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
|
From 4cf7afe98bc8cfef0cedc148cf9a834ed43f6c5f Mon Sep 17 00:00:00 2001
From: James Zhu <James.Zhu@amd.com>
Date: Fri, 5 Apr 2019 15:31:47 -0400
Subject: [PATCH 1907/2940] drm/amdgpu: add gfx9 gpr EDC workaround when RAS is
enabled
When RAS is enabled, initializes the VGPRs/LDS/SGPRs and
resets EDC error counts. This is done in late_init, before
RAS TA GFX enable.
Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 245 ++++++++++++++++++++++++++
drivers/gpu/drm/amd/amdgpu/soc15.h | 10 ++
2 files changed, 255 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index c05754e19beb..701a43563f5d 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -34,6 +34,7 @@
#include "vega10_enum.h"
#include "hdp/hdp_4_0_offset.h"
+#include "soc15.h"
#include "soc15_common.h"
#include "clearstate_gfx9.h"
#include "v9_structs.h"
@@ -3533,6 +3534,245 @@ static void gfx_v9_0_ring_emit_gds_switch(struct amdgpu_ring *ring,
(1 << (oa_size + oa_base)) - (1 << oa_base));
}
+static const u32 vgpr_init_compute_shader[] =
+{
+ 0xb07c0000, 0xbe8000ff,
+ 0x000000f8, 0xbf110800,
+ 0x7e000280, 0x7e020280,
+ 0x7e040280, 0x7e060280,
+ 0x7e080280, 0x7e0a0280,
+ 0x7e0c0280, 0x7e0e0280,
+ 0x80808800, 0xbe803200,
+ 0xbf84fff5, 0xbf9c0000,
+ 0xd28c0001, 0x0001007f,
+ 0xd28d0001, 0x0002027e,
+ 0x10020288, 0xb8810904,
+ 0xb7814000, 0xd1196a01,
+ 0x00000301, 0xbe800087,
+ 0xbefc00c1, 0xd89c4000,
+ 0x00020201, 0xd89cc080,
+ 0x00040401, 0x320202ff,
+ 0x00000800, 0x80808100,
+ 0xbf84fff8, 0x7e020280,
+ 0xbf810000, 0x00000000,
+};
+
+static const u32 sgpr_init_compute_shader[] =
+{
+ 0xb07c0000, 0xbe8000ff,
+ 0x0000005f, 0xbee50080,
+ 0xbe812c65, 0xbe822c65,
+ 0xbe832c65, 0xbe842c65,
+ 0xbe852c65, 0xb77c0005,
+ 0x80808500, 0xbf84fff8,
+ 0xbe800080, 0xbf810000,
+};
+
+static const struct soc15_reg_entry vgpr_init_regs[] = {
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x1000000 }, /* CU_GROUP_COUNT=1 */
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 256*2 },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 1 },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x100007f }, /* VGPRS=15 (256 logical VGPRs, SGPRS=1 (16 SGPRs, BULKY=1 */
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x400000 }, /* 64KB LDS */
+};
+
+static const struct soc15_reg_entry sgpr_init_regs[] = {
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE0), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE1), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE2), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_STATIC_THREAD_MGMT_SE3), 0xffffffff },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_RESOURCE_LIMITS), 0x1000000 }, /* CU_GROUP_COUNT=1 */
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_X), 256*2 },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Y), 1 },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_NUM_THREAD_Z), 1 },
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC1), 0x340 }, /* SGPRS=13 (112 GPRS) */
+ { SOC15_REG_ENTRY(GC, 0, mmCOMPUTE_PGM_RSRC2), 0x0 },
+};
+
+static const struct soc15_reg_entry sec_ded_counter_registers[] = {
+ { SOC15_REG_ENTRY(GC, 0, mmCPC_EDC_SCRATCH_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmCPC_EDC_UCODE_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmCPF_EDC_ROQ_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmCPF_EDC_TAG_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmCPG_EDC_DMA_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmCPG_EDC_TAG_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmDC_EDC_CSINVOC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmDC_EDC_RESTORE_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmDC_EDC_STATE_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmGDS_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmGDS_EDC_GRBM_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmGDS_EDC_OA_DED) },
+ { SOC15_REG_ENTRY(GC, 0, mmSPI_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmSQC_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmSQ_EDC_DED_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmSQ_EDC_INFO) },
+ { SOC15_REG_ENTRY(GC, 0, mmSQ_EDC_SEC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmTCC_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmTCP_ATC_EDC_GATCL1_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmTCP_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmTD_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmSQC_EDC_CNT2) },
+ { SOC15_REG_ENTRY(GC, 0, mmSQ_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmTA_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmGDS_EDC_OA_PHY_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmGDS_EDC_OA_PIPE_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmGCEA_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmGCEA_EDC_CNT2) },
+ { SOC15_REG_ENTRY(GC, 0, mmTCI_EDC_CNT) },
+ { SOC15_REG_ENTRY(GC, 0, mmTCC_EDC_CNT2) },
+ { SOC15_REG_ENTRY(GC, 0, mmTCA_EDC_CNT) },
+};
+
+static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
+{
+ struct amdgpu_ring *ring = &adev->gfx.compute_ring[0];
+ struct amdgpu_ib ib;
+ struct dma_fence *f = NULL;
+ int r, i, j;
+ u32 tmp;
+ unsigned total_size, vgpr_offset, sgpr_offset;
+ u64 gpu_addr;
+
+ /* only support when RAS is enabled */
+ if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
+ return 0;
+
+ /* bail if the compute ring is not ready */
+ if (!ring->sched.ready)
+ return 0;
+
+ tmp = RREG32_SOC15(GC, 0, mmGB_EDC_MODE);
+ WREG32_SOC15(GC, 0, mmGB_EDC_MODE, 0);
+
+ total_size =
+ ((ARRAY_SIZE(vgpr_init_regs) * 3) + 4 + 5 + 2) * 4;
+ total_size +=
+ ((ARRAY_SIZE(sgpr_init_regs) * 3) + 4 + 5 + 2) * 4;
+ total_size = ALIGN(total_size, 256);
+ vgpr_offset = total_size;
+ total_size += ALIGN(sizeof(vgpr_init_compute_shader), 256);
+ sgpr_offset = total_size;
+ total_size += sizeof(sgpr_init_compute_shader);
+
+ /* allocate an indirect buffer to put the commands in */
+ memset(&ib, 0, sizeof(ib));
+ r = amdgpu_ib_get(adev, NULL, total_size, &ib);
+ if (r) {
+ DRM_ERROR("amdgpu: failed to get ib (%d).\n", r);
+ return r;
+ }
+
+ /* load the compute shaders */
+ for (i = 0; i < ARRAY_SIZE(vgpr_init_compute_shader); i++)
+ ib.ptr[i + (vgpr_offset / 4)] = vgpr_init_compute_shader[i];
+
+ for (i = 0; i < ARRAY_SIZE(sgpr_init_compute_shader); i++)
+ ib.ptr[i + (sgpr_offset / 4)] = sgpr_init_compute_shader[i];
+
+ /* init the ib length to 0 */
+ ib.length_dw = 0;
+
+ /* VGPR */
+ /* write the register state for the compute dispatch */
+ for (i = 0; i < ARRAY_SIZE(vgpr_init_regs); i++) {
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1);
+ ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(vgpr_init_regs[i])
+ - PACKET3_SET_SH_REG_START;
+ ib.ptr[ib.length_dw++] = vgpr_init_regs[i].reg_value;
+ }
+ /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */
+ gpu_addr = (ib.gpu_addr + (u64)vgpr_offset) >> 8;
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 2);
+ ib.ptr[ib.length_dw++] = SOC15_REG_OFFSET(GC, 0, mmCOMPUTE_PGM_LO)
+ - PACKET3_SET_SH_REG_START;
+ ib.ptr[ib.length_dw++] = lower_32_bits(gpu_addr);
+ ib.ptr[ib.length_dw++] = upper_32_bits(gpu_addr);
+
+ /* write dispatch packet */
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3);
+ ib.ptr[ib.length_dw++] = 128; /* x */
+ ib.ptr[ib.length_dw++] = 1; /* y */
+ ib.ptr[ib.length_dw++] = 1; /* z */
+ ib.ptr[ib.length_dw++] =
+ REG_SET_FIELD(0, COMPUTE_DISPATCH_INITIATOR, COMPUTE_SHADER_EN, 1);
+
+ /* write CS partial flush packet */
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_EVENT_WRITE, 0);
+ ib.ptr[ib.length_dw++] = EVENT_TYPE(7) | EVENT_INDEX(4);
+
+ /* SGPR */
+ /* write the register state for the compute dispatch */
+ for (i = 0; i < ARRAY_SIZE(sgpr_init_regs); i++) {
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 1);
+ ib.ptr[ib.length_dw++] = SOC15_REG_ENTRY_OFFSET(sgpr_init_regs[i])
+ - PACKET3_SET_SH_REG_START;
+ ib.ptr[ib.length_dw++] = sgpr_init_regs[i].reg_value;
+ }
+ /* write the shader start address: mmCOMPUTE_PGM_LO, mmCOMPUTE_PGM_HI */
+ gpu_addr = (ib.gpu_addr + (u64)sgpr_offset) >> 8;
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_SET_SH_REG, 2);
+ ib.ptr[ib.length_dw++] = SOC15_REG_OFFSET(GC, 0, mmCOMPUTE_PGM_LO)
+ - PACKET3_SET_SH_REG_START;
+ ib.ptr[ib.length_dw++] = lower_32_bits(gpu_addr);
+ ib.ptr[ib.length_dw++] = upper_32_bits(gpu_addr);
+
+ /* write dispatch packet */
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_DISPATCH_DIRECT, 3);
+ ib.ptr[ib.length_dw++] = 128; /* x */
+ ib.ptr[ib.length_dw++] = 1; /* y */
+ ib.ptr[ib.length_dw++] = 1; /* z */
+ ib.ptr[ib.length_dw++] =
+ REG_SET_FIELD(0, COMPUTE_DISPATCH_INITIATOR, COMPUTE_SHADER_EN, 1);
+
+ /* write CS partial flush packet */
+ ib.ptr[ib.length_dw++] = PACKET3(PACKET3_EVENT_WRITE, 0);
+ ib.ptr[ib.length_dw++] = EVENT_TYPE(7) | EVENT_INDEX(4);
+
+ /* shedule the ib on the ring */
+ r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+ if (r) {
+ DRM_ERROR("amdgpu: ib submit failed (%d).\n", r);
+ goto fail;
+ }
+
+ /* wait for the GPU to finish processing the IB */
+ r = dma_fence_wait(f, false);
+ if (r) {
+ DRM_ERROR("amdgpu: fence wait failed (%d).\n", r);
+ goto fail;
+ }
+
+ /* read back registers to clear the counters */
+ mutex_lock(&adev->grbm_idx_mutex);
+ for (j = 0; j < 16; j++) {
+ gfx_v9_0_select_se_sh(adev, 0x01, 0x0, j);
+ for (i = 0; i < ARRAY_SIZE(sec_ded_counter_registers); i++)
+ RREG32(SOC15_REG_ENTRY_OFFSET(sec_ded_counter_registers[i]));
+ gfx_v9_0_select_se_sh(adev, 0x02, 0x0, j);
+ for (i = 0; i < ARRAY_SIZE(sec_ded_counter_registers); i++)
+ RREG32(SOC15_REG_ENTRY_OFFSET(sec_ded_counter_registers[i]));
+ gfx_v9_0_select_se_sh(adev, 0x03, 0x0, j);
+ for (i = 0; i < ARRAY_SIZE(sec_ded_counter_registers); i++)
+ RREG32(SOC15_REG_ENTRY_OFFSET(sec_ded_counter_registers[i]));
+ gfx_v9_0_select_se_sh(adev, 0x04, 0x0, j);
+ for (i = 0; i < ARRAY_SIZE(sec_ded_counter_registers); i++)
+ RREG32(SOC15_REG_ENTRY_OFFSET(sec_ded_counter_registers[i]));
+ }
+ WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xe0000000);
+ mutex_unlock(&adev->grbm_idx_mutex);
+
+fail:
+ amdgpu_ib_free(adev, &ib, NULL);
+ dma_fence_put(f);
+
+ return r;
+}
+
static int gfx_v9_0_early_init(void *handle)
{
struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -3581,6 +3821,11 @@ static int gfx_v9_0_ecc_late_init(void *handle)
if (!*ras_if)
return -ENOMEM;
+ /* requires IBs so do in late init after IB pool is initialized */
+ r = gfx_v9_0_do_edc_gpr_workarounds(adev);
+ if (r)
+ return r;
+
**ras_if = ras_block;
r = amdgpu_ras_feature_enable_on_boot(adev, *ras_if, 1);
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.h b/drivers/gpu/drm/amd/amdgpu/soc15.h
index a66c8bfbbaa6..06f39f5bbf76 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.h
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.h
@@ -42,8 +42,18 @@ struct soc15_reg_golden {
u32 or_mask;
};
+struct soc15_reg_entry {
+ uint32_t hwip;
+ uint32_t inst;
+ uint32_t seg;
+ uint32_t reg_offset;
+ uint32_t reg_value;
+};
+
#define SOC15_REG_ENTRY(ip, inst, reg) ip##_HWIP, inst, reg##_BASE_IDX, reg
+#define SOC15_REG_ENTRY_OFFSET(entry) (adev->reg_offset[entry.hwip][entry.inst][entry.seg] + entry.reg_offset)
+
#define SOC15_REG_GOLDEN_VALUE(ip, inst, reg, and_mask, or_mask) \
{ ip##_HWIP, inst, reg##_BASE_IDX, reg, and_mask, or_mask }
--
2.17.1
|
