[PATCH 2/5] xen/pvh*: Support > 32 VCPUs at domain restore

[Ankur Arora]

When Xen restores a PVHVM or PVH guest, its shared_info only holds
up to 32 CPUs. The hypercall VCPUOP_register_vcpu_info allows
us to setup per-page areas for VCPUs. This means we can boot
PVH* guests with more than 32 VCPUs. During restore the per-cpu
structure is allocated freshly by the hypervisor (vcpu_info_mfn is
set to INVALID_MFN) so that the newly restored guest can make a
VCPUOP_register_vcpu_info hypercall.

However, we end up triggering this condition in Xen:
/* Run this command on yourself or on other offline VCPUS. */
 if ( (v != current) && !test_bit(_VPF_down, &v->pause_flags) )

which means we are unable to setup the per-cpu VCPU structures
for running VCPUS. The Linux PV code paths makes this work by
iterating over cpu_possible in xen_vcpu_restore() with:

 1) is target CPU up (VCPUOP_is_up hypercall?)
 2) if yes, then VCPUOP_down to pause it
 3) VCPUOP_register_vcpu_info
 4) if it was down, then VCPUOP_up to bring it back up

With Xen commit 192df6f9122d ("xen/x86: allow HVM guests to use
hypercalls to bring up vCPUs") this is available for non-PV guests.
As such first check if VCPUOP_is_up is actually possible before
trying this dance.

As most of this dance code is done already in xen_vcpu_restore()
let's make it callable on PV, PVH and PVHVM.

Based-on-patch-by: Konrad Wilk <konrad.wilk@xxxxxxxxxx>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
Signed-off-by: Ankur Arora <ankur.a.arora@xxxxxxxxxx>
---
 arch/x86/xen/enlighten.c     | 45 +++++++++++++++++++++++++++++++-------------
 arch/x86/xen/enlighten_hvm.c | 20 +++++++-------------
 arch/x86/xen/smp_hvm.c       | 10 ++++++++++
 arch/x86/xen/suspend_hvm.c   | 11 +++--------
 include/xen/xen-ops.h        |  2 ++
 5 files changed, 54 insertions(+), 34 deletions(-)

diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 96b745e3f56c..276cc21619ec 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -106,6 +106,21 @@ int xen_cpuhp_setup(int (*cpu_up_prepare_cb)(unsigned int),
 	return rc >= 0 ? 0 : rc;
 }
 
+static void xen_vcpu_setup_restore(int cpu)
+{
+	/* Any per_cpu(xen_vcpu) is stale, so reset it */
+	xen_vcpu_info_reset(cpu);
+
+	/*
+	 * For PVH and PVHVM, setup online VCPUs only. The rest will
+	 * be handled by hotplug.
+	 */
+	if (xen_pv_domain() ||
+	    (xen_hvm_domain() && cpu_online(cpu))) {
+		xen_vcpu_setup(cpu);
+	}
+}
+
 /*
  * On restore, set the vcpu placement up again.
  * If it fails, then we're in a bad state, since
@@ -117,17 +132,23 @@ void xen_vcpu_restore(void)
 
 	for_each_possible_cpu(cpu) {
 		bool other_cpu = (cpu != smp_processor_id());
-		bool is_up = HYPERVISOR_vcpu_op(VCPUOP_is_up, xen_vcpu_nr(cpu),
-						NULL);
+		bool is_up;
+
+		if (xen_vcpu_nr(cpu) == XEN_VCPU_ID_INVALID)
+			continue;
+
+		/* Only Xen 4.5 and higher support this. */
+		is_up = HYPERVISOR_vcpu_op(VCPUOP_is_up,
+					   xen_vcpu_nr(cpu), NULL) > 0;
 
 		if (other_cpu && is_up &&
 		    HYPERVISOR_vcpu_op(VCPUOP_down, xen_vcpu_nr(cpu), NULL))
 			BUG();
 
-		xen_setup_runstate_info(cpu);
+		if (xen_pv_domain() || xen_feature(XENFEAT_hvm_safe_pvclock))
+			xen_setup_runstate_info(cpu);
 
-		if (xen_have_vcpu_info_placement)
-			xen_vcpu_setup(cpu);
+		xen_vcpu_setup_restore(cpu);
 
 		if (other_cpu && is_up &&
 		    HYPERVISOR_vcpu_op(VCPUOP_up, xen_vcpu_nr(cpu), NULL))
@@ -163,11 +184,11 @@ void xen_vcpu_setup(int cpu)
 	BUG_ON(HYPERVISOR_shared_info == &xen_dummy_shared_info);
 
 	/*
-	 * This path is called twice on PVHVM - first during bootup via
-	 * smp_init -> xen_hvm_cpu_notify, and then if the VCPU is being
-	 * hotplugged: cpu_up -> xen_hvm_cpu_notify.
-	 * As we can only do the VCPUOP_register_vcpu_info once lets
-	 * not over-write its result.
+	 * This path is called on PVHVM at bootup (xen_hvm_smp_prepare_boot_cpu)
+	 * and at restore (xen_vcpu_restore). Also called for hotplugged
+	 * VCPUs (cpu_init -> xen_hvm_cpu_prepare_hvm).
+	 * However, the hypercall can only be done once (see below) so if a VCPU
+	 * is offlined and comes back online then let's not redo the hypercall.
 	 *
 	 * For PV it is called during restore (xen_vcpu_restore) and bootup
 	 * (xen_setup_vcpu_info_placement). The hotplug mechanism does not
@@ -178,8 +199,6 @@ void xen_vcpu_setup(int cpu)
 			return;
 	}
 
-	xen_vcpu_info_reset(cpu);
-
 	if (xen_have_vcpu_info_placement) {
 		vcpup = &per_cpu(xen_vcpu_info, cpu);
 		info.mfn = arbitrary_virt_to_mfn(vcpup);
@@ -214,7 +233,7 @@ void xen_vcpu_setup(int cpu)
 	if (!xen_have_vcpu_info_placement) {
 		if (cpu >= MAX_VIRT_CPUS)
 			clamp_max_cpus();
-		return;
+		xen_vcpu_info_reset(cpu);
 	}
 }
 
diff --git a/arch/x86/xen/enlighten_hvm.c b/arch/x86/xen/enlighten_hvm.c
index eb53da6547ee..ba1afadb2512 100644
--- a/arch/x86/xen/enlighten_hvm.c
+++ b/arch/x86/xen/enlighten_hvm.c
@@ -20,7 +20,6 @@
 
 void __ref xen_hvm_init_shared_info(void)
 {
-	int cpu;
 	struct xen_add_to_physmap xatp;
 	static struct shared_info *shared_info_page;
 
@@ -35,18 +34,6 @@ void __ref xen_hvm_init_shared_info(void)
 		BUG();
 
 	HYPERVISOR_shared_info = (struct shared_info *)shared_info_page;
-
-	/* xen_vcpu is a pointer to the vcpu_info struct in the shared_info
-	 * page, we use it in the event channel upcall and in some pvclock
-	 * related functions. We don't need the vcpu_info placement
-	 * optimizations because we don't use any pv_mmu or pv_irq op on
-	 * HVM.
-	 * When xen_hvm_init_shared_info is run at boot time only vcpu 0 is
-	 * online but xen_hvm_init_shared_info is run at resume time too and
-	 * in that case multiple vcpus might be online. */
-	for_each_online_cpu(cpu) {
-		xen_vcpu_info_reset(cpu);
-	}
 }
 
 static void __init init_hvm_pv_info(void)
@@ -150,6 +137,13 @@ static void __init xen_hvm_guest_init(void)
 
 	xen_hvm_init_shared_info();
 
+	/*
+	 * xen_vcpu is a pointer to the vcpu_info struct in the shared_info
+	 * page, we use it in the event channel upcall and in some pvclock
+	 * related functions.
+	 */
+	xen_vcpu_info_reset(0);
+
 	xen_panic_handler_init();
 
 	if (xen_feature(XENFEAT_hvm_callback_vector))
diff --git a/arch/x86/xen/smp_hvm.c b/arch/x86/xen/smp_hvm.c
index 9e0fb9a015d4..6c8a805819ff 100644
--- a/arch/x86/xen/smp_hvm.c
+++ b/arch/x86/xen/smp_hvm.c
@@ -28,10 +28,20 @@ static void __init xen_hvm_smp_prepare_boot_cpu(void)
 
 static void __init xen_hvm_smp_prepare_cpus(unsigned int max_cpus)
 {
+	int cpu;
+
 	native_smp_prepare_cpus(max_cpus);
 	WARN_ON(xen_smp_intr_init(0));
 
 	xen_init_lock_cpu(0);
+
+	for_each_possible_cpu(cpu) {
+		if (cpu == 0)
+			continue;
+
+		/* Set default vcpu_id to make sure that we don't use cpu-0's */
+		per_cpu(xen_vcpu_id, cpu) = XEN_VCPU_ID_INVALID;
+	}
 }
 
 #ifdef CONFIG_HOTPLUG_CPU
diff --git a/arch/x86/xen/suspend_hvm.c b/arch/x86/xen/suspend_hvm.c
index 01afcadde50a..484999416d8b 100644
--- a/arch/x86/xen/suspend_hvm.c
+++ b/arch/x86/xen/suspend_hvm.c
@@ -8,15 +8,10 @@
 
 void xen_hvm_post_suspend(int suspend_cancelled)
 {
-	int cpu;
-
-	if (!suspend_cancelled)
+	if (!suspend_cancelled) {
 		xen_hvm_init_shared_info();
+		xen_vcpu_restore();
+	}
 	xen_callback_vector();
 	xen_unplug_emulated_devices();
-	if (xen_feature(XENFEAT_hvm_safe_pvclock)) {
-		for_each_online_cpu(cpu) {
-			xen_setup_runstate_info(cpu);
-		}
-	}
 }
diff --git a/include/xen/xen-ops.h b/include/xen/xen-ops.h
index c44a2ee8c8f8..218e6aae5433 100644
--- a/include/xen/xen-ops.h
+++ b/include/xen/xen-ops.h
@@ -15,6 +15,8 @@ static inline uint32_t xen_vcpu_nr(int cpu)
 	return per_cpu(xen_vcpu_id, cpu);
 }
 
+#define XEN_VCPU_ID_INVALID U32_MAX
+
 void xen_arch_pre_suspend(void);
 void xen_arch_post_suspend(int suspend_cancelled);
 
-- 
2.7.4