>>> On 01.09.16 at 11:13, <feng...@intel.com> wrote: >> From: Jan Beulich [mailto:jbeul...@suse.com] >> Sent: Thursday, September 1, 2016 4:17 PM >> >>> On 31.08.16 at 05:56, <feng...@intel.com> wrote: >> > --- a/xen/arch/x86/hvm/vmx/vmx.c >> > +++ b/xen/arch/x86/hvm/vmx/vmx.c >> > @@ -207,8 +207,6 @@ void vmx_pi_hooks_assign(struct domain *d) >> > ASSERT(!d->arch.hvm_domain.vmx.vcpu_block); >> > >> > d->arch.hvm_domain.vmx.vcpu_block = vmx_vcpu_block; >> > - d->arch.hvm_domain.vmx.pi_switch_from = vmx_pi_switch_from; >> > - d->arch.hvm_domain.vmx.pi_switch_to = vmx_pi_switch_to; >> > d->arch.hvm_domain.vmx.pi_do_resume = vmx_pi_do_resume; >> > } >> >> I'm sure I've said so before: While I can see why you want the >> adjustment to vmx_pi_hooks_deassign(), I don't see why leaving >> out this and ... > > As mentioned in the comments of this patch, these two hooks are also > needed for CPU side PI, so we cannot install here and ...
I've just read it again, and to me it doesn't say so. I.e. ... >> > @@ -236,6 +232,12 @@ static int vmx_domain_initialise(struct domain *d) >> > if ( (rc = vmx_alloc_vlapic_mapping(d)) != 0 ) >> > return rc; >> > >> > + if ( iommu_intpost ) >> > + { >> > + d->arch.hvm_domain.vmx.pi_switch_from = vmx_pi_switch_from; >> > + d->arch.hvm_domain.vmx.pi_switch_to = vmx_pi_switch_to; >> > + } >> > + >> > return 0; >> > } >> >> ... this would conflict with the goal of the patch. > > ... Yes this is incorrect, I should install these two hooks when CPU side PI > is > supported. ... you'd then additionally need to explain what bug there is without these hooks and with only CPU side PI present (iirc that's functionality which has been there before the VT-d part). Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel