Rev 1430 | Rev 1986 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
1128 | serge | 1 | /* |
2 | * Copyright 2008 Advanced Micro Devices, Inc. |
||
3 | * Copyright 2008 Red Hat Inc. |
||
4 | * Copyright 2009 Jerome Glisse. |
||
5 | * |
||
6 | * Permission is hereby granted, free of charge, to any person obtaining a |
||
7 | * copy of this software and associated documentation files (the "Software"), |
||
8 | * to deal in the Software without restriction, including without limitation |
||
9 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
||
10 | * and/or sell copies of the Software, and to permit persons to whom the |
||
11 | * Software is furnished to do so, subject to the following conditions: |
||
12 | * |
||
13 | * The above copyright notice and this permission notice shall be included in |
||
14 | * all copies or substantial portions of the Software. |
||
15 | * |
||
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
||
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
||
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
||
19 | * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR |
||
20 | * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, |
||
21 | * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
||
22 | * OTHER DEALINGS IN THE SOFTWARE. |
||
23 | * |
||
24 | * Authors: Dave Airlie |
||
25 | * Alex Deucher |
||
26 | * Jerome Glisse |
||
27 | */ |
||
1963 | serge | 28 | #include |
1233 | serge | 29 | #include |
1221 | serge | 30 | #include |
1128 | serge | 31 | #include "drmP.h" |
1221 | serge | 32 | #include "radeon_drm.h" |
1128 | serge | 33 | #include "radeon.h" |
1963 | serge | 34 | #include "radeon_asic.h" |
1221 | serge | 35 | #include "radeon_mode.h" |
36 | #include "r600d.h" |
||
37 | #include "atom.h" |
||
38 | #include "avivod.h" |
||
1128 | serge | 39 | |
1221 | serge | 40 | #define PFP_UCODE_SIZE 576 |
41 | #define PM4_UCODE_SIZE 1792 |
||
1321 | serge | 42 | #define RLC_UCODE_SIZE 768 |
1221 | serge | 43 | #define R700_PFP_UCODE_SIZE 848 |
44 | #define R700_PM4_UCODE_SIZE 1360 |
||
1321 | serge | 45 | #define R700_RLC_UCODE_SIZE 1024 |
1963 | serge | 46 | #define EVERGREEN_PFP_UCODE_SIZE 1120 |
47 | #define EVERGREEN_PM4_UCODE_SIZE 1376 |
||
48 | #define EVERGREEN_RLC_UCODE_SIZE 768 |
||
49 | #define CAYMAN_RLC_UCODE_SIZE 1024 |
||
1128 | serge | 50 | |
1221 | serge | 51 | /* Firmware Names */ |
52 | MODULE_FIRMWARE("radeon/R600_pfp.bin"); |
||
53 | MODULE_FIRMWARE("radeon/R600_me.bin"); |
||
54 | MODULE_FIRMWARE("radeon/RV610_pfp.bin"); |
||
55 | MODULE_FIRMWARE("radeon/RV610_me.bin"); |
||
56 | MODULE_FIRMWARE("radeon/RV630_pfp.bin"); |
||
57 | MODULE_FIRMWARE("radeon/RV630_me.bin"); |
||
58 | MODULE_FIRMWARE("radeon/RV620_pfp.bin"); |
||
59 | MODULE_FIRMWARE("radeon/RV620_me.bin"); |
||
60 | MODULE_FIRMWARE("radeon/RV635_pfp.bin"); |
||
61 | MODULE_FIRMWARE("radeon/RV635_me.bin"); |
||
62 | MODULE_FIRMWARE("radeon/RV670_pfp.bin"); |
||
63 | MODULE_FIRMWARE("radeon/RV670_me.bin"); |
||
64 | MODULE_FIRMWARE("radeon/RS780_pfp.bin"); |
||
65 | MODULE_FIRMWARE("radeon/RS780_me.bin"); |
||
66 | MODULE_FIRMWARE("radeon/RV770_pfp.bin"); |
||
67 | MODULE_FIRMWARE("radeon/RV770_me.bin"); |
||
68 | MODULE_FIRMWARE("radeon/RV730_pfp.bin"); |
||
69 | MODULE_FIRMWARE("radeon/RV730_me.bin"); |
||
70 | MODULE_FIRMWARE("radeon/RV710_pfp.bin"); |
||
71 | MODULE_FIRMWARE("radeon/RV710_me.bin"); |
||
1321 | serge | 72 | MODULE_FIRMWARE("radeon/R600_rlc.bin"); |
73 | MODULE_FIRMWARE("radeon/R700_rlc.bin"); |
||
1963 | serge | 74 | MODULE_FIRMWARE("radeon/CEDAR_pfp.bin"); |
75 | MODULE_FIRMWARE("radeon/CEDAR_me.bin"); |
||
76 | MODULE_FIRMWARE("radeon/CEDAR_rlc.bin"); |
||
77 | MODULE_FIRMWARE("radeon/REDWOOD_pfp.bin"); |
||
78 | MODULE_FIRMWARE("radeon/REDWOOD_me.bin"); |
||
79 | MODULE_FIRMWARE("radeon/REDWOOD_rlc.bin"); |
||
80 | MODULE_FIRMWARE("radeon/JUNIPER_pfp.bin"); |
||
81 | MODULE_FIRMWARE("radeon/JUNIPER_me.bin"); |
||
82 | MODULE_FIRMWARE("radeon/JUNIPER_rlc.bin"); |
||
83 | MODULE_FIRMWARE("radeon/CYPRESS_pfp.bin"); |
||
84 | MODULE_FIRMWARE("radeon/CYPRESS_me.bin"); |
||
85 | MODULE_FIRMWARE("radeon/CYPRESS_rlc.bin"); |
||
86 | MODULE_FIRMWARE("radeon/PALM_pfp.bin"); |
||
87 | MODULE_FIRMWARE("radeon/PALM_me.bin"); |
||
88 | MODULE_FIRMWARE("radeon/SUMO_rlc.bin"); |
||
89 | MODULE_FIRMWARE("radeon/SUMO_pfp.bin"); |
||
90 | MODULE_FIRMWARE("radeon/SUMO_me.bin"); |
||
91 | MODULE_FIRMWARE("radeon/SUMO2_pfp.bin"); |
||
92 | MODULE_FIRMWARE("radeon/SUMO2_me.bin"); |
||
1221 | serge | 93 | |
94 | int r600_debugfs_mc_info_init(struct radeon_device *rdev); |
||
95 | |||
96 | /* r600,rv610,rv630,rv620,rv635,rv670 */ |
||
1128 | serge | 97 | int r600_mc_wait_for_idle(struct radeon_device *rdev); |
98 | void r600_gpu_init(struct radeon_device *rdev); |
||
1221 | serge | 99 | void r600_fini(struct radeon_device *rdev); |
1963 | serge | 100 | void r600_irq_disable(struct radeon_device *rdev); |
101 | static void r600_pcie_gen2_enable(struct radeon_device *rdev); |
||
1128 | serge | 102 | |
1963 | serge | 103 | /* get temperature in millidegrees */ |
104 | int rv6xx_get_temp(struct radeon_device *rdev) |
||
105 | { |
||
106 | u32 temp = (RREG32(CG_THERMAL_STATUS) & ASIC_T_MASK) >> |
||
107 | ASIC_T_SHIFT; |
||
108 | int actual_temp = temp & 0xff; |
||
109 | |||
110 | if (temp & 0x100) |
||
111 | actual_temp -= 256; |
||
112 | |||
113 | return actual_temp * 1000; |
||
114 | } |
||
115 | |||
116 | |||
117 | |||
118 | |||
119 | |||
120 | |||
121 | bool r600_gui_idle(struct radeon_device *rdev) |
||
122 | { |
||
123 | if (RREG32(GRBM_STATUS) & GUI_ACTIVE) |
||
124 | return false; |
||
125 | else |
||
126 | return true; |
||
127 | } |
||
128 | |||
1321 | serge | 129 | /* hpd for digital panel detect/disconnect */ |
130 | bool r600_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd) |
||
131 | { |
||
132 | bool connected = false; |
||
133 | |||
134 | if (ASIC_IS_DCE3(rdev)) { |
||
135 | switch (hpd) { |
||
136 | case RADEON_HPD_1: |
||
137 | if (RREG32(DC_HPD1_INT_STATUS) & DC_HPDx_SENSE) |
||
138 | connected = true; |
||
139 | break; |
||
140 | case RADEON_HPD_2: |
||
141 | if (RREG32(DC_HPD2_INT_STATUS) & DC_HPDx_SENSE) |
||
142 | connected = true; |
||
143 | break; |
||
144 | case RADEON_HPD_3: |
||
145 | if (RREG32(DC_HPD3_INT_STATUS) & DC_HPDx_SENSE) |
||
146 | connected = true; |
||
147 | break; |
||
148 | case RADEON_HPD_4: |
||
149 | if (RREG32(DC_HPD4_INT_STATUS) & DC_HPDx_SENSE) |
||
150 | connected = true; |
||
151 | break; |
||
152 | /* DCE 3.2 */ |
||
153 | case RADEON_HPD_5: |
||
154 | if (RREG32(DC_HPD5_INT_STATUS) & DC_HPDx_SENSE) |
||
155 | connected = true; |
||
156 | break; |
||
157 | case RADEON_HPD_6: |
||
158 | if (RREG32(DC_HPD6_INT_STATUS) & DC_HPDx_SENSE) |
||
159 | connected = true; |
||
160 | break; |
||
161 | default: |
||
162 | break; |
||
163 | } |
||
164 | } else { |
||
165 | switch (hpd) { |
||
166 | case RADEON_HPD_1: |
||
167 | if (RREG32(DC_HOT_PLUG_DETECT1_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE) |
||
168 | connected = true; |
||
169 | break; |
||
170 | case RADEON_HPD_2: |
||
171 | if (RREG32(DC_HOT_PLUG_DETECT2_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE) |
||
172 | connected = true; |
||
173 | break; |
||
174 | case RADEON_HPD_3: |
||
175 | if (RREG32(DC_HOT_PLUG_DETECT3_INT_STATUS) & DC_HOT_PLUG_DETECTx_SENSE) |
||
176 | connected = true; |
||
177 | break; |
||
178 | default: |
||
179 | break; |
||
180 | } |
||
181 | } |
||
182 | return connected; |
||
183 | } |
||
184 | |||
185 | void r600_hpd_set_polarity(struct radeon_device *rdev, |
||
186 | enum radeon_hpd_id hpd) |
||
187 | { |
||
188 | u32 tmp; |
||
189 | bool connected = r600_hpd_sense(rdev, hpd); |
||
190 | |||
191 | if (ASIC_IS_DCE3(rdev)) { |
||
192 | switch (hpd) { |
||
193 | case RADEON_HPD_1: |
||
194 | tmp = RREG32(DC_HPD1_INT_CONTROL); |
||
195 | if (connected) |
||
196 | tmp &= ~DC_HPDx_INT_POLARITY; |
||
197 | else |
||
198 | tmp |= DC_HPDx_INT_POLARITY; |
||
199 | WREG32(DC_HPD1_INT_CONTROL, tmp); |
||
200 | break; |
||
201 | case RADEON_HPD_2: |
||
202 | tmp = RREG32(DC_HPD2_INT_CONTROL); |
||
203 | if (connected) |
||
204 | tmp &= ~DC_HPDx_INT_POLARITY; |
||
205 | else |
||
206 | tmp |= DC_HPDx_INT_POLARITY; |
||
207 | WREG32(DC_HPD2_INT_CONTROL, tmp); |
||
208 | break; |
||
209 | case RADEON_HPD_3: |
||
210 | tmp = RREG32(DC_HPD3_INT_CONTROL); |
||
211 | if (connected) |
||
212 | tmp &= ~DC_HPDx_INT_POLARITY; |
||
213 | else |
||
214 | tmp |= DC_HPDx_INT_POLARITY; |
||
215 | WREG32(DC_HPD3_INT_CONTROL, tmp); |
||
216 | break; |
||
217 | case RADEON_HPD_4: |
||
218 | tmp = RREG32(DC_HPD4_INT_CONTROL); |
||
219 | if (connected) |
||
220 | tmp &= ~DC_HPDx_INT_POLARITY; |
||
221 | else |
||
222 | tmp |= DC_HPDx_INT_POLARITY; |
||
223 | WREG32(DC_HPD4_INT_CONTROL, tmp); |
||
224 | break; |
||
225 | case RADEON_HPD_5: |
||
226 | tmp = RREG32(DC_HPD5_INT_CONTROL); |
||
227 | if (connected) |
||
228 | tmp &= ~DC_HPDx_INT_POLARITY; |
||
229 | else |
||
230 | tmp |= DC_HPDx_INT_POLARITY; |
||
231 | WREG32(DC_HPD5_INT_CONTROL, tmp); |
||
232 | break; |
||
233 | /* DCE 3.2 */ |
||
234 | case RADEON_HPD_6: |
||
235 | tmp = RREG32(DC_HPD6_INT_CONTROL); |
||
236 | if (connected) |
||
237 | tmp &= ~DC_HPDx_INT_POLARITY; |
||
238 | else |
||
239 | tmp |= DC_HPDx_INT_POLARITY; |
||
240 | WREG32(DC_HPD6_INT_CONTROL, tmp); |
||
241 | break; |
||
242 | default: |
||
243 | break; |
||
244 | } |
||
245 | } else { |
||
246 | switch (hpd) { |
||
247 | case RADEON_HPD_1: |
||
248 | tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL); |
||
249 | if (connected) |
||
250 | tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
251 | else |
||
252 | tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
253 | WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp); |
||
254 | break; |
||
255 | case RADEON_HPD_2: |
||
256 | tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL); |
||
257 | if (connected) |
||
258 | tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
259 | else |
||
260 | tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
261 | WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp); |
||
262 | break; |
||
263 | case RADEON_HPD_3: |
||
264 | tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL); |
||
265 | if (connected) |
||
266 | tmp &= ~DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
267 | else |
||
268 | tmp |= DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
269 | WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp); |
||
270 | break; |
||
271 | default: |
||
272 | break; |
||
273 | } |
||
274 | } |
||
275 | } |
||
276 | |||
277 | void r600_hpd_init(struct radeon_device *rdev) |
||
278 | { |
||
279 | struct drm_device *dev = rdev->ddev; |
||
280 | struct drm_connector *connector; |
||
281 | |||
282 | if (ASIC_IS_DCE3(rdev)) { |
||
283 | u32 tmp = DC_HPDx_CONNECTION_TIMER(0x9c4) | DC_HPDx_RX_INT_TIMER(0xfa); |
||
284 | if (ASIC_IS_DCE32(rdev)) |
||
285 | tmp |= DC_HPDx_EN; |
||
286 | |||
287 | list_for_each_entry(connector, &dev->mode_config.connector_list, head) { |
||
288 | struct radeon_connector *radeon_connector = to_radeon_connector(connector); |
||
289 | switch (radeon_connector->hpd.hpd) { |
||
290 | case RADEON_HPD_1: |
||
291 | WREG32(DC_HPD1_CONTROL, tmp); |
||
1403 | serge | 292 | // rdev->irq.hpd[0] = true; |
1321 | serge | 293 | break; |
294 | case RADEON_HPD_2: |
||
295 | WREG32(DC_HPD2_CONTROL, tmp); |
||
1403 | serge | 296 | // rdev->irq.hpd[1] = true; |
1321 | serge | 297 | break; |
298 | case RADEON_HPD_3: |
||
299 | WREG32(DC_HPD3_CONTROL, tmp); |
||
1403 | serge | 300 | // rdev->irq.hpd[2] = true; |
1321 | serge | 301 | break; |
302 | case RADEON_HPD_4: |
||
303 | WREG32(DC_HPD4_CONTROL, tmp); |
||
1403 | serge | 304 | // rdev->irq.hpd[3] = true; |
1321 | serge | 305 | break; |
306 | /* DCE 3.2 */ |
||
307 | case RADEON_HPD_5: |
||
308 | WREG32(DC_HPD5_CONTROL, tmp); |
||
1403 | serge | 309 | // rdev->irq.hpd[4] = true; |
1321 | serge | 310 | break; |
311 | case RADEON_HPD_6: |
||
312 | WREG32(DC_HPD6_CONTROL, tmp); |
||
1403 | serge | 313 | // rdev->irq.hpd[5] = true; |
1321 | serge | 314 | break; |
315 | default: |
||
316 | break; |
||
317 | } |
||
318 | } |
||
319 | } else { |
||
320 | list_for_each_entry(connector, &dev->mode_config.connector_list, head) { |
||
321 | struct radeon_connector *radeon_connector = to_radeon_connector(connector); |
||
322 | switch (radeon_connector->hpd.hpd) { |
||
323 | case RADEON_HPD_1: |
||
324 | WREG32(DC_HOT_PLUG_DETECT1_CONTROL, DC_HOT_PLUG_DETECTx_EN); |
||
1403 | serge | 325 | // rdev->irq.hpd[0] = true; |
1321 | serge | 326 | break; |
327 | case RADEON_HPD_2: |
||
328 | WREG32(DC_HOT_PLUG_DETECT2_CONTROL, DC_HOT_PLUG_DETECTx_EN); |
||
1403 | serge | 329 | // rdev->irq.hpd[1] = true; |
1321 | serge | 330 | break; |
331 | case RADEON_HPD_3: |
||
332 | WREG32(DC_HOT_PLUG_DETECT3_CONTROL, DC_HOT_PLUG_DETECTx_EN); |
||
1403 | serge | 333 | // rdev->irq.hpd[2] = true; |
1321 | serge | 334 | break; |
335 | default: |
||
336 | break; |
||
337 | } |
||
338 | } |
||
339 | } |
||
1403 | serge | 340 | // if (rdev->irq.installed) |
341 | // r600_irq_set(rdev); |
||
1321 | serge | 342 | } |
343 | |||
344 | void r600_hpd_fini(struct radeon_device *rdev) |
||
345 | { |
||
346 | struct drm_device *dev = rdev->ddev; |
||
347 | struct drm_connector *connector; |
||
348 | |||
349 | if (ASIC_IS_DCE3(rdev)) { |
||
350 | list_for_each_entry(connector, &dev->mode_config.connector_list, head) { |
||
351 | struct radeon_connector *radeon_connector = to_radeon_connector(connector); |
||
352 | switch (radeon_connector->hpd.hpd) { |
||
353 | case RADEON_HPD_1: |
||
354 | WREG32(DC_HPD1_CONTROL, 0); |
||
1403 | serge | 355 | // rdev->irq.hpd[0] = false; |
1321 | serge | 356 | break; |
357 | case RADEON_HPD_2: |
||
358 | WREG32(DC_HPD2_CONTROL, 0); |
||
1403 | serge | 359 | // rdev->irq.hpd[1] = false; |
1321 | serge | 360 | break; |
361 | case RADEON_HPD_3: |
||
362 | WREG32(DC_HPD3_CONTROL, 0); |
||
1403 | serge | 363 | // rdev->irq.hpd[2] = false; |
1321 | serge | 364 | break; |
365 | case RADEON_HPD_4: |
||
366 | WREG32(DC_HPD4_CONTROL, 0); |
||
1403 | serge | 367 | // rdev->irq.hpd[3] = false; |
1321 | serge | 368 | break; |
369 | /* DCE 3.2 */ |
||
370 | case RADEON_HPD_5: |
||
371 | WREG32(DC_HPD5_CONTROL, 0); |
||
1403 | serge | 372 | // rdev->irq.hpd[4] = false; |
1321 | serge | 373 | break; |
374 | case RADEON_HPD_6: |
||
375 | WREG32(DC_HPD6_CONTROL, 0); |
||
1403 | serge | 376 | // rdev->irq.hpd[5] = false; |
1321 | serge | 377 | break; |
378 | default: |
||
379 | break; |
||
380 | } |
||
381 | } |
||
382 | } else { |
||
383 | list_for_each_entry(connector, &dev->mode_config.connector_list, head) { |
||
384 | struct radeon_connector *radeon_connector = to_radeon_connector(connector); |
||
385 | switch (radeon_connector->hpd.hpd) { |
||
386 | case RADEON_HPD_1: |
||
387 | WREG32(DC_HOT_PLUG_DETECT1_CONTROL, 0); |
||
1403 | serge | 388 | // rdev->irq.hpd[0] = false; |
1321 | serge | 389 | break; |
390 | case RADEON_HPD_2: |
||
391 | WREG32(DC_HOT_PLUG_DETECT2_CONTROL, 0); |
||
1403 | serge | 392 | // rdev->irq.hpd[1] = false; |
1321 | serge | 393 | break; |
394 | case RADEON_HPD_3: |
||
395 | WREG32(DC_HOT_PLUG_DETECT3_CONTROL, 0); |
||
1403 | serge | 396 | // rdev->irq.hpd[2] = false; |
1321 | serge | 397 | break; |
398 | default: |
||
399 | break; |
||
400 | } |
||
401 | } |
||
402 | } |
||
403 | } |
||
404 | |||
1128 | serge | 405 | /* |
1221 | serge | 406 | * R600 PCIE GART |
1128 | serge | 407 | */ |
1221 | serge | 408 | void r600_pcie_gart_tlb_flush(struct radeon_device *rdev) |
409 | { |
||
410 | unsigned i; |
||
411 | u32 tmp; |
||
1128 | serge | 412 | |
1430 | serge | 413 | /* flush hdp cache so updates hit vram */ |
1963 | serge | 414 | if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_RV740) && |
415 | !(rdev->flags & RADEON_IS_AGP)) { |
||
416 | void __iomem *ptr = (void *)rdev->gart.table.vram.ptr; |
||
417 | u32 tmp; |
||
418 | |||
419 | /* r7xx hw bug. write to HDP_DEBUG1 followed by fb read |
||
420 | * rather than write to HDP_REG_COHERENCY_FLUSH_CNTL |
||
421 | * This seems to cause problems on some AGP cards. Just use the old |
||
422 | * method for them. |
||
423 | */ |
||
424 | WREG32(HDP_DEBUG1, 0); |
||
425 | tmp = readl((void __iomem *)ptr); |
||
426 | } else |
||
1430 | serge | 427 | WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1); |
428 | |||
1221 | serge | 429 | WREG32(VM_CONTEXT0_INVALIDATION_LOW_ADDR, rdev->mc.gtt_start >> 12); |
430 | WREG32(VM_CONTEXT0_INVALIDATION_HIGH_ADDR, (rdev->mc.gtt_end - 1) >> 12); |
||
431 | WREG32(VM_CONTEXT0_REQUEST_RESPONSE, REQUEST_TYPE(1)); |
||
432 | for (i = 0; i < rdev->usec_timeout; i++) { |
||
433 | /* read MC_STATUS */ |
||
434 | tmp = RREG32(VM_CONTEXT0_REQUEST_RESPONSE); |
||
435 | tmp = (tmp & RESPONSE_TYPE_MASK) >> RESPONSE_TYPE_SHIFT; |
||
436 | if (tmp == 2) { |
||
437 | printk(KERN_WARNING "[drm] r600 flush TLB failed\n"); |
||
438 | return; |
||
439 | } |
||
440 | if (tmp) { |
||
441 | return; |
||
442 | } |
||
443 | udelay(1); |
||
1128 | serge | 444 | } |
1221 | serge | 445 | } |
1128 | serge | 446 | |
1221 | serge | 447 | int r600_pcie_gart_init(struct radeon_device *rdev) |
448 | { |
||
449 | int r; |
||
450 | |||
451 | if (rdev->gart.table.vram.robj) { |
||
1963 | serge | 452 | WARN(1, "R600 PCIE GART already initialized\n"); |
1221 | serge | 453 | return 0; |
454 | } |
||
455 | /* Initialize common gart structure */ |
||
456 | r = radeon_gart_init(rdev); |
||
457 | if (r) |
||
458 | return r; |
||
459 | rdev->gart.table_size = rdev->gart.num_gpu_pages * 8; |
||
460 | return radeon_gart_table_vram_alloc(rdev); |
||
461 | } |
||
462 | |||
463 | int r600_pcie_gart_enable(struct radeon_device *rdev) |
||
464 | { |
||
465 | u32 tmp; |
||
466 | int r, i; |
||
467 | |||
468 | if (rdev->gart.table.vram.robj == NULL) { |
||
469 | dev_err(rdev->dev, "No VRAM object for PCIE GART.\n"); |
||
470 | return -EINVAL; |
||
471 | } |
||
472 | r = radeon_gart_table_vram_pin(rdev); |
||
473 | if (r) |
||
474 | return r; |
||
1430 | serge | 475 | radeon_gart_restore(rdev); |
1221 | serge | 476 | |
477 | /* Setup L2 cache */ |
||
478 | WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING | |
||
479 | ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE | |
||
480 | EFFECTIVE_L2_QUEUE_SIZE(7)); |
||
481 | WREG32(VM_L2_CNTL2, 0); |
||
482 | WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1)); |
||
483 | /* Setup TLB control */ |
||
484 | tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING | |
||
485 | SYSTEM_ACCESS_MODE_NOT_IN_SYS | |
||
486 | EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) | |
||
487 | ENABLE_WAIT_L2_QUERY; |
||
488 | WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp); |
||
489 | WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp); |
||
490 | WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp | ENABLE_L1_STRICT_ORDERING); |
||
491 | WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp); |
||
492 | WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp); |
||
493 | WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp); |
||
494 | WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp); |
||
495 | WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp); |
||
496 | WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp); |
||
497 | WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp); |
||
498 | WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp); |
||
499 | WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp); |
||
500 | WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE); |
||
501 | WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE); |
||
502 | WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12); |
||
503 | WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12); |
||
504 | WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12); |
||
505 | WREG32(VM_CONTEXT0_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) | |
||
506 | RANGE_PROTECTION_FAULT_ENABLE_DEFAULT); |
||
507 | WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, |
||
508 | (u32)(rdev->dummy_page.addr >> 12)); |
||
509 | for (i = 1; i < 7; i++) |
||
510 | WREG32(VM_CONTEXT0_CNTL + (i * 4), 0); |
||
511 | |||
512 | r600_pcie_gart_tlb_flush(rdev); |
||
513 | rdev->gart.ready = true; |
||
1128 | serge | 514 | return 0; |
515 | } |
||
516 | |||
1221 | serge | 517 | void r600_pcie_gart_disable(struct radeon_device *rdev) |
1128 | serge | 518 | { |
1221 | serge | 519 | u32 tmp; |
1321 | serge | 520 | int i, r; |
1221 | serge | 521 | |
522 | /* Disable all tables */ |
||
523 | for (i = 0; i < 7; i++) |
||
524 | WREG32(VM_CONTEXT0_CNTL + (i * 4), 0); |
||
525 | |||
526 | /* Disable L2 cache */ |
||
527 | WREG32(VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING | |
||
528 | EFFECTIVE_L2_QUEUE_SIZE(7)); |
||
529 | WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1)); |
||
530 | /* Setup L1 TLB control */ |
||
531 | tmp = EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) | |
||
532 | ENABLE_WAIT_L2_QUERY; |
||
533 | WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp); |
||
534 | WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp); |
||
535 | WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp); |
||
536 | WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp); |
||
537 | WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp); |
||
538 | WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp); |
||
539 | WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp); |
||
540 | WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp); |
||
541 | WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp); |
||
542 | WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp); |
||
543 | WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp); |
||
544 | WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp); |
||
545 | WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp); |
||
546 | WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp); |
||
547 | if (rdev->gart.table.vram.robj) { |
||
1403 | serge | 548 | r = radeon_bo_reserve(rdev->gart.table.vram.robj, false); |
549 | if (likely(r == 0)) { |
||
550 | radeon_bo_kunmap(rdev->gart.table.vram.robj); |
||
551 | radeon_bo_unpin(rdev->gart.table.vram.robj); |
||
552 | radeon_bo_unreserve(rdev->gart.table.vram.robj); |
||
553 | } |
||
1221 | serge | 554 | } |
1128 | serge | 555 | } |
556 | |||
1221 | serge | 557 | void r600_pcie_gart_fini(struct radeon_device *rdev) |
558 | { |
||
1963 | serge | 559 | radeon_gart_fini(rdev); |
1221 | serge | 560 | r600_pcie_gart_disable(rdev); |
561 | radeon_gart_table_vram_free(rdev); |
||
562 | } |
||
1128 | serge | 563 | |
1221 | serge | 564 | void r600_agp_enable(struct radeon_device *rdev) |
1128 | serge | 565 | { |
1221 | serge | 566 | u32 tmp; |
567 | int i; |
||
568 | |||
569 | /* Setup L2 cache */ |
||
570 | WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | ENABLE_L2_FRAGMENT_PROCESSING | |
||
571 | ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE | |
||
572 | EFFECTIVE_L2_QUEUE_SIZE(7)); |
||
573 | WREG32(VM_L2_CNTL2, 0); |
||
574 | WREG32(VM_L2_CNTL3, BANK_SELECT_0(0) | BANK_SELECT_1(1)); |
||
575 | /* Setup TLB control */ |
||
576 | tmp = ENABLE_L1_TLB | ENABLE_L1_FRAGMENT_PROCESSING | |
||
577 | SYSTEM_ACCESS_MODE_NOT_IN_SYS | |
||
578 | EFFECTIVE_L1_TLB_SIZE(5) | EFFECTIVE_L1_QUEUE_SIZE(5) | |
||
579 | ENABLE_WAIT_L2_QUERY; |
||
580 | WREG32(MC_VM_L1_TLB_MCB_RD_SYS_CNTL, tmp); |
||
581 | WREG32(MC_VM_L1_TLB_MCB_WR_SYS_CNTL, tmp); |
||
582 | WREG32(MC_VM_L1_TLB_MCB_RD_HDP_CNTL, tmp | ENABLE_L1_STRICT_ORDERING); |
||
583 | WREG32(MC_VM_L1_TLB_MCB_WR_HDP_CNTL, tmp); |
||
584 | WREG32(MC_VM_L1_TLB_MCD_RD_A_CNTL, tmp); |
||
585 | WREG32(MC_VM_L1_TLB_MCD_WR_A_CNTL, tmp); |
||
586 | WREG32(MC_VM_L1_TLB_MCD_RD_B_CNTL, tmp); |
||
587 | WREG32(MC_VM_L1_TLB_MCD_WR_B_CNTL, tmp); |
||
588 | WREG32(MC_VM_L1_TLB_MCB_RD_GFX_CNTL, tmp); |
||
589 | WREG32(MC_VM_L1_TLB_MCB_WR_GFX_CNTL, tmp); |
||
590 | WREG32(MC_VM_L1_TLB_MCB_RD_PDMA_CNTL, tmp); |
||
591 | WREG32(MC_VM_L1_TLB_MCB_WR_PDMA_CNTL, tmp); |
||
592 | WREG32(MC_VM_L1_TLB_MCB_RD_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE); |
||
593 | WREG32(MC_VM_L1_TLB_MCB_WR_SEM_CNTL, tmp | ENABLE_SEMAPHORE_MODE); |
||
594 | for (i = 0; i < 7; i++) |
||
595 | WREG32(VM_CONTEXT0_CNTL + (i * 4), 0); |
||
1128 | serge | 596 | } |
597 | |||
598 | int r600_mc_wait_for_idle(struct radeon_device *rdev) |
||
599 | { |
||
1221 | serge | 600 | unsigned i; |
601 | u32 tmp; |
||
602 | |||
603 | for (i = 0; i < rdev->usec_timeout; i++) { |
||
604 | /* read MC_STATUS */ |
||
605 | tmp = RREG32(R_000E50_SRBM_STATUS) & 0x3F00; |
||
606 | if (!tmp) |
||
1128 | serge | 607 | return 0; |
1221 | serge | 608 | udelay(1); |
609 | } |
||
610 | return -1; |
||
1128 | serge | 611 | } |
612 | |||
1221 | serge | 613 | static void r600_mc_program(struct radeon_device *rdev) |
1128 | serge | 614 | { |
1221 | serge | 615 | struct rv515_mc_save save; |
616 | u32 tmp; |
||
617 | int i, j; |
||
618 | |||
619 | /* Initialize HDP */ |
||
620 | for (i = 0, j = 0; i < 32; i++, j += 0x18) { |
||
621 | WREG32((0x2c14 + j), 0x00000000); |
||
622 | WREG32((0x2c18 + j), 0x00000000); |
||
623 | WREG32((0x2c1c + j), 0x00000000); |
||
624 | WREG32((0x2c20 + j), 0x00000000); |
||
625 | WREG32((0x2c24 + j), 0x00000000); |
||
626 | } |
||
627 | WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0); |
||
628 | |||
629 | rv515_mc_stop(rdev, &save); |
||
630 | if (r600_mc_wait_for_idle(rdev)) { |
||
631 | dev_warn(rdev->dev, "Wait for MC idle timedout !\n"); |
||
632 | } |
||
633 | /* Lockout access through VGA aperture (doesn't exist before R600) */ |
||
634 | WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE); |
||
635 | /* Update configuration */ |
||
636 | if (rdev->flags & RADEON_IS_AGP) { |
||
637 | if (rdev->mc.vram_start < rdev->mc.gtt_start) { |
||
638 | /* VRAM before AGP */ |
||
639 | WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, |
||
640 | rdev->mc.vram_start >> 12); |
||
641 | WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, |
||
642 | rdev->mc.gtt_end >> 12); |
||
643 | } else { |
||
644 | /* VRAM after AGP */ |
||
645 | WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, |
||
646 | rdev->mc.gtt_start >> 12); |
||
647 | WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, |
||
648 | rdev->mc.vram_end >> 12); |
||
649 | } |
||
650 | } else { |
||
651 | WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, rdev->mc.vram_start >> 12); |
||
652 | WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, rdev->mc.vram_end >> 12); |
||
653 | } |
||
654 | WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, 0); |
||
655 | tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16; |
||
656 | tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF); |
||
657 | WREG32(MC_VM_FB_LOCATION, tmp); |
||
658 | WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8)); |
||
659 | WREG32(HDP_NONSURFACE_INFO, (2 << 7)); |
||
1963 | serge | 660 | WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF); |
1221 | serge | 661 | if (rdev->flags & RADEON_IS_AGP) { |
662 | WREG32(MC_VM_AGP_TOP, rdev->mc.gtt_end >> 22); |
||
663 | WREG32(MC_VM_AGP_BOT, rdev->mc.gtt_start >> 22); |
||
664 | WREG32(MC_VM_AGP_BASE, rdev->mc.agp_base >> 22); |
||
665 | } else { |
||
666 | WREG32(MC_VM_AGP_BASE, 0); |
||
667 | WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF); |
||
668 | WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF); |
||
669 | } |
||
670 | if (r600_mc_wait_for_idle(rdev)) { |
||
671 | dev_warn(rdev->dev, "Wait for MC idle timedout !\n"); |
||
672 | } |
||
673 | rv515_mc_resume(rdev, &save); |
||
674 | /* we need to own VRAM, so turn off the VGA renderer here |
||
675 | * to stop it overwriting our objects */ |
||
676 | rv515_vga_render_disable(rdev); |
||
1128 | serge | 677 | } |
678 | |||
1430 | serge | 679 | /** |
680 | * r600_vram_gtt_location - try to find VRAM & GTT location |
||
681 | * @rdev: radeon device structure holding all necessary informations |
||
682 | * @mc: memory controller structure holding memory informations |
||
683 | * |
||
684 | * Function will place try to place VRAM at same place as in CPU (PCI) |
||
685 | * address space as some GPU seems to have issue when we reprogram at |
||
686 | * different address space. |
||
687 | * |
||
688 | * If there is not enough space to fit the unvisible VRAM after the |
||
689 | * aperture then we limit the VRAM size to the aperture. |
||
690 | * |
||
691 | * If we are using AGP then place VRAM adjacent to AGP aperture are we need |
||
692 | * them to be in one from GPU point of view so that we can program GPU to |
||
693 | * catch access outside them (weird GPU policy see ??). |
||
694 | * |
||
695 | * This function will never fails, worst case are limiting VRAM or GTT. |
||
696 | * |
||
697 | * Note: GTT start, end, size should be initialized before calling this |
||
698 | * function on AGP platform. |
||
699 | */ |
||
1963 | serge | 700 | static void r600_vram_gtt_location(struct radeon_device *rdev, struct radeon_mc *mc) |
1430 | serge | 701 | { |
702 | u64 size_bf, size_af; |
||
703 | |||
704 | if (mc->mc_vram_size > 0xE0000000) { |
||
705 | /* leave room for at least 512M GTT */ |
||
706 | dev_warn(rdev->dev, "limiting VRAM\n"); |
||
707 | mc->real_vram_size = 0xE0000000; |
||
708 | mc->mc_vram_size = 0xE0000000; |
||
709 | } |
||
710 | if (rdev->flags & RADEON_IS_AGP) { |
||
711 | size_bf = mc->gtt_start; |
||
712 | size_af = 0xFFFFFFFF - mc->gtt_end + 1; |
||
713 | if (size_bf > size_af) { |
||
714 | if (mc->mc_vram_size > size_bf) { |
||
715 | dev_warn(rdev->dev, "limiting VRAM\n"); |
||
716 | mc->real_vram_size = size_bf; |
||
717 | mc->mc_vram_size = size_bf; |
||
718 | } |
||
719 | mc->vram_start = mc->gtt_start - mc->mc_vram_size; |
||
720 | } else { |
||
721 | if (mc->mc_vram_size > size_af) { |
||
722 | dev_warn(rdev->dev, "limiting VRAM\n"); |
||
723 | mc->real_vram_size = size_af; |
||
724 | mc->mc_vram_size = size_af; |
||
725 | } |
||
726 | mc->vram_start = mc->gtt_end; |
||
727 | } |
||
728 | mc->vram_end = mc->vram_start + mc->mc_vram_size - 1; |
||
729 | dev_info(rdev->dev, "VRAM: %lluM 0x%08llX - 0x%08llX (%lluM used)\n", |
||
730 | mc->mc_vram_size >> 20, mc->vram_start, |
||
731 | mc->vram_end, mc->real_vram_size >> 20); |
||
732 | } else { |
||
733 | u64 base = 0; |
||
1963 | serge | 734 | if (rdev->flags & RADEON_IS_IGP) { |
735 | base = RREG32(MC_VM_FB_LOCATION) & 0xFFFF; |
||
736 | base <<= 24; |
||
737 | } |
||
1430 | serge | 738 | radeon_vram_location(rdev, &rdev->mc, base); |
1963 | serge | 739 | rdev->mc.gtt_base_align = 0; |
1430 | serge | 740 | radeon_gtt_location(rdev, mc); |
741 | } |
||
742 | } |
||
743 | |||
1221 | serge | 744 | int r600_mc_init(struct radeon_device *rdev) |
1128 | serge | 745 | { |
1221 | serge | 746 | u32 tmp; |
1268 | serge | 747 | int chansize, numchan; |
1128 | serge | 748 | |
1221 | serge | 749 | /* Get VRAM informations */ |
1128 | serge | 750 | rdev->mc.vram_is_ddr = true; |
1221 | serge | 751 | tmp = RREG32(RAMCFG); |
752 | if (tmp & CHANSIZE_OVERRIDE) { |
||
1128 | serge | 753 | chansize = 16; |
1221 | serge | 754 | } else if (tmp & CHANSIZE_MASK) { |
1128 | serge | 755 | chansize = 64; |
756 | } else { |
||
757 | chansize = 32; |
||
758 | } |
||
1268 | serge | 759 | tmp = RREG32(CHMAP); |
760 | switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) { |
||
761 | case 0: |
||
762 | default: |
||
763 | numchan = 1; |
||
764 | break; |
||
765 | case 1: |
||
766 | numchan = 2; |
||
767 | break; |
||
768 | case 2: |
||
769 | numchan = 4; |
||
770 | break; |
||
771 | case 3: |
||
772 | numchan = 8; |
||
773 | break; |
||
1128 | serge | 774 | } |
1268 | serge | 775 | rdev->mc.vram_width = numchan * chansize; |
1221 | serge | 776 | /* Could aper size report 0 ? */ |
1963 | serge | 777 | rdev->mc.aper_base = pci_resource_start(rdev->pdev, 0); |
778 | rdev->mc.aper_size = pci_resource_len(rdev->pdev, 0); |
||
1221 | serge | 779 | /* Setup GPU memory space */ |
780 | rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE); |
||
781 | rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE); |
||
1430 | serge | 782 | rdev->mc.visible_vram_size = rdev->mc.aper_size; |
783 | r600_vram_gtt_location(rdev, &rdev->mc); |
||
1963 | serge | 784 | |
785 | if (rdev->flags & RADEON_IS_IGP) { |
||
786 | rs690_pm_info(rdev); |
||
1403 | serge | 787 | rdev->mc.igp_sideport_enabled = radeon_atombios_sideport_present(rdev); |
1963 | serge | 788 | } |
789 | radeon_update_bandwidth_info(rdev); |
||
1221 | serge | 790 | return 0; |
1128 | serge | 791 | } |
792 | |||
1221 | serge | 793 | /* We doesn't check that the GPU really needs a reset we simply do the |
794 | * reset, it's up to the caller to determine if the GPU needs one. We |
||
795 | * might add an helper function to check that. |
||
796 | */ |
||
797 | int r600_gpu_soft_reset(struct radeon_device *rdev) |
||
1128 | serge | 798 | { |
1221 | serge | 799 | struct rv515_mc_save save; |
800 | u32 grbm_busy_mask = S_008010_VC_BUSY(1) | S_008010_VGT_BUSY_NO_DMA(1) | |
||
801 | S_008010_VGT_BUSY(1) | S_008010_TA03_BUSY(1) | |
||
802 | S_008010_TC_BUSY(1) | S_008010_SX_BUSY(1) | |
||
803 | S_008010_SH_BUSY(1) | S_008010_SPI03_BUSY(1) | |
||
804 | S_008010_SMX_BUSY(1) | S_008010_SC_BUSY(1) | |
||
805 | S_008010_PA_BUSY(1) | S_008010_DB03_BUSY(1) | |
||
806 | S_008010_CR_BUSY(1) | S_008010_CB03_BUSY(1) | |
||
807 | S_008010_GUI_ACTIVE(1); |
||
808 | u32 grbm2_busy_mask = S_008014_SPI0_BUSY(1) | S_008014_SPI1_BUSY(1) | |
||
809 | S_008014_SPI2_BUSY(1) | S_008014_SPI3_BUSY(1) | |
||
810 | S_008014_TA0_BUSY(1) | S_008014_TA1_BUSY(1) | |
||
811 | S_008014_TA2_BUSY(1) | S_008014_TA3_BUSY(1) | |
||
812 | S_008014_DB0_BUSY(1) | S_008014_DB1_BUSY(1) | |
||
813 | S_008014_DB2_BUSY(1) | S_008014_DB3_BUSY(1) | |
||
814 | S_008014_CB0_BUSY(1) | S_008014_CB1_BUSY(1) | |
||
815 | S_008014_CB2_BUSY(1) | S_008014_CB3_BUSY(1); |
||
816 | u32 tmp; |
||
1128 | serge | 817 | |
1963 | serge | 818 | if (!(RREG32(GRBM_STATUS) & GUI_ACTIVE)) |
819 | return 0; |
||
820 | |||
1221 | serge | 821 | dev_info(rdev->dev, "GPU softreset \n"); |
822 | dev_info(rdev->dev, " R_008010_GRBM_STATUS=0x%08X\n", |
||
823 | RREG32(R_008010_GRBM_STATUS)); |
||
824 | dev_info(rdev->dev, " R_008014_GRBM_STATUS2=0x%08X\n", |
||
825 | RREG32(R_008014_GRBM_STATUS2)); |
||
826 | dev_info(rdev->dev, " R_000E50_SRBM_STATUS=0x%08X\n", |
||
827 | RREG32(R_000E50_SRBM_STATUS)); |
||
828 | rv515_mc_stop(rdev, &save); |
||
829 | if (r600_mc_wait_for_idle(rdev)) { |
||
830 | dev_warn(rdev->dev, "Wait for MC idle timedout !\n"); |
||
831 | } |
||
832 | /* Disable CP parsing/prefetching */ |
||
1963 | serge | 833 | WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1)); |
1221 | serge | 834 | /* Check if any of the rendering block is busy and reset it */ |
835 | if ((RREG32(R_008010_GRBM_STATUS) & grbm_busy_mask) || |
||
836 | (RREG32(R_008014_GRBM_STATUS2) & grbm2_busy_mask)) { |
||
837 | tmp = S_008020_SOFT_RESET_CR(1) | |
||
838 | S_008020_SOFT_RESET_DB(1) | |
||
839 | S_008020_SOFT_RESET_CB(1) | |
||
840 | S_008020_SOFT_RESET_PA(1) | |
||
841 | S_008020_SOFT_RESET_SC(1) | |
||
842 | S_008020_SOFT_RESET_SMX(1) | |
||
843 | S_008020_SOFT_RESET_SPI(1) | |
||
844 | S_008020_SOFT_RESET_SX(1) | |
||
845 | S_008020_SOFT_RESET_SH(1) | |
||
846 | S_008020_SOFT_RESET_TC(1) | |
||
847 | S_008020_SOFT_RESET_TA(1) | |
||
848 | S_008020_SOFT_RESET_VC(1) | |
||
849 | S_008020_SOFT_RESET_VGT(1); |
||
850 | dev_info(rdev->dev, " R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp); |
||
851 | WREG32(R_008020_GRBM_SOFT_RESET, tmp); |
||
1963 | serge | 852 | RREG32(R_008020_GRBM_SOFT_RESET); |
853 | mdelay(15); |
||
1221 | serge | 854 | WREG32(R_008020_GRBM_SOFT_RESET, 0); |
855 | } |
||
856 | /* Reset CP (we always reset CP) */ |
||
857 | tmp = S_008020_SOFT_RESET_CP(1); |
||
858 | dev_info(rdev->dev, "R_008020_GRBM_SOFT_RESET=0x%08X\n", tmp); |
||
859 | WREG32(R_008020_GRBM_SOFT_RESET, tmp); |
||
1963 | serge | 860 | RREG32(R_008020_GRBM_SOFT_RESET); |
861 | mdelay(15); |
||
1221 | serge | 862 | WREG32(R_008020_GRBM_SOFT_RESET, 0); |
863 | /* Wait a little for things to settle down */ |
||
1963 | serge | 864 | mdelay(1); |
1221 | serge | 865 | dev_info(rdev->dev, " R_008010_GRBM_STATUS=0x%08X\n", |
866 | RREG32(R_008010_GRBM_STATUS)); |
||
867 | dev_info(rdev->dev, " R_008014_GRBM_STATUS2=0x%08X\n", |
||
868 | RREG32(R_008014_GRBM_STATUS2)); |
||
869 | dev_info(rdev->dev, " R_000E50_SRBM_STATUS=0x%08X\n", |
||
870 | RREG32(R_000E50_SRBM_STATUS)); |
||
871 | rv515_mc_resume(rdev, &save); |
||
872 | return 0; |
||
1128 | serge | 873 | } |
874 | |||
1963 | serge | 875 | bool r600_gpu_is_lockup(struct radeon_device *rdev) |
1221 | serge | 876 | { |
1963 | serge | 877 | u32 srbm_status; |
878 | u32 grbm_status; |
||
879 | u32 grbm_status2; |
||
880 | struct r100_gpu_lockup *lockup; |
||
881 | int r; |
||
882 | |||
883 | if (rdev->family >= CHIP_RV770) |
||
884 | lockup = &rdev->config.rv770.lockup; |
||
885 | else |
||
886 | lockup = &rdev->config.r600.lockup; |
||
887 | |||
888 | srbm_status = RREG32(R_000E50_SRBM_STATUS); |
||
889 | grbm_status = RREG32(R_008010_GRBM_STATUS); |
||
890 | grbm_status2 = RREG32(R_008014_GRBM_STATUS2); |
||
891 | if (!G_008010_GUI_ACTIVE(grbm_status)) { |
||
892 | r100_gpu_lockup_update(lockup, &rdev->cp); |
||
893 | return false; |
||
894 | } |
||
895 | /* force CP activities */ |
||
896 | r = radeon_ring_lock(rdev, 2); |
||
897 | if (!r) { |
||
898 | /* PACKET2 NOP */ |
||
899 | radeon_ring_write(rdev, 0x80000000); |
||
900 | radeon_ring_write(rdev, 0x80000000); |
||
901 | radeon_ring_unlock_commit(rdev); |
||
902 | } |
||
903 | rdev->cp.rptr = RREG32(R600_CP_RB_RPTR); |
||
904 | return r100_gpu_cp_is_lockup(rdev, lockup, &rdev->cp); |
||
905 | } |
||
906 | |||
907 | int r600_asic_reset(struct radeon_device *rdev) |
||
908 | { |
||
1221 | serge | 909 | return r600_gpu_soft_reset(rdev); |
910 | } |
||
911 | |||
912 | static u32 r600_get_tile_pipe_to_backend_map(u32 num_tile_pipes, |
||
913 | u32 num_backends, |
||
914 | u32 backend_disable_mask) |
||
915 | { |
||
916 | u32 backend_map = 0; |
||
917 | u32 enabled_backends_mask; |
||
918 | u32 enabled_backends_count; |
||
919 | u32 cur_pipe; |
||
920 | u32 swizzle_pipe[R6XX_MAX_PIPES]; |
||
921 | u32 cur_backend; |
||
922 | u32 i; |
||
923 | |||
924 | if (num_tile_pipes > R6XX_MAX_PIPES) |
||
925 | num_tile_pipes = R6XX_MAX_PIPES; |
||
926 | if (num_tile_pipes < 1) |
||
927 | num_tile_pipes = 1; |
||
928 | if (num_backends > R6XX_MAX_BACKENDS) |
||
929 | num_backends = R6XX_MAX_BACKENDS; |
||
930 | if (num_backends < 1) |
||
931 | num_backends = 1; |
||
932 | |||
933 | enabled_backends_mask = 0; |
||
934 | enabled_backends_count = 0; |
||
935 | for (i = 0; i < R6XX_MAX_BACKENDS; ++i) { |
||
936 | if (((backend_disable_mask >> i) & 1) == 0) { |
||
937 | enabled_backends_mask |= (1 << i); |
||
938 | ++enabled_backends_count; |
||
939 | } |
||
940 | if (enabled_backends_count == num_backends) |
||
941 | break; |
||
942 | } |
||
943 | |||
944 | if (enabled_backends_count == 0) { |
||
945 | enabled_backends_mask = 1; |
||
946 | enabled_backends_count = 1; |
||
947 | } |
||
948 | |||
949 | if (enabled_backends_count != num_backends) |
||
950 | num_backends = enabled_backends_count; |
||
951 | |||
952 | memset((uint8_t *)&swizzle_pipe[0], 0, sizeof(u32) * R6XX_MAX_PIPES); |
||
953 | switch (num_tile_pipes) { |
||
954 | case 1: |
||
955 | swizzle_pipe[0] = 0; |
||
956 | break; |
||
957 | case 2: |
||
958 | swizzle_pipe[0] = 0; |
||
959 | swizzle_pipe[1] = 1; |
||
960 | break; |
||
961 | case 3: |
||
962 | swizzle_pipe[0] = 0; |
||
963 | swizzle_pipe[1] = 1; |
||
964 | swizzle_pipe[2] = 2; |
||
965 | break; |
||
966 | case 4: |
||
967 | swizzle_pipe[0] = 0; |
||
968 | swizzle_pipe[1] = 1; |
||
969 | swizzle_pipe[2] = 2; |
||
970 | swizzle_pipe[3] = 3; |
||
971 | break; |
||
972 | case 5: |
||
973 | swizzle_pipe[0] = 0; |
||
974 | swizzle_pipe[1] = 1; |
||
975 | swizzle_pipe[2] = 2; |
||
976 | swizzle_pipe[3] = 3; |
||
977 | swizzle_pipe[4] = 4; |
||
978 | break; |
||
979 | case 6: |
||
980 | swizzle_pipe[0] = 0; |
||
981 | swizzle_pipe[1] = 2; |
||
982 | swizzle_pipe[2] = 4; |
||
983 | swizzle_pipe[3] = 5; |
||
984 | swizzle_pipe[4] = 1; |
||
985 | swizzle_pipe[5] = 3; |
||
986 | break; |
||
987 | case 7: |
||
988 | swizzle_pipe[0] = 0; |
||
989 | swizzle_pipe[1] = 2; |
||
990 | swizzle_pipe[2] = 4; |
||
991 | swizzle_pipe[3] = 6; |
||
992 | swizzle_pipe[4] = 1; |
||
993 | swizzle_pipe[5] = 3; |
||
994 | swizzle_pipe[6] = 5; |
||
995 | break; |
||
996 | case 8: |
||
997 | swizzle_pipe[0] = 0; |
||
998 | swizzle_pipe[1] = 2; |
||
999 | swizzle_pipe[2] = 4; |
||
1000 | swizzle_pipe[3] = 6; |
||
1001 | swizzle_pipe[4] = 1; |
||
1002 | swizzle_pipe[5] = 3; |
||
1003 | swizzle_pipe[6] = 5; |
||
1004 | swizzle_pipe[7] = 7; |
||
1005 | break; |
||
1006 | } |
||
1007 | |||
1008 | cur_backend = 0; |
||
1009 | for (cur_pipe = 0; cur_pipe < num_tile_pipes; ++cur_pipe) { |
||
1010 | while (((1 << cur_backend) & enabled_backends_mask) == 0) |
||
1011 | cur_backend = (cur_backend + 1) % R6XX_MAX_BACKENDS; |
||
1012 | |||
1013 | backend_map |= (u32)(((cur_backend & 3) << (swizzle_pipe[cur_pipe] * 2))); |
||
1014 | |||
1015 | cur_backend = (cur_backend + 1) % R6XX_MAX_BACKENDS; |
||
1016 | } |
||
1017 | |||
1018 | return backend_map; |
||
1019 | } |
||
1020 | |||
1021 | int r600_count_pipe_bits(uint32_t val) |
||
1022 | { |
||
1023 | int i, ret = 0; |
||
1024 | |||
1025 | for (i = 0; i < 32; i++) { |
||
1026 | ret += val & 1; |
||
1027 | val >>= 1; |
||
1028 | } |
||
1029 | return ret; |
||
1030 | } |
||
1031 | |||
1032 | void r600_gpu_init(struct radeon_device *rdev) |
||
1033 | { |
||
1034 | u32 tiling_config; |
||
1035 | u32 ramcfg; |
||
1430 | serge | 1036 | u32 backend_map; |
1037 | u32 cc_rb_backend_disable; |
||
1038 | u32 cc_gc_shader_pipe_config; |
||
1221 | serge | 1039 | u32 tmp; |
1040 | int i, j; |
||
1041 | u32 sq_config; |
||
1042 | u32 sq_gpr_resource_mgmt_1 = 0; |
||
1043 | u32 sq_gpr_resource_mgmt_2 = 0; |
||
1044 | u32 sq_thread_resource_mgmt = 0; |
||
1045 | u32 sq_stack_resource_mgmt_1 = 0; |
||
1046 | u32 sq_stack_resource_mgmt_2 = 0; |
||
1047 | |||
1048 | /* FIXME: implement */ |
||
1049 | switch (rdev->family) { |
||
1050 | case CHIP_R600: |
||
1051 | rdev->config.r600.max_pipes = 4; |
||
1052 | rdev->config.r600.max_tile_pipes = 8; |
||
1053 | rdev->config.r600.max_simds = 4; |
||
1054 | rdev->config.r600.max_backends = 4; |
||
1055 | rdev->config.r600.max_gprs = 256; |
||
1056 | rdev->config.r600.max_threads = 192; |
||
1057 | rdev->config.r600.max_stack_entries = 256; |
||
1058 | rdev->config.r600.max_hw_contexts = 8; |
||
1059 | rdev->config.r600.max_gs_threads = 16; |
||
1060 | rdev->config.r600.sx_max_export_size = 128; |
||
1061 | rdev->config.r600.sx_max_export_pos_size = 16; |
||
1062 | rdev->config.r600.sx_max_export_smx_size = 128; |
||
1063 | rdev->config.r600.sq_num_cf_insts = 2; |
||
1064 | break; |
||
1065 | case CHIP_RV630: |
||
1066 | case CHIP_RV635: |
||
1067 | rdev->config.r600.max_pipes = 2; |
||
1068 | rdev->config.r600.max_tile_pipes = 2; |
||
1069 | rdev->config.r600.max_simds = 3; |
||
1070 | rdev->config.r600.max_backends = 1; |
||
1071 | rdev->config.r600.max_gprs = 128; |
||
1072 | rdev->config.r600.max_threads = 192; |
||
1073 | rdev->config.r600.max_stack_entries = 128; |
||
1074 | rdev->config.r600.max_hw_contexts = 8; |
||
1075 | rdev->config.r600.max_gs_threads = 4; |
||
1076 | rdev->config.r600.sx_max_export_size = 128; |
||
1077 | rdev->config.r600.sx_max_export_pos_size = 16; |
||
1078 | rdev->config.r600.sx_max_export_smx_size = 128; |
||
1079 | rdev->config.r600.sq_num_cf_insts = 2; |
||
1080 | break; |
||
1081 | case CHIP_RV610: |
||
1082 | case CHIP_RV620: |
||
1083 | case CHIP_RS780: |
||
1084 | case CHIP_RS880: |
||
1085 | rdev->config.r600.max_pipes = 1; |
||
1086 | rdev->config.r600.max_tile_pipes = 1; |
||
1087 | rdev->config.r600.max_simds = 2; |
||
1088 | rdev->config.r600.max_backends = 1; |
||
1089 | rdev->config.r600.max_gprs = 128; |
||
1090 | rdev->config.r600.max_threads = 192; |
||
1091 | rdev->config.r600.max_stack_entries = 128; |
||
1092 | rdev->config.r600.max_hw_contexts = 4; |
||
1093 | rdev->config.r600.max_gs_threads = 4; |
||
1094 | rdev->config.r600.sx_max_export_size = 128; |
||
1095 | rdev->config.r600.sx_max_export_pos_size = 16; |
||
1096 | rdev->config.r600.sx_max_export_smx_size = 128; |
||
1097 | rdev->config.r600.sq_num_cf_insts = 1; |
||
1098 | break; |
||
1099 | case CHIP_RV670: |
||
1100 | rdev->config.r600.max_pipes = 4; |
||
1101 | rdev->config.r600.max_tile_pipes = 4; |
||
1102 | rdev->config.r600.max_simds = 4; |
||
1103 | rdev->config.r600.max_backends = 4; |
||
1104 | rdev->config.r600.max_gprs = 192; |
||
1105 | rdev->config.r600.max_threads = 192; |
||
1106 | rdev->config.r600.max_stack_entries = 256; |
||
1107 | rdev->config.r600.max_hw_contexts = 8; |
||
1108 | rdev->config.r600.max_gs_threads = 16; |
||
1109 | rdev->config.r600.sx_max_export_size = 128; |
||
1110 | rdev->config.r600.sx_max_export_pos_size = 16; |
||
1111 | rdev->config.r600.sx_max_export_smx_size = 128; |
||
1112 | rdev->config.r600.sq_num_cf_insts = 2; |
||
1113 | break; |
||
1114 | default: |
||
1115 | break; |
||
1116 | } |
||
1117 | |||
1118 | /* Initialize HDP */ |
||
1119 | for (i = 0, j = 0; i < 32; i++, j += 0x18) { |
||
1120 | WREG32((0x2c14 + j), 0x00000000); |
||
1121 | WREG32((0x2c18 + j), 0x00000000); |
||
1122 | WREG32((0x2c1c + j), 0x00000000); |
||
1123 | WREG32((0x2c20 + j), 0x00000000); |
||
1124 | WREG32((0x2c24 + j), 0x00000000); |
||
1125 | } |
||
1126 | |||
1127 | WREG32(GRBM_CNTL, GRBM_READ_TIMEOUT(0xff)); |
||
1128 | |||
1129 | /* Setup tiling */ |
||
1130 | tiling_config = 0; |
||
1131 | ramcfg = RREG32(RAMCFG); |
||
1132 | switch (rdev->config.r600.max_tile_pipes) { |
||
1133 | case 1: |
||
1134 | tiling_config |= PIPE_TILING(0); |
||
1135 | break; |
||
1136 | case 2: |
||
1137 | tiling_config |= PIPE_TILING(1); |
||
1138 | break; |
||
1139 | case 4: |
||
1140 | tiling_config |= PIPE_TILING(2); |
||
1141 | break; |
||
1142 | case 8: |
||
1143 | tiling_config |= PIPE_TILING(3); |
||
1144 | break; |
||
1145 | default: |
||
1146 | break; |
||
1147 | } |
||
1430 | serge | 1148 | rdev->config.r600.tiling_npipes = rdev->config.r600.max_tile_pipes; |
1149 | rdev->config.r600.tiling_nbanks = 4 << ((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT); |
||
1221 | serge | 1150 | tiling_config |= BANK_TILING((ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT); |
1963 | serge | 1151 | tiling_config |= GROUP_SIZE((ramcfg & BURSTLENGTH_MASK) >> BURSTLENGTH_SHIFT); |
1152 | if ((ramcfg & BURSTLENGTH_MASK) >> BURSTLENGTH_SHIFT) |
||
1153 | rdev->config.r600.tiling_group_size = 512; |
||
1154 | else |
||
1430 | serge | 1155 | rdev->config.r600.tiling_group_size = 256; |
1221 | serge | 1156 | tmp = (ramcfg & NOOFROWS_MASK) >> NOOFROWS_SHIFT; |
1157 | if (tmp > 3) { |
||
1158 | tiling_config |= ROW_TILING(3); |
||
1159 | tiling_config |= SAMPLE_SPLIT(3); |
||
1160 | } else { |
||
1161 | tiling_config |= ROW_TILING(tmp); |
||
1162 | tiling_config |= SAMPLE_SPLIT(tmp); |
||
1163 | } |
||
1164 | tiling_config |= BANK_SWAPS(1); |
||
1430 | serge | 1165 | |
1166 | cc_rb_backend_disable = RREG32(CC_RB_BACKEND_DISABLE) & 0x00ff0000; |
||
1167 | cc_rb_backend_disable |= |
||
1168 | BACKEND_DISABLE((R6XX_MAX_BACKENDS_MASK << rdev->config.r600.max_backends) & R6XX_MAX_BACKENDS_MASK); |
||
1169 | |||
1170 | cc_gc_shader_pipe_config = RREG32(CC_GC_SHADER_PIPE_CONFIG) & 0xffffff00; |
||
1171 | cc_gc_shader_pipe_config |= |
||
1172 | INACTIVE_QD_PIPES((R6XX_MAX_PIPES_MASK << rdev->config.r600.max_pipes) & R6XX_MAX_PIPES_MASK); |
||
1173 | cc_gc_shader_pipe_config |= |
||
1174 | INACTIVE_SIMDS((R6XX_MAX_SIMDS_MASK << rdev->config.r600.max_simds) & R6XX_MAX_SIMDS_MASK); |
||
1175 | |||
1176 | backend_map = r600_get_tile_pipe_to_backend_map(rdev->config.r600.max_tile_pipes, |
||
1177 | (R6XX_MAX_BACKENDS - |
||
1178 | r600_count_pipe_bits((cc_rb_backend_disable & |
||
1179 | R6XX_MAX_BACKENDS_MASK) >> 16)), |
||
1180 | (cc_rb_backend_disable >> 16)); |
||
1963 | serge | 1181 | rdev->config.r600.tile_config = tiling_config; |
1430 | serge | 1182 | tiling_config |= BACKEND_MAP(backend_map); |
1221 | serge | 1183 | WREG32(GB_TILING_CONFIG, tiling_config); |
1184 | WREG32(DCP_TILING_CONFIG, tiling_config & 0xffff); |
||
1185 | WREG32(HDP_TILING_CONFIG, tiling_config & 0xffff); |
||
1186 | |||
1187 | /* Setup pipes */ |
||
1430 | serge | 1188 | WREG32(CC_RB_BACKEND_DISABLE, cc_rb_backend_disable); |
1189 | WREG32(CC_GC_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config); |
||
1963 | serge | 1190 | WREG32(GC_USER_SHADER_PIPE_CONFIG, cc_gc_shader_pipe_config); |
1221 | serge | 1191 | |
1430 | serge | 1192 | tmp = R6XX_MAX_PIPES - r600_count_pipe_bits((cc_gc_shader_pipe_config & INACTIVE_QD_PIPES_MASK) >> 8); |
1221 | serge | 1193 | WREG32(VGT_OUT_DEALLOC_CNTL, (tmp * 4) & DEALLOC_DIST_MASK); |
1194 | WREG32(VGT_VERTEX_REUSE_BLOCK_CNTL, ((tmp * 4) - 2) & VTX_REUSE_DEPTH_MASK); |
||
1195 | |||
1196 | /* Setup some CP states */ |
||
1197 | WREG32(CP_QUEUE_THRESHOLDS, (ROQ_IB1_START(0x16) | ROQ_IB2_START(0x2b))); |
||
1198 | WREG32(CP_MEQ_THRESHOLDS, (MEQ_END(0x40) | ROQ_END(0x40))); |
||
1199 | |||
1200 | WREG32(TA_CNTL_AUX, (DISABLE_CUBE_ANISO | SYNC_GRADIENT | |
||
1201 | SYNC_WALKER | SYNC_ALIGNER)); |
||
1202 | /* Setup various GPU states */ |
||
1203 | if (rdev->family == CHIP_RV670) |
||
1204 | WREG32(ARB_GDEC_RD_CNTL, 0x00000021); |
||
1205 | |||
1206 | tmp = RREG32(SX_DEBUG_1); |
||
1207 | tmp |= SMX_EVENT_RELEASE; |
||
1208 | if ((rdev->family > CHIP_R600)) |
||
1209 | tmp |= ENABLE_NEW_SMX_ADDRESS; |
||
1210 | WREG32(SX_DEBUG_1, tmp); |
||
1211 | |||
1212 | if (((rdev->family) == CHIP_R600) || |
||
1213 | ((rdev->family) == CHIP_RV630) || |
||
1214 | ((rdev->family) == CHIP_RV610) || |
||
1215 | ((rdev->family) == CHIP_RV620) || |
||
1268 | serge | 1216 | ((rdev->family) == CHIP_RS780) || |
1217 | ((rdev->family) == CHIP_RS880)) { |
||
1221 | serge | 1218 | WREG32(DB_DEBUG, PREZ_MUST_WAIT_FOR_POSTZ_DONE); |
1219 | } else { |
||
1220 | WREG32(DB_DEBUG, 0); |
||
1221 | } |
||
1222 | WREG32(DB_WATERMARKS, (DEPTH_FREE(4) | DEPTH_CACHELINE_FREE(16) | |
||
1223 | DEPTH_FLUSH(16) | DEPTH_PENDING_FREE(4))); |
||
1224 | |||
1225 | WREG32(PA_SC_MULTI_CHIP_CNTL, 0); |
||
1226 | WREG32(VGT_NUM_INSTANCES, 0); |
||
1227 | |||
1228 | WREG32(SPI_CONFIG_CNTL, GPR_WRITE_PRIORITY(0)); |
||
1229 | WREG32(SPI_CONFIG_CNTL_1, VTX_DONE_DELAY(0)); |
||
1230 | |||
1231 | tmp = RREG32(SQ_MS_FIFO_SIZES); |
||
1232 | if (((rdev->family) == CHIP_RV610) || |
||
1233 | ((rdev->family) == CHIP_RV620) || |
||
1268 | serge | 1234 | ((rdev->family) == CHIP_RS780) || |
1235 | ((rdev->family) == CHIP_RS880)) { |
||
1221 | serge | 1236 | tmp = (CACHE_FIFO_SIZE(0xa) | |
1237 | FETCH_FIFO_HIWATER(0xa) | |
||
1238 | DONE_FIFO_HIWATER(0xe0) | |
||
1239 | ALU_UPDATE_FIFO_HIWATER(0x8)); |
||
1240 | } else if (((rdev->family) == CHIP_R600) || |
||
1241 | ((rdev->family) == CHIP_RV630)) { |
||
1242 | tmp &= ~DONE_FIFO_HIWATER(0xff); |
||
1243 | tmp |= DONE_FIFO_HIWATER(0x4); |
||
1244 | } |
||
1245 | WREG32(SQ_MS_FIFO_SIZES, tmp); |
||
1246 | |||
1247 | /* SQ_CONFIG, SQ_GPR_RESOURCE_MGMT, SQ_THREAD_RESOURCE_MGMT, SQ_STACK_RESOURCE_MGMT |
||
1248 | * should be adjusted as needed by the 2D/3D drivers. This just sets default values |
||
1249 | */ |
||
1250 | sq_config = RREG32(SQ_CONFIG); |
||
1251 | sq_config &= ~(PS_PRIO(3) | |
||
1252 | VS_PRIO(3) | |
||
1253 | GS_PRIO(3) | |
||
1254 | ES_PRIO(3)); |
||
1255 | sq_config |= (DX9_CONSTS | |
||
1256 | VC_ENABLE | |
||
1257 | PS_PRIO(0) | |
||
1258 | VS_PRIO(1) | |
||
1259 | GS_PRIO(2) | |
||
1260 | ES_PRIO(3)); |
||
1261 | |||
1262 | if ((rdev->family) == CHIP_R600) { |
||
1263 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(124) | |
||
1264 | NUM_VS_GPRS(124) | |
||
1265 | NUM_CLAUSE_TEMP_GPRS(4)); |
||
1266 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(0) | |
||
1267 | NUM_ES_GPRS(0)); |
||
1268 | sq_thread_resource_mgmt = (NUM_PS_THREADS(136) | |
||
1269 | NUM_VS_THREADS(48) | |
||
1270 | NUM_GS_THREADS(4) | |
||
1271 | NUM_ES_THREADS(4)); |
||
1272 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(128) | |
||
1273 | NUM_VS_STACK_ENTRIES(128)); |
||
1274 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(0) | |
||
1275 | NUM_ES_STACK_ENTRIES(0)); |
||
1276 | } else if (((rdev->family) == CHIP_RV610) || |
||
1277 | ((rdev->family) == CHIP_RV620) || |
||
1268 | serge | 1278 | ((rdev->family) == CHIP_RS780) || |
1279 | ((rdev->family) == CHIP_RS880)) { |
||
1221 | serge | 1280 | /* no vertex cache */ |
1281 | sq_config &= ~VC_ENABLE; |
||
1282 | |||
1283 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44) | |
||
1284 | NUM_VS_GPRS(44) | |
||
1285 | NUM_CLAUSE_TEMP_GPRS(2)); |
||
1286 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(17) | |
||
1287 | NUM_ES_GPRS(17)); |
||
1288 | sq_thread_resource_mgmt = (NUM_PS_THREADS(79) | |
||
1289 | NUM_VS_THREADS(78) | |
||
1290 | NUM_GS_THREADS(4) | |
||
1291 | NUM_ES_THREADS(31)); |
||
1292 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(40) | |
||
1293 | NUM_VS_STACK_ENTRIES(40)); |
||
1294 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(32) | |
||
1295 | NUM_ES_STACK_ENTRIES(16)); |
||
1296 | } else if (((rdev->family) == CHIP_RV630) || |
||
1297 | ((rdev->family) == CHIP_RV635)) { |
||
1298 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44) | |
||
1299 | NUM_VS_GPRS(44) | |
||
1300 | NUM_CLAUSE_TEMP_GPRS(2)); |
||
1301 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(18) | |
||
1302 | NUM_ES_GPRS(18)); |
||
1303 | sq_thread_resource_mgmt = (NUM_PS_THREADS(79) | |
||
1304 | NUM_VS_THREADS(78) | |
||
1305 | NUM_GS_THREADS(4) | |
||
1306 | NUM_ES_THREADS(31)); |
||
1307 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(40) | |
||
1308 | NUM_VS_STACK_ENTRIES(40)); |
||
1309 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(32) | |
||
1310 | NUM_ES_STACK_ENTRIES(16)); |
||
1311 | } else if ((rdev->family) == CHIP_RV670) { |
||
1312 | sq_gpr_resource_mgmt_1 = (NUM_PS_GPRS(44) | |
||
1313 | NUM_VS_GPRS(44) | |
||
1314 | NUM_CLAUSE_TEMP_GPRS(2)); |
||
1315 | sq_gpr_resource_mgmt_2 = (NUM_GS_GPRS(17) | |
||
1316 | NUM_ES_GPRS(17)); |
||
1317 | sq_thread_resource_mgmt = (NUM_PS_THREADS(79) | |
||
1318 | NUM_VS_THREADS(78) | |
||
1319 | NUM_GS_THREADS(4) | |
||
1320 | NUM_ES_THREADS(31)); |
||
1321 | sq_stack_resource_mgmt_1 = (NUM_PS_STACK_ENTRIES(64) | |
||
1322 | NUM_VS_STACK_ENTRIES(64)); |
||
1323 | sq_stack_resource_mgmt_2 = (NUM_GS_STACK_ENTRIES(64) | |
||
1324 | NUM_ES_STACK_ENTRIES(64)); |
||
1325 | } |
||
1326 | |||
1327 | WREG32(SQ_CONFIG, sq_config); |
||
1328 | WREG32(SQ_GPR_RESOURCE_MGMT_1, sq_gpr_resource_mgmt_1); |
||
1329 | WREG32(SQ_GPR_RESOURCE_MGMT_2, sq_gpr_resource_mgmt_2); |
||
1330 | WREG32(SQ_THREAD_RESOURCE_MGMT, sq_thread_resource_mgmt); |
||
1331 | WREG32(SQ_STACK_RESOURCE_MGMT_1, sq_stack_resource_mgmt_1); |
||
1332 | WREG32(SQ_STACK_RESOURCE_MGMT_2, sq_stack_resource_mgmt_2); |
||
1333 | |||
1334 | if (((rdev->family) == CHIP_RV610) || |
||
1335 | ((rdev->family) == CHIP_RV620) || |
||
1268 | serge | 1336 | ((rdev->family) == CHIP_RS780) || |
1337 | ((rdev->family) == CHIP_RS880)) { |
||
1221 | serge | 1338 | WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(TC_ONLY)); |
1339 | } else { |
||
1340 | WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC)); |
||
1341 | } |
||
1342 | |||
1343 | /* More default values. 2D/3D driver should adjust as needed */ |
||
1344 | WREG32(PA_SC_AA_SAMPLE_LOCS_2S, (S0_X(0xc) | S0_Y(0x4) | |
||
1345 | S1_X(0x4) | S1_Y(0xc))); |
||
1346 | WREG32(PA_SC_AA_SAMPLE_LOCS_4S, (S0_X(0xe) | S0_Y(0xe) | |
||
1347 | S1_X(0x2) | S1_Y(0x2) | |
||
1348 | S2_X(0xa) | S2_Y(0x6) | |
||
1349 | S3_X(0x6) | S3_Y(0xa))); |
||
1350 | WREG32(PA_SC_AA_SAMPLE_LOCS_8S_WD0, (S0_X(0xe) | S0_Y(0xb) | |
||
1351 | S1_X(0x4) | S1_Y(0xc) | |
||
1352 | S2_X(0x1) | S2_Y(0x6) | |
||
1353 | S3_X(0xa) | S3_Y(0xe))); |
||
1354 | WREG32(PA_SC_AA_SAMPLE_LOCS_8S_WD1, (S4_X(0x6) | S4_Y(0x1) | |
||
1355 | S5_X(0x0) | S5_Y(0x0) | |
||
1356 | S6_X(0xb) | S6_Y(0x4) | |
||
1357 | S7_X(0x7) | S7_Y(0x8))); |
||
1358 | |||
1359 | WREG32(VGT_STRMOUT_EN, 0); |
||
1360 | tmp = rdev->config.r600.max_pipes * 16; |
||
1361 | switch (rdev->family) { |
||
1362 | case CHIP_RV610: |
||
1268 | serge | 1363 | case CHIP_RV620: |
1221 | serge | 1364 | case CHIP_RS780: |
1268 | serge | 1365 | case CHIP_RS880: |
1221 | serge | 1366 | tmp += 32; |
1367 | break; |
||
1368 | case CHIP_RV670: |
||
1369 | tmp += 128; |
||
1370 | break; |
||
1371 | default: |
||
1372 | break; |
||
1373 | } |
||
1374 | if (tmp > 256) { |
||
1375 | tmp = 256; |
||
1376 | } |
||
1377 | WREG32(VGT_ES_PER_GS, 128); |
||
1378 | WREG32(VGT_GS_PER_ES, tmp); |
||
1379 | WREG32(VGT_GS_PER_VS, 2); |
||
1380 | WREG32(VGT_GS_VERTEX_REUSE, 16); |
||
1381 | |||
1382 | /* more default values. 2D/3D driver should adjust as needed */ |
||
1383 | WREG32(PA_SC_LINE_STIPPLE_STATE, 0); |
||
1384 | WREG32(VGT_STRMOUT_EN, 0); |
||
1385 | WREG32(SX_MISC, 0); |
||
1386 | WREG32(PA_SC_MODE_CNTL, 0); |
||
1387 | WREG32(PA_SC_AA_CONFIG, 0); |
||
1388 | WREG32(PA_SC_LINE_STIPPLE, 0); |
||
1389 | WREG32(SPI_INPUT_Z, 0); |
||
1390 | WREG32(SPI_PS_IN_CONTROL_0, NUM_INTERP(2)); |
||
1391 | WREG32(CB_COLOR7_FRAG, 0); |
||
1392 | |||
1393 | /* Clear render buffer base addresses */ |
||
1394 | WREG32(CB_COLOR0_BASE, 0); |
||
1395 | WREG32(CB_COLOR1_BASE, 0); |
||
1396 | WREG32(CB_COLOR2_BASE, 0); |
||
1397 | WREG32(CB_COLOR3_BASE, 0); |
||
1398 | WREG32(CB_COLOR4_BASE, 0); |
||
1399 | WREG32(CB_COLOR5_BASE, 0); |
||
1400 | WREG32(CB_COLOR6_BASE, 0); |
||
1401 | WREG32(CB_COLOR7_BASE, 0); |
||
1402 | WREG32(CB_COLOR7_FRAG, 0); |
||
1403 | |||
1404 | switch (rdev->family) { |
||
1405 | case CHIP_RV610: |
||
1268 | serge | 1406 | case CHIP_RV620: |
1221 | serge | 1407 | case CHIP_RS780: |
1268 | serge | 1408 | case CHIP_RS880: |
1221 | serge | 1409 | tmp = TC_L2_SIZE(8); |
1410 | break; |
||
1411 | case CHIP_RV630: |
||
1412 | case CHIP_RV635: |
||
1413 | tmp = TC_L2_SIZE(4); |
||
1414 | break; |
||
1415 | case CHIP_R600: |
||
1416 | tmp = TC_L2_SIZE(0) | L2_DISABLE_LATE_HIT; |
||
1417 | break; |
||
1418 | default: |
||
1419 | tmp = TC_L2_SIZE(0); |
||
1420 | break; |
||
1421 | } |
||
1422 | WREG32(TC_CNTL, tmp); |
||
1423 | |||
1424 | tmp = RREG32(HDP_HOST_PATH_CNTL); |
||
1425 | WREG32(HDP_HOST_PATH_CNTL, tmp); |
||
1426 | |||
1427 | tmp = RREG32(ARB_POP); |
||
1428 | tmp |= ENABLE_TC128; |
||
1429 | WREG32(ARB_POP, tmp); |
||
1430 | |||
1431 | WREG32(PA_SC_MULTI_CHIP_CNTL, 0); |
||
1432 | WREG32(PA_CL_ENHANCE, (CLIP_VTX_REORDER_ENA | |
||
1433 | NUM_CLIP_SEQ(3))); |
||
1434 | WREG32(PA_SC_ENHANCE, FORCE_EOV_MAX_CLK_CNT(4095)); |
||
1435 | } |
||
1436 | |||
1437 | |||
1128 | serge | 1438 | /* |
1439 | * Indirect registers accessor |
||
1440 | */ |
||
1221 | serge | 1441 | u32 r600_pciep_rreg(struct radeon_device *rdev, u32 reg) |
1128 | serge | 1442 | { |
1221 | serge | 1443 | u32 r; |
1128 | serge | 1444 | |
1221 | serge | 1445 | WREG32(PCIE_PORT_INDEX, ((reg) & 0xff)); |
1446 | (void)RREG32(PCIE_PORT_INDEX); |
||
1447 | r = RREG32(PCIE_PORT_DATA); |
||
1128 | serge | 1448 | return r; |
1449 | } |
||
1450 | |||
1221 | serge | 1451 | void r600_pciep_wreg(struct radeon_device *rdev, u32 reg, u32 v) |
1128 | serge | 1452 | { |
1221 | serge | 1453 | WREG32(PCIE_PORT_INDEX, ((reg) & 0xff)); |
1454 | (void)RREG32(PCIE_PORT_INDEX); |
||
1455 | WREG32(PCIE_PORT_DATA, (v)); |
||
1456 | (void)RREG32(PCIE_PORT_DATA); |
||
1128 | serge | 1457 | } |
1221 | serge | 1458 | |
1459 | /* |
||
1460 | * CP & Ring |
||
1461 | */ |
||
1462 | void r600_cp_stop(struct radeon_device *rdev) |
||
1463 | { |
||
1963 | serge | 1464 | // radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); |
1221 | serge | 1465 | WREG32(R_0086D8_CP_ME_CNTL, S_0086D8_CP_ME_HALT(1)); |
1963 | serge | 1466 | WREG32(SCRATCH_UMSK, 0); |
1221 | serge | 1467 | } |
1413 | serge | 1468 | |
1469 | int r600_init_microcode(struct radeon_device *rdev) |
||
1470 | { |
||
1471 | struct platform_device *pdev; |
||
1472 | const char *chip_name; |
||
1473 | const char *rlc_chip_name; |
||
1474 | size_t pfp_req_size, me_req_size, rlc_req_size; |
||
1475 | char fw_name[30]; |
||
1476 | int err; |
||
1477 | |||
1478 | DRM_DEBUG("\n"); |
||
1479 | |||
1480 | pdev = platform_device_register_simple("radeon_cp", 0, NULL, 0); |
||
1481 | err = IS_ERR(pdev); |
||
1482 | if (err) { |
||
1483 | printk(KERN_ERR "radeon_cp: Failed to register firmware\n"); |
||
1484 | return -EINVAL; |
||
1485 | } |
||
1486 | |||
1487 | switch (rdev->family) { |
||
1488 | case CHIP_R600: |
||
1489 | chip_name = "R600"; |
||
1490 | rlc_chip_name = "R600"; |
||
1491 | break; |
||
1492 | case CHIP_RV610: |
||
1493 | chip_name = "RV610"; |
||
1494 | rlc_chip_name = "R600"; |
||
1495 | break; |
||
1496 | case CHIP_RV630: |
||
1497 | chip_name = "RV630"; |
||
1498 | rlc_chip_name = "R600"; |
||
1499 | break; |
||
1500 | case CHIP_RV620: |
||
1501 | chip_name = "RV620"; |
||
1502 | rlc_chip_name = "R600"; |
||
1503 | break; |
||
1504 | case CHIP_RV635: |
||
1505 | chip_name = "RV635"; |
||
1506 | rlc_chip_name = "R600"; |
||
1507 | break; |
||
1508 | case CHIP_RV670: |
||
1509 | chip_name = "RV670"; |
||
1510 | rlc_chip_name = "R600"; |
||
1511 | break; |
||
1512 | case CHIP_RS780: |
||
1513 | case CHIP_RS880: |
||
1514 | chip_name = "RS780"; |
||
1515 | rlc_chip_name = "R600"; |
||
1516 | break; |
||
1517 | case CHIP_RV770: |
||
1518 | chip_name = "RV770"; |
||
1519 | rlc_chip_name = "R700"; |
||
1520 | break; |
||
1521 | case CHIP_RV730: |
||
1522 | case CHIP_RV740: |
||
1523 | chip_name = "RV730"; |
||
1524 | rlc_chip_name = "R700"; |
||
1525 | break; |
||
1526 | case CHIP_RV710: |
||
1527 | chip_name = "RV710"; |
||
1528 | rlc_chip_name = "R700"; |
||
1529 | break; |
||
1963 | serge | 1530 | case CHIP_CEDAR: |
1531 | chip_name = "CEDAR"; |
||
1532 | rlc_chip_name = "CEDAR"; |
||
1533 | break; |
||
1534 | case CHIP_REDWOOD: |
||
1535 | chip_name = "REDWOOD"; |
||
1536 | rlc_chip_name = "REDWOOD"; |
||
1537 | break; |
||
1538 | case CHIP_JUNIPER: |
||
1539 | chip_name = "JUNIPER"; |
||
1540 | rlc_chip_name = "JUNIPER"; |
||
1541 | break; |
||
1542 | case CHIP_CYPRESS: |
||
1543 | case CHIP_HEMLOCK: |
||
1544 | chip_name = "CYPRESS"; |
||
1545 | rlc_chip_name = "CYPRESS"; |
||
1546 | break; |
||
1547 | case CHIP_PALM: |
||
1548 | chip_name = "PALM"; |
||
1549 | rlc_chip_name = "SUMO"; |
||
1550 | break; |
||
1413 | serge | 1551 | default: BUG(); |
1552 | } |
||
1553 | |||
1963 | serge | 1554 | if (rdev->family >= CHIP_CEDAR) { |
1555 | pfp_req_size = EVERGREEN_PFP_UCODE_SIZE * 4; |
||
1556 | me_req_size = EVERGREEN_PM4_UCODE_SIZE * 4; |
||
1557 | rlc_req_size = EVERGREEN_RLC_UCODE_SIZE * 4; |
||
1558 | } else if (rdev->family >= CHIP_RV770) { |
||
1413 | serge | 1559 | pfp_req_size = R700_PFP_UCODE_SIZE * 4; |
1560 | me_req_size = R700_PM4_UCODE_SIZE * 4; |
||
1561 | rlc_req_size = R700_RLC_UCODE_SIZE * 4; |
||
1562 | } else { |
||
1563 | pfp_req_size = PFP_UCODE_SIZE * 4; |
||
1564 | me_req_size = PM4_UCODE_SIZE * 12; |
||
1565 | rlc_req_size = RLC_UCODE_SIZE * 4; |
||
1566 | } |
||
1567 | |||
1568 | DRM_INFO("Loading %s Microcode\n", chip_name); |
||
1569 | |||
1570 | snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name); |
||
1571 | err = request_firmware(&rdev->pfp_fw, fw_name, &pdev->dev); |
||
1572 | if (err) |
||
1573 | goto out; |
||
1574 | if (rdev->pfp_fw->size != pfp_req_size) { |
||
1575 | printk(KERN_ERR |
||
1576 | "r600_cp: Bogus length %zu in firmware \"%s\"\n", |
||
1577 | rdev->pfp_fw->size, fw_name); |
||
1578 | err = -EINVAL; |
||
1579 | goto out; |
||
1580 | } |
||
1581 | |||
1582 | snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", chip_name); |
||
1583 | err = request_firmware(&rdev->me_fw, fw_name, &pdev->dev); |
||
1584 | if (err) |
||
1585 | goto out; |
||
1586 | if (rdev->me_fw->size != me_req_size) { |
||
1587 | printk(KERN_ERR |
||
1588 | "r600_cp: Bogus length %zu in firmware \"%s\"\n", |
||
1589 | rdev->me_fw->size, fw_name); |
||
1590 | err = -EINVAL; |
||
1591 | } |
||
1592 | |||
1593 | snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", rlc_chip_name); |
||
1594 | err = request_firmware(&rdev->rlc_fw, fw_name, &pdev->dev); |
||
1595 | if (err) |
||
1596 | goto out; |
||
1597 | if (rdev->rlc_fw->size != rlc_req_size) { |
||
1598 | printk(KERN_ERR |
||
1599 | "r600_rlc: Bogus length %zu in firmware \"%s\"\n", |
||
1600 | rdev->rlc_fw->size, fw_name); |
||
1601 | err = -EINVAL; |
||
1602 | } |
||
1603 | |||
1604 | out: |
||
1605 | platform_device_unregister(pdev); |
||
1606 | |||
1607 | if (err) { |
||
1608 | if (err != -EINVAL) |
||
1609 | printk(KERN_ERR |
||
1610 | "r600_cp: Failed to load firmware \"%s\"\n", |
||
1611 | fw_name); |
||
1612 | release_firmware(rdev->pfp_fw); |
||
1613 | rdev->pfp_fw = NULL; |
||
1614 | release_firmware(rdev->me_fw); |
||
1615 | rdev->me_fw = NULL; |
||
1616 | release_firmware(rdev->rlc_fw); |
||
1617 | rdev->rlc_fw = NULL; |
||
1618 | } |
||
1619 | return err; |
||
1620 | } |
||
1621 | |||
1622 | static int r600_cp_load_microcode(struct radeon_device *rdev) |
||
1623 | { |
||
1624 | const __be32 *fw_data; |
||
1625 | int i; |
||
1626 | |||
1627 | if (!rdev->me_fw || !rdev->pfp_fw) |
||
1628 | return -EINVAL; |
||
1629 | |||
1630 | r600_cp_stop(rdev); |
||
1631 | |||
1963 | serge | 1632 | WREG32(CP_RB_CNTL, |
1633 | #ifdef __BIG_ENDIAN |
||
1634 | BUF_SWAP_32BIT | |
||
1635 | #endif |
||
1636 | RB_NO_UPDATE | RB_BLKSZ(15) | RB_BUFSZ(3)); |
||
1413 | serge | 1637 | |
1638 | /* Reset cp */ |
||
1639 | WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP); |
||
1640 | RREG32(GRBM_SOFT_RESET); |
||
1641 | mdelay(15); |
||
1642 | WREG32(GRBM_SOFT_RESET, 0); |
||
1643 | |||
1644 | WREG32(CP_ME_RAM_WADDR, 0); |
||
1645 | |||
1646 | fw_data = (const __be32 *)rdev->me_fw->data; |
||
1647 | WREG32(CP_ME_RAM_WADDR, 0); |
||
1648 | for (i = 0; i < PM4_UCODE_SIZE * 3; i++) |
||
1649 | WREG32(CP_ME_RAM_DATA, |
||
1650 | be32_to_cpup(fw_data++)); |
||
1651 | |||
1652 | fw_data = (const __be32 *)rdev->pfp_fw->data; |
||
1653 | WREG32(CP_PFP_UCODE_ADDR, 0); |
||
1654 | for (i = 0; i < PFP_UCODE_SIZE; i++) |
||
1655 | WREG32(CP_PFP_UCODE_DATA, |
||
1656 | be32_to_cpup(fw_data++)); |
||
1657 | |||
1658 | WREG32(CP_PFP_UCODE_ADDR, 0); |
||
1659 | WREG32(CP_ME_RAM_WADDR, 0); |
||
1660 | WREG32(CP_ME_RAM_RADDR, 0); |
||
1661 | return 0; |
||
1662 | } |
||
1663 | |||
1221 | serge | 1664 | int r600_cp_start(struct radeon_device *rdev) |
1665 | { |
||
1666 | int r; |
||
1667 | uint32_t cp_me; |
||
1668 | |||
1669 | r = radeon_ring_lock(rdev, 7); |
||
1670 | if (r) { |
||
1671 | DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); |
||
1672 | return r; |
||
1673 | } |
||
1674 | radeon_ring_write(rdev, PACKET3(PACKET3_ME_INITIALIZE, 5)); |
||
1675 | radeon_ring_write(rdev, 0x1); |
||
1963 | serge | 1676 | if (rdev->family >= CHIP_RV770) { |
1677 | radeon_ring_write(rdev, 0x0); |
||
1678 | radeon_ring_write(rdev, rdev->config.rv770.max_hw_contexts - 1); |
||
1679 | } else { |
||
1221 | serge | 1680 | radeon_ring_write(rdev, 0x3); |
1681 | radeon_ring_write(rdev, rdev->config.r600.max_hw_contexts - 1); |
||
1682 | } |
||
1683 | radeon_ring_write(rdev, PACKET3_ME_INITIALIZE_DEVICE_ID(1)); |
||
1684 | radeon_ring_write(rdev, 0); |
||
1685 | radeon_ring_write(rdev, 0); |
||
1686 | radeon_ring_unlock_commit(rdev); |
||
1687 | |||
1688 | cp_me = 0xff; |
||
1689 | WREG32(R_0086D8_CP_ME_CNTL, cp_me); |
||
1690 | return 0; |
||
1691 | } |
||
1413 | serge | 1692 | |
1693 | int r600_cp_resume(struct radeon_device *rdev) |
||
1694 | { |
||
1695 | u32 tmp; |
||
1696 | u32 rb_bufsz; |
||
1697 | int r; |
||
1698 | |||
1699 | /* Reset cp */ |
||
1700 | WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP); |
||
1701 | RREG32(GRBM_SOFT_RESET); |
||
1702 | mdelay(15); |
||
1703 | WREG32(GRBM_SOFT_RESET, 0); |
||
1704 | |||
1705 | /* Set ring buffer size */ |
||
1706 | rb_bufsz = drm_order(rdev->cp.ring_size / 8); |
||
1963 | serge | 1707 | tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz; |
1413 | serge | 1708 | #ifdef __BIG_ENDIAN |
1709 | tmp |= BUF_SWAP_32BIT; |
||
1710 | #endif |
||
1711 | WREG32(CP_RB_CNTL, tmp); |
||
1712 | WREG32(CP_SEM_WAIT_TIMER, 0x4); |
||
1713 | |||
1714 | /* Set the write pointer delay */ |
||
1715 | WREG32(CP_RB_WPTR_DELAY, 0); |
||
1716 | |||
1717 | /* Initialize the ring buffer's read and write pointers */ |
||
1718 | WREG32(CP_RB_CNTL, tmp | RB_RPTR_WR_ENA); |
||
1719 | WREG32(CP_RB_RPTR_WR, 0); |
||
1720 | WREG32(CP_RB_WPTR, 0); |
||
1963 | serge | 1721 | |
1722 | /* set the wb address whether it's enabled or not */ |
||
1723 | WREG32(CP_RB_RPTR_ADDR, |
||
1724 | #ifdef __BIG_ENDIAN |
||
1725 | RB_RPTR_SWAP(2) | |
||
1726 | #endif |
||
1727 | ((rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC)); |
||
1728 | WREG32(CP_RB_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF); |
||
1729 | WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF); |
||
1730 | |||
1731 | if (rdev->wb.enabled) |
||
1732 | WREG32(SCRATCH_UMSK, 0xff); |
||
1733 | else { |
||
1734 | tmp |= RB_NO_UPDATE; |
||
1735 | WREG32(SCRATCH_UMSK, 0); |
||
1736 | } |
||
1737 | |||
1413 | serge | 1738 | mdelay(1); |
1739 | WREG32(CP_RB_CNTL, tmp); |
||
1740 | |||
1741 | WREG32(CP_RB_BASE, rdev->cp.gpu_addr >> 8); |
||
1742 | WREG32(CP_DEBUG, (1 << 27) | (1 << 28)); |
||
1743 | |||
1744 | rdev->cp.rptr = RREG32(CP_RB_RPTR); |
||
1745 | rdev->cp.wptr = RREG32(CP_RB_WPTR); |
||
1746 | |||
1747 | r600_cp_start(rdev); |
||
1748 | rdev->cp.ready = true; |
||
1749 | r = radeon_ring_test(rdev); |
||
1750 | if (r) { |
||
1751 | rdev->cp.ready = false; |
||
1752 | return r; |
||
1753 | } |
||
1754 | return 0; |
||
1755 | } |
||
1756 | |||
1221 | serge | 1757 | void r600_cp_commit(struct radeon_device *rdev) |
1758 | { |
||
1759 | WREG32(CP_RB_WPTR, rdev->cp.wptr); |
||
1760 | (void)RREG32(CP_RB_WPTR); |
||
1761 | } |
||
1762 | |||
1233 | serge | 1763 | void r600_ring_init(struct radeon_device *rdev, unsigned ring_size) |
1764 | { |
||
1765 | u32 rb_bufsz; |
||
1221 | serge | 1766 | |
1233 | serge | 1767 | /* Align ring size */ |
1768 | rb_bufsz = drm_order(ring_size / 8); |
||
1769 | ring_size = (1 << (rb_bufsz + 1)) * 4; |
||
1770 | rdev->cp.ring_size = ring_size; |
||
1771 | rdev->cp.align_mask = 16 - 1; |
||
1772 | } |
||
1773 | |||
1963 | serge | 1774 | void r600_cp_fini(struct radeon_device *rdev) |
1775 | { |
||
1776 | r600_cp_stop(rdev); |
||
1777 | radeon_ring_fini(rdev); |
||
1778 | } |
||
1233 | serge | 1779 | |
1963 | serge | 1780 | |
1233 | serge | 1781 | /* |
1782 | * GPU scratch registers helpers function. |
||
1783 | */ |
||
1784 | void r600_scratch_init(struct radeon_device *rdev) |
||
1785 | { |
||
1786 | int i; |
||
1787 | |||
1788 | rdev->scratch.num_reg = 7; |
||
1963 | serge | 1789 | rdev->scratch.reg_base = SCRATCH_REG0; |
1233 | serge | 1790 | for (i = 0; i < rdev->scratch.num_reg; i++) { |
1791 | rdev->scratch.free[i] = true; |
||
1963 | serge | 1792 | rdev->scratch.reg[i] = rdev->scratch.reg_base + (i * 4); |
1233 | serge | 1793 | } |
1794 | } |
||
1413 | serge | 1795 | |
1796 | int r600_ring_test(struct radeon_device *rdev) |
||
1797 | { |
||
1798 | uint32_t scratch; |
||
1799 | uint32_t tmp = 0; |
||
1800 | unsigned i; |
||
1801 | int r; |
||
1802 | |||
1803 | r = radeon_scratch_get(rdev, &scratch); |
||
1804 | if (r) { |
||
1805 | DRM_ERROR("radeon: cp failed to get scratch reg (%d).\n", r); |
||
1806 | return r; |
||
1807 | } |
||
1808 | WREG32(scratch, 0xCAFEDEAD); |
||
1809 | r = radeon_ring_lock(rdev, 3); |
||
1810 | if (r) { |
||
1811 | DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); |
||
1812 | radeon_scratch_free(rdev, scratch); |
||
1813 | return r; |
||
1814 | } |
||
1815 | radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1)); |
||
1816 | radeon_ring_write(rdev, ((scratch - PACKET3_SET_CONFIG_REG_OFFSET) >> 2)); |
||
1817 | radeon_ring_write(rdev, 0xDEADBEEF); |
||
1818 | radeon_ring_unlock_commit(rdev); |
||
1819 | for (i = 0; i < rdev->usec_timeout; i++) { |
||
1820 | tmp = RREG32(scratch); |
||
1821 | if (tmp == 0xDEADBEEF) |
||
1822 | break; |
||
1823 | DRM_UDELAY(1); |
||
1824 | } |
||
1825 | if (i < rdev->usec_timeout) { |
||
1826 | DRM_INFO("ring test succeeded in %d usecs\n", i); |
||
1827 | } else { |
||
1828 | DRM_ERROR("radeon: ring test failed (scratch(0x%04X)=0x%08X)\n", |
||
1829 | scratch, tmp); |
||
1830 | r = -EINVAL; |
||
1831 | } |
||
1832 | radeon_scratch_free(rdev, scratch); |
||
1833 | return r; |
||
1834 | } |
||
1963 | serge | 1835 | |
1413 | serge | 1836 | void r600_fence_ring_emit(struct radeon_device *rdev, |
1837 | struct radeon_fence *fence) |
||
1838 | { |
||
1963 | serge | 1839 | if (rdev->wb.use_event) { |
1840 | u64 addr = rdev->wb.gpu_addr + R600_WB_EVENT_OFFSET + |
||
1841 | (u64)(rdev->fence_drv.scratch_reg - rdev->scratch.reg_base); |
||
1842 | /* EVENT_WRITE_EOP - flush caches, send int */ |
||
1843 | radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE_EOP, 4)); |
||
1844 | radeon_ring_write(rdev, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT_TS) | EVENT_INDEX(5)); |
||
1845 | radeon_ring_write(rdev, addr & 0xffffffff); |
||
1846 | radeon_ring_write(rdev, (upper_32_bits(addr) & 0xff) | DATA_SEL(1) | INT_SEL(2)); |
||
1847 | radeon_ring_write(rdev, fence->seq); |
||
1848 | radeon_ring_write(rdev, 0); |
||
1849 | } else { |
||
1430 | serge | 1850 | radeon_ring_write(rdev, PACKET3(PACKET3_EVENT_WRITE, 0)); |
1963 | serge | 1851 | radeon_ring_write(rdev, EVENT_TYPE(CACHE_FLUSH_AND_INV_EVENT) | EVENT_INDEX(0)); |
1430 | serge | 1852 | /* wait for 3D idle clean */ |
1853 | radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1)); |
||
1854 | radeon_ring_write(rdev, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); |
||
1855 | radeon_ring_write(rdev, WAIT_3D_IDLE_bit | WAIT_3D_IDLECLEAN_bit); |
||
1413 | serge | 1856 | /* Emit fence sequence & fire IRQ */ |
1857 | radeon_ring_write(rdev, PACKET3(PACKET3_SET_CONFIG_REG, 1)); |
||
1858 | radeon_ring_write(rdev, ((rdev->fence_drv.scratch_reg - PACKET3_SET_CONFIG_REG_OFFSET) >> 2)); |
||
1859 | radeon_ring_write(rdev, fence->seq); |
||
1860 | /* CP_INTERRUPT packet 3 no longer exists, use packet 0 */ |
||
1861 | radeon_ring_write(rdev, PACKET0(CP_INT_STATUS, 0)); |
||
1862 | radeon_ring_write(rdev, RB_INT_STAT); |
||
1963 | serge | 1863 | } |
1413 | serge | 1864 | } |
1963 | serge | 1865 | |
1866 | |||
1221 | serge | 1867 | int r600_set_surface_reg(struct radeon_device *rdev, int reg, |
1868 | uint32_t tiling_flags, uint32_t pitch, |
||
1869 | uint32_t offset, uint32_t obj_size) |
||
1870 | { |
||
1871 | /* FIXME: implement */ |
||
1872 | return 0; |
||
1873 | } |
||
1874 | |||
1875 | void r600_clear_surface_reg(struct radeon_device *rdev, int reg) |
||
1876 | { |
||
1877 | /* FIXME: implement */ |
||
1878 | } |
||
1879 | |||
1880 | int r600_startup(struct radeon_device *rdev) |
||
1881 | { |
||
1882 | int r; |
||
1883 | |||
1963 | serge | 1884 | /* enable pcie gen2 link */ |
1885 | r600_pcie_gen2_enable(rdev); |
||
1886 | |||
1413 | serge | 1887 | if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { |
1888 | r = r600_init_microcode(rdev); |
||
1889 | if (r) { |
||
1890 | DRM_ERROR("Failed to load firmware!\n"); |
||
1891 | return r; |
||
1892 | } |
||
1893 | } |
||
1894 | |||
1221 | serge | 1895 | r600_mc_program(rdev); |
1896 | if (rdev->flags & RADEON_IS_AGP) { |
||
1897 | r600_agp_enable(rdev); |
||
1898 | } else { |
||
1899 | r = r600_pcie_gart_enable(rdev); |
||
1900 | if (r) |
||
1901 | return r; |
||
1902 | } |
||
1903 | r600_gpu_init(rdev); |
||
1904 | |||
1413 | serge | 1905 | r = radeon_ring_init(rdev, rdev->cp.ring_size); |
1906 | if (r) |
||
1907 | return r; |
||
1908 | r = r600_cp_load_microcode(rdev); |
||
1909 | if (r) |
||
1910 | return r; |
||
1911 | r = r600_cp_resume(rdev); |
||
1912 | if (r) |
||
1913 | return r; |
||
1963 | serge | 1914 | |
1221 | serge | 1915 | return 0; |
1916 | } |
||
1917 | |||
1918 | void r600_vga_set_state(struct radeon_device *rdev, bool state) |
||
1919 | { |
||
1920 | uint32_t temp; |
||
1921 | |||
1922 | temp = RREG32(CONFIG_CNTL); |
||
1923 | if (state == false) { |
||
1924 | temp &= ~(1<<0); |
||
1925 | temp |= (1<<1); |
||
1926 | } else { |
||
1927 | temp &= ~(1<<1); |
||
1928 | } |
||
1929 | WREG32(CONFIG_CNTL, temp); |
||
1930 | } |
||
1931 | |||
1932 | |||
1933 | |||
1934 | |||
1935 | |||
1936 | /* Plan is to move initialization in that function and use |
||
1937 | * helper function so that radeon_device_init pretty much |
||
1938 | * do nothing more than calling asic specific function. This |
||
1939 | * should also allow to remove a bunch of callback function |
||
1940 | * like vram_info. |
||
1941 | */ |
||
1942 | int r600_init(struct radeon_device *rdev) |
||
1943 | { |
||
1944 | int r; |
||
1945 | |||
1946 | if (r600_debugfs_mc_info_init(rdev)) { |
||
1947 | DRM_ERROR("Failed to register debugfs file for mc !\n"); |
||
1948 | } |
||
1949 | /* This don't do much */ |
||
1963 | serge | 1950 | // r = radeon_gem_init(rdev); |
1951 | // if (r) |
||
1952 | // return r; |
||
1221 | serge | 1953 | /* Read BIOS */ |
1954 | if (!radeon_get_bios(rdev)) { |
||
1955 | if (ASIC_IS_AVIVO(rdev)) |
||
1956 | return -EINVAL; |
||
1957 | } |
||
1958 | /* Must be an ATOMBIOS */ |
||
1959 | if (!rdev->is_atom_bios) { |
||
1960 | dev_err(rdev->dev, "Expecting atombios for R600 GPU\n"); |
||
1961 | return -EINVAL; |
||
1962 | } |
||
1963 | r = radeon_atombios_init(rdev); |
||
1964 | if (r) |
||
1965 | return r; |
||
1966 | /* Post card if necessary */ |
||
1963 | serge | 1967 | if (!radeon_card_posted(rdev)) { |
1321 | serge | 1968 | if (!rdev->bios) { |
1969 | dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n"); |
||
1970 | return -EINVAL; |
||
1971 | } |
||
1221 | serge | 1972 | DRM_INFO("GPU not posted. posting now...\n"); |
1973 | atom_asic_init(rdev->mode_info.atom_context); |
||
1974 | } |
||
1975 | /* Initialize scratch registers */ |
||
1976 | r600_scratch_init(rdev); |
||
1977 | /* Initialize surface registers */ |
||
1978 | radeon_surface_init(rdev); |
||
1268 | serge | 1979 | /* Initialize clocks */ |
1221 | serge | 1980 | radeon_get_clock_info(rdev->ddev); |
1981 | /* Fence driver */ |
||
1963 | serge | 1982 | // r = radeon_fence_driver_init(rdev); |
1983 | // if (r) |
||
1984 | // return r; |
||
1403 | serge | 1985 | if (rdev->flags & RADEON_IS_AGP) { |
1986 | r = radeon_agp_init(rdev); |
||
1987 | if (r) |
||
1988 | radeon_agp_disable(rdev); |
||
1989 | } |
||
1221 | serge | 1990 | r = r600_mc_init(rdev); |
1991 | if (r) |
||
1992 | return r; |
||
1993 | /* Memory manager */ |
||
1321 | serge | 1994 | r = radeon_bo_init(rdev); |
1221 | serge | 1995 | if (r) |
1996 | return r; |
||
1321 | serge | 1997 | |
1998 | // r = radeon_irq_kms_init(rdev); |
||
1999 | // if (r) |
||
2000 | // return r; |
||
2001 | |||
1413 | serge | 2002 | rdev->cp.ring_obj = NULL; |
2003 | r600_ring_init(rdev, 1024 * 1024); |
||
1221 | serge | 2004 | |
1321 | serge | 2005 | // rdev->ih.ring_obj = NULL; |
2006 | // r600_ih_ring_init(rdev, 64 * 1024); |
||
1221 | serge | 2007 | |
2008 | r = r600_pcie_gart_init(rdev); |
||
2009 | if (r) |
||
2010 | return r; |
||
2011 | |||
1321 | serge | 2012 | rdev->accel_working = true; |
1221 | serge | 2013 | r = r600_startup(rdev); |
2014 | if (r) { |
||
1428 | serge | 2015 | dev_err(rdev->dev, "disabling GPU acceleration\n"); |
1221 | serge | 2016 | // r600_suspend(rdev); |
2017 | // r600_wb_fini(rdev); |
||
2018 | // radeon_ring_fini(rdev); |
||
2019 | r600_pcie_gart_fini(rdev); |
||
2020 | rdev->accel_working = false; |
||
2021 | } |
||
2022 | if (rdev->accel_working) { |
||
2023 | // r = radeon_ib_pool_init(rdev); |
||
2024 | // if (r) { |
||
2025 | // DRM_ERROR("radeon: failled initializing IB pool (%d).\n", r); |
||
2026 | // rdev->accel_working = false; |
||
2027 | // } |
||
2028 | // r = r600_ib_test(rdev); |
||
2029 | // if (r) { |
||
2030 | // DRM_ERROR("radeon: failled testing IB (%d).\n", r); |
||
2031 | // rdev->accel_working = false; |
||
2032 | // } |
||
2033 | } |
||
2034 | return 0; |
||
2035 | } |
||
2036 | |||
1963 | serge | 2037 | static void r600_disable_interrupt_state(struct radeon_device *rdev) |
2038 | { |
||
2039 | u32 tmp; |
||
1221 | serge | 2040 | |
1963 | serge | 2041 | WREG32(CP_INT_CNTL, CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE); |
2042 | WREG32(GRBM_INT_CNTL, 0); |
||
2043 | WREG32(DxMODE_INT_MASK, 0); |
||
2044 | WREG32(D1GRPH_INTERRUPT_CONTROL, 0); |
||
2045 | WREG32(D2GRPH_INTERRUPT_CONTROL, 0); |
||
2046 | if (ASIC_IS_DCE3(rdev)) { |
||
2047 | WREG32(DCE3_DACA_AUTODETECT_INT_CONTROL, 0); |
||
2048 | WREG32(DCE3_DACB_AUTODETECT_INT_CONTROL, 0); |
||
2049 | tmp = RREG32(DC_HPD1_INT_CONTROL) & DC_HPDx_INT_POLARITY; |
||
2050 | WREG32(DC_HPD1_INT_CONTROL, tmp); |
||
2051 | tmp = RREG32(DC_HPD2_INT_CONTROL) & DC_HPDx_INT_POLARITY; |
||
2052 | WREG32(DC_HPD2_INT_CONTROL, tmp); |
||
2053 | tmp = RREG32(DC_HPD3_INT_CONTROL) & DC_HPDx_INT_POLARITY; |
||
2054 | WREG32(DC_HPD3_INT_CONTROL, tmp); |
||
2055 | tmp = RREG32(DC_HPD4_INT_CONTROL) & DC_HPDx_INT_POLARITY; |
||
2056 | WREG32(DC_HPD4_INT_CONTROL, tmp); |
||
2057 | if (ASIC_IS_DCE32(rdev)) { |
||
2058 | tmp = RREG32(DC_HPD5_INT_CONTROL) & DC_HPDx_INT_POLARITY; |
||
2059 | WREG32(DC_HPD5_INT_CONTROL, tmp); |
||
2060 | tmp = RREG32(DC_HPD6_INT_CONTROL) & DC_HPDx_INT_POLARITY; |
||
2061 | WREG32(DC_HPD6_INT_CONTROL, tmp); |
||
2062 | } |
||
2063 | } else { |
||
2064 | WREG32(DACA_AUTODETECT_INT_CONTROL, 0); |
||
2065 | WREG32(DACB_AUTODETECT_INT_CONTROL, 0); |
||
2066 | tmp = RREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
2067 | WREG32(DC_HOT_PLUG_DETECT1_INT_CONTROL, tmp); |
||
2068 | tmp = RREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
2069 | WREG32(DC_HOT_PLUG_DETECT2_INT_CONTROL, tmp); |
||
2070 | tmp = RREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL) & DC_HOT_PLUG_DETECTx_INT_POLARITY; |
||
2071 | WREG32(DC_HOT_PLUG_DETECT3_INT_CONTROL, tmp); |
||
2072 | } |
||
2073 | } |
||
1221 | serge | 2074 | |
2075 | |||
2076 | |||
2077 | |||
2078 | |||
2079 | |||
2080 | |||
1963 | serge | 2081 | |
1221 | serge | 2082 | /* |
2083 | * Debugfs info |
||
2084 | */ |
||
2085 | #if defined(CONFIG_DEBUG_FS) |
||
2086 | |||
2087 | static int r600_debugfs_cp_ring_info(struct seq_file *m, void *data) |
||
2088 | { |
||
2089 | struct drm_info_node *node = (struct drm_info_node *) m->private; |
||
2090 | struct drm_device *dev = node->minor->dev; |
||
2091 | struct radeon_device *rdev = dev->dev_private; |
||
2092 | unsigned count, i, j; |
||
2093 | |||
2094 | radeon_ring_free_size(rdev); |
||
1321 | serge | 2095 | count = (rdev->cp.ring_size / 4) - rdev->cp.ring_free_dw; |
1221 | serge | 2096 | seq_printf(m, "CP_STAT 0x%08x\n", RREG32(CP_STAT)); |
1321 | serge | 2097 | seq_printf(m, "CP_RB_WPTR 0x%08x\n", RREG32(CP_RB_WPTR)); |
2098 | seq_printf(m, "CP_RB_RPTR 0x%08x\n", RREG32(CP_RB_RPTR)); |
||
2099 | seq_printf(m, "driver's copy of the CP_RB_WPTR 0x%08x\n", rdev->cp.wptr); |
||
2100 | seq_printf(m, "driver's copy of the CP_RB_RPTR 0x%08x\n", rdev->cp.rptr); |
||
1221 | serge | 2101 | seq_printf(m, "%u free dwords in ring\n", rdev->cp.ring_free_dw); |
2102 | seq_printf(m, "%u dwords in ring\n", count); |
||
1321 | serge | 2103 | i = rdev->cp.rptr; |
1221 | serge | 2104 | for (j = 0; j <= count; j++) { |
2105 | seq_printf(m, "r[%04d]=0x%08x\n", i, rdev->cp.ring[i]); |
||
1321 | serge | 2106 | i = (i + 1) & rdev->cp.ptr_mask; |
1221 | serge | 2107 | } |
2108 | return 0; |
||
2109 | } |
||
2110 | |||
2111 | static int r600_debugfs_mc_info(struct seq_file *m, void *data) |
||
2112 | { |
||
2113 | struct drm_info_node *node = (struct drm_info_node *) m->private; |
||
2114 | struct drm_device *dev = node->minor->dev; |
||
2115 | struct radeon_device *rdev = dev->dev_private; |
||
2116 | |||
2117 | DREG32_SYS(m, rdev, R_000E50_SRBM_STATUS); |
||
2118 | DREG32_SYS(m, rdev, VM_L2_STATUS); |
||
2119 | return 0; |
||
2120 | } |
||
2121 | |||
2122 | static struct drm_info_list r600_mc_info_list[] = { |
||
2123 | {"r600_mc_info", r600_debugfs_mc_info, 0, NULL}, |
||
2124 | {"r600_ring_info", r600_debugfs_cp_ring_info, 0, NULL}, |
||
2125 | }; |
||
2126 | #endif |
||
2127 | |||
2128 | int r600_debugfs_mc_info_init(struct radeon_device *rdev) |
||
2129 | { |
||
2130 | #if defined(CONFIG_DEBUG_FS) |
||
2131 | return radeon_debugfs_add_files(rdev, r600_mc_info_list, ARRAY_SIZE(r600_mc_info_list)); |
||
2132 | #else |
||
2133 | return 0; |
||
2134 | #endif |
||
2135 | } |
||
1404 | serge | 2136 | |
2137 | /** |
||
2138 | * r600_ioctl_wait_idle - flush host path cache on wait idle ioctl |
||
2139 | * rdev: radeon device structure |
||
2140 | * bo: buffer object struct which userspace is waiting for idle |
||
2141 | * |
||
2142 | * Some R6XX/R7XX doesn't seems to take into account HDP flush performed |
||
2143 | * through ring buffer, this leads to corruption in rendering, see |
||
2144 | * http://bugzilla.kernel.org/show_bug.cgi?id=15186 to avoid this we |
||
2145 | * directly perform HDP flush by writing register through MMIO. |
||
2146 | */ |
||
2147 | void r600_ioctl_wait_idle(struct radeon_device *rdev, struct radeon_bo *bo) |
||
2148 | { |
||
1963 | serge | 2149 | /* r7xx hw bug. write to HDP_DEBUG1 followed by fb read |
2150 | * rather than write to HDP_REG_COHERENCY_FLUSH_CNTL. |
||
2151 | * This seems to cause problems on some AGP cards. Just use the old |
||
2152 | * method for them. |
||
2153 | */ |
||
2154 | if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_RV740) && |
||
2155 | rdev->vram_scratch.ptr && !(rdev->flags & RADEON_IS_AGP)) { |
||
2156 | void __iomem *ptr = (void *)rdev->vram_scratch.ptr; |
||
2157 | u32 tmp; |
||
2158 | |||
2159 | WREG32(HDP_DEBUG1, 0); |
||
2160 | tmp = readl((void __iomem *)ptr); |
||
2161 | } else |
||
1404 | serge | 2162 | WREG32(R_005480_HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1); |
2163 | } |
||
1963 | serge | 2164 | |
2165 | void r600_set_pcie_lanes(struct radeon_device *rdev, int lanes) |
||
2166 | { |
||
2167 | u32 link_width_cntl, mask, target_reg; |
||
2168 | |||
2169 | if (rdev->flags & RADEON_IS_IGP) |
||
2170 | return; |
||
2171 | |||
2172 | if (!(rdev->flags & RADEON_IS_PCIE)) |
||
2173 | return; |
||
2174 | |||
2175 | /* x2 cards have a special sequence */ |
||
2176 | if (ASIC_IS_X2(rdev)) |
||
2177 | return; |
||
2178 | |||
2179 | /* FIXME wait for idle */ |
||
2180 | |||
2181 | switch (lanes) { |
||
2182 | case 0: |
||
2183 | mask = RADEON_PCIE_LC_LINK_WIDTH_X0; |
||
2184 | break; |
||
2185 | case 1: |
||
2186 | mask = RADEON_PCIE_LC_LINK_WIDTH_X1; |
||
2187 | break; |
||
2188 | case 2: |
||
2189 | mask = RADEON_PCIE_LC_LINK_WIDTH_X2; |
||
2190 | break; |
||
2191 | case 4: |
||
2192 | mask = RADEON_PCIE_LC_LINK_WIDTH_X4; |
||
2193 | break; |
||
2194 | case 8: |
||
2195 | mask = RADEON_PCIE_LC_LINK_WIDTH_X8; |
||
2196 | break; |
||
2197 | case 12: |
||
2198 | mask = RADEON_PCIE_LC_LINK_WIDTH_X12; |
||
2199 | break; |
||
2200 | case 16: |
||
2201 | default: |
||
2202 | mask = RADEON_PCIE_LC_LINK_WIDTH_X16; |
||
2203 | break; |
||
2204 | } |
||
2205 | |||
2206 | link_width_cntl = RREG32_PCIE_P(RADEON_PCIE_LC_LINK_WIDTH_CNTL); |
||
2207 | |||
2208 | if ((link_width_cntl & RADEON_PCIE_LC_LINK_WIDTH_RD_MASK) == |
||
2209 | (mask << RADEON_PCIE_LC_LINK_WIDTH_RD_SHIFT)) |
||
2210 | return; |
||
2211 | |||
2212 | if (link_width_cntl & R600_PCIE_LC_UPCONFIGURE_DIS) |
||
2213 | return; |
||
2214 | |||
2215 | link_width_cntl &= ~(RADEON_PCIE_LC_LINK_WIDTH_MASK | |
||
2216 | RADEON_PCIE_LC_RECONFIG_NOW | |
||
2217 | R600_PCIE_LC_RENEGOTIATE_EN | |
||
2218 | R600_PCIE_LC_RECONFIG_ARC_MISSING_ESCAPE); |
||
2219 | link_width_cntl |= mask; |
||
2220 | |||
2221 | WREG32_PCIE_P(RADEON_PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); |
||
2222 | |||
2223 | /* some northbridges can renegotiate the link rather than requiring |
||
2224 | * a complete re-config. |
||
2225 | * e.g., AMD 780/790 northbridges (pci ids: 0x5956, 0x5957, 0x5958, etc.) |
||
2226 | */ |
||
2227 | if (link_width_cntl & R600_PCIE_LC_RENEGOTIATION_SUPPORT) |
||
2228 | link_width_cntl |= R600_PCIE_LC_RENEGOTIATE_EN | R600_PCIE_LC_UPCONFIGURE_SUPPORT; |
||
2229 | else |
||
2230 | link_width_cntl |= R600_PCIE_LC_RECONFIG_ARC_MISSING_ESCAPE; |
||
2231 | |||
2232 | WREG32_PCIE_P(RADEON_PCIE_LC_LINK_WIDTH_CNTL, (link_width_cntl | |
||
2233 | RADEON_PCIE_LC_RECONFIG_NOW)); |
||
2234 | |||
2235 | if (rdev->family >= CHIP_RV770) |
||
2236 | target_reg = R700_TARGET_AND_CURRENT_PROFILE_INDEX; |
||
2237 | else |
||
2238 | target_reg = R600_TARGET_AND_CURRENT_PROFILE_INDEX; |
||
2239 | |||
2240 | /* wait for lane set to complete */ |
||
2241 | link_width_cntl = RREG32(target_reg); |
||
2242 | while (link_width_cntl == 0xffffffff) |
||
2243 | link_width_cntl = RREG32(target_reg); |
||
2244 | |||
2245 | } |
||
2246 | |||
2247 | int r600_get_pcie_lanes(struct radeon_device *rdev) |
||
2248 | { |
||
2249 | u32 link_width_cntl; |
||
2250 | |||
2251 | if (rdev->flags & RADEON_IS_IGP) |
||
2252 | return 0; |
||
2253 | |||
2254 | if (!(rdev->flags & RADEON_IS_PCIE)) |
||
2255 | return 0; |
||
2256 | |||
2257 | /* x2 cards have a special sequence */ |
||
2258 | if (ASIC_IS_X2(rdev)) |
||
2259 | return 0; |
||
2260 | |||
2261 | /* FIXME wait for idle */ |
||
2262 | |||
2263 | link_width_cntl = RREG32_PCIE_P(RADEON_PCIE_LC_LINK_WIDTH_CNTL); |
||
2264 | |||
2265 | switch ((link_width_cntl & RADEON_PCIE_LC_LINK_WIDTH_RD_MASK) >> RADEON_PCIE_LC_LINK_WIDTH_RD_SHIFT) { |
||
2266 | case RADEON_PCIE_LC_LINK_WIDTH_X0: |
||
2267 | return 0; |
||
2268 | case RADEON_PCIE_LC_LINK_WIDTH_X1: |
||
2269 | return 1; |
||
2270 | case RADEON_PCIE_LC_LINK_WIDTH_X2: |
||
2271 | return 2; |
||
2272 | case RADEON_PCIE_LC_LINK_WIDTH_X4: |
||
2273 | return 4; |
||
2274 | case RADEON_PCIE_LC_LINK_WIDTH_X8: |
||
2275 | return 8; |
||
2276 | case RADEON_PCIE_LC_LINK_WIDTH_X16: |
||
2277 | default: |
||
2278 | return 16; |
||
2279 | } |
||
2280 | } |
||
2281 | |||
2282 | static void r600_pcie_gen2_enable(struct radeon_device *rdev) |
||
2283 | { |
||
2284 | u32 link_width_cntl, lanes, speed_cntl, training_cntl, tmp; |
||
2285 | u16 link_cntl2; |
||
2286 | |||
2287 | if (radeon_pcie_gen2 == 0) |
||
2288 | return; |
||
2289 | |||
2290 | if (rdev->flags & RADEON_IS_IGP) |
||
2291 | return; |
||
2292 | |||
2293 | if (!(rdev->flags & RADEON_IS_PCIE)) |
||
2294 | return; |
||
2295 | |||
2296 | /* x2 cards have a special sequence */ |
||
2297 | if (ASIC_IS_X2(rdev)) |
||
2298 | return; |
||
2299 | |||
2300 | /* only RV6xx+ chips are supported */ |
||
2301 | if (rdev->family <= CHIP_R600) |
||
2302 | return; |
||
2303 | |||
2304 | /* 55 nm r6xx asics */ |
||
2305 | if ((rdev->family == CHIP_RV670) || |
||
2306 | (rdev->family == CHIP_RV620) || |
||
2307 | (rdev->family == CHIP_RV635)) { |
||
2308 | /* advertise upconfig capability */ |
||
2309 | link_width_cntl = RREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL); |
||
2310 | link_width_cntl &= ~LC_UPCONFIGURE_DIS; |
||
2311 | WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); |
||
2312 | link_width_cntl = RREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL); |
||
2313 | if (link_width_cntl & LC_RENEGOTIATION_SUPPORT) { |
||
2314 | lanes = (link_width_cntl & LC_LINK_WIDTH_RD_MASK) >> LC_LINK_WIDTH_RD_SHIFT; |
||
2315 | link_width_cntl &= ~(LC_LINK_WIDTH_MASK | |
||
2316 | LC_RECONFIG_ARC_MISSING_ESCAPE); |
||
2317 | link_width_cntl |= lanes | LC_RECONFIG_NOW | LC_RENEGOTIATE_EN; |
||
2318 | WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); |
||
2319 | } else { |
||
2320 | link_width_cntl |= LC_UPCONFIGURE_DIS; |
||
2321 | WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); |
||
2322 | } |
||
2323 | } |
||
2324 | |||
2325 | speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); |
||
2326 | if ((speed_cntl & LC_OTHER_SIDE_EVER_SENT_GEN2) && |
||
2327 | (speed_cntl & LC_OTHER_SIDE_SUPPORTS_GEN2)) { |
||
2328 | |||
2329 | /* 55 nm r6xx asics */ |
||
2330 | if ((rdev->family == CHIP_RV670) || |
||
2331 | (rdev->family == CHIP_RV620) || |
||
2332 | (rdev->family == CHIP_RV635)) { |
||
2333 | WREG32(MM_CFGREGS_CNTL, 0x8); |
||
2334 | link_cntl2 = RREG32(0x4088); |
||
2335 | WREG32(MM_CFGREGS_CNTL, 0); |
||
2336 | /* not supported yet */ |
||
2337 | if (link_cntl2 & SELECTABLE_DEEMPHASIS) |
||
2338 | return; |
||
2339 | } |
||
2340 | |||
2341 | speed_cntl &= ~LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK; |
||
2342 | speed_cntl |= (0x3 << LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT); |
||
2343 | speed_cntl &= ~LC_VOLTAGE_TIMER_SEL_MASK; |
||
2344 | speed_cntl &= ~LC_FORCE_DIS_HW_SPEED_CHANGE; |
||
2345 | speed_cntl |= LC_FORCE_EN_HW_SPEED_CHANGE; |
||
2346 | WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl); |
||
2347 | |||
2348 | tmp = RREG32(0x541c); |
||
2349 | WREG32(0x541c, tmp | 0x8); |
||
2350 | WREG32(MM_CFGREGS_CNTL, MM_WR_TO_CFG_EN); |
||
2351 | link_cntl2 = RREG16(0x4088); |
||
2352 | link_cntl2 &= ~TARGET_LINK_SPEED_MASK; |
||
2353 | link_cntl2 |= 0x2; |
||
2354 | WREG16(0x4088, link_cntl2); |
||
2355 | WREG32(MM_CFGREGS_CNTL, 0); |
||
2356 | |||
2357 | if ((rdev->family == CHIP_RV670) || |
||
2358 | (rdev->family == CHIP_RV620) || |
||
2359 | (rdev->family == CHIP_RV635)) { |
||
2360 | training_cntl = RREG32_PCIE_P(PCIE_LC_TRAINING_CNTL); |
||
2361 | training_cntl &= ~LC_POINT_7_PLUS_EN; |
||
2362 | WREG32_PCIE_P(PCIE_LC_TRAINING_CNTL, training_cntl); |
||
2363 | } else { |
||
2364 | speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); |
||
2365 | speed_cntl &= ~LC_TARGET_LINK_SPEED_OVERRIDE_EN; |
||
2366 | WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl); |
||
2367 | } |
||
2368 | |||
2369 | speed_cntl = RREG32_PCIE_P(PCIE_LC_SPEED_CNTL); |
||
2370 | speed_cntl |= LC_GEN2_EN_STRAP; |
||
2371 | WREG32_PCIE_P(PCIE_LC_SPEED_CNTL, speed_cntl); |
||
2372 | |||
2373 | } else { |
||
2374 | link_width_cntl = RREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL); |
||
2375 | /* XXX: only disable it if gen1 bridge vendor == 0x111d or 0x1106 */ |
||
2376 | if (1) |
||
2377 | link_width_cntl |= LC_UPCONFIGURE_DIS; |
||
2378 | else |
||
2379 | link_width_cntl &= ~LC_UPCONFIGURE_DIS; |
||
2380 | WREG32_PCIE_P(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); |
||
2381 | } |
||
2382 | }><>=>><>=>=>1); |