1 /*
2 * Copyright 2016 Advanced Micro Devices, Inc.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 *
22 */
23 #include "amdgpu.h"
24 #include "amdgpu_atombios.h"
25 #include "nbio_v7_0.h"
26
27 #include "nbio/nbio_7_0_default.h"
28 #include "nbio/nbio_7_0_offset.h"
29 #include "nbio/nbio_7_0_sh_mask.h"
30 #include "nbio/nbio_7_0_smn.h"
31 #include "vega10_enum.h"
32 #include <uapi/linux/kfd_ioctl.h>
33
34 #define smnNBIF_MGCG_CTRL_LCLK 0x1013a05c
35
nbio_v7_0_remap_hdp_registers(struct amdgpu_device * adev)36 static void nbio_v7_0_remap_hdp_registers(struct amdgpu_device *adev)
37 {
38 WREG32_SOC15(NBIO, 0, mmREMAP_HDP_MEM_FLUSH_CNTL,
39 adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL);
40 WREG32_SOC15(NBIO, 0, mmREMAP_HDP_REG_FLUSH_CNTL,
41 adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL);
42 }
43
nbio_v7_0_get_rev_id(struct amdgpu_device * adev)44 static u32 nbio_v7_0_get_rev_id(struct amdgpu_device *adev)
45 {
46 u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0);
47
48 tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
49 tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
50
51 return tmp;
52 }
53
nbio_v7_0_mc_access_enable(struct amdgpu_device * adev,bool enable)54 static void nbio_v7_0_mc_access_enable(struct amdgpu_device *adev, bool enable)
55 {
56 if (enable)
57 WREG32_SOC15(NBIO, 0, mmBIF_FB_EN,
58 BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
59 else
60 WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, 0);
61 }
62
nbio_v7_0_get_memsize(struct amdgpu_device * adev)63 static u32 nbio_v7_0_get_memsize(struct amdgpu_device *adev)
64 {
65 return RREG32_SOC15(NBIO, 0, mmRCC_CONFIG_MEMSIZE);
66 }
67
nbio_v7_0_sdma_doorbell_range(struct amdgpu_device * adev,int instance,bool use_doorbell,int doorbell_index,int doorbell_size)68 static void nbio_v7_0_sdma_doorbell_range(struct amdgpu_device *adev, int instance,
69 bool use_doorbell, int doorbell_index, int doorbell_size)
70 {
71 u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) :
72 SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE);
73
74 u32 doorbell_range = RREG32(reg);
75
76 if (use_doorbell) {
77 doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index);
78 doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, doorbell_size);
79 } else
80 doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0);
81
82 WREG32(reg, doorbell_range);
83 }
84
nbio_v7_0_vcn_doorbell_range(struct amdgpu_device * adev,bool use_doorbell,int doorbell_index,int instance)85 static void nbio_v7_0_vcn_doorbell_range(struct amdgpu_device *adev, bool use_doorbell,
86 int doorbell_index, int instance)
87 {
88 u32 reg = SOC15_REG_OFFSET(NBIO, 0, mmBIF_MMSCH0_DOORBELL_RANGE);
89
90 u32 doorbell_range = RREG32(reg);
91
92 if (use_doorbell) {
93 doorbell_range = REG_SET_FIELD(doorbell_range,
94 BIF_MMSCH0_DOORBELL_RANGE, OFFSET,
95 doorbell_index);
96 doorbell_range = REG_SET_FIELD(doorbell_range,
97 BIF_MMSCH0_DOORBELL_RANGE, SIZE, 8);
98 } else
99 doorbell_range = REG_SET_FIELD(doorbell_range,
100 BIF_MMSCH0_DOORBELL_RANGE, SIZE, 0);
101
102 WREG32(reg, doorbell_range);
103 }
104
nbio_v7_0_enable_doorbell_aperture(struct amdgpu_device * adev,bool enable)105 static void nbio_v7_0_enable_doorbell_aperture(struct amdgpu_device *adev,
106 bool enable)
107 {
108 WREG32_FIELD15(NBIO, 0, RCC_DOORBELL_APER_EN, BIF_DOORBELL_APER_EN, enable ? 1 : 0);
109 }
110
nbio_v7_0_enable_doorbell_selfring_aperture(struct amdgpu_device * adev,bool enable)111 static void nbio_v7_0_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
112 bool enable)
113 {
114
115 }
116
nbio_v7_0_ih_doorbell_range(struct amdgpu_device * adev,bool use_doorbell,int doorbell_index)117 static void nbio_v7_0_ih_doorbell_range(struct amdgpu_device *adev,
118 bool use_doorbell, int doorbell_index)
119 {
120 u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE);
121
122 if (use_doorbell) {
123 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, OFFSET, doorbell_index);
124 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 2);
125 } else
126 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 0);
127
128 WREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE, ih_doorbell_range);
129 }
130
nbio_7_0_read_syshub_ind_mmr(struct amdgpu_device * adev,uint32_t offset)131 static uint32_t nbio_7_0_read_syshub_ind_mmr(struct amdgpu_device *adev, uint32_t offset)
132 {
133 uint32_t data;
134
135 WREG32_SOC15(NBIO, 0, mmSYSHUB_INDEX, offset);
136 data = RREG32_SOC15(NBIO, 0, mmSYSHUB_DATA);
137
138 return data;
139 }
140
nbio_7_0_write_syshub_ind_mmr(struct amdgpu_device * adev,uint32_t offset,uint32_t data)141 static void nbio_7_0_write_syshub_ind_mmr(struct amdgpu_device *adev, uint32_t offset,
142 uint32_t data)
143 {
144 WREG32_SOC15(NBIO, 0, mmSYSHUB_INDEX, offset);
145 WREG32_SOC15(NBIO, 0, mmSYSHUB_DATA, data);
146 }
147
nbio_v7_0_update_medium_grain_clock_gating(struct amdgpu_device * adev,bool enable)148 static void nbio_v7_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
149 bool enable)
150 {
151 uint32_t def, data;
152
153 /* NBIF_MGCG_CTRL_LCLK */
154 def = data = RREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK);
155
156 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG))
157 data |= NBIF_MGCG_CTRL_LCLK__NBIF_MGCG_EN_LCLK_MASK;
158 else
159 data &= ~NBIF_MGCG_CTRL_LCLK__NBIF_MGCG_EN_LCLK_MASK;
160
161 if (def != data)
162 WREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK, data);
163
164 /* SYSHUB_MGCG_CTRL_SOCCLK */
165 def = data = nbio_7_0_read_syshub_ind_mmr(adev, ixSYSHUB_MMREG_IND_SYSHUB_MGCG_CTRL_SOCCLK);
166
167 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG))
168 data |= SYSHUB_MMREG_DIRECT_SYSHUB_MGCG_CTRL_SOCCLK__SYSHUB_MGCG_EN_SOCCLK_MASK;
169 else
170 data &= ~SYSHUB_MMREG_DIRECT_SYSHUB_MGCG_CTRL_SOCCLK__SYSHUB_MGCG_EN_SOCCLK_MASK;
171
172 if (def != data)
173 nbio_7_0_write_syshub_ind_mmr(adev, ixSYSHUB_MMREG_IND_SYSHUB_MGCG_CTRL_SOCCLK, data);
174
175 /* SYSHUB_MGCG_CTRL_SHUBCLK */
176 def = data = nbio_7_0_read_syshub_ind_mmr(adev, ixSYSHUB_MMREG_IND_SYSHUB_MGCG_CTRL_SHUBCLK);
177
178 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG))
179 data |= SYSHUB_MMREG_DIRECT_SYSHUB_MGCG_CTRL_SHUBCLK__SYSHUB_MGCG_EN_SHUBCLK_MASK;
180 else
181 data &= ~SYSHUB_MMREG_DIRECT_SYSHUB_MGCG_CTRL_SHUBCLK__SYSHUB_MGCG_EN_SHUBCLK_MASK;
182
183 if (def != data)
184 nbio_7_0_write_syshub_ind_mmr(adev, ixSYSHUB_MMREG_IND_SYSHUB_MGCG_CTRL_SHUBCLK, data);
185 }
186
nbio_v7_0_update_medium_grain_light_sleep(struct amdgpu_device * adev,bool enable)187 static void nbio_v7_0_update_medium_grain_light_sleep(struct amdgpu_device *adev,
188 bool enable)
189 {
190 uint32_t def, data;
191
192 def = data = RREG32_PCIE(smnPCIE_CNTL2);
193 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) {
194 data |= (PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
195 PCIE_CNTL2__MST_MEM_LS_EN_MASK |
196 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
197 } else {
198 data &= ~(PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
199 PCIE_CNTL2__MST_MEM_LS_EN_MASK |
200 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
201 }
202
203 if (def != data)
204 WREG32_PCIE(smnPCIE_CNTL2, data);
205 }
206
nbio_v7_0_get_clockgating_state(struct amdgpu_device * adev,u64 * flags)207 static void nbio_v7_0_get_clockgating_state(struct amdgpu_device *adev,
208 u64 *flags)
209 {
210 int data;
211
212 /* AMD_CG_SUPPORT_BIF_MGCG */
213 data = RREG32_PCIE(smnCPM_CONTROL);
214 if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK)
215 *flags |= AMD_CG_SUPPORT_BIF_MGCG;
216
217 /* AMD_CG_SUPPORT_BIF_LS */
218 data = RREG32_PCIE(smnPCIE_CNTL2);
219 if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK)
220 *flags |= AMD_CG_SUPPORT_BIF_LS;
221 }
222
nbio_v7_0_ih_control(struct amdgpu_device * adev)223 static void nbio_v7_0_ih_control(struct amdgpu_device *adev)
224 {
225 u32 interrupt_cntl;
226
227 /* setup interrupt control */
228 WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL2, adev->dummy_page_addr >> 8);
229 interrupt_cntl = RREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL);
230 /* INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
231 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
232 */
233 interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_DUMMY_RD_OVERRIDE, 0);
234 /* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
235 interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_REQ_NONSNOOP_EN, 0);
236 WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl);
237 }
238
nbio_v7_0_get_hdp_flush_req_offset(struct amdgpu_device * adev)239 static u32 nbio_v7_0_get_hdp_flush_req_offset(struct amdgpu_device *adev)
240 {
241 return SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_REQ);
242 }
243
nbio_v7_0_get_hdp_flush_done_offset(struct amdgpu_device * adev)244 static u32 nbio_v7_0_get_hdp_flush_done_offset(struct amdgpu_device *adev)
245 {
246 return SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_DONE);
247 }
248
nbio_v7_0_get_pcie_index_offset(struct amdgpu_device * adev)249 static u32 nbio_v7_0_get_pcie_index_offset(struct amdgpu_device *adev)
250 {
251 return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2);
252 }
253
nbio_v7_0_get_pcie_data_offset(struct amdgpu_device * adev)254 static u32 nbio_v7_0_get_pcie_data_offset(struct amdgpu_device *adev)
255 {
256 return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2);
257 }
258
259 const struct nbio_hdp_flush_reg nbio_v7_0_hdp_flush_reg = {
260 .ref_and_mask_cp0 = GPU_HDP_FLUSH_DONE__CP0_MASK,
261 .ref_and_mask_cp1 = GPU_HDP_FLUSH_DONE__CP1_MASK,
262 .ref_and_mask_cp2 = GPU_HDP_FLUSH_DONE__CP2_MASK,
263 .ref_and_mask_cp3 = GPU_HDP_FLUSH_DONE__CP3_MASK,
264 .ref_and_mask_cp4 = GPU_HDP_FLUSH_DONE__CP4_MASK,
265 .ref_and_mask_cp5 = GPU_HDP_FLUSH_DONE__CP5_MASK,
266 .ref_and_mask_cp6 = GPU_HDP_FLUSH_DONE__CP6_MASK,
267 .ref_and_mask_cp7 = GPU_HDP_FLUSH_DONE__CP7_MASK,
268 .ref_and_mask_cp8 = GPU_HDP_FLUSH_DONE__CP8_MASK,
269 .ref_and_mask_cp9 = GPU_HDP_FLUSH_DONE__CP9_MASK,
270 .ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__SDMA0_MASK,
271 .ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK,
272 };
273
274 #define regRCC_DEV0_EPF6_STRAP4 0xd304
275 #define regRCC_DEV0_EPF6_STRAP4_BASE_IDX 5
276
nbio_v7_0_init_registers(struct amdgpu_device * adev)277 static void nbio_v7_0_init_registers(struct amdgpu_device *adev)
278 {
279 uint32_t data;
280
281 switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
282 case IP_VERSION(2, 5, 0):
283 data = RREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF6_STRAP4) & ~BIT(23);
284 WREG32_SOC15(NBIO, 0, regRCC_DEV0_EPF6_STRAP4, data);
285 break;
286 }
287 }
288
289 #define MMIO_REG_HOLE_OFFSET (0x80000 - PAGE_SIZE)
290
nbio_v7_0_set_reg_remap(struct amdgpu_device * adev)291 static void nbio_v7_0_set_reg_remap(struct amdgpu_device *adev)
292 {
293 if (!amdgpu_sriov_vf(adev) && (PAGE_SIZE <= 4096)) {
294 adev->rmmio_remap.reg_offset = MMIO_REG_HOLE_OFFSET;
295 adev->rmmio_remap.bus_addr = adev->rmmio_base + MMIO_REG_HOLE_OFFSET;
296 } else {
297 adev->rmmio_remap.reg_offset =
298 SOC15_REG_OFFSET(NBIO, 0, mmHDP_MEM_COHERENCY_FLUSH_CNTL) << 2;
299 adev->rmmio_remap.bus_addr = 0;
300 }
301 }
302
303 const struct amdgpu_nbio_funcs nbio_v7_0_funcs = {
304 .get_hdp_flush_req_offset = nbio_v7_0_get_hdp_flush_req_offset,
305 .get_hdp_flush_done_offset = nbio_v7_0_get_hdp_flush_done_offset,
306 .get_pcie_index_offset = nbio_v7_0_get_pcie_index_offset,
307 .get_pcie_data_offset = nbio_v7_0_get_pcie_data_offset,
308 .get_rev_id = nbio_v7_0_get_rev_id,
309 .mc_access_enable = nbio_v7_0_mc_access_enable,
310 .get_memsize = nbio_v7_0_get_memsize,
311 .sdma_doorbell_range = nbio_v7_0_sdma_doorbell_range,
312 .vcn_doorbell_range = nbio_v7_0_vcn_doorbell_range,
313 .enable_doorbell_aperture = nbio_v7_0_enable_doorbell_aperture,
314 .enable_doorbell_selfring_aperture = nbio_v7_0_enable_doorbell_selfring_aperture,
315 .ih_doorbell_range = nbio_v7_0_ih_doorbell_range,
316 .update_medium_grain_clock_gating = nbio_v7_0_update_medium_grain_clock_gating,
317 .update_medium_grain_light_sleep = nbio_v7_0_update_medium_grain_light_sleep,
318 .get_clockgating_state = nbio_v7_0_get_clockgating_state,
319 .ih_control = nbio_v7_0_ih_control,
320 .init_registers = nbio_v7_0_init_registers,
321 .remap_hdp_registers = nbio_v7_0_remap_hdp_registers,
322 .set_reg_remap = nbio_v7_0_set_reg_remap,
323 };
324