Loading...
Note: File does not exist in v3.1.
1/*
2 * Copyright 2011 Advanced Micro Devices, Inc.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 *
22 * Authors: Alex Deucher
23 */
24#include <linux/firmware.h>
25#include <linux/slab.h>
26#include <linux/module.h>
27#include <drm/drmP.h>
28#include "radeon.h"
29#include "radeon_asic.h"
30#include "radeon_audio.h"
31#include <drm/radeon_drm.h>
32#include "sid.h"
33#include "atom.h"
34#include "si_blit_shaders.h"
35#include "clearstate_si.h"
36#include "radeon_ucode.h"
37
38
39MODULE_FIRMWARE("radeon/TAHITI_pfp.bin");
40MODULE_FIRMWARE("radeon/TAHITI_me.bin");
41MODULE_FIRMWARE("radeon/TAHITI_ce.bin");
42MODULE_FIRMWARE("radeon/TAHITI_mc.bin");
43MODULE_FIRMWARE("radeon/TAHITI_mc2.bin");
44MODULE_FIRMWARE("radeon/TAHITI_rlc.bin");
45MODULE_FIRMWARE("radeon/TAHITI_smc.bin");
46
47MODULE_FIRMWARE("radeon/tahiti_pfp.bin");
48MODULE_FIRMWARE("radeon/tahiti_me.bin");
49MODULE_FIRMWARE("radeon/tahiti_ce.bin");
50MODULE_FIRMWARE("radeon/tahiti_mc.bin");
51MODULE_FIRMWARE("radeon/tahiti_rlc.bin");
52MODULE_FIRMWARE("radeon/tahiti_smc.bin");
53
54MODULE_FIRMWARE("radeon/PITCAIRN_pfp.bin");
55MODULE_FIRMWARE("radeon/PITCAIRN_me.bin");
56MODULE_FIRMWARE("radeon/PITCAIRN_ce.bin");
57MODULE_FIRMWARE("radeon/PITCAIRN_mc.bin");
58MODULE_FIRMWARE("radeon/PITCAIRN_mc2.bin");
59MODULE_FIRMWARE("radeon/PITCAIRN_rlc.bin");
60MODULE_FIRMWARE("radeon/PITCAIRN_smc.bin");
61
62MODULE_FIRMWARE("radeon/pitcairn_pfp.bin");
63MODULE_FIRMWARE("radeon/pitcairn_me.bin");
64MODULE_FIRMWARE("radeon/pitcairn_ce.bin");
65MODULE_FIRMWARE("radeon/pitcairn_mc.bin");
66MODULE_FIRMWARE("radeon/pitcairn_rlc.bin");
67MODULE_FIRMWARE("radeon/pitcairn_smc.bin");
68MODULE_FIRMWARE("radeon/pitcairn_k_smc.bin");
69
70MODULE_FIRMWARE("radeon/VERDE_pfp.bin");
71MODULE_FIRMWARE("radeon/VERDE_me.bin");
72MODULE_FIRMWARE("radeon/VERDE_ce.bin");
73MODULE_FIRMWARE("radeon/VERDE_mc.bin");
74MODULE_FIRMWARE("radeon/VERDE_mc2.bin");
75MODULE_FIRMWARE("radeon/VERDE_rlc.bin");
76MODULE_FIRMWARE("radeon/VERDE_smc.bin");
77
78MODULE_FIRMWARE("radeon/verde_pfp.bin");
79MODULE_FIRMWARE("radeon/verde_me.bin");
80MODULE_FIRMWARE("radeon/verde_ce.bin");
81MODULE_FIRMWARE("radeon/verde_mc.bin");
82MODULE_FIRMWARE("radeon/verde_rlc.bin");
83MODULE_FIRMWARE("radeon/verde_smc.bin");
84MODULE_FIRMWARE("radeon/verde_k_smc.bin");
85
86MODULE_FIRMWARE("radeon/OLAND_pfp.bin");
87MODULE_FIRMWARE("radeon/OLAND_me.bin");
88MODULE_FIRMWARE("radeon/OLAND_ce.bin");
89MODULE_FIRMWARE("radeon/OLAND_mc.bin");
90MODULE_FIRMWARE("radeon/OLAND_mc2.bin");
91MODULE_FIRMWARE("radeon/OLAND_rlc.bin");
92MODULE_FIRMWARE("radeon/OLAND_smc.bin");
93
94MODULE_FIRMWARE("radeon/oland_pfp.bin");
95MODULE_FIRMWARE("radeon/oland_me.bin");
96MODULE_FIRMWARE("radeon/oland_ce.bin");
97MODULE_FIRMWARE("radeon/oland_mc.bin");
98MODULE_FIRMWARE("radeon/oland_rlc.bin");
99MODULE_FIRMWARE("radeon/oland_smc.bin");
100MODULE_FIRMWARE("radeon/oland_k_smc.bin");
101
102MODULE_FIRMWARE("radeon/HAINAN_pfp.bin");
103MODULE_FIRMWARE("radeon/HAINAN_me.bin");
104MODULE_FIRMWARE("radeon/HAINAN_ce.bin");
105MODULE_FIRMWARE("radeon/HAINAN_mc.bin");
106MODULE_FIRMWARE("radeon/HAINAN_mc2.bin");
107MODULE_FIRMWARE("radeon/HAINAN_rlc.bin");
108MODULE_FIRMWARE("radeon/HAINAN_smc.bin");
109
110MODULE_FIRMWARE("radeon/hainan_pfp.bin");
111MODULE_FIRMWARE("radeon/hainan_me.bin");
112MODULE_FIRMWARE("radeon/hainan_ce.bin");
113MODULE_FIRMWARE("radeon/hainan_mc.bin");
114MODULE_FIRMWARE("radeon/hainan_rlc.bin");
115MODULE_FIRMWARE("radeon/hainan_smc.bin");
116MODULE_FIRMWARE("radeon/hainan_k_smc.bin");
117MODULE_FIRMWARE("radeon/banks_k_2_smc.bin");
118
119MODULE_FIRMWARE("radeon/si58_mc.bin");
120
121static u32 si_get_cu_active_bitmap(struct radeon_device *rdev, u32 se, u32 sh);
122static void si_pcie_gen3_enable(struct radeon_device *rdev);
123static void si_program_aspm(struct radeon_device *rdev);
124extern void sumo_rlc_fini(struct radeon_device *rdev);
125extern int sumo_rlc_init(struct radeon_device *rdev);
126extern int r600_ih_ring_alloc(struct radeon_device *rdev);
127extern void r600_ih_ring_fini(struct radeon_device *rdev);
128extern void evergreen_fix_pci_max_read_req_size(struct radeon_device *rdev);
129extern void evergreen_mc_stop(struct radeon_device *rdev, struct evergreen_mc_save *save);
130extern void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_save *save);
131extern u32 evergreen_get_number_of_dram_channels(struct radeon_device *rdev);
132extern void evergreen_print_gpu_status_regs(struct radeon_device *rdev);
133extern bool evergreen_is_display_hung(struct radeon_device *rdev);
134static void si_enable_gui_idle_interrupt(struct radeon_device *rdev,
135 bool enable);
136static void si_init_pg(struct radeon_device *rdev);
137static void si_init_cg(struct radeon_device *rdev);
138static void si_fini_pg(struct radeon_device *rdev);
139static void si_fini_cg(struct radeon_device *rdev);
140static void si_rlc_stop(struct radeon_device *rdev);
141
142static const u32 verde_rlc_save_restore_register_list[] =
143{
144 (0x8000 << 16) | (0x98f4 >> 2),
145 0x00000000,
146 (0x8040 << 16) | (0x98f4 >> 2),
147 0x00000000,
148 (0x8000 << 16) | (0xe80 >> 2),
149 0x00000000,
150 (0x8040 << 16) | (0xe80 >> 2),
151 0x00000000,
152 (0x8000 << 16) | (0x89bc >> 2),
153 0x00000000,
154 (0x8040 << 16) | (0x89bc >> 2),
155 0x00000000,
156 (0x8000 << 16) | (0x8c1c >> 2),
157 0x00000000,
158 (0x8040 << 16) | (0x8c1c >> 2),
159 0x00000000,
160 (0x9c00 << 16) | (0x98f0 >> 2),
161 0x00000000,
162 (0x9c00 << 16) | (0xe7c >> 2),
163 0x00000000,
164 (0x8000 << 16) | (0x9148 >> 2),
165 0x00000000,
166 (0x8040 << 16) | (0x9148 >> 2),
167 0x00000000,
168 (0x9c00 << 16) | (0x9150 >> 2),
169 0x00000000,
170 (0x9c00 << 16) | (0x897c >> 2),
171 0x00000000,
172 (0x9c00 << 16) | (0x8d8c >> 2),
173 0x00000000,
174 (0x9c00 << 16) | (0xac54 >> 2),
175 0X00000000,
176 0x3,
177 (0x9c00 << 16) | (0x98f8 >> 2),
178 0x00000000,
179 (0x9c00 << 16) | (0x9910 >> 2),
180 0x00000000,
181 (0x9c00 << 16) | (0x9914 >> 2),
182 0x00000000,
183 (0x9c00 << 16) | (0x9918 >> 2),
184 0x00000000,
185 (0x9c00 << 16) | (0x991c >> 2),
186 0x00000000,
187 (0x9c00 << 16) | (0x9920 >> 2),
188 0x00000000,
189 (0x9c00 << 16) | (0x9924 >> 2),
190 0x00000000,
191 (0x9c00 << 16) | (0x9928 >> 2),
192 0x00000000,
193 (0x9c00 << 16) | (0x992c >> 2),
194 0x00000000,
195 (0x9c00 << 16) | (0x9930 >> 2),
196 0x00000000,
197 (0x9c00 << 16) | (0x9934 >> 2),
198 0x00000000,
199 (0x9c00 << 16) | (0x9938 >> 2),
200 0x00000000,
201 (0x9c00 << 16) | (0x993c >> 2),
202 0x00000000,
203 (0x9c00 << 16) | (0x9940 >> 2),
204 0x00000000,
205 (0x9c00 << 16) | (0x9944 >> 2),
206 0x00000000,
207 (0x9c00 << 16) | (0x9948 >> 2),
208 0x00000000,
209 (0x9c00 << 16) | (0x994c >> 2),
210 0x00000000,
211 (0x9c00 << 16) | (0x9950 >> 2),
212 0x00000000,
213 (0x9c00 << 16) | (0x9954 >> 2),
214 0x00000000,
215 (0x9c00 << 16) | (0x9958 >> 2),
216 0x00000000,
217 (0x9c00 << 16) | (0x995c >> 2),
218 0x00000000,
219 (0x9c00 << 16) | (0x9960 >> 2),
220 0x00000000,
221 (0x9c00 << 16) | (0x9964 >> 2),
222 0x00000000,
223 (0x9c00 << 16) | (0x9968 >> 2),
224 0x00000000,
225 (0x9c00 << 16) | (0x996c >> 2),
226 0x00000000,
227 (0x9c00 << 16) | (0x9970 >> 2),
228 0x00000000,
229 (0x9c00 << 16) | (0x9974 >> 2),
230 0x00000000,
231 (0x9c00 << 16) | (0x9978 >> 2),
232 0x00000000,
233 (0x9c00 << 16) | (0x997c >> 2),
234 0x00000000,
235 (0x9c00 << 16) | (0x9980 >> 2),
236 0x00000000,
237 (0x9c00 << 16) | (0x9984 >> 2),
238 0x00000000,
239 (0x9c00 << 16) | (0x9988 >> 2),
240 0x00000000,
241 (0x9c00 << 16) | (0x998c >> 2),
242 0x00000000,
243 (0x9c00 << 16) | (0x8c00 >> 2),
244 0x00000000,
245 (0x9c00 << 16) | (0x8c14 >> 2),
246 0x00000000,
247 (0x9c00 << 16) | (0x8c04 >> 2),
248 0x00000000,
249 (0x9c00 << 16) | (0x8c08 >> 2),
250 0x00000000,
251 (0x8000 << 16) | (0x9b7c >> 2),
252 0x00000000,
253 (0x8040 << 16) | (0x9b7c >> 2),
254 0x00000000,
255 (0x8000 << 16) | (0xe84 >> 2),
256 0x00000000,
257 (0x8040 << 16) | (0xe84 >> 2),
258 0x00000000,
259 (0x8000 << 16) | (0x89c0 >> 2),
260 0x00000000,
261 (0x8040 << 16) | (0x89c0 >> 2),
262 0x00000000,
263 (0x8000 << 16) | (0x914c >> 2),
264 0x00000000,
265 (0x8040 << 16) | (0x914c >> 2),
266 0x00000000,
267 (0x8000 << 16) | (0x8c20 >> 2),
268 0x00000000,
269 (0x8040 << 16) | (0x8c20 >> 2),
270 0x00000000,
271 (0x8000 << 16) | (0x9354 >> 2),
272 0x00000000,
273 (0x8040 << 16) | (0x9354 >> 2),
274 0x00000000,
275 (0x9c00 << 16) | (0x9060 >> 2),
276 0x00000000,
277 (0x9c00 << 16) | (0x9364 >> 2),
278 0x00000000,
279 (0x9c00 << 16) | (0x9100 >> 2),
280 0x00000000,
281 (0x9c00 << 16) | (0x913c >> 2),
282 0x00000000,
283 (0x8000 << 16) | (0x90e0 >> 2),
284 0x00000000,
285 (0x8000 << 16) | (0x90e4 >> 2),
286 0x00000000,
287 (0x8000 << 16) | (0x90e8 >> 2),
288 0x00000000,
289 (0x8040 << 16) | (0x90e0 >> 2),
290 0x00000000,
291 (0x8040 << 16) | (0x90e4 >> 2),
292 0x00000000,
293 (0x8040 << 16) | (0x90e8 >> 2),
294 0x00000000,
295 (0x9c00 << 16) | (0x8bcc >> 2),
296 0x00000000,
297 (0x9c00 << 16) | (0x8b24 >> 2),
298 0x00000000,
299 (0x9c00 << 16) | (0x88c4 >> 2),
300 0x00000000,
301 (0x9c00 << 16) | (0x8e50 >> 2),
302 0x00000000,
303 (0x9c00 << 16) | (0x8c0c >> 2),
304 0x00000000,
305 (0x9c00 << 16) | (0x8e58 >> 2),
306 0x00000000,
307 (0x9c00 << 16) | (0x8e5c >> 2),
308 0x00000000,
309 (0x9c00 << 16) | (0x9508 >> 2),
310 0x00000000,
311 (0x9c00 << 16) | (0x950c >> 2),
312 0x00000000,
313 (0x9c00 << 16) | (0x9494 >> 2),
314 0x00000000,
315 (0x9c00 << 16) | (0xac0c >> 2),
316 0x00000000,
317 (0x9c00 << 16) | (0xac10 >> 2),
318 0x00000000,
319 (0x9c00 << 16) | (0xac14 >> 2),
320 0x00000000,
321 (0x9c00 << 16) | (0xae00 >> 2),
322 0x00000000,
323 (0x9c00 << 16) | (0xac08 >> 2),
324 0x00000000,
325 (0x9c00 << 16) | (0x88d4 >> 2),
326 0x00000000,
327 (0x9c00 << 16) | (0x88c8 >> 2),
328 0x00000000,
329 (0x9c00 << 16) | (0x88cc >> 2),
330 0x00000000,
331 (0x9c00 << 16) | (0x89b0 >> 2),
332 0x00000000,
333 (0x9c00 << 16) | (0x8b10 >> 2),
334 0x00000000,
335 (0x9c00 << 16) | (0x8a14 >> 2),
336 0x00000000,
337 (0x9c00 << 16) | (0x9830 >> 2),
338 0x00000000,
339 (0x9c00 << 16) | (0x9834 >> 2),
340 0x00000000,
341 (0x9c00 << 16) | (0x9838 >> 2),
342 0x00000000,
343 (0x9c00 << 16) | (0x9a10 >> 2),
344 0x00000000,
345 (0x8000 << 16) | (0x9870 >> 2),
346 0x00000000,
347 (0x8000 << 16) | (0x9874 >> 2),
348 0x00000000,
349 (0x8001 << 16) | (0x9870 >> 2),
350 0x00000000,
351 (0x8001 << 16) | (0x9874 >> 2),
352 0x00000000,
353 (0x8040 << 16) | (0x9870 >> 2),
354 0x00000000,
355 (0x8040 << 16) | (0x9874 >> 2),
356 0x00000000,
357 (0x8041 << 16) | (0x9870 >> 2),
358 0x00000000,
359 (0x8041 << 16) | (0x9874 >> 2),
360 0x00000000,
361 0x00000000
362};
363
364static const u32 tahiti_golden_rlc_registers[] =
365{
366 0xc424, 0xffffffff, 0x00601005,
367 0xc47c, 0xffffffff, 0x10104040,
368 0xc488, 0xffffffff, 0x0100000a,
369 0xc314, 0xffffffff, 0x00000800,
370 0xc30c, 0xffffffff, 0x800000f4,
371 0xf4a8, 0xffffffff, 0x00000000
372};
373
374static const u32 tahiti_golden_registers[] =
375{
376 0x9a10, 0x00010000, 0x00018208,
377 0x9830, 0xffffffff, 0x00000000,
378 0x9834, 0xf00fffff, 0x00000400,
379 0x9838, 0x0002021c, 0x00020200,
380 0xc78, 0x00000080, 0x00000000,
381 0xd030, 0x000300c0, 0x00800040,
382 0xd830, 0x000300c0, 0x00800040,
383 0x5bb0, 0x000000f0, 0x00000070,
384 0x5bc0, 0x00200000, 0x50100000,
385 0x7030, 0x31000311, 0x00000011,
386 0x277c, 0x00000003, 0x000007ff,
387 0x240c, 0x000007ff, 0x00000000,
388 0x8a14, 0xf000001f, 0x00000007,
389 0x8b24, 0xffffffff, 0x00ffffff,
390 0x8b10, 0x0000ff0f, 0x00000000,
391 0x28a4c, 0x07ffffff, 0x4e000000,
392 0x28350, 0x3f3f3fff, 0x2a00126a,
393 0x30, 0x000000ff, 0x0040,
394 0x34, 0x00000040, 0x00004040,
395 0x9100, 0x07ffffff, 0x03000000,
396 0x8e88, 0x01ff1f3f, 0x00000000,
397 0x8e84, 0x01ff1f3f, 0x00000000,
398 0x9060, 0x0000007f, 0x00000020,
399 0x9508, 0x00010000, 0x00010000,
400 0xac14, 0x00000200, 0x000002fb,
401 0xac10, 0xffffffff, 0x0000543b,
402 0xac0c, 0xffffffff, 0xa9210876,
403 0x88d0, 0xffffffff, 0x000fff40,
404 0x88d4, 0x0000001f, 0x00000010,
405 0x1410, 0x20000000, 0x20fffed8,
406 0x15c0, 0x000c0fc0, 0x000c0400
407};
408
409static const u32 tahiti_golden_registers2[] =
410{
411 0xc64, 0x00000001, 0x00000001
412};
413
414static const u32 pitcairn_golden_rlc_registers[] =
415{
416 0xc424, 0xffffffff, 0x00601004,
417 0xc47c, 0xffffffff, 0x10102020,
418 0xc488, 0xffffffff, 0x01000020,
419 0xc314, 0xffffffff, 0x00000800,
420 0xc30c, 0xffffffff, 0x800000a4
421};
422
423static const u32 pitcairn_golden_registers[] =
424{
425 0x9a10, 0x00010000, 0x00018208,
426 0x9830, 0xffffffff, 0x00000000,
427 0x9834, 0xf00fffff, 0x00000400,
428 0x9838, 0x0002021c, 0x00020200,
429 0xc78, 0x00000080, 0x00000000,
430 0xd030, 0x000300c0, 0x00800040,
431 0xd830, 0x000300c0, 0x00800040,
432 0x5bb0, 0x000000f0, 0x00000070,
433 0x5bc0, 0x00200000, 0x50100000,
434 0x7030, 0x31000311, 0x00000011,
435 0x2ae4, 0x00073ffe, 0x000022a2,
436 0x240c, 0x000007ff, 0x00000000,
437 0x8a14, 0xf000001f, 0x00000007,
438 0x8b24, 0xffffffff, 0x00ffffff,
439 0x8b10, 0x0000ff0f, 0x00000000,
440 0x28a4c, 0x07ffffff, 0x4e000000,
441 0x28350, 0x3f3f3fff, 0x2a00126a,
442 0x30, 0x000000ff, 0x0040,
443 0x34, 0x00000040, 0x00004040,
444 0x9100, 0x07ffffff, 0x03000000,
445 0x9060, 0x0000007f, 0x00000020,
446 0x9508, 0x00010000, 0x00010000,
447 0xac14, 0x000003ff, 0x000000f7,
448 0xac10, 0xffffffff, 0x00000000,
449 0xac0c, 0xffffffff, 0x32761054,
450 0x88d4, 0x0000001f, 0x00000010,
451 0x15c0, 0x000c0fc0, 0x000c0400
452};
453
454static const u32 verde_golden_rlc_registers[] =
455{
456 0xc424, 0xffffffff, 0x033f1005,
457 0xc47c, 0xffffffff, 0x10808020,
458 0xc488, 0xffffffff, 0x00800008,
459 0xc314, 0xffffffff, 0x00001000,
460 0xc30c, 0xffffffff, 0x80010014
461};
462
463static const u32 verde_golden_registers[] =
464{
465 0x9a10, 0x00010000, 0x00018208,
466 0x9830, 0xffffffff, 0x00000000,
467 0x9834, 0xf00fffff, 0x00000400,
468 0x9838, 0x0002021c, 0x00020200,
469 0xc78, 0x00000080, 0x00000000,
470 0xd030, 0x000300c0, 0x00800040,
471 0xd030, 0x000300c0, 0x00800040,
472 0xd830, 0x000300c0, 0x00800040,
473 0xd830, 0x000300c0, 0x00800040,
474 0x5bb0, 0x000000f0, 0x00000070,
475 0x5bc0, 0x00200000, 0x50100000,
476 0x7030, 0x31000311, 0x00000011,
477 0x2ae4, 0x00073ffe, 0x000022a2,
478 0x2ae4, 0x00073ffe, 0x000022a2,
479 0x2ae4, 0x00073ffe, 0x000022a2,
480 0x240c, 0x000007ff, 0x00000000,
481 0x240c, 0x000007ff, 0x00000000,
482 0x240c, 0x000007ff, 0x00000000,
483 0x8a14, 0xf000001f, 0x00000007,
484 0x8a14, 0xf000001f, 0x00000007,
485 0x8a14, 0xf000001f, 0x00000007,
486 0x8b24, 0xffffffff, 0x00ffffff,
487 0x8b10, 0x0000ff0f, 0x00000000,
488 0x28a4c, 0x07ffffff, 0x4e000000,
489 0x28350, 0x3f3f3fff, 0x0000124a,
490 0x28350, 0x3f3f3fff, 0x0000124a,
491 0x28350, 0x3f3f3fff, 0x0000124a,
492 0x30, 0x000000ff, 0x0040,
493 0x34, 0x00000040, 0x00004040,
494 0x9100, 0x07ffffff, 0x03000000,
495 0x9100, 0x07ffffff, 0x03000000,
496 0x8e88, 0x01ff1f3f, 0x00000000,
497 0x8e88, 0x01ff1f3f, 0x00000000,
498 0x8e88, 0x01ff1f3f, 0x00000000,
499 0x8e84, 0x01ff1f3f, 0x00000000,
500 0x8e84, 0x01ff1f3f, 0x00000000,
501 0x8e84, 0x01ff1f3f, 0x00000000,
502 0x9060, 0x0000007f, 0x00000020,
503 0x9508, 0x00010000, 0x00010000,
504 0xac14, 0x000003ff, 0x00000003,
505 0xac14, 0x000003ff, 0x00000003,
506 0xac14, 0x000003ff, 0x00000003,
507 0xac10, 0xffffffff, 0x00000000,
508 0xac10, 0xffffffff, 0x00000000,
509 0xac10, 0xffffffff, 0x00000000,
510 0xac0c, 0xffffffff, 0x00001032,
511 0xac0c, 0xffffffff, 0x00001032,
512 0xac0c, 0xffffffff, 0x00001032,
513 0x88d4, 0x0000001f, 0x00000010,
514 0x88d4, 0x0000001f, 0x00000010,
515 0x88d4, 0x0000001f, 0x00000010,
516 0x15c0, 0x000c0fc0, 0x000c0400
517};
518
519static const u32 oland_golden_rlc_registers[] =
520{
521 0xc424, 0xffffffff, 0x00601005,
522 0xc47c, 0xffffffff, 0x10104040,
523 0xc488, 0xffffffff, 0x0100000a,
524 0xc314, 0xffffffff, 0x00000800,
525 0xc30c, 0xffffffff, 0x800000f4
526};
527
528static const u32 oland_golden_registers[] =
529{
530 0x9a10, 0x00010000, 0x00018208,
531 0x9830, 0xffffffff, 0x00000000,
532 0x9834, 0xf00fffff, 0x00000400,
533 0x9838, 0x0002021c, 0x00020200,
534 0xc78, 0x00000080, 0x00000000,
535 0xd030, 0x000300c0, 0x00800040,
536 0xd830, 0x000300c0, 0x00800040,
537 0x5bb0, 0x000000f0, 0x00000070,
538 0x5bc0, 0x00200000, 0x50100000,
539 0x7030, 0x31000311, 0x00000011,
540 0x2ae4, 0x00073ffe, 0x000022a2,
541 0x240c, 0x000007ff, 0x00000000,
542 0x8a14, 0xf000001f, 0x00000007,
543 0x8b24, 0xffffffff, 0x00ffffff,
544 0x8b10, 0x0000ff0f, 0x00000000,
545 0x28a4c, 0x07ffffff, 0x4e000000,
546 0x28350, 0x3f3f3fff, 0x00000082,
547 0x30, 0x000000ff, 0x0040,
548 0x34, 0x00000040, 0x00004040,
549 0x9100, 0x07ffffff, 0x03000000,
550 0x9060, 0x0000007f, 0x00000020,
551 0x9508, 0x00010000, 0x00010000,
552 0xac14, 0x000003ff, 0x000000f3,
553 0xac10, 0xffffffff, 0x00000000,
554 0xac0c, 0xffffffff, 0x00003210,
555 0x88d4, 0x0000001f, 0x00000010,
556 0x15c0, 0x000c0fc0, 0x000c0400
557};
558
559static const u32 hainan_golden_registers[] =
560{
561 0x9a10, 0x00010000, 0x00018208,
562 0x9830, 0xffffffff, 0x00000000,
563 0x9834, 0xf00fffff, 0x00000400,
564 0x9838, 0x0002021c, 0x00020200,
565 0xd0c0, 0xff000fff, 0x00000100,
566 0xd030, 0x000300c0, 0x00800040,
567 0xd8c0, 0xff000fff, 0x00000100,
568 0xd830, 0x000300c0, 0x00800040,
569 0x2ae4, 0x00073ffe, 0x000022a2,
570 0x240c, 0x000007ff, 0x00000000,
571 0x8a14, 0xf000001f, 0x00000007,
572 0x8b24, 0xffffffff, 0x00ffffff,
573 0x8b10, 0x0000ff0f, 0x00000000,
574 0x28a4c, 0x07ffffff, 0x4e000000,
575 0x28350, 0x3f3f3fff, 0x00000000,
576 0x30, 0x000000ff, 0x0040,
577 0x34, 0x00000040, 0x00004040,
578 0x9100, 0x03e00000, 0x03600000,
579 0x9060, 0x0000007f, 0x00000020,
580 0x9508, 0x00010000, 0x00010000,
581 0xac14, 0x000003ff, 0x000000f1,
582 0xac10, 0xffffffff, 0x00000000,
583 0xac0c, 0xffffffff, 0x00003210,
584 0x88d4, 0x0000001f, 0x00000010,
585 0x15c0, 0x000c0fc0, 0x000c0400
586};
587
588static const u32 hainan_golden_registers2[] =
589{
590 0x98f8, 0xffffffff, 0x02010001
591};
592
593static const u32 tahiti_mgcg_cgcg_init[] =
594{
595 0xc400, 0xffffffff, 0xfffffffc,
596 0x802c, 0xffffffff, 0xe0000000,
597 0x9a60, 0xffffffff, 0x00000100,
598 0x92a4, 0xffffffff, 0x00000100,
599 0xc164, 0xffffffff, 0x00000100,
600 0x9774, 0xffffffff, 0x00000100,
601 0x8984, 0xffffffff, 0x06000100,
602 0x8a18, 0xffffffff, 0x00000100,
603 0x92a0, 0xffffffff, 0x00000100,
604 0xc380, 0xffffffff, 0x00000100,
605 0x8b28, 0xffffffff, 0x00000100,
606 0x9144, 0xffffffff, 0x00000100,
607 0x8d88, 0xffffffff, 0x00000100,
608 0x8d8c, 0xffffffff, 0x00000100,
609 0x9030, 0xffffffff, 0x00000100,
610 0x9034, 0xffffffff, 0x00000100,
611 0x9038, 0xffffffff, 0x00000100,
612 0x903c, 0xffffffff, 0x00000100,
613 0xad80, 0xffffffff, 0x00000100,
614 0xac54, 0xffffffff, 0x00000100,
615 0x897c, 0xffffffff, 0x06000100,
616 0x9868, 0xffffffff, 0x00000100,
617 0x9510, 0xffffffff, 0x00000100,
618 0xaf04, 0xffffffff, 0x00000100,
619 0xae04, 0xffffffff, 0x00000100,
620 0x949c, 0xffffffff, 0x00000100,
621 0x802c, 0xffffffff, 0xe0000000,
622 0x9160, 0xffffffff, 0x00010000,
623 0x9164, 0xffffffff, 0x00030002,
624 0x9168, 0xffffffff, 0x00040007,
625 0x916c, 0xffffffff, 0x00060005,
626 0x9170, 0xffffffff, 0x00090008,
627 0x9174, 0xffffffff, 0x00020001,
628 0x9178, 0xffffffff, 0x00040003,
629 0x917c, 0xffffffff, 0x00000007,
630 0x9180, 0xffffffff, 0x00060005,
631 0x9184, 0xffffffff, 0x00090008,
632 0x9188, 0xffffffff, 0x00030002,
633 0x918c, 0xffffffff, 0x00050004,
634 0x9190, 0xffffffff, 0x00000008,
635 0x9194, 0xffffffff, 0x00070006,
636 0x9198, 0xffffffff, 0x000a0009,
637 0x919c, 0xffffffff, 0x00040003,
638 0x91a0, 0xffffffff, 0x00060005,
639 0x91a4, 0xffffffff, 0x00000009,
640 0x91a8, 0xffffffff, 0x00080007,
641 0x91ac, 0xffffffff, 0x000b000a,
642 0x91b0, 0xffffffff, 0x00050004,
643 0x91b4, 0xffffffff, 0x00070006,
644 0x91b8, 0xffffffff, 0x0008000b,
645 0x91bc, 0xffffffff, 0x000a0009,
646 0x91c0, 0xffffffff, 0x000d000c,
647 0x91c4, 0xffffffff, 0x00060005,
648 0x91c8, 0xffffffff, 0x00080007,
649 0x91cc, 0xffffffff, 0x0000000b,
650 0x91d0, 0xffffffff, 0x000a0009,
651 0x91d4, 0xffffffff, 0x000d000c,
652 0x91d8, 0xffffffff, 0x00070006,
653 0x91dc, 0xffffffff, 0x00090008,
654 0x91e0, 0xffffffff, 0x0000000c,
655 0x91e4, 0xffffffff, 0x000b000a,
656 0x91e8, 0xffffffff, 0x000e000d,
657 0x91ec, 0xffffffff, 0x00080007,
658 0x91f0, 0xffffffff, 0x000a0009,
659 0x91f4, 0xffffffff, 0x0000000d,
660 0x91f8, 0xffffffff, 0x000c000b,
661 0x91fc, 0xffffffff, 0x000f000e,
662 0x9200, 0xffffffff, 0x00090008,
663 0x9204, 0xffffffff, 0x000b000a,
664 0x9208, 0xffffffff, 0x000c000f,
665 0x920c, 0xffffffff, 0x000e000d,
666 0x9210, 0xffffffff, 0x00110010,
667 0x9214, 0xffffffff, 0x000a0009,
668 0x9218, 0xffffffff, 0x000c000b,
669 0x921c, 0xffffffff, 0x0000000f,
670 0x9220, 0xffffffff, 0x000e000d,
671 0x9224, 0xffffffff, 0x00110010,
672 0x9228, 0xffffffff, 0x000b000a,
673 0x922c, 0xffffffff, 0x000d000c,
674 0x9230, 0xffffffff, 0x00000010,
675 0x9234, 0xffffffff, 0x000f000e,
676 0x9238, 0xffffffff, 0x00120011,
677 0x923c, 0xffffffff, 0x000c000b,
678 0x9240, 0xffffffff, 0x000e000d,
679 0x9244, 0xffffffff, 0x00000011,
680 0x9248, 0xffffffff, 0x0010000f,
681 0x924c, 0xffffffff, 0x00130012,
682 0x9250, 0xffffffff, 0x000d000c,
683 0x9254, 0xffffffff, 0x000f000e,
684 0x9258, 0xffffffff, 0x00100013,
685 0x925c, 0xffffffff, 0x00120011,
686 0x9260, 0xffffffff, 0x00150014,
687 0x9264, 0xffffffff, 0x000e000d,
688 0x9268, 0xffffffff, 0x0010000f,
689 0x926c, 0xffffffff, 0x00000013,
690 0x9270, 0xffffffff, 0x00120011,
691 0x9274, 0xffffffff, 0x00150014,
692 0x9278, 0xffffffff, 0x000f000e,
693 0x927c, 0xffffffff, 0x00110010,
694 0x9280, 0xffffffff, 0x00000014,
695 0x9284, 0xffffffff, 0x00130012,
696 0x9288, 0xffffffff, 0x00160015,
697 0x928c, 0xffffffff, 0x0010000f,
698 0x9290, 0xffffffff, 0x00120011,
699 0x9294, 0xffffffff, 0x00000015,
700 0x9298, 0xffffffff, 0x00140013,
701 0x929c, 0xffffffff, 0x00170016,
702 0x9150, 0xffffffff, 0x96940200,
703 0x8708, 0xffffffff, 0x00900100,
704 0xc478, 0xffffffff, 0x00000080,
705 0xc404, 0xffffffff, 0x0020003f,
706 0x30, 0xffffffff, 0x0000001c,
707 0x34, 0x000f0000, 0x000f0000,
708 0x160c, 0xffffffff, 0x00000100,
709 0x1024, 0xffffffff, 0x00000100,
710 0x102c, 0x00000101, 0x00000000,
711 0x20a8, 0xffffffff, 0x00000104,
712 0x264c, 0x000c0000, 0x000c0000,
713 0x2648, 0x000c0000, 0x000c0000,
714 0x55e4, 0xff000fff, 0x00000100,
715 0x55e8, 0x00000001, 0x00000001,
716 0x2f50, 0x00000001, 0x00000001,
717 0x30cc, 0xc0000fff, 0x00000104,
718 0xc1e4, 0x00000001, 0x00000001,
719 0xd0c0, 0xfffffff0, 0x00000100,
720 0xd8c0, 0xfffffff0, 0x00000100
721};
722
723static const u32 pitcairn_mgcg_cgcg_init[] =
724{
725 0xc400, 0xffffffff, 0xfffffffc,
726 0x802c, 0xffffffff, 0xe0000000,
727 0x9a60, 0xffffffff, 0x00000100,
728 0x92a4, 0xffffffff, 0x00000100,
729 0xc164, 0xffffffff, 0x00000100,
730 0x9774, 0xffffffff, 0x00000100,
731 0x8984, 0xffffffff, 0x06000100,
732 0x8a18, 0xffffffff, 0x00000100,
733 0x92a0, 0xffffffff, 0x00000100,
734 0xc380, 0xffffffff, 0x00000100,
735 0x8b28, 0xffffffff, 0x00000100,
736 0x9144, 0xffffffff, 0x00000100,
737 0x8d88, 0xffffffff, 0x00000100,
738 0x8d8c, 0xffffffff, 0x00000100,
739 0x9030, 0xffffffff, 0x00000100,
740 0x9034, 0xffffffff, 0x00000100,
741 0x9038, 0xffffffff, 0x00000100,
742 0x903c, 0xffffffff, 0x00000100,
743 0xad80, 0xffffffff, 0x00000100,
744 0xac54, 0xffffffff, 0x00000100,
745 0x897c, 0xffffffff, 0x06000100,
746 0x9868, 0xffffffff, 0x00000100,
747 0x9510, 0xffffffff, 0x00000100,
748 0xaf04, 0xffffffff, 0x00000100,
749 0xae04, 0xffffffff, 0x00000100,
750 0x949c, 0xffffffff, 0x00000100,
751 0x802c, 0xffffffff, 0xe0000000,
752 0x9160, 0xffffffff, 0x00010000,
753 0x9164, 0xffffffff, 0x00030002,
754 0x9168, 0xffffffff, 0x00040007,
755 0x916c, 0xffffffff, 0x00060005,
756 0x9170, 0xffffffff, 0x00090008,
757 0x9174, 0xffffffff, 0x00020001,
758 0x9178, 0xffffffff, 0x00040003,
759 0x917c, 0xffffffff, 0x00000007,
760 0x9180, 0xffffffff, 0x00060005,
761 0x9184, 0xffffffff, 0x00090008,
762 0x9188, 0xffffffff, 0x00030002,
763 0x918c, 0xffffffff, 0x00050004,
764 0x9190, 0xffffffff, 0x00000008,
765 0x9194, 0xffffffff, 0x00070006,
766 0x9198, 0xffffffff, 0x000a0009,
767 0x919c, 0xffffffff, 0x00040003,
768 0x91a0, 0xffffffff, 0x00060005,
769 0x91a4, 0xffffffff, 0x00000009,
770 0x91a8, 0xffffffff, 0x00080007,
771 0x91ac, 0xffffffff, 0x000b000a,
772 0x91b0, 0xffffffff, 0x00050004,
773 0x91b4, 0xffffffff, 0x00070006,
774 0x91b8, 0xffffffff, 0x0008000b,
775 0x91bc, 0xffffffff, 0x000a0009,
776 0x91c0, 0xffffffff, 0x000d000c,
777 0x9200, 0xffffffff, 0x00090008,
778 0x9204, 0xffffffff, 0x000b000a,
779 0x9208, 0xffffffff, 0x000c000f,
780 0x920c, 0xffffffff, 0x000e000d,
781 0x9210, 0xffffffff, 0x00110010,
782 0x9214, 0xffffffff, 0x000a0009,
783 0x9218, 0xffffffff, 0x000c000b,
784 0x921c, 0xffffffff, 0x0000000f,
785 0x9220, 0xffffffff, 0x000e000d,
786 0x9224, 0xffffffff, 0x00110010,
787 0x9228, 0xffffffff, 0x000b000a,
788 0x922c, 0xffffffff, 0x000d000c,
789 0x9230, 0xffffffff, 0x00000010,
790 0x9234, 0xffffffff, 0x000f000e,
791 0x9238, 0xffffffff, 0x00120011,
792 0x923c, 0xffffffff, 0x000c000b,
793 0x9240, 0xffffffff, 0x000e000d,
794 0x9244, 0xffffffff, 0x00000011,
795 0x9248, 0xffffffff, 0x0010000f,
796 0x924c, 0xffffffff, 0x00130012,
797 0x9250, 0xffffffff, 0x000d000c,
798 0x9254, 0xffffffff, 0x000f000e,
799 0x9258, 0xffffffff, 0x00100013,
800 0x925c, 0xffffffff, 0x00120011,
801 0x9260, 0xffffffff, 0x00150014,
802 0x9150, 0xffffffff, 0x96940200,
803 0x8708, 0xffffffff, 0x00900100,
804 0xc478, 0xffffffff, 0x00000080,
805 0xc404, 0xffffffff, 0x0020003f,
806 0x30, 0xffffffff, 0x0000001c,
807 0x34, 0x000f0000, 0x000f0000,
808 0x160c, 0xffffffff, 0x00000100,
809 0x1024, 0xffffffff, 0x00000100,
810 0x102c, 0x00000101, 0x00000000,
811 0x20a8, 0xffffffff, 0x00000104,
812 0x55e4, 0xff000fff, 0x00000100,
813 0x55e8, 0x00000001, 0x00000001,
814 0x2f50, 0x00000001, 0x00000001,
815 0x30cc, 0xc0000fff, 0x00000104,
816 0xc1e4, 0x00000001, 0x00000001,
817 0xd0c0, 0xfffffff0, 0x00000100,
818 0xd8c0, 0xfffffff0, 0x00000100
819};
820
821static const u32 verde_mgcg_cgcg_init[] =
822{
823 0xc400, 0xffffffff, 0xfffffffc,
824 0x802c, 0xffffffff, 0xe0000000,
825 0x9a60, 0xffffffff, 0x00000100,
826 0x92a4, 0xffffffff, 0x00000100,
827 0xc164, 0xffffffff, 0x00000100,
828 0x9774, 0xffffffff, 0x00000100,
829 0x8984, 0xffffffff, 0x06000100,
830 0x8a18, 0xffffffff, 0x00000100,
831 0x92a0, 0xffffffff, 0x00000100,
832 0xc380, 0xffffffff, 0x00000100,
833 0x8b28, 0xffffffff, 0x00000100,
834 0x9144, 0xffffffff, 0x00000100,
835 0x8d88, 0xffffffff, 0x00000100,
836 0x8d8c, 0xffffffff, 0x00000100,
837 0x9030, 0xffffffff, 0x00000100,
838 0x9034, 0xffffffff, 0x00000100,
839 0x9038, 0xffffffff, 0x00000100,
840 0x903c, 0xffffffff, 0x00000100,
841 0xad80, 0xffffffff, 0x00000100,
842 0xac54, 0xffffffff, 0x00000100,
843 0x897c, 0xffffffff, 0x06000100,
844 0x9868, 0xffffffff, 0x00000100,
845 0x9510, 0xffffffff, 0x00000100,
846 0xaf04, 0xffffffff, 0x00000100,
847 0xae04, 0xffffffff, 0x00000100,
848 0x949c, 0xffffffff, 0x00000100,
849 0x802c, 0xffffffff, 0xe0000000,
850 0x9160, 0xffffffff, 0x00010000,
851 0x9164, 0xffffffff, 0x00030002,
852 0x9168, 0xffffffff, 0x00040007,
853 0x916c, 0xffffffff, 0x00060005,
854 0x9170, 0xffffffff, 0x00090008,
855 0x9174, 0xffffffff, 0x00020001,
856 0x9178, 0xffffffff, 0x00040003,
857 0x917c, 0xffffffff, 0x00000007,
858 0x9180, 0xffffffff, 0x00060005,
859 0x9184, 0xffffffff, 0x00090008,
860 0x9188, 0xffffffff, 0x00030002,
861 0x918c, 0xffffffff, 0x00050004,
862 0x9190, 0xffffffff, 0x00000008,
863 0x9194, 0xffffffff, 0x00070006,
864 0x9198, 0xffffffff, 0x000a0009,
865 0x919c, 0xffffffff, 0x00040003,
866 0x91a0, 0xffffffff, 0x00060005,
867 0x91a4, 0xffffffff, 0x00000009,
868 0x91a8, 0xffffffff, 0x00080007,
869 0x91ac, 0xffffffff, 0x000b000a,
870 0x91b0, 0xffffffff, 0x00050004,
871 0x91b4, 0xffffffff, 0x00070006,
872 0x91b8, 0xffffffff, 0x0008000b,
873 0x91bc, 0xffffffff, 0x000a0009,
874 0x91c0, 0xffffffff, 0x000d000c,
875 0x9200, 0xffffffff, 0x00090008,
876 0x9204, 0xffffffff, 0x000b000a,
877 0x9208, 0xffffffff, 0x000c000f,
878 0x920c, 0xffffffff, 0x000e000d,
879 0x9210, 0xffffffff, 0x00110010,
880 0x9214, 0xffffffff, 0x000a0009,
881 0x9218, 0xffffffff, 0x000c000b,
882 0x921c, 0xffffffff, 0x0000000f,
883 0x9220, 0xffffffff, 0x000e000d,
884 0x9224, 0xffffffff, 0x00110010,
885 0x9228, 0xffffffff, 0x000b000a,
886 0x922c, 0xffffffff, 0x000d000c,
887 0x9230, 0xffffffff, 0x00000010,
888 0x9234, 0xffffffff, 0x000f000e,
889 0x9238, 0xffffffff, 0x00120011,
890 0x923c, 0xffffffff, 0x000c000b,
891 0x9240, 0xffffffff, 0x000e000d,
892 0x9244, 0xffffffff, 0x00000011,
893 0x9248, 0xffffffff, 0x0010000f,
894 0x924c, 0xffffffff, 0x00130012,
895 0x9250, 0xffffffff, 0x000d000c,
896 0x9254, 0xffffffff, 0x000f000e,
897 0x9258, 0xffffffff, 0x00100013,
898 0x925c, 0xffffffff, 0x00120011,
899 0x9260, 0xffffffff, 0x00150014,
900 0x9150, 0xffffffff, 0x96940200,
901 0x8708, 0xffffffff, 0x00900100,
902 0xc478, 0xffffffff, 0x00000080,
903 0xc404, 0xffffffff, 0x0020003f,
904 0x30, 0xffffffff, 0x0000001c,
905 0x34, 0x000f0000, 0x000f0000,
906 0x160c, 0xffffffff, 0x00000100,
907 0x1024, 0xffffffff, 0x00000100,
908 0x102c, 0x00000101, 0x00000000,
909 0x20a8, 0xffffffff, 0x00000104,
910 0x264c, 0x000c0000, 0x000c0000,
911 0x2648, 0x000c0000, 0x000c0000,
912 0x55e4, 0xff000fff, 0x00000100,
913 0x55e8, 0x00000001, 0x00000001,
914 0x2f50, 0x00000001, 0x00000001,
915 0x30cc, 0xc0000fff, 0x00000104,
916 0xc1e4, 0x00000001, 0x00000001,
917 0xd0c0, 0xfffffff0, 0x00000100,
918 0xd8c0, 0xfffffff0, 0x00000100
919};
920
921static const u32 oland_mgcg_cgcg_init[] =
922{
923 0xc400, 0xffffffff, 0xfffffffc,
924 0x802c, 0xffffffff, 0xe0000000,
925 0x9a60, 0xffffffff, 0x00000100,
926 0x92a4, 0xffffffff, 0x00000100,
927 0xc164, 0xffffffff, 0x00000100,
928 0x9774, 0xffffffff, 0x00000100,
929 0x8984, 0xffffffff, 0x06000100,
930 0x8a18, 0xffffffff, 0x00000100,
931 0x92a0, 0xffffffff, 0x00000100,
932 0xc380, 0xffffffff, 0x00000100,
933 0x8b28, 0xffffffff, 0x00000100,
934 0x9144, 0xffffffff, 0x00000100,
935 0x8d88, 0xffffffff, 0x00000100,
936 0x8d8c, 0xffffffff, 0x00000100,
937 0x9030, 0xffffffff, 0x00000100,
938 0x9034, 0xffffffff, 0x00000100,
939 0x9038, 0xffffffff, 0x00000100,
940 0x903c, 0xffffffff, 0x00000100,
941 0xad80, 0xffffffff, 0x00000100,
942 0xac54, 0xffffffff, 0x00000100,
943 0x897c, 0xffffffff, 0x06000100,
944 0x9868, 0xffffffff, 0x00000100,
945 0x9510, 0xffffffff, 0x00000100,
946 0xaf04, 0xffffffff, 0x00000100,
947 0xae04, 0xffffffff, 0x00000100,
948 0x949c, 0xffffffff, 0x00000100,
949 0x802c, 0xffffffff, 0xe0000000,
950 0x9160, 0xffffffff, 0x00010000,
951 0x9164, 0xffffffff, 0x00030002,
952 0x9168, 0xffffffff, 0x00040007,
953 0x916c, 0xffffffff, 0x00060005,
954 0x9170, 0xffffffff, 0x00090008,
955 0x9174, 0xffffffff, 0x00020001,
956 0x9178, 0xffffffff, 0x00040003,
957 0x917c, 0xffffffff, 0x00000007,
958 0x9180, 0xffffffff, 0x00060005,
959 0x9184, 0xffffffff, 0x00090008,
960 0x9188, 0xffffffff, 0x00030002,
961 0x918c, 0xffffffff, 0x00050004,
962 0x9190, 0xffffffff, 0x00000008,
963 0x9194, 0xffffffff, 0x00070006,
964 0x9198, 0xffffffff, 0x000a0009,
965 0x919c, 0xffffffff, 0x00040003,
966 0x91a0, 0xffffffff, 0x00060005,
967 0x91a4, 0xffffffff, 0x00000009,
968 0x91a8, 0xffffffff, 0x00080007,
969 0x91ac, 0xffffffff, 0x000b000a,
970 0x91b0, 0xffffffff, 0x00050004,
971 0x91b4, 0xffffffff, 0x00070006,
972 0x91b8, 0xffffffff, 0x0008000b,
973 0x91bc, 0xffffffff, 0x000a0009,
974 0x91c0, 0xffffffff, 0x000d000c,
975 0x91c4, 0xffffffff, 0x00060005,
976 0x91c8, 0xffffffff, 0x00080007,
977 0x91cc, 0xffffffff, 0x0000000b,
978 0x91d0, 0xffffffff, 0x000a0009,
979 0x91d4, 0xffffffff, 0x000d000c,
980 0x9150, 0xffffffff, 0x96940200,
981 0x8708, 0xffffffff, 0x00900100,
982 0xc478, 0xffffffff, 0x00000080,
983 0xc404, 0xffffffff, 0x0020003f,
984 0x30, 0xffffffff, 0x0000001c,
985 0x34, 0x000f0000, 0x000f0000,
986 0x160c, 0xffffffff, 0x00000100,
987 0x1024, 0xffffffff, 0x00000100,
988 0x102c, 0x00000101, 0x00000000,
989 0x20a8, 0xffffffff, 0x00000104,
990 0x264c, 0x000c0000, 0x000c0000,
991 0x2648, 0x000c0000, 0x000c0000,
992 0x55e4, 0xff000fff, 0x00000100,
993 0x55e8, 0x00000001, 0x00000001,
994 0x2f50, 0x00000001, 0x00000001,
995 0x30cc, 0xc0000fff, 0x00000104,
996 0xc1e4, 0x00000001, 0x00000001,
997 0xd0c0, 0xfffffff0, 0x00000100,
998 0xd8c0, 0xfffffff0, 0x00000100
999};
1000
1001static const u32 hainan_mgcg_cgcg_init[] =
1002{
1003 0xc400, 0xffffffff, 0xfffffffc,
1004 0x802c, 0xffffffff, 0xe0000000,
1005 0x9a60, 0xffffffff, 0x00000100,
1006 0x92a4, 0xffffffff, 0x00000100,
1007 0xc164, 0xffffffff, 0x00000100,
1008 0x9774, 0xffffffff, 0x00000100,
1009 0x8984, 0xffffffff, 0x06000100,
1010 0x8a18, 0xffffffff, 0x00000100,
1011 0x92a0, 0xffffffff, 0x00000100,
1012 0xc380, 0xffffffff, 0x00000100,
1013 0x8b28, 0xffffffff, 0x00000100,
1014 0x9144, 0xffffffff, 0x00000100,
1015 0x8d88, 0xffffffff, 0x00000100,
1016 0x8d8c, 0xffffffff, 0x00000100,
1017 0x9030, 0xffffffff, 0x00000100,
1018 0x9034, 0xffffffff, 0x00000100,
1019 0x9038, 0xffffffff, 0x00000100,
1020 0x903c, 0xffffffff, 0x00000100,
1021 0xad80, 0xffffffff, 0x00000100,
1022 0xac54, 0xffffffff, 0x00000100,
1023 0x897c, 0xffffffff, 0x06000100,
1024 0x9868, 0xffffffff, 0x00000100,
1025 0x9510, 0xffffffff, 0x00000100,
1026 0xaf04, 0xffffffff, 0x00000100,
1027 0xae04, 0xffffffff, 0x00000100,
1028 0x949c, 0xffffffff, 0x00000100,
1029 0x802c, 0xffffffff, 0xe0000000,
1030 0x9160, 0xffffffff, 0x00010000,
1031 0x9164, 0xffffffff, 0x00030002,
1032 0x9168, 0xffffffff, 0x00040007,
1033 0x916c, 0xffffffff, 0x00060005,
1034 0x9170, 0xffffffff, 0x00090008,
1035 0x9174, 0xffffffff, 0x00020001,
1036 0x9178, 0xffffffff, 0x00040003,
1037 0x917c, 0xffffffff, 0x00000007,
1038 0x9180, 0xffffffff, 0x00060005,
1039 0x9184, 0xffffffff, 0x00090008,
1040 0x9188, 0xffffffff, 0x00030002,
1041 0x918c, 0xffffffff, 0x00050004,
1042 0x9190, 0xffffffff, 0x00000008,
1043 0x9194, 0xffffffff, 0x00070006,
1044 0x9198, 0xffffffff, 0x000a0009,
1045 0x919c, 0xffffffff, 0x00040003,
1046 0x91a0, 0xffffffff, 0x00060005,
1047 0x91a4, 0xffffffff, 0x00000009,
1048 0x91a8, 0xffffffff, 0x00080007,
1049 0x91ac, 0xffffffff, 0x000b000a,
1050 0x91b0, 0xffffffff, 0x00050004,
1051 0x91b4, 0xffffffff, 0x00070006,
1052 0x91b8, 0xffffffff, 0x0008000b,
1053 0x91bc, 0xffffffff, 0x000a0009,
1054 0x91c0, 0xffffffff, 0x000d000c,
1055 0x91c4, 0xffffffff, 0x00060005,
1056 0x91c8, 0xffffffff, 0x00080007,
1057 0x91cc, 0xffffffff, 0x0000000b,
1058 0x91d0, 0xffffffff, 0x000a0009,
1059 0x91d4, 0xffffffff, 0x000d000c,
1060 0x9150, 0xffffffff, 0x96940200,
1061 0x8708, 0xffffffff, 0x00900100,
1062 0xc478, 0xffffffff, 0x00000080,
1063 0xc404, 0xffffffff, 0x0020003f,
1064 0x30, 0xffffffff, 0x0000001c,
1065 0x34, 0x000f0000, 0x000f0000,
1066 0x160c, 0xffffffff, 0x00000100,
1067 0x1024, 0xffffffff, 0x00000100,
1068 0x20a8, 0xffffffff, 0x00000104,
1069 0x264c, 0x000c0000, 0x000c0000,
1070 0x2648, 0x000c0000, 0x000c0000,
1071 0x2f50, 0x00000001, 0x00000001,
1072 0x30cc, 0xc0000fff, 0x00000104,
1073 0xc1e4, 0x00000001, 0x00000001,
1074 0xd0c0, 0xfffffff0, 0x00000100,
1075 0xd8c0, 0xfffffff0, 0x00000100
1076};
1077
1078static u32 verde_pg_init[] =
1079{
1080 0x353c, 0xffffffff, 0x40000,
1081 0x3538, 0xffffffff, 0x200010ff,
1082 0x353c, 0xffffffff, 0x0,
1083 0x353c, 0xffffffff, 0x0,
1084 0x353c, 0xffffffff, 0x0,
1085 0x353c, 0xffffffff, 0x0,
1086 0x353c, 0xffffffff, 0x0,
1087 0x353c, 0xffffffff, 0x7007,
1088 0x3538, 0xffffffff, 0x300010ff,
1089 0x353c, 0xffffffff, 0x0,
1090 0x353c, 0xffffffff, 0x0,
1091 0x353c, 0xffffffff, 0x0,
1092 0x353c, 0xffffffff, 0x0,
1093 0x353c, 0xffffffff, 0x0,
1094 0x353c, 0xffffffff, 0x400000,
1095 0x3538, 0xffffffff, 0x100010ff,
1096 0x353c, 0xffffffff, 0x0,
1097 0x353c, 0xffffffff, 0x0,
1098 0x353c, 0xffffffff, 0x0,
1099 0x353c, 0xffffffff, 0x0,
1100 0x353c, 0xffffffff, 0x0,
1101 0x353c, 0xffffffff, 0x120200,
1102 0x3538, 0xffffffff, 0x500010ff,
1103 0x353c, 0xffffffff, 0x0,
1104 0x353c, 0xffffffff, 0x0,
1105 0x353c, 0xffffffff, 0x0,
1106 0x353c, 0xffffffff, 0x0,
1107 0x353c, 0xffffffff, 0x0,
1108 0x353c, 0xffffffff, 0x1e1e16,
1109 0x3538, 0xffffffff, 0x600010ff,
1110 0x353c, 0xffffffff, 0x0,
1111 0x353c, 0xffffffff, 0x0,
1112 0x353c, 0xffffffff, 0x0,
1113 0x353c, 0xffffffff, 0x0,
1114 0x353c, 0xffffffff, 0x0,
1115 0x353c, 0xffffffff, 0x171f1e,
1116 0x3538, 0xffffffff, 0x700010ff,
1117 0x353c, 0xffffffff, 0x0,
1118 0x353c, 0xffffffff, 0x0,
1119 0x353c, 0xffffffff, 0x0,
1120 0x353c, 0xffffffff, 0x0,
1121 0x353c, 0xffffffff, 0x0,
1122 0x353c, 0xffffffff, 0x0,
1123 0x3538, 0xffffffff, 0x9ff,
1124 0x3500, 0xffffffff, 0x0,
1125 0x3504, 0xffffffff, 0x10000800,
1126 0x3504, 0xffffffff, 0xf,
1127 0x3504, 0xffffffff, 0xf,
1128 0x3500, 0xffffffff, 0x4,
1129 0x3504, 0xffffffff, 0x1000051e,
1130 0x3504, 0xffffffff, 0xffff,
1131 0x3504, 0xffffffff, 0xffff,
1132 0x3500, 0xffffffff, 0x8,
1133 0x3504, 0xffffffff, 0x80500,
1134 0x3500, 0xffffffff, 0x12,
1135 0x3504, 0xffffffff, 0x9050c,
1136 0x3500, 0xffffffff, 0x1d,
1137 0x3504, 0xffffffff, 0xb052c,
1138 0x3500, 0xffffffff, 0x2a,
1139 0x3504, 0xffffffff, 0x1053e,
1140 0x3500, 0xffffffff, 0x2d,
1141 0x3504, 0xffffffff, 0x10546,
1142 0x3500, 0xffffffff, 0x30,
1143 0x3504, 0xffffffff, 0xa054e,
1144 0x3500, 0xffffffff, 0x3c,
1145 0x3504, 0xffffffff, 0x1055f,
1146 0x3500, 0xffffffff, 0x3f,
1147 0x3504, 0xffffffff, 0x10567,
1148 0x3500, 0xffffffff, 0x42,
1149 0x3504, 0xffffffff, 0x1056f,
1150 0x3500, 0xffffffff, 0x45,
1151 0x3504, 0xffffffff, 0x10572,
1152 0x3500, 0xffffffff, 0x48,
1153 0x3504, 0xffffffff, 0x20575,
1154 0x3500, 0xffffffff, 0x4c,
1155 0x3504, 0xffffffff, 0x190801,
1156 0x3500, 0xffffffff, 0x67,
1157 0x3504, 0xffffffff, 0x1082a,
1158 0x3500, 0xffffffff, 0x6a,
1159 0x3504, 0xffffffff, 0x1b082d,
1160 0x3500, 0xffffffff, 0x87,
1161 0x3504, 0xffffffff, 0x310851,
1162 0x3500, 0xffffffff, 0xba,
1163 0x3504, 0xffffffff, 0x891,
1164 0x3500, 0xffffffff, 0xbc,
1165 0x3504, 0xffffffff, 0x893,
1166 0x3500, 0xffffffff, 0xbe,
1167 0x3504, 0xffffffff, 0x20895,
1168 0x3500, 0xffffffff, 0xc2,
1169 0x3504, 0xffffffff, 0x20899,
1170 0x3500, 0xffffffff, 0xc6,
1171 0x3504, 0xffffffff, 0x2089d,
1172 0x3500, 0xffffffff, 0xca,
1173 0x3504, 0xffffffff, 0x8a1,
1174 0x3500, 0xffffffff, 0xcc,
1175 0x3504, 0xffffffff, 0x8a3,
1176 0x3500, 0xffffffff, 0xce,
1177 0x3504, 0xffffffff, 0x308a5,
1178 0x3500, 0xffffffff, 0xd3,
1179 0x3504, 0xffffffff, 0x6d08cd,
1180 0x3500, 0xffffffff, 0x142,
1181 0x3504, 0xffffffff, 0x2000095a,
1182 0x3504, 0xffffffff, 0x1,
1183 0x3500, 0xffffffff, 0x144,
1184 0x3504, 0xffffffff, 0x301f095b,
1185 0x3500, 0xffffffff, 0x165,
1186 0x3504, 0xffffffff, 0xc094d,
1187 0x3500, 0xffffffff, 0x173,
1188 0x3504, 0xffffffff, 0xf096d,
1189 0x3500, 0xffffffff, 0x184,
1190 0x3504, 0xffffffff, 0x15097f,
1191 0x3500, 0xffffffff, 0x19b,
1192 0x3504, 0xffffffff, 0xc0998,
1193 0x3500, 0xffffffff, 0x1a9,
1194 0x3504, 0xffffffff, 0x409a7,
1195 0x3500, 0xffffffff, 0x1af,
1196 0x3504, 0xffffffff, 0xcdc,
1197 0x3500, 0xffffffff, 0x1b1,
1198 0x3504, 0xffffffff, 0x800,
1199 0x3508, 0xffffffff, 0x6c9b2000,
1200 0x3510, 0xfc00, 0x2000,
1201 0x3544, 0xffffffff, 0xfc0,
1202 0x28d4, 0x00000100, 0x100
1203};
1204
1205static void si_init_golden_registers(struct radeon_device *rdev)
1206{
1207 switch (rdev->family) {
1208 case CHIP_TAHITI:
1209 radeon_program_register_sequence(rdev,
1210 tahiti_golden_registers,
1211 (const u32)ARRAY_SIZE(tahiti_golden_registers));
1212 radeon_program_register_sequence(rdev,
1213 tahiti_golden_rlc_registers,
1214 (const u32)ARRAY_SIZE(tahiti_golden_rlc_registers));
1215 radeon_program_register_sequence(rdev,
1216 tahiti_mgcg_cgcg_init,
1217 (const u32)ARRAY_SIZE(tahiti_mgcg_cgcg_init));
1218 radeon_program_register_sequence(rdev,
1219 tahiti_golden_registers2,
1220 (const u32)ARRAY_SIZE(tahiti_golden_registers2));
1221 break;
1222 case CHIP_PITCAIRN:
1223 radeon_program_register_sequence(rdev,
1224 pitcairn_golden_registers,
1225 (const u32)ARRAY_SIZE(pitcairn_golden_registers));
1226 radeon_program_register_sequence(rdev,
1227 pitcairn_golden_rlc_registers,
1228 (const u32)ARRAY_SIZE(pitcairn_golden_rlc_registers));
1229 radeon_program_register_sequence(rdev,
1230 pitcairn_mgcg_cgcg_init,
1231 (const u32)ARRAY_SIZE(pitcairn_mgcg_cgcg_init));
1232 break;
1233 case CHIP_VERDE:
1234 radeon_program_register_sequence(rdev,
1235 verde_golden_registers,
1236 (const u32)ARRAY_SIZE(verde_golden_registers));
1237 radeon_program_register_sequence(rdev,
1238 verde_golden_rlc_registers,
1239 (const u32)ARRAY_SIZE(verde_golden_rlc_registers));
1240 radeon_program_register_sequence(rdev,
1241 verde_mgcg_cgcg_init,
1242 (const u32)ARRAY_SIZE(verde_mgcg_cgcg_init));
1243 radeon_program_register_sequence(rdev,
1244 verde_pg_init,
1245 (const u32)ARRAY_SIZE(verde_pg_init));
1246 break;
1247 case CHIP_OLAND:
1248 radeon_program_register_sequence(rdev,
1249 oland_golden_registers,
1250 (const u32)ARRAY_SIZE(oland_golden_registers));
1251 radeon_program_register_sequence(rdev,
1252 oland_golden_rlc_registers,
1253 (const u32)ARRAY_SIZE(oland_golden_rlc_registers));
1254 radeon_program_register_sequence(rdev,
1255 oland_mgcg_cgcg_init,
1256 (const u32)ARRAY_SIZE(oland_mgcg_cgcg_init));
1257 break;
1258 case CHIP_HAINAN:
1259 radeon_program_register_sequence(rdev,
1260 hainan_golden_registers,
1261 (const u32)ARRAY_SIZE(hainan_golden_registers));
1262 radeon_program_register_sequence(rdev,
1263 hainan_golden_registers2,
1264 (const u32)ARRAY_SIZE(hainan_golden_registers2));
1265 radeon_program_register_sequence(rdev,
1266 hainan_mgcg_cgcg_init,
1267 (const u32)ARRAY_SIZE(hainan_mgcg_cgcg_init));
1268 break;
1269 default:
1270 break;
1271 }
1272}
1273
1274/**
1275 * si_get_allowed_info_register - fetch the register for the info ioctl
1276 *
1277 * @rdev: radeon_device pointer
1278 * @reg: register offset in bytes
1279 * @val: register value
1280 *
1281 * Returns 0 for success or -EINVAL for an invalid register
1282 *
1283 */
1284int si_get_allowed_info_register(struct radeon_device *rdev,
1285 u32 reg, u32 *val)
1286{
1287 switch (reg) {
1288 case GRBM_STATUS:
1289 case GRBM_STATUS2:
1290 case GRBM_STATUS_SE0:
1291 case GRBM_STATUS_SE1:
1292 case SRBM_STATUS:
1293 case SRBM_STATUS2:
1294 case (DMA_STATUS_REG + DMA0_REGISTER_OFFSET):
1295 case (DMA_STATUS_REG + DMA1_REGISTER_OFFSET):
1296 case UVD_STATUS:
1297 *val = RREG32(reg);
1298 return 0;
1299 default:
1300 return -EINVAL;
1301 }
1302}
1303
1304#define PCIE_BUS_CLK 10000
1305#define TCLK (PCIE_BUS_CLK / 10)
1306
1307/**
1308 * si_get_xclk - get the xclk
1309 *
1310 * @rdev: radeon_device pointer
1311 *
1312 * Returns the reference clock used by the gfx engine
1313 * (SI).
1314 */
1315u32 si_get_xclk(struct radeon_device *rdev)
1316{
1317 u32 reference_clock = rdev->clock.spll.reference_freq;
1318 u32 tmp;
1319
1320 tmp = RREG32(CG_CLKPIN_CNTL_2);
1321 if (tmp & MUX_TCLK_TO_XCLK)
1322 return TCLK;
1323
1324 tmp = RREG32(CG_CLKPIN_CNTL);
1325 if (tmp & XTALIN_DIVIDE)
1326 return reference_clock / 4;
1327
1328 return reference_clock;
1329}
1330
1331/* get temperature in millidegrees */
1332int si_get_temp(struct radeon_device *rdev)
1333{
1334 u32 temp;
1335 int actual_temp = 0;
1336
1337 temp = (RREG32(CG_MULT_THERMAL_STATUS) & CTF_TEMP_MASK) >>
1338 CTF_TEMP_SHIFT;
1339
1340 if (temp & 0x200)
1341 actual_temp = 255;
1342 else
1343 actual_temp = temp & 0x1ff;
1344
1345 actual_temp = (actual_temp * 1000);
1346
1347 return actual_temp;
1348}
1349
1350#define TAHITI_IO_MC_REGS_SIZE 36
1351
1352static const u32 tahiti_io_mc_regs[TAHITI_IO_MC_REGS_SIZE][2] = {
1353 {0x0000006f, 0x03044000},
1354 {0x00000070, 0x0480c018},
1355 {0x00000071, 0x00000040},
1356 {0x00000072, 0x01000000},
1357 {0x00000074, 0x000000ff},
1358 {0x00000075, 0x00143400},
1359 {0x00000076, 0x08ec0800},
1360 {0x00000077, 0x040000cc},
1361 {0x00000079, 0x00000000},
1362 {0x0000007a, 0x21000409},
1363 {0x0000007c, 0x00000000},
1364 {0x0000007d, 0xe8000000},
1365 {0x0000007e, 0x044408a8},
1366 {0x0000007f, 0x00000003},
1367 {0x00000080, 0x00000000},
1368 {0x00000081, 0x01000000},
1369 {0x00000082, 0x02000000},
1370 {0x00000083, 0x00000000},
1371 {0x00000084, 0xe3f3e4f4},
1372 {0x00000085, 0x00052024},
1373 {0x00000087, 0x00000000},
1374 {0x00000088, 0x66036603},
1375 {0x00000089, 0x01000000},
1376 {0x0000008b, 0x1c0a0000},
1377 {0x0000008c, 0xff010000},
1378 {0x0000008e, 0xffffefff},
1379 {0x0000008f, 0xfff3efff},
1380 {0x00000090, 0xfff3efbf},
1381 {0x00000094, 0x00101101},
1382 {0x00000095, 0x00000fff},
1383 {0x00000096, 0x00116fff},
1384 {0x00000097, 0x60010000},
1385 {0x00000098, 0x10010000},
1386 {0x00000099, 0x00006000},
1387 {0x0000009a, 0x00001000},
1388 {0x0000009f, 0x00a77400}
1389};
1390
1391static const u32 pitcairn_io_mc_regs[TAHITI_IO_MC_REGS_SIZE][2] = {
1392 {0x0000006f, 0x03044000},
1393 {0x00000070, 0x0480c018},
1394 {0x00000071, 0x00000040},
1395 {0x00000072, 0x01000000},
1396 {0x00000074, 0x000000ff},
1397 {0x00000075, 0x00143400},
1398 {0x00000076, 0x08ec0800},
1399 {0x00000077, 0x040000cc},
1400 {0x00000079, 0x00000000},
1401 {0x0000007a, 0x21000409},
1402 {0x0000007c, 0x00000000},
1403 {0x0000007d, 0xe8000000},
1404 {0x0000007e, 0x044408a8},
1405 {0x0000007f, 0x00000003},
1406 {0x00000080, 0x00000000},
1407 {0x00000081, 0x01000000},
1408 {0x00000082, 0x02000000},
1409 {0x00000083, 0x00000000},
1410 {0x00000084, 0xe3f3e4f4},
1411 {0x00000085, 0x00052024},
1412 {0x00000087, 0x00000000},
1413 {0x00000088, 0x66036603},
1414 {0x00000089, 0x01000000},
1415 {0x0000008b, 0x1c0a0000},
1416 {0x0000008c, 0xff010000},
1417 {0x0000008e, 0xffffefff},
1418 {0x0000008f, 0xfff3efff},
1419 {0x00000090, 0xfff3efbf},
1420 {0x00000094, 0x00101101},
1421 {0x00000095, 0x00000fff},
1422 {0x00000096, 0x00116fff},
1423 {0x00000097, 0x60010000},
1424 {0x00000098, 0x10010000},
1425 {0x00000099, 0x00006000},
1426 {0x0000009a, 0x00001000},
1427 {0x0000009f, 0x00a47400}
1428};
1429
1430static const u32 verde_io_mc_regs[TAHITI_IO_MC_REGS_SIZE][2] = {
1431 {0x0000006f, 0x03044000},
1432 {0x00000070, 0x0480c018},
1433 {0x00000071, 0x00000040},
1434 {0x00000072, 0x01000000},
1435 {0x00000074, 0x000000ff},
1436 {0x00000075, 0x00143400},
1437 {0x00000076, 0x08ec0800},
1438 {0x00000077, 0x040000cc},
1439 {0x00000079, 0x00000000},
1440 {0x0000007a, 0x21000409},
1441 {0x0000007c, 0x00000000},
1442 {0x0000007d, 0xe8000000},
1443 {0x0000007e, 0x044408a8},
1444 {0x0000007f, 0x00000003},
1445 {0x00000080, 0x00000000},
1446 {0x00000081, 0x01000000},
1447 {0x00000082, 0x02000000},
1448 {0x00000083, 0x00000000},
1449 {0x00000084, 0xe3f3e4f4},
1450 {0x00000085, 0x00052024},
1451 {0x00000087, 0x00000000},
1452 {0x00000088, 0x66036603},
1453 {0x00000089, 0x01000000},
1454 {0x0000008b, 0x1c0a0000},
1455 {0x0000008c, 0xff010000},
1456 {0x0000008e, 0xffffefff},
1457 {0x0000008f, 0xfff3efff},
1458 {0x00000090, 0xfff3efbf},
1459 {0x00000094, 0x00101101},
1460 {0x00000095, 0x00000fff},
1461 {0x00000096, 0x00116fff},
1462 {0x00000097, 0x60010000},
1463 {0x00000098, 0x10010000},
1464 {0x00000099, 0x00006000},
1465 {0x0000009a, 0x00001000},
1466 {0x0000009f, 0x00a37400}
1467};
1468
1469static const u32 oland_io_mc_regs[TAHITI_IO_MC_REGS_SIZE][2] = {
1470 {0x0000006f, 0x03044000},
1471 {0x00000070, 0x0480c018},
1472 {0x00000071, 0x00000040},
1473 {0x00000072, 0x01000000},
1474 {0x00000074, 0x000000ff},
1475 {0x00000075, 0x00143400},
1476 {0x00000076, 0x08ec0800},
1477 {0x00000077, 0x040000cc},
1478 {0x00000079, 0x00000000},
1479 {0x0000007a, 0x21000409},
1480 {0x0000007c, 0x00000000},
1481 {0x0000007d, 0xe8000000},
1482 {0x0000007e, 0x044408a8},
1483 {0x0000007f, 0x00000003},
1484 {0x00000080, 0x00000000},
1485 {0x00000081, 0x01000000},
1486 {0x00000082, 0x02000000},
1487 {0x00000083, 0x00000000},
1488 {0x00000084, 0xe3f3e4f4},
1489 {0x00000085, 0x00052024},
1490 {0x00000087, 0x00000000},
1491 {0x00000088, 0x66036603},
1492 {0x00000089, 0x01000000},
1493 {0x0000008b, 0x1c0a0000},
1494 {0x0000008c, 0xff010000},
1495 {0x0000008e, 0xffffefff},
1496 {0x0000008f, 0xfff3efff},
1497 {0x00000090, 0xfff3efbf},
1498 {0x00000094, 0x00101101},
1499 {0x00000095, 0x00000fff},
1500 {0x00000096, 0x00116fff},
1501 {0x00000097, 0x60010000},
1502 {0x00000098, 0x10010000},
1503 {0x00000099, 0x00006000},
1504 {0x0000009a, 0x00001000},
1505 {0x0000009f, 0x00a17730}
1506};
1507
1508static const u32 hainan_io_mc_regs[TAHITI_IO_MC_REGS_SIZE][2] = {
1509 {0x0000006f, 0x03044000},
1510 {0x00000070, 0x0480c018},
1511 {0x00000071, 0x00000040},
1512 {0x00000072, 0x01000000},
1513 {0x00000074, 0x000000ff},
1514 {0x00000075, 0x00143400},
1515 {0x00000076, 0x08ec0800},
1516 {0x00000077, 0x040000cc},
1517 {0x00000079, 0x00000000},
1518 {0x0000007a, 0x21000409},
1519 {0x0000007c, 0x00000000},
1520 {0x0000007d, 0xe8000000},
1521 {0x0000007e, 0x044408a8},
1522 {0x0000007f, 0x00000003},
1523 {0x00000080, 0x00000000},
1524 {0x00000081, 0x01000000},
1525 {0x00000082, 0x02000000},
1526 {0x00000083, 0x00000000},
1527 {0x00000084, 0xe3f3e4f4},
1528 {0x00000085, 0x00052024},
1529 {0x00000087, 0x00000000},
1530 {0x00000088, 0x66036603},
1531 {0x00000089, 0x01000000},
1532 {0x0000008b, 0x1c0a0000},
1533 {0x0000008c, 0xff010000},
1534 {0x0000008e, 0xffffefff},
1535 {0x0000008f, 0xfff3efff},
1536 {0x00000090, 0xfff3efbf},
1537 {0x00000094, 0x00101101},
1538 {0x00000095, 0x00000fff},
1539 {0x00000096, 0x00116fff},
1540 {0x00000097, 0x60010000},
1541 {0x00000098, 0x10010000},
1542 {0x00000099, 0x00006000},
1543 {0x0000009a, 0x00001000},
1544 {0x0000009f, 0x00a07730}
1545};
1546
1547/* ucode loading */
1548int si_mc_load_microcode(struct radeon_device *rdev)
1549{
1550 const __be32 *fw_data = NULL;
1551 const __le32 *new_fw_data = NULL;
1552 u32 running;
1553 u32 *io_mc_regs = NULL;
1554 const __le32 *new_io_mc_regs = NULL;
1555 int i, regs_size, ucode_size;
1556
1557 if (!rdev->mc_fw)
1558 return -EINVAL;
1559
1560 if (rdev->new_fw) {
1561 const struct mc_firmware_header_v1_0 *hdr =
1562 (const struct mc_firmware_header_v1_0 *)rdev->mc_fw->data;
1563
1564 radeon_ucode_print_mc_hdr(&hdr->header);
1565 regs_size = le32_to_cpu(hdr->io_debug_size_bytes) / (4 * 2);
1566 new_io_mc_regs = (const __le32 *)
1567 (rdev->mc_fw->data + le32_to_cpu(hdr->io_debug_array_offset_bytes));
1568 ucode_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
1569 new_fw_data = (const __le32 *)
1570 (rdev->mc_fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes));
1571 } else {
1572 ucode_size = rdev->mc_fw->size / 4;
1573
1574 switch (rdev->family) {
1575 case CHIP_TAHITI:
1576 io_mc_regs = (u32 *)&tahiti_io_mc_regs;
1577 regs_size = TAHITI_IO_MC_REGS_SIZE;
1578 break;
1579 case CHIP_PITCAIRN:
1580 io_mc_regs = (u32 *)&pitcairn_io_mc_regs;
1581 regs_size = TAHITI_IO_MC_REGS_SIZE;
1582 break;
1583 case CHIP_VERDE:
1584 default:
1585 io_mc_regs = (u32 *)&verde_io_mc_regs;
1586 regs_size = TAHITI_IO_MC_REGS_SIZE;
1587 break;
1588 case CHIP_OLAND:
1589 io_mc_regs = (u32 *)&oland_io_mc_regs;
1590 regs_size = TAHITI_IO_MC_REGS_SIZE;
1591 break;
1592 case CHIP_HAINAN:
1593 io_mc_regs = (u32 *)&hainan_io_mc_regs;
1594 regs_size = TAHITI_IO_MC_REGS_SIZE;
1595 break;
1596 }
1597 fw_data = (const __be32 *)rdev->mc_fw->data;
1598 }
1599
1600 running = RREG32(MC_SEQ_SUP_CNTL) & RUN_MASK;
1601
1602 if (running == 0) {
1603 /* reset the engine and set to writable */
1604 WREG32(MC_SEQ_SUP_CNTL, 0x00000008);
1605 WREG32(MC_SEQ_SUP_CNTL, 0x00000010);
1606
1607 /* load mc io regs */
1608 for (i = 0; i < regs_size; i++) {
1609 if (rdev->new_fw) {
1610 WREG32(MC_SEQ_IO_DEBUG_INDEX, le32_to_cpup(new_io_mc_regs++));
1611 WREG32(MC_SEQ_IO_DEBUG_DATA, le32_to_cpup(new_io_mc_regs++));
1612 } else {
1613 WREG32(MC_SEQ_IO_DEBUG_INDEX, io_mc_regs[(i << 1)]);
1614 WREG32(MC_SEQ_IO_DEBUG_DATA, io_mc_regs[(i << 1) + 1]);
1615 }
1616 }
1617 /* load the MC ucode */
1618 for (i = 0; i < ucode_size; i++) {
1619 if (rdev->new_fw)
1620 WREG32(MC_SEQ_SUP_PGM, le32_to_cpup(new_fw_data++));
1621 else
1622 WREG32(MC_SEQ_SUP_PGM, be32_to_cpup(fw_data++));
1623 }
1624
1625 /* put the engine back into the active state */
1626 WREG32(MC_SEQ_SUP_CNTL, 0x00000008);
1627 WREG32(MC_SEQ_SUP_CNTL, 0x00000004);
1628 WREG32(MC_SEQ_SUP_CNTL, 0x00000001);
1629
1630 /* wait for training to complete */
1631 for (i = 0; i < rdev->usec_timeout; i++) {
1632 if (RREG32(MC_SEQ_TRAIN_WAKEUP_CNTL) & TRAIN_DONE_D0)
1633 break;
1634 udelay(1);
1635 }
1636 for (i = 0; i < rdev->usec_timeout; i++) {
1637 if (RREG32(MC_SEQ_TRAIN_WAKEUP_CNTL) & TRAIN_DONE_D1)
1638 break;
1639 udelay(1);
1640 }
1641 }
1642
1643 return 0;
1644}
1645
1646static int si_init_microcode(struct radeon_device *rdev)
1647{
1648 const char *chip_name;
1649 const char *new_chip_name;
1650 size_t pfp_req_size, me_req_size, ce_req_size, rlc_req_size, mc_req_size;
1651 size_t smc_req_size, mc2_req_size;
1652 char fw_name[30];
1653 int err;
1654 int new_fw = 0;
1655 bool new_smc = false;
1656 bool si58_fw = false;
1657 bool banks2_fw = false;
1658
1659 DRM_DEBUG("\n");
1660
1661 switch (rdev->family) {
1662 case CHIP_TAHITI:
1663 chip_name = "TAHITI";
1664 new_chip_name = "tahiti";
1665 pfp_req_size = SI_PFP_UCODE_SIZE * 4;
1666 me_req_size = SI_PM4_UCODE_SIZE * 4;
1667 ce_req_size = SI_CE_UCODE_SIZE * 4;
1668 rlc_req_size = SI_RLC_UCODE_SIZE * 4;
1669 mc_req_size = SI_MC_UCODE_SIZE * 4;
1670 mc2_req_size = TAHITI_MC_UCODE_SIZE * 4;
1671 smc_req_size = ALIGN(TAHITI_SMC_UCODE_SIZE, 4);
1672 break;
1673 case CHIP_PITCAIRN:
1674 chip_name = "PITCAIRN";
1675 if ((rdev->pdev->revision == 0x81) &&
1676 ((rdev->pdev->device == 0x6810) ||
1677 (rdev->pdev->device == 0x6811)))
1678 new_smc = true;
1679 new_chip_name = "pitcairn";
1680 pfp_req_size = SI_PFP_UCODE_SIZE * 4;
1681 me_req_size = SI_PM4_UCODE_SIZE * 4;
1682 ce_req_size = SI_CE_UCODE_SIZE * 4;
1683 rlc_req_size = SI_RLC_UCODE_SIZE * 4;
1684 mc_req_size = SI_MC_UCODE_SIZE * 4;
1685 mc2_req_size = PITCAIRN_MC_UCODE_SIZE * 4;
1686 smc_req_size = ALIGN(PITCAIRN_SMC_UCODE_SIZE, 4);
1687 break;
1688 case CHIP_VERDE:
1689 chip_name = "VERDE";
1690 if (((rdev->pdev->device == 0x6820) &&
1691 ((rdev->pdev->revision == 0x81) ||
1692 (rdev->pdev->revision == 0x83))) ||
1693 ((rdev->pdev->device == 0x6821) &&
1694 ((rdev->pdev->revision == 0x83) ||
1695 (rdev->pdev->revision == 0x87))) ||
1696 ((rdev->pdev->revision == 0x87) &&
1697 ((rdev->pdev->device == 0x6823) ||
1698 (rdev->pdev->device == 0x682b))))
1699 new_smc = true;
1700 new_chip_name = "verde";
1701 pfp_req_size = SI_PFP_UCODE_SIZE * 4;
1702 me_req_size = SI_PM4_UCODE_SIZE * 4;
1703 ce_req_size = SI_CE_UCODE_SIZE * 4;
1704 rlc_req_size = SI_RLC_UCODE_SIZE * 4;
1705 mc_req_size = SI_MC_UCODE_SIZE * 4;
1706 mc2_req_size = VERDE_MC_UCODE_SIZE * 4;
1707 smc_req_size = ALIGN(VERDE_SMC_UCODE_SIZE, 4);
1708 break;
1709 case CHIP_OLAND:
1710 chip_name = "OLAND";
1711 if (((rdev->pdev->revision == 0x81) &&
1712 ((rdev->pdev->device == 0x6600) ||
1713 (rdev->pdev->device == 0x6604) ||
1714 (rdev->pdev->device == 0x6605) ||
1715 (rdev->pdev->device == 0x6610))) ||
1716 ((rdev->pdev->revision == 0x83) &&
1717 (rdev->pdev->device == 0x6610)))
1718 new_smc = true;
1719 new_chip_name = "oland";
1720 pfp_req_size = SI_PFP_UCODE_SIZE * 4;
1721 me_req_size = SI_PM4_UCODE_SIZE * 4;
1722 ce_req_size = SI_CE_UCODE_SIZE * 4;
1723 rlc_req_size = SI_RLC_UCODE_SIZE * 4;
1724 mc_req_size = mc2_req_size = OLAND_MC_UCODE_SIZE * 4;
1725 smc_req_size = ALIGN(OLAND_SMC_UCODE_SIZE, 4);
1726 break;
1727 case CHIP_HAINAN:
1728 chip_name = "HAINAN";
1729 if (((rdev->pdev->revision == 0x81) &&
1730 (rdev->pdev->device == 0x6660)) ||
1731 ((rdev->pdev->revision == 0x83) &&
1732 ((rdev->pdev->device == 0x6660) ||
1733 (rdev->pdev->device == 0x6663) ||
1734 (rdev->pdev->device == 0x6665) ||
1735 (rdev->pdev->device == 0x6667))))
1736 new_smc = true;
1737 else if ((rdev->pdev->revision == 0xc3) &&
1738 (rdev->pdev->device == 0x6665))
1739 banks2_fw = true;
1740 new_chip_name = "hainan";
1741 pfp_req_size = SI_PFP_UCODE_SIZE * 4;
1742 me_req_size = SI_PM4_UCODE_SIZE * 4;
1743 ce_req_size = SI_CE_UCODE_SIZE * 4;
1744 rlc_req_size = SI_RLC_UCODE_SIZE * 4;
1745 mc_req_size = mc2_req_size = OLAND_MC_UCODE_SIZE * 4;
1746 smc_req_size = ALIGN(HAINAN_SMC_UCODE_SIZE, 4);
1747 break;
1748 default: BUG();
1749 }
1750
1751 /* this memory configuration requires special firmware */
1752 if (((RREG32(MC_SEQ_MISC0) & 0xff000000) >> 24) == 0x58)
1753 si58_fw = true;
1754
1755 DRM_INFO("Loading %s Microcode\n", new_chip_name);
1756
1757 snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", new_chip_name);
1758 err = request_firmware(&rdev->pfp_fw, fw_name, rdev->dev);
1759 if (err) {
1760 snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name);
1761 err = request_firmware(&rdev->pfp_fw, fw_name, rdev->dev);
1762 if (err)
1763 goto out;
1764 if (rdev->pfp_fw->size != pfp_req_size) {
1765 printk(KERN_ERR
1766 "si_cp: Bogus length %zu in firmware \"%s\"\n",
1767 rdev->pfp_fw->size, fw_name);
1768 err = -EINVAL;
1769 goto out;
1770 }
1771 } else {
1772 err = radeon_ucode_validate(rdev->pfp_fw);
1773 if (err) {
1774 printk(KERN_ERR
1775 "si_cp: validation failed for firmware \"%s\"\n",
1776 fw_name);
1777 goto out;
1778 } else {
1779 new_fw++;
1780 }
1781 }
1782
1783 snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", new_chip_name);
1784 err = request_firmware(&rdev->me_fw, fw_name, rdev->dev);
1785 if (err) {
1786 snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", chip_name);
1787 err = request_firmware(&rdev->me_fw, fw_name, rdev->dev);
1788 if (err)
1789 goto out;
1790 if (rdev->me_fw->size != me_req_size) {
1791 printk(KERN_ERR
1792 "si_cp: Bogus length %zu in firmware \"%s\"\n",
1793 rdev->me_fw->size, fw_name);
1794 err = -EINVAL;
1795 }
1796 } else {
1797 err = radeon_ucode_validate(rdev->me_fw);
1798 if (err) {
1799 printk(KERN_ERR
1800 "si_cp: validation failed for firmware \"%s\"\n",
1801 fw_name);
1802 goto out;
1803 } else {
1804 new_fw++;
1805 }
1806 }
1807
1808 snprintf(fw_name, sizeof(fw_name), "radeon/%s_ce.bin", new_chip_name);
1809 err = request_firmware(&rdev->ce_fw, fw_name, rdev->dev);
1810 if (err) {
1811 snprintf(fw_name, sizeof(fw_name), "radeon/%s_ce.bin", chip_name);
1812 err = request_firmware(&rdev->ce_fw, fw_name, rdev->dev);
1813 if (err)
1814 goto out;
1815 if (rdev->ce_fw->size != ce_req_size) {
1816 printk(KERN_ERR
1817 "si_cp: Bogus length %zu in firmware \"%s\"\n",
1818 rdev->ce_fw->size, fw_name);
1819 err = -EINVAL;
1820 }
1821 } else {
1822 err = radeon_ucode_validate(rdev->ce_fw);
1823 if (err) {
1824 printk(KERN_ERR
1825 "si_cp: validation failed for firmware \"%s\"\n",
1826 fw_name);
1827 goto out;
1828 } else {
1829 new_fw++;
1830 }
1831 }
1832
1833 snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", new_chip_name);
1834 err = request_firmware(&rdev->rlc_fw, fw_name, rdev->dev);
1835 if (err) {
1836 snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", chip_name);
1837 err = request_firmware(&rdev->rlc_fw, fw_name, rdev->dev);
1838 if (err)
1839 goto out;
1840 if (rdev->rlc_fw->size != rlc_req_size) {
1841 printk(KERN_ERR
1842 "si_rlc: Bogus length %zu in firmware \"%s\"\n",
1843 rdev->rlc_fw->size, fw_name);
1844 err = -EINVAL;
1845 }
1846 } else {
1847 err = radeon_ucode_validate(rdev->rlc_fw);
1848 if (err) {
1849 printk(KERN_ERR
1850 "si_cp: validation failed for firmware \"%s\"\n",
1851 fw_name);
1852 goto out;
1853 } else {
1854 new_fw++;
1855 }
1856 }
1857
1858 if (si58_fw)
1859 snprintf(fw_name, sizeof(fw_name), "radeon/si58_mc.bin");
1860 else
1861 snprintf(fw_name, sizeof(fw_name), "radeon/%s_mc.bin", new_chip_name);
1862 err = request_firmware(&rdev->mc_fw, fw_name, rdev->dev);
1863 if (err) {
1864 snprintf(fw_name, sizeof(fw_name), "radeon/%s_mc2.bin", chip_name);
1865 err = request_firmware(&rdev->mc_fw, fw_name, rdev->dev);
1866 if (err) {
1867 snprintf(fw_name, sizeof(fw_name), "radeon/%s_mc.bin", chip_name);
1868 err = request_firmware(&rdev->mc_fw, fw_name, rdev->dev);
1869 if (err)
1870 goto out;
1871 }
1872 if ((rdev->mc_fw->size != mc_req_size) &&
1873 (rdev->mc_fw->size != mc2_req_size)) {
1874 printk(KERN_ERR
1875 "si_mc: Bogus length %zu in firmware \"%s\"\n",
1876 rdev->mc_fw->size, fw_name);
1877 err = -EINVAL;
1878 }
1879 DRM_INFO("%s: %zu bytes\n", fw_name, rdev->mc_fw->size);
1880 } else {
1881 err = radeon_ucode_validate(rdev->mc_fw);
1882 if (err) {
1883 printk(KERN_ERR
1884 "si_cp: validation failed for firmware \"%s\"\n",
1885 fw_name);
1886 goto out;
1887 } else {
1888 new_fw++;
1889 }
1890 }
1891
1892 if (banks2_fw)
1893 snprintf(fw_name, sizeof(fw_name), "radeon/banks_k_2_smc.bin");
1894 else if (new_smc)
1895 snprintf(fw_name, sizeof(fw_name), "radeon/%s_k_smc.bin", new_chip_name);
1896 else
1897 snprintf(fw_name, sizeof(fw_name), "radeon/%s_smc.bin", new_chip_name);
1898 err = request_firmware(&rdev->smc_fw, fw_name, rdev->dev);
1899 if (err) {
1900 snprintf(fw_name, sizeof(fw_name), "radeon/%s_smc.bin", chip_name);
1901 err = request_firmware(&rdev->smc_fw, fw_name, rdev->dev);
1902 if (err) {
1903 printk(KERN_ERR
1904 "smc: error loading firmware \"%s\"\n",
1905 fw_name);
1906 release_firmware(rdev->smc_fw);
1907 rdev->smc_fw = NULL;
1908 err = 0;
1909 } else if (rdev->smc_fw->size != smc_req_size) {
1910 printk(KERN_ERR
1911 "si_smc: Bogus length %zu in firmware \"%s\"\n",
1912 rdev->smc_fw->size, fw_name);
1913 err = -EINVAL;
1914 }
1915 } else {
1916 err = radeon_ucode_validate(rdev->smc_fw);
1917 if (err) {
1918 printk(KERN_ERR
1919 "si_cp: validation failed for firmware \"%s\"\n",
1920 fw_name);
1921 goto out;
1922 } else {
1923 new_fw++;
1924 }
1925 }
1926
1927 if (new_fw == 0) {
1928 rdev->new_fw = false;
1929 } else if (new_fw < 6) {
1930 printk(KERN_ERR "si_fw: mixing new and old firmware!\n");
1931 err = -EINVAL;
1932 } else {
1933 rdev->new_fw = true;
1934 }
1935out:
1936 if (err) {
1937 if (err != -EINVAL)
1938 printk(KERN_ERR
1939 "si_cp: Failed to load firmware \"%s\"\n",
1940 fw_name);
1941 release_firmware(rdev->pfp_fw);
1942 rdev->pfp_fw = NULL;
1943 release_firmware(rdev->me_fw);
1944 rdev->me_fw = NULL;
1945 release_firmware(rdev->ce_fw);
1946 rdev->ce_fw = NULL;
1947 release_firmware(rdev->rlc_fw);
1948 rdev->rlc_fw = NULL;
1949 release_firmware(rdev->mc_fw);
1950 rdev->mc_fw = NULL;
1951 release_firmware(rdev->smc_fw);
1952 rdev->smc_fw = NULL;
1953 }
1954 return err;
1955}
1956
1957/* watermark setup */
1958static u32 dce6_line_buffer_adjust(struct radeon_device *rdev,
1959 struct radeon_crtc *radeon_crtc,
1960 struct drm_display_mode *mode,
1961 struct drm_display_mode *other_mode)
1962{
1963 u32 tmp, buffer_alloc, i;
1964 u32 pipe_offset = radeon_crtc->crtc_id * 0x20;
1965 /*
1966 * Line Buffer Setup
1967 * There are 3 line buffers, each one shared by 2 display controllers.
1968 * DC_LB_MEMORY_SPLIT controls how that line buffer is shared between
1969 * the display controllers. The paritioning is done via one of four
1970 * preset allocations specified in bits 21:20:
1971 * 0 - half lb
1972 * 2 - whole lb, other crtc must be disabled
1973 */
1974 /* this can get tricky if we have two large displays on a paired group
1975 * of crtcs. Ideally for multiple large displays we'd assign them to
1976 * non-linked crtcs for maximum line buffer allocation.
1977 */
1978 if (radeon_crtc->base.enabled && mode) {
1979 if (other_mode) {
1980 tmp = 0; /* 1/2 */
1981 buffer_alloc = 1;
1982 } else {
1983 tmp = 2; /* whole */
1984 buffer_alloc = 2;
1985 }
1986 } else {
1987 tmp = 0;
1988 buffer_alloc = 0;
1989 }
1990
1991 WREG32(DC_LB_MEMORY_SPLIT + radeon_crtc->crtc_offset,
1992 DC_LB_MEMORY_CONFIG(tmp));
1993
1994 WREG32(PIPE0_DMIF_BUFFER_CONTROL + pipe_offset,
1995 DMIF_BUFFERS_ALLOCATED(buffer_alloc));
1996 for (i = 0; i < rdev->usec_timeout; i++) {
1997 if (RREG32(PIPE0_DMIF_BUFFER_CONTROL + pipe_offset) &
1998 DMIF_BUFFERS_ALLOCATED_COMPLETED)
1999 break;
2000 udelay(1);
2001 }
2002
2003 if (radeon_crtc->base.enabled && mode) {
2004 switch (tmp) {
2005 case 0:
2006 default:
2007 return 4096 * 2;
2008 case 2:
2009 return 8192 * 2;
2010 }
2011 }
2012
2013 /* controller not enabled, so no lb used */
2014 return 0;
2015}
2016
2017static u32 si_get_number_of_dram_channels(struct radeon_device *rdev)
2018{
2019 u32 tmp = RREG32(MC_SHARED_CHMAP);
2020
2021 switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
2022 case 0:
2023 default:
2024 return 1;
2025 case 1:
2026 return 2;
2027 case 2:
2028 return 4;
2029 case 3:
2030 return 8;
2031 case 4:
2032 return 3;
2033 case 5:
2034 return 6;
2035 case 6:
2036 return 10;
2037 case 7:
2038 return 12;
2039 case 8:
2040 return 16;
2041 }
2042}
2043
2044struct dce6_wm_params {
2045 u32 dram_channels; /* number of dram channels */
2046 u32 yclk; /* bandwidth per dram data pin in kHz */
2047 u32 sclk; /* engine clock in kHz */
2048 u32 disp_clk; /* display clock in kHz */
2049 u32 src_width; /* viewport width */
2050 u32 active_time; /* active display time in ns */
2051 u32 blank_time; /* blank time in ns */
2052 bool interlaced; /* mode is interlaced */
2053 fixed20_12 vsc; /* vertical scale ratio */
2054 u32 num_heads; /* number of active crtcs */
2055 u32 bytes_per_pixel; /* bytes per pixel display + overlay */
2056 u32 lb_size; /* line buffer allocated to pipe */
2057 u32 vtaps; /* vertical scaler taps */
2058};
2059
2060static u32 dce6_dram_bandwidth(struct dce6_wm_params *wm)
2061{
2062 /* Calculate raw DRAM Bandwidth */
2063 fixed20_12 dram_efficiency; /* 0.7 */
2064 fixed20_12 yclk, dram_channels, bandwidth;
2065 fixed20_12 a;
2066
2067 a.full = dfixed_const(1000);
2068 yclk.full = dfixed_const(wm->yclk);
2069 yclk.full = dfixed_div(yclk, a);
2070 dram_channels.full = dfixed_const(wm->dram_channels * 4);
2071 a.full = dfixed_const(10);
2072 dram_efficiency.full = dfixed_const(7);
2073 dram_efficiency.full = dfixed_div(dram_efficiency, a);
2074 bandwidth.full = dfixed_mul(dram_channels, yclk);
2075 bandwidth.full = dfixed_mul(bandwidth, dram_efficiency);
2076
2077 return dfixed_trunc(bandwidth);
2078}
2079
2080static u32 dce6_dram_bandwidth_for_display(struct dce6_wm_params *wm)
2081{
2082 /* Calculate DRAM Bandwidth and the part allocated to display. */
2083 fixed20_12 disp_dram_allocation; /* 0.3 to 0.7 */
2084 fixed20_12 yclk, dram_channels, bandwidth;
2085 fixed20_12 a;
2086
2087 a.full = dfixed_const(1000);
2088 yclk.full = dfixed_const(wm->yclk);
2089 yclk.full = dfixed_div(yclk, a);
2090 dram_channels.full = dfixed_const(wm->dram_channels * 4);
2091 a.full = dfixed_const(10);
2092 disp_dram_allocation.full = dfixed_const(3); /* XXX worse case value 0.3 */
2093 disp_dram_allocation.full = dfixed_div(disp_dram_allocation, a);
2094 bandwidth.full = dfixed_mul(dram_channels, yclk);
2095 bandwidth.full = dfixed_mul(bandwidth, disp_dram_allocation);
2096
2097 return dfixed_trunc(bandwidth);
2098}
2099
2100static u32 dce6_data_return_bandwidth(struct dce6_wm_params *wm)
2101{
2102 /* Calculate the display Data return Bandwidth */
2103 fixed20_12 return_efficiency; /* 0.8 */
2104 fixed20_12 sclk, bandwidth;
2105 fixed20_12 a;
2106
2107 a.full = dfixed_const(1000);
2108 sclk.full = dfixed_const(wm->sclk);
2109 sclk.full = dfixed_div(sclk, a);
2110 a.full = dfixed_const(10);
2111 return_efficiency.full = dfixed_const(8);
2112 return_efficiency.full = dfixed_div(return_efficiency, a);
2113 a.full = dfixed_const(32);
2114 bandwidth.full = dfixed_mul(a, sclk);
2115 bandwidth.full = dfixed_mul(bandwidth, return_efficiency);
2116
2117 return dfixed_trunc(bandwidth);
2118}
2119
2120static u32 dce6_get_dmif_bytes_per_request(struct dce6_wm_params *wm)
2121{
2122 return 32;
2123}
2124
2125static u32 dce6_dmif_request_bandwidth(struct dce6_wm_params *wm)
2126{
2127 /* Calculate the DMIF Request Bandwidth */
2128 fixed20_12 disp_clk_request_efficiency; /* 0.8 */
2129 fixed20_12 disp_clk, sclk, bandwidth;
2130 fixed20_12 a, b1, b2;
2131 u32 min_bandwidth;
2132
2133 a.full = dfixed_const(1000);
2134 disp_clk.full = dfixed_const(wm->disp_clk);
2135 disp_clk.full = dfixed_div(disp_clk, a);
2136 a.full = dfixed_const(dce6_get_dmif_bytes_per_request(wm) / 2);
2137 b1.full = dfixed_mul(a, disp_clk);
2138
2139 a.full = dfixed_const(1000);
2140 sclk.full = dfixed_const(wm->sclk);
2141 sclk.full = dfixed_div(sclk, a);
2142 a.full = dfixed_const(dce6_get_dmif_bytes_per_request(wm));
2143 b2.full = dfixed_mul(a, sclk);
2144
2145 a.full = dfixed_const(10);
2146 disp_clk_request_efficiency.full = dfixed_const(8);
2147 disp_clk_request_efficiency.full = dfixed_div(disp_clk_request_efficiency, a);
2148
2149 min_bandwidth = min(dfixed_trunc(b1), dfixed_trunc(b2));
2150
2151 a.full = dfixed_const(min_bandwidth);
2152 bandwidth.full = dfixed_mul(a, disp_clk_request_efficiency);
2153
2154 return dfixed_trunc(bandwidth);
2155}
2156
2157static u32 dce6_available_bandwidth(struct dce6_wm_params *wm)
2158{
2159 /* Calculate the Available bandwidth. Display can use this temporarily but not in average. */
2160 u32 dram_bandwidth = dce6_dram_bandwidth(wm);
2161 u32 data_return_bandwidth = dce6_data_return_bandwidth(wm);
2162 u32 dmif_req_bandwidth = dce6_dmif_request_bandwidth(wm);
2163
2164 return min(dram_bandwidth, min(data_return_bandwidth, dmif_req_bandwidth));
2165}
2166
2167static u32 dce6_average_bandwidth(struct dce6_wm_params *wm)
2168{
2169 /* Calculate the display mode Average Bandwidth
2170 * DisplayMode should contain the source and destination dimensions,
2171 * timing, etc.
2172 */
2173 fixed20_12 bpp;
2174 fixed20_12 line_time;
2175 fixed20_12 src_width;
2176 fixed20_12 bandwidth;
2177 fixed20_12 a;
2178
2179 a.full = dfixed_const(1000);
2180 line_time.full = dfixed_const(wm->active_time + wm->blank_time);
2181 line_time.full = dfixed_div(line_time, a);
2182 bpp.full = dfixed_const(wm->bytes_per_pixel);
2183 src_width.full = dfixed_const(wm->src_width);
2184 bandwidth.full = dfixed_mul(src_width, bpp);
2185 bandwidth.full = dfixed_mul(bandwidth, wm->vsc);
2186 bandwidth.full = dfixed_div(bandwidth, line_time);
2187
2188 return dfixed_trunc(bandwidth);
2189}
2190
2191static u32 dce6_latency_watermark(struct dce6_wm_params *wm)
2192{
2193 /* First calcualte the latency in ns */
2194 u32 mc_latency = 2000; /* 2000 ns. */
2195 u32 available_bandwidth = dce6_available_bandwidth(wm);
2196 u32 worst_chunk_return_time = (512 * 8 * 1000) / available_bandwidth;
2197 u32 cursor_line_pair_return_time = (128 * 4 * 1000) / available_bandwidth;
2198 u32 dc_latency = 40000000 / wm->disp_clk; /* dc pipe latency */
2199 u32 other_heads_data_return_time = ((wm->num_heads + 1) * worst_chunk_return_time) +
2200 (wm->num_heads * cursor_line_pair_return_time);
2201 u32 latency = mc_latency + other_heads_data_return_time + dc_latency;
2202 u32 max_src_lines_per_dst_line, lb_fill_bw, line_fill_time;
2203 u32 tmp, dmif_size = 12288;
2204 fixed20_12 a, b, c;
2205
2206 if (wm->num_heads == 0)
2207 return 0;
2208
2209 a.full = dfixed_const(2);
2210 b.full = dfixed_const(1);
2211 if ((wm->vsc.full > a.full) ||
2212 ((wm->vsc.full > b.full) && (wm->vtaps >= 3)) ||
2213 (wm->vtaps >= 5) ||
2214 ((wm->vsc.full >= a.full) && wm->interlaced))
2215 max_src_lines_per_dst_line = 4;
2216 else
2217 max_src_lines_per_dst_line = 2;
2218
2219 a.full = dfixed_const(available_bandwidth);
2220 b.full = dfixed_const(wm->num_heads);
2221 a.full = dfixed_div(a, b);
2222
2223 b.full = dfixed_const(mc_latency + 512);
2224 c.full = dfixed_const(wm->disp_clk);
2225 b.full = dfixed_div(b, c);
2226
2227 c.full = dfixed_const(dmif_size);
2228 b.full = dfixed_div(c, b);
2229
2230 tmp = min(dfixed_trunc(a), dfixed_trunc(b));
2231
2232 b.full = dfixed_const(1000);
2233 c.full = dfixed_const(wm->disp_clk);
2234 b.full = dfixed_div(c, b);
2235 c.full = dfixed_const(wm->bytes_per_pixel);
2236 b.full = dfixed_mul(b, c);
2237
2238 lb_fill_bw = min(tmp, dfixed_trunc(b));
2239
2240 a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
2241 b.full = dfixed_const(1000);
2242 c.full = dfixed_const(lb_fill_bw);
2243 b.full = dfixed_div(c, b);
2244 a.full = dfixed_div(a, b);
2245 line_fill_time = dfixed_trunc(a);
2246
2247 if (line_fill_time < wm->active_time)
2248 return latency;
2249 else
2250 return latency + (line_fill_time - wm->active_time);
2251
2252}
2253
2254static bool dce6_average_bandwidth_vs_dram_bandwidth_for_display(struct dce6_wm_params *wm)
2255{
2256 if (dce6_average_bandwidth(wm) <=
2257 (dce6_dram_bandwidth_for_display(wm) / wm->num_heads))
2258 return true;
2259 else
2260 return false;
2261};
2262
2263static bool dce6_average_bandwidth_vs_available_bandwidth(struct dce6_wm_params *wm)
2264{
2265 if (dce6_average_bandwidth(wm) <=
2266 (dce6_available_bandwidth(wm) / wm->num_heads))
2267 return true;
2268 else
2269 return false;
2270};
2271
2272static bool dce6_check_latency_hiding(struct dce6_wm_params *wm)
2273{
2274 u32 lb_partitions = wm->lb_size / wm->src_width;
2275 u32 line_time = wm->active_time + wm->blank_time;
2276 u32 latency_tolerant_lines;
2277 u32 latency_hiding;
2278 fixed20_12 a;
2279
2280 a.full = dfixed_const(1);
2281 if (wm->vsc.full > a.full)
2282 latency_tolerant_lines = 1;
2283 else {
2284 if (lb_partitions <= (wm->vtaps + 1))
2285 latency_tolerant_lines = 1;
2286 else
2287 latency_tolerant_lines = 2;
2288 }
2289
2290 latency_hiding = (latency_tolerant_lines * line_time + wm->blank_time);
2291
2292 if (dce6_latency_watermark(wm) <= latency_hiding)
2293 return true;
2294 else
2295 return false;
2296}
2297
2298static void dce6_program_watermarks(struct radeon_device *rdev,
2299 struct radeon_crtc *radeon_crtc,
2300 u32 lb_size, u32 num_heads)
2301{
2302 struct drm_display_mode *mode = &radeon_crtc->base.mode;
2303 struct dce6_wm_params wm_low, wm_high;
2304 u32 dram_channels;
2305 u32 pixel_period;
2306 u32 line_time = 0;
2307 u32 latency_watermark_a = 0, latency_watermark_b = 0;
2308 u32 priority_a_mark = 0, priority_b_mark = 0;
2309 u32 priority_a_cnt = PRIORITY_OFF;
2310 u32 priority_b_cnt = PRIORITY_OFF;
2311 u32 tmp, arb_control3;
2312 fixed20_12 a, b, c;
2313
2314 if (radeon_crtc->base.enabled && num_heads && mode) {
2315 pixel_period = 1000000 / (u32)mode->clock;
2316 line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535);
2317 priority_a_cnt = 0;
2318 priority_b_cnt = 0;
2319
2320 if (rdev->family == CHIP_ARUBA)
2321 dram_channels = evergreen_get_number_of_dram_channels(rdev);
2322 else
2323 dram_channels = si_get_number_of_dram_channels(rdev);
2324
2325 /* watermark for high clocks */
2326 if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) {
2327 wm_high.yclk =
2328 radeon_dpm_get_mclk(rdev, false) * 10;
2329 wm_high.sclk =
2330 radeon_dpm_get_sclk(rdev, false) * 10;
2331 } else {
2332 wm_high.yclk = rdev->pm.current_mclk * 10;
2333 wm_high.sclk = rdev->pm.current_sclk * 10;
2334 }
2335
2336 wm_high.disp_clk = mode->clock;
2337 wm_high.src_width = mode->crtc_hdisplay;
2338 wm_high.active_time = mode->crtc_hdisplay * pixel_period;
2339 wm_high.blank_time = line_time - wm_high.active_time;
2340 wm_high.interlaced = false;
2341 if (mode->flags & DRM_MODE_FLAG_INTERLACE)
2342 wm_high.interlaced = true;
2343 wm_high.vsc = radeon_crtc->vsc;
2344 wm_high.vtaps = 1;
2345 if (radeon_crtc->rmx_type != RMX_OFF)
2346 wm_high.vtaps = 2;
2347 wm_high.bytes_per_pixel = 4; /* XXX: get this from fb config */
2348 wm_high.lb_size = lb_size;
2349 wm_high.dram_channels = dram_channels;
2350 wm_high.num_heads = num_heads;
2351
2352 /* watermark for low clocks */
2353 if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) {
2354 wm_low.yclk =
2355 radeon_dpm_get_mclk(rdev, true) * 10;
2356 wm_low.sclk =
2357 radeon_dpm_get_sclk(rdev, true) * 10;
2358 } else {
2359 wm_low.yclk = rdev->pm.current_mclk * 10;
2360 wm_low.sclk = rdev->pm.current_sclk * 10;
2361 }
2362
2363 wm_low.disp_clk = mode->clock;
2364 wm_low.src_width = mode->crtc_hdisplay;
2365 wm_low.active_time = mode->crtc_hdisplay * pixel_period;
2366 wm_low.blank_time = line_time - wm_low.active_time;
2367 wm_low.interlaced = false;
2368 if (mode->flags & DRM_MODE_FLAG_INTERLACE)
2369 wm_low.interlaced = true;
2370 wm_low.vsc = radeon_crtc->vsc;
2371 wm_low.vtaps = 1;
2372 if (radeon_crtc->rmx_type != RMX_OFF)
2373 wm_low.vtaps = 2;
2374 wm_low.bytes_per_pixel = 4; /* XXX: get this from fb config */
2375 wm_low.lb_size = lb_size;
2376 wm_low.dram_channels = dram_channels;
2377 wm_low.num_heads = num_heads;
2378
2379 /* set for high clocks */
2380 latency_watermark_a = min(dce6_latency_watermark(&wm_high), (u32)65535);
2381 /* set for low clocks */
2382 latency_watermark_b = min(dce6_latency_watermark(&wm_low), (u32)65535);
2383
2384 /* possibly force display priority to high */
2385 /* should really do this at mode validation time... */
2386 if (!dce6_average_bandwidth_vs_dram_bandwidth_for_display(&wm_high) ||
2387 !dce6_average_bandwidth_vs_available_bandwidth(&wm_high) ||
2388 !dce6_check_latency_hiding(&wm_high) ||
2389 (rdev->disp_priority == 2)) {
2390 DRM_DEBUG_KMS("force priority to high\n");
2391 priority_a_cnt |= PRIORITY_ALWAYS_ON;
2392 priority_b_cnt |= PRIORITY_ALWAYS_ON;
2393 }
2394 if (!dce6_average_bandwidth_vs_dram_bandwidth_for_display(&wm_low) ||
2395 !dce6_average_bandwidth_vs_available_bandwidth(&wm_low) ||
2396 !dce6_check_latency_hiding(&wm_low) ||
2397 (rdev->disp_priority == 2)) {
2398 DRM_DEBUG_KMS("force priority to high\n");
2399 priority_a_cnt |= PRIORITY_ALWAYS_ON;
2400 priority_b_cnt |= PRIORITY_ALWAYS_ON;
2401 }
2402
2403 a.full = dfixed_const(1000);
2404 b.full = dfixed_const(mode->clock);
2405 b.full = dfixed_div(b, a);
2406 c.full = dfixed_const(latency_watermark_a);
2407 c.full = dfixed_mul(c, b);
2408 c.full = dfixed_mul(c, radeon_crtc->hsc);
2409 c.full = dfixed_div(c, a);
2410 a.full = dfixed_const(16);
2411 c.full = dfixed_div(c, a);
2412 priority_a_mark = dfixed_trunc(c);
2413 priority_a_cnt |= priority_a_mark & PRIORITY_MARK_MASK;
2414
2415 a.full = dfixed_const(1000);
2416 b.full = dfixed_const(mode->clock);
2417 b.full = dfixed_div(b, a);
2418 c.full = dfixed_const(latency_watermark_b);
2419 c.full = dfixed_mul(c, b);
2420 c.full = dfixed_mul(c, radeon_crtc->hsc);
2421 c.full = dfixed_div(c, a);
2422 a.full = dfixed_const(16);
2423 c.full = dfixed_div(c, a);
2424 priority_b_mark = dfixed_trunc(c);
2425 priority_b_cnt |= priority_b_mark & PRIORITY_MARK_MASK;
2426
2427 /* Save number of lines the linebuffer leads before the scanout */
2428 radeon_crtc->lb_vblank_lead_lines = DIV_ROUND_UP(lb_size, mode->crtc_hdisplay);
2429 }
2430
2431 /* select wm A */
2432 arb_control3 = RREG32(DPG_PIPE_ARBITRATION_CONTROL3 + radeon_crtc->crtc_offset);
2433 tmp = arb_control3;
2434 tmp &= ~LATENCY_WATERMARK_MASK(3);
2435 tmp |= LATENCY_WATERMARK_MASK(1);
2436 WREG32(DPG_PIPE_ARBITRATION_CONTROL3 + radeon_crtc->crtc_offset, tmp);
2437 WREG32(DPG_PIPE_LATENCY_CONTROL + radeon_crtc->crtc_offset,
2438 (LATENCY_LOW_WATERMARK(latency_watermark_a) |
2439 LATENCY_HIGH_WATERMARK(line_time)));
2440 /* select wm B */
2441 tmp = RREG32(DPG_PIPE_ARBITRATION_CONTROL3 + radeon_crtc->crtc_offset);
2442 tmp &= ~LATENCY_WATERMARK_MASK(3);
2443 tmp |= LATENCY_WATERMARK_MASK(2);
2444 WREG32(DPG_PIPE_ARBITRATION_CONTROL3 + radeon_crtc->crtc_offset, tmp);
2445 WREG32(DPG_PIPE_LATENCY_CONTROL + radeon_crtc->crtc_offset,
2446 (LATENCY_LOW_WATERMARK(latency_watermark_b) |
2447 LATENCY_HIGH_WATERMARK(line_time)));
2448 /* restore original selection */
2449 WREG32(DPG_PIPE_ARBITRATION_CONTROL3 + radeon_crtc->crtc_offset, arb_control3);
2450
2451 /* write the priority marks */
2452 WREG32(PRIORITY_A_CNT + radeon_crtc->crtc_offset, priority_a_cnt);
2453 WREG32(PRIORITY_B_CNT + radeon_crtc->crtc_offset, priority_b_cnt);
2454
2455 /* save values for DPM */
2456 radeon_crtc->line_time = line_time;
2457 radeon_crtc->wm_high = latency_watermark_a;
2458 radeon_crtc->wm_low = latency_watermark_b;
2459}
2460
2461void dce6_bandwidth_update(struct radeon_device *rdev)
2462{
2463 struct drm_display_mode *mode0 = NULL;
2464 struct drm_display_mode *mode1 = NULL;
2465 u32 num_heads = 0, lb_size;
2466 int i;
2467
2468 if (!rdev->mode_info.mode_config_initialized)
2469 return;
2470
2471 radeon_update_display_priority(rdev);
2472
2473 for (i = 0; i < rdev->num_crtc; i++) {
2474 if (rdev->mode_info.crtcs[i]->base.enabled)
2475 num_heads++;
2476 }
2477 for (i = 0; i < rdev->num_crtc; i += 2) {
2478 mode0 = &rdev->mode_info.crtcs[i]->base.mode;
2479 mode1 = &rdev->mode_info.crtcs[i+1]->base.mode;
2480 lb_size = dce6_line_buffer_adjust(rdev, rdev->mode_info.crtcs[i], mode0, mode1);
2481 dce6_program_watermarks(rdev, rdev->mode_info.crtcs[i], lb_size, num_heads);
2482 lb_size = dce6_line_buffer_adjust(rdev, rdev->mode_info.crtcs[i+1], mode1, mode0);
2483 dce6_program_watermarks(rdev, rdev->mode_info.crtcs[i+1], lb_size, num_heads);
2484 }
2485}
2486
2487/*
2488 * Core functions
2489 */
2490static void si_tiling_mode_table_init(struct radeon_device *rdev)
2491{
2492 u32 *tile = rdev->config.si.tile_mode_array;
2493 const u32 num_tile_mode_states =
2494 ARRAY_SIZE(rdev->config.si.tile_mode_array);
2495 u32 reg_offset, split_equal_to_row_size;
2496
2497 switch (rdev->config.si.mem_row_size_in_kb) {
2498 case 1:
2499 split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_1KB;
2500 break;
2501 case 2:
2502 default:
2503 split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_2KB;
2504 break;
2505 case 4:
2506 split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_4KB;
2507 break;
2508 }
2509
2510 for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++)
2511 tile[reg_offset] = 0;
2512
2513 switch(rdev->family) {
2514 case CHIP_TAHITI:
2515 case CHIP_PITCAIRN:
2516 /* non-AA compressed depth or any compressed stencil */
2517 tile[0] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2518 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2519 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2520 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2521 NUM_BANKS(ADDR_SURF_16_BANK) |
2522 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2523 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2524 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2525 /* 2xAA/4xAA compressed depth only */
2526 tile[1] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2527 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2528 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2529 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) |
2530 NUM_BANKS(ADDR_SURF_16_BANK) |
2531 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2532 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2533 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2534 /* 8xAA compressed depth only */
2535 tile[2] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2536 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2537 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2538 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2539 NUM_BANKS(ADDR_SURF_16_BANK) |
2540 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2541 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2542 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2543 /* 2xAA/4xAA compressed depth with stencil (for depth buffer) */
2544 tile[3] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2545 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2546 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2547 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) |
2548 NUM_BANKS(ADDR_SURF_16_BANK) |
2549 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2550 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2551 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2552 /* Maps w/ a dimension less than the 2D macro-tile dimensions (for mipmapped depth textures) */
2553 tile[4] = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
2554 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2555 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2556 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2557 NUM_BANKS(ADDR_SURF_16_BANK) |
2558 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2559 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2560 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2561 /* Uncompressed 16bpp depth - and stencil buffer allocated with it */
2562 tile[5] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2563 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2564 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2565 TILE_SPLIT(split_equal_to_row_size) |
2566 NUM_BANKS(ADDR_SURF_16_BANK) |
2567 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2568 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2569 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2570 /* Uncompressed 32bpp depth - and stencil buffer allocated with it */
2571 tile[6] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2572 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2573 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2574 TILE_SPLIT(split_equal_to_row_size) |
2575 NUM_BANKS(ADDR_SURF_16_BANK) |
2576 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2577 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2578 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2579 /* Uncompressed 8bpp stencil without depth (drivers typically do not use) */
2580 tile[7] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2581 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2582 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2583 TILE_SPLIT(split_equal_to_row_size) |
2584 NUM_BANKS(ADDR_SURF_16_BANK) |
2585 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2586 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2587 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2588 /* 1D and 1D Array Surfaces */
2589 tile[8] = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) |
2590 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2591 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2592 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2593 NUM_BANKS(ADDR_SURF_16_BANK) |
2594 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2595 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2596 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2597 /* Displayable maps. */
2598 tile[9] = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
2599 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2600 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2601 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2602 NUM_BANKS(ADDR_SURF_16_BANK) |
2603 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2604 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2605 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2606 /* Display 8bpp. */
2607 tile[10] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2608 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2609 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2610 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2611 NUM_BANKS(ADDR_SURF_16_BANK) |
2612 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2613 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2614 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2615 /* Display 16bpp. */
2616 tile[11] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2617 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2618 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2619 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2620 NUM_BANKS(ADDR_SURF_16_BANK) |
2621 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2622 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2623 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2624 /* Display 32bpp. */
2625 tile[12] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2626 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2627 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2628 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
2629 NUM_BANKS(ADDR_SURF_16_BANK) |
2630 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2631 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2632 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2633 /* Thin. */
2634 tile[13] = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
2635 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2636 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2637 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2638 NUM_BANKS(ADDR_SURF_16_BANK) |
2639 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2640 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2641 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2642 /* Thin 8 bpp. */
2643 tile[14] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2644 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2645 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2646 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2647 NUM_BANKS(ADDR_SURF_16_BANK) |
2648 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2649 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2650 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2651 /* Thin 16 bpp. */
2652 tile[15] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2653 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2654 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2655 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2656 NUM_BANKS(ADDR_SURF_16_BANK) |
2657 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2658 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2659 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2660 /* Thin 32 bpp. */
2661 tile[16] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2662 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2663 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2664 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
2665 NUM_BANKS(ADDR_SURF_16_BANK) |
2666 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2667 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2668 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2669 /* Thin 64 bpp. */
2670 tile[17] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2671 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2672 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2673 TILE_SPLIT(split_equal_to_row_size) |
2674 NUM_BANKS(ADDR_SURF_16_BANK) |
2675 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2676 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2677 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2678 /* 8 bpp PRT. */
2679 tile[21] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2680 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2681 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2682 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2683 NUM_BANKS(ADDR_SURF_16_BANK) |
2684 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) |
2685 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2686 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2687 /* 16 bpp PRT */
2688 tile[22] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2689 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2690 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2691 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2692 NUM_BANKS(ADDR_SURF_16_BANK) |
2693 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2694 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2695 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2696 /* 32 bpp PRT */
2697 tile[23] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2698 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2699 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2700 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2701 NUM_BANKS(ADDR_SURF_16_BANK) |
2702 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2703 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2704 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2705 /* 64 bpp PRT */
2706 tile[24] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2707 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2708 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2709 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
2710 NUM_BANKS(ADDR_SURF_16_BANK) |
2711 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2712 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2713 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2714 /* 128 bpp PRT */
2715 tile[25] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2716 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2717 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2718 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_1KB) |
2719 NUM_BANKS(ADDR_SURF_8_BANK) |
2720 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2721 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2722 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2723
2724 for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++)
2725 WREG32(GB_TILE_MODE0 + (reg_offset * 4), tile[reg_offset]);
2726 break;
2727
2728 case CHIP_VERDE:
2729 case CHIP_OLAND:
2730 case CHIP_HAINAN:
2731 /* non-AA compressed depth or any compressed stencil */
2732 tile[0] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2733 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2734 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2735 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2736 NUM_BANKS(ADDR_SURF_16_BANK) |
2737 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2738 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2739 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2740 /* 2xAA/4xAA compressed depth only */
2741 tile[1] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2742 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2743 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2744 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) |
2745 NUM_BANKS(ADDR_SURF_16_BANK) |
2746 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2747 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2748 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2749 /* 8xAA compressed depth only */
2750 tile[2] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2751 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2752 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2753 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2754 NUM_BANKS(ADDR_SURF_16_BANK) |
2755 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2756 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2757 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2758 /* 2xAA/4xAA compressed depth with stencil (for depth buffer) */
2759 tile[3] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2760 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2761 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2762 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) |
2763 NUM_BANKS(ADDR_SURF_16_BANK) |
2764 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2765 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2766 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2767 /* Maps w/ a dimension less than the 2D macro-tile dimensions (for mipmapped depth textures) */
2768 tile[4] = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
2769 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2770 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2771 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2772 NUM_BANKS(ADDR_SURF_16_BANK) |
2773 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2774 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2775 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2776 /* Uncompressed 16bpp depth - and stencil buffer allocated with it */
2777 tile[5] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2778 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2779 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2780 TILE_SPLIT(split_equal_to_row_size) |
2781 NUM_BANKS(ADDR_SURF_16_BANK) |
2782 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2783 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2784 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2785 /* Uncompressed 32bpp depth - and stencil buffer allocated with it */
2786 tile[6] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2787 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2788 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2789 TILE_SPLIT(split_equal_to_row_size) |
2790 NUM_BANKS(ADDR_SURF_16_BANK) |
2791 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2792 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2793 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2794 /* Uncompressed 8bpp stencil without depth (drivers typically do not use) */
2795 tile[7] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2796 MICRO_TILE_MODE(ADDR_SURF_DEPTH_MICRO_TILING) |
2797 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2798 TILE_SPLIT(split_equal_to_row_size) |
2799 NUM_BANKS(ADDR_SURF_16_BANK) |
2800 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2801 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2802 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2803 /* 1D and 1D Array Surfaces */
2804 tile[8] = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) |
2805 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2806 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2807 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2808 NUM_BANKS(ADDR_SURF_16_BANK) |
2809 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2810 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2811 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2812 /* Displayable maps. */
2813 tile[9] = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
2814 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2815 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2816 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2817 NUM_BANKS(ADDR_SURF_16_BANK) |
2818 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2819 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2820 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2821 /* Display 8bpp. */
2822 tile[10] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2823 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2824 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2825 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2826 NUM_BANKS(ADDR_SURF_16_BANK) |
2827 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2828 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2829 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2830 /* Display 16bpp. */
2831 tile[11] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2832 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2833 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2834 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2835 NUM_BANKS(ADDR_SURF_16_BANK) |
2836 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2837 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2838 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2839 /* Display 32bpp. */
2840 tile[12] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2841 MICRO_TILE_MODE(ADDR_SURF_DISPLAY_MICRO_TILING) |
2842 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2843 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
2844 NUM_BANKS(ADDR_SURF_16_BANK) |
2845 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2846 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2847 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2848 /* Thin. */
2849 tile[13] = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
2850 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2851 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2852 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
2853 NUM_BANKS(ADDR_SURF_16_BANK) |
2854 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2855 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2856 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2857 /* Thin 8 bpp. */
2858 tile[14] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2859 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2860 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2861 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2862 NUM_BANKS(ADDR_SURF_16_BANK) |
2863 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2864 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2865 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2866 /* Thin 16 bpp. */
2867 tile[15] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2868 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2869 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2870 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2871 NUM_BANKS(ADDR_SURF_16_BANK) |
2872 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2873 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2874 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2875 /* Thin 32 bpp. */
2876 tile[16] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2877 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2878 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2879 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
2880 NUM_BANKS(ADDR_SURF_16_BANK) |
2881 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2882 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2883 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2884 /* Thin 64 bpp. */
2885 tile[17] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2886 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2887 PIPE_CONFIG(ADDR_SURF_P4_8x16) |
2888 TILE_SPLIT(split_equal_to_row_size) |
2889 NUM_BANKS(ADDR_SURF_16_BANK) |
2890 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2891 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2892 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2893 /* 8 bpp PRT. */
2894 tile[21] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2895 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2896 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2897 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2898 NUM_BANKS(ADDR_SURF_16_BANK) |
2899 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) |
2900 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2901 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2902 /* 16 bpp PRT */
2903 tile[22] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2904 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2905 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2906 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2907 NUM_BANKS(ADDR_SURF_16_BANK) |
2908 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2909 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
2910 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4));
2911 /* 32 bpp PRT */
2912 tile[23] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2913 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2914 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2915 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
2916 NUM_BANKS(ADDR_SURF_16_BANK) |
2917 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2918 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
2919 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2920 /* 64 bpp PRT */
2921 tile[24] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2922 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2923 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2924 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
2925 NUM_BANKS(ADDR_SURF_16_BANK) |
2926 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2927 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2928 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2));
2929 /* 128 bpp PRT */
2930 tile[25] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
2931 MICRO_TILE_MODE(ADDR_SURF_THIN_MICRO_TILING) |
2932 PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) |
2933 TILE_SPLIT(ADDR_SURF_TILE_SPLIT_1KB) |
2934 NUM_BANKS(ADDR_SURF_8_BANK) |
2935 BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
2936 BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
2937 MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1));
2938
2939 for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++)
2940 WREG32(GB_TILE_MODE0 + (reg_offset * 4), tile[reg_offset]);
2941 break;
2942
2943 default:
2944 DRM_ERROR("unknown asic: 0x%x\n", rdev->family);
2945 }
2946}
2947
2948static void si_select_se_sh(struct radeon_device *rdev,
2949 u32 se_num, u32 sh_num)
2950{
2951 u32 data = INSTANCE_BROADCAST_WRITES;
2952
2953 if ((se_num == 0xffffffff) && (sh_num == 0xffffffff))
2954 data |= SH_BROADCAST_WRITES | SE_BROADCAST_WRITES;
2955 else if (se_num == 0xffffffff)
2956 data |= SE_BROADCAST_WRITES | SH_INDEX(sh_num);
2957 else if (sh_num == 0xffffffff)
2958 data |= SH_BROADCAST_WRITES | SE_INDEX(se_num);
2959 else
2960 data |= SH_INDEX(sh_num) | SE_INDEX(se_num);
2961 WREG32(GRBM_GFX_INDEX, data);
2962}
2963
2964static u32 si_create_bitmask(u32 bit_width)
2965{
2966 u32 i, mask = 0;
2967
2968 for (i = 0; i < bit_width; i++) {
2969 mask <<= 1;
2970 mask |= 1;
2971 }
2972 return mask;
2973}
2974
2975static u32 si_get_cu_enabled(struct radeon_device *rdev, u32 cu_per_sh)
2976{
2977 u32 data, mask;
2978
2979 data = RREG32(CC_GC_SHADER_ARRAY_CONFIG);
2980 if (data & 1)
2981 data &= INACTIVE_CUS_MASK;
2982 else
2983 data = 0;
2984 data |= RREG32(GC_USER_SHADER_ARRAY_CONFIG);
2985
2986 data >>= INACTIVE_CUS_SHIFT;
2987
2988 mask = si_create_bitmask(cu_per_sh);
2989
2990 return ~data & mask;
2991}
2992
2993static void si_setup_spi(struct radeon_device *rdev,
2994 u32 se_num, u32 sh_per_se,
2995 u32 cu_per_sh)
2996{
2997 int i, j, k;
2998 u32 data, mask, active_cu;
2999
3000 for (i = 0; i < se_num; i++) {
3001 for (j = 0; j < sh_per_se; j++) {
3002 si_select_se_sh(rdev, i, j);
3003 data = RREG32(SPI_STATIC_THREAD_MGMT_3);
3004 active_cu = si_get_cu_enabled(rdev, cu_per_sh);
3005
3006 mask = 1;
3007 for (k = 0; k < 16; k++) {
3008 mask <<= k;
3009 if (active_cu & mask) {
3010 data &= ~mask;
3011 WREG32(SPI_STATIC_THREAD_MGMT_3, data);
3012 break;
3013 }
3014 }
3015 }
3016 }
3017 si_select_se_sh(rdev, 0xffffffff, 0xffffffff);
3018}
3019
3020static u32 si_get_rb_disabled(struct radeon_device *rdev,
3021 u32 max_rb_num_per_se,
3022 u32 sh_per_se)
3023{
3024 u32 data, mask;
3025
3026 data = RREG32(CC_RB_BACKEND_DISABLE);
3027 if (data & 1)
3028 data &= BACKEND_DISABLE_MASK;
3029 else
3030 data = 0;
3031 data |= RREG32(GC_USER_RB_BACKEND_DISABLE);
3032
3033 data >>= BACKEND_DISABLE_SHIFT;
3034
3035 mask = si_create_bitmask(max_rb_num_per_se / sh_per_se);
3036
3037 return data & mask;
3038}
3039
3040static void si_setup_rb(struct radeon_device *rdev,
3041 u32 se_num, u32 sh_per_se,
3042 u32 max_rb_num_per_se)
3043{
3044 int i, j;
3045 u32 data, mask;
3046 u32 disabled_rbs = 0;
3047 u32 enabled_rbs = 0;
3048
3049 for (i = 0; i < se_num; i++) {
3050 for (j = 0; j < sh_per_se; j++) {
3051 si_select_se_sh(rdev, i, j);
3052 data = si_get_rb_disabled(rdev, max_rb_num_per_se, sh_per_se);
3053 disabled_rbs |= data << ((i * sh_per_se + j) * TAHITI_RB_BITMAP_WIDTH_PER_SH);
3054 }
3055 }
3056 si_select_se_sh(rdev, 0xffffffff, 0xffffffff);
3057
3058 mask = 1;
3059 for (i = 0; i < max_rb_num_per_se * se_num; i++) {
3060 if (!(disabled_rbs & mask))
3061 enabled_rbs |= mask;
3062 mask <<= 1;
3063 }
3064
3065 rdev->config.si.backend_enable_mask = enabled_rbs;
3066
3067 for (i = 0; i < se_num; i++) {
3068 si_select_se_sh(rdev, i, 0xffffffff);
3069 data = 0;
3070 for (j = 0; j < sh_per_se; j++) {
3071 switch (enabled_rbs & 3) {
3072 case 1:
3073 data |= (RASTER_CONFIG_RB_MAP_0 << (i * sh_per_se + j) * 2);
3074 break;
3075 case 2:
3076 data |= (RASTER_CONFIG_RB_MAP_3 << (i * sh_per_se + j) * 2);
3077 break;
3078 case 3:
3079 default:
3080 data |= (RASTER_CONFIG_RB_MAP_2 << (i * sh_per_se + j) * 2);
3081 break;
3082 }
3083 enabled_rbs >>= 2;
3084 }
3085 WREG32(PA_SC_RASTER_CONFIG, data);
3086 }
3087 si_select_se_sh(rdev, 0xffffffff, 0xffffffff);
3088}
3089
3090static void si_gpu_init(struct radeon_device *rdev)
3091{
3092 u32 gb_addr_config = 0;
3093 u32 mc_shared_chmap, mc_arb_ramcfg;
3094 u32 sx_debug_1;
3095 u32 hdp_host_path_cntl;
3096 u32 tmp;
3097 int i, j;
3098
3099 switch (rdev->family) {
3100 case CHIP_TAHITI:
3101 rdev->config.si.max_shader_engines = 2;
3102 rdev->config.si.max_tile_pipes = 12;
3103 rdev->config.si.max_cu_per_sh = 8;
3104 rdev->config.si.max_sh_per_se = 2;
3105 rdev->config.si.max_backends_per_se = 4;
3106 rdev->config.si.max_texture_channel_caches = 12;
3107 rdev->config.si.max_gprs = 256;
3108 rdev->config.si.max_gs_threads = 32;
3109 rdev->config.si.max_hw_contexts = 8;
3110
3111 rdev->config.si.sc_prim_fifo_size_frontend = 0x20;
3112 rdev->config.si.sc_prim_fifo_size_backend = 0x100;
3113 rdev->config.si.sc_hiz_tile_fifo_size = 0x30;
3114 rdev->config.si.sc_earlyz_tile_fifo_size = 0x130;
3115 gb_addr_config = TAHITI_GB_ADDR_CONFIG_GOLDEN;
3116 break;
3117 case CHIP_PITCAIRN:
3118 rdev->config.si.max_shader_engines = 2;
3119 rdev->config.si.max_tile_pipes = 8;
3120 rdev->config.si.max_cu_per_sh = 5;
3121 rdev->config.si.max_sh_per_se = 2;
3122 rdev->config.si.max_backends_per_se = 4;
3123 rdev->config.si.max_texture_channel_caches = 8;
3124 rdev->config.si.max_gprs = 256;
3125 rdev->config.si.max_gs_threads = 32;
3126 rdev->config.si.max_hw_contexts = 8;
3127
3128 rdev->config.si.sc_prim_fifo_size_frontend = 0x20;
3129 rdev->config.si.sc_prim_fifo_size_backend = 0x100;
3130 rdev->config.si.sc_hiz_tile_fifo_size = 0x30;
3131 rdev->config.si.sc_earlyz_tile_fifo_size = 0x130;
3132 gb_addr_config = TAHITI_GB_ADDR_CONFIG_GOLDEN;
3133 break;
3134 case CHIP_VERDE:
3135 default:
3136 rdev->config.si.max_shader_engines = 1;
3137 rdev->config.si.max_tile_pipes = 4;
3138 rdev->config.si.max_cu_per_sh = 5;
3139 rdev->config.si.max_sh_per_se = 2;
3140 rdev->config.si.max_backends_per_se = 4;
3141 rdev->config.si.max_texture_channel_caches = 4;
3142 rdev->config.si.max_gprs = 256;
3143 rdev->config.si.max_gs_threads = 32;
3144 rdev->config.si.max_hw_contexts = 8;
3145
3146 rdev->config.si.sc_prim_fifo_size_frontend = 0x20;
3147 rdev->config.si.sc_prim_fifo_size_backend = 0x40;
3148 rdev->config.si.sc_hiz_tile_fifo_size = 0x30;
3149 rdev->config.si.sc_earlyz_tile_fifo_size = 0x130;
3150 gb_addr_config = VERDE_GB_ADDR_CONFIG_GOLDEN;
3151 break;
3152 case CHIP_OLAND:
3153 rdev->config.si.max_shader_engines = 1;
3154 rdev->config.si.max_tile_pipes = 4;
3155 rdev->config.si.max_cu_per_sh = 6;
3156 rdev->config.si.max_sh_per_se = 1;
3157 rdev->config.si.max_backends_per_se = 2;
3158 rdev->config.si.max_texture_channel_caches = 4;
3159 rdev->config.si.max_gprs = 256;
3160 rdev->config.si.max_gs_threads = 16;
3161 rdev->config.si.max_hw_contexts = 8;
3162
3163 rdev->config.si.sc_prim_fifo_size_frontend = 0x20;
3164 rdev->config.si.sc_prim_fifo_size_backend = 0x40;
3165 rdev->config.si.sc_hiz_tile_fifo_size = 0x30;
3166 rdev->config.si.sc_earlyz_tile_fifo_size = 0x130;
3167 gb_addr_config = VERDE_GB_ADDR_CONFIG_GOLDEN;
3168 break;
3169 case CHIP_HAINAN:
3170 rdev->config.si.max_shader_engines = 1;
3171 rdev->config.si.max_tile_pipes = 4;
3172 rdev->config.si.max_cu_per_sh = 5;
3173 rdev->config.si.max_sh_per_se = 1;
3174 rdev->config.si.max_backends_per_se = 1;
3175 rdev->config.si.max_texture_channel_caches = 2;
3176 rdev->config.si.max_gprs = 256;
3177 rdev->config.si.max_gs_threads = 16;
3178 rdev->config.si.max_hw_contexts = 8;
3179
3180 rdev->config.si.sc_prim_fifo_size_frontend = 0x20;
3181 rdev->config.si.sc_prim_fifo_size_backend = 0x40;
3182 rdev->config.si.sc_hiz_tile_fifo_size = 0x30;
3183 rdev->config.si.sc_earlyz_tile_fifo_size = 0x130;
3184 gb_addr_config = HAINAN_GB_ADDR_CONFIG_GOLDEN;
3185 break;
3186 }
3187
3188 /* Initialize HDP */
3189 for (i = 0, j = 0; i < 32; i++, j += 0x18) {
3190 WREG32((0x2c14 + j), 0x00000000);
3191 WREG32((0x2c18 + j), 0x00000000);
3192 WREG32((0x2c1c + j), 0x00000000);
3193 WREG32((0x2c20 + j), 0x00000000);
3194 WREG32((0x2c24 + j), 0x00000000);
3195 }
3196
3197 WREG32(GRBM_CNTL, GRBM_READ_TIMEOUT(0xff));
3198 WREG32(SRBM_INT_CNTL, 1);
3199 WREG32(SRBM_INT_ACK, 1);
3200
3201 evergreen_fix_pci_max_read_req_size(rdev);
3202
3203 WREG32(BIF_FB_EN, FB_READ_EN | FB_WRITE_EN);
3204
3205 mc_shared_chmap = RREG32(MC_SHARED_CHMAP);
3206 mc_arb_ramcfg = RREG32(MC_ARB_RAMCFG);
3207
3208 rdev->config.si.num_tile_pipes = rdev->config.si.max_tile_pipes;
3209 rdev->config.si.mem_max_burst_length_bytes = 256;
3210 tmp = (mc_arb_ramcfg & NOOFCOLS_MASK) >> NOOFCOLS_SHIFT;
3211 rdev->config.si.mem_row_size_in_kb = (4 * (1 << (8 + tmp))) / 1024;
3212 if (rdev->config.si.mem_row_size_in_kb > 4)
3213 rdev->config.si.mem_row_size_in_kb = 4;
3214 /* XXX use MC settings? */
3215 rdev->config.si.shader_engine_tile_size = 32;
3216 rdev->config.si.num_gpus = 1;
3217 rdev->config.si.multi_gpu_tile_size = 64;
3218
3219 /* fix up row size */
3220 gb_addr_config &= ~ROW_SIZE_MASK;
3221 switch (rdev->config.si.mem_row_size_in_kb) {
3222 case 1:
3223 default:
3224 gb_addr_config |= ROW_SIZE(0);
3225 break;
3226 case 2:
3227 gb_addr_config |= ROW_SIZE(1);
3228 break;
3229 case 4:
3230 gb_addr_config |= ROW_SIZE(2);
3231 break;
3232 }
3233
3234 /* setup tiling info dword. gb_addr_config is not adequate since it does
3235 * not have bank info, so create a custom tiling dword.
3236 * bits 3:0 num_pipes
3237 * bits 7:4 num_banks
3238 * bits 11:8 group_size
3239 * bits 15:12 row_size
3240 */
3241 rdev->config.si.tile_config = 0;
3242 switch (rdev->config.si.num_tile_pipes) {
3243 case 1:
3244 rdev->config.si.tile_config |= (0 << 0);
3245 break;
3246 case 2:
3247 rdev->config.si.tile_config |= (1 << 0);
3248 break;
3249 case 4:
3250 rdev->config.si.tile_config |= (2 << 0);
3251 break;
3252 case 8:
3253 default:
3254 /* XXX what about 12? */
3255 rdev->config.si.tile_config |= (3 << 0);
3256 break;
3257 }
3258 switch ((mc_arb_ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT) {
3259 case 0: /* four banks */
3260 rdev->config.si.tile_config |= 0 << 4;
3261 break;
3262 case 1: /* eight banks */
3263 rdev->config.si.tile_config |= 1 << 4;
3264 break;
3265 case 2: /* sixteen banks */
3266 default:
3267 rdev->config.si.tile_config |= 2 << 4;
3268 break;
3269 }
3270 rdev->config.si.tile_config |=
3271 ((gb_addr_config & PIPE_INTERLEAVE_SIZE_MASK) >> PIPE_INTERLEAVE_SIZE_SHIFT) << 8;
3272 rdev->config.si.tile_config |=
3273 ((gb_addr_config & ROW_SIZE_MASK) >> ROW_SIZE_SHIFT) << 12;
3274
3275 WREG32(GB_ADDR_CONFIG, gb_addr_config);
3276 WREG32(DMIF_ADDR_CONFIG, gb_addr_config);
3277 WREG32(DMIF_ADDR_CALC, gb_addr_config);
3278 WREG32(HDP_ADDR_CONFIG, gb_addr_config);
3279 WREG32(DMA_TILING_CONFIG + DMA0_REGISTER_OFFSET, gb_addr_config);
3280 WREG32(DMA_TILING_CONFIG + DMA1_REGISTER_OFFSET, gb_addr_config);
3281 if (rdev->has_uvd) {
3282 WREG32(UVD_UDEC_ADDR_CONFIG, gb_addr_config);
3283 WREG32(UVD_UDEC_DB_ADDR_CONFIG, gb_addr_config);
3284 WREG32(UVD_UDEC_DBW_ADDR_CONFIG, gb_addr_config);
3285 }
3286
3287 si_tiling_mode_table_init(rdev);
3288
3289 si_setup_rb(rdev, rdev->config.si.max_shader_engines,
3290 rdev->config.si.max_sh_per_se,
3291 rdev->config.si.max_backends_per_se);
3292
3293 si_setup_spi(rdev, rdev->config.si.max_shader_engines,
3294 rdev->config.si.max_sh_per_se,
3295 rdev->config.si.max_cu_per_sh);
3296
3297 rdev->config.si.active_cus = 0;
3298 for (i = 0; i < rdev->config.si.max_shader_engines; i++) {
3299 for (j = 0; j < rdev->config.si.max_sh_per_se; j++) {
3300 rdev->config.si.active_cus +=
3301 hweight32(si_get_cu_active_bitmap(rdev, i, j));
3302 }
3303 }
3304
3305 /* set HW defaults for 3D engine */
3306 WREG32(CP_QUEUE_THRESHOLDS, (ROQ_IB1_START(0x16) |
3307 ROQ_IB2_START(0x2b)));
3308 WREG32(CP_MEQ_THRESHOLDS, MEQ1_START(0x30) | MEQ2_START(0x60));
3309
3310 sx_debug_1 = RREG32(SX_DEBUG_1);
3311 WREG32(SX_DEBUG_1, sx_debug_1);
3312
3313 WREG32(SPI_CONFIG_CNTL_1, VTX_DONE_DELAY(4));
3314
3315 WREG32(PA_SC_FIFO_SIZE, (SC_FRONTEND_PRIM_FIFO_SIZE(rdev->config.si.sc_prim_fifo_size_frontend) |
3316 SC_BACKEND_PRIM_FIFO_SIZE(rdev->config.si.sc_prim_fifo_size_backend) |
3317 SC_HIZ_TILE_FIFO_SIZE(rdev->config.si.sc_hiz_tile_fifo_size) |
3318 SC_EARLYZ_TILE_FIFO_SIZE(rdev->config.si.sc_earlyz_tile_fifo_size)));
3319
3320 WREG32(VGT_NUM_INSTANCES, 1);
3321
3322 WREG32(CP_PERFMON_CNTL, 0);
3323
3324 WREG32(SQ_CONFIG, 0);
3325
3326 WREG32(PA_SC_FORCE_EOV_MAX_CNTS, (FORCE_EOV_MAX_CLK_CNT(4095) |
3327 FORCE_EOV_MAX_REZ_CNT(255)));
3328
3329 WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC) |
3330 AUTO_INVLD_EN(ES_AND_GS_AUTO));
3331
3332 WREG32(VGT_GS_VERTEX_REUSE, 16);
3333 WREG32(PA_SC_LINE_STIPPLE_STATE, 0);
3334
3335 WREG32(CB_PERFCOUNTER0_SELECT0, 0);
3336 WREG32(CB_PERFCOUNTER0_SELECT1, 0);
3337 WREG32(CB_PERFCOUNTER1_SELECT0, 0);
3338 WREG32(CB_PERFCOUNTER1_SELECT1, 0);
3339 WREG32(CB_PERFCOUNTER2_SELECT0, 0);
3340 WREG32(CB_PERFCOUNTER2_SELECT1, 0);
3341 WREG32(CB_PERFCOUNTER3_SELECT0, 0);
3342 WREG32(CB_PERFCOUNTER3_SELECT1, 0);
3343
3344 tmp = RREG32(HDP_MISC_CNTL);
3345 tmp |= HDP_FLUSH_INVALIDATE_CACHE;
3346 WREG32(HDP_MISC_CNTL, tmp);
3347
3348 hdp_host_path_cntl = RREG32(HDP_HOST_PATH_CNTL);
3349 WREG32(HDP_HOST_PATH_CNTL, hdp_host_path_cntl);
3350
3351 WREG32(PA_CL_ENHANCE, CLIP_VTX_REORDER_ENA | NUM_CLIP_SEQ(3));
3352
3353 udelay(50);
3354}
3355
3356/*
3357 * GPU scratch registers helpers function.
3358 */
3359static void si_scratch_init(struct radeon_device *rdev)
3360{
3361 int i;
3362
3363 rdev->scratch.num_reg = 7;
3364 rdev->scratch.reg_base = SCRATCH_REG0;
3365 for (i = 0; i < rdev->scratch.num_reg; i++) {
3366 rdev->scratch.free[i] = true;
3367 rdev->scratch.reg[i] = rdev->scratch.reg_base + (i * 4);
3368 }
3369}
3370
3371void si_fence_ring_emit(struct radeon_device *rdev,
3372 struct radeon_fence *fence)
3373{
3374 struct radeon_ring *ring = &rdev->ring[fence->ring];
3375 u64 addr = rdev->fence_drv[fence->ring].gpu_addr;
3376
3377 /* flush read cache over gart */
3378 radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
3379 radeon_ring_write(ring, (CP_COHER_CNTL2 - PACKET3_SET_CONFIG_REG_START) >> 2);
3380 radeon_ring_write(ring, 0);
3381 radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
3382 radeon_ring_write(ring, PACKET3_TCL1_ACTION_ENA |
3383 PACKET3_TC_ACTION_ENA |
3384 PACKET3_SH_KCACHE_ACTION_ENA |
3385 PACKET3_SH_ICACHE_ACTION_ENA);
3386 radeon_ring_write(ring, 0xFFFFFFFF);
3387 radeon_ring_write(ring, 0);
3388 radeon_ring_write(ring, 10); /* poll interval */
3389 /* EVENT_WRITE_EOP - flush caches, send int */
3390 radeon_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
3391 radeon_ring_write(ring, EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | EVENT_INDEX(5));
3392 radeon_ring_write(ring, lower_32_bits(addr));
3393 radeon_ring_write(ring, (upper_32_bits(addr) & 0xff) | DATA_SEL(1) | INT_SEL(2));
3394 radeon_ring_write(ring, fence->seq);
3395 radeon_ring_write(ring, 0);
3396}
3397
3398/*
3399 * IB stuff
3400 */
3401void si_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
3402{
3403 struct radeon_ring *ring = &rdev->ring[ib->ring];
3404 unsigned vm_id = ib->vm ? ib->vm->ids[ib->ring].id : 0;
3405 u32 header;
3406
3407 if (ib->is_const_ib) {
3408 /* set switch buffer packet before const IB */
3409 radeon_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
3410 radeon_ring_write(ring, 0);
3411
3412 header = PACKET3(PACKET3_INDIRECT_BUFFER_CONST, 2);
3413 } else {
3414 u32 next_rptr;
3415 if (ring->rptr_save_reg) {
3416 next_rptr = ring->wptr + 3 + 4 + 8;
3417 radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
3418 radeon_ring_write(ring, ((ring->rptr_save_reg -
3419 PACKET3_SET_CONFIG_REG_START) >> 2));
3420 radeon_ring_write(ring, next_rptr);
3421 } else if (rdev->wb.enabled) {
3422 next_rptr = ring->wptr + 5 + 4 + 8;
3423 radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
3424 radeon_ring_write(ring, (1 << 8));
3425 radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
3426 radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr));
3427 radeon_ring_write(ring, next_rptr);
3428 }
3429
3430 header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
3431 }
3432
3433 radeon_ring_write(ring, header);
3434 radeon_ring_write(ring,
3435#ifdef __BIG_ENDIAN
3436 (2 << 0) |
3437#endif
3438 (ib->gpu_addr & 0xFFFFFFFC));
3439 radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF);
3440 radeon_ring_write(ring, ib->length_dw | (vm_id << 24));
3441
3442 if (!ib->is_const_ib) {
3443 /* flush read cache over gart for this vmid */
3444 radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
3445 radeon_ring_write(ring, (CP_COHER_CNTL2 - PACKET3_SET_CONFIG_REG_START) >> 2);
3446 radeon_ring_write(ring, vm_id);
3447 radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
3448 radeon_ring_write(ring, PACKET3_TCL1_ACTION_ENA |
3449 PACKET3_TC_ACTION_ENA |
3450 PACKET3_SH_KCACHE_ACTION_ENA |
3451 PACKET3_SH_ICACHE_ACTION_ENA);
3452 radeon_ring_write(ring, 0xFFFFFFFF);
3453 radeon_ring_write(ring, 0);
3454 radeon_ring_write(ring, 10); /* poll interval */
3455 }
3456}
3457
3458/*
3459 * CP.
3460 */
3461static void si_cp_enable(struct radeon_device *rdev, bool enable)
3462{
3463 if (enable)
3464 WREG32(CP_ME_CNTL, 0);
3465 else {
3466 if (rdev->asic->copy.copy_ring_index == RADEON_RING_TYPE_GFX_INDEX)
3467 radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size);
3468 WREG32(CP_ME_CNTL, (CP_ME_HALT | CP_PFP_HALT | CP_CE_HALT));
3469 WREG32(SCRATCH_UMSK, 0);
3470 rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false;
3471 rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX].ready = false;
3472 rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX].ready = false;
3473 }
3474 udelay(50);
3475}
3476
3477static int si_cp_load_microcode(struct radeon_device *rdev)
3478{
3479 int i;
3480
3481 if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw)
3482 return -EINVAL;
3483
3484 si_cp_enable(rdev, false);
3485
3486 if (rdev->new_fw) {
3487 const struct gfx_firmware_header_v1_0 *pfp_hdr =
3488 (const struct gfx_firmware_header_v1_0 *)rdev->pfp_fw->data;
3489 const struct gfx_firmware_header_v1_0 *ce_hdr =
3490 (const struct gfx_firmware_header_v1_0 *)rdev->ce_fw->data;
3491 const struct gfx_firmware_header_v1_0 *me_hdr =
3492 (const struct gfx_firmware_header_v1_0 *)rdev->me_fw->data;
3493 const __le32 *fw_data;
3494 u32 fw_size;
3495
3496 radeon_ucode_print_gfx_hdr(&pfp_hdr->header);
3497 radeon_ucode_print_gfx_hdr(&ce_hdr->header);
3498 radeon_ucode_print_gfx_hdr(&me_hdr->header);
3499
3500 /* PFP */
3501 fw_data = (const __le32 *)
3502 (rdev->pfp_fw->data + le32_to_cpu(pfp_hdr->header.ucode_array_offset_bytes));
3503 fw_size = le32_to_cpu(pfp_hdr->header.ucode_size_bytes) / 4;
3504 WREG32(CP_PFP_UCODE_ADDR, 0);
3505 for (i = 0; i < fw_size; i++)
3506 WREG32(CP_PFP_UCODE_DATA, le32_to_cpup(fw_data++));
3507 WREG32(CP_PFP_UCODE_ADDR, 0);
3508
3509 /* CE */
3510 fw_data = (const __le32 *)
3511 (rdev->ce_fw->data + le32_to_cpu(ce_hdr->header.ucode_array_offset_bytes));
3512 fw_size = le32_to_cpu(ce_hdr->header.ucode_size_bytes) / 4;
3513 WREG32(CP_CE_UCODE_ADDR, 0);
3514 for (i = 0; i < fw_size; i++)
3515 WREG32(CP_CE_UCODE_DATA, le32_to_cpup(fw_data++));
3516 WREG32(CP_CE_UCODE_ADDR, 0);
3517
3518 /* ME */
3519 fw_data = (const __be32 *)
3520 (rdev->me_fw->data + le32_to_cpu(me_hdr->header.ucode_array_offset_bytes));
3521 fw_size = le32_to_cpu(me_hdr->header.ucode_size_bytes) / 4;
3522 WREG32(CP_ME_RAM_WADDR, 0);
3523 for (i = 0; i < fw_size; i++)
3524 WREG32(CP_ME_RAM_DATA, le32_to_cpup(fw_data++));
3525 WREG32(CP_ME_RAM_WADDR, 0);
3526 } else {
3527 const __be32 *fw_data;
3528
3529 /* PFP */
3530 fw_data = (const __be32 *)rdev->pfp_fw->data;
3531 WREG32(CP_PFP_UCODE_ADDR, 0);
3532 for (i = 0; i < SI_PFP_UCODE_SIZE; i++)
3533 WREG32(CP_PFP_UCODE_DATA, be32_to_cpup(fw_data++));
3534 WREG32(CP_PFP_UCODE_ADDR, 0);
3535
3536 /* CE */
3537 fw_data = (const __be32 *)rdev->ce_fw->data;
3538 WREG32(CP_CE_UCODE_ADDR, 0);
3539 for (i = 0; i < SI_CE_UCODE_SIZE; i++)
3540 WREG32(CP_CE_UCODE_DATA, be32_to_cpup(fw_data++));
3541 WREG32(CP_CE_UCODE_ADDR, 0);
3542
3543 /* ME */
3544 fw_data = (const __be32 *)rdev->me_fw->data;
3545 WREG32(CP_ME_RAM_WADDR, 0);
3546 for (i = 0; i < SI_PM4_UCODE_SIZE; i++)
3547 WREG32(CP_ME_RAM_DATA, be32_to_cpup(fw_data++));
3548 WREG32(CP_ME_RAM_WADDR, 0);
3549 }
3550
3551 WREG32(CP_PFP_UCODE_ADDR, 0);
3552 WREG32(CP_CE_UCODE_ADDR, 0);
3553 WREG32(CP_ME_RAM_WADDR, 0);
3554 WREG32(CP_ME_RAM_RADDR, 0);
3555 return 0;
3556}
3557
3558static int si_cp_start(struct radeon_device *rdev)
3559{
3560 struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
3561 int r, i;
3562
3563 r = radeon_ring_lock(rdev, ring, 7 + 4);
3564 if (r) {
3565 DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
3566 return r;
3567 }
3568 /* init the CP */
3569 radeon_ring_write(ring, PACKET3(PACKET3_ME_INITIALIZE, 5));
3570 radeon_ring_write(ring, 0x1);
3571 radeon_ring_write(ring, 0x0);
3572 radeon_ring_write(ring, rdev->config.si.max_hw_contexts - 1);
3573 radeon_ring_write(ring, PACKET3_ME_INITIALIZE_DEVICE_ID(1));
3574 radeon_ring_write(ring, 0);
3575 radeon_ring_write(ring, 0);
3576
3577 /* init the CE partitions */
3578 radeon_ring_write(ring, PACKET3(PACKET3_SET_BASE, 2));
3579 radeon_ring_write(ring, PACKET3_BASE_INDEX(CE_PARTITION_BASE));
3580 radeon_ring_write(ring, 0xc000);
3581 radeon_ring_write(ring, 0xe000);
3582 radeon_ring_unlock_commit(rdev, ring, false);
3583
3584 si_cp_enable(rdev, true);
3585
3586 r = radeon_ring_lock(rdev, ring, si_default_size + 10);
3587 if (r) {
3588 DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r);
3589 return r;
3590 }
3591
3592 /* setup clear context state */
3593 radeon_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
3594 radeon_ring_write(ring, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
3595
3596 for (i = 0; i < si_default_size; i++)
3597 radeon_ring_write(ring, si_default_state[i]);
3598
3599 radeon_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
3600 radeon_ring_write(ring, PACKET3_PREAMBLE_END_CLEAR_STATE);
3601
3602 /* set clear context state */
3603 radeon_ring_write(ring, PACKET3(PACKET3_CLEAR_STATE, 0));
3604 radeon_ring_write(ring, 0);
3605
3606 radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
3607 radeon_ring_write(ring, 0x00000316);
3608 radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
3609 radeon_ring_write(ring, 0x00000010); /* VGT_OUT_DEALLOC_CNTL */
3610
3611 radeon_ring_unlock_commit(rdev, ring, false);
3612
3613 for (i = RADEON_RING_TYPE_GFX_INDEX; i <= CAYMAN_RING_TYPE_CP2_INDEX; ++i) {
3614 ring = &rdev->ring[i];
3615 r = radeon_ring_lock(rdev, ring, 2);
3616
3617 /* clear the compute context state */
3618 radeon_ring_write(ring, PACKET3_COMPUTE(PACKET3_CLEAR_STATE, 0));
3619 radeon_ring_write(ring, 0);
3620
3621 radeon_ring_unlock_commit(rdev, ring, false);
3622 }
3623
3624 return 0;
3625}
3626
3627static void si_cp_fini(struct radeon_device *rdev)
3628{
3629 struct radeon_ring *ring;
3630 si_cp_enable(rdev, false);
3631
3632 ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
3633 radeon_ring_fini(rdev, ring);
3634 radeon_scratch_free(rdev, ring->rptr_save_reg);
3635
3636 ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX];
3637 radeon_ring_fini(rdev, ring);
3638 radeon_scratch_free(rdev, ring->rptr_save_reg);
3639
3640 ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX];
3641 radeon_ring_fini(rdev, ring);
3642 radeon_scratch_free(rdev, ring->rptr_save_reg);
3643}
3644
3645static int si_cp_resume(struct radeon_device *rdev)
3646{
3647 struct radeon_ring *ring;
3648 u32 tmp;
3649 u32 rb_bufsz;
3650 int r;
3651
3652 si_enable_gui_idle_interrupt(rdev, false);
3653
3654 WREG32(CP_SEM_WAIT_TIMER, 0x0);
3655 WREG32(CP_SEM_INCOMPLETE_TIMER_CNTL, 0x0);
3656
3657 /* Set the write pointer delay */
3658 WREG32(CP_RB_WPTR_DELAY, 0);
3659
3660 WREG32(CP_DEBUG, 0);
3661 WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF);
3662
3663 /* ring 0 - compute and gfx */
3664 /* Set ring buffer size */
3665 ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
3666 rb_bufsz = order_base_2(ring->ring_size / 8);
3667 tmp = (order_base_2(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
3668#ifdef __BIG_ENDIAN
3669 tmp |= BUF_SWAP_32BIT;
3670#endif
3671 WREG32(CP_RB0_CNTL, tmp);
3672
3673 /* Initialize the ring buffer's read and write pointers */
3674 WREG32(CP_RB0_CNTL, tmp | RB_RPTR_WR_ENA);
3675 ring->wptr = 0;
3676 WREG32(CP_RB0_WPTR, ring->wptr);
3677
3678 /* set the wb address whether it's enabled or not */
3679 WREG32(CP_RB0_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC);
3680 WREG32(CP_RB0_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF);
3681
3682 if (rdev->wb.enabled)
3683 WREG32(SCRATCH_UMSK, 0xff);
3684 else {
3685 tmp |= RB_NO_UPDATE;
3686 WREG32(SCRATCH_UMSK, 0);
3687 }
3688
3689 mdelay(1);
3690 WREG32(CP_RB0_CNTL, tmp);
3691
3692 WREG32(CP_RB0_BASE, ring->gpu_addr >> 8);
3693
3694 /* ring1 - compute only */
3695 /* Set ring buffer size */
3696 ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX];
3697 rb_bufsz = order_base_2(ring->ring_size / 8);
3698 tmp = (order_base_2(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
3699#ifdef __BIG_ENDIAN
3700 tmp |= BUF_SWAP_32BIT;
3701#endif
3702 WREG32(CP_RB1_CNTL, tmp);
3703
3704 /* Initialize the ring buffer's read and write pointers */
3705 WREG32(CP_RB1_CNTL, tmp | RB_RPTR_WR_ENA);
3706 ring->wptr = 0;
3707 WREG32(CP_RB1_WPTR, ring->wptr);
3708
3709 /* set the wb address whether it's enabled or not */
3710 WREG32(CP_RB1_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP1_RPTR_OFFSET) & 0xFFFFFFFC);
3711 WREG32(CP_RB1_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP1_RPTR_OFFSET) & 0xFF);
3712
3713 mdelay(1);
3714 WREG32(CP_RB1_CNTL, tmp);
3715
3716 WREG32(CP_RB1_BASE, ring->gpu_addr >> 8);
3717
3718 /* ring2 - compute only */
3719 /* Set ring buffer size */
3720 ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX];
3721 rb_bufsz = order_base_2(ring->ring_size / 8);
3722 tmp = (order_base_2(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
3723#ifdef __BIG_ENDIAN
3724 tmp |= BUF_SWAP_32BIT;
3725#endif
3726 WREG32(CP_RB2_CNTL, tmp);
3727
3728 /* Initialize the ring buffer's read and write pointers */
3729 WREG32(CP_RB2_CNTL, tmp | RB_RPTR_WR_ENA);
3730 ring->wptr = 0;
3731 WREG32(CP_RB2_WPTR, ring->wptr);
3732
3733 /* set the wb address whether it's enabled or not */
3734 WREG32(CP_RB2_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP2_RPTR_OFFSET) & 0xFFFFFFFC);
3735 WREG32(CP_RB2_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP2_RPTR_OFFSET) & 0xFF);
3736
3737 mdelay(1);
3738 WREG32(CP_RB2_CNTL, tmp);
3739
3740 WREG32(CP_RB2_BASE, ring->gpu_addr >> 8);
3741
3742 /* start the rings */
3743 si_cp_start(rdev);
3744 rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = true;
3745 rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX].ready = true;
3746 rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX].ready = true;
3747 r = radeon_ring_test(rdev, RADEON_RING_TYPE_GFX_INDEX, &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]);
3748 if (r) {
3749 rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false;
3750 rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX].ready = false;
3751 rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX].ready = false;
3752 return r;
3753 }
3754 r = radeon_ring_test(rdev, CAYMAN_RING_TYPE_CP1_INDEX, &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]);
3755 if (r) {
3756 rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX].ready = false;
3757 }
3758 r = radeon_ring_test(rdev, CAYMAN_RING_TYPE_CP2_INDEX, &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]);
3759 if (r) {
3760 rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX].ready = false;
3761 }
3762
3763 si_enable_gui_idle_interrupt(rdev, true);
3764
3765 if (rdev->asic->copy.copy_ring_index == RADEON_RING_TYPE_GFX_INDEX)
3766 radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size);
3767
3768 return 0;
3769}
3770
3771u32 si_gpu_check_soft_reset(struct radeon_device *rdev)
3772{
3773 u32 reset_mask = 0;
3774 u32 tmp;
3775
3776 /* GRBM_STATUS */
3777 tmp = RREG32(GRBM_STATUS);
3778 if (tmp & (PA_BUSY | SC_BUSY |
3779 BCI_BUSY | SX_BUSY |
3780 TA_BUSY | VGT_BUSY |
3781 DB_BUSY | CB_BUSY |
3782 GDS_BUSY | SPI_BUSY |
3783 IA_BUSY | IA_BUSY_NO_DMA))
3784 reset_mask |= RADEON_RESET_GFX;
3785
3786 if (tmp & (CF_RQ_PENDING | PF_RQ_PENDING |
3787 CP_BUSY | CP_COHERENCY_BUSY))
3788 reset_mask |= RADEON_RESET_CP;
3789
3790 if (tmp & GRBM_EE_BUSY)
3791 reset_mask |= RADEON_RESET_GRBM | RADEON_RESET_GFX | RADEON_RESET_CP;
3792
3793 /* GRBM_STATUS2 */
3794 tmp = RREG32(GRBM_STATUS2);
3795 if (tmp & (RLC_RQ_PENDING | RLC_BUSY))
3796 reset_mask |= RADEON_RESET_RLC;
3797
3798 /* DMA_STATUS_REG 0 */
3799 tmp = RREG32(DMA_STATUS_REG + DMA0_REGISTER_OFFSET);
3800 if (!(tmp & DMA_IDLE))
3801 reset_mask |= RADEON_RESET_DMA;
3802
3803 /* DMA_STATUS_REG 1 */
3804 tmp = RREG32(DMA_STATUS_REG + DMA1_REGISTER_OFFSET);
3805 if (!(tmp & DMA_IDLE))
3806 reset_mask |= RADEON_RESET_DMA1;
3807
3808 /* SRBM_STATUS2 */
3809 tmp = RREG32(SRBM_STATUS2);
3810 if (tmp & DMA_BUSY)
3811 reset_mask |= RADEON_RESET_DMA;
3812
3813 if (tmp & DMA1_BUSY)
3814 reset_mask |= RADEON_RESET_DMA1;
3815
3816 /* SRBM_STATUS */
3817 tmp = RREG32(SRBM_STATUS);
3818
3819 if (tmp & IH_BUSY)
3820 reset_mask |= RADEON_RESET_IH;
3821
3822 if (tmp & SEM_BUSY)
3823 reset_mask |= RADEON_RESET_SEM;
3824
3825 if (tmp & GRBM_RQ_PENDING)
3826 reset_mask |= RADEON_RESET_GRBM;
3827
3828 if (tmp & VMC_BUSY)
3829 reset_mask |= RADEON_RESET_VMC;
3830
3831 if (tmp & (MCB_BUSY | MCB_NON_DISPLAY_BUSY |
3832 MCC_BUSY | MCD_BUSY))
3833 reset_mask |= RADEON_RESET_MC;
3834
3835 if (evergreen_is_display_hung(rdev))
3836 reset_mask |= RADEON_RESET_DISPLAY;
3837
3838 /* VM_L2_STATUS */
3839 tmp = RREG32(VM_L2_STATUS);
3840 if (tmp & L2_BUSY)
3841 reset_mask |= RADEON_RESET_VMC;
3842
3843 /* Skip MC reset as it's mostly likely not hung, just busy */
3844 if (reset_mask & RADEON_RESET_MC) {
3845 DRM_DEBUG("MC busy: 0x%08X, clearing.\n", reset_mask);
3846 reset_mask &= ~RADEON_RESET_MC;
3847 }
3848
3849 return reset_mask;
3850}
3851
3852static void si_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask)
3853{
3854 struct evergreen_mc_save save;
3855 u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
3856 u32 tmp;
3857
3858 if (reset_mask == 0)
3859 return;
3860
3861 dev_info(rdev->dev, "GPU softreset: 0x%08X\n", reset_mask);
3862
3863 evergreen_print_gpu_status_regs(rdev);
3864 dev_info(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x%08X\n",
3865 RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR));
3866 dev_info(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
3867 RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS));
3868
3869 /* disable PG/CG */
3870 si_fini_pg(rdev);
3871 si_fini_cg(rdev);
3872
3873 /* stop the rlc */
3874 si_rlc_stop(rdev);
3875
3876 /* Disable CP parsing/prefetching */
3877 WREG32(CP_ME_CNTL, CP_ME_HALT | CP_PFP_HALT | CP_CE_HALT);
3878
3879 if (reset_mask & RADEON_RESET_DMA) {
3880 /* dma0 */
3881 tmp = RREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET);
3882 tmp &= ~DMA_RB_ENABLE;
3883 WREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET, tmp);
3884 }
3885 if (reset_mask & RADEON_RESET_DMA1) {
3886 /* dma1 */
3887 tmp = RREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET);
3888 tmp &= ~DMA_RB_ENABLE;
3889 WREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET, tmp);
3890 }
3891
3892 udelay(50);
3893
3894 evergreen_mc_stop(rdev, &save);
3895 if (evergreen_mc_wait_for_idle(rdev)) {
3896 dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
3897 }
3898
3899 if (reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP)) {
3900 grbm_soft_reset = SOFT_RESET_CB |
3901 SOFT_RESET_DB |
3902 SOFT_RESET_GDS |
3903 SOFT_RESET_PA |
3904 SOFT_RESET_SC |
3905 SOFT_RESET_BCI |
3906 SOFT_RESET_SPI |
3907 SOFT_RESET_SX |
3908 SOFT_RESET_TC |
3909 SOFT_RESET_TA |
3910 SOFT_RESET_VGT |
3911 SOFT_RESET_IA;
3912 }
3913
3914 if (reset_mask & RADEON_RESET_CP) {
3915 grbm_soft_reset |= SOFT_RESET_CP | SOFT_RESET_VGT;
3916
3917 srbm_soft_reset |= SOFT_RESET_GRBM;
3918 }
3919
3920 if (reset_mask & RADEON_RESET_DMA)
3921 srbm_soft_reset |= SOFT_RESET_DMA;
3922
3923 if (reset_mask & RADEON_RESET_DMA1)
3924 srbm_soft_reset |= SOFT_RESET_DMA1;
3925
3926 if (reset_mask & RADEON_RESET_DISPLAY)
3927 srbm_soft_reset |= SOFT_RESET_DC;
3928
3929 if (reset_mask & RADEON_RESET_RLC)
3930 grbm_soft_reset |= SOFT_RESET_RLC;
3931
3932 if (reset_mask & RADEON_RESET_SEM)
3933 srbm_soft_reset |= SOFT_RESET_SEM;
3934
3935 if (reset_mask & RADEON_RESET_IH)
3936 srbm_soft_reset |= SOFT_RESET_IH;
3937
3938 if (reset_mask & RADEON_RESET_GRBM)
3939 srbm_soft_reset |= SOFT_RESET_GRBM;
3940
3941 if (reset_mask & RADEON_RESET_VMC)
3942 srbm_soft_reset |= SOFT_RESET_VMC;
3943
3944 if (reset_mask & RADEON_RESET_MC)
3945 srbm_soft_reset |= SOFT_RESET_MC;
3946
3947 if (grbm_soft_reset) {
3948 tmp = RREG32(GRBM_SOFT_RESET);
3949 tmp |= grbm_soft_reset;
3950 dev_info(rdev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
3951 WREG32(GRBM_SOFT_RESET, tmp);
3952 tmp = RREG32(GRBM_SOFT_RESET);
3953
3954 udelay(50);
3955
3956 tmp &= ~grbm_soft_reset;
3957 WREG32(GRBM_SOFT_RESET, tmp);
3958 tmp = RREG32(GRBM_SOFT_RESET);
3959 }
3960
3961 if (srbm_soft_reset) {
3962 tmp = RREG32(SRBM_SOFT_RESET);
3963 tmp |= srbm_soft_reset;
3964 dev_info(rdev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
3965 WREG32(SRBM_SOFT_RESET, tmp);
3966 tmp = RREG32(SRBM_SOFT_RESET);
3967
3968 udelay(50);
3969
3970 tmp &= ~srbm_soft_reset;
3971 WREG32(SRBM_SOFT_RESET, tmp);
3972 tmp = RREG32(SRBM_SOFT_RESET);
3973 }
3974
3975 /* Wait a little for things to settle down */
3976 udelay(50);
3977
3978 evergreen_mc_resume(rdev, &save);
3979 udelay(50);
3980
3981 evergreen_print_gpu_status_regs(rdev);
3982}
3983
3984static void si_set_clk_bypass_mode(struct radeon_device *rdev)
3985{
3986 u32 tmp, i;
3987
3988 tmp = RREG32(CG_SPLL_FUNC_CNTL);
3989 tmp |= SPLL_BYPASS_EN;
3990 WREG32(CG_SPLL_FUNC_CNTL, tmp);
3991
3992 tmp = RREG32(CG_SPLL_FUNC_CNTL_2);
3993 tmp |= SPLL_CTLREQ_CHG;
3994 WREG32(CG_SPLL_FUNC_CNTL_2, tmp);
3995
3996 for (i = 0; i < rdev->usec_timeout; i++) {
3997 if (RREG32(SPLL_STATUS) & SPLL_CHG_STATUS)
3998 break;
3999 udelay(1);
4000 }
4001
4002 tmp = RREG32(CG_SPLL_FUNC_CNTL_2);
4003 tmp &= ~(SPLL_CTLREQ_CHG | SCLK_MUX_UPDATE);
4004 WREG32(CG_SPLL_FUNC_CNTL_2, tmp);
4005
4006 tmp = RREG32(MPLL_CNTL_MODE);
4007 tmp &= ~MPLL_MCLK_SEL;
4008 WREG32(MPLL_CNTL_MODE, tmp);
4009}
4010
4011static void si_spll_powerdown(struct radeon_device *rdev)
4012{
4013 u32 tmp;
4014
4015 tmp = RREG32(SPLL_CNTL_MODE);
4016 tmp |= SPLL_SW_DIR_CONTROL;
4017 WREG32(SPLL_CNTL_MODE, tmp);
4018
4019 tmp = RREG32(CG_SPLL_FUNC_CNTL);
4020 tmp |= SPLL_RESET;
4021 WREG32(CG_SPLL_FUNC_CNTL, tmp);
4022
4023 tmp = RREG32(CG_SPLL_FUNC_CNTL);
4024 tmp |= SPLL_SLEEP;
4025 WREG32(CG_SPLL_FUNC_CNTL, tmp);
4026
4027 tmp = RREG32(SPLL_CNTL_MODE);
4028 tmp &= ~SPLL_SW_DIR_CONTROL;
4029 WREG32(SPLL_CNTL_MODE, tmp);
4030}
4031
4032static void si_gpu_pci_config_reset(struct radeon_device *rdev)
4033{
4034 struct evergreen_mc_save save;
4035 u32 tmp, i;
4036
4037 dev_info(rdev->dev, "GPU pci config reset\n");
4038
4039 /* disable dpm? */
4040
4041 /* disable cg/pg */
4042 si_fini_pg(rdev);
4043 si_fini_cg(rdev);
4044
4045 /* Disable CP parsing/prefetching */
4046 WREG32(CP_ME_CNTL, CP_ME_HALT | CP_PFP_HALT | CP_CE_HALT);
4047 /* dma0 */
4048 tmp = RREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET);
4049 tmp &= ~DMA_RB_ENABLE;
4050 WREG32(DMA_RB_CNTL + DMA0_REGISTER_OFFSET, tmp);
4051 /* dma1 */
4052 tmp = RREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET);
4053 tmp &= ~DMA_RB_ENABLE;
4054 WREG32(DMA_RB_CNTL + DMA1_REGISTER_OFFSET, tmp);
4055 /* XXX other engines? */
4056
4057 /* halt the rlc, disable cp internal ints */
4058 si_rlc_stop(rdev);
4059
4060 udelay(50);
4061
4062 /* disable mem access */
4063 evergreen_mc_stop(rdev, &save);
4064 if (evergreen_mc_wait_for_idle(rdev)) {
4065 dev_warn(rdev->dev, "Wait for MC idle timed out !\n");
4066 }
4067
4068 /* set mclk/sclk to bypass */
4069 si_set_clk_bypass_mode(rdev);
4070 /* powerdown spll */
4071 si_spll_powerdown(rdev);
4072 /* disable BM */
4073 pci_clear_master(rdev->pdev);
4074 /* reset */
4075 radeon_pci_config_reset(rdev);
4076 /* wait for asic to come out of reset */
4077 for (i = 0; i < rdev->usec_timeout; i++) {
4078 if (RREG32(CONFIG_MEMSIZE) != 0xffffffff)
4079 break;
4080 udelay(1);
4081 }
4082}
4083
4084int si_asic_reset(struct radeon_device *rdev, bool hard)
4085{
4086 u32 reset_mask;
4087
4088 if (hard) {
4089 si_gpu_pci_config_reset(rdev);
4090 return 0;
4091 }
4092
4093 reset_mask = si_gpu_check_soft_reset(rdev);
4094
4095 if (reset_mask)
4096 r600_set_bios_scratch_engine_hung(rdev, true);
4097
4098 /* try soft reset */
4099 si_gpu_soft_reset(rdev, reset_mask);
4100
4101 reset_mask = si_gpu_check_soft_reset(rdev);
4102
4103 /* try pci config reset */
4104 if (reset_mask && radeon_hard_reset)
4105 si_gpu_pci_config_reset(rdev);
4106
4107 reset_mask = si_gpu_check_soft_reset(rdev);
4108
4109 if (!reset_mask)
4110 r600_set_bios_scratch_engine_hung(rdev, false);
4111
4112 return 0;
4113}
4114
4115/**
4116 * si_gfx_is_lockup - Check if the GFX engine is locked up
4117 *
4118 * @rdev: radeon_device pointer
4119 * @ring: radeon_ring structure holding ring information
4120 *
4121 * Check if the GFX engine is locked up.
4122 * Returns true if the engine appears to be locked up, false if not.
4123 */
4124bool si_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring)
4125{
4126 u32 reset_mask = si_gpu_check_soft_reset(rdev);
4127
4128 if (!(reset_mask & (RADEON_RESET_GFX |
4129 RADEON_RESET_COMPUTE |
4130 RADEON_RESET_CP))) {
4131 radeon_ring_lockup_update(rdev, ring);
4132 return false;
4133 }
4134 return radeon_ring_test_lockup(rdev, ring);
4135}
4136
4137/* MC */
4138static void si_mc_program(struct radeon_device *rdev)
4139{
4140 struct evergreen_mc_save save;
4141 u32 tmp;
4142 int i, j;
4143
4144 /* Initialize HDP */
4145 for (i = 0, j = 0; i < 32; i++, j += 0x18) {
4146 WREG32((0x2c14 + j), 0x00000000);
4147 WREG32((0x2c18 + j), 0x00000000);
4148 WREG32((0x2c1c + j), 0x00000000);
4149 WREG32((0x2c20 + j), 0x00000000);
4150 WREG32((0x2c24 + j), 0x00000000);
4151 }
4152 WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0);
4153
4154 evergreen_mc_stop(rdev, &save);
4155 if (radeon_mc_wait_for_idle(rdev)) {
4156 dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
4157 }
4158 if (!ASIC_IS_NODCE(rdev))
4159 /* Lockout access through VGA aperture*/
4160 WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE);
4161 /* Update configuration */
4162 WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR,
4163 rdev->mc.vram_start >> 12);
4164 WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR,
4165 rdev->mc.vram_end >> 12);
4166 WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
4167 rdev->vram_scratch.gpu_addr >> 12);
4168 tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16;
4169 tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF);
4170 WREG32(MC_VM_FB_LOCATION, tmp);
4171 /* XXX double check these! */
4172 WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8));
4173 WREG32(HDP_NONSURFACE_INFO, (2 << 7) | (1 << 30));
4174 WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF);
4175 WREG32(MC_VM_AGP_BASE, 0);
4176 WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF);
4177 WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF);
4178 if (radeon_mc_wait_for_idle(rdev)) {
4179 dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
4180 }
4181 evergreen_mc_resume(rdev, &save);
4182 if (!ASIC_IS_NODCE(rdev)) {
4183 /* we need to own VRAM, so turn off the VGA renderer here
4184 * to stop it overwriting our objects */
4185 rv515_vga_render_disable(rdev);
4186 }
4187}
4188
4189void si_vram_gtt_location(struct radeon_device *rdev,
4190 struct radeon_mc *mc)
4191{
4192 if (mc->mc_vram_size > 0xFFC0000000ULL) {
4193 /* leave room for at least 1024M GTT */
4194 dev_warn(rdev->dev, "limiting VRAM\n");
4195 mc->real_vram_size = 0xFFC0000000ULL;
4196 mc->mc_vram_size = 0xFFC0000000ULL;
4197 }
4198 radeon_vram_location(rdev, &rdev->mc, 0);
4199 rdev->mc.gtt_base_align = 0;
4200 radeon_gtt_location(rdev, mc);
4201}
4202
4203static int si_mc_init(struct radeon_device *rdev)
4204{
4205 u32 tmp;
4206 int chansize, numchan;
4207
4208 /* Get VRAM informations */
4209 rdev->mc.vram_is_ddr = true;
4210 tmp = RREG32(MC_ARB_RAMCFG);
4211 if (tmp & CHANSIZE_OVERRIDE) {
4212 chansize = 16;
4213 } else if (tmp & CHANSIZE_MASK) {
4214 chansize = 64;
4215 } else {
4216 chansize = 32;
4217 }
4218 tmp = RREG32(MC_SHARED_CHMAP);
4219 switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) {
4220 case 0:
4221 default:
4222 numchan = 1;
4223 break;
4224 case 1:
4225 numchan = 2;
4226 break;
4227 case 2:
4228 numchan = 4;
4229 break;
4230 case 3:
4231 numchan = 8;
4232 break;
4233 case 4:
4234 numchan = 3;
4235 break;
4236 case 5:
4237 numchan = 6;
4238 break;
4239 case 6:
4240 numchan = 10;
4241 break;
4242 case 7:
4243 numchan = 12;
4244 break;
4245 case 8:
4246 numchan = 16;
4247 break;
4248 }
4249 rdev->mc.vram_width = numchan * chansize;
4250 /* Could aper size report 0 ? */
4251 rdev->mc.aper_base = pci_resource_start(rdev->pdev, 0);
4252 rdev->mc.aper_size = pci_resource_len(rdev->pdev, 0);
4253 /* size in MB on si */
4254 tmp = RREG32(CONFIG_MEMSIZE);
4255 /* some boards may have garbage in the upper 16 bits */
4256 if (tmp & 0xffff0000) {
4257 DRM_INFO("Probable bad vram size: 0x%08x\n", tmp);
4258 if (tmp & 0xffff)
4259 tmp &= 0xffff;
4260 }
4261 rdev->mc.mc_vram_size = tmp * 1024ULL * 1024ULL;
4262 rdev->mc.real_vram_size = rdev->mc.mc_vram_size;
4263 rdev->mc.visible_vram_size = rdev->mc.aper_size;
4264 si_vram_gtt_location(rdev, &rdev->mc);
4265 radeon_update_bandwidth_info(rdev);
4266
4267 return 0;
4268}
4269
4270/*
4271 * GART
4272 */
4273void si_pcie_gart_tlb_flush(struct radeon_device *rdev)
4274{
4275 /* flush hdp cache */
4276 WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);
4277
4278 /* bits 0-15 are the VM contexts0-15 */
4279 WREG32(VM_INVALIDATE_REQUEST, 1);
4280}
4281
4282static int si_pcie_gart_enable(struct radeon_device *rdev)
4283{
4284 int r, i;
4285
4286 if (rdev->gart.robj == NULL) {
4287 dev_err(rdev->dev, "No VRAM object for PCIE GART.\n");
4288 return -EINVAL;
4289 }
4290 r = radeon_gart_table_vram_pin(rdev);
4291 if (r)
4292 return r;
4293 /* Setup TLB control */
4294 WREG32(MC_VM_MX_L1_TLB_CNTL,
4295 (0xA << 7) |
4296 ENABLE_L1_TLB |
4297 ENABLE_L1_FRAGMENT_PROCESSING |
4298 SYSTEM_ACCESS_MODE_NOT_IN_SYS |
4299 ENABLE_ADVANCED_DRIVER_MODEL |
4300 SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU);
4301 /* Setup L2 cache */
4302 WREG32(VM_L2_CNTL, ENABLE_L2_CACHE |
4303 ENABLE_L2_FRAGMENT_PROCESSING |
4304 ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
4305 ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE |
4306 EFFECTIVE_L2_QUEUE_SIZE(7) |
4307 CONTEXT1_IDENTITY_ACCESS_MODE(1));
4308 WREG32(VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS | INVALIDATE_L2_CACHE);
4309 WREG32(VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY |
4310 BANK_SELECT(4) |
4311 L2_CACHE_BIGK_FRAGMENT_SIZE(4));
4312 /* setup context0 */
4313 WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12);
4314 WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12);
4315 WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12);
4316 WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,
4317 (u32)(rdev->dummy_page.addr >> 12));
4318 WREG32(VM_CONTEXT0_CNTL2, 0);
4319 WREG32(VM_CONTEXT0_CNTL, (ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
4320 RANGE_PROTECTION_FAULT_ENABLE_DEFAULT));
4321
4322 WREG32(0x15D4, 0);
4323 WREG32(0x15D8, 0);
4324 WREG32(0x15DC, 0);
4325
4326 /* empty context1-15 */
4327 /* set vm size, must be a multiple of 4 */
4328 WREG32(VM_CONTEXT1_PAGE_TABLE_START_ADDR, 0);
4329 WREG32(VM_CONTEXT1_PAGE_TABLE_END_ADDR, rdev->vm_manager.max_pfn - 1);
4330 /* Assign the pt base to something valid for now; the pts used for
4331 * the VMs are determined by the application and setup and assigned
4332 * on the fly in the vm part of radeon_gart.c
4333 */
4334 for (i = 1; i < 16; i++) {
4335 if (i < 8)
4336 WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (i << 2),
4337 rdev->vm_manager.saved_table_addr[i]);
4338 else
4339 WREG32(VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((i - 8) << 2),
4340 rdev->vm_manager.saved_table_addr[i]);
4341 }
4342
4343 /* enable context1-15 */
4344 WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
4345 (u32)(rdev->dummy_page.addr >> 12));
4346 WREG32(VM_CONTEXT1_CNTL2, 4);
4347 WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(1) |
4348 PAGE_TABLE_BLOCK_SIZE(radeon_vm_block_size - 9) |
4349 RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
4350 RANGE_PROTECTION_FAULT_ENABLE_DEFAULT |
4351 DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
4352 DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT |
4353 PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT |
4354 PDE0_PROTECTION_FAULT_ENABLE_DEFAULT |
4355 VALID_PROTECTION_FAULT_ENABLE_INTERRUPT |
4356 VALID_PROTECTION_FAULT_ENABLE_DEFAULT |
4357 READ_PROTECTION_FAULT_ENABLE_INTERRUPT |
4358 READ_PROTECTION_FAULT_ENABLE_DEFAULT |
4359 WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT |
4360 WRITE_PROTECTION_FAULT_ENABLE_DEFAULT);
4361
4362 si_pcie_gart_tlb_flush(rdev);
4363 DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
4364 (unsigned)(rdev->mc.gtt_size >> 20),
4365 (unsigned long long)rdev->gart.table_addr);
4366 rdev->gart.ready = true;
4367 return 0;
4368}
4369
4370static void si_pcie_gart_disable(struct radeon_device *rdev)
4371{
4372 unsigned i;
4373
4374 for (i = 1; i < 16; ++i) {
4375 uint32_t reg;
4376 if (i < 8)
4377 reg = VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (i << 2);
4378 else
4379 reg = VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((i - 8) << 2);
4380 rdev->vm_manager.saved_table_addr[i] = RREG32(reg);
4381 }
4382
4383 /* Disable all tables */
4384 WREG32(VM_CONTEXT0_CNTL, 0);
4385 WREG32(VM_CONTEXT1_CNTL, 0);
4386 /* Setup TLB control */
4387 WREG32(MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE_NOT_IN_SYS |
4388 SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU);
4389 /* Setup L2 cache */
4390 WREG32(VM_L2_CNTL, ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE |
4391 ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE |
4392 EFFECTIVE_L2_QUEUE_SIZE(7) |
4393 CONTEXT1_IDENTITY_ACCESS_MODE(1));
4394 WREG32(VM_L2_CNTL2, 0);
4395 WREG32(VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY |
4396 L2_CACHE_BIGK_FRAGMENT_SIZE(0));
4397 radeon_gart_table_vram_unpin(rdev);
4398}
4399
4400static void si_pcie_gart_fini(struct radeon_device *rdev)
4401{
4402 si_pcie_gart_disable(rdev);
4403 radeon_gart_table_vram_free(rdev);
4404 radeon_gart_fini(rdev);
4405}
4406
4407/* vm parser */
4408static bool si_vm_reg_valid(u32 reg)
4409{
4410 /* context regs are fine */
4411 if (reg >= 0x28000)
4412 return true;
4413
4414 /* shader regs are also fine */
4415 if (reg >= 0xB000 && reg < 0xC000)
4416 return true;
4417
4418 /* check config regs */
4419 switch (reg) {
4420 case GRBM_GFX_INDEX:
4421 case CP_STRMOUT_CNTL:
4422 case VGT_VTX_VECT_EJECT_REG:
4423 case VGT_CACHE_INVALIDATION:
4424 case VGT_ESGS_RING_SIZE:
4425 case VGT_GSVS_RING_SIZE:
4426 case VGT_GS_VERTEX_REUSE:
4427 case VGT_PRIMITIVE_TYPE:
4428 case VGT_INDEX_TYPE:
4429 case VGT_NUM_INDICES:
4430 case VGT_NUM_INSTANCES:
4431 case VGT_TF_RING_SIZE:
4432 case VGT_HS_OFFCHIP_PARAM:
4433 case VGT_TF_MEMORY_BASE:
4434 case PA_CL_ENHANCE:
4435 case PA_SU_LINE_STIPPLE_VALUE:
4436 case PA_SC_LINE_STIPPLE_STATE:
4437 case PA_SC_ENHANCE:
4438 case SQC_CACHES:
4439 case SPI_STATIC_THREAD_MGMT_1:
4440 case SPI_STATIC_THREAD_MGMT_2:
4441 case SPI_STATIC_THREAD_MGMT_3:
4442 case SPI_PS_MAX_WAVE_ID:
4443 case SPI_CONFIG_CNTL:
4444 case SPI_CONFIG_CNTL_1:
4445 case TA_CNTL_AUX:
4446 case TA_CS_BC_BASE_ADDR:
4447 return true;
4448 default:
4449 DRM_ERROR("Invalid register 0x%x in CS\n", reg);
4450 return false;
4451 }
4452}
4453
4454static int si_vm_packet3_ce_check(struct radeon_device *rdev,
4455 u32 *ib, struct radeon_cs_packet *pkt)
4456{
4457 switch (pkt->opcode) {
4458 case PACKET3_NOP:
4459 case PACKET3_SET_BASE:
4460 case PACKET3_SET_CE_DE_COUNTERS:
4461 case PACKET3_LOAD_CONST_RAM:
4462 case PACKET3_WRITE_CONST_RAM:
4463 case PACKET3_WRITE_CONST_RAM_OFFSET:
4464 case PACKET3_DUMP_CONST_RAM:
4465 case PACKET3_INCREMENT_CE_COUNTER:
4466 case PACKET3_WAIT_ON_DE_COUNTER:
4467 case PACKET3_CE_WRITE:
4468 break;
4469 default:
4470 DRM_ERROR("Invalid CE packet3: 0x%x\n", pkt->opcode);
4471 return -EINVAL;
4472 }
4473 return 0;
4474}
4475
4476static int si_vm_packet3_cp_dma_check(u32 *ib, u32 idx)
4477{
4478 u32 start_reg, reg, i;
4479 u32 command = ib[idx + 4];
4480 u32 info = ib[idx + 1];
4481 u32 idx_value = ib[idx];
4482 if (command & PACKET3_CP_DMA_CMD_SAS) {
4483 /* src address space is register */
4484 if (((info & 0x60000000) >> 29) == 0) {
4485 start_reg = idx_value << 2;
4486 if (command & PACKET3_CP_DMA_CMD_SAIC) {
4487 reg = start_reg;
4488 if (!si_vm_reg_valid(reg)) {
4489 DRM_ERROR("CP DMA Bad SRC register\n");
4490 return -EINVAL;
4491 }
4492 } else {
4493 for (i = 0; i < (command & 0x1fffff); i++) {
4494 reg = start_reg + (4 * i);
4495 if (!si_vm_reg_valid(reg)) {
4496 DRM_ERROR("CP DMA Bad SRC register\n");
4497 return -EINVAL;
4498 }
4499 }
4500 }
4501 }
4502 }
4503 if (command & PACKET3_CP_DMA_CMD_DAS) {
4504 /* dst address space is register */
4505 if (((info & 0x00300000) >> 20) == 0) {
4506 start_reg = ib[idx + 2];
4507 if (command & PACKET3_CP_DMA_CMD_DAIC) {
4508 reg = start_reg;
4509 if (!si_vm_reg_valid(reg)) {
4510 DRM_ERROR("CP DMA Bad DST register\n");
4511 return -EINVAL;
4512 }
4513 } else {
4514 for (i = 0; i < (command & 0x1fffff); i++) {
4515 reg = start_reg + (4 * i);
4516 if (!si_vm_reg_valid(reg)) {
4517 DRM_ERROR("CP DMA Bad DST register\n");
4518 return -EINVAL;
4519 }
4520 }
4521 }
4522 }
4523 }
4524 return 0;
4525}
4526
4527static int si_vm_packet3_gfx_check(struct radeon_device *rdev,
4528 u32 *ib, struct radeon_cs_packet *pkt)
4529{
4530 int r;
4531 u32 idx = pkt->idx + 1;
4532 u32 idx_value = ib[idx];
4533 u32 start_reg, end_reg, reg, i;
4534
4535 switch (pkt->opcode) {
4536 case PACKET3_NOP:
4537 case PACKET3_SET_BASE:
4538 case PACKET3_CLEAR_STATE:
4539 case PACKET3_INDEX_BUFFER_SIZE:
4540 case PACKET3_DISPATCH_DIRECT:
4541 case PACKET3_DISPATCH_INDIRECT:
4542 case PACKET3_ALLOC_GDS:
4543 case PACKET3_WRITE_GDS_RAM:
4544 case PACKET3_ATOMIC_GDS:
4545 case PACKET3_ATOMIC:
4546 case PACKET3_OCCLUSION_QUERY:
4547 case PACKET3_SET_PREDICATION:
4548 case PACKET3_COND_EXEC:
4549 case PACKET3_PRED_EXEC:
4550 case PACKET3_DRAW_INDIRECT:
4551 case PACKET3_DRAW_INDEX_INDIRECT:
4552 case PACKET3_INDEX_BASE:
4553 case PACKET3_DRAW_INDEX_2:
4554 case PACKET3_CONTEXT_CONTROL:
4555 case PACKET3_INDEX_TYPE:
4556 case PACKET3_DRAW_INDIRECT_MULTI:
4557 case PACKET3_DRAW_INDEX_AUTO:
4558 case PACKET3_DRAW_INDEX_IMMD:
4559 case PACKET3_NUM_INSTANCES:
4560 case PACKET3_DRAW_INDEX_MULTI_AUTO:
4561 case PACKET3_STRMOUT_BUFFER_UPDATE:
4562 case PACKET3_DRAW_INDEX_OFFSET_2:
4563 case PACKET3_DRAW_INDEX_MULTI_ELEMENT:
4564 case PACKET3_DRAW_INDEX_INDIRECT_MULTI:
4565 case PACKET3_MPEG_INDEX:
4566 case PACKET3_WAIT_REG_MEM:
4567 case PACKET3_MEM_WRITE:
4568 case PACKET3_PFP_SYNC_ME:
4569 case PACKET3_SURFACE_SYNC:
4570 case PACKET3_EVENT_WRITE:
4571 case PACKET3_EVENT_WRITE_EOP:
4572 case PACKET3_EVENT_WRITE_EOS:
4573 case PACKET3_SET_CONTEXT_REG:
4574 case PACKET3_SET_CONTEXT_REG_INDIRECT:
4575 case PACKET3_SET_SH_REG:
4576 case PACKET3_SET_SH_REG_OFFSET:
4577 case PACKET3_INCREMENT_DE_COUNTER:
4578 case PACKET3_WAIT_ON_CE_COUNTER:
4579 case PACKET3_WAIT_ON_AVAIL_BUFFER:
4580 case PACKET3_ME_WRITE:
4581 break;
4582 case PACKET3_COPY_DATA:
4583 if ((idx_value & 0xf00) == 0) {
4584 reg = ib[idx + 3] * 4;
4585 if (!si_vm_reg_valid(reg))
4586 return -EINVAL;
4587 }
4588 break;
4589 case PACKET3_WRITE_DATA:
4590 if ((idx_value & 0xf00) == 0) {
4591 start_reg = ib[idx + 1] * 4;
4592 if (idx_value & 0x10000) {
4593 if (!si_vm_reg_valid(start_reg))
4594 return -EINVAL;
4595 } else {
4596 for (i = 0; i < (pkt->count - 2); i++) {
4597 reg = start_reg + (4 * i);
4598 if (!si_vm_reg_valid(reg))
4599 return -EINVAL;
4600 }
4601 }
4602 }
4603 break;
4604 case PACKET3_COND_WRITE:
4605 if (idx_value & 0x100) {
4606 reg = ib[idx + 5] * 4;
4607 if (!si_vm_reg_valid(reg))
4608 return -EINVAL;
4609 }
4610 break;
4611 case PACKET3_COPY_DW:
4612 if (idx_value & 0x2) {
4613 reg = ib[idx + 3] * 4;
4614 if (!si_vm_reg_valid(reg))
4615 return -EINVAL;
4616 }
4617 break;
4618 case PACKET3_SET_CONFIG_REG:
4619 start_reg = (idx_value << 2) + PACKET3_SET_CONFIG_REG_START;
4620 end_reg = 4 * pkt->count + start_reg - 4;
4621 if ((start_reg < PACKET3_SET_CONFIG_REG_START) ||
4622 (start_reg >= PACKET3_SET_CONFIG_REG_END) ||
4623 (end_reg >= PACKET3_SET_CONFIG_REG_END)) {
4624 DRM_ERROR("bad PACKET3_SET_CONFIG_REG\n");
4625 return -EINVAL;
4626 }
4627 for (i = 0; i < pkt->count; i++) {
4628 reg = start_reg + (4 * i);
4629 if (!si_vm_reg_valid(reg))
4630 return -EINVAL;
4631 }
4632 break;
4633 case PACKET3_CP_DMA:
4634 r = si_vm_packet3_cp_dma_check(ib, idx);
4635 if (r)
4636 return r;
4637 break;
4638 default:
4639 DRM_ERROR("Invalid GFX packet3: 0x%x\n", pkt->opcode);
4640 return -EINVAL;
4641 }
4642 return 0;
4643}
4644
4645static int si_vm_packet3_compute_check(struct radeon_device *rdev,
4646 u32 *ib, struct radeon_cs_packet *pkt)
4647{
4648 int r;
4649 u32 idx = pkt->idx + 1;
4650 u32 idx_value = ib[idx];
4651 u32 start_reg, reg, i;
4652
4653 switch (pkt->opcode) {
4654 case PACKET3_NOP:
4655 case PACKET3_SET_BASE:
4656 case PACKET3_CLEAR_STATE:
4657 case PACKET3_DISPATCH_DIRECT:
4658 case PACKET3_DISPATCH_INDIRECT:
4659 case PACKET3_ALLOC_GDS:
4660 case PACKET3_WRITE_GDS_RAM:
4661 case PACKET3_ATOMIC_GDS:
4662 case PACKET3_ATOMIC:
4663 case PACKET3_OCCLUSION_QUERY:
4664 case PACKET3_SET_PREDICATION:
4665 case PACKET3_COND_EXEC:
4666 case PACKET3_PRED_EXEC:
4667 case PACKET3_CONTEXT_CONTROL:
4668 case PACKET3_STRMOUT_BUFFER_UPDATE:
4669 case PACKET3_WAIT_REG_MEM:
4670 case PACKET3_MEM_WRITE:
4671 case PACKET3_PFP_SYNC_ME:
4672 case PACKET3_SURFACE_SYNC:
4673 case PACKET3_EVENT_WRITE:
4674 case PACKET3_EVENT_WRITE_EOP:
4675 case PACKET3_EVENT_WRITE_EOS:
4676 case PACKET3_SET_CONTEXT_REG:
4677 case PACKET3_SET_CONTEXT_REG_INDIRECT:
4678 case PACKET3_SET_SH_REG:
4679 case PACKET3_SET_SH_REG_OFFSET:
4680 case PACKET3_INCREMENT_DE_COUNTER:
4681 case PACKET3_WAIT_ON_CE_COUNTER:
4682 case PACKET3_WAIT_ON_AVAIL_BUFFER:
4683 case PACKET3_ME_WRITE:
4684 break;
4685 case PACKET3_COPY_DATA:
4686 if ((idx_value & 0xf00) == 0) {
4687 reg = ib[idx + 3] * 4;
4688 if (!si_vm_reg_valid(reg))
4689 return -EINVAL;
4690 }
4691 break;
4692 case PACKET3_WRITE_DATA:
4693 if ((idx_value & 0xf00) == 0) {
4694 start_reg = ib[idx + 1] * 4;
4695 if (idx_value & 0x10000) {
4696 if (!si_vm_reg_valid(start_reg))
4697 return -EINVAL;
4698 } else {
4699 for (i = 0; i < (pkt->count - 2); i++) {
4700 reg = start_reg + (4 * i);
4701 if (!si_vm_reg_valid(reg))
4702 return -EINVAL;
4703 }
4704 }
4705 }
4706 break;
4707 case PACKET3_COND_WRITE:
4708 if (idx_value & 0x100) {
4709 reg = ib[idx + 5] * 4;
4710 if (!si_vm_reg_valid(reg))
4711 return -EINVAL;
4712 }
4713 break;
4714 case PACKET3_COPY_DW:
4715 if (idx_value & 0x2) {
4716 reg = ib[idx + 3] * 4;
4717 if (!si_vm_reg_valid(reg))
4718 return -EINVAL;
4719 }
4720 break;
4721 case PACKET3_CP_DMA:
4722 r = si_vm_packet3_cp_dma_check(ib, idx);
4723 if (r)
4724 return r;
4725 break;
4726 default:
4727 DRM_ERROR("Invalid Compute packet3: 0x%x\n", pkt->opcode);
4728 return -EINVAL;
4729 }
4730 return 0;
4731}
4732
4733int si_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib)
4734{
4735 int ret = 0;
4736 u32 idx = 0, i;
4737 struct radeon_cs_packet pkt;
4738
4739 do {
4740 pkt.idx = idx;
4741 pkt.type = RADEON_CP_PACKET_GET_TYPE(ib->ptr[idx]);
4742 pkt.count = RADEON_CP_PACKET_GET_COUNT(ib->ptr[idx]);
4743 pkt.one_reg_wr = 0;
4744 switch (pkt.type) {
4745 case RADEON_PACKET_TYPE0:
4746 dev_err(rdev->dev, "Packet0 not allowed!\n");
4747 ret = -EINVAL;
4748 break;
4749 case RADEON_PACKET_TYPE2:
4750 idx += 1;
4751 break;
4752 case RADEON_PACKET_TYPE3:
4753 pkt.opcode = RADEON_CP_PACKET3_GET_OPCODE(ib->ptr[idx]);
4754 if (ib->is_const_ib)
4755 ret = si_vm_packet3_ce_check(rdev, ib->ptr, &pkt);
4756 else {
4757 switch (ib->ring) {
4758 case RADEON_RING_TYPE_GFX_INDEX:
4759 ret = si_vm_packet3_gfx_check(rdev, ib->ptr, &pkt);
4760 break;
4761 case CAYMAN_RING_TYPE_CP1_INDEX:
4762 case CAYMAN_RING_TYPE_CP2_INDEX:
4763 ret = si_vm_packet3_compute_check(rdev, ib->ptr, &pkt);
4764 break;
4765 default:
4766 dev_err(rdev->dev, "Non-PM4 ring %d !\n", ib->ring);
4767 ret = -EINVAL;
4768 break;
4769 }
4770 }
4771 idx += pkt.count + 2;
4772 break;
4773 default:
4774 dev_err(rdev->dev, "Unknown packet type %d !\n", pkt.type);
4775 ret = -EINVAL;
4776 break;
4777 }
4778 if (ret) {
4779 for (i = 0; i < ib->length_dw; i++) {
4780 if (i == idx)
4781 printk("\t0x%08x <---\n", ib->ptr[i]);
4782 else
4783 printk("\t0x%08x\n", ib->ptr[i]);
4784 }
4785 break;
4786 }
4787 } while (idx < ib->length_dw);
4788
4789 return ret;
4790}
4791
4792/*
4793 * vm
4794 */
4795int si_vm_init(struct radeon_device *rdev)
4796{
4797 /* number of VMs */
4798 rdev->vm_manager.nvm = 16;
4799 /* base offset of vram pages */
4800 rdev->vm_manager.vram_base_offset = 0;
4801
4802 return 0;
4803}
4804
4805void si_vm_fini(struct radeon_device *rdev)
4806{
4807}
4808
4809/**
4810 * si_vm_decode_fault - print human readable fault info
4811 *
4812 * @rdev: radeon_device pointer
4813 * @status: VM_CONTEXT1_PROTECTION_FAULT_STATUS register value
4814 * @addr: VM_CONTEXT1_PROTECTION_FAULT_ADDR register value
4815 *
4816 * Print human readable fault information (SI).
4817 */
4818static void si_vm_decode_fault(struct radeon_device *rdev,
4819 u32 status, u32 addr)
4820{
4821 u32 mc_id = (status & MEMORY_CLIENT_ID_MASK) >> MEMORY_CLIENT_ID_SHIFT;
4822 u32 vmid = (status & FAULT_VMID_MASK) >> FAULT_VMID_SHIFT;
4823 u32 protections = (status & PROTECTIONS_MASK) >> PROTECTIONS_SHIFT;
4824 char *block;
4825
4826 if (rdev->family == CHIP_TAHITI) {
4827 switch (mc_id) {
4828 case 160:
4829 case 144:
4830 case 96:
4831 case 80:
4832 case 224:
4833 case 208:
4834 case 32:
4835 case 16:
4836 block = "CB";
4837 break;
4838 case 161:
4839 case 145:
4840 case 97:
4841 case 81:
4842 case 225:
4843 case 209:
4844 case 33:
4845 case 17:
4846 block = "CB_FMASK";
4847 break;
4848 case 162:
4849 case 146:
4850 case 98:
4851 case 82:
4852 case 226:
4853 case 210:
4854 case 34:
4855 case 18:
4856 block = "CB_CMASK";
4857 break;
4858 case 163:
4859 case 147:
4860 case 99:
4861 case 83:
4862 case 227:
4863 case 211:
4864 case 35:
4865 case 19:
4866 block = "CB_IMMED";
4867 break;
4868 case 164:
4869 case 148:
4870 case 100:
4871 case 84:
4872 case 228:
4873 case 212:
4874 case 36:
4875 case 20:
4876 block = "DB";
4877 break;
4878 case 165:
4879 case 149:
4880 case 101:
4881 case 85:
4882 case 229:
4883 case 213:
4884 case 37:
4885 case 21:
4886 block = "DB_HTILE";
4887 break;
4888 case 167:
4889 case 151:
4890 case 103:
4891 case 87:
4892 case 231:
4893 case 215:
4894 case 39:
4895 case 23:
4896 block = "DB_STEN";
4897 break;
4898 case 72:
4899 case 68:
4900 case 64:
4901 case 8:
4902 case 4:
4903 case 0:
4904 case 136:
4905 case 132:
4906 case 128:
4907 case 200:
4908 case 196:
4909 case 192:
4910 block = "TC";
4911 break;
4912 case 112:
4913 case 48:
4914 block = "CP";
4915 break;
4916 case 49:
4917 case 177:
4918 case 50:
4919 case 178:
4920 block = "SH";
4921 break;
4922 case 53:
4923 case 190:
4924 block = "VGT";
4925 break;
4926 case 117:
4927 block = "IH";
4928 break;
4929 case 51:
4930 case 115:
4931 block = "RLC";
4932 break;
4933 case 119:
4934 case 183:
4935 block = "DMA0";
4936 break;
4937 case 61:
4938 block = "DMA1";
4939 break;
4940 case 248:
4941 case 120:
4942 block = "HDP";
4943 break;
4944 default:
4945 block = "unknown";
4946 break;
4947 }
4948 } else {
4949 switch (mc_id) {
4950 case 32:
4951 case 16:
4952 case 96:
4953 case 80:
4954 case 160:
4955 case 144:
4956 case 224:
4957 case 208:
4958 block = "CB";
4959 break;
4960 case 33:
4961 case 17:
4962 case 97:
4963 case 81:
4964 case 161:
4965 case 145:
4966 case 225:
4967 case 209:
4968 block = "CB_FMASK";
4969 break;
4970 case 34:
4971 case 18:
4972 case 98:
4973 case 82:
4974 case 162:
4975 case 146:
4976 case 226:
4977 case 210:
4978 block = "CB_CMASK";
4979 break;
4980 case 35:
4981 case 19:
4982 case 99:
4983 case 83:
4984 case 163:
4985 case 147:
4986 case 227:
4987 case 211:
4988 block = "CB_IMMED";
4989 break;
4990 case 36:
4991 case 20:
4992 case 100:
4993 case 84:
4994 case 164:
4995 case 148:
4996 case 228:
4997 case 212:
4998 block = "DB";
4999 break;
5000 case 37:
5001 case 21:
5002 case 101:
5003 case 85:
5004 case 165:
5005 case 149:
5006 case 229:
5007 case 213:
5008 block = "DB_HTILE";
5009 break;
5010 case 39:
5011 case 23:
5012 case 103:
5013 case 87:
5014 case 167:
5015 case 151:
5016 case 231:
5017 case 215:
5018 block = "DB_STEN";
5019 break;
5020 case 72:
5021 case 68:
5022 case 8:
5023 case 4:
5024 case 136:
5025 case 132:
5026 case 200:
5027 case 196:
5028 block = "TC";
5029 break;
5030 case 112:
5031 case 48:
5032 block = "CP";
5033 break;
5034 case 49:
5035 case 177:
5036 case 50:
5037 case 178:
5038 block = "SH";
5039 break;
5040 case 53:
5041 block = "VGT";
5042 break;
5043 case 117:
5044 block = "IH";
5045 break;
5046 case 51:
5047 case 115:
5048 block = "RLC";
5049 break;
5050 case 119:
5051 case 183:
5052 block = "DMA0";
5053 break;
5054 case 61:
5055 block = "DMA1";
5056 break;
5057 case 248:
5058 case 120:
5059 block = "HDP";
5060 break;
5061 default:
5062 block = "unknown";
5063 break;
5064 }
5065 }
5066
5067 printk("VM fault (0x%02x, vmid %d) at page %u, %s from %s (%d)\n",
5068 protections, vmid, addr,
5069 (status & MEMORY_CLIENT_RW_MASK) ? "write" : "read",
5070 block, mc_id);
5071}
5072
5073void si_vm_flush(struct radeon_device *rdev, struct radeon_ring *ring,
5074 unsigned vm_id, uint64_t pd_addr)
5075{
5076 /* write new base address */
5077 radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
5078 radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(1) |
5079 WRITE_DATA_DST_SEL(0)));
5080
5081 if (vm_id < 8) {
5082 radeon_ring_write(ring,
5083 (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm_id << 2)) >> 2);
5084 } else {
5085 radeon_ring_write(ring,
5086 (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm_id - 8) << 2)) >> 2);
5087 }
5088 radeon_ring_write(ring, 0);
5089 radeon_ring_write(ring, pd_addr >> 12);
5090
5091 /* flush hdp cache */
5092 radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
5093 radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(1) |
5094 WRITE_DATA_DST_SEL(0)));
5095 radeon_ring_write(ring, HDP_MEM_COHERENCY_FLUSH_CNTL >> 2);
5096 radeon_ring_write(ring, 0);
5097 radeon_ring_write(ring, 0x1);
5098
5099 /* bits 0-15 are the VM contexts0-15 */
5100 radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
5101 radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(1) |
5102 WRITE_DATA_DST_SEL(0)));
5103 radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2);
5104 radeon_ring_write(ring, 0);
5105 radeon_ring_write(ring, 1 << vm_id);
5106
5107 /* wait for the invalidate to complete */
5108 radeon_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
5109 radeon_ring_write(ring, (WAIT_REG_MEM_FUNCTION(0) | /* always */
5110 WAIT_REG_MEM_ENGINE(0))); /* me */
5111 radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2);
5112 radeon_ring_write(ring, 0);
5113 radeon_ring_write(ring, 0); /* ref */
5114 radeon_ring_write(ring, 0); /* mask */
5115 radeon_ring_write(ring, 0x20); /* poll interval */
5116
5117 /* sync PFP to ME, otherwise we might get invalid PFP reads */
5118 radeon_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
5119 radeon_ring_write(ring, 0x0);
5120}
5121
5122/*
5123 * Power and clock gating
5124 */
5125static void si_wait_for_rlc_serdes(struct radeon_device *rdev)
5126{
5127 int i;
5128
5129 for (i = 0; i < rdev->usec_timeout; i++) {
5130 if (RREG32(RLC_SERDES_MASTER_BUSY_0) == 0)
5131 break;
5132 udelay(1);
5133 }
5134
5135 for (i = 0; i < rdev->usec_timeout; i++) {
5136 if (RREG32(RLC_SERDES_MASTER_BUSY_1) == 0)
5137 break;
5138 udelay(1);
5139 }
5140}
5141
5142static void si_enable_gui_idle_interrupt(struct radeon_device *rdev,
5143 bool enable)
5144{
5145 u32 tmp = RREG32(CP_INT_CNTL_RING0);
5146 u32 mask;
5147 int i;
5148
5149 if (enable)
5150 tmp |= (CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE);
5151 else
5152 tmp &= ~(CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE);
5153 WREG32(CP_INT_CNTL_RING0, tmp);
5154
5155 if (!enable) {
5156 /* read a gfx register */
5157 tmp = RREG32(DB_DEPTH_INFO);
5158
5159 mask = RLC_BUSY_STATUS | GFX_POWER_STATUS | GFX_CLOCK_STATUS | GFX_LS_STATUS;
5160 for (i = 0; i < rdev->usec_timeout; i++) {
5161 if ((RREG32(RLC_STAT) & mask) == (GFX_CLOCK_STATUS | GFX_POWER_STATUS))
5162 break;
5163 udelay(1);
5164 }
5165 }
5166}
5167
5168static void si_set_uvd_dcm(struct radeon_device *rdev,
5169 bool sw_mode)
5170{
5171 u32 tmp, tmp2;
5172
5173 tmp = RREG32(UVD_CGC_CTRL);
5174 tmp &= ~(CLK_OD_MASK | CG_DT_MASK);
5175 tmp |= DCM | CG_DT(1) | CLK_OD(4);
5176
5177 if (sw_mode) {
5178 tmp &= ~0x7ffff800;
5179 tmp2 = DYN_OR_EN | DYN_RR_EN | G_DIV_ID(7);
5180 } else {
5181 tmp |= 0x7ffff800;
5182 tmp2 = 0;
5183 }
5184
5185 WREG32(UVD_CGC_CTRL, tmp);
5186 WREG32_UVD_CTX(UVD_CGC_CTRL2, tmp2);
5187}
5188
5189void si_init_uvd_internal_cg(struct radeon_device *rdev)
5190{
5191 bool hw_mode = true;
5192
5193 if (hw_mode) {
5194 si_set_uvd_dcm(rdev, false);
5195 } else {
5196 u32 tmp = RREG32(UVD_CGC_CTRL);
5197 tmp &= ~DCM;
5198 WREG32(UVD_CGC_CTRL, tmp);
5199 }
5200}
5201
5202static u32 si_halt_rlc(struct radeon_device *rdev)
5203{
5204 u32 data, orig;
5205
5206 orig = data = RREG32(RLC_CNTL);
5207
5208 if (data & RLC_ENABLE) {
5209 data &= ~RLC_ENABLE;
5210 WREG32(RLC_CNTL, data);
5211
5212 si_wait_for_rlc_serdes(rdev);
5213 }
5214
5215 return orig;
5216}
5217
5218static void si_update_rlc(struct radeon_device *rdev, u32 rlc)
5219{
5220 u32 tmp;
5221
5222 tmp = RREG32(RLC_CNTL);
5223 if (tmp != rlc)
5224 WREG32(RLC_CNTL, rlc);
5225}
5226
5227static void si_enable_dma_pg(struct radeon_device *rdev, bool enable)
5228{
5229 u32 data, orig;
5230
5231 orig = data = RREG32(DMA_PG);
5232 if (enable && (rdev->pg_flags & RADEON_PG_SUPPORT_SDMA))
5233 data |= PG_CNTL_ENABLE;
5234 else
5235 data &= ~PG_CNTL_ENABLE;
5236 if (orig != data)
5237 WREG32(DMA_PG, data);
5238}
5239
5240static void si_init_dma_pg(struct radeon_device *rdev)
5241{
5242 u32 tmp;
5243
5244 WREG32(DMA_PGFSM_WRITE, 0x00002000);
5245 WREG32(DMA_PGFSM_CONFIG, 0x100010ff);
5246
5247 for (tmp = 0; tmp < 5; tmp++)
5248 WREG32(DMA_PGFSM_WRITE, 0);
5249}
5250
5251static void si_enable_gfx_cgpg(struct radeon_device *rdev,
5252 bool enable)
5253{
5254 u32 tmp;
5255
5256 if (enable && (rdev->pg_flags & RADEON_PG_SUPPORT_GFX_PG)) {
5257 tmp = RLC_PUD(0x10) | RLC_PDD(0x10) | RLC_TTPD(0x10) | RLC_MSD(0x10);
5258 WREG32(RLC_TTOP_D, tmp);
5259
5260 tmp = RREG32(RLC_PG_CNTL);
5261 tmp |= GFX_PG_ENABLE;
5262 WREG32(RLC_PG_CNTL, tmp);
5263
5264 tmp = RREG32(RLC_AUTO_PG_CTRL);
5265 tmp |= AUTO_PG_EN;
5266 WREG32(RLC_AUTO_PG_CTRL, tmp);
5267 } else {
5268 tmp = RREG32(RLC_AUTO_PG_CTRL);
5269 tmp &= ~AUTO_PG_EN;
5270 WREG32(RLC_AUTO_PG_CTRL, tmp);
5271
5272 tmp = RREG32(DB_RENDER_CONTROL);
5273 }
5274}
5275
5276static void si_init_gfx_cgpg(struct radeon_device *rdev)
5277{
5278 u32 tmp;
5279
5280 WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8);
5281
5282 tmp = RREG32(RLC_PG_CNTL);
5283 tmp |= GFX_PG_SRC;
5284 WREG32(RLC_PG_CNTL, tmp);
5285
5286 WREG32(RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8);
5287
5288 tmp = RREG32(RLC_AUTO_PG_CTRL);
5289
5290 tmp &= ~GRBM_REG_SGIT_MASK;
5291 tmp |= GRBM_REG_SGIT(0x700);
5292 tmp &= ~PG_AFTER_GRBM_REG_ST_MASK;
5293 WREG32(RLC_AUTO_PG_CTRL, tmp);
5294}
5295
5296static u32 si_get_cu_active_bitmap(struct radeon_device *rdev, u32 se, u32 sh)
5297{
5298 u32 mask = 0, tmp, tmp1;
5299 int i;
5300
5301 si_select_se_sh(rdev, se, sh);
5302 tmp = RREG32(CC_GC_SHADER_ARRAY_CONFIG);
5303 tmp1 = RREG32(GC_USER_SHADER_ARRAY_CONFIG);
5304 si_select_se_sh(rdev, 0xffffffff, 0xffffffff);
5305
5306 tmp &= 0xffff0000;
5307
5308 tmp |= tmp1;
5309 tmp >>= 16;
5310
5311 for (i = 0; i < rdev->config.si.max_cu_per_sh; i ++) {
5312 mask <<= 1;
5313 mask |= 1;
5314 }
5315
5316 return (~tmp) & mask;
5317}
5318
5319static void si_init_ao_cu_mask(struct radeon_device *rdev)
5320{
5321 u32 i, j, k, active_cu_number = 0;
5322 u32 mask, counter, cu_bitmap;
5323 u32 tmp = 0;
5324
5325 for (i = 0; i < rdev->config.si.max_shader_engines; i++) {
5326 for (j = 0; j < rdev->config.si.max_sh_per_se; j++) {
5327 mask = 1;
5328 cu_bitmap = 0;
5329 counter = 0;
5330 for (k = 0; k < rdev->config.si.max_cu_per_sh; k++) {
5331 if (si_get_cu_active_bitmap(rdev, i, j) & mask) {
5332 if (counter < 2)
5333 cu_bitmap |= mask;
5334 counter++;
5335 }
5336 mask <<= 1;
5337 }
5338
5339 active_cu_number += counter;
5340 tmp |= (cu_bitmap << (i * 16 + j * 8));
5341 }
5342 }
5343
5344 WREG32(RLC_PG_AO_CU_MASK, tmp);
5345
5346 tmp = RREG32(RLC_MAX_PG_CU);
5347 tmp &= ~MAX_PU_CU_MASK;
5348 tmp |= MAX_PU_CU(active_cu_number);
5349 WREG32(RLC_MAX_PG_CU, tmp);
5350}
5351
5352static void si_enable_cgcg(struct radeon_device *rdev,
5353 bool enable)
5354{
5355 u32 data, orig, tmp;
5356
5357 orig = data = RREG32(RLC_CGCG_CGLS_CTRL);
5358
5359 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_GFX_CGCG)) {
5360 si_enable_gui_idle_interrupt(rdev, true);
5361
5362 WREG32(RLC_GCPM_GENERAL_3, 0x00000080);
5363
5364 tmp = si_halt_rlc(rdev);
5365
5366 WREG32(RLC_SERDES_WR_MASTER_MASK_0, 0xffffffff);
5367 WREG32(RLC_SERDES_WR_MASTER_MASK_1, 0xffffffff);
5368 WREG32(RLC_SERDES_WR_CTRL, 0x00b000ff);
5369
5370 si_wait_for_rlc_serdes(rdev);
5371
5372 si_update_rlc(rdev, tmp);
5373
5374 WREG32(RLC_SERDES_WR_CTRL, 0x007000ff);
5375
5376 data |= CGCG_EN | CGLS_EN;
5377 } else {
5378 si_enable_gui_idle_interrupt(rdev, false);
5379
5380 RREG32(CB_CGTT_SCLK_CTRL);
5381 RREG32(CB_CGTT_SCLK_CTRL);
5382 RREG32(CB_CGTT_SCLK_CTRL);
5383 RREG32(CB_CGTT_SCLK_CTRL);
5384
5385 data &= ~(CGCG_EN | CGLS_EN);
5386 }
5387
5388 if (orig != data)
5389 WREG32(RLC_CGCG_CGLS_CTRL, data);
5390}
5391
5392static void si_enable_mgcg(struct radeon_device *rdev,
5393 bool enable)
5394{
5395 u32 data, orig, tmp = 0;
5396
5397 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_GFX_MGCG)) {
5398 orig = data = RREG32(CGTS_SM_CTRL_REG);
5399 data = 0x96940200;
5400 if (orig != data)
5401 WREG32(CGTS_SM_CTRL_REG, data);
5402
5403 if (rdev->cg_flags & RADEON_CG_SUPPORT_GFX_CP_LS) {
5404 orig = data = RREG32(CP_MEM_SLP_CNTL);
5405 data |= CP_MEM_LS_EN;
5406 if (orig != data)
5407 WREG32(CP_MEM_SLP_CNTL, data);
5408 }
5409
5410 orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE);
5411 data &= 0xffffffc0;
5412 if (orig != data)
5413 WREG32(RLC_CGTT_MGCG_OVERRIDE, data);
5414
5415 tmp = si_halt_rlc(rdev);
5416
5417 WREG32(RLC_SERDES_WR_MASTER_MASK_0, 0xffffffff);
5418 WREG32(RLC_SERDES_WR_MASTER_MASK_1, 0xffffffff);
5419 WREG32(RLC_SERDES_WR_CTRL, 0x00d000ff);
5420
5421 si_update_rlc(rdev, tmp);
5422 } else {
5423 orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE);
5424 data |= 0x00000003;
5425 if (orig != data)
5426 WREG32(RLC_CGTT_MGCG_OVERRIDE, data);
5427
5428 data = RREG32(CP_MEM_SLP_CNTL);
5429 if (data & CP_MEM_LS_EN) {
5430 data &= ~CP_MEM_LS_EN;
5431 WREG32(CP_MEM_SLP_CNTL, data);
5432 }
5433 orig = data = RREG32(CGTS_SM_CTRL_REG);
5434 data |= LS_OVERRIDE | OVERRIDE;
5435 if (orig != data)
5436 WREG32(CGTS_SM_CTRL_REG, data);
5437
5438 tmp = si_halt_rlc(rdev);
5439
5440 WREG32(RLC_SERDES_WR_MASTER_MASK_0, 0xffffffff);
5441 WREG32(RLC_SERDES_WR_MASTER_MASK_1, 0xffffffff);
5442 WREG32(RLC_SERDES_WR_CTRL, 0x00e000ff);
5443
5444 si_update_rlc(rdev, tmp);
5445 }
5446}
5447
5448static void si_enable_uvd_mgcg(struct radeon_device *rdev,
5449 bool enable)
5450{
5451 u32 orig, data, tmp;
5452
5453 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_UVD_MGCG)) {
5454 tmp = RREG32_UVD_CTX(UVD_CGC_MEM_CTRL);
5455 tmp |= 0x3fff;
5456 WREG32_UVD_CTX(UVD_CGC_MEM_CTRL, tmp);
5457
5458 orig = data = RREG32(UVD_CGC_CTRL);
5459 data |= DCM;
5460 if (orig != data)
5461 WREG32(UVD_CGC_CTRL, data);
5462
5463 WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_0, 0);
5464 WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_1, 0);
5465 } else {
5466 tmp = RREG32_UVD_CTX(UVD_CGC_MEM_CTRL);
5467 tmp &= ~0x3fff;
5468 WREG32_UVD_CTX(UVD_CGC_MEM_CTRL, tmp);
5469
5470 orig = data = RREG32(UVD_CGC_CTRL);
5471 data &= ~DCM;
5472 if (orig != data)
5473 WREG32(UVD_CGC_CTRL, data);
5474
5475 WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_0, 0xffffffff);
5476 WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_1, 0xffffffff);
5477 }
5478}
5479
5480static const u32 mc_cg_registers[] =
5481{
5482 MC_HUB_MISC_HUB_CG,
5483 MC_HUB_MISC_SIP_CG,
5484 MC_HUB_MISC_VM_CG,
5485 MC_XPB_CLK_GAT,
5486 ATC_MISC_CG,
5487 MC_CITF_MISC_WR_CG,
5488 MC_CITF_MISC_RD_CG,
5489 MC_CITF_MISC_VM_CG,
5490 VM_L2_CG,
5491};
5492
5493static void si_enable_mc_ls(struct radeon_device *rdev,
5494 bool enable)
5495{
5496 int i;
5497 u32 orig, data;
5498
5499 for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
5500 orig = data = RREG32(mc_cg_registers[i]);
5501 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_MC_LS))
5502 data |= MC_LS_ENABLE;
5503 else
5504 data &= ~MC_LS_ENABLE;
5505 if (data != orig)
5506 WREG32(mc_cg_registers[i], data);
5507 }
5508}
5509
5510static void si_enable_mc_mgcg(struct radeon_device *rdev,
5511 bool enable)
5512{
5513 int i;
5514 u32 orig, data;
5515
5516 for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
5517 orig = data = RREG32(mc_cg_registers[i]);
5518 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_MC_MGCG))
5519 data |= MC_CG_ENABLE;
5520 else
5521 data &= ~MC_CG_ENABLE;
5522 if (data != orig)
5523 WREG32(mc_cg_registers[i], data);
5524 }
5525}
5526
5527static void si_enable_dma_mgcg(struct radeon_device *rdev,
5528 bool enable)
5529{
5530 u32 orig, data, offset;
5531 int i;
5532
5533 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_SDMA_MGCG)) {
5534 for (i = 0; i < 2; i++) {
5535 if (i == 0)
5536 offset = DMA0_REGISTER_OFFSET;
5537 else
5538 offset = DMA1_REGISTER_OFFSET;
5539 orig = data = RREG32(DMA_POWER_CNTL + offset);
5540 data &= ~MEM_POWER_OVERRIDE;
5541 if (data != orig)
5542 WREG32(DMA_POWER_CNTL + offset, data);
5543 WREG32(DMA_CLK_CTRL + offset, 0x00000100);
5544 }
5545 } else {
5546 for (i = 0; i < 2; i++) {
5547 if (i == 0)
5548 offset = DMA0_REGISTER_OFFSET;
5549 else
5550 offset = DMA1_REGISTER_OFFSET;
5551 orig = data = RREG32(DMA_POWER_CNTL + offset);
5552 data |= MEM_POWER_OVERRIDE;
5553 if (data != orig)
5554 WREG32(DMA_POWER_CNTL + offset, data);
5555
5556 orig = data = RREG32(DMA_CLK_CTRL + offset);
5557 data = 0xff000000;
5558 if (data != orig)
5559 WREG32(DMA_CLK_CTRL + offset, data);
5560 }
5561 }
5562}
5563
5564static void si_enable_bif_mgls(struct radeon_device *rdev,
5565 bool enable)
5566{
5567 u32 orig, data;
5568
5569 orig = data = RREG32_PCIE(PCIE_CNTL2);
5570
5571 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_BIF_LS))
5572 data |= SLV_MEM_LS_EN | MST_MEM_LS_EN |
5573 REPLAY_MEM_LS_EN | SLV_MEM_AGGRESSIVE_LS_EN;
5574 else
5575 data &= ~(SLV_MEM_LS_EN | MST_MEM_LS_EN |
5576 REPLAY_MEM_LS_EN | SLV_MEM_AGGRESSIVE_LS_EN);
5577
5578 if (orig != data)
5579 WREG32_PCIE(PCIE_CNTL2, data);
5580}
5581
5582static void si_enable_hdp_mgcg(struct radeon_device *rdev,
5583 bool enable)
5584{
5585 u32 orig, data;
5586
5587 orig = data = RREG32(HDP_HOST_PATH_CNTL);
5588
5589 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_HDP_MGCG))
5590 data &= ~CLOCK_GATING_DIS;
5591 else
5592 data |= CLOCK_GATING_DIS;
5593
5594 if (orig != data)
5595 WREG32(HDP_HOST_PATH_CNTL, data);
5596}
5597
5598static void si_enable_hdp_ls(struct radeon_device *rdev,
5599 bool enable)
5600{
5601 u32 orig, data;
5602
5603 orig = data = RREG32(HDP_MEM_POWER_LS);
5604
5605 if (enable && (rdev->cg_flags & RADEON_CG_SUPPORT_HDP_LS))
5606 data |= HDP_LS_ENABLE;
5607 else
5608 data &= ~HDP_LS_ENABLE;
5609
5610 if (orig != data)
5611 WREG32(HDP_MEM_POWER_LS, data);
5612}
5613
5614static void si_update_cg(struct radeon_device *rdev,
5615 u32 block, bool enable)
5616{
5617 if (block & RADEON_CG_BLOCK_GFX) {
5618 si_enable_gui_idle_interrupt(rdev, false);
5619 /* order matters! */
5620 if (enable) {
5621 si_enable_mgcg(rdev, true);
5622 si_enable_cgcg(rdev, true);
5623 } else {
5624 si_enable_cgcg(rdev, false);
5625 si_enable_mgcg(rdev, false);
5626 }
5627 si_enable_gui_idle_interrupt(rdev, true);
5628 }
5629
5630 if (block & RADEON_CG_BLOCK_MC) {
5631 si_enable_mc_mgcg(rdev, enable);
5632 si_enable_mc_ls(rdev, enable);
5633 }
5634
5635 if (block & RADEON_CG_BLOCK_SDMA) {
5636 si_enable_dma_mgcg(rdev, enable);
5637 }
5638
5639 if (block & RADEON_CG_BLOCK_BIF) {
5640 si_enable_bif_mgls(rdev, enable);
5641 }
5642
5643 if (block & RADEON_CG_BLOCK_UVD) {
5644 if (rdev->has_uvd) {
5645 si_enable_uvd_mgcg(rdev, enable);
5646 }
5647 }
5648
5649 if (block & RADEON_CG_BLOCK_HDP) {
5650 si_enable_hdp_mgcg(rdev, enable);
5651 si_enable_hdp_ls(rdev, enable);
5652 }
5653}
5654
5655static void si_init_cg(struct radeon_device *rdev)
5656{
5657 si_update_cg(rdev, (RADEON_CG_BLOCK_GFX |
5658 RADEON_CG_BLOCK_MC |
5659 RADEON_CG_BLOCK_SDMA |
5660 RADEON_CG_BLOCK_BIF |
5661 RADEON_CG_BLOCK_HDP), true);
5662 if (rdev->has_uvd) {
5663 si_update_cg(rdev, RADEON_CG_BLOCK_UVD, true);
5664 si_init_uvd_internal_cg(rdev);
5665 }
5666}
5667
5668static void si_fini_cg(struct radeon_device *rdev)
5669{
5670 if (rdev->has_uvd) {
5671 si_update_cg(rdev, RADEON_CG_BLOCK_UVD, false);
5672 }
5673 si_update_cg(rdev, (RADEON_CG_BLOCK_GFX |
5674 RADEON_CG_BLOCK_MC |
5675 RADEON_CG_BLOCK_SDMA |
5676 RADEON_CG_BLOCK_BIF |
5677 RADEON_CG_BLOCK_HDP), false);
5678}
5679
5680u32 si_get_csb_size(struct radeon_device *rdev)
5681{
5682 u32 count = 0;
5683 const struct cs_section_def *sect = NULL;
5684 const struct cs_extent_def *ext = NULL;
5685
5686 if (rdev->rlc.cs_data == NULL)
5687 return 0;
5688
5689 /* begin clear state */
5690 count += 2;
5691 /* context control state */
5692 count += 3;
5693
5694 for (sect = rdev->rlc.cs_data; sect->section != NULL; ++sect) {
5695 for (ext = sect->section; ext->extent != NULL; ++ext) {
5696 if (sect->id == SECT_CONTEXT)
5697 count += 2 + ext->reg_count;
5698 else
5699 return 0;
5700 }
5701 }
5702 /* pa_sc_raster_config */
5703 count += 3;
5704 /* end clear state */
5705 count += 2;
5706 /* clear state */
5707 count += 2;
5708
5709 return count;
5710}
5711
5712void si_get_csb_buffer(struct radeon_device *rdev, volatile u32 *buffer)
5713{
5714 u32 count = 0, i;
5715 const struct cs_section_def *sect = NULL;
5716 const struct cs_extent_def *ext = NULL;
5717
5718 if (rdev->rlc.cs_data == NULL)
5719 return;
5720 if (buffer == NULL)
5721 return;
5722
5723 buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
5724 buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
5725
5726 buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
5727 buffer[count++] = cpu_to_le32(0x80000000);
5728 buffer[count++] = cpu_to_le32(0x80000000);
5729
5730 for (sect = rdev->rlc.cs_data; sect->section != NULL; ++sect) {
5731 for (ext = sect->section; ext->extent != NULL; ++ext) {
5732 if (sect->id == SECT_CONTEXT) {
5733 buffer[count++] =
5734 cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
5735 buffer[count++] = cpu_to_le32(ext->reg_index - 0xa000);
5736 for (i = 0; i < ext->reg_count; i++)
5737 buffer[count++] = cpu_to_le32(ext->extent[i]);
5738 } else {
5739 return;
5740 }
5741 }
5742 }
5743
5744 buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1));
5745 buffer[count++] = cpu_to_le32(PA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
5746 switch (rdev->family) {
5747 case CHIP_TAHITI:
5748 case CHIP_PITCAIRN:
5749 buffer[count++] = cpu_to_le32(0x2a00126a);
5750 break;
5751 case CHIP_VERDE:
5752 buffer[count++] = cpu_to_le32(0x0000124a);
5753 break;
5754 case CHIP_OLAND:
5755 buffer[count++] = cpu_to_le32(0x00000082);
5756 break;
5757 case CHIP_HAINAN:
5758 buffer[count++] = cpu_to_le32(0x00000000);
5759 break;
5760 default:
5761 buffer[count++] = cpu_to_le32(0x00000000);
5762 break;
5763 }
5764
5765 buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
5766 buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
5767
5768 buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
5769 buffer[count++] = cpu_to_le32(0);
5770}
5771
5772static void si_init_pg(struct radeon_device *rdev)
5773{
5774 if (rdev->pg_flags) {
5775 if (rdev->pg_flags & RADEON_PG_SUPPORT_SDMA) {
5776 si_init_dma_pg(rdev);
5777 }
5778 si_init_ao_cu_mask(rdev);
5779 if (rdev->pg_flags & RADEON_PG_SUPPORT_GFX_PG) {
5780 si_init_gfx_cgpg(rdev);
5781 } else {
5782 WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8);
5783 WREG32(RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8);
5784 }
5785 si_enable_dma_pg(rdev, true);
5786 si_enable_gfx_cgpg(rdev, true);
5787 } else {
5788 WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8);
5789 WREG32(RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8);
5790 }
5791}
5792
5793static void si_fini_pg(struct radeon_device *rdev)
5794{
5795 if (rdev->pg_flags) {
5796 si_enable_dma_pg(rdev, false);
5797 si_enable_gfx_cgpg(rdev, false);
5798 }
5799}
5800
5801/*
5802 * RLC
5803 */
5804void si_rlc_reset(struct radeon_device *rdev)
5805{
5806 u32 tmp = RREG32(GRBM_SOFT_RESET);
5807
5808 tmp |= SOFT_RESET_RLC;
5809 WREG32(GRBM_SOFT_RESET, tmp);
5810 udelay(50);
5811 tmp &= ~SOFT_RESET_RLC;
5812 WREG32(GRBM_SOFT_RESET, tmp);
5813 udelay(50);
5814}
5815
5816static void si_rlc_stop(struct radeon_device *rdev)
5817{
5818 WREG32(RLC_CNTL, 0);
5819
5820 si_enable_gui_idle_interrupt(rdev, false);
5821
5822 si_wait_for_rlc_serdes(rdev);
5823}
5824
5825static void si_rlc_start(struct radeon_device *rdev)
5826{
5827 WREG32(RLC_CNTL, RLC_ENABLE);
5828
5829 si_enable_gui_idle_interrupt(rdev, true);
5830
5831 udelay(50);
5832}
5833
5834static bool si_lbpw_supported(struct radeon_device *rdev)
5835{
5836 u32 tmp;
5837
5838 /* Enable LBPW only for DDR3 */
5839 tmp = RREG32(MC_SEQ_MISC0);
5840 if ((tmp & 0xF0000000) == 0xB0000000)
5841 return true;
5842 return false;
5843}
5844
5845static void si_enable_lbpw(struct radeon_device *rdev, bool enable)
5846{
5847 u32 tmp;
5848
5849 tmp = RREG32(RLC_LB_CNTL);
5850 if (enable)
5851 tmp |= LOAD_BALANCE_ENABLE;
5852 else
5853 tmp &= ~LOAD_BALANCE_ENABLE;
5854 WREG32(RLC_LB_CNTL, tmp);
5855
5856 if (!enable) {
5857 si_select_se_sh(rdev, 0xffffffff, 0xffffffff);
5858 WREG32(SPI_LB_CU_MASK, 0x00ff);
5859 }
5860}
5861
5862static int si_rlc_resume(struct radeon_device *rdev)
5863{
5864 u32 i;
5865
5866 if (!rdev->rlc_fw)
5867 return -EINVAL;
5868
5869 si_rlc_stop(rdev);
5870
5871 si_rlc_reset(rdev);
5872
5873 si_init_pg(rdev);
5874
5875 si_init_cg(rdev);
5876
5877 WREG32(RLC_RL_BASE, 0);
5878 WREG32(RLC_RL_SIZE, 0);
5879 WREG32(RLC_LB_CNTL, 0);
5880 WREG32(RLC_LB_CNTR_MAX, 0xffffffff);
5881 WREG32(RLC_LB_CNTR_INIT, 0);
5882 WREG32(RLC_LB_INIT_CU_MASK, 0xffffffff);
5883
5884 WREG32(RLC_MC_CNTL, 0);
5885 WREG32(RLC_UCODE_CNTL, 0);
5886
5887 if (rdev->new_fw) {
5888 const struct rlc_firmware_header_v1_0 *hdr =
5889 (const struct rlc_firmware_header_v1_0 *)rdev->rlc_fw->data;
5890 u32 fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
5891 const __le32 *fw_data = (const __le32 *)
5892 (rdev->rlc_fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes));
5893
5894 radeon_ucode_print_rlc_hdr(&hdr->header);
5895
5896 for (i = 0; i < fw_size; i++) {
5897 WREG32(RLC_UCODE_ADDR, i);
5898 WREG32(RLC_UCODE_DATA, le32_to_cpup(fw_data++));
5899 }
5900 } else {
5901 const __be32 *fw_data =
5902 (const __be32 *)rdev->rlc_fw->data;
5903 for (i = 0; i < SI_RLC_UCODE_SIZE; i++) {
5904 WREG32(RLC_UCODE_ADDR, i);
5905 WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++));
5906 }
5907 }
5908 WREG32(RLC_UCODE_ADDR, 0);
5909
5910 si_enable_lbpw(rdev, si_lbpw_supported(rdev));
5911
5912 si_rlc_start(rdev);
5913
5914 return 0;
5915}
5916
5917static void si_enable_interrupts(struct radeon_device *rdev)
5918{
5919 u32 ih_cntl = RREG32(IH_CNTL);
5920 u32 ih_rb_cntl = RREG32(IH_RB_CNTL);
5921
5922 ih_cntl |= ENABLE_INTR;
5923 ih_rb_cntl |= IH_RB_ENABLE;
5924 WREG32(IH_CNTL, ih_cntl);
5925 WREG32(IH_RB_CNTL, ih_rb_cntl);
5926 rdev->ih.enabled = true;
5927}
5928
5929static void si_disable_interrupts(struct radeon_device *rdev)
5930{
5931 u32 ih_rb_cntl = RREG32(IH_RB_CNTL);
5932 u32 ih_cntl = RREG32(IH_CNTL);
5933
5934 ih_rb_cntl &= ~IH_RB_ENABLE;
5935 ih_cntl &= ~ENABLE_INTR;
5936 WREG32(IH_RB_CNTL, ih_rb_cntl);
5937 WREG32(IH_CNTL, ih_cntl);
5938 /* set rptr, wptr to 0 */
5939 WREG32(IH_RB_RPTR, 0);
5940 WREG32(IH_RB_WPTR, 0);
5941 rdev->ih.enabled = false;
5942 rdev->ih.rptr = 0;
5943}
5944
5945static void si_disable_interrupt_state(struct radeon_device *rdev)
5946{
5947 u32 tmp;
5948
5949 tmp = RREG32(CP_INT_CNTL_RING0) &
5950 (CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE);
5951 WREG32(CP_INT_CNTL_RING0, tmp);
5952 WREG32(CP_INT_CNTL_RING1, 0);
5953 WREG32(CP_INT_CNTL_RING2, 0);
5954 tmp = RREG32(DMA_CNTL + DMA0_REGISTER_OFFSET) & ~TRAP_ENABLE;
5955 WREG32(DMA_CNTL + DMA0_REGISTER_OFFSET, tmp);
5956 tmp = RREG32(DMA_CNTL + DMA1_REGISTER_OFFSET) & ~TRAP_ENABLE;
5957 WREG32(DMA_CNTL + DMA1_REGISTER_OFFSET, tmp);
5958 WREG32(GRBM_INT_CNTL, 0);
5959 WREG32(SRBM_INT_CNTL, 0);
5960 if (rdev->num_crtc >= 2) {
5961 WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
5962 WREG32(INT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
5963 }
5964 if (rdev->num_crtc >= 4) {
5965 WREG32(INT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
5966 WREG32(INT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
5967 }
5968 if (rdev->num_crtc >= 6) {
5969 WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
5970 WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
5971 }
5972
5973 if (rdev->num_crtc >= 2) {
5974 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, 0);
5975 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, 0);
5976 }
5977 if (rdev->num_crtc >= 4) {
5978 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET, 0);
5979 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET, 0);
5980 }
5981 if (rdev->num_crtc >= 6) {
5982 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET, 0);
5983 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET, 0);
5984 }
5985
5986 if (!ASIC_IS_NODCE(rdev)) {
5987 WREG32(DAC_AUTODETECT_INT_CONTROL, 0);
5988
5989 tmp = RREG32(DC_HPD1_INT_CONTROL) & DC_HPDx_INT_POLARITY;
5990 WREG32(DC_HPD1_INT_CONTROL, tmp);
5991 tmp = RREG32(DC_HPD2_INT_CONTROL) & DC_HPDx_INT_POLARITY;
5992 WREG32(DC_HPD2_INT_CONTROL, tmp);
5993 tmp = RREG32(DC_HPD3_INT_CONTROL) & DC_HPDx_INT_POLARITY;
5994 WREG32(DC_HPD3_INT_CONTROL, tmp);
5995 tmp = RREG32(DC_HPD4_INT_CONTROL) & DC_HPDx_INT_POLARITY;
5996 WREG32(DC_HPD4_INT_CONTROL, tmp);
5997 tmp = RREG32(DC_HPD5_INT_CONTROL) & DC_HPDx_INT_POLARITY;
5998 WREG32(DC_HPD5_INT_CONTROL, tmp);
5999 tmp = RREG32(DC_HPD6_INT_CONTROL) & DC_HPDx_INT_POLARITY;
6000 WREG32(DC_HPD6_INT_CONTROL, tmp);
6001 }
6002}
6003
6004static int si_irq_init(struct radeon_device *rdev)
6005{
6006 int ret = 0;
6007 int rb_bufsz;
6008 u32 interrupt_cntl, ih_cntl, ih_rb_cntl;
6009
6010 /* allocate ring */
6011 ret = r600_ih_ring_alloc(rdev);
6012 if (ret)
6013 return ret;
6014
6015 /* disable irqs */
6016 si_disable_interrupts(rdev);
6017
6018 /* init rlc */
6019 ret = si_rlc_resume(rdev);
6020 if (ret) {
6021 r600_ih_ring_fini(rdev);
6022 return ret;
6023 }
6024
6025 /* setup interrupt control */
6026 /* set dummy read address to ring address */
6027 WREG32(INTERRUPT_CNTL2, rdev->ih.gpu_addr >> 8);
6028 interrupt_cntl = RREG32(INTERRUPT_CNTL);
6029 /* IH_DUMMY_RD_OVERRIDE=0 - dummy read disabled with msi, enabled without msi
6030 * IH_DUMMY_RD_OVERRIDE=1 - dummy read controlled by IH_DUMMY_RD_EN
6031 */
6032 interrupt_cntl &= ~IH_DUMMY_RD_OVERRIDE;
6033 /* IH_REQ_NONSNOOP_EN=1 if ring is in non-cacheable memory, e.g., vram */
6034 interrupt_cntl &= ~IH_REQ_NONSNOOP_EN;
6035 WREG32(INTERRUPT_CNTL, interrupt_cntl);
6036
6037 WREG32(IH_RB_BASE, rdev->ih.gpu_addr >> 8);
6038 rb_bufsz = order_base_2(rdev->ih.ring_size / 4);
6039
6040 ih_rb_cntl = (IH_WPTR_OVERFLOW_ENABLE |
6041 IH_WPTR_OVERFLOW_CLEAR |
6042 (rb_bufsz << 1));
6043
6044 if (rdev->wb.enabled)
6045 ih_rb_cntl |= IH_WPTR_WRITEBACK_ENABLE;
6046
6047 /* set the writeback address whether it's enabled or not */
6048 WREG32(IH_RB_WPTR_ADDR_LO, (rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFFFFFFFC);
6049 WREG32(IH_RB_WPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFF);
6050
6051 WREG32(IH_RB_CNTL, ih_rb_cntl);
6052
6053 /* set rptr, wptr to 0 */
6054 WREG32(IH_RB_RPTR, 0);
6055 WREG32(IH_RB_WPTR, 0);
6056
6057 /* Default settings for IH_CNTL (disabled at first) */
6058 ih_cntl = MC_WRREQ_CREDIT(0x10) | MC_WR_CLEAN_CNT(0x10) | MC_VMID(0);
6059 /* RPTR_REARM only works if msi's are enabled */
6060 if (rdev->msi_enabled)
6061 ih_cntl |= RPTR_REARM;
6062 WREG32(IH_CNTL, ih_cntl);
6063
6064 /* force the active interrupt state to all disabled */
6065 si_disable_interrupt_state(rdev);
6066
6067 pci_set_master(rdev->pdev);
6068
6069 /* enable irqs */
6070 si_enable_interrupts(rdev);
6071
6072 return ret;
6073}
6074
6075int si_irq_set(struct radeon_device *rdev)
6076{
6077 u32 cp_int_cntl;
6078 u32 cp_int_cntl1 = 0, cp_int_cntl2 = 0;
6079 u32 crtc1 = 0, crtc2 = 0, crtc3 = 0, crtc4 = 0, crtc5 = 0, crtc6 = 0;
6080 u32 hpd1 = 0, hpd2 = 0, hpd3 = 0, hpd4 = 0, hpd5 = 0, hpd6 = 0;
6081 u32 grbm_int_cntl = 0;
6082 u32 dma_cntl, dma_cntl1;
6083 u32 thermal_int = 0;
6084
6085 if (!rdev->irq.installed) {
6086 WARN(1, "Can't enable IRQ/MSI because no handler is installed\n");
6087 return -EINVAL;
6088 }
6089 /* don't enable anything if the ih is disabled */
6090 if (!rdev->ih.enabled) {
6091 si_disable_interrupts(rdev);
6092 /* force the active interrupt state to all disabled */
6093 si_disable_interrupt_state(rdev);
6094 return 0;
6095 }
6096
6097 cp_int_cntl = RREG32(CP_INT_CNTL_RING0) &
6098 (CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE);
6099
6100 if (!ASIC_IS_NODCE(rdev)) {
6101 hpd1 = RREG32(DC_HPD1_INT_CONTROL) & ~(DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN);
6102 hpd2 = RREG32(DC_HPD2_INT_CONTROL) & ~(DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN);
6103 hpd3 = RREG32(DC_HPD3_INT_CONTROL) & ~(DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN);
6104 hpd4 = RREG32(DC_HPD4_INT_CONTROL) & ~(DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN);
6105 hpd5 = RREG32(DC_HPD5_INT_CONTROL) & ~(DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN);
6106 hpd6 = RREG32(DC_HPD6_INT_CONTROL) & ~(DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN);
6107 }
6108
6109 dma_cntl = RREG32(DMA_CNTL + DMA0_REGISTER_OFFSET) & ~TRAP_ENABLE;
6110 dma_cntl1 = RREG32(DMA_CNTL + DMA1_REGISTER_OFFSET) & ~TRAP_ENABLE;
6111
6112 thermal_int = RREG32(CG_THERMAL_INT) &
6113 ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW);
6114
6115 /* enable CP interrupts on all rings */
6116 if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) {
6117 DRM_DEBUG("si_irq_set: sw int gfx\n");
6118 cp_int_cntl |= TIME_STAMP_INT_ENABLE;
6119 }
6120 if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP1_INDEX])) {
6121 DRM_DEBUG("si_irq_set: sw int cp1\n");
6122 cp_int_cntl1 |= TIME_STAMP_INT_ENABLE;
6123 }
6124 if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP2_INDEX])) {
6125 DRM_DEBUG("si_irq_set: sw int cp2\n");
6126 cp_int_cntl2 |= TIME_STAMP_INT_ENABLE;
6127 }
6128 if (atomic_read(&rdev->irq.ring_int[R600_RING_TYPE_DMA_INDEX])) {
6129 DRM_DEBUG("si_irq_set: sw int dma\n");
6130 dma_cntl |= TRAP_ENABLE;
6131 }
6132
6133 if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_DMA1_INDEX])) {
6134 DRM_DEBUG("si_irq_set: sw int dma1\n");
6135 dma_cntl1 |= TRAP_ENABLE;
6136 }
6137 if (rdev->irq.crtc_vblank_int[0] ||
6138 atomic_read(&rdev->irq.pflip[0])) {
6139 DRM_DEBUG("si_irq_set: vblank 0\n");
6140 crtc1 |= VBLANK_INT_MASK;
6141 }
6142 if (rdev->irq.crtc_vblank_int[1] ||
6143 atomic_read(&rdev->irq.pflip[1])) {
6144 DRM_DEBUG("si_irq_set: vblank 1\n");
6145 crtc2 |= VBLANK_INT_MASK;
6146 }
6147 if (rdev->irq.crtc_vblank_int[2] ||
6148 atomic_read(&rdev->irq.pflip[2])) {
6149 DRM_DEBUG("si_irq_set: vblank 2\n");
6150 crtc3 |= VBLANK_INT_MASK;
6151 }
6152 if (rdev->irq.crtc_vblank_int[3] ||
6153 atomic_read(&rdev->irq.pflip[3])) {
6154 DRM_DEBUG("si_irq_set: vblank 3\n");
6155 crtc4 |= VBLANK_INT_MASK;
6156 }
6157 if (rdev->irq.crtc_vblank_int[4] ||
6158 atomic_read(&rdev->irq.pflip[4])) {
6159 DRM_DEBUG("si_irq_set: vblank 4\n");
6160 crtc5 |= VBLANK_INT_MASK;
6161 }
6162 if (rdev->irq.crtc_vblank_int[5] ||
6163 atomic_read(&rdev->irq.pflip[5])) {
6164 DRM_DEBUG("si_irq_set: vblank 5\n");
6165 crtc6 |= VBLANK_INT_MASK;
6166 }
6167 if (rdev->irq.hpd[0]) {
6168 DRM_DEBUG("si_irq_set: hpd 1\n");
6169 hpd1 |= DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN;
6170 }
6171 if (rdev->irq.hpd[1]) {
6172 DRM_DEBUG("si_irq_set: hpd 2\n");
6173 hpd2 |= DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN;
6174 }
6175 if (rdev->irq.hpd[2]) {
6176 DRM_DEBUG("si_irq_set: hpd 3\n");
6177 hpd3 |= DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN;
6178 }
6179 if (rdev->irq.hpd[3]) {
6180 DRM_DEBUG("si_irq_set: hpd 4\n");
6181 hpd4 |= DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN;
6182 }
6183 if (rdev->irq.hpd[4]) {
6184 DRM_DEBUG("si_irq_set: hpd 5\n");
6185 hpd5 |= DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN;
6186 }
6187 if (rdev->irq.hpd[5]) {
6188 DRM_DEBUG("si_irq_set: hpd 6\n");
6189 hpd6 |= DC_HPDx_INT_EN | DC_HPDx_RX_INT_EN;
6190 }
6191
6192 WREG32(CP_INT_CNTL_RING0, cp_int_cntl);
6193 WREG32(CP_INT_CNTL_RING1, cp_int_cntl1);
6194 WREG32(CP_INT_CNTL_RING2, cp_int_cntl2);
6195
6196 WREG32(DMA_CNTL + DMA0_REGISTER_OFFSET, dma_cntl);
6197 WREG32(DMA_CNTL + DMA1_REGISTER_OFFSET, dma_cntl1);
6198
6199 WREG32(GRBM_INT_CNTL, grbm_int_cntl);
6200
6201 if (rdev->irq.dpm_thermal) {
6202 DRM_DEBUG("dpm thermal\n");
6203 thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW;
6204 }
6205
6206 if (rdev->num_crtc >= 2) {
6207 WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, crtc1);
6208 WREG32(INT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, crtc2);
6209 }
6210 if (rdev->num_crtc >= 4) {
6211 WREG32(INT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, crtc3);
6212 WREG32(INT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, crtc4);
6213 }
6214 if (rdev->num_crtc >= 6) {
6215 WREG32(INT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, crtc5);
6216 WREG32(INT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, crtc6);
6217 }
6218
6219 if (rdev->num_crtc >= 2) {
6220 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET,
6221 GRPH_PFLIP_INT_MASK);
6222 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET,
6223 GRPH_PFLIP_INT_MASK);
6224 }
6225 if (rdev->num_crtc >= 4) {
6226 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC2_REGISTER_OFFSET,
6227 GRPH_PFLIP_INT_MASK);
6228 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC3_REGISTER_OFFSET,
6229 GRPH_PFLIP_INT_MASK);
6230 }
6231 if (rdev->num_crtc >= 6) {
6232 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC4_REGISTER_OFFSET,
6233 GRPH_PFLIP_INT_MASK);
6234 WREG32(GRPH_INT_CONTROL + EVERGREEN_CRTC5_REGISTER_OFFSET,
6235 GRPH_PFLIP_INT_MASK);
6236 }
6237
6238 if (!ASIC_IS_NODCE(rdev)) {
6239 WREG32(DC_HPD1_INT_CONTROL, hpd1);
6240 WREG32(DC_HPD2_INT_CONTROL, hpd2);
6241 WREG32(DC_HPD3_INT_CONTROL, hpd3);
6242 WREG32(DC_HPD4_INT_CONTROL, hpd4);
6243 WREG32(DC_HPD5_INT_CONTROL, hpd5);
6244 WREG32(DC_HPD6_INT_CONTROL, hpd6);
6245 }
6246
6247 WREG32(CG_THERMAL_INT, thermal_int);
6248
6249 /* posting read */
6250 RREG32(SRBM_STATUS);
6251
6252 return 0;
6253}
6254
6255static inline void si_irq_ack(struct radeon_device *rdev)
6256{
6257 u32 tmp;
6258
6259 if (ASIC_IS_NODCE(rdev))
6260 return;
6261
6262 rdev->irq.stat_regs.evergreen.disp_int = RREG32(DISP_INTERRUPT_STATUS);
6263 rdev->irq.stat_regs.evergreen.disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE);
6264 rdev->irq.stat_regs.evergreen.disp_int_cont2 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE2);
6265 rdev->irq.stat_regs.evergreen.disp_int_cont3 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE3);
6266 rdev->irq.stat_regs.evergreen.disp_int_cont4 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE4);
6267 rdev->irq.stat_regs.evergreen.disp_int_cont5 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE5);
6268 rdev->irq.stat_regs.evergreen.d1grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET);
6269 rdev->irq.stat_regs.evergreen.d2grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET);
6270 if (rdev->num_crtc >= 4) {
6271 rdev->irq.stat_regs.evergreen.d3grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET);
6272 rdev->irq.stat_regs.evergreen.d4grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET);
6273 }
6274 if (rdev->num_crtc >= 6) {
6275 rdev->irq.stat_regs.evergreen.d5grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET);
6276 rdev->irq.stat_regs.evergreen.d6grph_int = RREG32(GRPH_INT_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET);
6277 }
6278
6279 if (rdev->irq.stat_regs.evergreen.d1grph_int & GRPH_PFLIP_INT_OCCURRED)
6280 WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
6281 if (rdev->irq.stat_regs.evergreen.d2grph_int & GRPH_PFLIP_INT_OCCURRED)
6282 WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
6283 if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VBLANK_INTERRUPT)
6284 WREG32(VBLANK_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VBLANK_ACK);
6285 if (rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VLINE_INTERRUPT)
6286 WREG32(VLINE_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VLINE_ACK);
6287 if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VBLANK_INTERRUPT)
6288 WREG32(VBLANK_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VBLANK_ACK);
6289 if (rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VLINE_INTERRUPT)
6290 WREG32(VLINE_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VLINE_ACK);
6291
6292 if (rdev->num_crtc >= 4) {
6293 if (rdev->irq.stat_regs.evergreen.d3grph_int & GRPH_PFLIP_INT_OCCURRED)
6294 WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
6295 if (rdev->irq.stat_regs.evergreen.d4grph_int & GRPH_PFLIP_INT_OCCURRED)
6296 WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
6297 if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT)
6298 WREG32(VBLANK_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VBLANK_ACK);
6299 if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VLINE_INTERRUPT)
6300 WREG32(VLINE_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VLINE_ACK);
6301 if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT)
6302 WREG32(VBLANK_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VBLANK_ACK);
6303 if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VLINE_INTERRUPT)
6304 WREG32(VLINE_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VLINE_ACK);
6305 }
6306
6307 if (rdev->num_crtc >= 6) {
6308 if (rdev->irq.stat_regs.evergreen.d5grph_int & GRPH_PFLIP_INT_OCCURRED)
6309 WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
6310 if (rdev->irq.stat_regs.evergreen.d6grph_int & GRPH_PFLIP_INT_OCCURRED)
6311 WREG32(GRPH_INT_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, GRPH_PFLIP_INT_CLEAR);
6312 if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT)
6313 WREG32(VBLANK_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VBLANK_ACK);
6314 if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VLINE_INTERRUPT)
6315 WREG32(VLINE_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VLINE_ACK);
6316 if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT)
6317 WREG32(VBLANK_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VBLANK_ACK);
6318 if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VLINE_INTERRUPT)
6319 WREG32(VLINE_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VLINE_ACK);
6320 }
6321
6322 if (rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_INTERRUPT) {
6323 tmp = RREG32(DC_HPD1_INT_CONTROL);
6324 tmp |= DC_HPDx_INT_ACK;
6325 WREG32(DC_HPD1_INT_CONTROL, tmp);
6326 }
6327 if (rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_INTERRUPT) {
6328 tmp = RREG32(DC_HPD2_INT_CONTROL);
6329 tmp |= DC_HPDx_INT_ACK;
6330 WREG32(DC_HPD2_INT_CONTROL, tmp);
6331 }
6332 if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_INTERRUPT) {
6333 tmp = RREG32(DC_HPD3_INT_CONTROL);
6334 tmp |= DC_HPDx_INT_ACK;
6335 WREG32(DC_HPD3_INT_CONTROL, tmp);
6336 }
6337 if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_INTERRUPT) {
6338 tmp = RREG32(DC_HPD4_INT_CONTROL);
6339 tmp |= DC_HPDx_INT_ACK;
6340 WREG32(DC_HPD4_INT_CONTROL, tmp);
6341 }
6342 if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_INTERRUPT) {
6343 tmp = RREG32(DC_HPD5_INT_CONTROL);
6344 tmp |= DC_HPDx_INT_ACK;
6345 WREG32(DC_HPD5_INT_CONTROL, tmp);
6346 }
6347 if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_INTERRUPT) {
6348 tmp = RREG32(DC_HPD5_INT_CONTROL);
6349 tmp |= DC_HPDx_INT_ACK;
6350 WREG32(DC_HPD6_INT_CONTROL, tmp);
6351 }
6352
6353 if (rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_RX_INTERRUPT) {
6354 tmp = RREG32(DC_HPD1_INT_CONTROL);
6355 tmp |= DC_HPDx_RX_INT_ACK;
6356 WREG32(DC_HPD1_INT_CONTROL, tmp);
6357 }
6358 if (rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_RX_INTERRUPT) {
6359 tmp = RREG32(DC_HPD2_INT_CONTROL);
6360 tmp |= DC_HPDx_RX_INT_ACK;
6361 WREG32(DC_HPD2_INT_CONTROL, tmp);
6362 }
6363 if (rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_RX_INTERRUPT) {
6364 tmp = RREG32(DC_HPD3_INT_CONTROL);
6365 tmp |= DC_HPDx_RX_INT_ACK;
6366 WREG32(DC_HPD3_INT_CONTROL, tmp);
6367 }
6368 if (rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_RX_INTERRUPT) {
6369 tmp = RREG32(DC_HPD4_INT_CONTROL);
6370 tmp |= DC_HPDx_RX_INT_ACK;
6371 WREG32(DC_HPD4_INT_CONTROL, tmp);
6372 }
6373 if (rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_RX_INTERRUPT) {
6374 tmp = RREG32(DC_HPD5_INT_CONTROL);
6375 tmp |= DC_HPDx_RX_INT_ACK;
6376 WREG32(DC_HPD5_INT_CONTROL, tmp);
6377 }
6378 if (rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_RX_INTERRUPT) {
6379 tmp = RREG32(DC_HPD5_INT_CONTROL);
6380 tmp |= DC_HPDx_RX_INT_ACK;
6381 WREG32(DC_HPD6_INT_CONTROL, tmp);
6382 }
6383}
6384
6385static void si_irq_disable(struct radeon_device *rdev)
6386{
6387 si_disable_interrupts(rdev);
6388 /* Wait and acknowledge irq */
6389 mdelay(1);
6390 si_irq_ack(rdev);
6391 si_disable_interrupt_state(rdev);
6392}
6393
6394static void si_irq_suspend(struct radeon_device *rdev)
6395{
6396 si_irq_disable(rdev);
6397 si_rlc_stop(rdev);
6398}
6399
6400static void si_irq_fini(struct radeon_device *rdev)
6401{
6402 si_irq_suspend(rdev);
6403 r600_ih_ring_fini(rdev);
6404}
6405
6406static inline u32 si_get_ih_wptr(struct radeon_device *rdev)
6407{
6408 u32 wptr, tmp;
6409
6410 if (rdev->wb.enabled)
6411 wptr = le32_to_cpu(rdev->wb.wb[R600_WB_IH_WPTR_OFFSET/4]);
6412 else
6413 wptr = RREG32(IH_RB_WPTR);
6414
6415 if (wptr & RB_OVERFLOW) {
6416 wptr &= ~RB_OVERFLOW;
6417 /* When a ring buffer overflow happen start parsing interrupt
6418 * from the last not overwritten vector (wptr + 16). Hopefully
6419 * this should allow us to catchup.
6420 */
6421 dev_warn(rdev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",
6422 wptr, rdev->ih.rptr, (wptr + 16) & rdev->ih.ptr_mask);
6423 rdev->ih.rptr = (wptr + 16) & rdev->ih.ptr_mask;
6424 tmp = RREG32(IH_RB_CNTL);
6425 tmp |= IH_WPTR_OVERFLOW_CLEAR;
6426 WREG32(IH_RB_CNTL, tmp);
6427 }
6428 return (wptr & rdev->ih.ptr_mask);
6429}
6430
6431/* SI IV Ring
6432 * Each IV ring entry is 128 bits:
6433 * [7:0] - interrupt source id
6434 * [31:8] - reserved
6435 * [59:32] - interrupt source data
6436 * [63:60] - reserved
6437 * [71:64] - RINGID
6438 * [79:72] - VMID
6439 * [127:80] - reserved
6440 */
6441int si_irq_process(struct radeon_device *rdev)
6442{
6443 u32 wptr;
6444 u32 rptr;
6445 u32 src_id, src_data, ring_id;
6446 u32 ring_index;
6447 bool queue_hotplug = false;
6448 bool queue_dp = false;
6449 bool queue_thermal = false;
6450 u32 status, addr;
6451
6452 if (!rdev->ih.enabled || rdev->shutdown)
6453 return IRQ_NONE;
6454
6455 wptr = si_get_ih_wptr(rdev);
6456
6457restart_ih:
6458 /* is somebody else already processing irqs? */
6459 if (atomic_xchg(&rdev->ih.lock, 1))
6460 return IRQ_NONE;
6461
6462 rptr = rdev->ih.rptr;
6463 DRM_DEBUG("si_irq_process start: rptr %d, wptr %d\n", rptr, wptr);
6464
6465 /* Order reading of wptr vs. reading of IH ring data */
6466 rmb();
6467
6468 /* display interrupts */
6469 si_irq_ack(rdev);
6470
6471 while (rptr != wptr) {
6472 /* wptr/rptr are in bytes! */
6473 ring_index = rptr / 4;
6474 src_id = le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff;
6475 src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff;
6476 ring_id = le32_to_cpu(rdev->ih.ring[ring_index + 2]) & 0xff;
6477
6478 switch (src_id) {
6479 case 1: /* D1 vblank/vline */
6480 switch (src_data) {
6481 case 0: /* D1 vblank */
6482 if (!(rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VBLANK_INTERRUPT))
6483 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6484
6485 if (rdev->irq.crtc_vblank_int[0]) {
6486 drm_handle_vblank(rdev->ddev, 0);
6487 rdev->pm.vblank_sync = true;
6488 wake_up(&rdev->irq.vblank_queue);
6489 }
6490 if (atomic_read(&rdev->irq.pflip[0]))
6491 radeon_crtc_handle_vblank(rdev, 0);
6492 rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VBLANK_INTERRUPT;
6493 DRM_DEBUG("IH: D1 vblank\n");
6494
6495 break;
6496 case 1: /* D1 vline */
6497 if (!(rdev->irq.stat_regs.evergreen.disp_int & LB_D1_VLINE_INTERRUPT))
6498 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6499
6500 rdev->irq.stat_regs.evergreen.disp_int &= ~LB_D1_VLINE_INTERRUPT;
6501 DRM_DEBUG("IH: D1 vline\n");
6502
6503 break;
6504 default:
6505 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6506 break;
6507 }
6508 break;
6509 case 2: /* D2 vblank/vline */
6510 switch (src_data) {
6511 case 0: /* D2 vblank */
6512 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VBLANK_INTERRUPT))
6513 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6514
6515 if (rdev->irq.crtc_vblank_int[1]) {
6516 drm_handle_vblank(rdev->ddev, 1);
6517 rdev->pm.vblank_sync = true;
6518 wake_up(&rdev->irq.vblank_queue);
6519 }
6520 if (atomic_read(&rdev->irq.pflip[1]))
6521 radeon_crtc_handle_vblank(rdev, 1);
6522 rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT;
6523 DRM_DEBUG("IH: D2 vblank\n");
6524
6525 break;
6526 case 1: /* D2 vline */
6527 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont & LB_D2_VLINE_INTERRUPT))
6528 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6529
6530 rdev->irq.stat_regs.evergreen.disp_int_cont &= ~LB_D2_VLINE_INTERRUPT;
6531 DRM_DEBUG("IH: D2 vline\n");
6532
6533 break;
6534 default:
6535 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6536 break;
6537 }
6538 break;
6539 case 3: /* D3 vblank/vline */
6540 switch (src_data) {
6541 case 0: /* D3 vblank */
6542 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT))
6543 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6544
6545 if (rdev->irq.crtc_vblank_int[2]) {
6546 drm_handle_vblank(rdev->ddev, 2);
6547 rdev->pm.vblank_sync = true;
6548 wake_up(&rdev->irq.vblank_queue);
6549 }
6550 if (atomic_read(&rdev->irq.pflip[2]))
6551 radeon_crtc_handle_vblank(rdev, 2);
6552 rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT;
6553 DRM_DEBUG("IH: D3 vblank\n");
6554
6555 break;
6556 case 1: /* D3 vline */
6557 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont2 & LB_D3_VLINE_INTERRUPT))
6558 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6559
6560 rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~LB_D3_VLINE_INTERRUPT;
6561 DRM_DEBUG("IH: D3 vline\n");
6562
6563 break;
6564 default:
6565 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6566 break;
6567 }
6568 break;
6569 case 4: /* D4 vblank/vline */
6570 switch (src_data) {
6571 case 0: /* D4 vblank */
6572 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT))
6573 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6574
6575 if (rdev->irq.crtc_vblank_int[3]) {
6576 drm_handle_vblank(rdev->ddev, 3);
6577 rdev->pm.vblank_sync = true;
6578 wake_up(&rdev->irq.vblank_queue);
6579 }
6580 if (atomic_read(&rdev->irq.pflip[3]))
6581 radeon_crtc_handle_vblank(rdev, 3);
6582 rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT;
6583 DRM_DEBUG("IH: D4 vblank\n");
6584
6585 break;
6586 case 1: /* D4 vline */
6587 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont3 & LB_D4_VLINE_INTERRUPT))
6588 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6589
6590 rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~LB_D4_VLINE_INTERRUPT;
6591 DRM_DEBUG("IH: D4 vline\n");
6592
6593 break;
6594 default:
6595 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6596 break;
6597 }
6598 break;
6599 case 5: /* D5 vblank/vline */
6600 switch (src_data) {
6601 case 0: /* D5 vblank */
6602 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT))
6603 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6604
6605 if (rdev->irq.crtc_vblank_int[4]) {
6606 drm_handle_vblank(rdev->ddev, 4);
6607 rdev->pm.vblank_sync = true;
6608 wake_up(&rdev->irq.vblank_queue);
6609 }
6610 if (atomic_read(&rdev->irq.pflip[4]))
6611 radeon_crtc_handle_vblank(rdev, 4);
6612 rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT;
6613 DRM_DEBUG("IH: D5 vblank\n");
6614
6615 break;
6616 case 1: /* D5 vline */
6617 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont4 & LB_D5_VLINE_INTERRUPT))
6618 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6619
6620 rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~LB_D5_VLINE_INTERRUPT;
6621 DRM_DEBUG("IH: D5 vline\n");
6622
6623 break;
6624 default:
6625 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6626 break;
6627 }
6628 break;
6629 case 6: /* D6 vblank/vline */
6630 switch (src_data) {
6631 case 0: /* D6 vblank */
6632 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT))
6633 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6634
6635 if (rdev->irq.crtc_vblank_int[5]) {
6636 drm_handle_vblank(rdev->ddev, 5);
6637 rdev->pm.vblank_sync = true;
6638 wake_up(&rdev->irq.vblank_queue);
6639 }
6640 if (atomic_read(&rdev->irq.pflip[5]))
6641 radeon_crtc_handle_vblank(rdev, 5);
6642 rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT;
6643 DRM_DEBUG("IH: D6 vblank\n");
6644
6645 break;
6646 case 1: /* D6 vline */
6647 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont5 & LB_D6_VLINE_INTERRUPT))
6648 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6649
6650 rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~LB_D6_VLINE_INTERRUPT;
6651 DRM_DEBUG("IH: D6 vline\n");
6652
6653 break;
6654 default:
6655 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6656 break;
6657 }
6658 break;
6659 case 8: /* D1 page flip */
6660 case 10: /* D2 page flip */
6661 case 12: /* D3 page flip */
6662 case 14: /* D4 page flip */
6663 case 16: /* D5 page flip */
6664 case 18: /* D6 page flip */
6665 DRM_DEBUG("IH: D%d flip\n", ((src_id - 8) >> 1) + 1);
6666 if (radeon_use_pflipirq > 0)
6667 radeon_crtc_handle_flip(rdev, (src_id - 8) >> 1);
6668 break;
6669 case 42: /* HPD hotplug */
6670 switch (src_data) {
6671 case 0:
6672 if (!(rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_INTERRUPT))
6673 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6674
6675 rdev->irq.stat_regs.evergreen.disp_int &= ~DC_HPD1_INTERRUPT;
6676 queue_hotplug = true;
6677 DRM_DEBUG("IH: HPD1\n");
6678
6679 break;
6680 case 1:
6681 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_INTERRUPT))
6682 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6683
6684 rdev->irq.stat_regs.evergreen.disp_int_cont &= ~DC_HPD2_INTERRUPT;
6685 queue_hotplug = true;
6686 DRM_DEBUG("IH: HPD2\n");
6687
6688 break;
6689 case 2:
6690 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_INTERRUPT))
6691 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6692
6693 rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~DC_HPD3_INTERRUPT;
6694 queue_hotplug = true;
6695 DRM_DEBUG("IH: HPD3\n");
6696
6697 break;
6698 case 3:
6699 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_INTERRUPT))
6700 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6701
6702 rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~DC_HPD4_INTERRUPT;
6703 queue_hotplug = true;
6704 DRM_DEBUG("IH: HPD4\n");
6705
6706 break;
6707 case 4:
6708 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_INTERRUPT))
6709 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6710
6711 rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~DC_HPD5_INTERRUPT;
6712 queue_hotplug = true;
6713 DRM_DEBUG("IH: HPD5\n");
6714
6715 break;
6716 case 5:
6717 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_INTERRUPT))
6718 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6719
6720 rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~DC_HPD6_INTERRUPT;
6721 queue_hotplug = true;
6722 DRM_DEBUG("IH: HPD6\n");
6723
6724 break;
6725 case 6:
6726 if (!(rdev->irq.stat_regs.evergreen.disp_int & DC_HPD1_RX_INTERRUPT))
6727 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6728
6729 rdev->irq.stat_regs.evergreen.disp_int &= ~DC_HPD1_RX_INTERRUPT;
6730 queue_dp = true;
6731 DRM_DEBUG("IH: HPD_RX 1\n");
6732
6733 break;
6734 case 7:
6735 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont & DC_HPD2_RX_INTERRUPT))
6736 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6737
6738 rdev->irq.stat_regs.evergreen.disp_int_cont &= ~DC_HPD2_RX_INTERRUPT;
6739 queue_dp = true;
6740 DRM_DEBUG("IH: HPD_RX 2\n");
6741
6742 break;
6743 case 8:
6744 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont2 & DC_HPD3_RX_INTERRUPT))
6745 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6746
6747 rdev->irq.stat_regs.evergreen.disp_int_cont2 &= ~DC_HPD3_RX_INTERRUPT;
6748 queue_dp = true;
6749 DRM_DEBUG("IH: HPD_RX 3\n");
6750
6751 break;
6752 case 9:
6753 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont3 & DC_HPD4_RX_INTERRUPT))
6754 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6755
6756 rdev->irq.stat_regs.evergreen.disp_int_cont3 &= ~DC_HPD4_RX_INTERRUPT;
6757 queue_dp = true;
6758 DRM_DEBUG("IH: HPD_RX 4\n");
6759
6760 break;
6761 case 10:
6762 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont4 & DC_HPD5_RX_INTERRUPT))
6763 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6764
6765 rdev->irq.stat_regs.evergreen.disp_int_cont4 &= ~DC_HPD5_RX_INTERRUPT;
6766 queue_dp = true;
6767 DRM_DEBUG("IH: HPD_RX 5\n");
6768
6769 break;
6770 case 11:
6771 if (!(rdev->irq.stat_regs.evergreen.disp_int_cont5 & DC_HPD6_RX_INTERRUPT))
6772 DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
6773
6774 rdev->irq.stat_regs.evergreen.disp_int_cont5 &= ~DC_HPD6_RX_INTERRUPT;
6775 queue_dp = true;
6776 DRM_DEBUG("IH: HPD_RX 6\n");
6777
6778 break;
6779 default:
6780 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6781 break;
6782 }
6783 break;
6784 case 96:
6785 DRM_ERROR("SRBM_READ_ERROR: 0x%x\n", RREG32(SRBM_READ_ERROR));
6786 WREG32(SRBM_INT_ACK, 0x1);
6787 break;
6788 case 124: /* UVD */
6789 DRM_DEBUG("IH: UVD int: 0x%08x\n", src_data);
6790 radeon_fence_process(rdev, R600_RING_TYPE_UVD_INDEX);
6791 break;
6792 case 146:
6793 case 147:
6794 addr = RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR);
6795 status = RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS);
6796 /* reset addr and status */
6797 WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1);
6798 if (addr == 0x0 && status == 0x0)
6799 break;
6800 dev_err(rdev->dev, "GPU fault detected: %d 0x%08x\n", src_id, src_data);
6801 dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x%08X\n",
6802 addr);
6803 dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
6804 status);
6805 si_vm_decode_fault(rdev, status, addr);
6806 break;
6807 case 176: /* RINGID0 CP_INT */
6808 radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
6809 break;
6810 case 177: /* RINGID1 CP_INT */
6811 radeon_fence_process(rdev, CAYMAN_RING_TYPE_CP1_INDEX);
6812 break;
6813 case 178: /* RINGID2 CP_INT */
6814 radeon_fence_process(rdev, CAYMAN_RING_TYPE_CP2_INDEX);
6815 break;
6816 case 181: /* CP EOP event */
6817 DRM_DEBUG("IH: CP EOP\n");
6818 switch (ring_id) {
6819 case 0:
6820 radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
6821 break;
6822 case 1:
6823 radeon_fence_process(rdev, CAYMAN_RING_TYPE_CP1_INDEX);
6824 break;
6825 case 2:
6826 radeon_fence_process(rdev, CAYMAN_RING_TYPE_CP2_INDEX);
6827 break;
6828 }
6829 break;
6830 case 224: /* DMA trap event */
6831 DRM_DEBUG("IH: DMA trap\n");
6832 radeon_fence_process(rdev, R600_RING_TYPE_DMA_INDEX);
6833 break;
6834 case 230: /* thermal low to high */
6835 DRM_DEBUG("IH: thermal low to high\n");
6836 rdev->pm.dpm.thermal.high_to_low = false;
6837 queue_thermal = true;
6838 break;
6839 case 231: /* thermal high to low */
6840 DRM_DEBUG("IH: thermal high to low\n");
6841 rdev->pm.dpm.thermal.high_to_low = true;
6842 queue_thermal = true;
6843 break;
6844 case 233: /* GUI IDLE */
6845 DRM_DEBUG("IH: GUI idle\n");
6846 break;
6847 case 244: /* DMA trap event */
6848 DRM_DEBUG("IH: DMA1 trap\n");
6849 radeon_fence_process(rdev, CAYMAN_RING_TYPE_DMA1_INDEX);
6850 break;
6851 default:
6852 DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data);
6853 break;
6854 }
6855
6856 /* wptr/rptr are in bytes! */
6857 rptr += 16;
6858 rptr &= rdev->ih.ptr_mask;
6859 WREG32(IH_RB_RPTR, rptr);
6860 }
6861 if (queue_dp)
6862 schedule_work(&rdev->dp_work);
6863 if (queue_hotplug)
6864 schedule_delayed_work(&rdev->hotplug_work, 0);
6865 if (queue_thermal && rdev->pm.dpm_enabled)
6866 schedule_work(&rdev->pm.dpm.thermal.work);
6867 rdev->ih.rptr = rptr;
6868 atomic_set(&rdev->ih.lock, 0);
6869
6870 /* make sure wptr hasn't changed while processing */
6871 wptr = si_get_ih_wptr(rdev);
6872 if (wptr != rptr)
6873 goto restart_ih;
6874
6875 return IRQ_HANDLED;
6876}
6877
6878/*
6879 * startup/shutdown callbacks
6880 */
6881static void si_uvd_init(struct radeon_device *rdev)
6882{
6883 int r;
6884
6885 if (!rdev->has_uvd)
6886 return;
6887
6888 r = radeon_uvd_init(rdev);
6889 if (r) {
6890 dev_err(rdev->dev, "failed UVD (%d) init.\n", r);
6891 /*
6892 * At this point rdev->uvd.vcpu_bo is NULL which trickles down
6893 * to early fails uvd_v2_2_resume() and thus nothing happens
6894 * there. So it is pointless to try to go through that code
6895 * hence why we disable uvd here.
6896 */
6897 rdev->has_uvd = 0;
6898 return;
6899 }
6900 rdev->ring[R600_RING_TYPE_UVD_INDEX].ring_obj = NULL;
6901 r600_ring_init(rdev, &rdev->ring[R600_RING_TYPE_UVD_INDEX], 4096);
6902}
6903
6904static void si_uvd_start(struct radeon_device *rdev)
6905{
6906 int r;
6907
6908 if (!rdev->has_uvd)
6909 return;
6910
6911 r = uvd_v2_2_resume(rdev);
6912 if (r) {
6913 dev_err(rdev->dev, "failed UVD resume (%d).\n", r);
6914 goto error;
6915 }
6916 r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX);
6917 if (r) {
6918 dev_err(rdev->dev, "failed initializing UVD fences (%d).\n", r);
6919 goto error;
6920 }
6921 return;
6922
6923error:
6924 rdev->ring[R600_RING_TYPE_UVD_INDEX].ring_size = 0;
6925}
6926
6927static void si_uvd_resume(struct radeon_device *rdev)
6928{
6929 struct radeon_ring *ring;
6930 int r;
6931
6932 if (!rdev->has_uvd || !rdev->ring[R600_RING_TYPE_UVD_INDEX].ring_size)
6933 return;
6934
6935 ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX];
6936 r = radeon_ring_init(rdev, ring, ring->ring_size, 0, PACKET0(UVD_NO_OP, 0));
6937 if (r) {
6938 dev_err(rdev->dev, "failed initializing UVD ring (%d).\n", r);
6939 return;
6940 }
6941 r = uvd_v1_0_init(rdev);
6942 if (r) {
6943 dev_err(rdev->dev, "failed initializing UVD (%d).\n", r);
6944 return;
6945 }
6946}
6947
6948static void si_vce_init(struct radeon_device *rdev)
6949{
6950 int r;
6951
6952 if (!rdev->has_vce)
6953 return;
6954
6955 r = radeon_vce_init(rdev);
6956 if (r) {
6957 dev_err(rdev->dev, "failed VCE (%d) init.\n", r);
6958 /*
6959 * At this point rdev->vce.vcpu_bo is NULL which trickles down
6960 * to early fails si_vce_start() and thus nothing happens
6961 * there. So it is pointless to try to go through that code
6962 * hence why we disable vce here.
6963 */
6964 rdev->has_vce = 0;
6965 return;
6966 }
6967 rdev->ring[TN_RING_TYPE_VCE1_INDEX].ring_obj = NULL;
6968 r600_ring_init(rdev, &rdev->ring[TN_RING_TYPE_VCE1_INDEX], 4096);
6969 rdev->ring[TN_RING_TYPE_VCE2_INDEX].ring_obj = NULL;
6970 r600_ring_init(rdev, &rdev->ring[TN_RING_TYPE_VCE2_INDEX], 4096);
6971}
6972
6973static void si_vce_start(struct radeon_device *rdev)
6974{
6975 int r;
6976
6977 if (!rdev->has_vce)
6978 return;
6979
6980 r = radeon_vce_resume(rdev);
6981 if (r) {
6982 dev_err(rdev->dev, "failed VCE resume (%d).\n", r);
6983 goto error;
6984 }
6985 r = vce_v1_0_resume(rdev);
6986 if (r) {
6987 dev_err(rdev->dev, "failed VCE resume (%d).\n", r);
6988 goto error;
6989 }
6990 r = radeon_fence_driver_start_ring(rdev, TN_RING_TYPE_VCE1_INDEX);
6991 if (r) {
6992 dev_err(rdev->dev, "failed initializing VCE1 fences (%d).\n", r);
6993 goto error;
6994 }
6995 r = radeon_fence_driver_start_ring(rdev, TN_RING_TYPE_VCE2_INDEX);
6996 if (r) {
6997 dev_err(rdev->dev, "failed initializing VCE2 fences (%d).\n", r);
6998 goto error;
6999 }
7000 return;
7001
7002error:
7003 rdev->ring[TN_RING_TYPE_VCE1_INDEX].ring_size = 0;
7004 rdev->ring[TN_RING_TYPE_VCE2_INDEX].ring_size = 0;
7005}
7006
7007static void si_vce_resume(struct radeon_device *rdev)
7008{
7009 struct radeon_ring *ring;
7010 int r;
7011
7012 if (!rdev->has_vce || !rdev->ring[TN_RING_TYPE_VCE1_INDEX].ring_size)
7013 return;
7014
7015 ring = &rdev->ring[TN_RING_TYPE_VCE1_INDEX];
7016 r = radeon_ring_init(rdev, ring, ring->ring_size, 0, VCE_CMD_NO_OP);
7017 if (r) {
7018 dev_err(rdev->dev, "failed initializing VCE1 ring (%d).\n", r);
7019 return;
7020 }
7021 ring = &rdev->ring[TN_RING_TYPE_VCE2_INDEX];
7022 r = radeon_ring_init(rdev, ring, ring->ring_size, 0, VCE_CMD_NO_OP);
7023 if (r) {
7024 dev_err(rdev->dev, "failed initializing VCE1 ring (%d).\n", r);
7025 return;
7026 }
7027 r = vce_v1_0_init(rdev);
7028 if (r) {
7029 dev_err(rdev->dev, "failed initializing VCE (%d).\n", r);
7030 return;
7031 }
7032}
7033
7034static int si_startup(struct radeon_device *rdev)
7035{
7036 struct radeon_ring *ring;
7037 int r;
7038
7039 /* enable pcie gen2/3 link */
7040 si_pcie_gen3_enable(rdev);
7041 /* enable aspm */
7042 si_program_aspm(rdev);
7043
7044 /* scratch needs to be initialized before MC */
7045 r = r600_vram_scratch_init(rdev);
7046 if (r)
7047 return r;
7048
7049 si_mc_program(rdev);
7050
7051 if (!rdev->pm.dpm_enabled) {
7052 r = si_mc_load_microcode(rdev);
7053 if (r) {
7054 DRM_ERROR("Failed to load MC firmware!\n");
7055 return r;
7056 }
7057 }
7058
7059 r = si_pcie_gart_enable(rdev);
7060 if (r)
7061 return r;
7062 si_gpu_init(rdev);
7063
7064 /* allocate rlc buffers */
7065 if (rdev->family == CHIP_VERDE) {
7066 rdev->rlc.reg_list = verde_rlc_save_restore_register_list;
7067 rdev->rlc.reg_list_size =
7068 (u32)ARRAY_SIZE(verde_rlc_save_restore_register_list);
7069 }
7070 rdev->rlc.cs_data = si_cs_data;
7071 r = sumo_rlc_init(rdev);
7072 if (r) {
7073 DRM_ERROR("Failed to init rlc BOs!\n");
7074 return r;
7075 }
7076
7077 /* allocate wb buffer */
7078 r = radeon_wb_init(rdev);
7079 if (r)
7080 return r;
7081
7082 r = radeon_fence_driver_start_ring(rdev, RADEON_RING_TYPE_GFX_INDEX);
7083 if (r) {
7084 dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r);
7085 return r;
7086 }
7087
7088 r = radeon_fence_driver_start_ring(rdev, CAYMAN_RING_TYPE_CP1_INDEX);
7089 if (r) {
7090 dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r);
7091 return r;
7092 }
7093
7094 r = radeon_fence_driver_start_ring(rdev, CAYMAN_RING_TYPE_CP2_INDEX);
7095 if (r) {
7096 dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r);
7097 return r;
7098 }
7099
7100 r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_DMA_INDEX);
7101 if (r) {
7102 dev_err(rdev->dev, "failed initializing DMA fences (%d).\n", r);
7103 return r;
7104 }
7105
7106 r = radeon_fence_driver_start_ring(rdev, CAYMAN_RING_TYPE_DMA1_INDEX);
7107 if (r) {
7108 dev_err(rdev->dev, "failed initializing DMA fences (%d).\n", r);
7109 return r;
7110 }
7111
7112 si_uvd_start(rdev);
7113 si_vce_start(rdev);
7114
7115 /* Enable IRQ */
7116 if (!rdev->irq.installed) {
7117 r = radeon_irq_kms_init(rdev);
7118 if (r)
7119 return r;
7120 }
7121
7122 r = si_irq_init(rdev);
7123 if (r) {
7124 DRM_ERROR("radeon: IH init failed (%d).\n", r);
7125 radeon_irq_kms_fini(rdev);
7126 return r;
7127 }
7128 si_irq_set(rdev);
7129
7130 ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
7131 r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET,
7132 RADEON_CP_PACKET2);
7133 if (r)
7134 return r;
7135
7136 ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX];
7137 r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP1_RPTR_OFFSET,
7138 RADEON_CP_PACKET2);
7139 if (r)
7140 return r;
7141
7142 ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX];
7143 r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP2_RPTR_OFFSET,
7144 RADEON_CP_PACKET2);
7145 if (r)
7146 return r;
7147
7148 ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX];
7149 r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET,
7150 DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0));
7151 if (r)
7152 return r;
7153
7154 ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX];
7155 r = radeon_ring_init(rdev, ring, ring->ring_size, CAYMAN_WB_DMA1_RPTR_OFFSET,
7156 DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0, 0));
7157 if (r)
7158 return r;
7159
7160 r = si_cp_load_microcode(rdev);
7161 if (r)
7162 return r;
7163 r = si_cp_resume(rdev);
7164 if (r)
7165 return r;
7166
7167 r = cayman_dma_resume(rdev);
7168 if (r)
7169 return r;
7170
7171 si_uvd_resume(rdev);
7172 si_vce_resume(rdev);
7173
7174 r = radeon_ib_pool_init(rdev);
7175 if (r) {
7176 dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
7177 return r;
7178 }
7179
7180 r = radeon_vm_manager_init(rdev);
7181 if (r) {
7182 dev_err(rdev->dev, "vm manager initialization failed (%d).\n", r);
7183 return r;
7184 }
7185
7186 r = radeon_audio_init(rdev);
7187 if (r)
7188 return r;
7189
7190 return 0;
7191}
7192
7193int si_resume(struct radeon_device *rdev)
7194{
7195 int r;
7196
7197 /* Do not reset GPU before posting, on rv770 hw unlike on r500 hw,
7198 * posting will perform necessary task to bring back GPU into good
7199 * shape.
7200 */
7201 /* post card */
7202 atom_asic_init(rdev->mode_info.atom_context);
7203
7204 /* init golden registers */
7205 si_init_golden_registers(rdev);
7206
7207 if (rdev->pm.pm_method == PM_METHOD_DPM)
7208 radeon_pm_resume(rdev);
7209
7210 rdev->accel_working = true;
7211 r = si_startup(rdev);
7212 if (r) {
7213 DRM_ERROR("si startup failed on resume\n");
7214 rdev->accel_working = false;
7215 return r;
7216 }
7217
7218 return r;
7219
7220}
7221
7222int si_suspend(struct radeon_device *rdev)
7223{
7224 radeon_pm_suspend(rdev);
7225 radeon_audio_fini(rdev);
7226 radeon_vm_manager_fini(rdev);
7227 si_cp_enable(rdev, false);
7228 cayman_dma_stop(rdev);
7229 if (rdev->has_uvd) {
7230 uvd_v1_0_fini(rdev);
7231 radeon_uvd_suspend(rdev);
7232 }
7233 if (rdev->has_vce)
7234 radeon_vce_suspend(rdev);
7235 si_fini_pg(rdev);
7236 si_fini_cg(rdev);
7237 si_irq_suspend(rdev);
7238 radeon_wb_disable(rdev);
7239 si_pcie_gart_disable(rdev);
7240 return 0;
7241}
7242
7243/* Plan is to move initialization in that function and use
7244 * helper function so that radeon_device_init pretty much
7245 * do nothing more than calling asic specific function. This
7246 * should also allow to remove a bunch of callback function
7247 * like vram_info.
7248 */
7249int si_init(struct radeon_device *rdev)
7250{
7251 struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
7252 int r;
7253
7254 /* Read BIOS */
7255 if (!radeon_get_bios(rdev)) {
7256 if (ASIC_IS_AVIVO(rdev))
7257 return -EINVAL;
7258 }
7259 /* Must be an ATOMBIOS */
7260 if (!rdev->is_atom_bios) {
7261 dev_err(rdev->dev, "Expecting atombios for cayman GPU\n");
7262 return -EINVAL;
7263 }
7264 r = radeon_atombios_init(rdev);
7265 if (r)
7266 return r;
7267
7268 /* Post card if necessary */
7269 if (!radeon_card_posted(rdev)) {
7270 if (!rdev->bios) {
7271 dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n");
7272 return -EINVAL;
7273 }
7274 DRM_INFO("GPU not posted. posting now...\n");
7275 atom_asic_init(rdev->mode_info.atom_context);
7276 }
7277 /* init golden registers */
7278 si_init_golden_registers(rdev);
7279 /* Initialize scratch registers */
7280 si_scratch_init(rdev);
7281 /* Initialize surface registers */
7282 radeon_surface_init(rdev);
7283 /* Initialize clocks */
7284 radeon_get_clock_info(rdev->ddev);
7285
7286 /* Fence driver */
7287 r = radeon_fence_driver_init(rdev);
7288 if (r)
7289 return r;
7290
7291 /* initialize memory controller */
7292 r = si_mc_init(rdev);
7293 if (r)
7294 return r;
7295 /* Memory manager */
7296 r = radeon_bo_init(rdev);
7297 if (r)
7298 return r;
7299
7300 if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw ||
7301 !rdev->rlc_fw || !rdev->mc_fw) {
7302 r = si_init_microcode(rdev);
7303 if (r) {
7304 DRM_ERROR("Failed to load firmware!\n");
7305 return r;
7306 }
7307 }
7308
7309 /* Initialize power management */
7310 radeon_pm_init(rdev);
7311
7312 ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX];
7313 ring->ring_obj = NULL;
7314 r600_ring_init(rdev, ring, 1024 * 1024);
7315
7316 ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX];
7317 ring->ring_obj = NULL;
7318 r600_ring_init(rdev, ring, 1024 * 1024);
7319
7320 ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX];
7321 ring->ring_obj = NULL;
7322 r600_ring_init(rdev, ring, 1024 * 1024);
7323
7324 ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX];
7325 ring->ring_obj = NULL;
7326 r600_ring_init(rdev, ring, 64 * 1024);
7327
7328 ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX];
7329 ring->ring_obj = NULL;
7330 r600_ring_init(rdev, ring, 64 * 1024);
7331
7332 si_uvd_init(rdev);
7333 si_vce_init(rdev);
7334
7335 rdev->ih.ring_obj = NULL;
7336 r600_ih_ring_init(rdev, 64 * 1024);
7337
7338 r = r600_pcie_gart_init(rdev);
7339 if (r)
7340 return r;
7341
7342 rdev->accel_working = true;
7343 r = si_startup(rdev);
7344 if (r) {
7345 dev_err(rdev->dev, "disabling GPU acceleration\n");
7346 si_cp_fini(rdev);
7347 cayman_dma_fini(rdev);
7348 si_irq_fini(rdev);
7349 sumo_rlc_fini(rdev);
7350 radeon_wb_fini(rdev);
7351 radeon_ib_pool_fini(rdev);
7352 radeon_vm_manager_fini(rdev);
7353 radeon_irq_kms_fini(rdev);
7354 si_pcie_gart_fini(rdev);
7355 rdev->accel_working = false;
7356 }
7357
7358 /* Don't start up if the MC ucode is missing.
7359 * The default clocks and voltages before the MC ucode
7360 * is loaded are not suffient for advanced operations.
7361 */
7362 if (!rdev->mc_fw) {
7363 DRM_ERROR("radeon: MC ucode required for NI+.\n");
7364 return -EINVAL;
7365 }
7366
7367 return 0;
7368}
7369
7370void si_fini(struct radeon_device *rdev)
7371{
7372 radeon_pm_fini(rdev);
7373 si_cp_fini(rdev);
7374 cayman_dma_fini(rdev);
7375 si_fini_pg(rdev);
7376 si_fini_cg(rdev);
7377 si_irq_fini(rdev);
7378 sumo_rlc_fini(rdev);
7379 radeon_wb_fini(rdev);
7380 radeon_vm_manager_fini(rdev);
7381 radeon_ib_pool_fini(rdev);
7382 radeon_irq_kms_fini(rdev);
7383 if (rdev->has_uvd) {
7384 uvd_v1_0_fini(rdev);
7385 radeon_uvd_fini(rdev);
7386 }
7387 if (rdev->has_vce)
7388 radeon_vce_fini(rdev);
7389 si_pcie_gart_fini(rdev);
7390 r600_vram_scratch_fini(rdev);
7391 radeon_gem_fini(rdev);
7392 radeon_fence_driver_fini(rdev);
7393 radeon_bo_fini(rdev);
7394 radeon_atombios_fini(rdev);
7395 kfree(rdev->bios);
7396 rdev->bios = NULL;
7397}
7398
7399/**
7400 * si_get_gpu_clock_counter - return GPU clock counter snapshot
7401 *
7402 * @rdev: radeon_device pointer
7403 *
7404 * Fetches a GPU clock counter snapshot (SI).
7405 * Returns the 64 bit clock counter snapshot.
7406 */
7407uint64_t si_get_gpu_clock_counter(struct radeon_device *rdev)
7408{
7409 uint64_t clock;
7410
7411 mutex_lock(&rdev->gpu_clock_mutex);
7412 WREG32(RLC_CAPTURE_GPU_CLOCK_COUNT, 1);
7413 clock = (uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_LSB) |
7414 ((uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
7415 mutex_unlock(&rdev->gpu_clock_mutex);
7416 return clock;
7417}
7418
7419int si_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk)
7420{
7421 unsigned fb_div = 0, vclk_div = 0, dclk_div = 0;
7422 int r;
7423
7424 /* bypass vclk and dclk with bclk */
7425 WREG32_P(CG_UPLL_FUNC_CNTL_2,
7426 VCLK_SRC_SEL(1) | DCLK_SRC_SEL(1),
7427 ~(VCLK_SRC_SEL_MASK | DCLK_SRC_SEL_MASK));
7428
7429 /* put PLL in bypass mode */
7430 WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_BYPASS_EN_MASK, ~UPLL_BYPASS_EN_MASK);
7431
7432 if (!vclk || !dclk) {
7433 /* keep the Bypass mode */
7434 return 0;
7435 }
7436
7437 r = radeon_uvd_calc_upll_dividers(rdev, vclk, dclk, 125000, 250000,
7438 16384, 0x03FFFFFF, 0, 128, 5,
7439 &fb_div, &vclk_div, &dclk_div);
7440 if (r)
7441 return r;
7442
7443 /* set RESET_ANTI_MUX to 0 */
7444 WREG32_P(CG_UPLL_FUNC_CNTL_5, 0, ~RESET_ANTI_MUX_MASK);
7445
7446 /* set VCO_MODE to 1 */
7447 WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_VCO_MODE_MASK, ~UPLL_VCO_MODE_MASK);
7448
7449 /* disable sleep mode */
7450 WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_SLEEP_MASK);
7451
7452 /* deassert UPLL_RESET */
7453 WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_RESET_MASK);
7454
7455 mdelay(1);
7456
7457 r = radeon_uvd_send_upll_ctlreq(rdev, CG_UPLL_FUNC_CNTL);
7458 if (r)
7459 return r;
7460
7461 /* assert UPLL_RESET again */
7462 WREG32_P(CG_UPLL_FUNC_CNTL, UPLL_RESET_MASK, ~UPLL_RESET_MASK);
7463
7464 /* disable spread spectrum. */
7465 WREG32_P(CG_UPLL_SPREAD_SPECTRUM, 0, ~SSEN_MASK);
7466
7467 /* set feedback divider */
7468 WREG32_P(CG_UPLL_FUNC_CNTL_3, UPLL_FB_DIV(fb_div), ~UPLL_FB_DIV_MASK);
7469
7470 /* set ref divider to 0 */
7471 WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_REF_DIV_MASK);
7472
7473 if (fb_div < 307200)
7474 WREG32_P(CG_UPLL_FUNC_CNTL_4, 0, ~UPLL_SPARE_ISPARE9);
7475 else
7476 WREG32_P(CG_UPLL_FUNC_CNTL_4, UPLL_SPARE_ISPARE9, ~UPLL_SPARE_ISPARE9);
7477
7478 /* set PDIV_A and PDIV_B */
7479 WREG32_P(CG_UPLL_FUNC_CNTL_2,
7480 UPLL_PDIV_A(vclk_div) | UPLL_PDIV_B(dclk_div),
7481 ~(UPLL_PDIV_A_MASK | UPLL_PDIV_B_MASK));
7482
7483 /* give the PLL some time to settle */
7484 mdelay(15);
7485
7486 /* deassert PLL_RESET */
7487 WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_RESET_MASK);
7488
7489 mdelay(15);
7490
7491 /* switch from bypass mode to normal mode */
7492 WREG32_P(CG_UPLL_FUNC_CNTL, 0, ~UPLL_BYPASS_EN_MASK);
7493
7494 r = radeon_uvd_send_upll_ctlreq(rdev, CG_UPLL_FUNC_CNTL);
7495 if (r)
7496 return r;
7497
7498 /* switch VCLK and DCLK selection */
7499 WREG32_P(CG_UPLL_FUNC_CNTL_2,
7500 VCLK_SRC_SEL(2) | DCLK_SRC_SEL(2),
7501 ~(VCLK_SRC_SEL_MASK | DCLK_SRC_SEL_MASK));
7502
7503 mdelay(100);
7504
7505 return 0;
7506}
7507
7508static void si_pcie_gen3_enable(struct radeon_device *rdev)
7509{
7510 struct pci_dev *root = rdev->pdev->bus->self;
7511 int bridge_pos, gpu_pos;
7512 u32 speed_cntl, mask, current_data_rate;
7513 int ret, i;
7514 u16 tmp16;
7515
7516 if (pci_is_root_bus(rdev->pdev->bus))
7517 return;
7518
7519 if (radeon_pcie_gen2 == 0)
7520 return;
7521
7522 if (rdev->flags & RADEON_IS_IGP)
7523 return;
7524
7525 if (!(rdev->flags & RADEON_IS_PCIE))
7526 return;
7527
7528 ret = drm_pcie_get_speed_cap_mask(rdev->ddev, &mask);
7529 if (ret != 0)
7530 return;
7531
7532 if (!(mask & (DRM_PCIE_SPEED_50 | DRM_PCIE_SPEED_80)))
7533 return;
7534
7535 speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL);
7536 current_data_rate = (speed_cntl & LC_CURRENT_DATA_RATE_MASK) >>
7537 LC_CURRENT_DATA_RATE_SHIFT;
7538 if (mask & DRM_PCIE_SPEED_80) {
7539 if (current_data_rate == 2) {
7540 DRM_INFO("PCIE gen 3 link speeds already enabled\n");
7541 return;
7542 }
7543 DRM_INFO("enabling PCIE gen 3 link speeds, disable with radeon.pcie_gen2=0\n");
7544 } else if (mask & DRM_PCIE_SPEED_50) {
7545 if (current_data_rate == 1) {
7546 DRM_INFO("PCIE gen 2 link speeds already enabled\n");
7547 return;
7548 }
7549 DRM_INFO("enabling PCIE gen 2 link speeds, disable with radeon.pcie_gen2=0\n");
7550 }
7551
7552 bridge_pos = pci_pcie_cap(root);
7553 if (!bridge_pos)
7554 return;
7555
7556 gpu_pos = pci_pcie_cap(rdev->pdev);
7557 if (!gpu_pos)
7558 return;
7559
7560 if (mask & DRM_PCIE_SPEED_80) {
7561 /* re-try equalization if gen3 is not already enabled */
7562 if (current_data_rate != 2) {
7563 u16 bridge_cfg, gpu_cfg;
7564 u16 bridge_cfg2, gpu_cfg2;
7565 u32 max_lw, current_lw, tmp;
7566
7567 pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL, &bridge_cfg);
7568 pci_read_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL, &gpu_cfg);
7569
7570 tmp16 = bridge_cfg | PCI_EXP_LNKCTL_HAWD;
7571 pci_write_config_word(root, bridge_pos + PCI_EXP_LNKCTL, tmp16);
7572
7573 tmp16 = gpu_cfg | PCI_EXP_LNKCTL_HAWD;
7574 pci_write_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL, tmp16);
7575
7576 tmp = RREG32_PCIE(PCIE_LC_STATUS1);
7577 max_lw = (tmp & LC_DETECTED_LINK_WIDTH_MASK) >> LC_DETECTED_LINK_WIDTH_SHIFT;
7578 current_lw = (tmp & LC_OPERATING_LINK_WIDTH_MASK) >> LC_OPERATING_LINK_WIDTH_SHIFT;
7579
7580 if (current_lw < max_lw) {
7581 tmp = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL);
7582 if (tmp & LC_RENEGOTIATION_SUPPORT) {
7583 tmp &= ~(LC_LINK_WIDTH_MASK | LC_UPCONFIGURE_DIS);
7584 tmp |= (max_lw << LC_LINK_WIDTH_SHIFT);
7585 tmp |= LC_UPCONFIGURE_SUPPORT | LC_RENEGOTIATE_EN | LC_RECONFIG_NOW;
7586 WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, tmp);
7587 }
7588 }
7589
7590 for (i = 0; i < 10; i++) {
7591 /* check status */
7592 pci_read_config_word(rdev->pdev, gpu_pos + PCI_EXP_DEVSTA, &tmp16);
7593 if (tmp16 & PCI_EXP_DEVSTA_TRPND)
7594 break;
7595
7596 pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL, &bridge_cfg);
7597 pci_read_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL, &gpu_cfg);
7598
7599 pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL2, &bridge_cfg2);
7600 pci_read_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL2, &gpu_cfg2);
7601
7602 tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4);
7603 tmp |= LC_SET_QUIESCE;
7604 WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp);
7605
7606 tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4);
7607 tmp |= LC_REDO_EQ;
7608 WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp);
7609
7610 mdelay(100);
7611
7612 /* linkctl */
7613 pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL, &tmp16);
7614 tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
7615 tmp16 |= (bridge_cfg & PCI_EXP_LNKCTL_HAWD);
7616 pci_write_config_word(root, bridge_pos + PCI_EXP_LNKCTL, tmp16);
7617
7618 pci_read_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL, &tmp16);
7619 tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
7620 tmp16 |= (gpu_cfg & PCI_EXP_LNKCTL_HAWD);
7621 pci_write_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL, tmp16);
7622
7623 /* linkctl2 */
7624 pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL2, &tmp16);
7625 tmp16 &= ~((1 << 4) | (7 << 9));
7626 tmp16 |= (bridge_cfg2 & ((1 << 4) | (7 << 9)));
7627 pci_write_config_word(root, bridge_pos + PCI_EXP_LNKCTL2, tmp16);
7628
7629 pci_read_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL2, &tmp16);
7630 tmp16 &= ~((1 << 4) | (7 << 9));
7631 tmp16 |= (gpu_cfg2 & ((1 << 4) | (7 << 9)));
7632 pci_write_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL2, tmp16);
7633
7634 tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4);
7635 tmp &= ~LC_SET_QUIESCE;
7636 WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp);
7637 }
7638 }
7639 }
7640
7641 /* set the link speed */
7642 speed_cntl |= LC_FORCE_EN_SW_SPEED_CHANGE | LC_FORCE_DIS_HW_SPEED_CHANGE;
7643 speed_cntl &= ~LC_FORCE_DIS_SW_SPEED_CHANGE;
7644 WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl);
7645
7646 pci_read_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL2, &tmp16);
7647 tmp16 &= ~0xf;
7648 if (mask & DRM_PCIE_SPEED_80)
7649 tmp16 |= 3; /* gen3 */
7650 else if (mask & DRM_PCIE_SPEED_50)
7651 tmp16 |= 2; /* gen2 */
7652 else
7653 tmp16 |= 1; /* gen1 */
7654 pci_write_config_word(rdev->pdev, gpu_pos + PCI_EXP_LNKCTL2, tmp16);
7655
7656 speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL);
7657 speed_cntl |= LC_INITIATE_LINK_SPEED_CHANGE;
7658 WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl);
7659
7660 for (i = 0; i < rdev->usec_timeout; i++) {
7661 speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL);
7662 if ((speed_cntl & LC_INITIATE_LINK_SPEED_CHANGE) == 0)
7663 break;
7664 udelay(1);
7665 }
7666}
7667
7668static void si_program_aspm(struct radeon_device *rdev)
7669{
7670 u32 data, orig;
7671 bool disable_l0s = false, disable_l1 = false, disable_plloff_in_l1 = false;
7672 bool disable_clkreq = false;
7673
7674 if (radeon_aspm == 0)
7675 return;
7676
7677 if (!(rdev->flags & RADEON_IS_PCIE))
7678 return;
7679
7680 orig = data = RREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL);
7681 data &= ~LC_XMIT_N_FTS_MASK;
7682 data |= LC_XMIT_N_FTS(0x24) | LC_XMIT_N_FTS_OVERRIDE_EN;
7683 if (orig != data)
7684 WREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL, data);
7685
7686 orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL3);
7687 data |= LC_GO_TO_RECOVERY;
7688 if (orig != data)
7689 WREG32_PCIE_PORT(PCIE_LC_CNTL3, data);
7690
7691 orig = data = RREG32_PCIE(PCIE_P_CNTL);
7692 data |= P_IGNORE_EDB_ERR;
7693 if (orig != data)
7694 WREG32_PCIE(PCIE_P_CNTL, data);
7695
7696 orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL);
7697 data &= ~(LC_L0S_INACTIVITY_MASK | LC_L1_INACTIVITY_MASK);
7698 data |= LC_PMI_TO_L1_DIS;
7699 if (!disable_l0s)
7700 data |= LC_L0S_INACTIVITY(7);
7701
7702 if (!disable_l1) {
7703 data |= LC_L1_INACTIVITY(7);
7704 data &= ~LC_PMI_TO_L1_DIS;
7705 if (orig != data)
7706 WREG32_PCIE_PORT(PCIE_LC_CNTL, data);
7707
7708 if (!disable_plloff_in_l1) {
7709 bool clk_req_support;
7710
7711 orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0);
7712 data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK);
7713 data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7);
7714 if (orig != data)
7715 WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0, data);
7716
7717 orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1);
7718 data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK);
7719 data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7);
7720 if (orig != data)
7721 WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1, data);
7722
7723 orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0);
7724 data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK);
7725 data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7);
7726 if (orig != data)
7727 WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0, data);
7728
7729 orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1);
7730 data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK);
7731 data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7);
7732 if (orig != data)
7733 WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1, data);
7734
7735 if ((rdev->family != CHIP_OLAND) && (rdev->family != CHIP_HAINAN)) {
7736 orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0);
7737 data &= ~PLL_RAMP_UP_TIME_0_MASK;
7738 if (orig != data)
7739 WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0, data);
7740
7741 orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1);
7742 data &= ~PLL_RAMP_UP_TIME_1_MASK;
7743 if (orig != data)
7744 WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1, data);
7745
7746 orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_2);
7747 data &= ~PLL_RAMP_UP_TIME_2_MASK;
7748 if (orig != data)
7749 WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_2, data);
7750
7751 orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_3);
7752 data &= ~PLL_RAMP_UP_TIME_3_MASK;
7753 if (orig != data)
7754 WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_3, data);
7755
7756 orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0);
7757 data &= ~PLL_RAMP_UP_TIME_0_MASK;
7758 if (orig != data)
7759 WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0, data);
7760
7761 orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1);
7762 data &= ~PLL_RAMP_UP_TIME_1_MASK;
7763 if (orig != data)
7764 WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1, data);
7765
7766 orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_2);
7767 data &= ~PLL_RAMP_UP_TIME_2_MASK;
7768 if (orig != data)
7769 WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_2, data);
7770
7771 orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_3);
7772 data &= ~PLL_RAMP_UP_TIME_3_MASK;
7773 if (orig != data)
7774 WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_3, data);
7775 }
7776 orig = data = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL);
7777 data &= ~LC_DYN_LANES_PWR_STATE_MASK;
7778 data |= LC_DYN_LANES_PWR_STATE(3);
7779 if (orig != data)
7780 WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, data);
7781
7782 orig = data = RREG32_PIF_PHY0(PB0_PIF_CNTL);
7783 data &= ~LS2_EXIT_TIME_MASK;
7784 if ((rdev->family == CHIP_OLAND) || (rdev->family == CHIP_HAINAN))
7785 data |= LS2_EXIT_TIME(5);
7786 if (orig != data)
7787 WREG32_PIF_PHY0(PB0_PIF_CNTL, data);
7788
7789 orig = data = RREG32_PIF_PHY1(PB1_PIF_CNTL);
7790 data &= ~LS2_EXIT_TIME_MASK;
7791 if ((rdev->family == CHIP_OLAND) || (rdev->family == CHIP_HAINAN))
7792 data |= LS2_EXIT_TIME(5);
7793 if (orig != data)
7794 WREG32_PIF_PHY1(PB1_PIF_CNTL, data);
7795
7796 if (!disable_clkreq &&
7797 !pci_is_root_bus(rdev->pdev->bus)) {
7798 struct pci_dev *root = rdev->pdev->bus->self;
7799 u32 lnkcap;
7800
7801 clk_req_support = false;
7802 pcie_capability_read_dword(root, PCI_EXP_LNKCAP, &lnkcap);
7803 if (lnkcap & PCI_EXP_LNKCAP_CLKPM)
7804 clk_req_support = true;
7805 } else {
7806 clk_req_support = false;
7807 }
7808
7809 if (clk_req_support) {
7810 orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL2);
7811 data |= LC_ALLOW_PDWN_IN_L1 | LC_ALLOW_PDWN_IN_L23;
7812 if (orig != data)
7813 WREG32_PCIE_PORT(PCIE_LC_CNTL2, data);
7814
7815 orig = data = RREG32(THM_CLK_CNTL);
7816 data &= ~(CMON_CLK_SEL_MASK | TMON_CLK_SEL_MASK);
7817 data |= CMON_CLK_SEL(1) | TMON_CLK_SEL(1);
7818 if (orig != data)
7819 WREG32(THM_CLK_CNTL, data);
7820
7821 orig = data = RREG32(MISC_CLK_CNTL);
7822 data &= ~(DEEP_SLEEP_CLK_SEL_MASK | ZCLK_SEL_MASK);
7823 data |= DEEP_SLEEP_CLK_SEL(1) | ZCLK_SEL(1);
7824 if (orig != data)
7825 WREG32(MISC_CLK_CNTL, data);
7826
7827 orig = data = RREG32(CG_CLKPIN_CNTL);
7828 data &= ~BCLK_AS_XCLK;
7829 if (orig != data)
7830 WREG32(CG_CLKPIN_CNTL, data);
7831
7832 orig = data = RREG32(CG_CLKPIN_CNTL_2);
7833 data &= ~FORCE_BIF_REFCLK_EN;
7834 if (orig != data)
7835 WREG32(CG_CLKPIN_CNTL_2, data);
7836
7837 orig = data = RREG32(MPLL_BYPASSCLK_SEL);
7838 data &= ~MPLL_CLKOUT_SEL_MASK;
7839 data |= MPLL_CLKOUT_SEL(4);
7840 if (orig != data)
7841 WREG32(MPLL_BYPASSCLK_SEL, data);
7842
7843 orig = data = RREG32(SPLL_CNTL_MODE);
7844 data &= ~SPLL_REFCLK_SEL_MASK;
7845 if (orig != data)
7846 WREG32(SPLL_CNTL_MODE, data);
7847 }
7848 }
7849 } else {
7850 if (orig != data)
7851 WREG32_PCIE_PORT(PCIE_LC_CNTL, data);
7852 }
7853
7854 orig = data = RREG32_PCIE(PCIE_CNTL2);
7855 data |= SLV_MEM_LS_EN | MST_MEM_LS_EN | REPLAY_MEM_LS_EN;
7856 if (orig != data)
7857 WREG32_PCIE(PCIE_CNTL2, data);
7858
7859 if (!disable_l0s) {
7860 data = RREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL);
7861 if((data & LC_N_FTS_MASK) == LC_N_FTS_MASK) {
7862 data = RREG32_PCIE(PCIE_LC_STATUS1);
7863 if ((data & LC_REVERSE_XMIT) && (data & LC_REVERSE_RCVR)) {
7864 orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL);
7865 data &= ~LC_L0S_INACTIVITY_MASK;
7866 if (orig != data)
7867 WREG32_PCIE_PORT(PCIE_LC_CNTL, data);
7868 }
7869 }
7870 }
7871}
7872
7873static int si_vce_send_vcepll_ctlreq(struct radeon_device *rdev)
7874{
7875 unsigned i;
7876
7877 /* make sure VCEPLL_CTLREQ is deasserted */
7878 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~UPLL_CTLREQ_MASK);
7879
7880 mdelay(10);
7881
7882 /* assert UPLL_CTLREQ */
7883 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, UPLL_CTLREQ_MASK, ~UPLL_CTLREQ_MASK);
7884
7885 /* wait for CTLACK and CTLACK2 to get asserted */
7886 for (i = 0; i < 100; ++i) {
7887 uint32_t mask = UPLL_CTLACK_MASK | UPLL_CTLACK2_MASK;
7888 if ((RREG32_SMC(CG_VCEPLL_FUNC_CNTL) & mask) == mask)
7889 break;
7890 mdelay(10);
7891 }
7892
7893 /* deassert UPLL_CTLREQ */
7894 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~UPLL_CTLREQ_MASK);
7895
7896 if (i == 100) {
7897 DRM_ERROR("Timeout setting UVD clocks!\n");
7898 return -ETIMEDOUT;
7899 }
7900
7901 return 0;
7902}
7903
7904int si_set_vce_clocks(struct radeon_device *rdev, u32 evclk, u32 ecclk)
7905{
7906 unsigned fb_div = 0, evclk_div = 0, ecclk_div = 0;
7907 int r;
7908
7909 /* bypass evclk and ecclk with bclk */
7910 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL_2,
7911 EVCLK_SRC_SEL(1) | ECCLK_SRC_SEL(1),
7912 ~(EVCLK_SRC_SEL_MASK | ECCLK_SRC_SEL_MASK));
7913
7914 /* put PLL in bypass mode */
7915 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, VCEPLL_BYPASS_EN_MASK,
7916 ~VCEPLL_BYPASS_EN_MASK);
7917
7918 if (!evclk || !ecclk) {
7919 /* keep the Bypass mode, put PLL to sleep */
7920 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, VCEPLL_SLEEP_MASK,
7921 ~VCEPLL_SLEEP_MASK);
7922 return 0;
7923 }
7924
7925 r = radeon_uvd_calc_upll_dividers(rdev, evclk, ecclk, 125000, 250000,
7926 16384, 0x03FFFFFF, 0, 128, 5,
7927 &fb_div, &evclk_div, &ecclk_div);
7928 if (r)
7929 return r;
7930
7931 /* set RESET_ANTI_MUX to 0 */
7932 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL_5, 0, ~RESET_ANTI_MUX_MASK);
7933
7934 /* set VCO_MODE to 1 */
7935 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, VCEPLL_VCO_MODE_MASK,
7936 ~VCEPLL_VCO_MODE_MASK);
7937
7938 /* toggle VCEPLL_SLEEP to 1 then back to 0 */
7939 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, VCEPLL_SLEEP_MASK,
7940 ~VCEPLL_SLEEP_MASK);
7941 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~VCEPLL_SLEEP_MASK);
7942
7943 /* deassert VCEPLL_RESET */
7944 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~VCEPLL_RESET_MASK);
7945
7946 mdelay(1);
7947
7948 r = si_vce_send_vcepll_ctlreq(rdev);
7949 if (r)
7950 return r;
7951
7952 /* assert VCEPLL_RESET again */
7953 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, VCEPLL_RESET_MASK, ~VCEPLL_RESET_MASK);
7954
7955 /* disable spread spectrum. */
7956 WREG32_SMC_P(CG_VCEPLL_SPREAD_SPECTRUM, 0, ~SSEN_MASK);
7957
7958 /* set feedback divider */
7959 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL_3, VCEPLL_FB_DIV(fb_div), ~VCEPLL_FB_DIV_MASK);
7960
7961 /* set ref divider to 0 */
7962 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~VCEPLL_REF_DIV_MASK);
7963
7964 /* set PDIV_A and PDIV_B */
7965 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL_2,
7966 VCEPLL_PDIV_A(evclk_div) | VCEPLL_PDIV_B(ecclk_div),
7967 ~(VCEPLL_PDIV_A_MASK | VCEPLL_PDIV_B_MASK));
7968
7969 /* give the PLL some time to settle */
7970 mdelay(15);
7971
7972 /* deassert PLL_RESET */
7973 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~VCEPLL_RESET_MASK);
7974
7975 mdelay(15);
7976
7977 /* switch from bypass mode to normal mode */
7978 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL, 0, ~VCEPLL_BYPASS_EN_MASK);
7979
7980 r = si_vce_send_vcepll_ctlreq(rdev);
7981 if (r)
7982 return r;
7983
7984 /* switch VCLK and DCLK selection */
7985 WREG32_SMC_P(CG_VCEPLL_FUNC_CNTL_2,
7986 EVCLK_SRC_SEL(16) | ECCLK_SRC_SEL(16),
7987 ~(EVCLK_SRC_SEL_MASK | ECCLK_SRC_SEL_MASK));
7988
7989 mdelay(100);
7990
7991 return 0;
7992}