aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorAlexander Shishkin <alexander.shishkin@linux.intel.com>2016-03-29 10:43:10 -0400
committerIngo Molnar <mingo@kernel.org>2016-04-28 04:32:42 -0400
commit1c5ac21a0e9bab7fc45d0ba9e11623e9ad99d02e (patch)
treee17bdbe1029aa8ff7a8efa77874947e5b3caa901
parent79c9ce57eb2d5f1497546a3946b4ae21b6fdc438 (diff)
perf/x86/intel/pt: Don't die on VMXON
Some versions of Intel PT do not support tracing across VMXON, more specifically, VMXON will clear TraceEn control bit and any attempt to set it before VMXOFF will throw a #GP, which in the current state of things will crash the kernel. Namely: $ perf record -e intel_pt// kvm -nographic on such a machine will kill it. To avoid this, notify the intel_pt driver before VMXON and after VMXOFF so that it knows when not to enable itself. Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: Gleb Natapov <gleb@kernel.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vince Weaver <vincent.weaver@maine.edu> Cc: hpa@zytor.com Link: http://lkml.kernel.org/r/87oa9dwrfk.fsf@ashishki-desk.ger.corp.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
Diffstat
-rw-r--r--arch/x86/events/intel/pt.c75
-rw-r--r--arch/x86/events/intel/pt.h3
-rw-r--r--arch/x86/include/asm/perf_event.h4
-rw-r--r--arch/x86/kvm/vmx.c4
4 files changed, 75 insertions, 11 deletions
diff --git a/arch/x86/events/intel/pt.c b/arch/x86/events/intel/pt.c
index 6af7cf71d6b2..09a77dbc73c9 100644
--- a/arch/x86/events/intel/pt.c
+++ b/arch/x86/events/intel/pt.c
@@ -136,9 +136,21 @@ static int __init pt_pmu_hw_init(void)
136 struct dev_ext_attribute *de_attrs; 136 struct dev_ext_attribute *de_attrs;
137 struct attribute **attrs; 137 struct attribute **attrs;
138 size_t size; 138 size_t size;
139 u64 reg;
139 int ret; 140 int ret;
140 long i; 141 long i;
141 142
143 if (boot_cpu_has(X86_FEATURE_VMX)) {
144 /*
145 * Intel SDM, 36.5 "Tracing post-VMXON" says that
146 * "IA32_VMX_MISC[bit 14]" being 1 means PT can trace
147 * post-VMXON.
148 */
149 rdmsrl(MSR_IA32_VMX_MISC, reg);
150 if (reg & BIT(14))
151 pt_pmu.vmx = true;
152 }
153
142 attrs = NULL; 154 attrs = NULL;
143 155
144 for (i = 0; i < PT_CPUID_LEAVES; i++) { 156 for (i = 0; i < PT_CPUID_LEAVES; i++) {
@@ -269,20 +281,23 @@ static void pt_config(struct perf_event *event)
269 281
270 reg |= (event->attr.config & PT_CONFIG_MASK); 282 reg |= (event->attr.config & PT_CONFIG_MASK);
271 283
284 event->hw.config = reg;
272 wrmsrl(MSR_IA32_RTIT_CTL, reg); 285 wrmsrl(MSR_IA32_RTIT_CTL, reg);
273} 286}
274 287
275static void pt_config_start(bool start) 288static void pt_config_stop(struct perf_event *event)
276{ 289{
277 u64 ctl; 290 u64 ctl = READ_ONCE(event->hw.config);
291
292 /* may be already stopped by a PMI */
293 if (!(ctl & RTIT_CTL_TRACEEN))
294 return;
278 295
279 rdmsrl(MSR_IA32_RTIT_CTL, ctl); 296 ctl &= ~RTIT_CTL_TRACEEN;
280 if (start)
281 ctl |= RTIT_CTL_TRACEEN;
282 else
283 ctl &= ~RTIT_CTL_TRACEEN;
284 wrmsrl(MSR_IA32_RTIT_CTL, ctl); 297 wrmsrl(MSR_IA32_RTIT_CTL, ctl);
285 298
299 WRITE_ONCE(event->hw.config, ctl);
300
286 /* 301 /*
287 * A wrmsr that disables trace generation serializes other PT 302 * A wrmsr that disables trace generation serializes other PT
288 * registers and causes all data packets to be written to memory, 303 * registers and causes all data packets to be written to memory,
@@ -291,8 +306,7 @@ static void pt_config_start(bool start)
291 * The below WMB, separating data store and aux_head store matches 306 * The below WMB, separating data store and aux_head store matches
292 * the consumer's RMB that separates aux_head load and data load. 307 * the consumer's RMB that separates aux_head load and data load.
293 */ 308 */
294 if (!start) 309 wmb();
295 wmb();
296} 310}
297 311
298static void pt_config_buffer(void *buf, unsigned int topa_idx, 312static void pt_config_buffer(void *buf, unsigned int topa_idx,
@@ -942,11 +956,17 @@ void intel_pt_interrupt(void)
942 if (!ACCESS_ONCE(pt->handle_nmi)) 956 if (!ACCESS_ONCE(pt->handle_nmi))
943 return; 957 return;
944 958
945 pt_config_start(false); 959 /*
960 * If VMX is on and PT does not support it, don't touch anything.
961 */
962 if (READ_ONCE(pt->vmx_on))
963 return;
946 964
947 if (!event) 965 if (!event)
948 return; 966 return;
949 967
968 pt_config_stop(event);
969
950 buf = perf_get_aux(&pt->handle); 970 buf = perf_get_aux(&pt->handle);
951 if (!buf) 971 if (!buf)
952 return; 972 return;
@@ -983,6 +1003,35 @@ void intel_pt_interrupt(void)
983 } 1003 }
984} 1004}
985 1005
1006void intel_pt_handle_vmx(int on)
1007{
1008 struct pt *pt = this_cpu_ptr(&pt_ctx);
1009 struct perf_event *event;
1010 unsigned long flags;
1011
1012 /* PT plays nice with VMX, do nothing */
1013 if (pt_pmu.vmx)
1014 return;
1015
1016 /*
1017 * VMXON will clear RTIT_CTL.TraceEn; we need to make
1018 * sure to not try to set it while VMX is on. Disable
1019 * interrupts to avoid racing with pmu callbacks;
1020 * concurrent PMI should be handled fine.
1021 */
1022 local_irq_save(flags);
1023 WRITE_ONCE(pt->vmx_on, on);
1024
1025 if (on) {
1026 /* prevent pt_config_stop() from writing RTIT_CTL */
1027 event = pt->handle.event;
1028 if (event)
1029 event->hw.config = 0;
1030 }
1031 local_irq_restore(flags);
1032}
1033EXPORT_SYMBOL_GPL(intel_pt_handle_vmx);
1034
986/* 1035/*
987 * PMU callbacks 1036 * PMU callbacks
988 */ 1037 */
@@ -992,6 +1041,9 @@ static void pt_event_start(struct perf_event *event, int mode)
992 struct pt *pt = this_cpu_ptr(&pt_ctx); 1041 struct pt *pt = this_cpu_ptr(&pt_ctx);
993 struct pt_buffer *buf = perf_get_aux(&pt->handle); 1042 struct pt_buffer *buf = perf_get_aux(&pt->handle);
994 1043
1044 if (READ_ONCE(pt->vmx_on))
1045 return;
1046
995 if (!buf || pt_buffer_is_full(buf, pt)) { 1047 if (!buf || pt_buffer_is_full(buf, pt)) {
996 event->hw.state = PERF_HES_STOPPED; 1048 event->hw.state = PERF_HES_STOPPED;
997 return; 1049 return;
@@ -1014,7 +1066,8 @@ static void pt_event_stop(struct perf_event *event, int mode)
1014 * see comment in intel_pt_interrupt(). 1066 * see comment in intel_pt_interrupt().
1015 */ 1067 */
1016 ACCESS_ONCE(pt->handle_nmi) = 0; 1068 ACCESS_ONCE(pt->handle_nmi) = 0;
1017 pt_config_start(false); 1069
1070 pt_config_stop(event);
1018 1071
1019 if (event->hw.state == PERF_HES_STOPPED) 1072 if (event->hw.state == PERF_HES_STOPPED)
1020 return; 1073 return;
diff --git a/arch/x86/events/intel/pt.h b/arch/x86/events/intel/pt.h
index 336878a5d205..3abb5f5cccc8 100644
--- a/arch/x86/events/intel/pt.h
+++ b/arch/x86/events/intel/pt.h
@@ -65,6 +65,7 @@ enum pt_capabilities {
65struct pt_pmu { 65struct pt_pmu {
66 struct pmu pmu; 66 struct pmu pmu;
67 u32 caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES]; 67 u32 caps[PT_CPUID_REGS_NUM * PT_CPUID_LEAVES];
68 bool vmx;
68}; 69};
69 70
70/** 71/**
@@ -107,10 +108,12 @@ struct pt_buffer {
107 * struct pt - per-cpu pt context 108 * struct pt - per-cpu pt context
108 * @handle: perf output handle 109 * @handle: perf output handle
109 * @handle_nmi: do handle PT PMI on this cpu, there's an active event 110 * @handle_nmi: do handle PT PMI on this cpu, there's an active event
111 * @vmx_on: 1 if VMX is ON on this cpu
110 */ 112 */
111struct pt { 113struct pt {
112 struct perf_output_handle handle; 114 struct perf_output_handle handle;
113 int handle_nmi; 115 int handle_nmi;
116 int vmx_on;
114}; 117};
115 118
116#endif /* __INTEL_PT_H__ */ 119#endif /* __INTEL_PT_H__ */
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index 5a2ed3ed2f26..f353061bba1d 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -285,6 +285,10 @@ static inline void perf_events_lapic_init(void) { }
285static inline void perf_check_microcode(void) { } 285static inline void perf_check_microcode(void) { }
286#endif 286#endif
287 287
288#ifdef CONFIG_CPU_SUP_INTEL
289 extern void intel_pt_handle_vmx(int on);
290#endif
291
288#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD) 292#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD)
289 extern void amd_pmu_enable_virt(void); 293 extern void amd_pmu_enable_virt(void);
290 extern void amd_pmu_disable_virt(void); 294 extern void amd_pmu_disable_virt(void);
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index ee1c8a93871c..133679d520af 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -3103,6 +3103,8 @@ static __init int vmx_disabled_by_bios(void)
3103 3103
3104static void kvm_cpu_vmxon(u64 addr) 3104static void kvm_cpu_vmxon(u64 addr)
3105{ 3105{
3106 intel_pt_handle_vmx(1);
3107
3106 asm volatile (ASM_VMX_VMXON_RAX 3108 asm volatile (ASM_VMX_VMXON_RAX
3107 : : "a"(&addr), "m"(addr) 3109 : : "a"(&addr), "m"(addr)
3108 : "memory", "cc"); 3110 : "memory", "cc");
@@ -3172,6 +3174,8 @@ static void vmclear_local_loaded_vmcss(void)
3172static void kvm_cpu_vmxoff(void) 3174static void kvm_cpu_vmxoff(void)
3173{ 3175{
3174 asm volatile (__ex(ASM_VMX_VMXOFF) : : : "cc"); 3176 asm volatile (__ex(ASM_VMX_VMXOFF) : : : "cc");
3177
3178 intel_pt_handle_vmx(0);
3175} 3179}
3176 3180
3177static void hardware_disable(void) 3181static void hardware_disable(void)
generated by cgit v1.2.2 (git 2.25.0) at 2025-03-22 06:03:49 -0400