From mboxrd@z Thu Jan 1 00:00:00 1970 From: Boris Ostrovsky Subject: [PATCH v6 11/19] x86/VPMU: Initialize PMU for PV guests Date: Tue, 13 May 2014 11:53:25 -0400 Message-ID: <1399996413-1998-12-git-send-email-boris.ostrovsky@oracle.com> References: <1399996413-1998-1-git-send-email-boris.ostrovsky@oracle.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1399996413-1998-1-git-send-email-boris.ostrovsky@oracle.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: JBeulich@suse.com, kevin.tian@intel.com, dietmar.hahn@ts.fujitsu.com, suravee.suthikulpanit@amd.com Cc: keir@xen.org, andrew.cooper3@citrix.com, donald.d.dugger@intel.com, xen-devel@lists.xen.org, jun.nakajima@intel.com, boris.ostrovsky@oracle.com List-Id: xen-devel@lists.xenproject.org Code for initializing/tearing down PMU for PV guests Signed-off-by: Boris Ostrovsky Acked-by: Kevin Tian Reviewed-by: Dietmar Hahn Tested-by: Dietmar Hahn --- xen/arch/x86/hvm/svm/svm.c | 6 ++- xen/arch/x86/hvm/svm/vpmu.c | 37 +++++++++-------- xen/arch/x86/hvm/vmx/vmx.c | 6 ++- xen/arch/x86/hvm/vmx/vpmu_core2.c | 64 +++++++++++++++++++---------- xen/arch/x86/hvm/vpmu.c | 85 ++++++++++++++++++++++++++++++++++++++- 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 + xen/include/xen/softirq.h | 1 + 10 files changed, 161 insertions(+), 43 deletions(-) diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c index 38d7923..c23db32 100644 --- a/xen/arch/x86/hvm/svm/svm.c +++ b/xen/arch/x86/hvm/svm/svm.c @@ -1150,7 +1150,8 @@ static int svm_vcpu_initialise(struct vcpu *v) return rc; } - vpmu_initialise(v); + if ( is_hvm_domain(v->domain) ) + vpmu_initialise(v); svm_guest_osvw_init(v); @@ -1159,7 +1160,8 @@ static int svm_vcpu_initialise(struct vcpu *v) static void svm_vcpu_destroy(struct vcpu *v) { - vpmu_destroy(v); + if ( is_hvm_domain(v->domain) ) + vpmu_destroy(v); svm_destroy_vmcb(v); passive_domain_destroy(v); } diff --git a/xen/arch/x86/hvm/svm/vpmu.c b/xen/arch/x86/hvm/svm/vpmu.c index 3828c4b..42c3530 100644 --- a/xen/arch/x86/hvm/svm/vpmu.c +++ b/xen/arch/x86/hvm/svm/vpmu.c @@ -381,16 +381,21 @@ static int amd_vpmu_initialise(struct vcpu *v) } } - ctxt = xzalloc_bytes(sizeof(struct xen_pmu_amd_ctxt) + - sizeof(uint64_t) * AMD_MAX_COUNTERS + - sizeof(uint64_t) * AMD_MAX_COUNTERS); - if ( !ctxt ) + if ( !is_pv_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) + + sizeof(uint64_t) * AMD_MAX_COUNTERS + + sizeof(uint64_t) * AMD_MAX_COUNTERS); + if ( !ctxt ) + { + 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; + } } + else + ctxt = &v->arch.vpmu.xenpmu_data->pmu.c.amd; ctxt->counters = sizeof(struct xen_pmu_amd_ctxt); ctxt->ctrls = ctxt->counters + sizeof(uint64_t) * AMD_MAX_COUNTERS; @@ -408,17 +413,17 @@ static void amd_vpmu_destroy(struct vcpu *v) if ( !vpmu_is_set(vpmu, VPMU_CONTEXT_ALLOCATED) ) return; - if ( is_hvm_domain(v->domain) && is_msr_bitmap_on(vpmu) ) - amd_vpmu_unset_msr_bitmap(v); - - xfree(vpmu->context); - vpmu_reset(vpmu, VPMU_CONTEXT_ALLOCATED); - - if ( vpmu_is_set(vpmu, VPMU_RUNNING) ) + if ( is_hvm_domain(v->domain) ) { - vpmu_reset(vpmu, VPMU_RUNNING); + if ( is_msr_bitmap_on(vpmu) ) + amd_vpmu_unset_msr_bitmap(v); + + xfree(vpmu->context); 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 23d58d9..1c9e742 100644 --- a/xen/arch/x86/hvm/vmx/vmx.c +++ b/xen/arch/x86/hvm/vmx/vmx.c @@ -115,7 +115,8 @@ static int vmx_vcpu_initialise(struct vcpu *v) return rc; } - vpmu_initialise(v); + if ( is_hvm_domain(v->domain) ) + vpmu_initialise(v); vmx_install_vlapic_mapping(v); @@ -129,7 +130,8 @@ static int vmx_vcpu_initialise(struct vcpu *v) static void vmx_vcpu_destroy(struct vcpu *v) { vmx_destroy_vmcs(v); - vpmu_destroy(v); + if ( is_hvm_domain(v->domain) ) + vpmu_destroy(v); passive_domain_destroy(v); } diff --git a/xen/arch/x86/hvm/vmx/vpmu_core2.c b/xen/arch/x86/hvm/vmx/vpmu_core2.c index 8d5b26e..b048180 100644 --- a/xen/arch/x86/hvm/vmx/vpmu_core2.c +++ b/xen/arch/x86/hvm/vmx/vpmu_core2.c @@ -360,25 +360,34 @@ static int core2_vpmu_alloc_resource(struct vcpu *v) struct xen_pmu_intel_ctxt *core2_vpmu_cxt = NULL; uint64_t *p = NULL; - 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) ) - 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_bytes(sizeof(uint64_t)); - if ( !core2_vpmu_cxt || !p ) + if ( !p ) goto out_err; + if ( !is_pv_domain(v->domain) ) + { + if ( !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); + + 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); + if ( !core2_vpmu_cxt ) + goto out_err_hvm; + } + else + { + 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 + sizeof(uint64_t) * fixed_pmc_cnt; @@ -390,7 +399,7 @@ static int core2_vpmu_alloc_resource(struct vcpu *v) return 1; -out_err: +out_err_hvm: vmx_rm_msr(MSR_CORE_PERF_GLOBAL_CTRL, VMX_HOST_MSR); vmx_rm_msr(MSR_CORE_PERF_GLOBAL_CTRL, VMX_GUEST_MSR); release_pmu_ownship(PMU_OWNER_HVM); @@ -398,6 +407,7 @@ out_err: xfree(core2_vpmu_cxt); xfree(p); +out_err: printk("Failed to allocate VPMU resources for domain %u vcpu %u\n", v->vcpu_id, v->domain->domain_id); @@ -753,6 +763,10 @@ func_out: 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 1; + return 0; } @@ -764,11 +778,17 @@ static void core2_vpmu_destroy(struct vcpu *v) return; xfree(vpmu->context); - xfree(vpmu->priv_context); - if ( cpu_has_vmx_msr_bitmap && is_hvm_domain(v->domain) ) - core2_vpmu_unset_msr_bitmap(v->arch.hvm_vmx.msr_bitmap); - release_pmu_ownship(PMU_OWNER_HVM); - vpmu_reset(vpmu, VPMU_CONTEXT_ALLOCATED); + + if ( is_hvm_domain(v->domain) ) + { + xfree(vpmu->priv_context); + if ( cpu_has_vmx_msr_bitmap ) + core2_vpmu_unset_msr_bitmap(v->arch.hvm_vmx.msr_bitmap); + release_pmu_ownship(PMU_OWNER_HVM); + } + + 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 d3de8a6..2d152f7 100644 --- a/xen/arch/x86/hvm/vpmu.c +++ b/xen/arch/x86/hvm/vpmu.c @@ -21,10 +21,14 @@ #include #include #include +#include +#include +#include #include #include #include #include +#include #include #include #include @@ -267,7 +271,13 @@ 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, (void *)v, 1); + vpmu->arch_vpmu_ops->arch_vpmu_destroy(v); + } } /* Dump some vpmu informations on console. Used in keyhandler dump_domains(). */ @@ -312,6 +322,67 @@ static void vpmu_unload_all(void) } } +static int pvpmu_init(struct domain *d, xen_pmu_params_t *params) +{ + struct vcpu *v; + struct page_info *page; + uint64_t gfn = params->d.val; + + if ( !is_pv_domain(d) ) + return -EINVAL; + + if ( params->vcpu < 0 || params->vcpu >= d->max_vcpus ) + 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 < 0 || params->vcpu >= d->max_vcpus ) + 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); +} long do_xenpmu_op(int op, XEN_GUEST_HANDLE_PARAM(xen_pmu_params_t) arg) { @@ -372,7 +443,19 @@ 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 6853842..7f9b8c9 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 3dfdab7..438a913 100644 --- a/xen/include/asm-x86/hvm/vpmu.h +++ b/xen/include/asm-x86/hvm/vpmu.h @@ -61,6 +61,7 @@ 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 f91d935..814e061 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 diff --git a/xen/include/xen/softirq.h b/xen/include/xen/softirq.h index 0c0d481..5829fa4 100644 --- a/xen/include/xen/softirq.h +++ b/xen/include/xen/softirq.h @@ -8,6 +8,7 @@ enum { NEW_TLBFLUSH_CLOCK_PERIOD_SOFTIRQ, RCU_SOFTIRQ, TASKLET_SOFTIRQ, + PMU_SOFTIRQ, NR_COMMON_SOFTIRQS }; -- 1.8.1.4