xref: /openbmc/linux/arch/x86/include/asm/perf_event.h (revision 6df696cd)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _ASM_X86_PERF_EVENT_H
3 #define _ASM_X86_PERF_EVENT_H
4 
5 #include <linux/static_call.h>
6 
7 /*
8  * Performance event hw details:
9  */
10 
11 #define INTEL_PMC_MAX_GENERIC				       32
12 #define INTEL_PMC_MAX_FIXED				       16
13 #define INTEL_PMC_IDX_FIXED				       32
14 
15 #define X86_PMC_IDX_MAX					       64
16 
17 #define MSR_ARCH_PERFMON_PERFCTR0			      0xc1
18 #define MSR_ARCH_PERFMON_PERFCTR1			      0xc2
19 
20 #define MSR_ARCH_PERFMON_EVENTSEL0			     0x186
21 #define MSR_ARCH_PERFMON_EVENTSEL1			     0x187
22 
23 #define ARCH_PERFMON_EVENTSEL_EVENT			0x000000FFULL
24 #define ARCH_PERFMON_EVENTSEL_UMASK			0x0000FF00ULL
25 #define ARCH_PERFMON_EVENTSEL_USR			(1ULL << 16)
26 #define ARCH_PERFMON_EVENTSEL_OS			(1ULL << 17)
27 #define ARCH_PERFMON_EVENTSEL_EDGE			(1ULL << 18)
28 #define ARCH_PERFMON_EVENTSEL_PIN_CONTROL		(1ULL << 19)
29 #define ARCH_PERFMON_EVENTSEL_INT			(1ULL << 20)
30 #define ARCH_PERFMON_EVENTSEL_ANY			(1ULL << 21)
31 #define ARCH_PERFMON_EVENTSEL_ENABLE			(1ULL << 22)
32 #define ARCH_PERFMON_EVENTSEL_INV			(1ULL << 23)
33 #define ARCH_PERFMON_EVENTSEL_CMASK			0xFF000000ULL
34 
35 #define HSW_IN_TX					(1ULL << 32)
36 #define HSW_IN_TX_CHECKPOINTED				(1ULL << 33)
37 #define ICL_EVENTSEL_ADAPTIVE				(1ULL << 34)
38 #define ICL_FIXED_0_ADAPTIVE				(1ULL << 32)
39 
40 #define AMD64_EVENTSEL_INT_CORE_ENABLE			(1ULL << 36)
41 #define AMD64_EVENTSEL_GUESTONLY			(1ULL << 40)
42 #define AMD64_EVENTSEL_HOSTONLY				(1ULL << 41)
43 
44 #define AMD64_EVENTSEL_INT_CORE_SEL_SHIFT		37
45 #define AMD64_EVENTSEL_INT_CORE_SEL_MASK		\
46 	(0xFULL << AMD64_EVENTSEL_INT_CORE_SEL_SHIFT)
47 
48 #define AMD64_EVENTSEL_EVENT	\
49 	(ARCH_PERFMON_EVENTSEL_EVENT | (0x0FULL << 32))
50 #define INTEL_ARCH_EVENT_MASK	\
51 	(ARCH_PERFMON_EVENTSEL_UMASK | ARCH_PERFMON_EVENTSEL_EVENT)
52 
53 #define AMD64_L3_SLICE_SHIFT				48
54 #define AMD64_L3_SLICE_MASK				\
55 	(0xFULL << AMD64_L3_SLICE_SHIFT)
56 #define AMD64_L3_SLICEID_MASK				\
57 	(0x7ULL << AMD64_L3_SLICE_SHIFT)
58 
59 #define AMD64_L3_THREAD_SHIFT				56
60 #define AMD64_L3_THREAD_MASK				\
61 	(0xFFULL << AMD64_L3_THREAD_SHIFT)
62 #define AMD64_L3_F19H_THREAD_MASK			\
63 	(0x3ULL << AMD64_L3_THREAD_SHIFT)
64 
65 #define AMD64_L3_EN_ALL_CORES				BIT_ULL(47)
66 #define AMD64_L3_EN_ALL_SLICES				BIT_ULL(46)
67 
68 #define AMD64_L3_COREID_SHIFT				42
69 #define AMD64_L3_COREID_MASK				\
70 	(0x7ULL << AMD64_L3_COREID_SHIFT)
71 
72 #define X86_RAW_EVENT_MASK		\
73 	(ARCH_PERFMON_EVENTSEL_EVENT |	\
74 	 ARCH_PERFMON_EVENTSEL_UMASK |	\
75 	 ARCH_PERFMON_EVENTSEL_EDGE  |	\
76 	 ARCH_PERFMON_EVENTSEL_INV   |	\
77 	 ARCH_PERFMON_EVENTSEL_CMASK)
78 #define X86_ALL_EVENT_FLAGS  			\
79 	(ARCH_PERFMON_EVENTSEL_EDGE |  		\
80 	 ARCH_PERFMON_EVENTSEL_INV | 		\
81 	 ARCH_PERFMON_EVENTSEL_CMASK | 		\
82 	 ARCH_PERFMON_EVENTSEL_ANY | 		\
83 	 ARCH_PERFMON_EVENTSEL_PIN_CONTROL | 	\
84 	 HSW_IN_TX | 				\
85 	 HSW_IN_TX_CHECKPOINTED)
86 #define AMD64_RAW_EVENT_MASK		\
87 	(X86_RAW_EVENT_MASK          |  \
88 	 AMD64_EVENTSEL_EVENT)
89 #define AMD64_RAW_EVENT_MASK_NB		\
90 	(AMD64_EVENTSEL_EVENT        |  \
91 	 ARCH_PERFMON_EVENTSEL_UMASK)
92 
93 #define AMD64_PERFMON_V2_EVENTSEL_EVENT_NB	\
94 	(AMD64_EVENTSEL_EVENT	|		\
95 	 GENMASK_ULL(37, 36))
96 
97 #define AMD64_PERFMON_V2_EVENTSEL_UMASK_NB	\
98 	(ARCH_PERFMON_EVENTSEL_UMASK	|	\
99 	 GENMASK_ULL(27, 24))
100 
101 #define AMD64_PERFMON_V2_RAW_EVENT_MASK_NB		\
102 	(AMD64_PERFMON_V2_EVENTSEL_EVENT_NB	|	\
103 	 AMD64_PERFMON_V2_EVENTSEL_UMASK_NB)
104 
105 #define AMD64_NUM_COUNTERS				4
106 #define AMD64_NUM_COUNTERS_CORE				6
107 #define AMD64_NUM_COUNTERS_NB				4
108 
109 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL		0x3c
110 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK		(0x00 << 8)
111 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX		0
112 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT \
113 		(1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX))
114 
115 #define ARCH_PERFMON_BRANCH_MISSES_RETIRED		6
116 #define ARCH_PERFMON_EVENTS_COUNT			7
117 
118 #define PEBS_DATACFG_MEMINFO	BIT_ULL(0)
119 #define PEBS_DATACFG_GP	BIT_ULL(1)
120 #define PEBS_DATACFG_XMMS	BIT_ULL(2)
121 #define PEBS_DATACFG_LBRS	BIT_ULL(3)
122 #define PEBS_DATACFG_LBR_SHIFT	24
123 
124 /* Steal the highest bit of pebs_data_cfg for SW usage */
125 #define PEBS_UPDATE_DS_SW	BIT_ULL(63)
126 
127 /*
128  * Intel "Architectural Performance Monitoring" CPUID
129  * detection/enumeration details:
130  */
131 union cpuid10_eax {
132 	struct {
133 		unsigned int version_id:8;
134 		unsigned int num_counters:8;
135 		unsigned int bit_width:8;
136 		unsigned int mask_length:8;
137 	} split;
138 	unsigned int full;
139 };
140 
141 union cpuid10_ebx {
142 	struct {
143 		unsigned int no_unhalted_core_cycles:1;
144 		unsigned int no_instructions_retired:1;
145 		unsigned int no_unhalted_reference_cycles:1;
146 		unsigned int no_llc_reference:1;
147 		unsigned int no_llc_misses:1;
148 		unsigned int no_branch_instruction_retired:1;
149 		unsigned int no_branch_misses_retired:1;
150 	} split;
151 	unsigned int full;
152 };
153 
154 union cpuid10_edx {
155 	struct {
156 		unsigned int num_counters_fixed:5;
157 		unsigned int bit_width_fixed:8;
158 		unsigned int reserved1:2;
159 		unsigned int anythread_deprecated:1;
160 		unsigned int reserved2:16;
161 	} split;
162 	unsigned int full;
163 };
164 
165 /*
166  * Intel "Architectural Performance Monitoring extension" CPUID
167  * detection/enumeration details:
168  */
169 #define ARCH_PERFMON_EXT_LEAF			0x00000023
170 #define ARCH_PERFMON_NUM_COUNTER_LEAF_BIT	0x1
171 #define ARCH_PERFMON_NUM_COUNTER_LEAF		0x1
172 
173 /*
174  * Intel Architectural LBR CPUID detection/enumeration details:
175  */
176 union cpuid28_eax {
177 	struct {
178 		/* Supported LBR depth values */
179 		unsigned int	lbr_depth_mask:8;
180 		unsigned int	reserved:22;
181 		/* Deep C-state Reset */
182 		unsigned int	lbr_deep_c_reset:1;
183 		/* IP values contain LIP */
184 		unsigned int	lbr_lip:1;
185 	} split;
186 	unsigned int		full;
187 };
188 
189 union cpuid28_ebx {
190 	struct {
191 		/* CPL Filtering Supported */
192 		unsigned int    lbr_cpl:1;
193 		/* Branch Filtering Supported */
194 		unsigned int    lbr_filter:1;
195 		/* Call-stack Mode Supported */
196 		unsigned int    lbr_call_stack:1;
197 	} split;
198 	unsigned int            full;
199 };
200 
201 union cpuid28_ecx {
202 	struct {
203 		/* Mispredict Bit Supported */
204 		unsigned int    lbr_mispred:1;
205 		/* Timed LBRs Supported */
206 		unsigned int    lbr_timed_lbr:1;
207 		/* Branch Type Field Supported */
208 		unsigned int    lbr_br_type:1;
209 	} split;
210 	unsigned int            full;
211 };
212 
213 /*
214  * AMD "Extended Performance Monitoring and Debug" CPUID
215  * detection/enumeration details:
216  */
217 union cpuid_0x80000022_ebx {
218 	struct {
219 		/* Number of Core Performance Counters */
220 		unsigned int	num_core_pmc:4;
221 		/* Number of available LBR Stack Entries */
222 		unsigned int	lbr_v2_stack_sz:6;
223 		/* Number of Data Fabric Counters */
224 		unsigned int	num_df_pmc:6;
225 	} split;
226 	unsigned int		full;
227 };
228 
229 struct x86_pmu_capability {
230 	int		version;
231 	int		num_counters_gp;
232 	int		num_counters_fixed;
233 	int		bit_width_gp;
234 	int		bit_width_fixed;
235 	unsigned int	events_mask;
236 	int		events_mask_len;
237 	unsigned int	pebs_ept	:1;
238 };
239 
240 /*
241  * Fixed-purpose performance events:
242  */
243 
244 /* RDPMC offset for Fixed PMCs */
245 #define INTEL_PMC_FIXED_RDPMC_BASE		(1 << 30)
246 #define INTEL_PMC_FIXED_RDPMC_METRICS		(1 << 29)
247 
248 /*
249  * All the fixed-mode PMCs are configured via this single MSR:
250  */
251 #define MSR_ARCH_PERFMON_FIXED_CTR_CTRL	0x38d
252 
253 /*
254  * There is no event-code assigned to the fixed-mode PMCs.
255  *
256  * For a fixed-mode PMC, which has an equivalent event on a general-purpose
257  * PMC, the event-code of the equivalent event is used for the fixed-mode PMC,
258  * e.g., Instr_Retired.Any and CPU_CLK_Unhalted.Core.
259  *
260  * For a fixed-mode PMC, which doesn't have an equivalent event, a
261  * pseudo-encoding is used, e.g., CPU_CLK_Unhalted.Ref and TOPDOWN.SLOTS.
262  * The pseudo event-code for a fixed-mode PMC must be 0x00.
263  * The pseudo umask-code is 0xX. The X equals the index of the fixed
264  * counter + 1, e.g., the fixed counter 2 has the pseudo-encoding 0x0300.
265  *
266  * The counts are available in separate MSRs:
267  */
268 
269 /* Instr_Retired.Any: */
270 #define MSR_ARCH_PERFMON_FIXED_CTR0	0x309
271 #define INTEL_PMC_IDX_FIXED_INSTRUCTIONS	(INTEL_PMC_IDX_FIXED + 0)
272 
273 /* CPU_CLK_Unhalted.Core: */
274 #define MSR_ARCH_PERFMON_FIXED_CTR1	0x30a
275 #define INTEL_PMC_IDX_FIXED_CPU_CYCLES	(INTEL_PMC_IDX_FIXED + 1)
276 
277 /* CPU_CLK_Unhalted.Ref: event=0x00,umask=0x3 (pseudo-encoding) */
278 #define MSR_ARCH_PERFMON_FIXED_CTR2	0x30b
279 #define INTEL_PMC_IDX_FIXED_REF_CYCLES	(INTEL_PMC_IDX_FIXED + 2)
280 #define INTEL_PMC_MSK_FIXED_REF_CYCLES	(1ULL << INTEL_PMC_IDX_FIXED_REF_CYCLES)
281 
282 /* TOPDOWN.SLOTS: event=0x00,umask=0x4 (pseudo-encoding) */
283 #define MSR_ARCH_PERFMON_FIXED_CTR3	0x30c
284 #define INTEL_PMC_IDX_FIXED_SLOTS	(INTEL_PMC_IDX_FIXED + 3)
285 #define INTEL_PMC_MSK_FIXED_SLOTS	(1ULL << INTEL_PMC_IDX_FIXED_SLOTS)
286 
287 static inline bool use_fixed_pseudo_encoding(u64 code)
288 {
289 	return !(code & 0xff);
290 }
291 
292 /*
293  * We model BTS tracing as another fixed-mode PMC.
294  *
295  * We choose the value 47 for the fixed index of BTS, since lower
296  * values are used by actual fixed events and higher values are used
297  * to indicate other overflow conditions in the PERF_GLOBAL_STATUS msr.
298  */
299 #define INTEL_PMC_IDX_FIXED_BTS			(INTEL_PMC_IDX_FIXED + 15)
300 
301 /*
302  * The PERF_METRICS MSR is modeled as several magic fixed-mode PMCs, one for
303  * each TopDown metric event.
304  *
305  * Internally the TopDown metric events are mapped to the FxCtr 3 (SLOTS).
306  */
307 #define INTEL_PMC_IDX_METRIC_BASE		(INTEL_PMC_IDX_FIXED + 16)
308 #define INTEL_PMC_IDX_TD_RETIRING		(INTEL_PMC_IDX_METRIC_BASE + 0)
309 #define INTEL_PMC_IDX_TD_BAD_SPEC		(INTEL_PMC_IDX_METRIC_BASE + 1)
310 #define INTEL_PMC_IDX_TD_FE_BOUND		(INTEL_PMC_IDX_METRIC_BASE + 2)
311 #define INTEL_PMC_IDX_TD_BE_BOUND		(INTEL_PMC_IDX_METRIC_BASE + 3)
312 #define INTEL_PMC_IDX_TD_HEAVY_OPS		(INTEL_PMC_IDX_METRIC_BASE + 4)
313 #define INTEL_PMC_IDX_TD_BR_MISPREDICT		(INTEL_PMC_IDX_METRIC_BASE + 5)
314 #define INTEL_PMC_IDX_TD_FETCH_LAT		(INTEL_PMC_IDX_METRIC_BASE + 6)
315 #define INTEL_PMC_IDX_TD_MEM_BOUND		(INTEL_PMC_IDX_METRIC_BASE + 7)
316 #define INTEL_PMC_IDX_METRIC_END		INTEL_PMC_IDX_TD_MEM_BOUND
317 #define INTEL_PMC_MSK_TOPDOWN			((0xffull << INTEL_PMC_IDX_METRIC_BASE) | \
318 						INTEL_PMC_MSK_FIXED_SLOTS)
319 
320 /*
321  * There is no event-code assigned to the TopDown events.
322  *
323  * For the slots event, use the pseudo code of the fixed counter 3.
324  *
325  * For the metric events, the pseudo event-code is 0x00.
326  * The pseudo umask-code starts from the middle of the pseudo event
327  * space, 0x80.
328  */
329 #define INTEL_TD_SLOTS				0x0400	/* TOPDOWN.SLOTS */
330 /* Level 1 metrics */
331 #define INTEL_TD_METRIC_RETIRING		0x8000	/* Retiring metric */
332 #define INTEL_TD_METRIC_BAD_SPEC		0x8100	/* Bad speculation metric */
333 #define INTEL_TD_METRIC_FE_BOUND		0x8200	/* FE bound metric */
334 #define INTEL_TD_METRIC_BE_BOUND		0x8300	/* BE bound metric */
335 /* Level 2 metrics */
336 #define INTEL_TD_METRIC_HEAVY_OPS		0x8400  /* Heavy Operations metric */
337 #define INTEL_TD_METRIC_BR_MISPREDICT		0x8500  /* Branch Mispredict metric */
338 #define INTEL_TD_METRIC_FETCH_LAT		0x8600  /* Fetch Latency metric */
339 #define INTEL_TD_METRIC_MEM_BOUND		0x8700  /* Memory bound metric */
340 
341 #define INTEL_TD_METRIC_MAX			INTEL_TD_METRIC_MEM_BOUND
342 #define INTEL_TD_METRIC_NUM			8
343 
344 static inline bool is_metric_idx(int idx)
345 {
346 	return (unsigned)(idx - INTEL_PMC_IDX_METRIC_BASE) < INTEL_TD_METRIC_NUM;
347 }
348 
349 static inline bool is_topdown_idx(int idx)
350 {
351 	return is_metric_idx(idx) || idx == INTEL_PMC_IDX_FIXED_SLOTS;
352 }
353 
354 #define INTEL_PMC_OTHER_TOPDOWN_BITS(bit)	\
355 			(~(0x1ull << bit) & INTEL_PMC_MSK_TOPDOWN)
356 
357 #define GLOBAL_STATUS_COND_CHG			BIT_ULL(63)
358 #define GLOBAL_STATUS_BUFFER_OVF_BIT		62
359 #define GLOBAL_STATUS_BUFFER_OVF		BIT_ULL(GLOBAL_STATUS_BUFFER_OVF_BIT)
360 #define GLOBAL_STATUS_UNC_OVF			BIT_ULL(61)
361 #define GLOBAL_STATUS_ASIF			BIT_ULL(60)
362 #define GLOBAL_STATUS_COUNTERS_FROZEN		BIT_ULL(59)
363 #define GLOBAL_STATUS_LBRS_FROZEN_BIT		58
364 #define GLOBAL_STATUS_LBRS_FROZEN		BIT_ULL(GLOBAL_STATUS_LBRS_FROZEN_BIT)
365 #define GLOBAL_STATUS_TRACE_TOPAPMI_BIT		55
366 #define GLOBAL_STATUS_TRACE_TOPAPMI		BIT_ULL(GLOBAL_STATUS_TRACE_TOPAPMI_BIT)
367 #define GLOBAL_STATUS_PERF_METRICS_OVF_BIT	48
368 
369 #define GLOBAL_CTRL_EN_PERF_METRICS		48
370 /*
371  * We model guest LBR event tracing as another fixed-mode PMC like BTS.
372  *
373  * We choose bit 58 because it's used to indicate LBR stack frozen state
374  * for architectural perfmon v4, also we unconditionally mask that bit in
375  * the handle_pmi_common(), so it'll never be set in the overflow handling.
376  *
377  * With this fake counter assigned, the guest LBR event user (such as KVM),
378  * can program the LBR registers on its own, and we don't actually do anything
379  * with then in the host context.
380  */
381 #define INTEL_PMC_IDX_FIXED_VLBR	(GLOBAL_STATUS_LBRS_FROZEN_BIT)
382 
383 /*
384  * Pseudo-encoding the guest LBR event as event=0x00,umask=0x1b,
385  * since it would claim bit 58 which is effectively Fixed26.
386  */
387 #define INTEL_FIXED_VLBR_EVENT	0x1b00
388 
389 /*
390  * Adaptive PEBS v4
391  */
392 
393 struct pebs_basic {
394 	u64 format_size;
395 	u64 ip;
396 	u64 applicable_counters;
397 	u64 tsc;
398 };
399 
400 struct pebs_meminfo {
401 	u64 address;
402 	u64 aux;
403 	u64 latency;
404 	u64 tsx_tuning;
405 };
406 
407 struct pebs_gprs {
408 	u64 flags, ip, ax, cx, dx, bx, sp, bp, si, di;
409 	u64 r8, r9, r10, r11, r12, r13, r14, r15;
410 };
411 
412 struct pebs_xmm {
413 	u64 xmm[16*2];	/* two entries for each register */
414 };
415 
416 /*
417  * AMD Extended Performance Monitoring and Debug cpuid feature detection
418  */
419 #define EXT_PERFMON_DEBUG_FEATURES		0x80000022
420 
421 /*
422  * IBS cpuid feature detection
423  */
424 
425 #define IBS_CPUID_FEATURES		0x8000001b
426 
427 /*
428  * Same bit mask as for IBS cpuid feature flags (Fn8000_001B_EAX), but
429  * bit 0 is used to indicate the existence of IBS.
430  */
431 #define IBS_CAPS_AVAIL			(1U<<0)
432 #define IBS_CAPS_FETCHSAM		(1U<<1)
433 #define IBS_CAPS_OPSAM			(1U<<2)
434 #define IBS_CAPS_RDWROPCNT		(1U<<3)
435 #define IBS_CAPS_OPCNT			(1U<<4)
436 #define IBS_CAPS_BRNTRGT		(1U<<5)
437 #define IBS_CAPS_OPCNTEXT		(1U<<6)
438 #define IBS_CAPS_RIPINVALIDCHK		(1U<<7)
439 #define IBS_CAPS_OPBRNFUSE		(1U<<8)
440 #define IBS_CAPS_FETCHCTLEXTD		(1U<<9)
441 #define IBS_CAPS_OPDATA4		(1U<<10)
442 #define IBS_CAPS_ZEN4			(1U<<11)
443 
444 #define IBS_CAPS_DEFAULT		(IBS_CAPS_AVAIL		\
445 					 | IBS_CAPS_FETCHSAM	\
446 					 | IBS_CAPS_OPSAM)
447 
448 /*
449  * IBS APIC setup
450  */
451 #define IBSCTL				0x1cc
452 #define IBSCTL_LVT_OFFSET_VALID		(1ULL<<8)
453 #define IBSCTL_LVT_OFFSET_MASK		0x0F
454 
455 /* IBS fetch bits/masks */
456 #define IBS_FETCH_L3MISSONLY	(1ULL<<59)
457 #define IBS_FETCH_RAND_EN	(1ULL<<57)
458 #define IBS_FETCH_VAL		(1ULL<<49)
459 #define IBS_FETCH_ENABLE	(1ULL<<48)
460 #define IBS_FETCH_CNT		0xFFFF0000ULL
461 #define IBS_FETCH_MAX_CNT	0x0000FFFFULL
462 
463 /*
464  * IBS op bits/masks
465  * The lower 7 bits of the current count are random bits
466  * preloaded by hardware and ignored in software
467  */
468 #define IBS_OP_CUR_CNT		(0xFFF80ULL<<32)
469 #define IBS_OP_CUR_CNT_RAND	(0x0007FULL<<32)
470 #define IBS_OP_CNT_CTL		(1ULL<<19)
471 #define IBS_OP_VAL		(1ULL<<18)
472 #define IBS_OP_ENABLE		(1ULL<<17)
473 #define IBS_OP_L3MISSONLY	(1ULL<<16)
474 #define IBS_OP_MAX_CNT		0x0000FFFFULL
475 #define IBS_OP_MAX_CNT_EXT	0x007FFFFFULL	/* not a register bit mask */
476 #define IBS_OP_MAX_CNT_EXT_MASK	(0x7FULL<<20)	/* separate upper 7 bits */
477 #define IBS_RIP_INVALID		(1ULL<<38)
478 
479 #ifdef CONFIG_X86_LOCAL_APIC
480 extern u32 get_ibs_caps(void);
481 #else
482 static inline u32 get_ibs_caps(void) { return 0; }
483 #endif
484 
485 #ifdef CONFIG_PERF_EVENTS
486 extern void perf_events_lapic_init(void);
487 
488 /*
489  * Abuse bits {3,5} of the cpu eflags register. These flags are otherwise
490  * unused and ABI specified to be 0, so nobody should care what we do with
491  * them.
492  *
493  * EXACT - the IP points to the exact instruction that triggered the
494  *         event (HW bugs exempt).
495  * VM    - original X86_VM_MASK; see set_linear_ip().
496  */
497 #define PERF_EFLAGS_EXACT	(1UL << 3)
498 #define PERF_EFLAGS_VM		(1UL << 5)
499 
500 struct pt_regs;
501 struct x86_perf_regs {
502 	struct pt_regs	regs;
503 	u64		*xmm_regs;
504 };
505 
506 extern unsigned long perf_instruction_pointer(struct pt_regs *regs);
507 extern unsigned long perf_misc_flags(struct pt_regs *regs);
508 #define perf_misc_flags(regs)	perf_misc_flags(regs)
509 
510 #include <asm/stacktrace.h>
511 
512 /*
513  * We abuse bit 3 from flags to pass exact information, see perf_misc_flags
514  * and the comment with PERF_EFLAGS_EXACT.
515  */
516 #define perf_arch_fetch_caller_regs(regs, __ip)		{	\
517 	(regs)->ip = (__ip);					\
518 	(regs)->sp = (unsigned long)__builtin_frame_address(0);	\
519 	(regs)->cs = __KERNEL_CS;				\
520 	regs->flags = 0;					\
521 }
522 
523 struct perf_guest_switch_msr {
524 	unsigned msr;
525 	u64 host, guest;
526 };
527 
528 struct x86_pmu_lbr {
529 	unsigned int	nr;
530 	unsigned int	from;
531 	unsigned int	to;
532 	unsigned int	info;
533 };
534 
535 extern void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap);
536 extern u64 perf_get_hw_event_config(int hw_event);
537 extern void perf_check_microcode(void);
538 extern void perf_clear_dirty_counters(void);
539 extern int x86_perf_rdpmc_index(struct perf_event *event);
540 #else
541 static inline void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap)
542 {
543 	memset(cap, 0, sizeof(*cap));
544 }
545 
546 static inline u64 perf_get_hw_event_config(int hw_event)
547 {
548 	return 0;
549 }
550 
551 static inline void perf_events_lapic_init(void)	{ }
552 static inline void perf_check_microcode(void) { }
553 #endif
554 
555 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_INTEL)
556 extern struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr, void *data);
557 extern void x86_perf_get_lbr(struct x86_pmu_lbr *lbr);
558 #else
559 struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr, void *data);
560 static inline void x86_perf_get_lbr(struct x86_pmu_lbr *lbr)
561 {
562 	memset(lbr, 0, sizeof(*lbr));
563 }
564 #endif
565 
566 #ifdef CONFIG_CPU_SUP_INTEL
567  extern void intel_pt_handle_vmx(int on);
568 #else
569 static inline void intel_pt_handle_vmx(int on)
570 {
571 
572 }
573 #endif
574 
575 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD)
576  extern void amd_pmu_enable_virt(void);
577  extern void amd_pmu_disable_virt(void);
578 
579 #if defined(CONFIG_PERF_EVENTS_AMD_BRS)
580 
581 #define PERF_NEEDS_LOPWR_CB 1
582 
583 /*
584  * architectural low power callback impacts
585  * drivers/acpi/processor_idle.c
586  * drivers/acpi/acpi_pad.c
587  */
588 extern void perf_amd_brs_lopwr_cb(bool lopwr_in);
589 
590 DECLARE_STATIC_CALL(perf_lopwr_cb, perf_amd_brs_lopwr_cb);
591 
592 static __always_inline void perf_lopwr_cb(bool lopwr_in)
593 {
594 	static_call_mod(perf_lopwr_cb)(lopwr_in);
595 }
596 
597 #endif /* PERF_NEEDS_LOPWR_CB */
598 
599 #else
600  static inline void amd_pmu_enable_virt(void) { }
601  static inline void amd_pmu_disable_virt(void) { }
602 #endif
603 
604 #define arch_perf_out_copy_user copy_from_user_nmi
605 
606 #endif /* _ASM_X86_PERF_EVENT_H */
607