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
|
// SPDX-License-Identifier: MIT
/*
* Copyright © 2019 Intel Corporation
*/
#include "intel_context.h"
#include "intel_engine_pm.h"
#include "intel_gpu_commands.h"
#include "intel_gt_requests.h"
#include "intel_ring.h"
#include "selftest_rc6.h"
#include "selftests/i915_random.h"
#include "selftests/librapl.h"
static u64 rc6_residency(struct intel_rc6 *rc6)
{
u64 result;
/* XXX VLV_GT_MEDIA_RC6? */
result = intel_rc6_residency_ns(rc6, INTEL_RC6_RES_RC6);
if (HAS_RC6p(rc6_to_i915(rc6)))
result += intel_rc6_residency_ns(rc6, INTEL_RC6_RES_RC6p);
if (HAS_RC6pp(rc6_to_i915(rc6)))
result += intel_rc6_residency_ns(rc6, INTEL_RC6_RES_RC6pp);
return result;
}
int live_rc6_manual(void *arg)
{
struct intel_gt *gt = arg;
struct intel_rc6 *rc6 = >->rc6;
u64 rc0_power, rc6_power;
intel_wakeref_t wakeref;
bool has_power;
ktime_t dt;
u64 res[2];
int err = 0;
/*
* Our claim is that we can "encourage" the GPU to enter rc6 at will.
* Let's try it!
*/
if (!rc6->enabled)
return 0;
/* bsw/byt use a PCU and decouple RC6 from our manual control */
if (IS_VALLEYVIEW(gt->i915) || IS_CHERRYVIEW(gt->i915))
return 0;
has_power = librapl_supported(gt->i915);
wakeref = intel_runtime_pm_get(gt->uncore->rpm);
/* Force RC6 off for starters */
__intel_rc6_disable(rc6);
msleep(1); /* wakeup is not immediate, takes about 100us on icl */
res[0] = rc6_residency(rc6);
dt = ktime_get();
rc0_power = librapl_energy_uJ();
msleep(250);
rc0_power = librapl_energy_uJ() - rc0_power;
dt = ktime_sub(ktime_get(), dt);
res[1] = rc6_residency(rc6);
if ((res[1] - res[0]) >> 10) {
pr_err("RC6 residency increased by %lldus while disabled for 250ms!\n",
(res[1] - res[0]) >> 10);
err = -EINVAL;
goto out_unlock;
}
if (has_power) {
rc0_power = div64_u64(NSEC_PER_SEC * rc0_power,
ktime_to_ns(dt));
if (!rc0_power) {
pr_err("No power measured while in RC0\n");
err = -EINVAL;
goto out_unlock;
}
}
/* Manually enter RC6 */
intel_rc6_park(rc6);
res[0] = rc6_residency(rc6);
intel_uncore_forcewake_flush(rc6_to_uncore(rc6), FORCEWAKE_ALL);
dt = ktime_get();
rc6_power = librapl_energy_uJ();
msleep(100);
rc6_power = librapl_energy_uJ() - rc6_power;
dt = ktime_sub(ktime_get(), dt);
res[1] = rc6_residency(rc6);
if (res[1] == res[0]) {
pr_err("Did not enter RC6! RC6_STATE=%08x, RC6_CONTROL=%08x, residency=%lld\n",
intel_uncore_read_fw(gt->uncore, GEN6_RC_STATE),
intel_uncore_read_fw(gt->uncore, GEN6_RC_CONTROL),
res[0]);
err = -EINVAL;
}
if (has_power) {
rc6_power = div64_u64(NSEC_PER_SEC * rc6_power,
ktime_to_ns(dt));
pr_info("GPU consumed %llduW in RC0 and %llduW in RC6\n",
rc0_power, rc6_power);
if (2 * rc6_power > rc0_power) {
pr_err("GPU leaked energy while in RC6!\n");
err = -EINVAL;
goto out_unlock;
}
}
/* Restore what should have been the original state! */
intel_rc6_unpark(rc6);
out_unlock:
intel_runtime_pm_put(gt->uncore->rpm, wakeref);
return err;
}
static const u32 *__live_rc6_ctx(struct intel_context *ce)
{
struct i915_request *rq;
const u32 *result;
u32 cmd;
u32 *cs;
rq = intel_context_create_request(ce);
if (IS_ERR(rq))
return ERR_CAST(rq);
cs = intel_ring_begin(rq, 4);
if (IS_ERR(cs)) {
i915_request_add(rq);
return cs;
}
cmd = MI_STORE_REGISTER_MEM | MI_USE_GGTT;
if (GRAPHICS_VER(rq->i915) >= 8)
cmd++;
*cs++ = cmd;
*cs++ = i915_mmio_reg_offset(GEN8_RC6_CTX_INFO);
*cs++ = ce->timeline->hwsp_offset + 8;
*cs++ = 0;
intel_ring_advance(rq, cs);
result = rq->hwsp_seqno + 2;
i915_request_add(rq);
return result;
}
static struct intel_engine_cs **
randomised_engines(struct intel_gt *gt,
struct rnd_state *prng,
unsigned int *count)
{
struct intel_engine_cs *engine, **engines;
enum intel_engine_id id;
int n;
n = 0;
for_each_engine(engine, gt, id)
n++;
if (!n)
return NULL;
engines = kmalloc_array(n, sizeof(*engines), GFP_KERNEL);
if (!engines)
return NULL;
n = 0;
for_each_engine(engine, gt, id)
engines[n++] = engine;
i915_prandom_shuffle(engines, sizeof(*engines), n, prng);
*count = n;
return engines;
}
int live_rc6_ctx_wa(void *arg)
{
struct intel_gt *gt = arg;
struct intel_engine_cs **engines;
unsigned int n, count;
I915_RND_STATE(prng);
int err = 0;
/* A read of CTX_INFO upsets rc6. Poke the bear! */
if (GRAPHICS_VER(gt->i915) < 8)
return 0;
engines = randomised_engines(gt, &prng, &count);
if (!engines)
return 0;
for (n = 0; n < count; n++) {
struct intel_engine_cs *engine = engines[n];
int pass;
for (pass = 0; pass < 2; pass++) {
struct i915_gpu_error *error = >->i915->gpu_error;
struct intel_context *ce;
unsigned int resets =
i915_reset_engine_count(error, engine);
const u32 *res;
/* Use a sacrifical context */
ce = intel_context_create(engine);
if (IS_ERR(ce)) {
err = PTR_ERR(ce);
goto out;
}
intel_engine_pm_get(engine);
res = __live_rc6_ctx(ce);
intel_engine_pm_put(engine);
intel_context_put(ce);
if (IS_ERR(res)) {
err = PTR_ERR(res);
goto out;
}
if (intel_gt_wait_for_idle(gt, HZ / 5) == -ETIME) {
intel_gt_set_wedged(gt);
err = -ETIME;
goto out;
}
intel_gt_pm_wait_for_idle(gt);
pr_debug("%s: CTX_INFO=%0x\n",
engine->name, READ_ONCE(*res));
if (resets !=
i915_reset_engine_count(error, engine)) {
pr_err("%s: GPU reset required\n",
engine->name);
add_taint_for_CI(gt->i915, TAINT_WARN);
err = -EIO;
goto out;
}
}
}
out:
kfree(engines);
return err;
}
|
