xref: /linux/drivers/gpu/drm/amd/amdgpu/nbio_v6_1.c (revision b8d312aa075f33282565467662c4628dae0a2aff)
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_v6_1.h"
26 
27 #include "nbio/nbio_6_1_default.h"
28 #include "nbio/nbio_6_1_offset.h"
29 #include "nbio/nbio_6_1_sh_mask.h"
30 #include "nbio/nbio_6_1_smn.h"
31 #include "vega10_enum.h"
32 
33 static u32 nbio_v6_1_get_rev_id(struct amdgpu_device *adev)
34 {
35         u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0);
36 
37 	tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
38 	tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
39 
40 	return tmp;
41 }
42 
43 static void nbio_v6_1_mc_access_enable(struct amdgpu_device *adev, bool enable)
44 {
45 	if (enable)
46 		WREG32_SOC15(NBIO, 0, mmBIF_FB_EN,
47 			     BIF_FB_EN__FB_READ_EN_MASK |
48 			     BIF_FB_EN__FB_WRITE_EN_MASK);
49 	else
50 		WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, 0);
51 }
52 
53 static void nbio_v6_1_hdp_flush(struct amdgpu_device *adev,
54 				struct amdgpu_ring *ring)
55 {
56 	if (!ring || !ring->funcs->emit_wreg)
57 		WREG32_SOC15_NO_KIQ(NBIO, 0,
58 				    mmBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL,
59 				    0);
60 	else
61 		amdgpu_ring_emit_wreg(ring, SOC15_REG_OFFSET(
62 			NBIO, 0, mmBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL), 0);
63 }
64 
65 static u32 nbio_v6_1_get_memsize(struct amdgpu_device *adev)
66 {
67 	return RREG32_SOC15(NBIO, 0, mmRCC_PF_0_0_RCC_CONFIG_MEMSIZE);
68 }
69 
70 static void nbio_v6_1_sdma_doorbell_range(struct amdgpu_device *adev, int instance,
71 			bool use_doorbell, int doorbell_index, int doorbell_size)
72 {
73 	u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) :
74 			SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE);
75 
76 	u32 doorbell_range = RREG32(reg);
77 
78 	if (use_doorbell) {
79 		doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index);
80 		doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, doorbell_size);
81 	} else
82 		doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0);
83 
84 	WREG32(reg, doorbell_range);
85 
86 }
87 
88 static void nbio_v6_1_enable_doorbell_aperture(struct amdgpu_device *adev,
89 					       bool enable)
90 {
91 	WREG32_FIELD15(NBIO, 0, RCC_PF_0_0_RCC_DOORBELL_APER_EN, BIF_DOORBELL_APER_EN, enable ? 1 : 0);
92 }
93 
94 static void nbio_v6_1_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
95 							bool enable)
96 {
97 	u32 tmp = 0;
98 
99 	if (enable) {
100 		tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_EN, 1) |
101 		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_MODE, 1) |
102 		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_SIZE, 0);
103 
104 		WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW,
105 			     lower_32_bits(adev->doorbell.base));
106 		WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH,
107 			     upper_32_bits(adev->doorbell.base));
108 	}
109 
110 	WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, tmp);
111 }
112 
113 
114 static void nbio_v6_1_ih_doorbell_range(struct amdgpu_device *adev,
115 					bool use_doorbell, int doorbell_index)
116 {
117 	u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0 , mmBIF_IH_DOORBELL_RANGE);
118 
119 	if (use_doorbell) {
120 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, OFFSET, doorbell_index);
121 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range,
122 						  BIF_IH_DOORBELL_RANGE, SIZE, 6);
123 	} else
124 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 0);
125 
126 	WREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE, ih_doorbell_range);
127 }
128 
129 static void nbio_v6_1_ih_control(struct amdgpu_device *adev)
130 {
131 	u32 interrupt_cntl;
132 
133 	/* setup interrupt control */
134 	WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL2, adev->dummy_page_addr >> 8);
135 	interrupt_cntl = RREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL);
136 	/* INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
137 	 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
138 	 */
139 	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_DUMMY_RD_OVERRIDE, 0);
140 	/* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
141 	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_REQ_NONSNOOP_EN, 0);
142 	WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl);
143 }
144 
145 static void nbio_v6_1_update_medium_grain_clock_gating(struct amdgpu_device *adev,
146 						       bool enable)
147 {
148 	uint32_t def, data;
149 
150 	def = data = RREG32_PCIE(smnCPM_CONTROL);
151 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG)) {
152 		data |= (CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
153 			 CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
154 			 CPM_CONTROL__TXCLK_PERM_GATE_ENABLE_MASK |
155 			 CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
156 			 CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
157 			 CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
158 			 CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
159 	} else {
160 		data &= ~(CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
161 			  CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
162 			  CPM_CONTROL__TXCLK_PERM_GATE_ENABLE_MASK |
163 			  CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
164 			  CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
165 			  CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
166 			  CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
167 	}
168 
169 	if (def != data)
170 		WREG32_PCIE(smnCPM_CONTROL, data);
171 }
172 
173 static void nbio_v6_1_update_medium_grain_light_sleep(struct amdgpu_device *adev,
174 						      bool enable)
175 {
176 	uint32_t def, data;
177 
178 	def = data = RREG32_PCIE(smnPCIE_CNTL2);
179 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) {
180 		data |= (PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
181 			 PCIE_CNTL2__MST_MEM_LS_EN_MASK |
182 			 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
183 	} else {
184 		data &= ~(PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
185 			  PCIE_CNTL2__MST_MEM_LS_EN_MASK |
186 			  PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
187 	}
188 
189 	if (def != data)
190 		WREG32_PCIE(smnPCIE_CNTL2, data);
191 }
192 
193 static void nbio_v6_1_get_clockgating_state(struct amdgpu_device *adev,
194 					    u32 *flags)
195 {
196 	int data;
197 
198 	/* AMD_CG_SUPPORT_BIF_MGCG */
199 	data = RREG32_PCIE(smnCPM_CONTROL);
200 	if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK)
201 		*flags |= AMD_CG_SUPPORT_BIF_MGCG;
202 
203 	/* AMD_CG_SUPPORT_BIF_LS */
204 	data = RREG32_PCIE(smnPCIE_CNTL2);
205 	if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK)
206 		*flags |= AMD_CG_SUPPORT_BIF_LS;
207 }
208 
209 static u32 nbio_v6_1_get_hdp_flush_req_offset(struct amdgpu_device *adev)
210 {
211 	return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_REQ);
212 }
213 
214 static u32 nbio_v6_1_get_hdp_flush_done_offset(struct amdgpu_device *adev)
215 {
216 	return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_DONE);
217 }
218 
219 static u32 nbio_v6_1_get_pcie_index_offset(struct amdgpu_device *adev)
220 {
221 	return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2);
222 }
223 
224 static u32 nbio_v6_1_get_pcie_data_offset(struct amdgpu_device *adev)
225 {
226 	return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2);
227 }
228 
229 const struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg = {
230 	.ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK,
231 	.ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK,
232 	.ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK,
233 	.ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK,
234 	.ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK,
235 	.ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK,
236 	.ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK,
237 	.ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK,
238 	.ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK,
239 	.ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK,
240 	.ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK,
241 	.ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK
242 };
243 
244 static void nbio_v6_1_detect_hw_virt(struct amdgpu_device *adev)
245 {
246 	uint32_t reg;
247 
248 	reg = RREG32_SOC15(NBIO, 0, mmRCC_PF_0_0_RCC_IOV_FUNC_IDENTIFIER);
249 	if (reg & 1)
250 		adev->virt.caps |= AMDGPU_SRIOV_CAPS_IS_VF;
251 
252 	if (reg & 0x80000000)
253 		adev->virt.caps |= AMDGPU_SRIOV_CAPS_ENABLE_IOV;
254 
255 	if (!reg) {
256 		if (is_virtual_machine())	/* passthrough mode exclus sriov mod */
257 			adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE;
258 	}
259 }
260 
261 static void nbio_v6_1_init_registers(struct amdgpu_device *adev)
262 {
263 	uint32_t def, data;
264 
265 	def = data = RREG32_PCIE(smnPCIE_CONFIG_CNTL);
266 	data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1);
267 	data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_PRIV, 1);
268 
269 	if (def != data)
270 		WREG32_PCIE(smnPCIE_CONFIG_CNTL, data);
271 
272 	def = data = RREG32_PCIE(smnPCIE_CI_CNTL);
273 	data = REG_SET_FIELD(data, PCIE_CI_CNTL, CI_SLV_ORDERING_DIS, 1);
274 
275 	if (def != data)
276 		WREG32_PCIE(smnPCIE_CI_CNTL, data);
277 }
278 
279 const struct amdgpu_nbio_funcs nbio_v6_1_funcs = {
280 	.get_hdp_flush_req_offset = nbio_v6_1_get_hdp_flush_req_offset,
281 	.get_hdp_flush_done_offset = nbio_v6_1_get_hdp_flush_done_offset,
282 	.get_pcie_index_offset = nbio_v6_1_get_pcie_index_offset,
283 	.get_pcie_data_offset = nbio_v6_1_get_pcie_data_offset,
284 	.get_rev_id = nbio_v6_1_get_rev_id,
285 	.mc_access_enable = nbio_v6_1_mc_access_enable,
286 	.hdp_flush = nbio_v6_1_hdp_flush,
287 	.get_memsize = nbio_v6_1_get_memsize,
288 	.sdma_doorbell_range = nbio_v6_1_sdma_doorbell_range,
289 	.enable_doorbell_aperture = nbio_v6_1_enable_doorbell_aperture,
290 	.enable_doorbell_selfring_aperture = nbio_v6_1_enable_doorbell_selfring_aperture,
291 	.ih_doorbell_range = nbio_v6_1_ih_doorbell_range,
292 	.update_medium_grain_clock_gating = nbio_v6_1_update_medium_grain_clock_gating,
293 	.update_medium_grain_light_sleep = nbio_v6_1_update_medium_grain_light_sleep,
294 	.get_clockgating_state = nbio_v6_1_get_clockgating_state,
295 	.ih_control = nbio_v6_1_ih_control,
296 	.init_registers = nbio_v6_1_init_registers,
297 	.detect_hw_virt = nbio_v6_1_detect_hw_virt,
298 };
299