1 /*
2 * Copyright 2019 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_v2_3.h"
26
27 #include "nbio/nbio_2_3_default.h"
28 #include "nbio/nbio_2_3_offset.h"
29 #include "nbio/nbio_2_3_sh_mask.h"
30 #include <uapi/linux/kfd_ioctl.h>
31
32 #define smnPCIE_CONFIG_CNTL 0x11180044
33 #define smnCPM_CONTROL 0x11180460
34 #define smnPCIE_CNTL2 0x11180070
35
36 #define mmBIF_SDMA2_DOORBELL_RANGE 0x01d6
37 #define mmBIF_SDMA2_DOORBELL_RANGE_BASE_IDX 2
38 #define mmBIF_SDMA3_DOORBELL_RANGE 0x01d7
39 #define mmBIF_SDMA3_DOORBELL_RANGE_BASE_IDX 2
40
41 #define mmBIF_MMSCH1_DOORBELL_RANGE 0x01d8
42 #define mmBIF_MMSCH1_DOORBELL_RANGE_BASE_IDX 2
43
nbio_v2_3_remap_hdp_registers(struct amdgpu_device * adev)44 static void nbio_v2_3_remap_hdp_registers(struct amdgpu_device *adev)
45 {
46 WREG32_SOC15(NBIO, 0, mmREMAP_HDP_MEM_FLUSH_CNTL,
47 adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL);
48 WREG32_SOC15(NBIO, 0, mmREMAP_HDP_REG_FLUSH_CNTL,
49 adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_REG_FLUSH_CNTL);
50 }
51
nbio_v2_3_get_rev_id(struct amdgpu_device * adev)52 static u32 nbio_v2_3_get_rev_id(struct amdgpu_device *adev)
53 {
54 u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0);
55
56 tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
57 tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
58
59 return tmp;
60 }
61
nbio_v2_3_mc_access_enable(struct amdgpu_device * adev,bool enable)62 static void nbio_v2_3_mc_access_enable(struct amdgpu_device *adev, bool enable)
63 {
64 if (enable)
65 WREG32_SOC15(NBIO, 0, mmBIF_FB_EN,
66 BIF_FB_EN__FB_READ_EN_MASK |
67 BIF_FB_EN__FB_WRITE_EN_MASK);
68 else
69 WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, 0);
70 }
71
nbio_v2_3_hdp_flush(struct amdgpu_device * adev,struct amdgpu_ring * ring)72 static void nbio_v2_3_hdp_flush(struct amdgpu_device *adev,
73 struct amdgpu_ring *ring)
74 {
75 if (!ring || !ring->funcs->emit_wreg)
76 WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
77 else
78 amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
79 }
80
nbio_v2_3_get_memsize(struct amdgpu_device * adev)81 static u32 nbio_v2_3_get_memsize(struct amdgpu_device *adev)
82 {
83 return RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_RCC_CONFIG_MEMSIZE);
84 }
85
nbio_v2_3_sdma_doorbell_range(struct amdgpu_device * adev,int instance,bool use_doorbell,int doorbell_index,int doorbell_size)86 static void nbio_v2_3_sdma_doorbell_range(struct amdgpu_device *adev, int instance,
87 bool use_doorbell, int doorbell_index,
88 int doorbell_size)
89 {
90 u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) :
91 instance == 1 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE) :
92 instance == 2 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA2_DOORBELL_RANGE) :
93 SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA3_DOORBELL_RANGE);
94
95 u32 doorbell_range = RREG32(reg);
96
97 if (use_doorbell) {
98 doorbell_range = REG_SET_FIELD(doorbell_range,
99 BIF_SDMA0_DOORBELL_RANGE, OFFSET,
100 doorbell_index);
101 doorbell_range = REG_SET_FIELD(doorbell_range,
102 BIF_SDMA0_DOORBELL_RANGE, SIZE,
103 doorbell_size);
104 } else
105 doorbell_range = REG_SET_FIELD(doorbell_range,
106 BIF_SDMA0_DOORBELL_RANGE, SIZE,
107 0);
108
109 WREG32(reg, doorbell_range);
110 }
111
nbio_v2_3_vcn_doorbell_range(struct amdgpu_device * adev,bool use_doorbell,int doorbell_index,int instance)112 static void nbio_v2_3_vcn_doorbell_range(struct amdgpu_device *adev, bool use_doorbell,
113 int doorbell_index, int instance)
114 {
115 u32 reg = instance ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_MMSCH1_DOORBELL_RANGE) :
116 SOC15_REG_OFFSET(NBIO, 0, mmBIF_MMSCH0_DOORBELL_RANGE);
117
118 u32 doorbell_range = RREG32(reg);
119
120 if (use_doorbell) {
121 doorbell_range = REG_SET_FIELD(doorbell_range,
122 BIF_MMSCH0_DOORBELL_RANGE, OFFSET,
123 doorbell_index);
124 doorbell_range = REG_SET_FIELD(doorbell_range,
125 BIF_MMSCH0_DOORBELL_RANGE, SIZE, 8);
126 } else
127 doorbell_range = REG_SET_FIELD(doorbell_range,
128 BIF_MMSCH0_DOORBELL_RANGE, SIZE, 0);
129
130 WREG32(reg, doorbell_range);
131 }
132
nbio_v2_3_enable_doorbell_aperture(struct amdgpu_device * adev,bool enable)133 static void nbio_v2_3_enable_doorbell_aperture(struct amdgpu_device *adev,
134 bool enable)
135 {
136 WREG32_FIELD15(NBIO, 0, RCC_DEV0_EPF0_RCC_DOORBELL_APER_EN, BIF_DOORBELL_APER_EN,
137 enable ? 1 : 0);
138 }
139
nbio_v2_3_enable_doorbell_selfring_aperture(struct amdgpu_device * adev,bool enable)140 static void nbio_v2_3_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
141 bool enable)
142 {
143 u32 tmp = 0;
144
145 if (enable) {
146 tmp = REG_SET_FIELD(tmp, BIF_BX_PF_DOORBELL_SELFRING_GPA_APER_CNTL,
147 DOORBELL_SELFRING_GPA_APER_EN, 1) |
148 REG_SET_FIELD(tmp, BIF_BX_PF_DOORBELL_SELFRING_GPA_APER_CNTL,
149 DOORBELL_SELFRING_GPA_APER_MODE, 1) |
150 REG_SET_FIELD(tmp, BIF_BX_PF_DOORBELL_SELFRING_GPA_APER_CNTL,
151 DOORBELL_SELFRING_GPA_APER_SIZE, 0);
152
153 WREG32_SOC15(NBIO, 0, mmBIF_BX_PF_DOORBELL_SELFRING_GPA_APER_BASE_LOW,
154 lower_32_bits(adev->doorbell.base));
155 WREG32_SOC15(NBIO, 0, mmBIF_BX_PF_DOORBELL_SELFRING_GPA_APER_BASE_HIGH,
156 upper_32_bits(adev->doorbell.base));
157 }
158
159 WREG32_SOC15(NBIO, 0, mmBIF_BX_PF_DOORBELL_SELFRING_GPA_APER_CNTL,
160 tmp);
161 }
162
163
nbio_v2_3_ih_doorbell_range(struct amdgpu_device * adev,bool use_doorbell,int doorbell_index)164 static void nbio_v2_3_ih_doorbell_range(struct amdgpu_device *adev,
165 bool use_doorbell, int doorbell_index)
166 {
167 u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE);
168
169 if (use_doorbell) {
170 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
171 BIF_IH_DOORBELL_RANGE, OFFSET,
172 doorbell_index);
173 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
174 BIF_IH_DOORBELL_RANGE, SIZE,
175 2);
176 } else
177 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
178 BIF_IH_DOORBELL_RANGE, SIZE,
179 0);
180
181 WREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE, ih_doorbell_range);
182 }
183
nbio_v2_3_ih_control(struct amdgpu_device * adev)184 static void nbio_v2_3_ih_control(struct amdgpu_device *adev)
185 {
186 u32 interrupt_cntl;
187
188 /* setup interrupt control */
189 WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL2, adev->dummy_page_addr >> 8);
190
191 interrupt_cntl = RREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL);
192 /*
193 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
194 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
195 */
196 interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL,
197 IH_DUMMY_RD_OVERRIDE, 0);
198
199 /* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
200 interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL,
201 IH_REQ_NONSNOOP_EN, 0);
202
203 WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl);
204 }
205
nbio_v2_3_update_medium_grain_clock_gating(struct amdgpu_device * adev,bool enable)206 static void nbio_v2_3_update_medium_grain_clock_gating(struct amdgpu_device *adev,
207 bool enable)
208 {
209 uint32_t def, data;
210
211 def = data = RREG32_PCIE(smnCPM_CONTROL);
212 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG)) {
213 data |= (CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
214 CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
215 CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
216 CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
217 CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
218 CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
219 } else {
220 data &= ~(CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
221 CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
222 CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
223 CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
224 CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
225 CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
226 }
227
228 if (def != data)
229 WREG32_PCIE(smnCPM_CONTROL, data);
230 }
231
nbio_v2_3_update_medium_grain_light_sleep(struct amdgpu_device * adev,bool enable)232 static void nbio_v2_3_update_medium_grain_light_sleep(struct amdgpu_device *adev,
233 bool enable)
234 {
235 uint32_t def, data;
236
237 def = data = RREG32_PCIE(smnPCIE_CNTL2);
238 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) {
239 data |= (PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
240 PCIE_CNTL2__MST_MEM_LS_EN_MASK |
241 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
242 } else {
243 data &= ~(PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
244 PCIE_CNTL2__MST_MEM_LS_EN_MASK |
245 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
246 }
247
248 if (def != data)
249 WREG32_PCIE(smnPCIE_CNTL2, data);
250 }
251
nbio_v2_3_get_clockgating_state(struct amdgpu_device * adev,u32 * flags)252 static void nbio_v2_3_get_clockgating_state(struct amdgpu_device *adev,
253 u32 *flags)
254 {
255 int data;
256
257 /* AMD_CG_SUPPORT_BIF_MGCG */
258 data = RREG32_PCIE(smnCPM_CONTROL);
259 if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK)
260 *flags |= AMD_CG_SUPPORT_BIF_MGCG;
261
262 /* AMD_CG_SUPPORT_BIF_LS */
263 data = RREG32_PCIE(smnPCIE_CNTL2);
264 if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK)
265 *flags |= AMD_CG_SUPPORT_BIF_LS;
266 }
267
nbio_v2_3_get_hdp_flush_req_offset(struct amdgpu_device * adev)268 static u32 nbio_v2_3_get_hdp_flush_req_offset(struct amdgpu_device *adev)
269 {
270 return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF_GPU_HDP_FLUSH_REQ);
271 }
272
nbio_v2_3_get_hdp_flush_done_offset(struct amdgpu_device * adev)273 static u32 nbio_v2_3_get_hdp_flush_done_offset(struct amdgpu_device *adev)
274 {
275 return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF_GPU_HDP_FLUSH_DONE);
276 }
277
nbio_v2_3_get_pcie_index_offset(struct amdgpu_device * adev)278 static u32 nbio_v2_3_get_pcie_index_offset(struct amdgpu_device *adev)
279 {
280 return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2);
281 }
282
nbio_v2_3_get_pcie_data_offset(struct amdgpu_device * adev)283 static u32 nbio_v2_3_get_pcie_data_offset(struct amdgpu_device *adev)
284 {
285 return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2);
286 }
287
288 const struct nbio_hdp_flush_reg nbio_v2_3_hdp_flush_reg = {
289 .ref_and_mask_cp0 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP0_MASK,
290 .ref_and_mask_cp1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP1_MASK,
291 .ref_and_mask_cp2 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP2_MASK,
292 .ref_and_mask_cp3 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP3_MASK,
293 .ref_and_mask_cp4 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP4_MASK,
294 .ref_and_mask_cp5 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP5_MASK,
295 .ref_and_mask_cp6 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP6_MASK,
296 .ref_and_mask_cp7 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP7_MASK,
297 .ref_and_mask_cp8 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP8_MASK,
298 .ref_and_mask_cp9 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__CP9_MASK,
299 .ref_and_mask_sdma0 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__SDMA0_MASK,
300 .ref_and_mask_sdma1 = BIF_BX_PF_GPU_HDP_FLUSH_DONE__SDMA1_MASK,
301 };
302
nbio_v2_3_init_registers(struct amdgpu_device * adev)303 static void nbio_v2_3_init_registers(struct amdgpu_device *adev)
304 {
305 uint32_t def, data;
306
307 def = data = RREG32_PCIE(smnPCIE_CONFIG_CNTL);
308 data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1);
309 data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_PRIV, 1);
310
311 if (def != data)
312 WREG32_PCIE(smnPCIE_CONFIG_CNTL, data);
313 }
314
315 const struct amdgpu_nbio_funcs nbio_v2_3_funcs = {
316 .get_hdp_flush_req_offset = nbio_v2_3_get_hdp_flush_req_offset,
317 .get_hdp_flush_done_offset = nbio_v2_3_get_hdp_flush_done_offset,
318 .get_pcie_index_offset = nbio_v2_3_get_pcie_index_offset,
319 .get_pcie_data_offset = nbio_v2_3_get_pcie_data_offset,
320 .get_rev_id = nbio_v2_3_get_rev_id,
321 .mc_access_enable = nbio_v2_3_mc_access_enable,
322 .hdp_flush = nbio_v2_3_hdp_flush,
323 .get_memsize = nbio_v2_3_get_memsize,
324 .sdma_doorbell_range = nbio_v2_3_sdma_doorbell_range,
325 .vcn_doorbell_range = nbio_v2_3_vcn_doorbell_range,
326 .enable_doorbell_aperture = nbio_v2_3_enable_doorbell_aperture,
327 .enable_doorbell_selfring_aperture = nbio_v2_3_enable_doorbell_selfring_aperture,
328 .ih_doorbell_range = nbio_v2_3_ih_doorbell_range,
329 .update_medium_grain_clock_gating = nbio_v2_3_update_medium_grain_clock_gating,
330 .update_medium_grain_light_sleep = nbio_v2_3_update_medium_grain_light_sleep,
331 .get_clockgating_state = nbio_v2_3_get_clockgating_state,
332 .ih_control = nbio_v2_3_ih_control,
333 .init_registers = nbio_v2_3_init_registers,
334 .remap_hdp_registers = nbio_v2_3_remap_hdp_registers,
335 };
336