has been
> doing for many years when it never sent that data over, and I'm not
> aware of that ever causing us problems.
> 
> Dave

A bit more data on testing and/or code analysis would be helpful
I think.


> > > > 
> > > > 
> > > > So the problem is 2.3 -> 2.2 when serial is being used (notice that just
> > > > opening it it is using).  That is what we are differing about what is
> > > > the right thing to do.  As Paolo says, in upstream, we have done in the
> > > > past the correct thing, in downstream, it depends.
> > > > 
> > > > Notice that adding this patch makes that the three cases are bug
> > > > compatible, i.e. there is no way to detect breakage neither a way to fix
> > > > the issue (fix without the patch is just upgrade both binaries.
> > > > )
> > > > 
> > > > Later, Juan.
> > > --
> > > Dr. David Alan Gilbert / address@hidden / Manchester, UK
> --
> Dr. David Alan Gilbert / address@hidden / Manchester, UK