xref: /openbmc/linux/arch/x86/include/asm/perf_event.h (revision d37cf9b63113f13d742713881ce691fc615d8b3b)
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 INTEL_FIXED_BITS_MASK				0xFULL
36 #define INTEL_FIXED_BITS_STRIDE			4
37 #define INTEL_FIXED_0_KERNEL				(1ULL << 0)
38 #define INTEL_FIXED_0_USER				(1ULL << 1)
39 #define INTEL_FIXED_0_ANYTHREAD			(1ULL << 2)
40 #define INTEL_FIXED_0_ENABLE_PMI			(1ULL << 3)
41 
42 #define HSW_IN_TX					(1ULL << 32)
43 #define HSW_IN_TX_CHECKPOINTED				(1ULL << 33)
44 #define ICL_EVENTSEL_ADAPTIVE				(1ULL << 34)
45 #define ICL_FIXED_0_ADAPTIVE				(1ULL << 32)
46 
47 #define intel_fixed_bits_by_idx(_idx, _bits)			\
48 	((_bits) << ((_idx) * INTEL_FIXED_BITS_STRIDE))
49 
50 #define AMD64_EVENTSEL_INT_CORE_ENABLE			(1ULL << 36)
51 #define AMD64_EVENTSEL_GUESTONLY			(1ULL << 40)
52 #define AMD64_EVENTSEL_HOSTONLY				(1ULL << 41)
53 
54 #define AMD64_EVENTSEL_INT_CORE_SEL_SHIFT		37
55 #define AMD64_EVENTSEL_INT_CORE_SEL_MASK		\
56 	(0xFULL << AMD64_EVENTSEL_INT_CORE_SEL_SHIFT)
57 
58 #define AMD64_EVENTSEL_EVENT	\
59 	(ARCH_PERFMON_EVENTSEL_EVENT | (0x0FULL << 32))
60 #define INTEL_ARCH_EVENT_MASK	\
61 	(ARCH_PERFMON_EVENTSEL_UMASK | ARCH_PERFMON_EVENTSEL_EVENT)
62 
63 #define AMD64_L3_SLICE_SHIFT				48
64 #define AMD64_L3_SLICE_MASK				\
65 	(0xFULL << AMD64_L3_SLICE_SHIFT)
66 #define AMD64_L3_SLICEID_MASK				\
67 	(0x7ULL << AMD64_L3_SLICE_SHIFT)
68 
69 #define AMD64_L3_THREAD_SHIFT				56
70 #define AMD64_L3_THREAD_MASK				\
71 	(0xFFULL << AMD64_L3_THREAD_SHIFT)
72 #define AMD64_L3_F19H_THREAD_MASK			\
73 	(0x3ULL << AMD64_L3_THREAD_SHIFT)
74 
75 #define AMD64_L3_EN_ALL_CORES				BIT_ULL(47)
76 #define AMD64_L3_EN_ALL_SLICES				BIT_ULL(46)
77 
78 #define AMD64_L3_COREID_SHIFT				42
79 #define AMD64_L3_COREID_MASK				\
80 	(0x7ULL << AMD64_L3_COREID_SHIFT)
81 
82 #define X86_RAW_EVENT_MASK		\
83 	(ARCH_PERFMON_EVENTSEL_EVENT |	\
84 	 ARCH_PERFMON_EVENTSEL_UMASK |	\
85 	 ARCH_PERFMON_EVENTSEL_EDGE  |	\
86 	 ARCH_PERFMON_EVENTSEL_INV   |	\
87 	 ARCH_PERFMON_EVENTSEL_CMASK)
88 #define X86_ALL_EVENT_FLAGS  			\
89 	(ARCH_PERFMON_EVENTSEL_EDGE |  		\
90 	 ARCH_PERFMON_EVENTSEL_INV | 		\
91 	 ARCH_PERFMON_EVENTSEL_CMASK | 		\
92 	 ARCH_PERFMON_EVENTSEL_ANY | 		\
93 	 ARCH_PERFMON_EVENTSEL_PIN_CONTROL | 	\
94 	 HSW_IN_TX | 				\
95 	 HSW_IN_TX_CHECKPOINTED)
96 #define AMD64_RAW_EVENT_MASK		\
97 	(X86_RAW_EVENT_MASK          |  \
98 	 AMD64_EVENTSEL_EVENT)
99 #define AMD64_RAW_EVENT_MASK_NB		\
100 	(AMD64_EVENTSEL_EVENT        |  \
101 	 ARCH_PERFMON_EVENTSEL_UMASK)
102 
103 #define AMD64_PERFMON_V2_EVENTSEL_EVENT_NB	\
104 	(AMD64_EVENTSEL_EVENT	|		\
105 	 GENMASK_ULL(37, 36))
106 
107 #define AMD64_PERFMON_V2_EVENTSEL_UMASK_NB	\
108 	(ARCH_PERFMON_EVENTSEL_UMASK	|	\
109 	 GENMASK_ULL(27, 24))
110 
111 #define AMD64_PERFMON_V2_RAW_EVENT_MASK_NB		\
112 	(AMD64_PERFMON_V2_EVENTSEL_EVENT_NB	|	\
113 	 AMD64_PERFMON_V2_EVENTSEL_UMASK_NB)
114 
115 #define AMD64_NUM_COUNTERS				4
116 #define AMD64_NUM_COUNTERS_CORE				6
117 #define AMD64_NUM_COUNTERS_NB				4
118 
119 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL		0x3c
120 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK		(0x00 << 8)
121 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX		0
122 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT \
123 		(1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX))
124 
125 #define ARCH_PERFMON_BRANCH_MISSES_RETIRED		6
126 #define ARCH_PERFMON_EVENTS_COUNT			7
127 
128 #define PEBS_DATACFG_MEMINFO	BIT_ULL(0)
129 #define PEBS_DATACFG_GP	BIT_ULL(1)
130 #define PEBS_DATACFG_XMMS	BIT_ULL(2)
131 #define PEBS_DATACFG_LBRS	BIT_ULL(3)
132 #define PEBS_DATACFG_LBR_SHIFT	24
133 
134 /* Steal the highest bit of pebs_data_cfg for SW usage */
135 #define PEBS_UPDATE_DS_SW	BIT_ULL(63)
136 
137 /*
138  * Intel "Architectural Performance Monitoring" CPUID
139  * detection/enumeration details:
140  */
141 union cpuid10_eax {
142 	struct {
143 		unsigned int version_id:8;
144 		unsigned int num_counters:8;
145 		unsigned int bit_width:8;
146 		unsigned int mask_length:8;
147 	} split;
148 	unsigned int full;
149 };
150 
151 union cpuid10_ebx {
152 	struct {
153 		unsigned int no_unhalted_core_cycles:1;
154 		unsigned int no_instructions_retired:1;
155 		unsigned int no_unhalted_reference_cycles:1;
156 		unsigned int no_llc_reference:1;
157 		unsigned int no_llc_misses:1;
158 		unsigned int no_branch_instruction_retired:1;
159 		unsigned int no_branch_misses_retired:1;
160 	} split;
161 	unsigned int full;
162 };
163 
164 union cpuid10_edx {
165 	struct {
166 		unsigned int num_counters_fixed:5;
167 		unsigned int bit_width_fixed:8;
168 		unsigned int reserved1:2;
169 		unsigned int anythread_deprecated:1;
170 		unsigned int reserved2:16;
171 	} split;
172 	unsigned int full;
173 };
174 
175 /*
176  * Intel "Architectural Performance Monitoring extension" CPUID
177  * detection/enumeration details:
178  */
179 #define ARCH_PERFMON_EXT_LEAF			0x00000023
180 #define ARCH_PERFMON_NUM_COUNTER_LEAF		0x1
181 
182 union cpuid35_eax {
183 	struct {
184 		unsigned int	leaf0:1;
185 		/* Counters Sub-Leaf */
186 		unsigned int    cntr_subleaf:1;
187 		/* Auto Counter Reload Sub-Leaf */
188 		unsigned int    acr_subleaf:1;
189 		/* Events Sub-Leaf */
190 		unsigned int    events_subleaf:1;
191 		unsigned int	reserved:28;
192 	} split;
193 	unsigned int            full;
194 };
195 
196 union cpuid35_ebx {
197 	struct {
198 		/* UnitMask2 Supported */
199 		unsigned int    umask2:1;
200 		/* EQ-bit Supported */
201 		unsigned int    eq:1;
202 		unsigned int	reserved:30;
203 	} split;
204 	unsigned int            full;
205 };
206 
207 /*
208  * Intel Architectural LBR CPUID detection/enumeration details:
209  */
210 union cpuid28_eax {
211 	struct {
212 		/* Supported LBR depth values */
213 		unsigned int	lbr_depth_mask:8;
214 		unsigned int	reserved:22;
215 		/* Deep C-state Reset */
216 		unsigned int	lbr_deep_c_reset:1;
217 		/* IP values contain LIP */
218 		unsigned int	lbr_lip:1;
219 	} split;
220 	unsigned int		full;
221 };
222 
223 union cpuid28_ebx {
224 	struct {
225 		/* CPL Filtering Supported */
226 		unsigned int    lbr_cpl:1;
227 		/* Branch Filtering Supported */
228 		unsigned int    lbr_filter:1;
229 		/* Call-stack Mode Supported */
230 		unsigned int    lbr_call_stack:1;
231 	} split;
232 	unsigned int            full;
233 };
234 
235 union cpuid28_ecx {
236 	struct {
237 		/* Mispredict Bit Supported */
238 		unsigned int    lbr_mispred:1;
239 		/* Timed LBRs Supported */
240 		unsigned int    lbr_timed_lbr:1;
241 		/* Branch Type Field Supported */
242 		unsigned int    lbr_br_type:1;
243 	} split;
244 	unsigned int            full;
245 };
246 
247 /*
248  * AMD "Extended Performance Monitoring and Debug" CPUID
249  * detection/enumeration details:
250  */
251 union cpuid_0x80000022_ebx {
252 	struct {
253 		/* Number of Core Performance Counters */
254 		unsigned int	num_core_pmc:4;
255 		/* Number of available LBR Stack Entries */
256 		unsigned int	lbr_v2_stack_sz:6;
257 		/* Number of Data Fabric Counters */
258 		unsigned int	num_df_pmc:6;
259 	} split;
260 	unsigned int		full;
261 };
262 
263 struct x86_pmu_capability {
264 	int		version;
265 	int		num_counters_gp;
266 	int		num_counters_fixed;
267 	int		bit_width_gp;
268 	int		bit_width_fixed;
269 	unsigned int	events_mask;
270 	int		events_mask_len;
271 	unsigned int	pebs_ept	:1;
272 };
273 
274 /*
275  * Fixed-purpose performance events:
276  */
277 
278 /* RDPMC offset for Fixed PMCs */
279 #define INTEL_PMC_FIXED_RDPMC_BASE		(1 << 30)
280 #define INTEL_PMC_FIXED_RDPMC_METRICS		(1 << 29)
281 
282 /*
283  * All the fixed-mode PMCs are configured via this single MSR:
284  */
285 #define MSR_ARCH_PERFMON_FIXED_CTR_CTRL	0x38d
286 
287 /*
288  * There is no event-code assigned to the fixed-mode PMCs.
289  *
290  * For a fixed-mode PMC, which has an equivalent event on a general-purpose
291  * PMC, the event-code of the equivalent event is used for the fixed-mode PMC,
292  * e.g., Instr_Retired.Any and CPU_CLK_Unhalted.Core.
293  *
294  * For a fixed-mode PMC, which doesn't have an equivalent event, a
295  * pseudo-encoding is used, e.g., CPU_CLK_Unhalted.Ref and TOPDOWN.SLOTS.
296  * The pseudo event-code for a fixed-mode PMC must be 0x00.
297  * The pseudo umask-code is 0xX. The X equals the index of the fixed
298  * counter + 1, e.g., the fixed counter 2 has the pseudo-encoding 0x0300.
299  *
300  * The counts are available in separate MSRs:
301  */
302 
303 /* Instr_Retired.Any: */
304 #define MSR_ARCH_PERFMON_FIXED_CTR0	0x309
305 #define INTEL_PMC_IDX_FIXED_INSTRUCTIONS	(INTEL_PMC_IDX_FIXED + 0)
306 
307 /* CPU_CLK_Unhalted.Core: */
308 #define MSR_ARCH_PERFMON_FIXED_CTR1	0x30a
309 #define INTEL_PMC_IDX_FIXED_CPU_CYCLES	(INTEL_PMC_IDX_FIXED + 1)
310 
311 /* CPU_CLK_Unhalted.Ref: event=0x00,umask=0x3 (pseudo-encoding) */
312 #define MSR_ARCH_PERFMON_FIXED_CTR2	0x30b
313 #define INTEL_PMC_IDX_FIXED_REF_CYCLES	(INTEL_PMC_IDX_FIXED + 2)
314 #define INTEL_PMC_MSK_FIXED_REF_CYCLES	(1ULL << INTEL_PMC_IDX_FIXED_REF_CYCLES)
315 
316 /* TOPDOWN.SLOTS: event=0x00,umask=0x4 (pseudo-encoding) */
317 #define MSR_ARCH_PERFMON_FIXED_CTR3	0x30c
318 #define INTEL_PMC_IDX_FIXED_SLOTS	(INTEL_PMC_IDX_FIXED + 3)
319 #define INTEL_PMC_MSK_FIXED_SLOTS	(1ULL << INTEL_PMC_IDX_FIXED_SLOTS)
320 
use_fixed_pseudo_encoding(u64 code)321 static inline bool use_fixed_pseudo_encoding(u64 code)
322 {
323 	return !(code & 0xff);
324 }
325 
326 /*
327  * We model BTS tracing as another fixed-mode PMC.
328  *
329  * We choose the value 47 for the fixed index of BTS, since lower
330  * values are used by actual fixed events and higher values are used
331  * to indicate other overflow conditions in the PERF_GLOBAL_STATUS msr.
332  */
333 #define INTEL_PMC_IDX_FIXED_BTS			(INTEL_PMC_IDX_FIXED + 15)
334 
335 /*
336  * The PERF_METRICS MSR is modeled as several magic fixed-mode PMCs, one for
337  * each TopDown metric event.
338  *
339  * Internally the TopDown metric events are mapped to the FxCtr 3 (SLOTS).
340  */
341 #define INTEL_PMC_IDX_METRIC_BASE		(INTEL_PMC_IDX_FIXED + 16)
342 #define INTEL_PMC_IDX_TD_RETIRING		(INTEL_PMC_IDX_METRIC_BASE + 0)
343 #define INTEL_PMC_IDX_TD_BAD_SPEC		(INTEL_PMC_IDX_METRIC_BASE + 1)
344 #define INTEL_PMC_IDX_TD_FE_BOUND		(INTEL_PMC_IDX_METRIC_BASE + 2)
345 #define INTEL_PMC_IDX_TD_BE_BOUND		(INTEL_PMC_IDX_METRIC_BASE + 3)
346 #define INTEL_PMC_IDX_TD_HEAVY_OPS		(INTEL_PMC_IDX_METRIC_BASE + 4)
347 #define INTEL_PMC_IDX_TD_BR_MISPREDICT		(INTEL_PMC_IDX_METRIC_BASE + 5)
348 #define INTEL_PMC_IDX_TD_FETCH_LAT		(INTEL_PMC_IDX_METRIC_BASE + 6)
349 #define INTEL_PMC_IDX_TD_MEM_BOUND		(INTEL_PMC_IDX_METRIC_BASE + 7)
350 #define INTEL_PMC_IDX_METRIC_END		INTEL_PMC_IDX_TD_MEM_BOUND
351 #define INTEL_PMC_MSK_TOPDOWN			((0xffull << INTEL_PMC_IDX_METRIC_BASE) | \
352 						INTEL_PMC_MSK_FIXED_SLOTS)
353 
354 /*
355  * There is no event-code assigned to the TopDown events.
356  *
357  * For the slots event, use the pseudo code of the fixed counter 3.
358  *
359  * For the metric events, the pseudo event-code is 0x00.
360  * The pseudo umask-code starts from the middle of the pseudo event
361  * space, 0x80.
362  */
363 #define INTEL_TD_SLOTS				0x0400	/* TOPDOWN.SLOTS */
364 /* Level 1 metrics */
365 #define INTEL_TD_METRIC_RETIRING		0x8000	/* Retiring metric */
366 #define INTEL_TD_METRIC_BAD_SPEC		0x8100	/* Bad speculation metric */
367 #define INTEL_TD_METRIC_FE_BOUND		0x8200	/* FE bound metric */
368 #define INTEL_TD_METRIC_BE_BOUND		0x8300	/* BE bound metric */
369 /* Level 2 metrics */
370 #define INTEL_TD_METRIC_HEAVY_OPS		0x8400  /* Heavy Operations metric */
371 #define INTEL_TD_METRIC_BR_MISPREDICT		0x8500  /* Branch Mispredict metric */
372 #define INTEL_TD_METRIC_FETCH_LAT		0x8600  /* Fetch Latency metric */
373 #define INTEL_TD_METRIC_MEM_BOUND		0x8700  /* Memory bound metric */
374 
375 #define INTEL_TD_METRIC_MAX			INTEL_TD_METRIC_MEM_BOUND
376 #define INTEL_TD_METRIC_NUM			8
377 
is_metric_idx(int idx)378 static inline bool is_metric_idx(int idx)
379 {
380 	return (unsigned)(idx - INTEL_PMC_IDX_METRIC_BASE) < INTEL_TD_METRIC_NUM;
381 }
382 
is_topdown_idx(int idx)383 static inline bool is_topdown_idx(int idx)
384 {
385 	return is_metric_idx(idx) || idx == INTEL_PMC_IDX_FIXED_SLOTS;
386 }
387 
388 #define INTEL_PMC_OTHER_TOPDOWN_BITS(bit)	\
389 			(~(0x1ull << bit) & INTEL_PMC_MSK_TOPDOWN)
390 
391 #define GLOBAL_STATUS_COND_CHG			BIT_ULL(63)
392 #define GLOBAL_STATUS_BUFFER_OVF_BIT		62
393 #define GLOBAL_STATUS_BUFFER_OVF		BIT_ULL(GLOBAL_STATUS_BUFFER_OVF_BIT)
394 #define GLOBAL_STATUS_UNC_OVF			BIT_ULL(61)
395 #define GLOBAL_STATUS_ASIF			BIT_ULL(60)
396 #define GLOBAL_STATUS_COUNTERS_FROZEN		BIT_ULL(59)
397 #define GLOBAL_STATUS_LBRS_FROZEN_BIT		58
398 #define GLOBAL_STATUS_LBRS_FROZEN		BIT_ULL(GLOBAL_STATUS_LBRS_FROZEN_BIT)
399 #define GLOBAL_STATUS_TRACE_TOPAPMI_BIT		55
400 #define GLOBAL_STATUS_TRACE_TOPAPMI		BIT_ULL(GLOBAL_STATUS_TRACE_TOPAPMI_BIT)
401 #define GLOBAL_STATUS_PERF_METRICS_OVF_BIT	48
402 
403 #define GLOBAL_CTRL_EN_PERF_METRICS		48
404 /*
405  * We model guest LBR event tracing as another fixed-mode PMC like BTS.
406  *
407  * We choose bit 58 because it's used to indicate LBR stack frozen state
408  * for architectural perfmon v4, also we unconditionally mask that bit in
409  * the handle_pmi_common(), so it'll never be set in the overflow handling.
410  *
411  * With this fake counter assigned, the guest LBR event user (such as KVM),
412  * can program the LBR registers on its own, and we don't actually do anything
413  * with then in the host context.
414  */
415 #define INTEL_PMC_IDX_FIXED_VLBR	(GLOBAL_STATUS_LBRS_FROZEN_BIT)
416 
417 /*
418  * Pseudo-encoding the guest LBR event as event=0x00,umask=0x1b,
419  * since it would claim bit 58 which is effectively Fixed26.
420  */
421 #define INTEL_FIXED_VLBR_EVENT	0x1b00
422 
423 /*
424  * Adaptive PEBS v4
425  */
426 
427 struct pebs_basic {
428 	u64 format_size;
429 	u64 ip;
430 	u64 applicable_counters;
431 	u64 tsc;
432 };
433 
434 struct pebs_meminfo {
435 	u64 address;
436 	u64 aux;
437 	u64 latency;
438 	u64 tsx_tuning;
439 };
440 
441 struct pebs_gprs {
442 	u64 flags, ip, ax, cx, dx, bx, sp, bp, si, di;
443 	u64 r8, r9, r10, r11, r12, r13, r14, r15;
444 };
445 
446 struct pebs_xmm {
447 	u64 xmm[16*2];	/* two entries for each register */
448 };
449 
450 /*
451  * AMD Extended Performance Monitoring and Debug cpuid feature detection
452  */
453 #define EXT_PERFMON_DEBUG_FEATURES		0x80000022
454 
455 /*
456  * IBS cpuid feature detection
457  */
458 
459 #define IBS_CPUID_FEATURES		0x8000001b
460 
461 /*
462  * Same bit mask as for IBS cpuid feature flags (Fn8000_001B_EAX), but
463  * bit 0 is used to indicate the existence of IBS.
464  */
465 #define IBS_CAPS_AVAIL			(1U<<0)
466 #define IBS_CAPS_FETCHSAM		(1U<<1)
467 #define IBS_CAPS_OPSAM			(1U<<2)
468 #define IBS_CAPS_RDWROPCNT		(1U<<3)
469 #define IBS_CAPS_OPCNT			(1U<<4)
470 #define IBS_CAPS_BRNTRGT		(1U<<5)
471 #define IBS_CAPS_OPCNTEXT		(1U<<6)
472 #define IBS_CAPS_RIPINVALIDCHK		(1U<<7)
473 #define IBS_CAPS_OPBRNFUSE		(1U<<8)
474 #define IBS_CAPS_FETCHCTLEXTD		(1U<<9)
475 #define IBS_CAPS_OPDATA4		(1U<<10)
476 #define IBS_CAPS_ZEN4			(1U<<11)
477 
478 #define IBS_CAPS_DEFAULT		(IBS_CAPS_AVAIL		\
479 					 | IBS_CAPS_FETCHSAM	\
480 					 | IBS_CAPS_OPSAM)
481 
482 /*
483  * IBS APIC setup
484  */
485 #define IBSCTL				0x1cc
486 #define IBSCTL_LVT_OFFSET_VALID		(1ULL<<8)
487 #define IBSCTL_LVT_OFFSET_MASK		0x0F
488 
489 /* IBS fetch bits/masks */
490 #define IBS_FETCH_L3MISSONLY	(1ULL<<59)
491 #define IBS_FETCH_RAND_EN	(1ULL<<57)
492 #define IBS_FETCH_VAL		(1ULL<<49)
493 #define IBS_FETCH_ENABLE	(1ULL<<48)
494 #define IBS_FETCH_CNT		0xFFFF0000ULL
495 #define IBS_FETCH_MAX_CNT	0x0000FFFFULL
496 
497 /*
498  * IBS op bits/masks
499  * The lower 7 bits of the current count are random bits
500  * preloaded by hardware and ignored in software
501  */
502 #define IBS_OP_CUR_CNT		(0xFFF80ULL<<32)
503 #define IBS_OP_CUR_CNT_RAND	(0x0007FULL<<32)
504 #define IBS_OP_CNT_CTL		(1ULL<<19)
505 #define IBS_OP_VAL		(1ULL<<18)
506 #define IBS_OP_ENABLE		(1ULL<<17)
507 #define IBS_OP_L3MISSONLY	(1ULL<<16)
508 #define IBS_OP_MAX_CNT		0x0000FFFFULL
509 #define IBS_OP_MAX_CNT_EXT	0x007FFFFFULL	/* not a register bit mask */
510 #define IBS_OP_MAX_CNT_EXT_MASK	(0x7FULL<<20)	/* separate upper 7 bits */
511 #define IBS_RIP_INVALID		(1ULL<<38)
512 
513 #ifdef CONFIG_X86_LOCAL_APIC
514 extern u32 get_ibs_caps(void);
515 extern int forward_event_to_ibs(struct perf_event *event);
516 #else
get_ibs_caps(void)517 static inline u32 get_ibs_caps(void) { return 0; }
forward_event_to_ibs(struct perf_event * event)518 static inline int forward_event_to_ibs(struct perf_event *event) { return -ENOENT; }
519 #endif
520 
521 #ifdef CONFIG_PERF_EVENTS
522 extern void perf_events_lapic_init(void);
523 
524 /*
525  * Abuse bits {3,5} of the cpu eflags register. These flags are otherwise
526  * unused and ABI specified to be 0, so nobody should care what we do with
527  * them.
528  *
529  * EXACT - the IP points to the exact instruction that triggered the
530  *         event (HW bugs exempt).
531  * VM    - original X86_VM_MASK; see set_linear_ip().
532  */
533 #define PERF_EFLAGS_EXACT	(1UL << 3)
534 #define PERF_EFLAGS_VM		(1UL << 5)
535 
536 struct pt_regs;
537 struct x86_perf_regs {
538 	struct pt_regs	regs;
539 	u64		*xmm_regs;
540 };
541 
542 extern unsigned long perf_instruction_pointer(struct pt_regs *regs);
543 extern unsigned long perf_misc_flags(struct pt_regs *regs);
544 #define perf_misc_flags(regs)	perf_misc_flags(regs)
545 
546 #include <asm/stacktrace.h>
547 
548 /*
549  * We abuse bit 3 from flags to pass exact information, see perf_misc_flags
550  * and the comment with PERF_EFLAGS_EXACT.
551  */
552 #define perf_arch_fetch_caller_regs(regs, __ip)		{	\
553 	(regs)->ip = (__ip);					\
554 	(regs)->sp = (unsigned long)__builtin_frame_address(0);	\
555 	(regs)->cs = __KERNEL_CS;				\
556 	regs->flags = 0;					\
557 }
558 
559 struct perf_guest_switch_msr {
560 	unsigned msr;
561 	u64 host, guest;
562 };
563 
564 struct x86_pmu_lbr {
565 	unsigned int	nr;
566 	unsigned int	from;
567 	unsigned int	to;
568 	unsigned int	info;
569 };
570 
571 extern void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap);
572 extern u64 perf_get_hw_event_config(int hw_event);
573 extern void perf_check_microcode(void);
574 extern void perf_clear_dirty_counters(void);
575 extern int x86_perf_rdpmc_index(struct perf_event *event);
576 #else
perf_get_x86_pmu_capability(struct x86_pmu_capability * cap)577 static inline void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap)
578 {
579 	memset(cap, 0, sizeof(*cap));
580 }
581 
perf_get_hw_event_config(int hw_event)582 static inline u64 perf_get_hw_event_config(int hw_event)
583 {
584 	return 0;
585 }
586 
perf_events_lapic_init(void)587 static inline void perf_events_lapic_init(void)	{ }
perf_check_microcode(void)588 static inline void perf_check_microcode(void) { }
589 #endif
590 
591 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_INTEL)
592 extern struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr, void *data);
593 extern void x86_perf_get_lbr(struct x86_pmu_lbr *lbr);
594 #else
595 struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr, void *data);
x86_perf_get_lbr(struct x86_pmu_lbr * lbr)596 static inline void x86_perf_get_lbr(struct x86_pmu_lbr *lbr)
597 {
598 	memset(lbr, 0, sizeof(*lbr));
599 }
600 #endif
601 
602 #ifdef CONFIG_CPU_SUP_INTEL
603  extern void intel_pt_handle_vmx(int on);
604 #else
intel_pt_handle_vmx(int on)605 static inline void intel_pt_handle_vmx(int on)
606 {
607 
608 }
609 #endif
610 
611 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD)
612  extern void amd_pmu_enable_virt(void);
613  extern void amd_pmu_disable_virt(void);
614 
615 #if defined(CONFIG_PERF_EVENTS_AMD_BRS)
616 
617 #define PERF_NEEDS_LOPWR_CB 1
618 
619 /*
620  * architectural low power callback impacts
621  * drivers/acpi/processor_idle.c
622  * drivers/acpi/acpi_pad.c
623  */
624 extern void perf_amd_brs_lopwr_cb(bool lopwr_in);
625 
626 DECLARE_STATIC_CALL(perf_lopwr_cb, perf_amd_brs_lopwr_cb);
627 
perf_lopwr_cb(bool lopwr_in)628 static __always_inline void perf_lopwr_cb(bool lopwr_in)
629 {
630 	static_call_mod(perf_lopwr_cb)(lopwr_in);
631 }
632 
633 #endif /* PERF_NEEDS_LOPWR_CB */
634 
635 #else
amd_pmu_enable_virt(void)636  static inline void amd_pmu_enable_virt(void) { }
amd_pmu_disable_virt(void)637  static inline void amd_pmu_disable_virt(void) { }
638 #endif
639 
640 #define arch_perf_out_copy_user copy_from_user_nmi
641 
642 #endif /* _ASM_X86_PERF_EVENT_H */
643