| Message ID | 1467132449-1030-4-git-send-email-vkuznets@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 28/06/16 17:47, Vitaly Kuznetsov wrote: > HYPERVISOR_vcpu_op passes Linux's idea of vCPU id as a parameter while > Xen's idea is expected. In some cases these ideas diverge so we need to > do remapping. > > There is an issue, however. PV guests do VCPUOP_is_up very early > (see xen_fill_possible_map() and xen_filter_cpu_maps()) when we don't have > perpu areas initialized. While it could be solved with switching to > early_percpu for xen_vcpu_id I think it's not worth it: PV guests will > probably never get to the point where their idea of vCPU id diverges from > Xen's. [...] > static inline int > HYPERVISOR_vcpu_op(int cmd, int vcpuid, void *extra_args) > { > - return _hypercall3(int, vcpu_op, cmd, vcpuid, extra_args); > + /* > + * PV guests call HYPERVISOR_vcpu_op before percpu areas are > + * initialized. As we always use direct mapping for vCPU ids > + * for them we can simply use Linux vcpuid here. > + */ > + return _hypercall3(int, vcpu_op, cmd, > + per_cpu(xen_vcpu_id, vcpuid) != -1 ? > + per_cpu(xen_vcpu_id, vcpuid) : vcpuid, > + extra_args); > } HYPERVISOR_vcpu_op() should take Xen VCPUs, with the callers doing the mapping. David
diff --git a/arch/x86/include/asm/xen/hypercall.h b/arch/x86/include/asm/xen/hypercall.h index a12a047..20750c8 100644 --- a/arch/x86/include/asm/xen/hypercall.h +++ b/arch/x86/include/asm/xen/hypercall.h @@ -49,6 +49,7 @@ #include <xen/interface/physdev.h> #include <xen/interface/platform.h> #include <xen/interface/xen-mca.h> +#include <xen/xen-ops.h> /* * The hypercall asms have to meet several constraints: @@ -422,7 +423,15 @@ HYPERVISOR_vm_assist(unsigned int cmd, unsigned int type) static inline int HYPERVISOR_vcpu_op(int cmd, int vcpuid, void *extra_args) { - return _hypercall3(int, vcpu_op, cmd, vcpuid, extra_args); + /* + * PV guests call HYPERVISOR_vcpu_op before percpu areas are + * initialized. As we always use direct mapping for vCPU ids + * for them we can simply use Linux vcpuid here. + */ + return _hypercall3(int, vcpu_op, cmd, + per_cpu(xen_vcpu_id, vcpuid) != -1 ? + per_cpu(xen_vcpu_id, vcpuid) : vcpuid, + extra_args); } #ifdef CONFIG_X86_64
HYPERVISOR_vcpu_op passes Linux's idea of vCPU id as a parameter while Xen's idea is expected. In some cases these ideas diverge so we need to do remapping. There is an issue, however. PV guests do VCPUOP_is_up very early (see xen_fill_possible_map() and xen_filter_cpu_maps()) when we don't have perpu areas initialized. While it could be solved with switching to early_percpu for xen_vcpu_id I think it's not worth it: PV guests will probably never get to the point where their idea of vCPU id diverges from Xen's. Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com> --- arch/x86/include/asm/xen/hypercall.h | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)