Re: [PATCH 4/4] KVM/MIPS32: Bring in patch from David Daney with new 64 bit compatible ABI.

[Gleb Natapov <gleb@redhat.com>]

On Sun, May 19, 2013 at 02:17:33PM -0700, David Daney wrote:
> On 05/19/2013 07:17 AM, Gleb Natapov wrote:
> >On Sat, May 18, 2013 at 06:54:26AM -0700, Sanjay Lal wrote:
> >>From: David Daney <david.daney@cavium.com>
> >>
> >>There are several parts to this:
> >>
> >>o All registers are 64-bits wide, 32-bit guests use the least
> >>   significant portion of the register storage fields.
> >>
> >>o FPU register formats are defined.
> >>
> >>o CP0 Registers are manipulated via the KVM_GET_MSRS/KVM_SET_MSRS
> >>   mechanism.
> >>
> >>The vcpu_ioctl_get_regs and vcpu_ioctl_set_regs function pointers
> >>become unused so they were removed.
> >>
> >>Some IOCTL functions were moved to kvm_trap_emul because the
> >>implementations are only for that flavor of KVM host.  In the future, if
> >>hardware based virtualization is added, they can be hidden behind
> >>function pointers as appropriate.
> >>
> >David, can you please divide this one big patch to smaller patches
> >with each one having only one of the changes listed above?
> 
> Expanding the registers to 64 bits changes only four lines. Defining
> the FPU registers is an additional seven lines.  The rest really has
> to be an atomic change.
> 
It does not matter. If you have 10 logically unrelated one-liners (even
if they are all part of one big goal) I expect to get 10 patches.

> The point here is that we change the ABI.  Any userspace tools have
> to change too.  So is it better to have a multi-part patch set where
> the interface is unusable in the intermediate patches?  Or is it
> preferable to do an 'atomic' switch?
Are "The vcpu_ioctl_get_regs and vcpu_ioctl_set_regs function pointers
become unused so they were removed." and "Some IOCTL functions were
moved to kvm_trap_emul..." also changes ABI? Unlikely, and then I expect
to have two series: first one only have patches that change ABI and
another rearrange the code. First one will go into 3.10 second in 3.11.

> 
> It wasn't out of laziness that I chose to do it this way, it was
> because I thought it was cleaner.
> 
> So to directly answer your question:  I prefer not to split this up,
> and would want to have a better reason than an orthodox
> interpretation of SubmittingPatches sec. 3.
> 
It may seams orthodox interpretation if you are on a sender side, from
a reviewer point of view it is the interpretation that saves a lot of
time. I did looked into the patch before asking for split, not just
asked for it based on the description. And, in addition, in this case,
I want to have minimal set of changes that will go into 3.10.

--
			Gleb.