summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
blob: 2a0c0634d446ed0b59871374d25104c06214dc87 (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
/*
 * SPDX-License-Identifier: MIT
 *
 * Copyright © 2017 Intel Corporation
 */

#include <linux/prime_numbers.h>

#include "gt/intel_engine_pm.h"
#include "gt/intel_gpu_commands.h"
#include "gt/intel_gt.h"
#include "gt/intel_gt_pm.h"
#include "gt/intel_ring.h"

#include "i915_selftest.h"
#include "selftests/i915_random.h"

struct context {
	struct drm_i915_gem_object *obj;
	struct intel_engine_cs *engine;
};

static int cpu_set(struct context *ctx, unsigned long offset, u32 v)
{
	unsigned int needs_clflush;
	struct page *page;
	u32 *cpu;
	int err;

	i915_gem_object_lock(ctx->obj, NULL);
	err = i915_gem_object_prepare_write(ctx->obj, &needs_clflush);
	if (err)
		goto out;

	page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT);
	cpu = kmap_local_page(page) + offset_in_page(offset);

	if (needs_clflush & CLFLUSH_BEFORE)
		drm_clflush_virt_range(cpu, sizeof(*cpu));

	*cpu = v;

	if (needs_clflush & CLFLUSH_AFTER)
		drm_clflush_virt_range(cpu, sizeof(*cpu));

	kunmap_local(cpu);
	i915_gem_object_finish_access(ctx->obj);

out:
	i915_gem_object_unlock(ctx->obj);
	return err;
}

static int cpu_get(struct context *ctx, unsigned long offset, u32 *v)
{
	unsigned int needs_clflush;
	struct page *page;
	u32 *cpu;
	int err;

	i915_gem_object_lock(ctx->obj, NULL);
	err = i915_gem_object_prepare_read(ctx->obj, &needs_clflush);
	if (err)
		goto out;

	page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT);
	cpu = kmap_local_page(page) + offset_in_page(offset);

	if (needs_clflush & CLFLUSH_BEFORE)
		drm_clflush_virt_range(cpu, sizeof(*cpu));

	*v = *cpu;

	kunmap_local(cpu);
	i915_gem_object_finish_access(ctx->obj);

out:
	i915_gem_object_unlock(ctx->obj);
	return err;
}

static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
{
	intel_wakeref_t wakeref;
	struct i915_vma *vma;
	u32 __iomem *map;
	int err = 0;

	i915_gem_object_lock(ctx->obj, NULL);
	err = i915_gem_object_set_to_gtt_domain(ctx->obj, true);
	i915_gem_object_unlock(ctx->obj);
	if (err)
		return err;

	vma = i915_gem_object_ggtt_pin(ctx->obj, NULL, 0, 0, PIN_MAPPABLE);
	if (IS_ERR(vma))
		return PTR_ERR(vma);

	wakeref = intel_gt_pm_get(vma->vm->gt);

	map = i915_vma_pin_iomap(vma);
	i915_vma_unpin(vma);
	if (IS_ERR(map)) {
		err = PTR_ERR(map);
		goto out_rpm;
	}

	iowrite32(v, &map[offset / sizeof(*map)]);
	i915_vma_unpin_iomap(vma);

out_rpm:
	intel_gt_pm_put(vma->vm->gt, wakeref);
	return err;
}

static int gtt_get(struct context *ctx, unsigned long offset, u32 *v)
{
	intel_wakeref_t wakeref;
	struct i915_vma *vma;
	u32 __iomem *map;
	int err = 0;

	i915_gem_object_lock(ctx->obj, NULL);
	err = i915_gem_object_set_to_gtt_domain(ctx->obj, false);
	i915_gem_object_unlock(ctx->obj);
	if (err)
		return err;

	vma = i915_gem_object_ggtt_pin(ctx->obj, NULL, 0, 0, PIN_MAPPABLE);
	if (IS_ERR(vma))
		return PTR_ERR(vma);

	wakeref = intel_gt_pm_get(vma->vm->gt);

	map = i915_vma_pin_iomap(vma);
	i915_vma_unpin(vma);
	if (IS_ERR(map)) {
		err = PTR_ERR(map);
		goto out_rpm;
	}

	*v = ioread32(&map[offset / sizeof(*map)]);
	i915_vma_unpin_iomap(vma);

out_rpm:
	intel_gt_pm_put(vma->vm->gt, wakeref);
	return err;
}

static int wc_set(struct context *ctx, unsigned long offset, u32 v)
{
	u32 *map;
	int err;

	i915_gem_object_lock(ctx->obj, NULL);
	err = i915_gem_object_set_to_wc_domain(ctx->obj, true);
	i915_gem_object_unlock(ctx->obj);
	if (err)
		return err;

	map = i915_gem_object_pin_map_unlocked(ctx->obj, I915_MAP_WC);
	if (IS_ERR(map))
		return PTR_ERR(map);

	map[offset / sizeof(*map)] = v;

	__i915_gem_object_flush_map(ctx->obj, offset, sizeof(*map));
	i915_gem_object_unpin_map(ctx->obj);

	return 0;
}

static int wc_get(struct context *ctx, unsigned long offset, u32 *v)
{
	u32 *map;
	int err;

	i915_gem_object_lock(ctx->obj, NULL);
	err = i915_gem_object_set_to_wc_domain(ctx->obj, false);
	i915_gem_object_unlock(ctx->obj);
	if (err)
		return err;

	map = i915_gem_object_pin_map_unlocked(ctx->obj, I915_MAP_WC);
	if (IS_ERR(map))
		return PTR_ERR(map);

	*v = map[offset / sizeof(*map)];
	i915_gem_object_unpin_map(ctx->obj);

	return 0;
}

static int gpu_set(struct context *ctx, unsigned long offset, u32 v)
{
	struct i915_request *rq;
	struct i915_vma *vma;
	u32 *cs;
	int err;

	vma = i915_gem_object_ggtt_pin(ctx->obj, NULL, 0, 0, 0);
	if (IS_ERR(vma))
		return PTR_ERR(vma);

	i915_gem_object_lock(ctx->obj, NULL);
	err = i915_gem_object_set_to_gtt_domain(ctx->obj, true);
	if (err)
		goto out_unlock;

	rq = intel_engine_create_kernel_request(ctx->engine);
	if (IS_ERR(rq)) {
		err = PTR_ERR(rq);
		goto out_unpin;
	}

	cs = intel_ring_begin(rq, 4);
	if (IS_ERR(cs)) {
		err = PTR_ERR(cs);
		goto out_rq;
	}

	if (GRAPHICS_VER(ctx->engine->i915) >= 8) {
		*cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT;
		*cs++ = lower_32_bits(i915_ggtt_offset(vma) + offset);
		*cs++ = upper_32_bits(i915_ggtt_offset(vma) + offset);
		*cs++ = v;
	} else if (GRAPHICS_VER(ctx->engine->i915) >= 4) {
		*cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT;
		*cs++ = 0;
		*cs++ = i915_ggtt_offset(vma) + offset;
		*cs++ = v;
	} else {
		*cs++ = MI_STORE_DWORD_IMM | MI_MEM_VIRTUAL;
		*cs++ = i915_ggtt_offset(vma) + offset;
		*cs++ = v;
		*cs++ = MI_NOOP;
	}
	intel_ring_advance(rq, cs);

	err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);

out_rq:
	i915_request_add(rq);
out_unpin:
	i915_vma_unpin(vma);
out_unlock:
	i915_gem_object_unlock(ctx->obj);

	return err;
}

static bool always_valid(struct context *ctx)
{
	return true;
}

static bool needs_fence_registers(struct context *ctx)
{
	struct intel_gt *gt = ctx->engine->gt;

	if (intel_gt_is_wedged(gt))
		return false;

	return gt->ggtt->num_fences;
}

static bool needs_mi_store_dword(struct context *ctx)
{
	if (intel_gt_is_wedged(ctx->engine->gt))
		return false;

	return intel_engine_can_store_dword(ctx->engine);
}

static const struct igt_coherency_mode {
	const char *name;
	int (*set)(struct context *ctx, unsigned long offset, u32 v);
	int (*get)(struct context *ctx, unsigned long offset, u32 *v);
	bool (*valid)(struct context *ctx);
} igt_coherency_mode[] = {
	{ "cpu", cpu_set, cpu_get, always_valid },
	{ "gtt", gtt_set, gtt_get, needs_fence_registers },
	{ "wc", wc_set, wc_get, always_valid },
	{ "gpu", gpu_set, NULL, needs_mi_store_dword },
	{ },
};

static struct intel_engine_cs *
random_engine(struct drm_i915_private *i915, struct rnd_state *prng)
{
	struct intel_engine_cs *engine;
	unsigned int count;

	count = 0;
	for_each_uabi_engine(engine, i915)
		count++;

	count = i915_prandom_u32_max_state(count, prng);
	for_each_uabi_engine(engine, i915)
		if (count-- == 0)
			return engine;

	return NULL;
}

static int igt_gem_coherency(void *arg)
{
	const unsigned int ncachelines = PAGE_SIZE/64;
	struct drm_i915_private *i915 = arg;
	const struct igt_coherency_mode *read, *write, *over;
	unsigned long count, n;
	u32 *offsets, *values;
	I915_RND_STATE(prng);
	struct context ctx;
	int err = 0;

	/*
	 * We repeatedly write, overwrite and read from a sequence of
	 * cachelines in order to try and detect incoherency (unflushed writes
	 * from either the CPU or GPU). Each setter/getter uses our cache
	 * domain API which should prevent incoherency.
	 */

	offsets = kmalloc_array(ncachelines, 2*sizeof(u32), GFP_KERNEL);
	if (!offsets)
		return -ENOMEM;
	for (count = 0; count < ncachelines; count++)
		offsets[count] = count * 64 + 4 * (count % 16);

	values = offsets + ncachelines;

	ctx.engine = random_engine(i915, &prng);
	if (!ctx.engine) {
		err = -ENODEV;
		goto out_free;
	}
	pr_info("%s: using %s\n", __func__, ctx.engine->name);
	intel_engine_pm_get(ctx.engine);

	for (over = igt_coherency_mode; over->name; over++) {
		if (!over->set)
			continue;

		if (!over->valid(&ctx))
			continue;

		for (write = igt_coherency_mode; write->name; write++) {
			if (!write->set)
				continue;

			if (!write->valid(&ctx))
				continue;

			for (read = igt_coherency_mode; read->name; read++) {
				if (!read->get)
					continue;

				if (!read->valid(&ctx))
					continue;

				for_each_prime_number_from(count, 1, ncachelines) {
					ctx.obj = i915_gem_object_create_internal(i915, PAGE_SIZE);
					if (IS_ERR(ctx.obj)) {
						err = PTR_ERR(ctx.obj);
						goto out_pm;
					}

					i915_random_reorder(offsets, ncachelines, &prng);
					for (n = 0; n < count; n++)
						values[n] = prandom_u32_state(&prng);

					for (n = 0; n < count; n++) {
						err = over->set(&ctx, offsets[n], ~values[n]);
						if (err) {
							pr_err("Failed to set stale value[%ld/%ld] in object using %s, err=%d\n",
							       n, count, over->name, err);
							goto put_object;
						}
					}

					for (n = 0; n < count; n++) {
						err = write->set(&ctx, offsets[n], values[n]);
						if (err) {
							pr_err("Failed to set value[%ld/%ld] in object using %s, err=%d\n",
							       n, count, write->name, err);
							goto put_object;
						}
					}

					for (n = 0; n < count; n++) {
						u32 found;

						err = read->get(&ctx, offsets[n], &found);
						if (err) {
							pr_err("Failed to get value[%ld/%ld] in object using %s, err=%d\n",
							       n, count, read->name, err);
							goto put_object;
						}

						if (found != values[n]) {
							pr_err("Value[%ld/%ld] mismatch, (overwrite with %s) wrote [%s] %x read [%s] %x (inverse %x), at offset %x\n",
							       n, count, over->name,
							       write->name, values[n],
							       read->name, found,
							       ~values[n], offsets[n]);
							err = -EINVAL;
							goto put_object;
						}
					}

					i915_gem_object_put(ctx.obj);
				}
			}
		}
	}
out_pm:
	intel_engine_pm_put(ctx.engine);
out_free:
	kfree(offsets);
	return err;

put_object:
	i915_gem_object_put(ctx.obj);
	goto out_pm;
}

int i915_gem_coherency_live_selftests(struct drm_i915_private *i915)
{
	static const struct i915_subtest tests[] = {
		SUBTEST(igt_gem_coherency),
	};

	return i915_live_subtests(tests, i915);
}