Re: [PATCH v3] kvm: make vcpu life cycle separated from kvm instance

[Gleb Natapov <gleb@redhat.com>]

On Thu, Dec 15, 2011 at 05:06:09PM +0800, Liu ping fan wrote:
> 2011/12/15 Gleb Natapov <gleb@redhat.com>:
> > On Thu, Dec 15, 2011 at 11:21:37AM +0800, Liu ping fan wrote:
> >> On Tue, Dec 13, 2011 at 7:36 PM, Marcelo Tosatti <mtosatti@redhat.com> wrote:
> >> > On Mon, Dec 12, 2011 at 10:41:23AM +0800, Liu Ping Fan wrote:
> >> >> From: Liu Ping Fan <pingfank@linux.vnet.ibm.com>
> >> >>
> >> >> Currently, vcpu can be destructed only when kvm instance destroyed.
> >> >> Change this to vcpu's destruction taken when its refcnt is zero,
> >> >> and then vcpu MUST and CAN be destroyed before kvm's destroy.
> >> >>
> >> >> Signed-off-by: Liu Ping Fan <pingfank@linux.vnet.ibm.com>
> >> >> ---
> >> >>  arch/x86/kvm/i8254.c     |   10 ++++--
> >> >>  arch/x86/kvm/i8259.c     |   12 +++++--
> >> >>  arch/x86/kvm/mmu.c       |    7 ++--
> >> >>  arch/x86/kvm/x86.c       |   54 +++++++++++++++++++----------------
> >> >>  include/linux/kvm_host.h |   71 ++++++++++++++++++++++++++++++++++++++++++----
> >> >>  virt/kvm/irq_comm.c      |    7 +++-
> >> >>  virt/kvm/kvm_main.c      |   62 +++++++++++++++++++++++++++++++++------
> >> >>  7 files changed, 170 insertions(+), 53 deletions(-)
> >> >
> >> > This needs a full audit of paths that access vcpus. See for one example
> >> > bsp_vcpu pointer.
> >> >
> >> Yes, I had missed it and just paid attention to the access path to
> >> vcpu in kvm_lapic and the path used in async_pf. I will correct it
> >> later.
> >> BTW, I want to make it sure that because kvm_lapic will be destroyed
> >> before vcpu, so  it is safe to bypass the access path there, and the
> >> situation is the same in async_pf for we have called
> >> kvm_clear_async_pf_completion_queue before zapping vcpu.  Am I right?
> >>
> >> As to the scene like bsp_vcpu, I think that introducing refcount like
> >> in V2 can handle it easier. Please help to review these changes in V4
> >> which I will send a little later.
> >>
> > Since bsp_vcpu pointer will never be released or re-assigned introducing
> > reference count to keep the pointer valid is not necessary. The counter
> > will never reach 0 and bsp vcpu will never be freed. Just disallow
> 
> OK. And I have a question -- who will play the role to guard bsp_vcpu?
> kernel or qemu?  Must I add something in kernel to protect the
> bsp_vcpu
> 
Kernel of course. But I prefer just to rid of bsp_vcpu. I'll try to send
patch today.

> > removal of bsp_vcpu. Or better get rid of bsp_vcpu at all since its only
> > use is invalid anyway.
> >
> I will dig into it and see how to handle it.
> 
> Thanks and regards,
> ping fan
> > --
> >                        Gleb.

--
			Gleb.