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 #define ARCH_PERFMON_EVENTSEL_BR_CNTR (1ULL << 35)
35 #define ARCH_PERFMON_EVENTSEL_EQ (1ULL << 36)
36 #define ARCH_PERFMON_EVENTSEL_UMASK2 (0xFFULL << 40)
37
38 #define INTEL_FIXED_BITS_MASK 0xFULL
39 #define INTEL_FIXED_BITS_STRIDE 4
40 #define INTEL_FIXED_0_KERNEL (1ULL << 0)
41 #define INTEL_FIXED_0_USER (1ULL << 1)
42 #define INTEL_FIXED_0_ANYTHREAD (1ULL << 2)
43 #define INTEL_FIXED_0_ENABLE_PMI (1ULL << 3)
44
45 #define HSW_IN_TX (1ULL << 32)
46 #define HSW_IN_TX_CHECKPOINTED (1ULL << 33)
47 #define ICL_EVENTSEL_ADAPTIVE (1ULL << 34)
48 #define ICL_FIXED_0_ADAPTIVE (1ULL << 32)
49
50 #define intel_fixed_bits_by_idx(_idx, _bits) \
51 ((_bits) << ((_idx) * INTEL_FIXED_BITS_STRIDE))
52
53 #define AMD64_EVENTSEL_INT_CORE_ENABLE (1ULL << 36)
54 #define AMD64_EVENTSEL_GUESTONLY (1ULL << 40)
55 #define AMD64_EVENTSEL_HOSTONLY (1ULL << 41)
56
57 #define AMD64_EVENTSEL_INT_CORE_SEL_SHIFT 37
58 #define AMD64_EVENTSEL_INT_CORE_SEL_MASK \
59 (0xFULL << AMD64_EVENTSEL_INT_CORE_SEL_SHIFT)
60
61 #define AMD64_EVENTSEL_EVENT \
62 (ARCH_PERFMON_EVENTSEL_EVENT | (0x0FULL << 32))
63 #define INTEL_ARCH_EVENT_MASK \
64 (ARCH_PERFMON_EVENTSEL_UMASK | ARCH_PERFMON_EVENTSEL_EVENT)
65
66 #define AMD64_L3_SLICE_SHIFT 48
67 #define AMD64_L3_SLICE_MASK \
68 (0xFULL << AMD64_L3_SLICE_SHIFT)
69 #define AMD64_L3_SLICEID_MASK \
70 (0x7ULL << AMD64_L3_SLICE_SHIFT)
71
72 #define AMD64_L3_THREAD_SHIFT 56
73 #define AMD64_L3_THREAD_MASK \
74 (0xFFULL << AMD64_L3_THREAD_SHIFT)
75 #define AMD64_L3_F19H_THREAD_MASK \
76 (0x3ULL << AMD64_L3_THREAD_SHIFT)
77
78 #define AMD64_L3_EN_ALL_CORES BIT_ULL(47)
79 #define AMD64_L3_EN_ALL_SLICES BIT_ULL(46)
80
81 #define AMD64_L3_COREID_SHIFT 42
82 #define AMD64_L3_COREID_MASK \
83 (0x7ULL << AMD64_L3_COREID_SHIFT)
84
85 #define X86_RAW_EVENT_MASK \
86 (ARCH_PERFMON_EVENTSEL_EVENT | \
87 ARCH_PERFMON_EVENTSEL_UMASK | \
88 ARCH_PERFMON_EVENTSEL_EDGE | \
89 ARCH_PERFMON_EVENTSEL_INV | \
90 ARCH_PERFMON_EVENTSEL_CMASK)
91 #define X86_ALL_EVENT_FLAGS \
92 (ARCH_PERFMON_EVENTSEL_EDGE | \
93 ARCH_PERFMON_EVENTSEL_INV | \
94 ARCH_PERFMON_EVENTSEL_CMASK | \
95 ARCH_PERFMON_EVENTSEL_ANY | \
96 ARCH_PERFMON_EVENTSEL_PIN_CONTROL | \
97 HSW_IN_TX | \
98 HSW_IN_TX_CHECKPOINTED)
99 #define AMD64_RAW_EVENT_MASK \
100 (X86_RAW_EVENT_MASK | \
101 AMD64_EVENTSEL_EVENT)
102 #define AMD64_RAW_EVENT_MASK_NB \
103 (AMD64_EVENTSEL_EVENT | \
104 ARCH_PERFMON_EVENTSEL_UMASK)
105
106 #define AMD64_PERFMON_V2_EVENTSEL_EVENT_NB \
107 (AMD64_EVENTSEL_EVENT | \
108 GENMASK_ULL(37, 36))
109
110 #define AMD64_PERFMON_V2_EVENTSEL_UMASK_NB \
111 (ARCH_PERFMON_EVENTSEL_UMASK | \
112 GENMASK_ULL(27, 24))
113
114 #define AMD64_PERFMON_V2_RAW_EVENT_MASK_NB \
115 (AMD64_PERFMON_V2_EVENTSEL_EVENT_NB | \
116 AMD64_PERFMON_V2_EVENTSEL_UMASK_NB)
117
118 #define AMD64_PERFMON_V2_ENABLE_UMC BIT_ULL(31)
119 #define AMD64_PERFMON_V2_EVENTSEL_EVENT_UMC GENMASK_ULL(7, 0)
120 #define AMD64_PERFMON_V2_EVENTSEL_RDWRMASK_UMC GENMASK_ULL(9, 8)
121 #define AMD64_PERFMON_V2_RAW_EVENT_MASK_UMC \
122 (AMD64_PERFMON_V2_EVENTSEL_EVENT_UMC | \
123 AMD64_PERFMON_V2_EVENTSEL_RDWRMASK_UMC)
124
125 #define AMD64_NUM_COUNTERS 4
126 #define AMD64_NUM_COUNTERS_CORE 6
127 #define AMD64_NUM_COUNTERS_NB 4
128
129 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL 0x3c
130 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK (0x00 << 8)
131 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX 0
132 #define ARCH_PERFMON_UNHALTED_CORE_CYCLES_PRESENT \
133 (1 << (ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX))
134
135 #define ARCH_PERFMON_BRANCH_MISSES_RETIRED 6
136 #define ARCH_PERFMON_EVENTS_COUNT 7
137
138 #define PEBS_DATACFG_MEMINFO BIT_ULL(0)
139 #define PEBS_DATACFG_GP BIT_ULL(1)
140 #define PEBS_DATACFG_XMMS BIT_ULL(2)
141 #define PEBS_DATACFG_LBRS BIT_ULL(3)
142 #define PEBS_DATACFG_LBR_SHIFT 24
143
144 /* Steal the highest bit of pebs_data_cfg for SW usage */
145 #define PEBS_UPDATE_DS_SW BIT_ULL(63)
146
147 /*
148 * Intel "Architectural Performance Monitoring" CPUID
149 * detection/enumeration details:
150 */
151 union cpuid10_eax {
152 struct {
153 unsigned int version_id:8;
154 unsigned int num_counters:8;
155 unsigned int bit_width:8;
156 unsigned int mask_length:8;
157 } split;
158 unsigned int full;
159 };
160
161 union cpuid10_ebx {
162 struct {
163 unsigned int no_unhalted_core_cycles:1;
164 unsigned int no_instructions_retired:1;
165 unsigned int no_unhalted_reference_cycles:1;
166 unsigned int no_llc_reference:1;
167 unsigned int no_llc_misses:1;
168 unsigned int no_branch_instruction_retired:1;
169 unsigned int no_branch_misses_retired:1;
170 } split;
171 unsigned int full;
172 };
173
174 union cpuid10_edx {
175 struct {
176 unsigned int num_counters_fixed:5;
177 unsigned int bit_width_fixed:8;
178 unsigned int reserved1:2;
179 unsigned int anythread_deprecated:1;
180 unsigned int reserved2:16;
181 } split;
182 unsigned int full;
183 };
184
185 /*
186 * Intel "Architectural Performance Monitoring extension" CPUID
187 * detection/enumeration details:
188 */
189 #define ARCH_PERFMON_EXT_LEAF 0x00000023
190 #define ARCH_PERFMON_NUM_COUNTER_LEAF 0x1
191
192 union cpuid35_eax {
193 struct {
194 unsigned int leaf0:1;
195 /* Counters Sub-Leaf */
196 unsigned int cntr_subleaf:1;
197 /* Auto Counter Reload Sub-Leaf */
198 unsigned int acr_subleaf:1;
199 /* Events Sub-Leaf */
200 unsigned int events_subleaf:1;
201 unsigned int reserved:28;
202 } split;
203 unsigned int full;
204 };
205
206 union cpuid35_ebx {
207 struct {
208 /* UnitMask2 Supported */
209 unsigned int umask2:1;
210 /* EQ-bit Supported */
211 unsigned int eq:1;
212 unsigned int reserved:30;
213 } split;
214 unsigned int full;
215 };
216
217 /*
218 * Intel Architectural LBR CPUID detection/enumeration details:
219 */
220 union cpuid28_eax {
221 struct {
222 /* Supported LBR depth values */
223 unsigned int lbr_depth_mask:8;
224 unsigned int reserved:22;
225 /* Deep C-state Reset */
226 unsigned int lbr_deep_c_reset:1;
227 /* IP values contain LIP */
228 unsigned int lbr_lip:1;
229 } split;
230 unsigned int full;
231 };
232
233 union cpuid28_ebx {
234 struct {
235 /* CPL Filtering Supported */
236 unsigned int lbr_cpl:1;
237 /* Branch Filtering Supported */
238 unsigned int lbr_filter:1;
239 /* Call-stack Mode Supported */
240 unsigned int lbr_call_stack:1;
241 } split;
242 unsigned int full;
243 };
244
245 union cpuid28_ecx {
246 struct {
247 /* Mispredict Bit Supported */
248 unsigned int lbr_mispred:1;
249 /* Timed LBRs Supported */
250 unsigned int lbr_timed_lbr:1;
251 /* Branch Type Field Supported */
252 unsigned int lbr_br_type:1;
253 unsigned int reserved:13;
254 /* Branch counters (Event Logging) Supported */
255 unsigned int lbr_counters:4;
256 } split;
257 unsigned int full;
258 };
259
260 /*
261 * AMD "Extended Performance Monitoring and Debug" CPUID
262 * detection/enumeration details:
263 */
264 union cpuid_0x80000022_ebx {
265 struct {
266 /* Number of Core Performance Counters */
267 unsigned int num_core_pmc:4;
268 /* Number of available LBR Stack Entries */
269 unsigned int lbr_v2_stack_sz:6;
270 /* Number of Data Fabric Counters */
271 unsigned int num_df_pmc:6;
272 /* Number of Unified Memory Controller Counters */
273 unsigned int num_umc_pmc:6;
274 } split;
275 unsigned int full;
276 };
277
278 struct x86_pmu_capability {
279 int version;
280 int num_counters_gp;
281 int num_counters_fixed;
282 int bit_width_gp;
283 int bit_width_fixed;
284 unsigned int events_mask;
285 int events_mask_len;
286 unsigned int pebs_ept :1;
287 };
288
289 /*
290 * Fixed-purpose performance events:
291 */
292
293 /* RDPMC offset for Fixed PMCs */
294 #define INTEL_PMC_FIXED_RDPMC_BASE (1 << 30)
295 #define INTEL_PMC_FIXED_RDPMC_METRICS (1 << 29)
296
297 /*
298 * All the fixed-mode PMCs are configured via this single MSR:
299 */
300 #define MSR_ARCH_PERFMON_FIXED_CTR_CTRL 0x38d
301
302 /*
303 * There is no event-code assigned to the fixed-mode PMCs.
304 *
305 * For a fixed-mode PMC, which has an equivalent event on a general-purpose
306 * PMC, the event-code of the equivalent event is used for the fixed-mode PMC,
307 * e.g., Instr_Retired.Any and CPU_CLK_Unhalted.Core.
308 *
309 * For a fixed-mode PMC, which doesn't have an equivalent event, a
310 * pseudo-encoding is used, e.g., CPU_CLK_Unhalted.Ref and TOPDOWN.SLOTS.
311 * The pseudo event-code for a fixed-mode PMC must be 0x00.
312 * The pseudo umask-code is 0xX. The X equals the index of the fixed
313 * counter + 1, e.g., the fixed counter 2 has the pseudo-encoding 0x0300.
314 *
315 * The counts are available in separate MSRs:
316 */
317
318 /* Instr_Retired.Any: */
319 #define MSR_ARCH_PERFMON_FIXED_CTR0 0x309
320 #define INTEL_PMC_IDX_FIXED_INSTRUCTIONS (INTEL_PMC_IDX_FIXED + 0)
321
322 /* CPU_CLK_Unhalted.Core: */
323 #define MSR_ARCH_PERFMON_FIXED_CTR1 0x30a
324 #define INTEL_PMC_IDX_FIXED_CPU_CYCLES (INTEL_PMC_IDX_FIXED + 1)
325
326 /* CPU_CLK_Unhalted.Ref: event=0x00,umask=0x3 (pseudo-encoding) */
327 #define MSR_ARCH_PERFMON_FIXED_CTR2 0x30b
328 #define INTEL_PMC_IDX_FIXED_REF_CYCLES (INTEL_PMC_IDX_FIXED + 2)
329 #define INTEL_PMC_MSK_FIXED_REF_CYCLES (1ULL << INTEL_PMC_IDX_FIXED_REF_CYCLES)
330
331 /* TOPDOWN.SLOTS: event=0x00,umask=0x4 (pseudo-encoding) */
332 #define MSR_ARCH_PERFMON_FIXED_CTR3 0x30c
333 #define INTEL_PMC_IDX_FIXED_SLOTS (INTEL_PMC_IDX_FIXED + 3)
334 #define INTEL_PMC_MSK_FIXED_SLOTS (1ULL << INTEL_PMC_IDX_FIXED_SLOTS)
335
336 /* TOPDOWN_BAD_SPECULATION.ALL: fixed counter 4 (Atom only) */
337 /* TOPDOWN_FE_BOUND.ALL: fixed counter 5 (Atom only) */
338 /* TOPDOWN_RETIRING.ALL: fixed counter 6 (Atom only) */
339
use_fixed_pseudo_encoding(u64 code)340 static inline bool use_fixed_pseudo_encoding(u64 code)
341 {
342 return !(code & 0xff);
343 }
344
345 /*
346 * We model BTS tracing as another fixed-mode PMC.
347 *
348 * We choose the value 47 for the fixed index of BTS, since lower
349 * values are used by actual fixed events and higher values are used
350 * to indicate other overflow conditions in the PERF_GLOBAL_STATUS msr.
351 */
352 #define INTEL_PMC_IDX_FIXED_BTS (INTEL_PMC_IDX_FIXED + 15)
353
354 /*
355 * The PERF_METRICS MSR is modeled as several magic fixed-mode PMCs, one for
356 * each TopDown metric event.
357 *
358 * Internally the TopDown metric events are mapped to the FxCtr 3 (SLOTS).
359 */
360 #define INTEL_PMC_IDX_METRIC_BASE (INTEL_PMC_IDX_FIXED + 16)
361 #define INTEL_PMC_IDX_TD_RETIRING (INTEL_PMC_IDX_METRIC_BASE + 0)
362 #define INTEL_PMC_IDX_TD_BAD_SPEC (INTEL_PMC_IDX_METRIC_BASE + 1)
363 #define INTEL_PMC_IDX_TD_FE_BOUND (INTEL_PMC_IDX_METRIC_BASE + 2)
364 #define INTEL_PMC_IDX_TD_BE_BOUND (INTEL_PMC_IDX_METRIC_BASE + 3)
365 #define INTEL_PMC_IDX_TD_HEAVY_OPS (INTEL_PMC_IDX_METRIC_BASE + 4)
366 #define INTEL_PMC_IDX_TD_BR_MISPREDICT (INTEL_PMC_IDX_METRIC_BASE + 5)
367 #define INTEL_PMC_IDX_TD_FETCH_LAT (INTEL_PMC_IDX_METRIC_BASE + 6)
368 #define INTEL_PMC_IDX_TD_MEM_BOUND (INTEL_PMC_IDX_METRIC_BASE + 7)
369 #define INTEL_PMC_IDX_METRIC_END INTEL_PMC_IDX_TD_MEM_BOUND
370 #define INTEL_PMC_MSK_TOPDOWN ((0xffull << INTEL_PMC_IDX_METRIC_BASE) | \
371 INTEL_PMC_MSK_FIXED_SLOTS)
372
373 /*
374 * There is no event-code assigned to the TopDown events.
375 *
376 * For the slots event, use the pseudo code of the fixed counter 3.
377 *
378 * For the metric events, the pseudo event-code is 0x00.
379 * The pseudo umask-code starts from the middle of the pseudo event
380 * space, 0x80.
381 */
382 #define INTEL_TD_SLOTS 0x0400 /* TOPDOWN.SLOTS */
383 /* Level 1 metrics */
384 #define INTEL_TD_METRIC_RETIRING 0x8000 /* Retiring metric */
385 #define INTEL_TD_METRIC_BAD_SPEC 0x8100 /* Bad speculation metric */
386 #define INTEL_TD_METRIC_FE_BOUND 0x8200 /* FE bound metric */
387 #define INTEL_TD_METRIC_BE_BOUND 0x8300 /* BE bound metric */
388 /* Level 2 metrics */
389 #define INTEL_TD_METRIC_HEAVY_OPS 0x8400 /* Heavy Operations metric */
390 #define INTEL_TD_METRIC_BR_MISPREDICT 0x8500 /* Branch Mispredict metric */
391 #define INTEL_TD_METRIC_FETCH_LAT 0x8600 /* Fetch Latency metric */
392 #define INTEL_TD_METRIC_MEM_BOUND 0x8700 /* Memory bound metric */
393
394 #define INTEL_TD_METRIC_MAX INTEL_TD_METRIC_MEM_BOUND
395 #define INTEL_TD_METRIC_NUM 8
396
is_metric_idx(int idx)397 static inline bool is_metric_idx(int idx)
398 {
399 return (unsigned)(idx - INTEL_PMC_IDX_METRIC_BASE) < INTEL_TD_METRIC_NUM;
400 }
401
is_topdown_idx(int idx)402 static inline bool is_topdown_idx(int idx)
403 {
404 return is_metric_idx(idx) || idx == INTEL_PMC_IDX_FIXED_SLOTS;
405 }
406
407 #define INTEL_PMC_OTHER_TOPDOWN_BITS(bit) \
408 (~(0x1ull << bit) & INTEL_PMC_MSK_TOPDOWN)
409
410 #define GLOBAL_STATUS_COND_CHG BIT_ULL(63)
411 #define GLOBAL_STATUS_BUFFER_OVF_BIT 62
412 #define GLOBAL_STATUS_BUFFER_OVF BIT_ULL(GLOBAL_STATUS_BUFFER_OVF_BIT)
413 #define GLOBAL_STATUS_UNC_OVF BIT_ULL(61)
414 #define GLOBAL_STATUS_ASIF BIT_ULL(60)
415 #define GLOBAL_STATUS_COUNTERS_FROZEN BIT_ULL(59)
416 #define GLOBAL_STATUS_LBRS_FROZEN_BIT 58
417 #define GLOBAL_STATUS_LBRS_FROZEN BIT_ULL(GLOBAL_STATUS_LBRS_FROZEN_BIT)
418 #define GLOBAL_STATUS_TRACE_TOPAPMI_BIT 55
419 #define GLOBAL_STATUS_TRACE_TOPAPMI BIT_ULL(GLOBAL_STATUS_TRACE_TOPAPMI_BIT)
420 #define GLOBAL_STATUS_PERF_METRICS_OVF_BIT 48
421
422 #define GLOBAL_CTRL_EN_PERF_METRICS 48
423 /*
424 * We model guest LBR event tracing as another fixed-mode PMC like BTS.
425 *
426 * We choose bit 58 because it's used to indicate LBR stack frozen state
427 * for architectural perfmon v4, also we unconditionally mask that bit in
428 * the handle_pmi_common(), so it'll never be set in the overflow handling.
429 *
430 * With this fake counter assigned, the guest LBR event user (such as KVM),
431 * can program the LBR registers on its own, and we don't actually do anything
432 * with then in the host context.
433 */
434 #define INTEL_PMC_IDX_FIXED_VLBR (GLOBAL_STATUS_LBRS_FROZEN_BIT)
435
436 /*
437 * Pseudo-encoding the guest LBR event as event=0x00,umask=0x1b,
438 * since it would claim bit 58 which is effectively Fixed26.
439 */
440 #define INTEL_FIXED_VLBR_EVENT 0x1b00
441
442 /*
443 * Adaptive PEBS v4
444 */
445
446 struct pebs_basic {
447 u64 format_size;
448 u64 ip;
449 u64 applicable_counters;
450 u64 tsc;
451 };
452
453 struct pebs_meminfo {
454 u64 address;
455 u64 aux;
456 u64 latency;
457 u64 tsx_tuning;
458 };
459
460 struct pebs_gprs {
461 u64 flags, ip, ax, cx, dx, bx, sp, bp, si, di;
462 u64 r8, r9, r10, r11, r12, r13, r14, r15;
463 };
464
465 struct pebs_xmm {
466 u64 xmm[16*2]; /* two entries for each register */
467 };
468
469 /*
470 * AMD Extended Performance Monitoring and Debug cpuid feature detection
471 */
472 #define EXT_PERFMON_DEBUG_FEATURES 0x80000022
473
474 /*
475 * IBS cpuid feature detection
476 */
477
478 #define IBS_CPUID_FEATURES 0x8000001b
479
480 /*
481 * Same bit mask as for IBS cpuid feature flags (Fn8000_001B_EAX), but
482 * bit 0 is used to indicate the existence of IBS.
483 */
484 #define IBS_CAPS_AVAIL (1U<<0)
485 #define IBS_CAPS_FETCHSAM (1U<<1)
486 #define IBS_CAPS_OPSAM (1U<<2)
487 #define IBS_CAPS_RDWROPCNT (1U<<3)
488 #define IBS_CAPS_OPCNT (1U<<4)
489 #define IBS_CAPS_BRNTRGT (1U<<5)
490 #define IBS_CAPS_OPCNTEXT (1U<<6)
491 #define IBS_CAPS_RIPINVALIDCHK (1U<<7)
492 #define IBS_CAPS_OPBRNFUSE (1U<<8)
493 #define IBS_CAPS_FETCHCTLEXTD (1U<<9)
494 #define IBS_CAPS_OPDATA4 (1U<<10)
495 #define IBS_CAPS_ZEN4 (1U<<11)
496
497 #define IBS_CAPS_DEFAULT (IBS_CAPS_AVAIL \
498 | IBS_CAPS_FETCHSAM \
499 | IBS_CAPS_OPSAM)
500
501 /*
502 * IBS APIC setup
503 */
504 #define IBSCTL 0x1cc
505 #define IBSCTL_LVT_OFFSET_VALID (1ULL<<8)
506 #define IBSCTL_LVT_OFFSET_MASK 0x0F
507
508 /* IBS fetch bits/masks */
509 #define IBS_FETCH_L3MISSONLY (1ULL<<59)
510 #define IBS_FETCH_RAND_EN (1ULL<<57)
511 #define IBS_FETCH_VAL (1ULL<<49)
512 #define IBS_FETCH_ENABLE (1ULL<<48)
513 #define IBS_FETCH_CNT 0xFFFF0000ULL
514 #define IBS_FETCH_MAX_CNT 0x0000FFFFULL
515
516 /*
517 * IBS op bits/masks
518 * The lower 7 bits of the current count are random bits
519 * preloaded by hardware and ignored in software
520 */
521 #define IBS_OP_CUR_CNT (0xFFF80ULL<<32)
522 #define IBS_OP_CUR_CNT_RAND (0x0007FULL<<32)
523 #define IBS_OP_CUR_CNT_EXT_MASK (0x7FULL<<52)
524 #define IBS_OP_CNT_CTL (1ULL<<19)
525 #define IBS_OP_VAL (1ULL<<18)
526 #define IBS_OP_ENABLE (1ULL<<17)
527 #define IBS_OP_L3MISSONLY (1ULL<<16)
528 #define IBS_OP_MAX_CNT 0x0000FFFFULL
529 #define IBS_OP_MAX_CNT_EXT 0x007FFFFFULL /* not a register bit mask */
530 #define IBS_OP_MAX_CNT_EXT_MASK (0x7FULL<<20) /* separate upper 7 bits */
531 #define IBS_RIP_INVALID (1ULL<<38)
532
533 #ifdef CONFIG_X86_LOCAL_APIC
534 extern u32 get_ibs_caps(void);
535 extern int forward_event_to_ibs(struct perf_event *event);
536 #else
get_ibs_caps(void)537 static inline u32 get_ibs_caps(void) { return 0; }
forward_event_to_ibs(struct perf_event * event)538 static inline int forward_event_to_ibs(struct perf_event *event) { return -ENOENT; }
539 #endif
540
541 #ifdef CONFIG_PERF_EVENTS
542 extern void perf_events_lapic_init(void);
543
544 /*
545 * Abuse bits {3,5} of the cpu eflags register. These flags are otherwise
546 * unused and ABI specified to be 0, so nobody should care what we do with
547 * them.
548 *
549 * EXACT - the IP points to the exact instruction that triggered the
550 * event (HW bugs exempt).
551 * VM - original X86_VM_MASK; see set_linear_ip().
552 */
553 #define PERF_EFLAGS_EXACT (1UL << 3)
554 #define PERF_EFLAGS_VM (1UL << 5)
555
556 struct pt_regs;
557 struct x86_perf_regs {
558 struct pt_regs regs;
559 u64 *xmm_regs;
560 };
561
562 extern unsigned long perf_instruction_pointer(struct pt_regs *regs);
563 extern unsigned long perf_misc_flags(struct pt_regs *regs);
564 #define perf_misc_flags(regs) perf_misc_flags(regs)
565
566 #include <asm/stacktrace.h>
567
568 /*
569 * We abuse bit 3 from flags to pass exact information, see perf_misc_flags
570 * and the comment with PERF_EFLAGS_EXACT.
571 */
572 #define perf_arch_fetch_caller_regs(regs, __ip) { \
573 (regs)->ip = (__ip); \
574 (regs)->sp = (unsigned long)__builtin_frame_address(0); \
575 (regs)->cs = __KERNEL_CS; \
576 regs->flags = 0; \
577 }
578
579 struct perf_guest_switch_msr {
580 unsigned msr;
581 u64 host, guest;
582 };
583
584 struct x86_pmu_lbr {
585 unsigned int nr;
586 unsigned int from;
587 unsigned int to;
588 unsigned int info;
589 bool has_callstack;
590 };
591
592 extern void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap);
593 extern u64 perf_get_hw_event_config(int hw_event);
594 extern void perf_check_microcode(void);
595 extern void perf_clear_dirty_counters(void);
596 extern int x86_perf_rdpmc_index(struct perf_event *event);
597 #else
perf_get_x86_pmu_capability(struct x86_pmu_capability * cap)598 static inline void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap)
599 {
600 memset(cap, 0, sizeof(*cap));
601 }
602
perf_get_hw_event_config(int hw_event)603 static inline u64 perf_get_hw_event_config(int hw_event)
604 {
605 return 0;
606 }
607
perf_events_lapic_init(void)608 static inline void perf_events_lapic_init(void) { }
perf_check_microcode(void)609 static inline void perf_check_microcode(void) { }
610 #endif
611
612 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_INTEL)
613 extern struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr, void *data);
614 extern void x86_perf_get_lbr(struct x86_pmu_lbr *lbr);
615 #else
616 struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr, void *data);
x86_perf_get_lbr(struct x86_pmu_lbr * lbr)617 static inline void x86_perf_get_lbr(struct x86_pmu_lbr *lbr)
618 {
619 memset(lbr, 0, sizeof(*lbr));
620 }
621 #endif
622
623 #ifdef CONFIG_CPU_SUP_INTEL
624 extern void intel_pt_handle_vmx(int on);
625 #else
intel_pt_handle_vmx(int on)626 static inline void intel_pt_handle_vmx(int on)
627 {
628
629 }
630 #endif
631
632 #if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_CPU_SUP_AMD)
633 extern void amd_pmu_enable_virt(void);
634 extern void amd_pmu_disable_virt(void);
635
636 #if defined(CONFIG_PERF_EVENTS_AMD_BRS)
637
638 #define PERF_NEEDS_LOPWR_CB 1
639
640 /*
641 * architectural low power callback impacts
642 * drivers/acpi/processor_idle.c
643 * drivers/acpi/acpi_pad.c
644 */
645 extern void perf_amd_brs_lopwr_cb(bool lopwr_in);
646
647 DECLARE_STATIC_CALL(perf_lopwr_cb, perf_amd_brs_lopwr_cb);
648
perf_lopwr_cb(bool lopwr_in)649 static __always_inline void perf_lopwr_cb(bool lopwr_in)
650 {
651 static_call_mod(perf_lopwr_cb)(lopwr_in);
652 }
653
654 #endif /* PERF_NEEDS_LOPWR_CB */
655
656 #else
amd_pmu_enable_virt(void)657 static inline void amd_pmu_enable_virt(void) { }
amd_pmu_disable_virt(void)658 static inline void amd_pmu_disable_virt(void) { }
659 #endif
660
661 #define arch_perf_out_copy_user copy_from_user_nmi
662
663 #endif /* _ASM_X86_PERF_EVENT_H */
664