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
|
From 19681fc835b0ded70494c9663fc0a900e93da514 Mon Sep 17 00:00:00 2001
From: Alex Deucher <alexander.deucher@amd.com>
Date: Thu, 28 Sep 2017 09:47:32 -0400
Subject: [PATCH 1913/4131] drm/amdgpu: fix vf error handling
The error handling for virtual functions assumed a single
vf per VM and didn't properly account for bare metal. Make
the error arrays per device and add locking.
Reviewed-by: Gavin Wan <gavin.wan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 23 ++++++------
drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c | 54 +++++++++++++---------------
drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h | 5 ++-
drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h | 13 +++++++
4 files changed, 54 insertions(+), 41 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index cda0a88..b3d12c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -2119,6 +2119,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
mutex_init(&adev->srbm_mutex);
mutex_init(&adev->grbm_idx_mutex);
mutex_init(&adev->mn_lock);
+ mutex_init(&adev->virt.vf_errors.lock);
hash_init(adev->mn_hash);
amdgpu_check_arguments(adev);
@@ -2203,7 +2204,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_atombios_init(adev);
if (r) {
dev_err(adev->dev, "amdgpu_atombios_init failed\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
goto failed;
}
@@ -2214,7 +2215,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
if (amdgpu_vpost_needed(adev)) {
if (!adev->bios) {
dev_err(adev->dev, "no vBIOS found\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_NO_VBIOS, 0, 0);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_NO_VBIOS, 0, 0);
r = -EINVAL;
goto failed;
}
@@ -2222,7 +2223,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_atom_asic_init(adev->mode_info.atom_context);
if (r) {
dev_err(adev->dev, "gpu post error!\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_POST_ERROR, 0, 0);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_POST_ERROR, 0, 0);
goto failed;
}
} else {
@@ -2234,7 +2235,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_atomfirmware_get_clock_info(adev);
if (r) {
dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
goto failed;
}
} else {
@@ -2242,7 +2243,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_atombios_get_clock_info(adev);
if (r) {
dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
goto failed;
}
/* init i2c buses */
@@ -2254,7 +2255,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_fence_driver_init(adev);
if (r) {
dev_err(adev->dev, "amdgpu_fence_driver_init failed\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
goto failed;
}
@@ -2264,7 +2265,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_init(adev);
if (r) {
dev_err(adev->dev, "amdgpu_init failed\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
amdgpu_fini(adev);
goto failed;
}
@@ -2284,7 +2285,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_ib_pool_init(adev);
if (r) {
dev_err(adev->dev, "IB initialization failed (%d).\n", r);
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
goto failed;
}
@@ -2337,7 +2338,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
r = amdgpu_late_init(adev);
if (r) {
dev_err(adev->dev, "amdgpu_late_init failed\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
goto failed;
}
@@ -3046,7 +3047,7 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
}
} else {
dev_err(adev->dev, "asic resume failed (%d).\n", r);
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, 0, r);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ASIC_RESUME_FAIL, 0, r);
for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
if (adev->rings[i] && adev->rings[i]->sched.thread) {
kthread_unpark(adev->rings[i]->sched.thread);
@@ -3064,7 +3065,7 @@ int amdgpu_gpu_reset(struct amdgpu_device *adev)
if (r) {
/* bad news, how to tell it to userspace ? */
dev_info(adev->dev, "GPU reset failed\n");
- amdgpu_vf_error_put(AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
+ amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
}
else {
dev_info(adev->dev, "GPU reset successed!\n");
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c
index 45ac918..746b813 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.c
@@ -25,30 +25,21 @@
#include "amdgpu_vf_error.h"
#include "mxgpu_ai.h"
-#define AMDGPU_VF_ERROR_ENTRY_SIZE 16
-
-/* struct error_entry - amdgpu VF error information. */
-struct amdgpu_vf_error_buffer {
- int read_count;
- int write_count;
- uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE];
- uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE];
- uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE];
-};
-
-struct amdgpu_vf_error_buffer admgpu_vf_errors;
-
-
-void amdgpu_vf_error_put(uint16_t sub_error_code, uint16_t error_flags, uint64_t error_data)
+void amdgpu_vf_error_put(struct amdgpu_device *adev,
+ uint16_t sub_error_code,
+ uint16_t error_flags,
+ uint64_t error_data)
{
int index;
uint16_t error_code = AMDGIM_ERROR_CODE(AMDGIM_ERROR_CATEGORY_VF, sub_error_code);
- index = admgpu_vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
- admgpu_vf_errors.code [index] = error_code;
- admgpu_vf_errors.flags [index] = error_flags;
- admgpu_vf_errors.data [index] = error_data;
- admgpu_vf_errors.write_count ++;
+ mutex_lock(&adev->virt.vf_errors.lock);
+ index = adev->virt.vf_errors.write_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
+ adev->virt.vf_errors.code [index] = error_code;
+ adev->virt.vf_errors.flags [index] = error_flags;
+ adev->virt.vf_errors.data [index] = error_data;
+ adev->virt.vf_errors.write_count ++;
+ mutex_unlock(&adev->virt.vf_errors.lock);
}
@@ -58,7 +49,8 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev)
u32 data1, data2, data3;
int index;
- if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) || (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) {
+ if ((NULL == adev) || (!amdgpu_sriov_vf(adev)) ||
+ (!adev->virt.ops) || (!adev->virt.ops->trans_msg)) {
return;
}
/*
@@ -68,18 +60,22 @@ void amdgpu_vf_error_trans_all(struct amdgpu_device *adev)
return;
}
*/
+
+ mutex_lock(&adev->virt.vf_errors.lock);
/* The errors are overlay of array, correct read_count as full. */
- if (admgpu_vf_errors.write_count - admgpu_vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) {
- admgpu_vf_errors.read_count = admgpu_vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE;
+ if (adev->virt.vf_errors.write_count - adev->virt.vf_errors.read_count > AMDGPU_VF_ERROR_ENTRY_SIZE) {
+ adev->virt.vf_errors.read_count = adev->virt.vf_errors.write_count - AMDGPU_VF_ERROR_ENTRY_SIZE;
}
- while (admgpu_vf_errors.read_count < admgpu_vf_errors.write_count) {
- index =admgpu_vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
- data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX (admgpu_vf_errors.code[index], admgpu_vf_errors.flags[index]);
- data2 = admgpu_vf_errors.data[index] & 0xFFFFFFFF;
- data3 = (admgpu_vf_errors.data[index] >> 32) & 0xFFFFFFFF;
+ while (adev->virt.vf_errors.read_count < adev->virt.vf_errors.write_count) {
+ index =adev->virt.vf_errors.read_count % AMDGPU_VF_ERROR_ENTRY_SIZE;
+ data1 = AMDGIM_ERROR_CODE_FLAGS_TO_MAILBOX(adev->virt.vf_errors.code[index],
+ adev->virt.vf_errors.flags[index]);
+ data2 = adev->virt.vf_errors.data[index] & 0xFFFFFFFF;
+ data3 = (adev->virt.vf_errors.data[index] >> 32) & 0xFFFFFFFF;
adev->virt.ops->trans_msg(adev, IDH_LOG_VF_ERROR, data1, data2, data3);
- admgpu_vf_errors.read_count ++;
+ adev->virt.vf_errors.read_count ++;
}
+ mutex_unlock(&adev->virt.vf_errors.lock);
}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h
index 2a3278e..6436bd0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vf_error.h
@@ -56,7 +56,10 @@ enum AMDGIM_ERROR_CATEGORY {
AMDGIM_ERROR_CATEGORY_MAX
};
-void amdgpu_vf_error_put(uint16_t sub_error_code, uint16_t error_flags, uint64_t error_data);
+void amdgpu_vf_error_put(struct amdgpu_device *adev,
+ uint16_t sub_error_code,
+ uint16_t error_flags,
+ uint64_t error_data);
void amdgpu_vf_error_trans_all (struct amdgpu_device *adev);
#endif /* __VF_ERROR_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
index afcfb8b..e5fd0ff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
@@ -36,6 +36,18 @@ struct amdgpu_mm_table {
uint64_t gpu_addr;
};
+#define AMDGPU_VF_ERROR_ENTRY_SIZE 16
+
+/* struct error_entry - amdgpu VF error information. */
+struct amdgpu_vf_error_buffer {
+ struct mutex lock;
+ int read_count;
+ int write_count;
+ uint16_t code[AMDGPU_VF_ERROR_ENTRY_SIZE];
+ uint16_t flags[AMDGPU_VF_ERROR_ENTRY_SIZE];
+ uint64_t data[AMDGPU_VF_ERROR_ENTRY_SIZE];
+};
+
/**
* struct amdgpu_virt_ops - amdgpu device virt operations
*/
@@ -59,6 +71,7 @@ struct amdgpu_virt {
struct work_struct flr_work;
struct amdgpu_mm_table mm_table;
const struct amdgpu_virt_ops *ops;
+ struct amdgpu_vf_error_buffer vf_errors;
};
#define AMDGPU_CSA_SIZE (8 * 1024)
--
2.7.4
|
