From 8bf00a529967dafbbb210b377c38a15834d1e979 Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Wed, 5 Oct 2011 14:01:22 +0200 Subject: KVM: VMX: add support for switching of PERF_GLOBAL_CTRL Some cpus have special support for switching PERF_GLOBAL_CTRL msr. Add logic to detect if such support exists and works properly and extend msr switching code to use it if available. Also extend number of generic msr switching entries to 8. Signed-off-by: Gleb Natapov Signed-off-by: Avi Kivity --- arch/x86/kvm/vmx.c | 104 +++++++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 93 insertions(+), 11 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index a0d6bd9..55e849b 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -118,7 +118,7 @@ module_param(ple_gap, int, S_IRUGO); static int ple_window = KVM_VMX_DEFAULT_PLE_WINDOW; module_param(ple_window, int, S_IRUGO); -#define NR_AUTOLOAD_MSRS 1 +#define NR_AUTOLOAD_MSRS 8 #define VMCS02_POOL_SIZE 1 struct vmcs { @@ -622,6 +622,7 @@ static unsigned long *vmx_msr_bitmap_legacy; static unsigned long *vmx_msr_bitmap_longmode; static bool cpu_has_load_ia32_efer; +static bool cpu_has_load_perf_global_ctrl; static DECLARE_BITMAP(vmx_vpid_bitmap, VMX_NR_VPIDS); static DEFINE_SPINLOCK(vmx_vpid_lock); @@ -1191,15 +1192,34 @@ static void update_exception_bitmap(struct kvm_vcpu *vcpu) vmcs_write32(EXCEPTION_BITMAP, eb); } +static void clear_atomic_switch_msr_special(unsigned long entry, + unsigned long exit) +{ + vmcs_clear_bits(VM_ENTRY_CONTROLS, entry); + vmcs_clear_bits(VM_EXIT_CONTROLS, exit); +} + static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr) { unsigned i; struct msr_autoload *m = &vmx->msr_autoload; - if (msr == MSR_EFER && cpu_has_load_ia32_efer) { - vmcs_clear_bits(VM_ENTRY_CONTROLS, VM_ENTRY_LOAD_IA32_EFER); - vmcs_clear_bits(VM_EXIT_CONTROLS, VM_EXIT_LOAD_IA32_EFER); - return; + switch (msr) { + case MSR_EFER: + if (cpu_has_load_ia32_efer) { + clear_atomic_switch_msr_special(VM_ENTRY_LOAD_IA32_EFER, + VM_EXIT_LOAD_IA32_EFER); + return; + } + break; + case MSR_CORE_PERF_GLOBAL_CTRL: + if (cpu_has_load_perf_global_ctrl) { + clear_atomic_switch_msr_special( + VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, + VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL); + return; + } + break; } for (i = 0; i < m->nr; ++i) @@ -1215,18 +1235,44 @@ static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr) vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->nr); } +static void add_atomic_switch_msr_special(unsigned long entry, + unsigned long exit, unsigned long guest_val_vmcs, + unsigned long host_val_vmcs, u64 guest_val, u64 host_val) +{ + vmcs_write64(guest_val_vmcs, guest_val); + vmcs_write64(host_val_vmcs, host_val); + vmcs_set_bits(VM_ENTRY_CONTROLS, entry); + vmcs_set_bits(VM_EXIT_CONTROLS, exit); +} + static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr, u64 guest_val, u64 host_val) { unsigned i; struct msr_autoload *m = &vmx->msr_autoload; - if (msr == MSR_EFER && cpu_has_load_ia32_efer) { - vmcs_write64(GUEST_IA32_EFER, guest_val); - vmcs_write64(HOST_IA32_EFER, host_val); - vmcs_set_bits(VM_ENTRY_CONTROLS, VM_ENTRY_LOAD_IA32_EFER); - vmcs_set_bits(VM_EXIT_CONTROLS, VM_EXIT_LOAD_IA32_EFER); - return; + switch (msr) { + case MSR_EFER: + if (cpu_has_load_ia32_efer) { + add_atomic_switch_msr_special(VM_ENTRY_LOAD_IA32_EFER, + VM_EXIT_LOAD_IA32_EFER, + GUEST_IA32_EFER, + HOST_IA32_EFER, + guest_val, host_val); + return; + } + break; + case MSR_CORE_PERF_GLOBAL_CTRL: + if (cpu_has_load_perf_global_ctrl) { + add_atomic_switch_msr_special( + VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, + VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL, + GUEST_IA32_PERF_GLOBAL_CTRL, + HOST_IA32_PERF_GLOBAL_CTRL, + guest_val, host_val); + return; + } + break; } for (i = 0; i < m->nr; ++i) @@ -2455,6 +2501,42 @@ static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf) && allow_1_setting(MSR_IA32_VMX_EXIT_CTLS, VM_EXIT_LOAD_IA32_EFER); + cpu_has_load_perf_global_ctrl = + allow_1_setting(MSR_IA32_VMX_ENTRY_CTLS, + VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL) + && allow_1_setting(MSR_IA32_VMX_EXIT_CTLS, + VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL); + + /* + * Some cpus support VM_ENTRY_(LOAD|SAVE)_IA32_PERF_GLOBAL_CTRL + * but due to arrata below it can't be used. Workaround is to use + * msr load mechanism to switch IA32_PERF_GLOBAL_CTRL. + * + * VM Exit May Incorrectly Clear IA32_PERF_GLOBAL_CTRL [34:32] + * + * AAK155 (model 26) + * AAP115 (model 30) + * AAT100 (model 37) + * BC86,AAY89,BD102 (model 44) + * BA97 (model 46) + * + */ + if (cpu_has_load_perf_global_ctrl && boot_cpu_data.x86 == 0x6) { + switch (boot_cpu_data.x86_model) { + case 26: + case 30: + case 37: + case 44: + case 46: + cpu_has_load_perf_global_ctrl = false; + printk_once(KERN_WARNING"kvm: VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL " + "does not work properly. Using workaround\n"); + break; + default: + break; + } + } + return 0; } -- cgit v1.1 From d7cd97964ba6d70c558348bd2c87290dce885583 Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Wed, 5 Oct 2011 14:01:23 +0200 Subject: KVM: VMX: Add support for guest/host-only profiling Support guest/host-only profiling by switch perf msrs on a guest entry if needed. Signed-off-by: Gleb Natapov Signed-off-by: Avi Kivity --- arch/x86/kvm/vmx.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'arch/x86') diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 55e849b..98f4b0b 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -39,6 +39,7 @@ #include #include #include +#include #include "trace.h" @@ -6050,6 +6051,24 @@ static void vmx_cancel_injection(struct kvm_vcpu *vcpu) vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); } +static void atomic_switch_perf_msrs(struct vcpu_vmx *vmx) +{ + int i, nr_msrs; + struct perf_guest_switch_msr *msrs; + + msrs = perf_guest_get_msrs(&nr_msrs); + + if (!msrs) + return; + + for (i = 0; i < nr_msrs; i++) + if (msrs[i].host == msrs[i].guest) + clear_atomic_switch_msr(vmx, msrs[i].msr); + else + add_atomic_switch_msr(vmx, msrs[i].msr, msrs[i].guest, + msrs[i].host); +} + #ifdef CONFIG_X86_64 #define R "r" #define Q "q" @@ -6099,6 +6118,8 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu) if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) vmx_set_interrupt_shadow(vcpu, 0); + atomic_switch_perf_msrs(vmx); + vmx->__launched = vmx->loaded_vmcs->launched; asm( /* Store host registers */ -- cgit v1.1 From e7fc6f93b4242b2b566f0070709e27257d6da8a2 Mon Sep 17 00:00:00 2001 From: Gleb Natapov Date: Wed, 5 Oct 2011 14:01:24 +0200 Subject: KVM: VMX: Check for automatic switch msr table overflow Signed-off-by: Gleb Natapov Signed-off-by: Avi Kivity --- arch/x86/kvm/vmx.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'arch/x86') diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 98f4b0b..579a0b5 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -1280,7 +1280,11 @@ static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr, if (m->guest[i].index == msr) break; - if (i == m->nr) { + if (i == NR_AUTOLOAD_MSRS) { + printk_once(KERN_WARNING"Not enough mst switch entries. " + "Can't add msr %x\n", msr); + return; + } else if (i == m->nr) { ++m->nr; vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->nr); vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->nr); -- cgit v1.1 From 95ef1e52922cf75b1ea2eae54ef886f2cc47eecb Mon Sep 17 00:00:00 2001 From: Avi Kivity Date: Tue, 15 Nov 2011 14:59:07 +0200 Subject: KVM guest: prevent tracing recursion with kvmclock Prevent tracing of preempt_disable() in get_cpu_var() in kvm_clock_read(). When CONFIG_DEBUG_PREEMPT is enabled, preempt_disable/enable() are traced and this causes the function_graph tracer to go into an infinite recursion. By open coding the preempt_disable() around the get_cpu_var(), we can use the notrace version which prevents preempt_disable/enable() from being traced and prevents the recursion. Based on a similar patch for Xen from Jeremy Fitzhardinge. Tested-by: Gleb Natapov Acked-by: Steven Rostedt Signed-off-by: Avi Kivity --- arch/x86/kernel/kvmclock.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'arch/x86') diff --git a/arch/x86/kernel/kvmclock.c b/arch/x86/kernel/kvmclock.c index c1a0188..44842d7 100644 --- a/arch/x86/kernel/kvmclock.c +++ b/arch/x86/kernel/kvmclock.c @@ -74,9 +74,10 @@ static cycle_t kvm_clock_read(void) struct pvclock_vcpu_time_info *src; cycle_t ret; - src = &get_cpu_var(hv_clock); + preempt_disable_notrace(); + src = &__get_cpu_var(hv_clock); ret = pvclock_clocksource_read(src); - put_cpu_var(hv_clock); + preempt_enable_notrace(); return ret; } -- cgit v1.1