summaryrefslogtreecommitdiffstats
path: root/drivers/perf/arm_pmu_acpi.c
blob: 05dda19c5359a39849ac9a3c6b1a5cf14eb0614a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
// SPDX-License-Identifier: GPL-2.0-only
/*
 * ACPI probing code for ARM performance counters.
 *
 * Copyright (C) 2017 ARM Ltd.
 */

#include <linux/acpi.h>
#include <linux/cpumask.h>
#include <linux/init.h>
#include <linux/irq.h>
#include <linux/irqdesc.h>
#include <linux/percpu.h>
#include <linux/perf/arm_pmu.h>

#include <asm/cpu.h>
#include <asm/cputype.h>

static DEFINE_PER_CPU(struct arm_pmu *, probed_pmus);
static DEFINE_PER_CPU(int, pmu_irqs);

static int arm_pmu_acpi_register_irq(int cpu)
{
	struct acpi_madt_generic_interrupt *gicc;
	int gsi, trigger;

	gicc = acpi_cpu_get_madt_gicc(cpu);

	gsi = gicc->performance_interrupt;

	/*
	 * Per the ACPI spec, the MADT cannot describe a PMU that doesn't
	 * have an interrupt. QEMU advertises this by using a GSI of zero,
	 * which is not known to be valid on any hardware despite being
	 * valid per the spec. Take the pragmatic approach and reject a
	 * GSI of zero for now.
	 */
	if (!gsi)
		return 0;

	if (gicc->flags & ACPI_MADT_PERFORMANCE_IRQ_MODE)
		trigger = ACPI_EDGE_SENSITIVE;
	else
		trigger = ACPI_LEVEL_SENSITIVE;

	/*
	 * Helpfully, the MADT GICC doesn't have a polarity flag for the
	 * "performance interrupt". Luckily, on compliant GICs the polarity is
	 * a fixed value in HW (for both SPIs and PPIs) that we cannot change
	 * from SW.
	 *
	 * Here we pass in ACPI_ACTIVE_HIGH to keep the core code happy. This
	 * may not match the real polarity, but that should not matter.
	 *
	 * Other interrupt controllers are not supported with ACPI.
	 */
	return acpi_register_gsi(NULL, gsi, trigger, ACPI_ACTIVE_HIGH);
}

static void arm_pmu_acpi_unregister_irq(int cpu)
{
	struct acpi_madt_generic_interrupt *gicc;
	int gsi;

	gicc = acpi_cpu_get_madt_gicc(cpu);

	gsi = gicc->performance_interrupt;
	if (gsi)
		acpi_unregister_gsi(gsi);
}

static int __maybe_unused
arm_acpi_register_pmu_device(struct platform_device *pdev, u8 len,
			     u16 (*parse_gsi)(struct acpi_madt_generic_interrupt *))
{
	int cpu, this_hetid, hetid, irq, ret;
	u16 this_gsi = 0, gsi = 0;

	/*
	 * Ensure that platform device must have IORESOURCE_IRQ
	 * resource to hold gsi interrupt.
	 */
	if (pdev->num_resources != 1)
		return -ENXIO;

	if (pdev->resource[0].flags != IORESOURCE_IRQ)
		return -ENXIO;

	/*
	 * Sanity check all the GICC tables for the same interrupt
	 * number. For now, only support homogeneous ACPI machines.
	 */
	for_each_possible_cpu(cpu) {
		struct acpi_madt_generic_interrupt *gicc;

		gicc = acpi_cpu_get_madt_gicc(cpu);
		if (gicc->header.length < len)
			return gsi ? -ENXIO : 0;

		this_gsi = parse_gsi(gicc);
		this_hetid = find_acpi_cpu_topology_hetero_id(cpu);
		if (!gsi) {
			hetid = this_hetid;
			gsi = this_gsi;
		} else if (hetid != this_hetid || gsi != this_gsi) {
			pr_warn("ACPI: %s: must be homogeneous\n", pdev->name);
			return -ENXIO;
		}
	}

	if (!this_gsi)
		return 0;

	irq = acpi_register_gsi(NULL, gsi, ACPI_LEVEL_SENSITIVE, ACPI_ACTIVE_HIGH);
	if (irq < 0) {
		pr_warn("ACPI: %s Unable to register interrupt: %d\n", pdev->name, gsi);
		return -ENXIO;
	}

	pdev->resource[0].start = irq;
	ret = platform_device_register(pdev);
	if (ret)
		acpi_unregister_gsi(gsi);

	return ret;
}

#if IS_ENABLED(CONFIG_ARM_SPE_PMU)
static struct resource spe_resources[] = {
	{
		/* irq */
		.flags          = IORESOURCE_IRQ,
	}
};

static struct platform_device spe_dev = {
	.name = ARMV8_SPE_PDEV_NAME,
	.id = -1,
	.resource = spe_resources,
	.num_resources = ARRAY_SIZE(spe_resources)
};

static u16 arm_spe_parse_gsi(struct acpi_madt_generic_interrupt *gicc)
{
	return gicc->spe_interrupt;
}

/*
 * For lack of a better place, hook the normal PMU MADT walk
 * and create a SPE device if we detect a recent MADT with
 * a homogeneous PPI mapping.
 */
static void arm_spe_acpi_register_device(void)
{
	int ret = arm_acpi_register_pmu_device(&spe_dev, ACPI_MADT_GICC_SPE,
					       arm_spe_parse_gsi);
	if (ret)
		pr_warn("ACPI: SPE: Unable to register device\n");
}
#else
static inline void arm_spe_acpi_register_device(void)
{
}
#endif /* CONFIG_ARM_SPE_PMU */

#if IS_ENABLED(CONFIG_CORESIGHT_TRBE)
static struct resource trbe_resources[] = {
	{
		/* irq */
		.flags          = IORESOURCE_IRQ,
	}
};

static struct platform_device trbe_dev = {
	.name = ARMV8_TRBE_PDEV_NAME,
	.id = -1,
	.resource = trbe_resources,
	.num_resources = ARRAY_SIZE(trbe_resources)
};

static u16 arm_trbe_parse_gsi(struct acpi_madt_generic_interrupt *gicc)
{
	return gicc->trbe_interrupt;
}

static void arm_trbe_acpi_register_device(void)
{
	int ret = arm_acpi_register_pmu_device(&trbe_dev, ACPI_MADT_GICC_TRBE,
					       arm_trbe_parse_gsi);
	if (ret)
		pr_warn("ACPI: TRBE: Unable to register device\n");
}
#else
static inline void arm_trbe_acpi_register_device(void)
{

}
#endif /* CONFIG_CORESIGHT_TRBE */

static int arm_pmu_acpi_parse_irqs(void)
{
	int irq, cpu, irq_cpu, err;

	for_each_possible_cpu(cpu) {
		irq = arm_pmu_acpi_register_irq(cpu);
		if (irq < 0) {
			err = irq;
			pr_warn("Unable to parse ACPI PMU IRQ for CPU%d: %d\n",
				cpu, err);
			goto out_err;
		} else if (irq == 0) {
			pr_warn("No ACPI PMU IRQ for CPU%d\n", cpu);
		}

		/*
		 * Log and request the IRQ so the core arm_pmu code can manage
		 * it. We'll have to sanity-check IRQs later when we associate
		 * them with their PMUs.
		 */
		per_cpu(pmu_irqs, cpu) = irq;
		err = armpmu_request_irq(irq, cpu);
		if (err)
			goto out_err;
	}

	return 0;

out_err:
	for_each_possible_cpu(cpu) {
		irq = per_cpu(pmu_irqs, cpu);
		if (!irq)
			continue;

		arm_pmu_acpi_unregister_irq(cpu);

		/*
		 * Blat all copies of the IRQ so that we only unregister the
		 * corresponding GSI once (e.g. when we have PPIs).
		 */
		for_each_possible_cpu(irq_cpu) {
			if (per_cpu(pmu_irqs, irq_cpu) == irq)
				per_cpu(pmu_irqs, irq_cpu) = 0;
		}
	}

	return err;
}

static struct arm_pmu *arm_pmu_acpi_find_pmu(void)
{
	unsigned long cpuid = read_cpuid_id();
	struct arm_pmu *pmu;
	int cpu;

	for_each_possible_cpu(cpu) {
		pmu = per_cpu(probed_pmus, cpu);
		if (!pmu || pmu->acpi_cpuid != cpuid)
			continue;

		return pmu;
	}

	return NULL;
}

/*
 * Check whether the new IRQ is compatible with those already associated with
 * the PMU (e.g. we don't have mismatched PPIs).
 */
static bool pmu_irq_matches(struct arm_pmu *pmu, int irq)
{
	struct pmu_hw_events __percpu *hw_events = pmu->hw_events;
	int cpu;

	if (!irq)
		return true;

	for_each_cpu(cpu, &pmu->supported_cpus) {
		int other_irq = per_cpu(hw_events->irq, cpu);
		if (!other_irq)
			continue;

		if (irq == other_irq)
			continue;
		if (!irq_is_percpu_devid(irq) && !irq_is_percpu_devid(other_irq))
			continue;

		pr_warn("mismatched PPIs detected\n");
		return false;
	}

	return true;
}

static void arm_pmu_acpi_associate_pmu_cpu(struct arm_pmu *pmu,
					   unsigned int cpu)
{
	int irq = per_cpu(pmu_irqs, cpu);

	per_cpu(probed_pmus, cpu) = pmu;

	if (pmu_irq_matches(pmu, irq)) {
		struct pmu_hw_events __percpu *hw_events;
		hw_events = pmu->hw_events;
		per_cpu(hw_events->irq, cpu) = irq;
	}

	cpumask_set_cpu(cpu, &pmu->supported_cpus);
}

/*
 * This must run before the common arm_pmu hotplug logic, so that we can
 * associate a CPU and its interrupt before the common code tries to manage the
 * affinity and so on.
 *
 * Note that hotplug events are serialized, so we cannot race with another CPU
 * coming up. The perf core won't open events while a hotplug event is in
 * progress.
 */
static int arm_pmu_acpi_cpu_starting(unsigned int cpu)
{
	struct arm_pmu *pmu;

	/* If we've already probed this CPU, we have nothing to do */
	if (per_cpu(probed_pmus, cpu))
		return 0;

	pmu = arm_pmu_acpi_find_pmu();
	if (!pmu) {
		pr_warn_ratelimited("Unable to associate CPU%d with a PMU\n",
				    cpu);
		return 0;
	}

	arm_pmu_acpi_associate_pmu_cpu(pmu, cpu);
	return 0;
}

static void arm_pmu_acpi_probe_matching_cpus(struct arm_pmu *pmu,
					     unsigned long cpuid)
{
	int cpu;

	for_each_online_cpu(cpu) {
		unsigned long cpu_cpuid = per_cpu(cpu_data, cpu).reg_midr;

		if (cpu_cpuid == cpuid)
			arm_pmu_acpi_associate_pmu_cpu(pmu, cpu);
	}
}

int arm_pmu_acpi_probe(armpmu_init_fn init_fn)
{
	int pmu_idx = 0;
	unsigned int cpu;
	int ret;

	ret = arm_pmu_acpi_parse_irqs();
	if (ret)
		return ret;

	ret = cpuhp_setup_state_nocalls(CPUHP_AP_PERF_ARM_ACPI_STARTING,
					"perf/arm/pmu_acpi:starting",
					arm_pmu_acpi_cpu_starting, NULL);
	if (ret)
		return ret;

	/*
	 * Initialise and register the set of PMUs which we know about right
	 * now. Ideally we'd do this in arm_pmu_acpi_cpu_starting() so that we
	 * could handle late hotplug, but this may lead to deadlock since we
	 * might try to register a hotplug notifier instance from within a
	 * hotplug notifier.
	 *
	 * There's also the problem of having access to the right init_fn,
	 * without tying this too deeply into the "real" PMU driver.
	 *
	 * For the moment, as with the platform/DT case, we need at least one
	 * of a PMU's CPUs to be online at probe time.
	 */
	for_each_online_cpu(cpu) {
		struct arm_pmu *pmu = per_cpu(probed_pmus, cpu);
		unsigned long cpuid;
		char *base_name;

		/* If we've already probed this CPU, we have nothing to do */
		if (pmu)
			continue;

		pmu = armpmu_alloc();
		if (!pmu) {
			pr_warn("Unable to allocate PMU for CPU%d\n",
				cpu);
			return -ENOMEM;
		}

		cpuid = per_cpu(cpu_data, cpu).reg_midr;
		pmu->acpi_cpuid = cpuid;

		arm_pmu_acpi_probe_matching_cpus(pmu, cpuid);

		ret = init_fn(pmu);
		if (ret == -ENODEV) {
			/* PMU not handled by this driver, or not present */
			continue;
		} else if (ret) {
			pr_warn("Unable to initialise PMU for CPU%d\n", cpu);
			return ret;
		}

		base_name = pmu->name;
		pmu->name = kasprintf(GFP_KERNEL, "%s_%d", base_name, pmu_idx++);
		if (!pmu->name) {
			pr_warn("Unable to allocate PMU name for CPU%d\n", cpu);
			return -ENOMEM;
		}

		ret = armpmu_register(pmu);
		if (ret) {
			pr_warn("Failed to register PMU for CPU%d\n", cpu);
			kfree(pmu->name);
			return ret;
		}
	}

	return ret;
}

static int arm_pmu_acpi_init(void)
{
	if (acpi_disabled)
		return 0;

	arm_spe_acpi_register_device();
	arm_trbe_acpi_register_device();

	return 0;
}
subsys_initcall(arm_pmu_acpi_init)