[PATCH v10 12/20] x86/VPMU: Initialize PMU for PV(H) guests

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

Code for initializing/tearing down PMU for PV guests

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Acked-by: Kevin Tian <kevin.tian@intel.com>
Reviewed-by: Dietmar Hahn <dietmar.hahn@ts.fujitsu.com>
Tested-by: Dietmar Hahn <dietmar.hahn@ts.fujitsu.com>
---
 xen/arch/x86/hvm/hvm.c            |  3 +-
 xen/arch/x86/hvm/svm/svm.c        |  4 +-
 xen/arch/x86/hvm/svm/vpmu.c       | 43 +++++++++++++-------
 xen/arch/x86/hvm/vmx/vmx.c        |  4 +-
 xen/arch/x86/hvm/vmx/vpmu_core2.c | 83 ++++++++++++++++++++++++++++-----------
 xen/arch/x86/hvm/vpmu.c           | 83 ++++++++++++++++++++++++++++++++++++++-
 xen/common/event_channel.c        |  1 +
 xen/include/asm-x86/hvm/vpmu.h    |  1 +
 xen/include/public/pmu.h          |  2 +
 xen/include/public/xen.h          |  1 +
 10 files changed, 183 insertions(+), 42 deletions(-)

diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 83e6fae..5b34689 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -4832,7 +4832,8 @@ static hvm_hypercall_t *const pvh_hypercall64_table[NR_hypercalls] = {
     [ __HYPERVISOR_physdev_op ]      = (hvm_hypercall_t *)hvm_physdev_op,
     HYPERCALL(hvm_op),
     HYPERCALL(sysctl),
-    HYPERCALL(domctl)
+    HYPERCALL(domctl),
+    HYPERCALL(xenpmu_op)
 };
 
 int hvm_do_hypercall(struct cpu_user_regs *regs)
diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
index b5188e6..8b9a34e 100644
--- a/xen/arch/x86/hvm/svm/svm.c
+++ b/xen/arch/x86/hvm/svm/svm.c
@@ -1158,7 +1158,9 @@ static int svm_vcpu_initialise(struct vcpu *v)
         return rc;
     }
 
-    vpmu_initialise(v);
+    /* PVH's VPMU is initialized via hypercall */
+    if ( is_hvm_domain(v->domain) )
+        vpmu_initialise(v);
 
     svm_guest_osvw_init(v);
 
diff --git a/xen/arch/x86/hvm/svm/vpmu.c b/xen/arch/x86/hvm/svm/vpmu.c
index 37d8228..be3ab27 100644
--- a/xen/arch/x86/hvm/svm/vpmu.c
+++ b/xen/arch/x86/hvm/svm/vpmu.c
@@ -362,6 +362,7 @@ static int amd_vpmu_initialise(struct vcpu *v)
     struct xen_pmu_amd_ctxt *ctxt;
     struct vpmu_struct *vpmu = vcpu_vpmu(v);
     uint8_t family = current_cpu_data.x86;
+    unsigned int regs_size;
 
     if ( vpmu_is_set(vpmu, VPMU_CONTEXT_ALLOCATED) )
         return 0;
@@ -389,14 +390,26 @@ static int amd_vpmu_initialise(struct vcpu *v)
 	 }
     }
 
-    ctxt = xzalloc_bytes(sizeof(struct xen_pmu_amd_ctxt) +
-                         2 * sizeof(uint64_t) * AMD_MAX_COUNTERS);
-    if ( !ctxt )
+    regs_size = 2 * sizeof(uint64_t) * AMD_MAX_COUNTERS;
+    if ( is_hvm_domain(v->domain) )
     {
-        gdprintk(XENLOG_WARNING, "Insufficient memory for PMU, "
-            " PMU feature is unavailable on domain %d vcpu %d.\n",
-            v->vcpu_id, v->domain->domain_id);
-        return -ENOMEM;
+        ctxt = xzalloc_bytes(sizeof(struct xen_pmu_amd_ctxt) + regs_size);
+        if ( !ctxt )
+        {
+            gdprintk(XENLOG_WARNING, "Insufficient memory for PMU, "
+                "PMU feature is unavailable\n");
+            return -ENOMEM;
+        }
+    }
+    else
+    {
+        if ( sizeof(struct xen_pmu_data) + regs_size > PAGE_SIZE )
+        {
+            gdprintk(XENLOG_WARNING,
+                    "Register bank does not fit into VPMU shared page\n");
+            return -ENOSPC;
+        }
+        ctxt = &v->arch.vpmu.xenpmu_data->pmu.c.amd;
     }
 
     ctxt->counters = sizeof(struct xen_pmu_amd_ctxt);
@@ -415,17 +428,19 @@ static void amd_vpmu_destroy(struct vcpu *v)
     if ( !vpmu_is_set(vpmu, VPMU_CONTEXT_ALLOCATED) )
         return;
 
-    if ( has_hvm_container_domain(v->domain) && is_msr_bitmap_on(vpmu) )
-        amd_vpmu_unset_msr_bitmap(v);
+    if ( has_hvm_container_domain(v->domain) )
+    {
+        if ( is_msr_bitmap_on(vpmu) )
+            amd_vpmu_unset_msr_bitmap(v);
 
-    xfree(vpmu->context);
-    vpmu_reset(vpmu, VPMU_CONTEXT_ALLOCATED);
+        if ( is_hvm_domain(v->domain) )
+            xfree(vpmu->context);
 
-    if ( vpmu_is_set(vpmu, VPMU_RUNNING) )
-    {
-        vpmu_reset(vpmu, VPMU_RUNNING);
         release_pmu_ownship(PMU_OWNER_HVM);
     }
+
+    vpmu->context = NULL;
+    vpmu_clear(vpmu);
 }
 
 /* VPMU part of the 'q' keyhandler */
diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
index aaff87b..6ca7c32 100644
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -116,7 +116,9 @@ static int vmx_vcpu_initialise(struct vcpu *v)
         return rc;
     }
 
-    vpmu_initialise(v);
+    /* PVH's VPMU is initialized via hypercall */
+    if ( is_hvm_domain(v->domain) )
+        vpmu_initialise(v);
 
     vmx_install_vlapic_mapping(v);
 
diff --git a/xen/arch/x86/hvm/vmx/vpmu_core2.c b/xen/arch/x86/hvm/vmx/vpmu_core2.c
index f5d85e4..5c9fa19 100644
--- a/xen/arch/x86/hvm/vmx/vpmu_core2.c
+++ b/xen/arch/x86/hvm/vmx/vpmu_core2.c
@@ -356,25 +356,45 @@ static int core2_vpmu_alloc_resource(struct vcpu *v)
     struct vpmu_struct *vpmu = vcpu_vpmu(v);
     struct xen_pmu_intel_ctxt *core2_vpmu_cxt = NULL;
     uint64_t *p = NULL;
+    unsigned int regs_size;
 
-    if ( !acquire_pmu_ownership(PMU_OWNER_HVM) )
-        return 0;
-
-    wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0);
-    if ( vmx_add_msr(MSR_CORE_PERF_GLOBAL_CTRL, VMX_HOST_MSR) )
+    p = xzalloc_bytes(sizeof(uint64_t));
+    if ( !p )
         goto out_err;
 
-    if ( vmx_add_msr(MSR_CORE_PERF_GLOBAL_CTRL, VMX_GUEST_MSR) )
-        goto out_err;
-    vmx_write_guest_msr(MSR_CORE_PERF_GLOBAL_CTRL, 0);
-
-    core2_vpmu_cxt = xzalloc_bytes(sizeof(struct xen_pmu_intel_ctxt) +
-                                   sizeof(uint64_t) * fixed_pmc_cnt +
-                                   sizeof(struct xen_pmu_cntr_pair) *
-                                   arch_pmc_cnt);
-    p = xzalloc(uint64_t);
-    if ( !core2_vpmu_cxt || !p )
-        goto out_err;
+    if ( has_hvm_container_domain(v->domain) )
+    {
+        if ( is_hvm_domain(v->domain) && !acquire_pmu_ownership(PMU_OWNER_HVM) )
+            goto out_err;
+
+        wrmsrl(MSR_CORE_PERF_GLOBAL_CTRL, 0);
+        if ( vmx_add_msr(MSR_CORE_PERF_GLOBAL_CTRL, VMX_HOST_MSR) )
+            goto out_err_hvm;
+        if ( vmx_add_msr(MSR_CORE_PERF_GLOBAL_CTRL, VMX_GUEST_MSR) )
+            goto out_err_hvm;
+        vmx_write_guest_msr(MSR_CORE_PERF_GLOBAL_CTRL, 0);
+    }
+
+    regs_size = sizeof(uint64_t) * fixed_pmc_cnt +
+                sizeof(struct xen_pmu_cntr_pair) * arch_pmc_cnt;
+    if ( is_hvm_domain(v->domain) )
+    {
+        core2_vpmu_cxt = xzalloc_bytes(sizeof(struct xen_pmu_intel_ctxt) +
+                                       regs_size);
+        if ( !core2_vpmu_cxt )
+            goto out_err_hvm;
+    }
+    else
+    {
+        if ( sizeof(struct xen_pmu_data) + regs_size > PAGE_SIZE )
+        {
+            printk(XENLOG_WARNING
+                   "Register bank does not fit into VPMU share page\n");
+            goto out_err_hvm;
+        }
+
+        core2_vpmu_cxt = &v->arch.vpmu.xenpmu_data->pmu.c.intel;
+    }
 
     core2_vpmu_cxt->fixed_counters = sizeof(struct xen_pmu_intel_ctxt);
     core2_vpmu_cxt->arch_counters = core2_vpmu_cxt->fixed_counters +
@@ -387,10 +407,12 @@ static int core2_vpmu_alloc_resource(struct vcpu *v)
 
     return 1;
 
-out_err:
-    release_pmu_ownship(PMU_OWNER_HVM);
-
+ out_err_hvm:
     xfree(core2_vpmu_cxt);
+    if ( is_hvm_domain(v->domain) )
+        release_pmu_ownship(PMU_OWNER_HVM);
+
+ out_err:
     xfree(p);
 
     printk("Failed to allocate VPMU resources for domain %u vcpu %u\n",
@@ -756,6 +778,11 @@ static int core2_vpmu_initialise(struct vcpu *v)
     arch_pmc_cnt = core2_get_arch_pmc_count();
     fixed_pmc_cnt = core2_get_fixed_pmc_count();
     check_pmc_quirk();
+
+    /* PV domains can allocate resources immediately */
+    if ( is_pv_domain(v->domain) && !core2_vpmu_alloc_resource(v) )
+        return -EIO;
+
     return 0;
 }
 
@@ -766,12 +793,20 @@ static void core2_vpmu_destroy(struct vcpu *v)
     if ( !vpmu_is_set(vpmu, VPMU_CONTEXT_ALLOCATED) )
         return;
 
-    xfree(vpmu->context);
+    if ( has_hvm_container_domain(v->domain) )
+    {
+        if ( cpu_has_vmx_msr_bitmap )
+            core2_vpmu_unset_msr_bitmap(v->arch.hvm_vmx.msr_bitmap);
+
+        if ( is_hvm_domain(v->domain) )
+            xfree(vpmu->context);
+
+        release_pmu_ownship(PMU_OWNER_HVM);
+    }
+
     xfree(vpmu->priv_context);
-    if ( has_hvm_container_domain(v->domain) && cpu_has_vmx_msr_bitmap )
-        core2_vpmu_unset_msr_bitmap(v->arch.hvm_vmx.msr_bitmap);
-    release_pmu_ownship(PMU_OWNER_HVM);
-    vpmu_reset(vpmu, VPMU_CONTEXT_ALLOCATED);
+    vpmu->context = NULL;
+    vpmu_clear(vpmu);
 }
 
 struct arch_vpmu_ops core2_vpmu_ops = {
diff --git a/xen/arch/x86/hvm/vpmu.c b/xen/arch/x86/hvm/vpmu.c
index 39d13f6..9c3cf7c 100644
--- a/xen/arch/x86/hvm/vpmu.c
+++ b/xen/arch/x86/hvm/vpmu.c
@@ -26,6 +26,7 @@
 #include <asm/regs.h>
 #include <asm/types.h>
 #include <asm/msr.h>
+#include <asm/p2m.h>
 #include <asm/hvm/support.h>
 #include <asm/hvm/vmx/vmx.h>
 #include <asm/hvm/vmx/vmcs.h>
@@ -252,6 +253,7 @@ void vpmu_initialise(struct vcpu *v)
         vpmu_destroy(v);
     vpmu_clear(vpmu);
     vpmu->context = NULL;
+    vpmu->hw_lapic_lvtpc = PMU_APIC_VECTOR | APIC_LVT_MASKED;
 
     switch ( vendor )
     {
@@ -278,7 +280,74 @@ void vpmu_destroy(struct vcpu *v)
     struct vpmu_struct *vpmu = vcpu_vpmu(v);
 
     if ( vpmu->arch_vpmu_ops && vpmu->arch_vpmu_ops->arch_vpmu_destroy )
+    {
+        /* Unload VPMU first. This will stop counters */
+        on_selected_cpus(cpumask_of(vcpu_vpmu(v)->last_pcpu),
+                         vpmu_save_force, v, 1);
+
         vpmu->arch_vpmu_ops->arch_vpmu_destroy(v);
+    }
+}
+
+static int pvpmu_init(struct domain *d, xen_pmu_params_t *params)
+{
+    struct vcpu *v;
+    struct page_info *page;
+    uint64_t gfn = params->val;
+
+    if ( (params->vcpu >= d->max_vcpus) || (d->vcpu == NULL) ||
+         (d->vcpu[params->vcpu] == NULL) )
+        return -EINVAL;
+
+    page = get_page_from_gfn(d, gfn, NULL, P2M_ALLOC);
+    if ( !page )
+        return -EINVAL;
+
+    if ( !get_page_type(page, PGT_writable_page) )
+    {
+        put_page(page);
+        return -EINVAL;
+    }
+
+    v = d->vcpu[params->vcpu];
+    v->arch.vpmu.xenpmu_data = __map_domain_page_global(page);
+    if ( !v->arch.vpmu.xenpmu_data )
+    {
+        put_page_and_type(page);
+        return -EINVAL;
+    }
+
+    vpmu_initialise(v);
+
+    return 0;
+}
+
+static void pvpmu_finish(struct domain *d, xen_pmu_params_t *params)
+{
+    struct vcpu *v;
+    uint64_t mfn;
+
+    if ( (params->vcpu >= d->max_vcpus) || (d->vcpu == NULL) ||
+         (d->vcpu[params->vcpu] == NULL) )
+        return;
+
+    v = d->vcpu[params->vcpu];
+    if ( v != current )
+        vcpu_pause(v);
+
+    if ( v->arch.vpmu.xenpmu_data )
+    {
+        mfn = domain_page_map_to_mfn(v->arch.vpmu.xenpmu_data);
+        if ( mfn_valid(mfn) )
+        {
+            unmap_domain_page_global(v->arch.vpmu.xenpmu_data);
+            put_page_and_type(mfn_to_page(mfn));
+        }
+    }
+    vpmu_destroy(v);
+
+    if ( v != current )
+        vcpu_unpause(v);
 }
 
 /* Dump some vpmu informations on console. Used in keyhandler dump_domains(). */
@@ -468,7 +537,19 @@ long do_xenpmu_op(int op, XEN_GUEST_HANDLE_PARAM(xen_pmu_params_t) arg)
             return -EFAULT;
         ret = 0;
         break;
-     }
+
+    case XENPMU_init:
+        if ( copy_from_guest(&pmu_params, arg, 1) )
+            return -EFAULT;
+        ret = pvpmu_init(current->domain, &pmu_params);
+        break;
+
+    case XENPMU_finish:
+        if ( copy_from_guest(&pmu_params, arg, 1) )
+            return -EFAULT;
+        pvpmu_finish(current->domain, &pmu_params);
+        break;
+    }
 
     return ret;
 }
diff --git a/xen/common/event_channel.c b/xen/common/event_channel.c
index 7d6de54..a991b2d 100644
--- a/xen/common/event_channel.c
+++ b/xen/common/event_channel.c
@@ -108,6 +108,7 @@ static int virq_is_global(uint32_t virq)
     case VIRQ_TIMER:
     case VIRQ_DEBUG:
     case VIRQ_XENOPROF:
+    case VIRQ_XENPMU:
         rc = 0;
         break;
     case VIRQ_ARCH_0 ... VIRQ_ARCH_7:
diff --git a/xen/include/asm-x86/hvm/vpmu.h b/xen/include/asm-x86/hvm/vpmu.h
index 8572835..a70b1a4 100644
--- a/xen/include/asm-x86/hvm/vpmu.h
+++ b/xen/include/asm-x86/hvm/vpmu.h
@@ -62,6 +62,7 @@ struct vpmu_struct {
     void *context;      /* May be shared with PV guest */
     void *priv_context; /* hypervisor-only */
     struct arch_vpmu_ops *arch_vpmu_ops;
+    xen_pmu_data_t *xenpmu_data;
 };
 
 /* VPMU states */
diff --git a/xen/include/public/pmu.h b/xen/include/public/pmu.h
index 0855005..3d55eb6 100644
--- a/xen/include/public/pmu.h
+++ b/xen/include/public/pmu.h
@@ -25,6 +25,8 @@
 #define XENPMU_mode_set        1
 #define XENPMU_feature_get     2
 #define XENPMU_feature_set     3
+#define XENPMU_init            4
+#define XENPMU_finish          5
 /* ` } */
 
 /* Parameters structure for HYPERVISOR_xenpmu_op call */
diff --git a/xen/include/public/xen.h b/xen/include/public/xen.h
index 0766790..e4d0b79 100644
--- a/xen/include/public/xen.h
+++ b/xen/include/public/xen.h
@@ -161,6 +161,7 @@ DEFINE_XEN_GUEST_HANDLE(xen_ulong_t);
 #define VIRQ_MEM_EVENT  10 /* G. (DOM0) A memory event has occured           */
 #define VIRQ_XC_RESERVED 11 /* G. Reserved for XenClient                     */
 #define VIRQ_ENOMEM     12 /* G. (DOM0) Low on heap memory       */
+#define VIRQ_XENPMU     13 /* V.  PMC interrupt                              */
 
 /* Architecture-specific VIRQ definitions. */
 #define VIRQ_ARCH_0    16
-- 
1.8.1.4