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 "vega10_enum.h"
31 
32 #define smnCPM_CONTROL                                                                                  0x11180460
33 #define smnPCIE_CNTL2                                                                                   0x11180070
34 #define smnPCIE_CONFIG_CNTL                                                                             0x11180044
35 
36 static u32 nbio_v6_1_get_rev_id(struct amdgpu_device *adev)
37 {
38         u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0);
39 
40 	tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK;
41 	tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT;
42 
43 	return tmp;
44 }
45 
46 static void nbio_v6_1_mc_access_enable(struct amdgpu_device *adev, bool enable)
47 {
48 	if (enable)
49 		WREG32_SOC15(NBIO, 0, mmBIF_FB_EN,
50 			     BIF_FB_EN__FB_READ_EN_MASK |
51 			     BIF_FB_EN__FB_WRITE_EN_MASK);
52 	else
53 		WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, 0);
54 }
55 
56 static void nbio_v6_1_hdp_flush(struct amdgpu_device *adev)
57 {
58 	WREG32_SOC15_NO_KIQ(NBIO, 0, mmBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL, 0);
59 }
60 
61 static u32 nbio_v6_1_get_memsize(struct amdgpu_device *adev)
62 {
63 	return RREG32_SOC15(NBIO, 0, mmRCC_PF_0_0_RCC_CONFIG_MEMSIZE);
64 }
65 
66 static void nbio_v6_1_sdma_doorbell_range(struct amdgpu_device *adev, int instance,
67 				  bool use_doorbell, int doorbell_index)
68 {
69 	u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) :
70 			SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE);
71 
72 	u32 doorbell_range = RREG32(reg);
73 
74 	if (use_doorbell) {
75 		doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index);
76 		doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 2);
77 	} else
78 		doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0);
79 
80 	WREG32(reg, doorbell_range);
81 
82 }
83 
84 static void nbio_v6_1_enable_doorbell_aperture(struct amdgpu_device *adev,
85 					       bool enable)
86 {
87 	WREG32_FIELD15(NBIO, 0, RCC_PF_0_0_RCC_DOORBELL_APER_EN, BIF_DOORBELL_APER_EN, enable ? 1 : 0);
88 }
89 
90 static void nbio_v6_1_enable_doorbell_selfring_aperture(struct amdgpu_device *adev,
91 							bool enable)
92 {
93 	u32 tmp = 0;
94 
95 	if (enable) {
96 		tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_EN, 1) |
97 		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_MODE, 1) |
98 		      REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_SIZE, 0);
99 
100 		WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW,
101 			     lower_32_bits(adev->doorbell.base));
102 		WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH,
103 			     upper_32_bits(adev->doorbell.base));
104 	}
105 
106 	WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, tmp);
107 }
108 
109 
110 static void nbio_v6_1_ih_doorbell_range(struct amdgpu_device *adev,
111 					bool use_doorbell, int doorbell_index)
112 {
113 	u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0 , mmBIF_IH_DOORBELL_RANGE);
114 
115 	if (use_doorbell) {
116 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, OFFSET, doorbell_index);
117 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 2);
118 	} else
119 		ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 0);
120 
121 	WREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE, ih_doorbell_range);
122 }
123 
124 static void nbio_v6_1_ih_control(struct amdgpu_device *adev)
125 {
126 	u32 interrupt_cntl;
127 
128 	/* setup interrupt control */
129 	WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL2, adev->dummy_page.addr >> 8);
130 	interrupt_cntl = RREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL);
131 	/* INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
132 	 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
133 	 */
134 	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_DUMMY_RD_OVERRIDE, 0);
135 	/* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
136 	interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_REQ_NONSNOOP_EN, 0);
137 	WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl);
138 }
139 
140 static void nbio_v6_1_update_medium_grain_clock_gating(struct amdgpu_device *adev,
141 						       bool enable)
142 {
143 	uint32_t def, data;
144 
145 	def = data = RREG32_PCIE(smnCPM_CONTROL);
146 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG)) {
147 		data |= (CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
148 			 CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
149 			 CPM_CONTROL__TXCLK_PERM_GATE_ENABLE_MASK |
150 			 CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
151 			 CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
152 			 CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
153 			 CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
154 	} else {
155 		data &= ~(CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK |
156 			  CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK |
157 			  CPM_CONTROL__TXCLK_PERM_GATE_ENABLE_MASK |
158 			  CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK |
159 			  CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK |
160 			  CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK |
161 			  CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK);
162 	}
163 
164 	if (def != data)
165 		WREG32_PCIE(smnCPM_CONTROL, data);
166 }
167 
168 static void nbio_v6_1_update_medium_grain_light_sleep(struct amdgpu_device *adev,
169 						      bool enable)
170 {
171 	uint32_t def, data;
172 
173 	def = data = RREG32_PCIE(smnPCIE_CNTL2);
174 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) {
175 		data |= (PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
176 			 PCIE_CNTL2__MST_MEM_LS_EN_MASK |
177 			 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
178 	} else {
179 		data &= ~(PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
180 			  PCIE_CNTL2__MST_MEM_LS_EN_MASK |
181 			  PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK);
182 	}
183 
184 	if (def != data)
185 		WREG32_PCIE(smnPCIE_CNTL2, data);
186 }
187 
188 static void nbio_v6_1_get_clockgating_state(struct amdgpu_device *adev,
189 					    u32 *flags)
190 {
191 	int data;
192 
193 	/* AMD_CG_SUPPORT_BIF_MGCG */
194 	data = RREG32_PCIE(smnCPM_CONTROL);
195 	if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK)
196 		*flags |= AMD_CG_SUPPORT_BIF_MGCG;
197 
198 	/* AMD_CG_SUPPORT_BIF_LS */
199 	data = RREG32_PCIE(smnPCIE_CNTL2);
200 	if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK)
201 		*flags |= AMD_CG_SUPPORT_BIF_LS;
202 }
203 
204 static u32 nbio_v6_1_get_hdp_flush_req_offset(struct amdgpu_device *adev)
205 {
206 	return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_REQ);
207 }
208 
209 static u32 nbio_v6_1_get_hdp_flush_done_offset(struct amdgpu_device *adev)
210 {
211 	return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_DONE);
212 }
213 
214 static u32 nbio_v6_1_get_pcie_index_offset(struct amdgpu_device *adev)
215 {
216 	return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX);
217 }
218 
219 static u32 nbio_v6_1_get_pcie_data_offset(struct amdgpu_device *adev)
220 {
221 	return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA);
222 }
223 
224 static const struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg = {
225 	.ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK,
226 	.ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK,
227 	.ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK,
228 	.ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK,
229 	.ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK,
230 	.ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK,
231 	.ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK,
232 	.ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK,
233 	.ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK,
234 	.ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK,
235 	.ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK,
236 	.ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK
237 };
238 
239 static void nbio_v6_1_detect_hw_virt(struct amdgpu_device *adev)
240 {
241 	uint32_t reg;
242 
243 	reg = RREG32_SOC15(NBIO, 0, mmRCC_PF_0_0_RCC_IOV_FUNC_IDENTIFIER);
244 	if (reg & 1)
245 		adev->virt.caps |= AMDGPU_SRIOV_CAPS_IS_VF;
246 
247 	if (reg & 0x80000000)
248 		adev->virt.caps |= AMDGPU_SRIOV_CAPS_ENABLE_IOV;
249 
250 	if (!reg) {
251 		if (is_virtual_machine())	/* passthrough mode exclus sriov mod */
252 			adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE;
253 	}
254 }
255 
256 static void nbio_v6_1_init_registers(struct amdgpu_device *adev)
257 {
258 	uint32_t def, data;
259 
260 	def = data = RREG32_PCIE(smnPCIE_CONFIG_CNTL);
261 	data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1);
262 	data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_PRIV, 1);
263 
264 	if (def != data)
265 		WREG32_PCIE(smnPCIE_CONFIG_CNTL, data);
266 }
267 
268 const struct amdgpu_nbio_funcs nbio_v6_1_funcs = {
269 	.hdp_flush_reg = &nbio_v6_1_hdp_flush_reg,
270 	.get_hdp_flush_req_offset = nbio_v6_1_get_hdp_flush_req_offset,
271 	.get_hdp_flush_done_offset = nbio_v6_1_get_hdp_flush_done_offset,
272 	.get_pcie_index_offset = nbio_v6_1_get_pcie_index_offset,
273 	.get_pcie_data_offset = nbio_v6_1_get_pcie_data_offset,
274 	.get_rev_id = nbio_v6_1_get_rev_id,
275 	.mc_access_enable = nbio_v6_1_mc_access_enable,
276 	.hdp_flush = nbio_v6_1_hdp_flush,
277 	.get_memsize = nbio_v6_1_get_memsize,
278 	.sdma_doorbell_range = nbio_v6_1_sdma_doorbell_range,
279 	.enable_doorbell_aperture = nbio_v6_1_enable_doorbell_aperture,
280 	.enable_doorbell_selfring_aperture = nbio_v6_1_enable_doorbell_selfring_aperture,
281 	.ih_doorbell_range = nbio_v6_1_ih_doorbell_range,
282 	.update_medium_grain_clock_gating = nbio_v6_1_update_medium_grain_clock_gating,
283 	.update_medium_grain_light_sleep = nbio_v6_1_update_medium_grain_light_sleep,
284 	.get_clockgating_state = nbio_v6_1_get_clockgating_state,
285 	.ih_control = nbio_v6_1_ih_control,
286 	.init_registers = nbio_v6_1_init_registers,
287 	.detect_hw_virt = nbio_v6_1_detect_hw_virt,
288 };
289