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
|
From 73932b74c1fdca7a6b405f6d9f8d784c8ac261b9 Mon Sep 17 00:00:00 2001
From: Felix Kuehling <Felix.Kuehling@amd.com>
Date: Mon, 6 Jun 2016 17:15:39 -0400
Subject: [PATCH 1445/4131] drm/amdgpu: Add kfd2kgd APIs for dumping HQDs
This can be used by KFD for debugging features, such as dumping
HQDs in debugfs.
Change-Id: I5d42361730a61f388e83aecb5d0f9ea606b24ac6
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 71 ++++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 80 +++++++++++++++++++++++
drivers/gpu/drm/amd/include/kgd_kfd_interface.h | 14 ++++
3 files changed, 165 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
index b6559af..1e8ff65 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -105,7 +105,13 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
uint32_t queue_id, uint32_t __user *wptr,
uint32_t page_table_base);
+static int kgd_hqd_dump(struct kgd_dev *kgd,
+ uint32_t pipe_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd);
+static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+ uint32_t engine_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
uint32_t pipe_id, uint32_t queue_id);
static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
@@ -156,6 +162,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
.init_interrupts = kgd_init_interrupts,
.hqd_load = kgd_hqd_load,
.hqd_sdma_load = kgd_hqd_sdma_load,
+ .hqd_dump = kgd_hqd_dump,
+ .hqd_sdma_dump = kgd_hqd_sdma_dump,
.hqd_is_occupied = kgd_hqd_is_occupied,
.hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied,
.hqd_destroy = kgd_hqd_destroy,
@@ -397,6 +405,42 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
return 0;
}
+static int kgd_hqd_dump(struct kgd_dev *kgd,
+ uint32_t pipe_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs)
+{
+ struct amdgpu_device *adev = get_amdgpu_device(kgd);
+ uint32_t i = 0, reg;
+#define HQD_N_REGS (35+4)
+#define DUMP_REG(addr) do { \
+ if (WARN_ON_ONCE(i >= HQD_N_REGS)) \
+ break; \
+ (*dump)[i][0] = (addr) << 2; \
+ (*dump)[i++][1] = RREG32(addr); \
+ } while (0)
+
+ *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+ if (*dump == NULL)
+ return -ENOMEM;
+
+ acquire_queue(kgd, pipe_id, queue_id);
+
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE2);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE3);
+
+ for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_MQD_CONTROL; reg++)
+ DUMP_REG(reg);
+
+ release_queue(kgd);
+
+ WARN_ON_ONCE(i != HQD_N_REGS);
+ *n_regs = i;
+
+ return 0;
+}
+
static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
@@ -418,6 +462,33 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
return 0;
}
+static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+ uint32_t engine_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs)
+{
+ struct amdgpu_device *adev = get_amdgpu_device(kgd);
+ uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET +
+ queue_id * KFD_CIK_SDMA_QUEUE_OFFSET;
+ uint32_t i = 0, reg;
+#undef HQD_N_REGS
+#define HQD_N_REGS (19+4)
+
+ *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+ if (*dump == NULL)
+ return -ENOMEM;
+
+ for (reg = mmSDMA0_RLC0_RB_CNTL; reg <= mmSDMA0_RLC0_DOORBELL; reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_VIRTUAL_ADDR; reg <= mmSDMA0_RLC0_WATERMARK;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+
+ WARN_ON_ONCE(i != HQD_N_REGS);
+ *n_regs = i;
+
+ return 0;
+}
+
static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
uint32_t pipe_id, uint32_t queue_id)
{
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
index dc7d57d..f20498b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -81,7 +81,13 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
uint32_t queue_id, uint32_t __user *wptr,
uint32_t page_table_base);
+static int kgd_hqd_dump(struct kgd_dev *kgd,
+ uint32_t pipe_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd);
+static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+ uint32_t engine_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
uint32_t pipe_id, uint32_t queue_id);
static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd);
@@ -137,6 +143,8 @@ static const struct kfd2kgd_calls kfd2kgd = {
.init_interrupts = kgd_init_interrupts,
.hqd_load = kgd_hqd_load,
.hqd_sdma_load = kgd_hqd_sdma_load,
+ .hqd_dump = kgd_hqd_dump,
+ .hqd_sdma_dump = kgd_hqd_sdma_dump,
.hqd_is_occupied = kgd_hqd_is_occupied,
.hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied,
.hqd_destroy = kgd_hqd_destroy,
@@ -387,6 +395,42 @@ static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
return 0;
}
+static int kgd_hqd_dump(struct kgd_dev *kgd,
+ uint32_t pipe_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs)
+{
+ struct amdgpu_device *adev = get_amdgpu_device(kgd);
+ uint32_t i = 0, reg;
+#define HQD_N_REGS (54+4)
+#define DUMP_REG(addr) do { \
+ if (WARN_ON_ONCE(i >= HQD_N_REGS)) \
+ break; \
+ (*dump)[i][0] = (addr) << 2; \
+ (*dump)[i++][1] = RREG32(addr); \
+ } while (0)
+
+ *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+ if (*dump == NULL)
+ return -ENOMEM;
+
+ acquire_queue(kgd, pipe_id, queue_id);
+
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE0);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE1);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE2);
+ DUMP_REG(mmCOMPUTE_STATIC_THREAD_MGMT_SE3);
+
+ for (reg = mmCP_MQD_BASE_ADDR; reg <= mmCP_HQD_EOP_DONES; reg++)
+ DUMP_REG(reg);
+
+ release_queue(kgd);
+
+ WARN_ON_ONCE(i != HQD_N_REGS);
+ *n_regs = i;
+
+ return 0;
+}
+
static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
{
struct amdgpu_device *adev = get_amdgpu_device(kgd);
@@ -435,6 +479,42 @@ static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd)
return 0;
}
+static int kgd_hqd_sdma_dump(struct kgd_dev *kgd,
+ uint32_t engine_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs)
+{
+ struct amdgpu_device *adev = get_amdgpu_device(kgd);
+ uint32_t sdma_offset = engine_id * SDMA1_REGISTER_OFFSET +
+ queue_id * KFD_VI_SDMA_QUEUE_OFFSET;
+ uint32_t i = 0, reg;
+#undef HQD_N_REGS
+#define HQD_N_REGS (19+4+2+3+7)
+
+ *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+ if (*dump == NULL)
+ return -ENOMEM;
+
+ for (reg = mmSDMA0_RLC0_RB_CNTL; reg <= mmSDMA0_RLC0_DOORBELL; reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_VIRTUAL_ADDR; reg <= mmSDMA0_RLC0_WATERMARK;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_CSA_ADDR_LO; reg <= mmSDMA0_RLC0_CSA_ADDR_HI;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_IB_SUB_REMAIN; reg <= mmSDMA0_RLC0_DUMMY_REG;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+ for (reg = mmSDMA0_RLC0_MIDCMD_DATA0; reg <= mmSDMA0_RLC0_MIDCMD_CNTL;
+ reg++)
+ DUMP_REG(sdma_offset + reg);
+
+ WARN_ON_ONCE(i != HQD_N_REGS);
+ *n_regs = i;
+
+ return 0;
+}
+
static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address,
uint32_t pipe_id, uint32_t queue_id)
{
diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
index 3142551..3dca616 100644
--- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
+++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h
@@ -172,6 +172,12 @@ struct kgd2kfd_shared_resources {
* @hqd_sdma_load: Loads the SDMA mqd structure to a H/W SDMA hqd slot.
* used only for no HWS mode.
*
+ * @hqd_dump: Dumps CPC HQD registers to an array of address-value pairs.
+ * Array is allocated with kmalloc, needs to be freed with kfree by caller.
+ *
+ * @hqd_sdma_dump: Dumps SDMA HQD registers to an array of address-value pairs.
+ * Array is allocated with kmalloc, needs to be freed with kfree by caller.
+ *
* @hqd_is_occupies: Checks if a hqd slot is occupied.
*
* @hqd_destroy: Destructs and preempts the queue assigned to that hqd slot.
@@ -246,6 +252,14 @@ struct kfd2kgd_calls {
int (*hqd_sdma_load)(struct kgd_dev *kgd, void *mqd);
+ int (*hqd_dump)(struct kgd_dev *kgd,
+ uint32_t pipe_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
+
+ int (*hqd_sdma_dump)(struct kgd_dev *kgd,
+ uint32_t engine_id, uint32_t queue_id,
+ uint32_t (**dump)[2], uint32_t *n_regs);
+
bool (*hqd_is_occupied)(struct kgd_dev *kgd, uint64_t queue_address,
uint32_t pipe_id, uint32_t queue_id);
--
2.7.4
|
