]>
Commit | Line | Data |
---|---|---|
771fe6b9 JG |
1 | /* |
2 | * Copyright 2008 Advanced Micro Devices, Inc. | |
3 | * Copyright 2008 Red Hat Inc. | |
4 | * Copyright 2009 Jerome Glisse. | |
5 | * | |
6 | * Permission is hereby granted, free of charge, to any person obtaining a | |
7 | * copy of this software and associated documentation files (the "Software"), | |
8 | * to deal in the Software without restriction, including without limitation | |
9 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, | |
10 | * and/or sell copies of the Software, and to permit persons to whom the | |
11 | * Software is furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR | |
20 | * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, | |
21 | * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR | |
22 | * OTHER DEALINGS IN THE SOFTWARE. | |
23 | * | |
24 | * Authors: Dave Airlie | |
25 | * Alex Deucher | |
26 | * Jerome Glisse | |
27 | */ | |
28 | #include <linux/seq_file.h> | |
5a0e3ad6 | 29 | #include <linux/slab.h> |
771fe6b9 JG |
30 | #include "drmP.h" |
31 | #include "drm.h" | |
32 | #include "radeon_drm.h" | |
771fe6b9 JG |
33 | #include "radeon_reg.h" |
34 | #include "radeon.h" | |
3ce0a23d | 35 | #include "r100d.h" |
d4550907 JG |
36 | #include "rs100d.h" |
37 | #include "rv200d.h" | |
38 | #include "rv250d.h" | |
3ce0a23d | 39 | |
70967ab9 BH |
40 | #include <linux/firmware.h> |
41 | #include <linux/platform_device.h> | |
42 | ||
551ebd83 DA |
43 | #include "r100_reg_safe.h" |
44 | #include "rn50_reg_safe.h" | |
45 | ||
70967ab9 BH |
46 | /* Firmware Names */ |
47 | #define FIRMWARE_R100 "radeon/R100_cp.bin" | |
48 | #define FIRMWARE_R200 "radeon/R200_cp.bin" | |
49 | #define FIRMWARE_R300 "radeon/R300_cp.bin" | |
50 | #define FIRMWARE_R420 "radeon/R420_cp.bin" | |
51 | #define FIRMWARE_RS690 "radeon/RS690_cp.bin" | |
52 | #define FIRMWARE_RS600 "radeon/RS600_cp.bin" | |
53 | #define FIRMWARE_R520 "radeon/R520_cp.bin" | |
54 | ||
55 | MODULE_FIRMWARE(FIRMWARE_R100); | |
56 | MODULE_FIRMWARE(FIRMWARE_R200); | |
57 | MODULE_FIRMWARE(FIRMWARE_R300); | |
58 | MODULE_FIRMWARE(FIRMWARE_R420); | |
59 | MODULE_FIRMWARE(FIRMWARE_RS690); | |
60 | MODULE_FIRMWARE(FIRMWARE_RS600); | |
61 | MODULE_FIRMWARE(FIRMWARE_R520); | |
771fe6b9 | 62 | |
551ebd83 DA |
63 | #include "r100_track.h" |
64 | ||
771fe6b9 JG |
65 | /* This files gather functions specifics to: |
66 | * r100,rv100,rs100,rv200,rs200,r200,rv250,rs300,rv280 | |
771fe6b9 | 67 | */ |
771fe6b9 | 68 | |
05a05c50 AD |
69 | /* hpd for digital panel detect/disconnect */ |
70 | bool r100_hpd_sense(struct radeon_device *rdev, enum radeon_hpd_id hpd) | |
71 | { | |
72 | bool connected = false; | |
73 | ||
74 | switch (hpd) { | |
75 | case RADEON_HPD_1: | |
76 | if (RREG32(RADEON_FP_GEN_CNTL) & RADEON_FP_DETECT_SENSE) | |
77 | connected = true; | |
78 | break; | |
79 | case RADEON_HPD_2: | |
80 | if (RREG32(RADEON_FP2_GEN_CNTL) & RADEON_FP2_DETECT_SENSE) | |
81 | connected = true; | |
82 | break; | |
83 | default: | |
84 | break; | |
85 | } | |
86 | return connected; | |
87 | } | |
88 | ||
89 | void r100_hpd_set_polarity(struct radeon_device *rdev, | |
90 | enum radeon_hpd_id hpd) | |
91 | { | |
92 | u32 tmp; | |
93 | bool connected = r100_hpd_sense(rdev, hpd); | |
94 | ||
95 | switch (hpd) { | |
96 | case RADEON_HPD_1: | |
97 | tmp = RREG32(RADEON_FP_GEN_CNTL); | |
98 | if (connected) | |
99 | tmp &= ~RADEON_FP_DETECT_INT_POL; | |
100 | else | |
101 | tmp |= RADEON_FP_DETECT_INT_POL; | |
102 | WREG32(RADEON_FP_GEN_CNTL, tmp); | |
103 | break; | |
104 | case RADEON_HPD_2: | |
105 | tmp = RREG32(RADEON_FP2_GEN_CNTL); | |
106 | if (connected) | |
107 | tmp &= ~RADEON_FP2_DETECT_INT_POL; | |
108 | else | |
109 | tmp |= RADEON_FP2_DETECT_INT_POL; | |
110 | WREG32(RADEON_FP2_GEN_CNTL, tmp); | |
111 | break; | |
112 | default: | |
113 | break; | |
114 | } | |
115 | } | |
116 | ||
117 | void r100_hpd_init(struct radeon_device *rdev) | |
118 | { | |
119 | struct drm_device *dev = rdev->ddev; | |
120 | struct drm_connector *connector; | |
121 | ||
122 | list_for_each_entry(connector, &dev->mode_config.connector_list, head) { | |
123 | struct radeon_connector *radeon_connector = to_radeon_connector(connector); | |
124 | switch (radeon_connector->hpd.hpd) { | |
125 | case RADEON_HPD_1: | |
126 | rdev->irq.hpd[0] = true; | |
127 | break; | |
128 | case RADEON_HPD_2: | |
129 | rdev->irq.hpd[1] = true; | |
130 | break; | |
131 | default: | |
132 | break; | |
133 | } | |
134 | } | |
003e69f9 JG |
135 | if (rdev->irq.installed) |
136 | r100_irq_set(rdev); | |
05a05c50 AD |
137 | } |
138 | ||
139 | void r100_hpd_fini(struct radeon_device *rdev) | |
140 | { | |
141 | struct drm_device *dev = rdev->ddev; | |
142 | struct drm_connector *connector; | |
143 | ||
144 | list_for_each_entry(connector, &dev->mode_config.connector_list, head) { | |
145 | struct radeon_connector *radeon_connector = to_radeon_connector(connector); | |
146 | switch (radeon_connector->hpd.hpd) { | |
147 | case RADEON_HPD_1: | |
148 | rdev->irq.hpd[0] = false; | |
149 | break; | |
150 | case RADEON_HPD_2: | |
151 | rdev->irq.hpd[1] = false; | |
152 | break; | |
153 | default: | |
154 | break; | |
155 | } | |
156 | } | |
157 | } | |
158 | ||
771fe6b9 JG |
159 | /* |
160 | * PCI GART | |
161 | */ | |
162 | void r100_pci_gart_tlb_flush(struct radeon_device *rdev) | |
163 | { | |
164 | /* TODO: can we do somethings here ? */ | |
165 | /* It seems hw only cache one entry so we should discard this | |
166 | * entry otherwise if first GPU GART read hit this entry it | |
167 | * could end up in wrong address. */ | |
168 | } | |
169 | ||
4aac0473 | 170 | int r100_pci_gart_init(struct radeon_device *rdev) |
771fe6b9 | 171 | { |
771fe6b9 JG |
172 | int r; |
173 | ||
4aac0473 JG |
174 | if (rdev->gart.table.ram.ptr) { |
175 | WARN(1, "R100 PCI GART already initialized.\n"); | |
176 | return 0; | |
177 | } | |
771fe6b9 JG |
178 | /* Initialize common gart structure */ |
179 | r = radeon_gart_init(rdev); | |
4aac0473 | 180 | if (r) |
771fe6b9 | 181 | return r; |
4aac0473 JG |
182 | rdev->gart.table_size = rdev->gart.num_gpu_pages * 4; |
183 | rdev->asic->gart_tlb_flush = &r100_pci_gart_tlb_flush; | |
184 | rdev->asic->gart_set_page = &r100_pci_gart_set_page; | |
185 | return radeon_gart_table_ram_alloc(rdev); | |
186 | } | |
187 | ||
17e15b0c DA |
188 | /* required on r1xx, r2xx, r300, r(v)350, r420/r481, rs400/rs480 */ |
189 | void r100_enable_bm(struct radeon_device *rdev) | |
190 | { | |
191 | uint32_t tmp; | |
192 | /* Enable bus mastering */ | |
193 | tmp = RREG32(RADEON_BUS_CNTL) & ~RADEON_BUS_MASTER_DIS; | |
194 | WREG32(RADEON_BUS_CNTL, tmp); | |
195 | } | |
196 | ||
4aac0473 JG |
197 | int r100_pci_gart_enable(struct radeon_device *rdev) |
198 | { | |
199 | uint32_t tmp; | |
200 | ||
82568565 | 201 | radeon_gart_restore(rdev); |
771fe6b9 JG |
202 | /* discard memory request outside of configured range */ |
203 | tmp = RREG32(RADEON_AIC_CNTL) | RADEON_DIS_OUT_OF_PCI_GART_ACCESS; | |
204 | WREG32(RADEON_AIC_CNTL, tmp); | |
205 | /* set address range for PCI address translate */ | |
d594e46a JG |
206 | WREG32(RADEON_AIC_LO_ADDR, rdev->mc.gtt_start); |
207 | WREG32(RADEON_AIC_HI_ADDR, rdev->mc.gtt_end); | |
771fe6b9 JG |
208 | /* set PCI GART page-table base address */ |
209 | WREG32(RADEON_AIC_PT_BASE, rdev->gart.table_addr); | |
210 | tmp = RREG32(RADEON_AIC_CNTL) | RADEON_PCIGART_TRANSLATE_EN; | |
211 | WREG32(RADEON_AIC_CNTL, tmp); | |
212 | r100_pci_gart_tlb_flush(rdev); | |
213 | rdev->gart.ready = true; | |
214 | return 0; | |
215 | } | |
216 | ||
217 | void r100_pci_gart_disable(struct radeon_device *rdev) | |
218 | { | |
219 | uint32_t tmp; | |
220 | ||
221 | /* discard memory request outside of configured range */ | |
222 | tmp = RREG32(RADEON_AIC_CNTL) | RADEON_DIS_OUT_OF_PCI_GART_ACCESS; | |
223 | WREG32(RADEON_AIC_CNTL, tmp & ~RADEON_PCIGART_TRANSLATE_EN); | |
224 | WREG32(RADEON_AIC_LO_ADDR, 0); | |
225 | WREG32(RADEON_AIC_HI_ADDR, 0); | |
226 | } | |
227 | ||
228 | int r100_pci_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr) | |
229 | { | |
230 | if (i < 0 || i > rdev->gart.num_gpu_pages) { | |
231 | return -EINVAL; | |
232 | } | |
ed10f95d | 233 | rdev->gart.table.ram.ptr[i] = cpu_to_le32(lower_32_bits(addr)); |
771fe6b9 JG |
234 | return 0; |
235 | } | |
236 | ||
4aac0473 | 237 | void r100_pci_gart_fini(struct radeon_device *rdev) |
771fe6b9 | 238 | { |
4aac0473 JG |
239 | r100_pci_gart_disable(rdev); |
240 | radeon_gart_table_ram_free(rdev); | |
241 | radeon_gart_fini(rdev); | |
771fe6b9 JG |
242 | } |
243 | ||
7ed220d7 MD |
244 | int r100_irq_set(struct radeon_device *rdev) |
245 | { | |
246 | uint32_t tmp = 0; | |
247 | ||
003e69f9 JG |
248 | if (!rdev->irq.installed) { |
249 | WARN(1, "Can't enable IRQ/MSI because no handler is installed.\n"); | |
250 | WREG32(R_000040_GEN_INT_CNTL, 0); | |
251 | return -EINVAL; | |
252 | } | |
7ed220d7 MD |
253 | if (rdev->irq.sw_int) { |
254 | tmp |= RADEON_SW_INT_ENABLE; | |
255 | } | |
256 | if (rdev->irq.crtc_vblank_int[0]) { | |
257 | tmp |= RADEON_CRTC_VBLANK_MASK; | |
258 | } | |
259 | if (rdev->irq.crtc_vblank_int[1]) { | |
260 | tmp |= RADEON_CRTC2_VBLANK_MASK; | |
261 | } | |
05a05c50 AD |
262 | if (rdev->irq.hpd[0]) { |
263 | tmp |= RADEON_FP_DETECT_MASK; | |
264 | } | |
265 | if (rdev->irq.hpd[1]) { | |
266 | tmp |= RADEON_FP2_DETECT_MASK; | |
267 | } | |
7ed220d7 MD |
268 | WREG32(RADEON_GEN_INT_CNTL, tmp); |
269 | return 0; | |
270 | } | |
271 | ||
9f022ddf JG |
272 | void r100_irq_disable(struct radeon_device *rdev) |
273 | { | |
274 | u32 tmp; | |
275 | ||
276 | WREG32(R_000040_GEN_INT_CNTL, 0); | |
277 | /* Wait and acknowledge irq */ | |
278 | mdelay(1); | |
279 | tmp = RREG32(R_000044_GEN_INT_STATUS); | |
280 | WREG32(R_000044_GEN_INT_STATUS, tmp); | |
281 | } | |
282 | ||
7ed220d7 MD |
283 | static inline uint32_t r100_irq_ack(struct radeon_device *rdev) |
284 | { | |
285 | uint32_t irqs = RREG32(RADEON_GEN_INT_STATUS); | |
05a05c50 AD |
286 | uint32_t irq_mask = RADEON_SW_INT_TEST | |
287 | RADEON_CRTC_VBLANK_STAT | RADEON_CRTC2_VBLANK_STAT | | |
288 | RADEON_FP_DETECT_STAT | RADEON_FP2_DETECT_STAT; | |
7ed220d7 MD |
289 | |
290 | if (irqs) { | |
291 | WREG32(RADEON_GEN_INT_STATUS, irqs); | |
292 | } | |
293 | return irqs & irq_mask; | |
294 | } | |
295 | ||
296 | int r100_irq_process(struct radeon_device *rdev) | |
297 | { | |
3e5cb98d | 298 | uint32_t status, msi_rearm; |
d4877cf2 | 299 | bool queue_hotplug = false; |
7ed220d7 MD |
300 | |
301 | status = r100_irq_ack(rdev); | |
302 | if (!status) { | |
303 | return IRQ_NONE; | |
304 | } | |
a513c184 JG |
305 | if (rdev->shutdown) { |
306 | return IRQ_NONE; | |
307 | } | |
7ed220d7 MD |
308 | while (status) { |
309 | /* SW interrupt */ | |
310 | if (status & RADEON_SW_INT_TEST) { | |
311 | radeon_fence_process(rdev); | |
312 | } | |
313 | /* Vertical blank interrupts */ | |
314 | if (status & RADEON_CRTC_VBLANK_STAT) { | |
315 | drm_handle_vblank(rdev->ddev, 0); | |
73a6d3fc | 316 | wake_up(&rdev->irq.vblank_queue); |
7ed220d7 MD |
317 | } |
318 | if (status & RADEON_CRTC2_VBLANK_STAT) { | |
319 | drm_handle_vblank(rdev->ddev, 1); | |
73a6d3fc | 320 | wake_up(&rdev->irq.vblank_queue); |
7ed220d7 | 321 | } |
05a05c50 | 322 | if (status & RADEON_FP_DETECT_STAT) { |
d4877cf2 AD |
323 | queue_hotplug = true; |
324 | DRM_DEBUG("HPD1\n"); | |
05a05c50 AD |
325 | } |
326 | if (status & RADEON_FP2_DETECT_STAT) { | |
d4877cf2 AD |
327 | queue_hotplug = true; |
328 | DRM_DEBUG("HPD2\n"); | |
05a05c50 | 329 | } |
7ed220d7 MD |
330 | status = r100_irq_ack(rdev); |
331 | } | |
d4877cf2 AD |
332 | if (queue_hotplug) |
333 | queue_work(rdev->wq, &rdev->hotplug_work); | |
3e5cb98d AD |
334 | if (rdev->msi_enabled) { |
335 | switch (rdev->family) { | |
336 | case CHIP_RS400: | |
337 | case CHIP_RS480: | |
338 | msi_rearm = RREG32(RADEON_AIC_CNTL) & ~RS400_MSI_REARM; | |
339 | WREG32(RADEON_AIC_CNTL, msi_rearm); | |
340 | WREG32(RADEON_AIC_CNTL, msi_rearm | RS400_MSI_REARM); | |
341 | break; | |
342 | default: | |
343 | msi_rearm = RREG32(RADEON_MSI_REARM_EN) & ~RV370_MSI_REARM_EN; | |
344 | WREG32(RADEON_MSI_REARM_EN, msi_rearm); | |
345 | WREG32(RADEON_MSI_REARM_EN, msi_rearm | RV370_MSI_REARM_EN); | |
346 | break; | |
347 | } | |
348 | } | |
7ed220d7 MD |
349 | return IRQ_HANDLED; |
350 | } | |
351 | ||
352 | u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc) | |
353 | { | |
354 | if (crtc == 0) | |
355 | return RREG32(RADEON_CRTC_CRNT_FRAME); | |
356 | else | |
357 | return RREG32(RADEON_CRTC2_CRNT_FRAME); | |
358 | } | |
359 | ||
9e5b2af7 PN |
360 | /* Who ever call radeon_fence_emit should call ring_lock and ask |
361 | * for enough space (today caller are ib schedule and buffer move) */ | |
771fe6b9 JG |
362 | void r100_fence_ring_emit(struct radeon_device *rdev, |
363 | struct radeon_fence *fence) | |
364 | { | |
9e5b2af7 PN |
365 | /* We have to make sure that caches are flushed before |
366 | * CPU might read something from VRAM. */ | |
367 | radeon_ring_write(rdev, PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0)); | |
368 | radeon_ring_write(rdev, RADEON_RB3D_DC_FLUSH_ALL); | |
369 | radeon_ring_write(rdev, PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0)); | |
370 | radeon_ring_write(rdev, RADEON_RB3D_ZC_FLUSH_ALL); | |
771fe6b9 | 371 | /* Wait until IDLE & CLEAN */ |
4612dc97 AD |
372 | radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0)); |
373 | radeon_ring_write(rdev, RADEON_WAIT_2D_IDLECLEAN | RADEON_WAIT_3D_IDLECLEAN); | |
cafe6609 JG |
374 | radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0)); |
375 | radeon_ring_write(rdev, rdev->config.r100.hdp_cntl | | |
376 | RADEON_HDP_READ_BUFFER_INVALIDATE); | |
377 | radeon_ring_write(rdev, PACKET0(RADEON_HOST_PATH_CNTL, 0)); | |
378 | radeon_ring_write(rdev, rdev->config.r100.hdp_cntl); | |
771fe6b9 JG |
379 | /* Emit fence sequence & fire IRQ */ |
380 | radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0)); | |
381 | radeon_ring_write(rdev, fence->seq); | |
382 | radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0)); | |
383 | radeon_ring_write(rdev, RADEON_SW_INT_FIRE); | |
384 | } | |
385 | ||
771fe6b9 JG |
386 | int r100_wb_init(struct radeon_device *rdev) |
387 | { | |
388 | int r; | |
389 | ||
390 | if (rdev->wb.wb_obj == NULL) { | |
4c788679 JG |
391 | r = radeon_bo_create(rdev, NULL, RADEON_GPU_PAGE_SIZE, true, |
392 | RADEON_GEM_DOMAIN_GTT, | |
393 | &rdev->wb.wb_obj); | |
771fe6b9 | 394 | if (r) { |
4c788679 | 395 | dev_err(rdev->dev, "(%d) create WB buffer failed\n", r); |
771fe6b9 JG |
396 | return r; |
397 | } | |
4c788679 JG |
398 | r = radeon_bo_reserve(rdev->wb.wb_obj, false); |
399 | if (unlikely(r != 0)) | |
400 | return r; | |
401 | r = radeon_bo_pin(rdev->wb.wb_obj, RADEON_GEM_DOMAIN_GTT, | |
402 | &rdev->wb.gpu_addr); | |
771fe6b9 | 403 | if (r) { |
4c788679 JG |
404 | dev_err(rdev->dev, "(%d) pin WB buffer failed\n", r); |
405 | radeon_bo_unreserve(rdev->wb.wb_obj); | |
771fe6b9 JG |
406 | return r; |
407 | } | |
4c788679 JG |
408 | r = radeon_bo_kmap(rdev->wb.wb_obj, (void **)&rdev->wb.wb); |
409 | radeon_bo_unreserve(rdev->wb.wb_obj); | |
771fe6b9 | 410 | if (r) { |
4c788679 | 411 | dev_err(rdev->dev, "(%d) map WB buffer failed\n", r); |
771fe6b9 JG |
412 | return r; |
413 | } | |
414 | } | |
9f022ddf JG |
415 | WREG32(R_000774_SCRATCH_ADDR, rdev->wb.gpu_addr); |
416 | WREG32(R_00070C_CP_RB_RPTR_ADDR, | |
417 | S_00070C_RB_RPTR_ADDR((rdev->wb.gpu_addr + 1024) >> 2)); | |
418 | WREG32(R_000770_SCRATCH_UMSK, 0xff); | |
771fe6b9 JG |
419 | return 0; |
420 | } | |
421 | ||
9f022ddf JG |
422 | void r100_wb_disable(struct radeon_device *rdev) |
423 | { | |
424 | WREG32(R_000770_SCRATCH_UMSK, 0); | |
425 | } | |
426 | ||
771fe6b9 JG |
427 | void r100_wb_fini(struct radeon_device *rdev) |
428 | { | |
4c788679 JG |
429 | int r; |
430 | ||
9f022ddf | 431 | r100_wb_disable(rdev); |
771fe6b9 | 432 | if (rdev->wb.wb_obj) { |
4c788679 JG |
433 | r = radeon_bo_reserve(rdev->wb.wb_obj, false); |
434 | if (unlikely(r != 0)) { | |
435 | dev_err(rdev->dev, "(%d) can't finish WB\n", r); | |
436 | return; | |
437 | } | |
438 | radeon_bo_kunmap(rdev->wb.wb_obj); | |
439 | radeon_bo_unpin(rdev->wb.wb_obj); | |
440 | radeon_bo_unreserve(rdev->wb.wb_obj); | |
441 | radeon_bo_unref(&rdev->wb.wb_obj); | |
771fe6b9 JG |
442 | rdev->wb.wb = NULL; |
443 | rdev->wb.wb_obj = NULL; | |
444 | } | |
445 | } | |
446 | ||
447 | int r100_copy_blit(struct radeon_device *rdev, | |
448 | uint64_t src_offset, | |
449 | uint64_t dst_offset, | |
450 | unsigned num_pages, | |
451 | struct radeon_fence *fence) | |
452 | { | |
453 | uint32_t cur_pages; | |
454 | uint32_t stride_bytes = PAGE_SIZE; | |
455 | uint32_t pitch; | |
456 | uint32_t stride_pixels; | |
457 | unsigned ndw; | |
458 | int num_loops; | |
459 | int r = 0; | |
460 | ||
461 | /* radeon limited to 16k stride */ | |
462 | stride_bytes &= 0x3fff; | |
463 | /* radeon pitch is /64 */ | |
464 | pitch = stride_bytes / 64; | |
465 | stride_pixels = stride_bytes / 4; | |
466 | num_loops = DIV_ROUND_UP(num_pages, 8191); | |
467 | ||
468 | /* Ask for enough room for blit + flush + fence */ | |
469 | ndw = 64 + (10 * num_loops); | |
470 | r = radeon_ring_lock(rdev, ndw); | |
471 | if (r) { | |
472 | DRM_ERROR("radeon: moving bo (%d) asking for %u dw.\n", r, ndw); | |
473 | return -EINVAL; | |
474 | } | |
475 | while (num_pages > 0) { | |
476 | cur_pages = num_pages; | |
477 | if (cur_pages > 8191) { | |
478 | cur_pages = 8191; | |
479 | } | |
480 | num_pages -= cur_pages; | |
481 | ||
482 | /* pages are in Y direction - height | |
483 | page width in X direction - width */ | |
484 | radeon_ring_write(rdev, PACKET3(PACKET3_BITBLT_MULTI, 8)); | |
485 | radeon_ring_write(rdev, | |
486 | RADEON_GMC_SRC_PITCH_OFFSET_CNTL | | |
487 | RADEON_GMC_DST_PITCH_OFFSET_CNTL | | |
488 | RADEON_GMC_SRC_CLIPPING | | |
489 | RADEON_GMC_DST_CLIPPING | | |
490 | RADEON_GMC_BRUSH_NONE | | |
491 | (RADEON_COLOR_FORMAT_ARGB8888 << 8) | | |
492 | RADEON_GMC_SRC_DATATYPE_COLOR | | |
493 | RADEON_ROP3_S | | |
494 | RADEON_DP_SRC_SOURCE_MEMORY | | |
495 | RADEON_GMC_CLR_CMP_CNTL_DIS | | |
496 | RADEON_GMC_WR_MSK_DIS); | |
497 | radeon_ring_write(rdev, (pitch << 22) | (src_offset >> 10)); | |
498 | radeon_ring_write(rdev, (pitch << 22) | (dst_offset >> 10)); | |
499 | radeon_ring_write(rdev, (0x1fff) | (0x1fff << 16)); | |
500 | radeon_ring_write(rdev, 0); | |
501 | radeon_ring_write(rdev, (0x1fff) | (0x1fff << 16)); | |
502 | radeon_ring_write(rdev, num_pages); | |
503 | radeon_ring_write(rdev, num_pages); | |
504 | radeon_ring_write(rdev, cur_pages | (stride_pixels << 16)); | |
505 | } | |
506 | radeon_ring_write(rdev, PACKET0(RADEON_DSTCACHE_CTLSTAT, 0)); | |
507 | radeon_ring_write(rdev, RADEON_RB2D_DC_FLUSH_ALL); | |
508 | radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0)); | |
509 | radeon_ring_write(rdev, | |
510 | RADEON_WAIT_2D_IDLECLEAN | | |
511 | RADEON_WAIT_HOST_IDLECLEAN | | |
512 | RADEON_WAIT_DMA_GUI_IDLE); | |
513 | if (fence) { | |
514 | r = radeon_fence_emit(rdev, fence); | |
515 | } | |
516 | radeon_ring_unlock_commit(rdev); | |
517 | return r; | |
518 | } | |
519 | ||
45600232 JG |
520 | static int r100_cp_wait_for_idle(struct radeon_device *rdev) |
521 | { | |
522 | unsigned i; | |
523 | u32 tmp; | |
524 | ||
525 | for (i = 0; i < rdev->usec_timeout; i++) { | |
526 | tmp = RREG32(R_000E40_RBBM_STATUS); | |
527 | if (!G_000E40_CP_CMDSTRM_BUSY(tmp)) { | |
528 | return 0; | |
529 | } | |
530 | udelay(1); | |
531 | } | |
532 | return -1; | |
533 | } | |
534 | ||
771fe6b9 JG |
535 | void r100_ring_start(struct radeon_device *rdev) |
536 | { | |
537 | int r; | |
538 | ||
539 | r = radeon_ring_lock(rdev, 2); | |
540 | if (r) { | |
541 | return; | |
542 | } | |
543 | radeon_ring_write(rdev, PACKET0(RADEON_ISYNC_CNTL, 0)); | |
544 | radeon_ring_write(rdev, | |
545 | RADEON_ISYNC_ANY2D_IDLE3D | | |
546 | RADEON_ISYNC_ANY3D_IDLE2D | | |
547 | RADEON_ISYNC_WAIT_IDLEGUI | | |
548 | RADEON_ISYNC_CPSCRATCH_IDLEGUI); | |
549 | radeon_ring_unlock_commit(rdev); | |
550 | } | |
551 | ||
70967ab9 BH |
552 | |
553 | /* Load the microcode for the CP */ | |
554 | static int r100_cp_init_microcode(struct radeon_device *rdev) | |
771fe6b9 | 555 | { |
70967ab9 BH |
556 | struct platform_device *pdev; |
557 | const char *fw_name = NULL; | |
558 | int err; | |
771fe6b9 | 559 | |
70967ab9 | 560 | DRM_DEBUG("\n"); |
771fe6b9 | 561 | |
70967ab9 BH |
562 | pdev = platform_device_register_simple("radeon_cp", 0, NULL, 0); |
563 | err = IS_ERR(pdev); | |
564 | if (err) { | |
565 | printk(KERN_ERR "radeon_cp: Failed to register firmware\n"); | |
566 | return -EINVAL; | |
567 | } | |
771fe6b9 JG |
568 | if ((rdev->family == CHIP_R100) || (rdev->family == CHIP_RV100) || |
569 | (rdev->family == CHIP_RV200) || (rdev->family == CHIP_RS100) || | |
570 | (rdev->family == CHIP_RS200)) { | |
571 | DRM_INFO("Loading R100 Microcode\n"); | |
70967ab9 | 572 | fw_name = FIRMWARE_R100; |
771fe6b9 JG |
573 | } else if ((rdev->family == CHIP_R200) || |
574 | (rdev->family == CHIP_RV250) || | |
575 | (rdev->family == CHIP_RV280) || | |
576 | (rdev->family == CHIP_RS300)) { | |
577 | DRM_INFO("Loading R200 Microcode\n"); | |
70967ab9 | 578 | fw_name = FIRMWARE_R200; |
771fe6b9 JG |
579 | } else if ((rdev->family == CHIP_R300) || |
580 | (rdev->family == CHIP_R350) || | |
581 | (rdev->family == CHIP_RV350) || | |
582 | (rdev->family == CHIP_RV380) || | |
583 | (rdev->family == CHIP_RS400) || | |
584 | (rdev->family == CHIP_RS480)) { | |
585 | DRM_INFO("Loading R300 Microcode\n"); | |
70967ab9 | 586 | fw_name = FIRMWARE_R300; |
771fe6b9 JG |
587 | } else if ((rdev->family == CHIP_R420) || |
588 | (rdev->family == CHIP_R423) || | |
589 | (rdev->family == CHIP_RV410)) { | |
590 | DRM_INFO("Loading R400 Microcode\n"); | |
70967ab9 | 591 | fw_name = FIRMWARE_R420; |
771fe6b9 JG |
592 | } else if ((rdev->family == CHIP_RS690) || |
593 | (rdev->family == CHIP_RS740)) { | |
594 | DRM_INFO("Loading RS690/RS740 Microcode\n"); | |
70967ab9 | 595 | fw_name = FIRMWARE_RS690; |
771fe6b9 JG |
596 | } else if (rdev->family == CHIP_RS600) { |
597 | DRM_INFO("Loading RS600 Microcode\n"); | |
70967ab9 | 598 | fw_name = FIRMWARE_RS600; |
771fe6b9 JG |
599 | } else if ((rdev->family == CHIP_RV515) || |
600 | (rdev->family == CHIP_R520) || | |
601 | (rdev->family == CHIP_RV530) || | |
602 | (rdev->family == CHIP_R580) || | |
603 | (rdev->family == CHIP_RV560) || | |
604 | (rdev->family == CHIP_RV570)) { | |
605 | DRM_INFO("Loading R500 Microcode\n"); | |
70967ab9 BH |
606 | fw_name = FIRMWARE_R520; |
607 | } | |
608 | ||
3ce0a23d | 609 | err = request_firmware(&rdev->me_fw, fw_name, &pdev->dev); |
70967ab9 BH |
610 | platform_device_unregister(pdev); |
611 | if (err) { | |
612 | printk(KERN_ERR "radeon_cp: Failed to load firmware \"%s\"\n", | |
613 | fw_name); | |
3ce0a23d | 614 | } else if (rdev->me_fw->size % 8) { |
70967ab9 BH |
615 | printk(KERN_ERR |
616 | "radeon_cp: Bogus length %zu in firmware \"%s\"\n", | |
3ce0a23d | 617 | rdev->me_fw->size, fw_name); |
70967ab9 | 618 | err = -EINVAL; |
3ce0a23d JG |
619 | release_firmware(rdev->me_fw); |
620 | rdev->me_fw = NULL; | |
70967ab9 BH |
621 | } |
622 | return err; | |
623 | } | |
d4550907 | 624 | |
70967ab9 BH |
625 | static void r100_cp_load_microcode(struct radeon_device *rdev) |
626 | { | |
627 | const __be32 *fw_data; | |
628 | int i, size; | |
629 | ||
630 | if (r100_gui_wait_for_idle(rdev)) { | |
631 | printk(KERN_WARNING "Failed to wait GUI idle while " | |
632 | "programming pipes. Bad things might happen.\n"); | |
633 | } | |
634 | ||
3ce0a23d JG |
635 | if (rdev->me_fw) { |
636 | size = rdev->me_fw->size / 4; | |
637 | fw_data = (const __be32 *)&rdev->me_fw->data[0]; | |
70967ab9 BH |
638 | WREG32(RADEON_CP_ME_RAM_ADDR, 0); |
639 | for (i = 0; i < size; i += 2) { | |
640 | WREG32(RADEON_CP_ME_RAM_DATAH, | |
641 | be32_to_cpup(&fw_data[i])); | |
642 | WREG32(RADEON_CP_ME_RAM_DATAL, | |
643 | be32_to_cpup(&fw_data[i + 1])); | |
771fe6b9 JG |
644 | } |
645 | } | |
646 | } | |
647 | ||
648 | int r100_cp_init(struct radeon_device *rdev, unsigned ring_size) | |
649 | { | |
650 | unsigned rb_bufsz; | |
651 | unsigned rb_blksz; | |
652 | unsigned max_fetch; | |
653 | unsigned pre_write_timer; | |
654 | unsigned pre_write_limit; | |
655 | unsigned indirect2_start; | |
656 | unsigned indirect1_start; | |
657 | uint32_t tmp; | |
658 | int r; | |
659 | ||
660 | if (r100_debugfs_cp_init(rdev)) { | |
661 | DRM_ERROR("Failed to register debugfs file for CP !\n"); | |
662 | } | |
663 | /* Reset CP */ | |
664 | tmp = RREG32(RADEON_CP_CSQ_STAT); | |
665 | if ((tmp & (1 << 31))) { | |
666 | DRM_INFO("radeon: cp busy (0x%08X) resetting\n", tmp); | |
667 | WREG32(RADEON_CP_CSQ_MODE, 0); | |
668 | WREG32(RADEON_CP_CSQ_CNTL, 0); | |
669 | WREG32(RADEON_RBBM_SOFT_RESET, RADEON_SOFT_RESET_CP); | |
670 | tmp = RREG32(RADEON_RBBM_SOFT_RESET); | |
671 | mdelay(2); | |
672 | WREG32(RADEON_RBBM_SOFT_RESET, 0); | |
673 | tmp = RREG32(RADEON_RBBM_SOFT_RESET); | |
674 | mdelay(2); | |
675 | tmp = RREG32(RADEON_CP_CSQ_STAT); | |
676 | if ((tmp & (1 << 31))) { | |
677 | DRM_INFO("radeon: cp reset failed (0x%08X)\n", tmp); | |
678 | } | |
679 | } else { | |
680 | DRM_INFO("radeon: cp idle (0x%08X)\n", tmp); | |
681 | } | |
70967ab9 | 682 | |
3ce0a23d | 683 | if (!rdev->me_fw) { |
70967ab9 BH |
684 | r = r100_cp_init_microcode(rdev); |
685 | if (r) { | |
686 | DRM_ERROR("Failed to load firmware!\n"); | |
687 | return r; | |
688 | } | |
689 | } | |
690 | ||
771fe6b9 JG |
691 | /* Align ring size */ |
692 | rb_bufsz = drm_order(ring_size / 8); | |
693 | ring_size = (1 << (rb_bufsz + 1)) * 4; | |
694 | r100_cp_load_microcode(rdev); | |
695 | r = radeon_ring_init(rdev, ring_size); | |
696 | if (r) { | |
697 | return r; | |
698 | } | |
699 | /* Each time the cp read 1024 bytes (16 dword/quadword) update | |
700 | * the rptr copy in system ram */ | |
701 | rb_blksz = 9; | |
702 | /* cp will read 128bytes at a time (4 dwords) */ | |
703 | max_fetch = 1; | |
704 | rdev->cp.align_mask = 16 - 1; | |
705 | /* Write to CP_RB_WPTR will be delayed for pre_write_timer clocks */ | |
706 | pre_write_timer = 64; | |
707 | /* Force CP_RB_WPTR write if written more than one time before the | |
708 | * delay expire | |
709 | */ | |
710 | pre_write_limit = 0; | |
711 | /* Setup the cp cache like this (cache size is 96 dwords) : | |
712 | * RING 0 to 15 | |
713 | * INDIRECT1 16 to 79 | |
714 | * INDIRECT2 80 to 95 | |
715 | * So ring cache size is 16dwords (> (2 * max_fetch = 2 * 4dwords)) | |
716 | * indirect1 cache size is 64dwords (> (2 * max_fetch = 2 * 4dwords)) | |
717 | * indirect2 cache size is 16dwords (> (2 * max_fetch = 2 * 4dwords)) | |
718 | * Idea being that most of the gpu cmd will be through indirect1 buffer | |
719 | * so it gets the bigger cache. | |
720 | */ | |
721 | indirect2_start = 80; | |
722 | indirect1_start = 16; | |
723 | /* cp setup */ | |
724 | WREG32(0x718, pre_write_timer | (pre_write_limit << 28)); | |
d6f28938 | 725 | tmp = (REG_SET(RADEON_RB_BUFSZ, rb_bufsz) | |
771fe6b9 JG |
726 | REG_SET(RADEON_RB_BLKSZ, rb_blksz) | |
727 | REG_SET(RADEON_MAX_FETCH, max_fetch) | | |
728 | RADEON_RB_NO_UPDATE); | |
d6f28938 AD |
729 | #ifdef __BIG_ENDIAN |
730 | tmp |= RADEON_BUF_SWAP_32BIT; | |
731 | #endif | |
732 | WREG32(RADEON_CP_RB_CNTL, tmp); | |
733 | ||
771fe6b9 JG |
734 | /* Set ring address */ |
735 | DRM_INFO("radeon: ring at 0x%016lX\n", (unsigned long)rdev->cp.gpu_addr); | |
736 | WREG32(RADEON_CP_RB_BASE, rdev->cp.gpu_addr); | |
737 | /* Force read & write ptr to 0 */ | |
771fe6b9 JG |
738 | WREG32(RADEON_CP_RB_CNTL, tmp | RADEON_RB_RPTR_WR_ENA); |
739 | WREG32(RADEON_CP_RB_RPTR_WR, 0); | |
740 | WREG32(RADEON_CP_RB_WPTR, 0); | |
741 | WREG32(RADEON_CP_RB_CNTL, tmp); | |
742 | udelay(10); | |
743 | rdev->cp.rptr = RREG32(RADEON_CP_RB_RPTR); | |
744 | rdev->cp.wptr = RREG32(RADEON_CP_RB_WPTR); | |
745 | /* Set cp mode to bus mastering & enable cp*/ | |
746 | WREG32(RADEON_CP_CSQ_MODE, | |
747 | REG_SET(RADEON_INDIRECT2_START, indirect2_start) | | |
748 | REG_SET(RADEON_INDIRECT1_START, indirect1_start)); | |
749 | WREG32(0x718, 0); | |
750 | WREG32(0x744, 0x00004D4D); | |
751 | WREG32(RADEON_CP_CSQ_CNTL, RADEON_CSQ_PRIBM_INDBM); | |
752 | radeon_ring_start(rdev); | |
753 | r = radeon_ring_test(rdev); | |
754 | if (r) { | |
755 | DRM_ERROR("radeon: cp isn't working (%d).\n", r); | |
756 | return r; | |
757 | } | |
758 | rdev->cp.ready = true; | |
759 | return 0; | |
760 | } | |
761 | ||
762 | void r100_cp_fini(struct radeon_device *rdev) | |
763 | { | |
45600232 JG |
764 | if (r100_cp_wait_for_idle(rdev)) { |
765 | DRM_ERROR("Wait for CP idle timeout, shutting down CP.\n"); | |
766 | } | |
771fe6b9 | 767 | /* Disable ring */ |
a18d7ea1 | 768 | r100_cp_disable(rdev); |
771fe6b9 JG |
769 | radeon_ring_fini(rdev); |
770 | DRM_INFO("radeon: cp finalized\n"); | |
771 | } | |
772 | ||
773 | void r100_cp_disable(struct radeon_device *rdev) | |
774 | { | |
775 | /* Disable ring */ | |
776 | rdev->cp.ready = false; | |
777 | WREG32(RADEON_CP_CSQ_MODE, 0); | |
778 | WREG32(RADEON_CP_CSQ_CNTL, 0); | |
779 | if (r100_gui_wait_for_idle(rdev)) { | |
780 | printk(KERN_WARNING "Failed to wait GUI idle while " | |
781 | "programming pipes. Bad things might happen.\n"); | |
782 | } | |
783 | } | |
784 | ||
785 | int r100_cp_reset(struct radeon_device *rdev) | |
786 | { | |
787 | uint32_t tmp; | |
788 | bool reinit_cp; | |
789 | int i; | |
790 | ||
791 | reinit_cp = rdev->cp.ready; | |
792 | rdev->cp.ready = false; | |
793 | WREG32(RADEON_CP_CSQ_MODE, 0); | |
794 | WREG32(RADEON_CP_CSQ_CNTL, 0); | |
795 | WREG32(RADEON_RBBM_SOFT_RESET, RADEON_SOFT_RESET_CP); | |
796 | (void)RREG32(RADEON_RBBM_SOFT_RESET); | |
797 | udelay(200); | |
798 | WREG32(RADEON_RBBM_SOFT_RESET, 0); | |
799 | /* Wait to prevent race in RBBM_STATUS */ | |
800 | mdelay(1); | |
801 | for (i = 0; i < rdev->usec_timeout; i++) { | |
802 | tmp = RREG32(RADEON_RBBM_STATUS); | |
803 | if (!(tmp & (1 << 16))) { | |
804 | DRM_INFO("CP reset succeed (RBBM_STATUS=0x%08X)\n", | |
805 | tmp); | |
806 | if (reinit_cp) { | |
807 | return r100_cp_init(rdev, rdev->cp.ring_size); | |
808 | } | |
809 | return 0; | |
810 | } | |
811 | DRM_UDELAY(1); | |
812 | } | |
813 | tmp = RREG32(RADEON_RBBM_STATUS); | |
814 | DRM_ERROR("Failed to reset CP (RBBM_STATUS=0x%08X)!\n", tmp); | |
815 | return -1; | |
816 | } | |
817 | ||
3ce0a23d JG |
818 | void r100_cp_commit(struct radeon_device *rdev) |
819 | { | |
820 | WREG32(RADEON_CP_RB_WPTR, rdev->cp.wptr); | |
821 | (void)RREG32(RADEON_CP_RB_WPTR); | |
822 | } | |
823 | ||
771fe6b9 JG |
824 | |
825 | /* | |
826 | * CS functions | |
827 | */ | |
828 | int r100_cs_parse_packet0(struct radeon_cs_parser *p, | |
829 | struct radeon_cs_packet *pkt, | |
068a117c | 830 | const unsigned *auth, unsigned n, |
771fe6b9 JG |
831 | radeon_packet0_check_t check) |
832 | { | |
833 | unsigned reg; | |
834 | unsigned i, j, m; | |
835 | unsigned idx; | |
836 | int r; | |
837 | ||
838 | idx = pkt->idx + 1; | |
839 | reg = pkt->reg; | |
068a117c JG |
840 | /* Check that register fall into register range |
841 | * determined by the number of entry (n) in the | |
842 | * safe register bitmap. | |
843 | */ | |
771fe6b9 JG |
844 | if (pkt->one_reg_wr) { |
845 | if ((reg >> 7) > n) { | |
846 | return -EINVAL; | |
847 | } | |
848 | } else { | |
849 | if (((reg + (pkt->count << 2)) >> 7) > n) { | |
850 | return -EINVAL; | |
851 | } | |
852 | } | |
853 | for (i = 0; i <= pkt->count; i++, idx++) { | |
854 | j = (reg >> 7); | |
855 | m = 1 << ((reg >> 2) & 31); | |
856 | if (auth[j] & m) { | |
857 | r = check(p, pkt, idx, reg); | |
858 | if (r) { | |
859 | return r; | |
860 | } | |
861 | } | |
862 | if (pkt->one_reg_wr) { | |
863 | if (!(auth[j] & m)) { | |
864 | break; | |
865 | } | |
866 | } else { | |
867 | reg += 4; | |
868 | } | |
869 | } | |
870 | return 0; | |
871 | } | |
872 | ||
771fe6b9 JG |
873 | void r100_cs_dump_packet(struct radeon_cs_parser *p, |
874 | struct radeon_cs_packet *pkt) | |
875 | { | |
771fe6b9 JG |
876 | volatile uint32_t *ib; |
877 | unsigned i; | |
878 | unsigned idx; | |
879 | ||
880 | ib = p->ib->ptr; | |
771fe6b9 JG |
881 | idx = pkt->idx; |
882 | for (i = 0; i <= (pkt->count + 1); i++, idx++) { | |
883 | DRM_INFO("ib[%d]=0x%08X\n", idx, ib[idx]); | |
884 | } | |
885 | } | |
886 | ||
887 | /** | |
888 | * r100_cs_packet_parse() - parse cp packet and point ib index to next packet | |
889 | * @parser: parser structure holding parsing context. | |
890 | * @pkt: where to store packet informations | |
891 | * | |
892 | * Assume that chunk_ib_index is properly set. Will return -EINVAL | |
893 | * if packet is bigger than remaining ib size. or if packets is unknown. | |
894 | **/ | |
895 | int r100_cs_packet_parse(struct radeon_cs_parser *p, | |
896 | struct radeon_cs_packet *pkt, | |
897 | unsigned idx) | |
898 | { | |
899 | struct radeon_cs_chunk *ib_chunk = &p->chunks[p->chunk_ib_idx]; | |
fa99239c | 900 | uint32_t header; |
771fe6b9 JG |
901 | |
902 | if (idx >= ib_chunk->length_dw) { | |
903 | DRM_ERROR("Can not parse packet at %d after CS end %d !\n", | |
904 | idx, ib_chunk->length_dw); | |
905 | return -EINVAL; | |
906 | } | |
513bcb46 | 907 | header = radeon_get_ib_value(p, idx); |
771fe6b9 JG |
908 | pkt->idx = idx; |
909 | pkt->type = CP_PACKET_GET_TYPE(header); | |
910 | pkt->count = CP_PACKET_GET_COUNT(header); | |
911 | switch (pkt->type) { | |
912 | case PACKET_TYPE0: | |
913 | pkt->reg = CP_PACKET0_GET_REG(header); | |
914 | pkt->one_reg_wr = CP_PACKET0_GET_ONE_REG_WR(header); | |
915 | break; | |
916 | case PACKET_TYPE3: | |
917 | pkt->opcode = CP_PACKET3_GET_OPCODE(header); | |
918 | break; | |
919 | case PACKET_TYPE2: | |
920 | pkt->count = -1; | |
921 | break; | |
922 | default: | |
923 | DRM_ERROR("Unknown packet type %d at %d !\n", pkt->type, idx); | |
924 | return -EINVAL; | |
925 | } | |
926 | if ((pkt->count + 1 + pkt->idx) >= ib_chunk->length_dw) { | |
927 | DRM_ERROR("Packet (%d:%d:%d) end after CS buffer (%d) !\n", | |
928 | pkt->idx, pkt->type, pkt->count, ib_chunk->length_dw); | |
929 | return -EINVAL; | |
930 | } | |
931 | return 0; | |
932 | } | |
933 | ||
531369e6 DA |
934 | /** |
935 | * r100_cs_packet_next_vline() - parse userspace VLINE packet | |
936 | * @parser: parser structure holding parsing context. | |
937 | * | |
938 | * Userspace sends a special sequence for VLINE waits. | |
939 | * PACKET0 - VLINE_START_END + value | |
940 | * PACKET0 - WAIT_UNTIL +_value | |
941 | * RELOC (P3) - crtc_id in reloc. | |
942 | * | |
943 | * This function parses this and relocates the VLINE START END | |
944 | * and WAIT UNTIL packets to the correct crtc. | |
945 | * It also detects a switched off crtc and nulls out the | |
946 | * wait in that case. | |
947 | */ | |
948 | int r100_cs_packet_parse_vline(struct radeon_cs_parser *p) | |
949 | { | |
531369e6 DA |
950 | struct drm_mode_object *obj; |
951 | struct drm_crtc *crtc; | |
952 | struct radeon_crtc *radeon_crtc; | |
953 | struct radeon_cs_packet p3reloc, waitreloc; | |
954 | int crtc_id; | |
955 | int r; | |
956 | uint32_t header, h_idx, reg; | |
513bcb46 | 957 | volatile uint32_t *ib; |
531369e6 | 958 | |
513bcb46 | 959 | ib = p->ib->ptr; |
531369e6 DA |
960 | |
961 | /* parse the wait until */ | |
962 | r = r100_cs_packet_parse(p, &waitreloc, p->idx); | |
963 | if (r) | |
964 | return r; | |
965 | ||
966 | /* check its a wait until and only 1 count */ | |
967 | if (waitreloc.reg != RADEON_WAIT_UNTIL || | |
968 | waitreloc.count != 0) { | |
969 | DRM_ERROR("vline wait had illegal wait until segment\n"); | |
970 | r = -EINVAL; | |
971 | return r; | |
972 | } | |
973 | ||
513bcb46 | 974 | if (radeon_get_ib_value(p, waitreloc.idx + 1) != RADEON_WAIT_CRTC_VLINE) { |
531369e6 DA |
975 | DRM_ERROR("vline wait had illegal wait until\n"); |
976 | r = -EINVAL; | |
977 | return r; | |
978 | } | |
979 | ||
980 | /* jump over the NOP */ | |
90ebd065 | 981 | r = r100_cs_packet_parse(p, &p3reloc, p->idx + waitreloc.count + 2); |
531369e6 DA |
982 | if (r) |
983 | return r; | |
984 | ||
985 | h_idx = p->idx - 2; | |
90ebd065 AD |
986 | p->idx += waitreloc.count + 2; |
987 | p->idx += p3reloc.count + 2; | |
531369e6 | 988 | |
513bcb46 DA |
989 | header = radeon_get_ib_value(p, h_idx); |
990 | crtc_id = radeon_get_ib_value(p, h_idx + 5); | |
d4ac6a05 | 991 | reg = CP_PACKET0_GET_REG(header); |
531369e6 DA |
992 | mutex_lock(&p->rdev->ddev->mode_config.mutex); |
993 | obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC); | |
994 | if (!obj) { | |
995 | DRM_ERROR("cannot find crtc %d\n", crtc_id); | |
996 | r = -EINVAL; | |
997 | goto out; | |
998 | } | |
999 | crtc = obj_to_crtc(obj); | |
1000 | radeon_crtc = to_radeon_crtc(crtc); | |
1001 | crtc_id = radeon_crtc->crtc_id; | |
1002 | ||
1003 | if (!crtc->enabled) { | |
1004 | /* if the CRTC isn't enabled - we need to nop out the wait until */ | |
513bcb46 DA |
1005 | ib[h_idx + 2] = PACKET2(0); |
1006 | ib[h_idx + 3] = PACKET2(0); | |
531369e6 DA |
1007 | } else if (crtc_id == 1) { |
1008 | switch (reg) { | |
1009 | case AVIVO_D1MODE_VLINE_START_END: | |
90ebd065 | 1010 | header &= ~R300_CP_PACKET0_REG_MASK; |
531369e6 DA |
1011 | header |= AVIVO_D2MODE_VLINE_START_END >> 2; |
1012 | break; | |
1013 | case RADEON_CRTC_GUI_TRIG_VLINE: | |
90ebd065 | 1014 | header &= ~R300_CP_PACKET0_REG_MASK; |
531369e6 DA |
1015 | header |= RADEON_CRTC2_GUI_TRIG_VLINE >> 2; |
1016 | break; | |
1017 | default: | |
1018 | DRM_ERROR("unknown crtc reloc\n"); | |
1019 | r = -EINVAL; | |
1020 | goto out; | |
1021 | } | |
513bcb46 DA |
1022 | ib[h_idx] = header; |
1023 | ib[h_idx + 3] |= RADEON_ENG_DISPLAY_SELECT_CRTC1; | |
531369e6 DA |
1024 | } |
1025 | out: | |
1026 | mutex_unlock(&p->rdev->ddev->mode_config.mutex); | |
1027 | return r; | |
1028 | } | |
1029 | ||
771fe6b9 JG |
1030 | /** |
1031 | * r100_cs_packet_next_reloc() - parse next packet which should be reloc packet3 | |
1032 | * @parser: parser structure holding parsing context. | |
1033 | * @data: pointer to relocation data | |
1034 | * @offset_start: starting offset | |
1035 | * @offset_mask: offset mask (to align start offset on) | |
1036 | * @reloc: reloc informations | |
1037 | * | |
1038 | * Check next packet is relocation packet3, do bo validation and compute | |
1039 | * GPU offset using the provided start. | |
1040 | **/ | |
1041 | int r100_cs_packet_next_reloc(struct radeon_cs_parser *p, | |
1042 | struct radeon_cs_reloc **cs_reloc) | |
1043 | { | |
771fe6b9 JG |
1044 | struct radeon_cs_chunk *relocs_chunk; |
1045 | struct radeon_cs_packet p3reloc; | |
1046 | unsigned idx; | |
1047 | int r; | |
1048 | ||
1049 | if (p->chunk_relocs_idx == -1) { | |
1050 | DRM_ERROR("No relocation chunk !\n"); | |
1051 | return -EINVAL; | |
1052 | } | |
1053 | *cs_reloc = NULL; | |
771fe6b9 JG |
1054 | relocs_chunk = &p->chunks[p->chunk_relocs_idx]; |
1055 | r = r100_cs_packet_parse(p, &p3reloc, p->idx); | |
1056 | if (r) { | |
1057 | return r; | |
1058 | } | |
1059 | p->idx += p3reloc.count + 2; | |
1060 | if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) { | |
1061 | DRM_ERROR("No packet3 for relocation for packet at %d.\n", | |
1062 | p3reloc.idx); | |
1063 | r100_cs_dump_packet(p, &p3reloc); | |
1064 | return -EINVAL; | |
1065 | } | |
513bcb46 | 1066 | idx = radeon_get_ib_value(p, p3reloc.idx + 1); |
771fe6b9 JG |
1067 | if (idx >= relocs_chunk->length_dw) { |
1068 | DRM_ERROR("Relocs at %d after relocations chunk end %d !\n", | |
1069 | idx, relocs_chunk->length_dw); | |
1070 | r100_cs_dump_packet(p, &p3reloc); | |
1071 | return -EINVAL; | |
1072 | } | |
1073 | /* FIXME: we assume reloc size is 4 dwords */ | |
1074 | *cs_reloc = p->relocs_ptr[(idx / 4)]; | |
1075 | return 0; | |
1076 | } | |
1077 | ||
551ebd83 DA |
1078 | static int r100_get_vtx_size(uint32_t vtx_fmt) |
1079 | { | |
1080 | int vtx_size; | |
1081 | vtx_size = 2; | |
1082 | /* ordered according to bits in spec */ | |
1083 | if (vtx_fmt & RADEON_SE_VTX_FMT_W0) | |
1084 | vtx_size++; | |
1085 | if (vtx_fmt & RADEON_SE_VTX_FMT_FPCOLOR) | |
1086 | vtx_size += 3; | |
1087 | if (vtx_fmt & RADEON_SE_VTX_FMT_FPALPHA) | |
1088 | vtx_size++; | |
1089 | if (vtx_fmt & RADEON_SE_VTX_FMT_PKCOLOR) | |
1090 | vtx_size++; | |
1091 | if (vtx_fmt & RADEON_SE_VTX_FMT_FPSPEC) | |
1092 | vtx_size += 3; | |
1093 | if (vtx_fmt & RADEON_SE_VTX_FMT_FPFOG) | |
1094 | vtx_size++; | |
1095 | if (vtx_fmt & RADEON_SE_VTX_FMT_PKSPEC) | |
1096 | vtx_size++; | |
1097 | if (vtx_fmt & RADEON_SE_VTX_FMT_ST0) | |
1098 | vtx_size += 2; | |
1099 | if (vtx_fmt & RADEON_SE_VTX_FMT_ST1) | |
1100 | vtx_size += 2; | |
1101 | if (vtx_fmt & RADEON_SE_VTX_FMT_Q1) | |
1102 | vtx_size++; | |
1103 | if (vtx_fmt & RADEON_SE_VTX_FMT_ST2) | |
1104 | vtx_size += 2; | |
1105 | if (vtx_fmt & RADEON_SE_VTX_FMT_Q2) | |
1106 | vtx_size++; | |
1107 | if (vtx_fmt & RADEON_SE_VTX_FMT_ST3) | |
1108 | vtx_size += 2; | |
1109 | if (vtx_fmt & RADEON_SE_VTX_FMT_Q3) | |
1110 | vtx_size++; | |
1111 | if (vtx_fmt & RADEON_SE_VTX_FMT_Q0) | |
1112 | vtx_size++; | |
1113 | /* blend weight */ | |
1114 | if (vtx_fmt & (0x7 << 15)) | |
1115 | vtx_size += (vtx_fmt >> 15) & 0x7; | |
1116 | if (vtx_fmt & RADEON_SE_VTX_FMT_N0) | |
1117 | vtx_size += 3; | |
1118 | if (vtx_fmt & RADEON_SE_VTX_FMT_XY1) | |
1119 | vtx_size += 2; | |
1120 | if (vtx_fmt & RADEON_SE_VTX_FMT_Z1) | |
1121 | vtx_size++; | |
1122 | if (vtx_fmt & RADEON_SE_VTX_FMT_W1) | |
1123 | vtx_size++; | |
1124 | if (vtx_fmt & RADEON_SE_VTX_FMT_N1) | |
1125 | vtx_size++; | |
1126 | if (vtx_fmt & RADEON_SE_VTX_FMT_Z) | |
1127 | vtx_size++; | |
1128 | return vtx_size; | |
1129 | } | |
1130 | ||
771fe6b9 | 1131 | static int r100_packet0_check(struct radeon_cs_parser *p, |
551ebd83 DA |
1132 | struct radeon_cs_packet *pkt, |
1133 | unsigned idx, unsigned reg) | |
771fe6b9 | 1134 | { |
771fe6b9 | 1135 | struct radeon_cs_reloc *reloc; |
551ebd83 | 1136 | struct r100_cs_track *track; |
771fe6b9 JG |
1137 | volatile uint32_t *ib; |
1138 | uint32_t tmp; | |
771fe6b9 | 1139 | int r; |
551ebd83 | 1140 | int i, face; |
e024e110 | 1141 | u32 tile_flags = 0; |
513bcb46 | 1142 | u32 idx_value; |
771fe6b9 JG |
1143 | |
1144 | ib = p->ib->ptr; | |
551ebd83 DA |
1145 | track = (struct r100_cs_track *)p->track; |
1146 | ||
513bcb46 DA |
1147 | idx_value = radeon_get_ib_value(p, idx); |
1148 | ||
551ebd83 DA |
1149 | switch (reg) { |
1150 | case RADEON_CRTC_GUI_TRIG_VLINE: | |
1151 | r = r100_cs_packet_parse_vline(p); | |
1152 | if (r) { | |
1153 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1154 | idx, reg); | |
1155 | r100_cs_dump_packet(p, pkt); | |
1156 | return r; | |
1157 | } | |
1158 | break; | |
771fe6b9 JG |
1159 | /* FIXME: only allow PACKET3 blit? easier to check for out of |
1160 | * range access */ | |
551ebd83 DA |
1161 | case RADEON_DST_PITCH_OFFSET: |
1162 | case RADEON_SRC_PITCH_OFFSET: | |
1163 | r = r100_reloc_pitch_offset(p, pkt, idx, reg); | |
1164 | if (r) | |
1165 | return r; | |
1166 | break; | |
1167 | case RADEON_RB3D_DEPTHOFFSET: | |
1168 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1169 | if (r) { | |
1170 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1171 | idx, reg); | |
1172 | r100_cs_dump_packet(p, pkt); | |
1173 | return r; | |
1174 | } | |
1175 | track->zb.robj = reloc->robj; | |
513bcb46 DA |
1176 | track->zb.offset = idx_value; |
1177 | ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); | |
551ebd83 DA |
1178 | break; |
1179 | case RADEON_RB3D_COLOROFFSET: | |
1180 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1181 | if (r) { | |
1182 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1183 | idx, reg); | |
1184 | r100_cs_dump_packet(p, pkt); | |
1185 | return r; | |
1186 | } | |
1187 | track->cb[0].robj = reloc->robj; | |
513bcb46 DA |
1188 | track->cb[0].offset = idx_value; |
1189 | ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); | |
551ebd83 DA |
1190 | break; |
1191 | case RADEON_PP_TXOFFSET_0: | |
1192 | case RADEON_PP_TXOFFSET_1: | |
1193 | case RADEON_PP_TXOFFSET_2: | |
1194 | i = (reg - RADEON_PP_TXOFFSET_0) / 24; | |
1195 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1196 | if (r) { | |
1197 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1198 | idx, reg); | |
1199 | r100_cs_dump_packet(p, pkt); | |
1200 | return r; | |
1201 | } | |
513bcb46 | 1202 | ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); |
551ebd83 DA |
1203 | track->textures[i].robj = reloc->robj; |
1204 | break; | |
1205 | case RADEON_PP_CUBIC_OFFSET_T0_0: | |
1206 | case RADEON_PP_CUBIC_OFFSET_T0_1: | |
1207 | case RADEON_PP_CUBIC_OFFSET_T0_2: | |
1208 | case RADEON_PP_CUBIC_OFFSET_T0_3: | |
1209 | case RADEON_PP_CUBIC_OFFSET_T0_4: | |
1210 | i = (reg - RADEON_PP_CUBIC_OFFSET_T0_0) / 4; | |
1211 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1212 | if (r) { | |
1213 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1214 | idx, reg); | |
1215 | r100_cs_dump_packet(p, pkt); | |
1216 | return r; | |
1217 | } | |
513bcb46 DA |
1218 | track->textures[0].cube_info[i].offset = idx_value; |
1219 | ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); | |
551ebd83 DA |
1220 | track->textures[0].cube_info[i].robj = reloc->robj; |
1221 | break; | |
1222 | case RADEON_PP_CUBIC_OFFSET_T1_0: | |
1223 | case RADEON_PP_CUBIC_OFFSET_T1_1: | |
1224 | case RADEON_PP_CUBIC_OFFSET_T1_2: | |
1225 | case RADEON_PP_CUBIC_OFFSET_T1_3: | |
1226 | case RADEON_PP_CUBIC_OFFSET_T1_4: | |
1227 | i = (reg - RADEON_PP_CUBIC_OFFSET_T1_0) / 4; | |
1228 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1229 | if (r) { | |
1230 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1231 | idx, reg); | |
1232 | r100_cs_dump_packet(p, pkt); | |
1233 | return r; | |
1234 | } | |
513bcb46 DA |
1235 | track->textures[1].cube_info[i].offset = idx_value; |
1236 | ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); | |
551ebd83 DA |
1237 | track->textures[1].cube_info[i].robj = reloc->robj; |
1238 | break; | |
1239 | case RADEON_PP_CUBIC_OFFSET_T2_0: | |
1240 | case RADEON_PP_CUBIC_OFFSET_T2_1: | |
1241 | case RADEON_PP_CUBIC_OFFSET_T2_2: | |
1242 | case RADEON_PP_CUBIC_OFFSET_T2_3: | |
1243 | case RADEON_PP_CUBIC_OFFSET_T2_4: | |
1244 | i = (reg - RADEON_PP_CUBIC_OFFSET_T2_0) / 4; | |
1245 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1246 | if (r) { | |
1247 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1248 | idx, reg); | |
1249 | r100_cs_dump_packet(p, pkt); | |
1250 | return r; | |
1251 | } | |
513bcb46 DA |
1252 | track->textures[2].cube_info[i].offset = idx_value; |
1253 | ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); | |
551ebd83 DA |
1254 | track->textures[2].cube_info[i].robj = reloc->robj; |
1255 | break; | |
1256 | case RADEON_RE_WIDTH_HEIGHT: | |
513bcb46 | 1257 | track->maxy = ((idx_value >> 16) & 0x7FF); |
551ebd83 DA |
1258 | break; |
1259 | case RADEON_RB3D_COLORPITCH: | |
1260 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1261 | if (r) { | |
1262 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1263 | idx, reg); | |
1264 | r100_cs_dump_packet(p, pkt); | |
1265 | return r; | |
1266 | } | |
e024e110 | 1267 | |
551ebd83 DA |
1268 | if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) |
1269 | tile_flags |= RADEON_COLOR_TILE_ENABLE; | |
1270 | if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) | |
1271 | tile_flags |= RADEON_COLOR_MICROTILE_ENABLE; | |
e024e110 | 1272 | |
513bcb46 | 1273 | tmp = idx_value & ~(0x7 << 16); |
551ebd83 DA |
1274 | tmp |= tile_flags; |
1275 | ib[idx] = tmp; | |
e024e110 | 1276 | |
513bcb46 | 1277 | track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK; |
551ebd83 DA |
1278 | break; |
1279 | case RADEON_RB3D_DEPTHPITCH: | |
513bcb46 | 1280 | track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK; |
551ebd83 DA |
1281 | break; |
1282 | case RADEON_RB3D_CNTL: | |
513bcb46 | 1283 | switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) { |
551ebd83 DA |
1284 | case 7: |
1285 | case 8: | |
1286 | case 9: | |
1287 | case 11: | |
1288 | case 12: | |
1289 | track->cb[0].cpp = 1; | |
e024e110 | 1290 | break; |
551ebd83 DA |
1291 | case 3: |
1292 | case 4: | |
1293 | case 15: | |
1294 | track->cb[0].cpp = 2; | |
1295 | break; | |
1296 | case 6: | |
1297 | track->cb[0].cpp = 4; | |
1298 | break; | |
1299 | default: | |
1300 | DRM_ERROR("Invalid color buffer format (%d) !\n", | |
513bcb46 | 1301 | ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f)); |
551ebd83 DA |
1302 | return -EINVAL; |
1303 | } | |
513bcb46 | 1304 | track->z_enabled = !!(idx_value & RADEON_Z_ENABLE); |
551ebd83 DA |
1305 | break; |
1306 | case RADEON_RB3D_ZSTENCILCNTL: | |
513bcb46 | 1307 | switch (idx_value & 0xf) { |
551ebd83 DA |
1308 | case 0: |
1309 | track->zb.cpp = 2; | |
1310 | break; | |
1311 | case 2: | |
1312 | case 3: | |
1313 | case 4: | |
1314 | case 5: | |
1315 | case 9: | |
1316 | case 11: | |
1317 | track->zb.cpp = 4; | |
17782d99 | 1318 | break; |
771fe6b9 | 1319 | default: |
771fe6b9 JG |
1320 | break; |
1321 | } | |
551ebd83 DA |
1322 | break; |
1323 | case RADEON_RB3D_ZPASS_ADDR: | |
1324 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1325 | if (r) { | |
1326 | DRM_ERROR("No reloc for ib[%d]=0x%04X\n", | |
1327 | idx, reg); | |
1328 | r100_cs_dump_packet(p, pkt); | |
1329 | return r; | |
1330 | } | |
513bcb46 | 1331 | ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); |
551ebd83 DA |
1332 | break; |
1333 | case RADEON_PP_CNTL: | |
1334 | { | |
513bcb46 | 1335 | uint32_t temp = idx_value >> 4; |
551ebd83 DA |
1336 | for (i = 0; i < track->num_texture; i++) |
1337 | track->textures[i].enabled = !!(temp & (1 << i)); | |
1338 | } | |
1339 | break; | |
1340 | case RADEON_SE_VF_CNTL: | |
513bcb46 | 1341 | track->vap_vf_cntl = idx_value; |
551ebd83 DA |
1342 | break; |
1343 | case RADEON_SE_VTX_FMT: | |
513bcb46 | 1344 | track->vtx_size = r100_get_vtx_size(idx_value); |
551ebd83 DA |
1345 | break; |
1346 | case RADEON_PP_TEX_SIZE_0: | |
1347 | case RADEON_PP_TEX_SIZE_1: | |
1348 | case RADEON_PP_TEX_SIZE_2: | |
1349 | i = (reg - RADEON_PP_TEX_SIZE_0) / 8; | |
513bcb46 DA |
1350 | track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1; |
1351 | track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1; | |
551ebd83 DA |
1352 | break; |
1353 | case RADEON_PP_TEX_PITCH_0: | |
1354 | case RADEON_PP_TEX_PITCH_1: | |
1355 | case RADEON_PP_TEX_PITCH_2: | |
1356 | i = (reg - RADEON_PP_TEX_PITCH_0) / 8; | |
513bcb46 | 1357 | track->textures[i].pitch = idx_value + 32; |
551ebd83 DA |
1358 | break; |
1359 | case RADEON_PP_TXFILTER_0: | |
1360 | case RADEON_PP_TXFILTER_1: | |
1361 | case RADEON_PP_TXFILTER_2: | |
1362 | i = (reg - RADEON_PP_TXFILTER_0) / 24; | |
513bcb46 | 1363 | track->textures[i].num_levels = ((idx_value & RADEON_MAX_MIP_LEVEL_MASK) |
551ebd83 | 1364 | >> RADEON_MAX_MIP_LEVEL_SHIFT); |
513bcb46 | 1365 | tmp = (idx_value >> 23) & 0x7; |
551ebd83 DA |
1366 | if (tmp == 2 || tmp == 6) |
1367 | track->textures[i].roundup_w = false; | |
513bcb46 | 1368 | tmp = (idx_value >> 27) & 0x7; |
551ebd83 DA |
1369 | if (tmp == 2 || tmp == 6) |
1370 | track->textures[i].roundup_h = false; | |
1371 | break; | |
1372 | case RADEON_PP_TXFORMAT_0: | |
1373 | case RADEON_PP_TXFORMAT_1: | |
1374 | case RADEON_PP_TXFORMAT_2: | |
1375 | i = (reg - RADEON_PP_TXFORMAT_0) / 24; | |
513bcb46 | 1376 | if (idx_value & RADEON_TXFORMAT_NON_POWER2) { |
551ebd83 DA |
1377 | track->textures[i].use_pitch = 1; |
1378 | } else { | |
1379 | track->textures[i].use_pitch = 0; | |
513bcb46 DA |
1380 | track->textures[i].width = 1 << ((idx_value >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK); |
1381 | track->textures[i].height = 1 << ((idx_value >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK); | |
551ebd83 | 1382 | } |
513bcb46 | 1383 | if (idx_value & RADEON_TXFORMAT_CUBIC_MAP_ENABLE) |
551ebd83 | 1384 | track->textures[i].tex_coord_type = 2; |
513bcb46 | 1385 | switch ((idx_value & RADEON_TXFORMAT_FORMAT_MASK)) { |
551ebd83 DA |
1386 | case RADEON_TXFORMAT_I8: |
1387 | case RADEON_TXFORMAT_RGB332: | |
1388 | case RADEON_TXFORMAT_Y8: | |
1389 | track->textures[i].cpp = 1; | |
1390 | break; | |
1391 | case RADEON_TXFORMAT_AI88: | |
1392 | case RADEON_TXFORMAT_ARGB1555: | |
1393 | case RADEON_TXFORMAT_RGB565: | |
1394 | case RADEON_TXFORMAT_ARGB4444: | |
1395 | case RADEON_TXFORMAT_VYUY422: | |
1396 | case RADEON_TXFORMAT_YVYU422: | |
551ebd83 DA |
1397 | case RADEON_TXFORMAT_SHADOW16: |
1398 | case RADEON_TXFORMAT_LDUDV655: | |
1399 | case RADEON_TXFORMAT_DUDV88: | |
1400 | track->textures[i].cpp = 2; | |
771fe6b9 | 1401 | break; |
551ebd83 DA |
1402 | case RADEON_TXFORMAT_ARGB8888: |
1403 | case RADEON_TXFORMAT_RGBA8888: | |
551ebd83 DA |
1404 | case RADEON_TXFORMAT_SHADOW32: |
1405 | case RADEON_TXFORMAT_LDUDUV8888: | |
1406 | track->textures[i].cpp = 4; | |
1407 | break; | |
d785d78b DA |
1408 | case RADEON_TXFORMAT_DXT1: |
1409 | track->textures[i].cpp = 1; | |
1410 | track->textures[i].compress_format = R100_TRACK_COMP_DXT1; | |
1411 | break; | |
1412 | case RADEON_TXFORMAT_DXT23: | |
1413 | case RADEON_TXFORMAT_DXT45: | |
1414 | track->textures[i].cpp = 1; | |
1415 | track->textures[i].compress_format = R100_TRACK_COMP_DXT35; | |
1416 | break; | |
551ebd83 | 1417 | } |
513bcb46 DA |
1418 | track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf); |
1419 | track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf); | |
551ebd83 DA |
1420 | break; |
1421 | case RADEON_PP_CUBIC_FACES_0: | |
1422 | case RADEON_PP_CUBIC_FACES_1: | |
1423 | case RADEON_PP_CUBIC_FACES_2: | |
513bcb46 | 1424 | tmp = idx_value; |
551ebd83 DA |
1425 | i = (reg - RADEON_PP_CUBIC_FACES_0) / 4; |
1426 | for (face = 0; face < 4; face++) { | |
1427 | track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf); | |
1428 | track->textures[i].cube_info[face].height = 1 << ((tmp >> ((face * 8) + 4)) & 0xf); | |
771fe6b9 | 1429 | } |
551ebd83 DA |
1430 | break; |
1431 | default: | |
1432 | printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n", | |
1433 | reg, idx); | |
1434 | return -EINVAL; | |
771fe6b9 JG |
1435 | } |
1436 | return 0; | |
1437 | } | |
1438 | ||
068a117c JG |
1439 | int r100_cs_track_check_pkt3_indx_buffer(struct radeon_cs_parser *p, |
1440 | struct radeon_cs_packet *pkt, | |
4c788679 | 1441 | struct radeon_bo *robj) |
068a117c | 1442 | { |
068a117c | 1443 | unsigned idx; |
513bcb46 | 1444 | u32 value; |
068a117c | 1445 | idx = pkt->idx + 1; |
513bcb46 | 1446 | value = radeon_get_ib_value(p, idx + 2); |
4c788679 | 1447 | if ((value + 1) > radeon_bo_size(robj)) { |
068a117c JG |
1448 | DRM_ERROR("[drm] Buffer too small for PACKET3 INDX_BUFFER " |
1449 | "(need %u have %lu) !\n", | |
513bcb46 | 1450 | value + 1, |
4c788679 | 1451 | radeon_bo_size(robj)); |
068a117c JG |
1452 | return -EINVAL; |
1453 | } | |
1454 | return 0; | |
1455 | } | |
1456 | ||
771fe6b9 JG |
1457 | static int r100_packet3_check(struct radeon_cs_parser *p, |
1458 | struct radeon_cs_packet *pkt) | |
1459 | { | |
771fe6b9 | 1460 | struct radeon_cs_reloc *reloc; |
551ebd83 | 1461 | struct r100_cs_track *track; |
771fe6b9 | 1462 | unsigned idx; |
771fe6b9 JG |
1463 | volatile uint32_t *ib; |
1464 | int r; | |
1465 | ||
1466 | ib = p->ib->ptr; | |
771fe6b9 | 1467 | idx = pkt->idx + 1; |
551ebd83 | 1468 | track = (struct r100_cs_track *)p->track; |
771fe6b9 JG |
1469 | switch (pkt->opcode) { |
1470 | case PACKET3_3D_LOAD_VBPNTR: | |
513bcb46 DA |
1471 | r = r100_packet3_load_vbpntr(p, pkt, idx); |
1472 | if (r) | |
1473 | return r; | |
771fe6b9 JG |
1474 | break; |
1475 | case PACKET3_INDX_BUFFER: | |
1476 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1477 | if (r) { | |
1478 | DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode); | |
1479 | r100_cs_dump_packet(p, pkt); | |
1480 | return r; | |
1481 | } | |
513bcb46 | 1482 | ib[idx+1] = radeon_get_ib_value(p, idx+1) + ((u32)reloc->lobj.gpu_offset); |
068a117c JG |
1483 | r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj); |
1484 | if (r) { | |
1485 | return r; | |
1486 | } | |
771fe6b9 JG |
1487 | break; |
1488 | case 0x23: | |
771fe6b9 JG |
1489 | /* 3D_RNDR_GEN_INDX_PRIM on r100/r200 */ |
1490 | r = r100_cs_packet_next_reloc(p, &reloc); | |
1491 | if (r) { | |
1492 | DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode); | |
1493 | r100_cs_dump_packet(p, pkt); | |
1494 | return r; | |
1495 | } | |
513bcb46 | 1496 | ib[idx] = radeon_get_ib_value(p, idx) + ((u32)reloc->lobj.gpu_offset); |
551ebd83 | 1497 | track->num_arrays = 1; |
513bcb46 | 1498 | track->vtx_size = r100_get_vtx_size(radeon_get_ib_value(p, idx + 2)); |
551ebd83 DA |
1499 | |
1500 | track->arrays[0].robj = reloc->robj; | |
1501 | track->arrays[0].esize = track->vtx_size; | |
1502 | ||
513bcb46 | 1503 | track->max_indx = radeon_get_ib_value(p, idx+1); |
551ebd83 | 1504 | |
513bcb46 | 1505 | track->vap_vf_cntl = radeon_get_ib_value(p, idx+3); |
551ebd83 DA |
1506 | track->immd_dwords = pkt->count - 1; |
1507 | r = r100_cs_track_check(p->rdev, track); | |
1508 | if (r) | |
1509 | return r; | |
771fe6b9 JG |
1510 | break; |
1511 | case PACKET3_3D_DRAW_IMMD: | |
513bcb46 | 1512 | if (((radeon_get_ib_value(p, idx + 1) >> 4) & 0x3) != 3) { |
551ebd83 DA |
1513 | DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); |
1514 | return -EINVAL; | |
1515 | } | |
cf57fc7a | 1516 | track->vtx_size = r100_get_vtx_size(radeon_get_ib_value(p, idx + 0)); |
513bcb46 | 1517 | track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1); |
551ebd83 DA |
1518 | track->immd_dwords = pkt->count - 1; |
1519 | r = r100_cs_track_check(p->rdev, track); | |
1520 | if (r) | |
1521 | return r; | |
1522 | break; | |
771fe6b9 JG |
1523 | /* triggers drawing using in-packet vertex data */ |
1524 | case PACKET3_3D_DRAW_IMMD_2: | |
513bcb46 | 1525 | if (((radeon_get_ib_value(p, idx) >> 4) & 0x3) != 3) { |
551ebd83 DA |
1526 | DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n"); |
1527 | return -EINVAL; | |
1528 | } | |
513bcb46 | 1529 | track->vap_vf_cntl = radeon_get_ib_value(p, idx); |
551ebd83 DA |
1530 | track->immd_dwords = pkt->count; |
1531 | r = r100_cs_track_check(p->rdev, track); | |
1532 | if (r) | |
1533 | return r; | |
1534 | break; | |
771fe6b9 JG |
1535 | /* triggers drawing using in-packet vertex data */ |
1536 | case PACKET3_3D_DRAW_VBUF_2: | |
513bcb46 | 1537 | track->vap_vf_cntl = radeon_get_ib_value(p, idx); |
551ebd83 DA |
1538 | r = r100_cs_track_check(p->rdev, track); |
1539 | if (r) | |
1540 | return r; | |
1541 | break; | |
771fe6b9 JG |
1542 | /* triggers drawing of vertex buffers setup elsewhere */ |
1543 | case PACKET3_3D_DRAW_INDX_2: | |
513bcb46 | 1544 | track->vap_vf_cntl = radeon_get_ib_value(p, idx); |
551ebd83 DA |
1545 | r = r100_cs_track_check(p->rdev, track); |
1546 | if (r) | |
1547 | return r; | |
1548 | break; | |
771fe6b9 JG |
1549 | /* triggers drawing using indices to vertex buffer */ |
1550 | case PACKET3_3D_DRAW_VBUF: | |
513bcb46 | 1551 | track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1); |
551ebd83 DA |
1552 | r = r100_cs_track_check(p->rdev, track); |
1553 | if (r) | |
1554 | return r; | |
1555 | break; | |
771fe6b9 JG |
1556 | /* triggers drawing of vertex buffers setup elsewhere */ |
1557 | case PACKET3_3D_DRAW_INDX: | |
513bcb46 | 1558 | track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1); |
551ebd83 DA |
1559 | r = r100_cs_track_check(p->rdev, track); |
1560 | if (r) | |
1561 | return r; | |
1562 | break; | |
771fe6b9 JG |
1563 | /* triggers drawing using indices to vertex buffer */ |
1564 | case PACKET3_NOP: | |
1565 | break; | |
1566 | default: | |
1567 | DRM_ERROR("Packet3 opcode %x not supported\n", pkt->opcode); | |
1568 | return -EINVAL; | |
1569 | } | |
1570 | return 0; | |
1571 | } | |
1572 | ||
1573 | int r100_cs_parse(struct radeon_cs_parser *p) | |
1574 | { | |
1575 | struct radeon_cs_packet pkt; | |
9f022ddf | 1576 | struct r100_cs_track *track; |
771fe6b9 JG |
1577 | int r; |
1578 | ||
9f022ddf JG |
1579 | track = kzalloc(sizeof(*track), GFP_KERNEL); |
1580 | r100_cs_track_clear(p->rdev, track); | |
1581 | p->track = track; | |
771fe6b9 JG |
1582 | do { |
1583 | r = r100_cs_packet_parse(p, &pkt, p->idx); | |
1584 | if (r) { | |
1585 | return r; | |
1586 | } | |
1587 | p->idx += pkt.count + 2; | |
1588 | switch (pkt.type) { | |
068a117c | 1589 | case PACKET_TYPE0: |
551ebd83 DA |
1590 | if (p->rdev->family >= CHIP_R200) |
1591 | r = r100_cs_parse_packet0(p, &pkt, | |
1592 | p->rdev->config.r100.reg_safe_bm, | |
1593 | p->rdev->config.r100.reg_safe_bm_size, | |
1594 | &r200_packet0_check); | |
1595 | else | |
1596 | r = r100_cs_parse_packet0(p, &pkt, | |
1597 | p->rdev->config.r100.reg_safe_bm, | |
1598 | p->rdev->config.r100.reg_safe_bm_size, | |
1599 | &r100_packet0_check); | |
068a117c JG |
1600 | break; |
1601 | case PACKET_TYPE2: | |
1602 | break; | |
1603 | case PACKET_TYPE3: | |
1604 | r = r100_packet3_check(p, &pkt); | |
1605 | break; | |
1606 | default: | |
1607 | DRM_ERROR("Unknown packet type %d !\n", | |
1608 | pkt.type); | |
1609 | return -EINVAL; | |
771fe6b9 JG |
1610 | } |
1611 | if (r) { | |
1612 | return r; | |
1613 | } | |
1614 | } while (p->idx < p->chunks[p->chunk_ib_idx].length_dw); | |
1615 | return 0; | |
1616 | } | |
1617 | ||
1618 | ||
1619 | /* | |
1620 | * Global GPU functions | |
1621 | */ | |
1622 | void r100_errata(struct radeon_device *rdev) | |
1623 | { | |
1624 | rdev->pll_errata = 0; | |
1625 | ||
1626 | if (rdev->family == CHIP_RV200 || rdev->family == CHIP_RS200) { | |
1627 | rdev->pll_errata |= CHIP_ERRATA_PLL_DUMMYREADS; | |
1628 | } | |
1629 | ||
1630 | if (rdev->family == CHIP_RV100 || | |
1631 | rdev->family == CHIP_RS100 || | |
1632 | rdev->family == CHIP_RS200) { | |
1633 | rdev->pll_errata |= CHIP_ERRATA_PLL_DELAY; | |
1634 | } | |
1635 | } | |
1636 | ||
1637 | /* Wait for vertical sync on primary CRTC */ | |
1638 | void r100_gpu_wait_for_vsync(struct radeon_device *rdev) | |
1639 | { | |
1640 | uint32_t crtc_gen_cntl, tmp; | |
1641 | int i; | |
1642 | ||
1643 | crtc_gen_cntl = RREG32(RADEON_CRTC_GEN_CNTL); | |
1644 | if ((crtc_gen_cntl & RADEON_CRTC_DISP_REQ_EN_B) || | |
1645 | !(crtc_gen_cntl & RADEON_CRTC_EN)) { | |
1646 | return; | |
1647 | } | |
1648 | /* Clear the CRTC_VBLANK_SAVE bit */ | |
1649 | WREG32(RADEON_CRTC_STATUS, RADEON_CRTC_VBLANK_SAVE_CLEAR); | |
1650 | for (i = 0; i < rdev->usec_timeout; i++) { | |
1651 | tmp = RREG32(RADEON_CRTC_STATUS); | |
1652 | if (tmp & RADEON_CRTC_VBLANK_SAVE) { | |
1653 | return; | |
1654 | } | |
1655 | DRM_UDELAY(1); | |
1656 | } | |
1657 | } | |
1658 | ||
1659 | /* Wait for vertical sync on secondary CRTC */ | |
1660 | void r100_gpu_wait_for_vsync2(struct radeon_device *rdev) | |
1661 | { | |
1662 | uint32_t crtc2_gen_cntl, tmp; | |
1663 | int i; | |
1664 | ||
1665 | crtc2_gen_cntl = RREG32(RADEON_CRTC2_GEN_CNTL); | |
1666 | if ((crtc2_gen_cntl & RADEON_CRTC2_DISP_REQ_EN_B) || | |
1667 | !(crtc2_gen_cntl & RADEON_CRTC2_EN)) | |
1668 | return; | |
1669 | ||
1670 | /* Clear the CRTC_VBLANK_SAVE bit */ | |
1671 | WREG32(RADEON_CRTC2_STATUS, RADEON_CRTC2_VBLANK_SAVE_CLEAR); | |
1672 | for (i = 0; i < rdev->usec_timeout; i++) { | |
1673 | tmp = RREG32(RADEON_CRTC2_STATUS); | |
1674 | if (tmp & RADEON_CRTC2_VBLANK_SAVE) { | |
1675 | return; | |
1676 | } | |
1677 | DRM_UDELAY(1); | |
1678 | } | |
1679 | } | |
1680 | ||
1681 | int r100_rbbm_fifo_wait_for_entry(struct radeon_device *rdev, unsigned n) | |
1682 | { | |
1683 | unsigned i; | |
1684 | uint32_t tmp; | |
1685 | ||
1686 | for (i = 0; i < rdev->usec_timeout; i++) { | |
1687 | tmp = RREG32(RADEON_RBBM_STATUS) & RADEON_RBBM_FIFOCNT_MASK; | |
1688 | if (tmp >= n) { | |
1689 | return 0; | |
1690 | } | |
1691 | DRM_UDELAY(1); | |
1692 | } | |
1693 | return -1; | |
1694 | } | |
1695 | ||
1696 | int r100_gui_wait_for_idle(struct radeon_device *rdev) | |
1697 | { | |
1698 | unsigned i; | |
1699 | uint32_t tmp; | |
1700 | ||
1701 | if (r100_rbbm_fifo_wait_for_entry(rdev, 64)) { | |
1702 | printk(KERN_WARNING "radeon: wait for empty RBBM fifo failed !" | |
1703 | " Bad things might happen.\n"); | |
1704 | } | |
1705 | for (i = 0; i < rdev->usec_timeout; i++) { | |
1706 | tmp = RREG32(RADEON_RBBM_STATUS); | |
4612dc97 | 1707 | if (!(tmp & RADEON_RBBM_ACTIVE)) { |
771fe6b9 JG |
1708 | return 0; |
1709 | } | |
1710 | DRM_UDELAY(1); | |
1711 | } | |
1712 | return -1; | |
1713 | } | |
1714 | ||
1715 | int r100_mc_wait_for_idle(struct radeon_device *rdev) | |
1716 | { | |
1717 | unsigned i; | |
1718 | uint32_t tmp; | |
1719 | ||
1720 | for (i = 0; i < rdev->usec_timeout; i++) { | |
1721 | /* read MC_STATUS */ | |
4612dc97 AD |
1722 | tmp = RREG32(RADEON_MC_STATUS); |
1723 | if (tmp & RADEON_MC_IDLE) { | |
771fe6b9 JG |
1724 | return 0; |
1725 | } | |
1726 | DRM_UDELAY(1); | |
1727 | } | |
1728 | return -1; | |
1729 | } | |
1730 | ||
1731 | void r100_gpu_init(struct radeon_device *rdev) | |
1732 | { | |
1733 | /* TODO: anythings to do here ? pipes ? */ | |
1734 | r100_hdp_reset(rdev); | |
1735 | } | |
1736 | ||
1737 | void r100_hdp_reset(struct radeon_device *rdev) | |
1738 | { | |
1739 | uint32_t tmp; | |
1740 | ||
1741 | tmp = RREG32(RADEON_HOST_PATH_CNTL) & RADEON_HDP_APER_CNTL; | |
1742 | tmp |= (7 << 28); | |
1743 | WREG32(RADEON_HOST_PATH_CNTL, tmp | RADEON_HDP_SOFT_RESET | RADEON_HDP_READ_BUFFER_INVALIDATE); | |
1744 | (void)RREG32(RADEON_HOST_PATH_CNTL); | |
1745 | udelay(200); | |
1746 | WREG32(RADEON_RBBM_SOFT_RESET, 0); | |
1747 | WREG32(RADEON_HOST_PATH_CNTL, tmp); | |
1748 | (void)RREG32(RADEON_HOST_PATH_CNTL); | |
1749 | } | |
1750 | ||
1751 | int r100_rb2d_reset(struct radeon_device *rdev) | |
1752 | { | |
1753 | uint32_t tmp; | |
1754 | int i; | |
1755 | ||
1756 | WREG32(RADEON_RBBM_SOFT_RESET, RADEON_SOFT_RESET_E2); | |
1757 | (void)RREG32(RADEON_RBBM_SOFT_RESET); | |
1758 | udelay(200); | |
1759 | WREG32(RADEON_RBBM_SOFT_RESET, 0); | |
1760 | /* Wait to prevent race in RBBM_STATUS */ | |
1761 | mdelay(1); | |
1762 | for (i = 0; i < rdev->usec_timeout; i++) { | |
1763 | tmp = RREG32(RADEON_RBBM_STATUS); | |
1764 | if (!(tmp & (1 << 26))) { | |
1765 | DRM_INFO("RB2D reset succeed (RBBM_STATUS=0x%08X)\n", | |
1766 | tmp); | |
1767 | return 0; | |
1768 | } | |
1769 | DRM_UDELAY(1); | |
1770 | } | |
1771 | tmp = RREG32(RADEON_RBBM_STATUS); | |
1772 | DRM_ERROR("Failed to reset RB2D (RBBM_STATUS=0x%08X)!\n", tmp); | |
1773 | return -1; | |
1774 | } | |
1775 | ||
1776 | int r100_gpu_reset(struct radeon_device *rdev) | |
1777 | { | |
1778 | uint32_t status; | |
1779 | ||
1780 | /* reset order likely matter */ | |
1781 | status = RREG32(RADEON_RBBM_STATUS); | |
1782 | /* reset HDP */ | |
1783 | r100_hdp_reset(rdev); | |
1784 | /* reset rb2d */ | |
1785 | if (status & ((1 << 17) | (1 << 18) | (1 << 27))) { | |
1786 | r100_rb2d_reset(rdev); | |
1787 | } | |
1788 | /* TODO: reset 3D engine */ | |
1789 | /* reset CP */ | |
1790 | status = RREG32(RADEON_RBBM_STATUS); | |
1791 | if (status & (1 << 16)) { | |
1792 | r100_cp_reset(rdev); | |
1793 | } | |
1794 | /* Check if GPU is idle */ | |
1795 | status = RREG32(RADEON_RBBM_STATUS); | |
4612dc97 | 1796 | if (status & RADEON_RBBM_ACTIVE) { |
771fe6b9 JG |
1797 | DRM_ERROR("Failed to reset GPU (RBBM_STATUS=0x%08X)\n", status); |
1798 | return -1; | |
1799 | } | |
1800 | DRM_INFO("GPU reset succeed (RBBM_STATUS=0x%08X)\n", status); | |
1801 | return 0; | |
1802 | } | |
1803 | ||
92cde00c AD |
1804 | void r100_set_common_regs(struct radeon_device *rdev) |
1805 | { | |
2739d49c AD |
1806 | struct drm_device *dev = rdev->ddev; |
1807 | bool force_dac2 = false; | |
1808 | ||
92cde00c AD |
1809 | /* set these so they don't interfere with anything */ |
1810 | WREG32(RADEON_OV0_SCALE_CNTL, 0); | |
1811 | WREG32(RADEON_SUBPIC_CNTL, 0); | |
1812 | WREG32(RADEON_VIPH_CONTROL, 0); | |
1813 | WREG32(RADEON_I2C_CNTL_1, 0); | |
1814 | WREG32(RADEON_DVI_I2C_CNTL_1, 0); | |
1815 | WREG32(RADEON_CAP0_TRIG_CNTL, 0); | |
1816 | WREG32(RADEON_CAP1_TRIG_CNTL, 0); | |
2739d49c AD |
1817 | |
1818 | /* always set up dac2 on rn50 and some rv100 as lots | |
1819 | * of servers seem to wire it up to a VGA port but | |
1820 | * don't report it in the bios connector | |
1821 | * table. | |
1822 | */ | |
1823 | switch (dev->pdev->device) { | |
1824 | /* RN50 */ | |
1825 | case 0x515e: | |
1826 | case 0x5969: | |
1827 | force_dac2 = true; | |
1828 | break; | |
1829 | /* RV100*/ | |
1830 | case 0x5159: | |
1831 | case 0x515a: | |
1832 | /* DELL triple head servers */ | |
1833 | if ((dev->pdev->subsystem_vendor == 0x1028 /* DELL */) && | |
1834 | ((dev->pdev->subsystem_device == 0x016c) || | |
1835 | (dev->pdev->subsystem_device == 0x016d) || | |
1836 | (dev->pdev->subsystem_device == 0x016e) || | |
1837 | (dev->pdev->subsystem_device == 0x016f) || | |
1838 | (dev->pdev->subsystem_device == 0x0170) || | |
1839 | (dev->pdev->subsystem_device == 0x017d) || | |
1840 | (dev->pdev->subsystem_device == 0x017e) || | |
1841 | (dev->pdev->subsystem_device == 0x0183) || | |
1842 | (dev->pdev->subsystem_device == 0x018a) || | |
1843 | (dev->pdev->subsystem_device == 0x019a))) | |
1844 | force_dac2 = true; | |
1845 | break; | |
1846 | } | |
1847 | ||
1848 | if (force_dac2) { | |
1849 | u32 disp_hw_debug = RREG32(RADEON_DISP_HW_DEBUG); | |
1850 | u32 tv_dac_cntl = RREG32(RADEON_TV_DAC_CNTL); | |
1851 | u32 dac2_cntl = RREG32(RADEON_DAC_CNTL2); | |
1852 | ||
1853 | /* For CRT on DAC2, don't turn it on if BIOS didn't | |
1854 | enable it, even it's detected. | |
1855 | */ | |
1856 | ||
1857 | /* force it to crtc0 */ | |
1858 | dac2_cntl &= ~RADEON_DAC2_DAC_CLK_SEL; | |
1859 | dac2_cntl |= RADEON_DAC2_DAC2_CLK_SEL; | |
1860 | disp_hw_debug |= RADEON_CRT2_DISP1_SEL; | |
1861 | ||
1862 | /* set up the TV DAC */ | |
1863 | tv_dac_cntl &= ~(RADEON_TV_DAC_PEDESTAL | | |
1864 | RADEON_TV_DAC_STD_MASK | | |
1865 | RADEON_TV_DAC_RDACPD | | |
1866 | RADEON_TV_DAC_GDACPD | | |
1867 | RADEON_TV_DAC_BDACPD | | |
1868 | RADEON_TV_DAC_BGADJ_MASK | | |
1869 | RADEON_TV_DAC_DACADJ_MASK); | |
1870 | tv_dac_cntl |= (RADEON_TV_DAC_NBLANK | | |
1871 | RADEON_TV_DAC_NHOLD | | |
1872 | RADEON_TV_DAC_STD_PS2 | | |
1873 | (0x58 << 16)); | |
1874 | ||
1875 | WREG32(RADEON_TV_DAC_CNTL, tv_dac_cntl); | |
1876 | WREG32(RADEON_DISP_HW_DEBUG, disp_hw_debug); | |
1877 | WREG32(RADEON_DAC_CNTL2, dac2_cntl); | |
1878 | } | |
92cde00c | 1879 | } |
771fe6b9 JG |
1880 | |
1881 | /* | |
1882 | * VRAM info | |
1883 | */ | |
1884 | static void r100_vram_get_type(struct radeon_device *rdev) | |
1885 | { | |
1886 | uint32_t tmp; | |
1887 | ||
1888 | rdev->mc.vram_is_ddr = false; | |
1889 | if (rdev->flags & RADEON_IS_IGP) | |
1890 | rdev->mc.vram_is_ddr = true; | |
1891 | else if (RREG32(RADEON_MEM_SDRAM_MODE_REG) & RADEON_MEM_CFG_TYPE_DDR) | |
1892 | rdev->mc.vram_is_ddr = true; | |
1893 | if ((rdev->family == CHIP_RV100) || | |
1894 | (rdev->family == CHIP_RS100) || | |
1895 | (rdev->family == CHIP_RS200)) { | |
1896 | tmp = RREG32(RADEON_MEM_CNTL); | |
1897 | if (tmp & RV100_HALF_MODE) { | |
1898 | rdev->mc.vram_width = 32; | |
1899 | } else { | |
1900 | rdev->mc.vram_width = 64; | |
1901 | } | |
1902 | if (rdev->flags & RADEON_SINGLE_CRTC) { | |
1903 | rdev->mc.vram_width /= 4; | |
1904 | rdev->mc.vram_is_ddr = true; | |
1905 | } | |
1906 | } else if (rdev->family <= CHIP_RV280) { | |
1907 | tmp = RREG32(RADEON_MEM_CNTL); | |
1908 | if (tmp & RADEON_MEM_NUM_CHANNELS_MASK) { | |
1909 | rdev->mc.vram_width = 128; | |
1910 | } else { | |
1911 | rdev->mc.vram_width = 64; | |
1912 | } | |
1913 | } else { | |
1914 | /* newer IGPs */ | |
1915 | rdev->mc.vram_width = 128; | |
1916 | } | |
1917 | } | |
1918 | ||
2a0f8918 | 1919 | static u32 r100_get_accessible_vram(struct radeon_device *rdev) |
771fe6b9 | 1920 | { |
2a0f8918 DA |
1921 | u32 aper_size; |
1922 | u8 byte; | |
1923 | ||
1924 | aper_size = RREG32(RADEON_CONFIG_APER_SIZE); | |
1925 | ||
1926 | /* Set HDP_APER_CNTL only on cards that are known not to be broken, | |
1927 | * that is has the 2nd generation multifunction PCI interface | |
1928 | */ | |
1929 | if (rdev->family == CHIP_RV280 || | |
1930 | rdev->family >= CHIP_RV350) { | |
1931 | WREG32_P(RADEON_HOST_PATH_CNTL, RADEON_HDP_APER_CNTL, | |
1932 | ~RADEON_HDP_APER_CNTL); | |
1933 | DRM_INFO("Generation 2 PCI interface, using max accessible memory\n"); | |
1934 | return aper_size * 2; | |
1935 | } | |
1936 | ||
1937 | /* Older cards have all sorts of funny issues to deal with. First | |
1938 | * check if it's a multifunction card by reading the PCI config | |
1939 | * header type... Limit those to one aperture size | |
1940 | */ | |
1941 | pci_read_config_byte(rdev->pdev, 0xe, &byte); | |
1942 | if (byte & 0x80) { | |
1943 | DRM_INFO("Generation 1 PCI interface in multifunction mode\n"); | |
1944 | DRM_INFO("Limiting VRAM to one aperture\n"); | |
1945 | return aper_size; | |
1946 | } | |
1947 | ||
1948 | /* Single function older card. We read HDP_APER_CNTL to see how the BIOS | |
1949 | * have set it up. We don't write this as it's broken on some ASICs but | |
1950 | * we expect the BIOS to have done the right thing (might be too optimistic...) | |
1951 | */ | |
1952 | if (RREG32(RADEON_HOST_PATH_CNTL) & RADEON_HDP_APER_CNTL) | |
1953 | return aper_size * 2; | |
1954 | return aper_size; | |
1955 | } | |
1956 | ||
1957 | void r100_vram_init_sizes(struct radeon_device *rdev) | |
1958 | { | |
1959 | u64 config_aper_size; | |
2a0f8918 | 1960 | |
d594e46a | 1961 | /* work out accessible VRAM */ |
d594e46a JG |
1962 | rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); |
1963 | rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); | |
51e5fcd3 JG |
1964 | rdev->mc.visible_vram_size = r100_get_accessible_vram(rdev); |
1965 | /* FIXME we don't use the second aperture yet when we could use it */ | |
1966 | if (rdev->mc.visible_vram_size > rdev->mc.aper_size) | |
1967 | rdev->mc.visible_vram_size = rdev->mc.aper_size; | |
2a0f8918 | 1968 | config_aper_size = RREG32(RADEON_CONFIG_APER_SIZE); |
771fe6b9 JG |
1969 | if (rdev->flags & RADEON_IS_IGP) { |
1970 | uint32_t tom; | |
1971 | /* read NB_TOM to get the amount of ram stolen for the GPU */ | |
1972 | tom = RREG32(RADEON_NB_TOM); | |
7a50f01a | 1973 | rdev->mc.real_vram_size = (((tom >> 16) - (tom & 0xffff) + 1) << 16); |
7a50f01a DA |
1974 | WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); |
1975 | rdev->mc.mc_vram_size = rdev->mc.real_vram_size; | |
771fe6b9 | 1976 | } else { |
7a50f01a | 1977 | rdev->mc.real_vram_size = RREG32(RADEON_CONFIG_MEMSIZE); |
771fe6b9 JG |
1978 | /* Some production boards of m6 will report 0 |
1979 | * if it's 8 MB | |
1980 | */ | |
7a50f01a DA |
1981 | if (rdev->mc.real_vram_size == 0) { |
1982 | rdev->mc.real_vram_size = 8192 * 1024; | |
1983 | WREG32(RADEON_CONFIG_MEMSIZE, rdev->mc.real_vram_size); | |
771fe6b9 | 1984 | } |
d594e46a JG |
1985 | /* Fix for RN50, M6, M7 with 8/16/32(??) MBs of VRAM - |
1986 | * Novell bug 204882 + along with lots of ubuntu ones | |
1987 | */ | |
7a50f01a DA |
1988 | if (config_aper_size > rdev->mc.real_vram_size) |
1989 | rdev->mc.mc_vram_size = config_aper_size; | |
1990 | else | |
1991 | rdev->mc.mc_vram_size = rdev->mc.real_vram_size; | |
771fe6b9 | 1992 | } |
d594e46a JG |
1993 | /* FIXME remove this once we support unmappable VRAM */ |
1994 | if (rdev->mc.mc_vram_size > rdev->mc.aper_size) { | |
7a50f01a | 1995 | rdev->mc.mc_vram_size = rdev->mc.aper_size; |
7a50f01a | 1996 | rdev->mc.real_vram_size = rdev->mc.aper_size; |
d594e46a | 1997 | } |
2a0f8918 DA |
1998 | } |
1999 | ||
28d52043 DA |
2000 | void r100_vga_set_state(struct radeon_device *rdev, bool state) |
2001 | { | |
2002 | uint32_t temp; | |
2003 | ||
2004 | temp = RREG32(RADEON_CONFIG_CNTL); | |
2005 | if (state == false) { | |
2006 | temp &= ~(1<<8); | |
2007 | temp |= (1<<9); | |
2008 | } else { | |
2009 | temp &= ~(1<<9); | |
2010 | } | |
2011 | WREG32(RADEON_CONFIG_CNTL, temp); | |
2012 | } | |
2013 | ||
d594e46a | 2014 | void r100_mc_init(struct radeon_device *rdev) |
2a0f8918 | 2015 | { |
d594e46a | 2016 | u64 base; |
2a0f8918 | 2017 | |
d594e46a | 2018 | r100_vram_get_type(rdev); |
2a0f8918 | 2019 | r100_vram_init_sizes(rdev); |
d594e46a JG |
2020 | base = rdev->mc.aper_base; |
2021 | if (rdev->flags & RADEON_IS_IGP) | |
2022 | base = (RREG32(RADEON_NB_TOM) & 0xffff) << 16; | |
2023 | radeon_vram_location(rdev, &rdev->mc, base); | |
2024 | if (!(rdev->flags & RADEON_IS_AGP)) | |
2025 | radeon_gtt_location(rdev, &rdev->mc); | |
771fe6b9 JG |
2026 | } |
2027 | ||
2028 | ||
2029 | /* | |
2030 | * Indirect registers accessor | |
2031 | */ | |
2032 | void r100_pll_errata_after_index(struct radeon_device *rdev) | |
2033 | { | |
2034 | if (!(rdev->pll_errata & CHIP_ERRATA_PLL_DUMMYREADS)) { | |
2035 | return; | |
2036 | } | |
2037 | (void)RREG32(RADEON_CLOCK_CNTL_DATA); | |
2038 | (void)RREG32(RADEON_CRTC_GEN_CNTL); | |
2039 | } | |
2040 | ||
2041 | static void r100_pll_errata_after_data(struct radeon_device *rdev) | |
2042 | { | |
2043 | /* This workarounds is necessary on RV100, RS100 and RS200 chips | |
2044 | * or the chip could hang on a subsequent access | |
2045 | */ | |
2046 | if (rdev->pll_errata & CHIP_ERRATA_PLL_DELAY) { | |
2047 | udelay(5000); | |
2048 | } | |
2049 | ||
2050 | /* This function is required to workaround a hardware bug in some (all?) | |
2051 | * revisions of the R300. This workaround should be called after every | |
2052 | * CLOCK_CNTL_INDEX register access. If not, register reads afterward | |
2053 | * may not be correct. | |
2054 | */ | |
2055 | if (rdev->pll_errata & CHIP_ERRATA_R300_CG) { | |
2056 | uint32_t save, tmp; | |
2057 | ||
2058 | save = RREG32(RADEON_CLOCK_CNTL_INDEX); | |
2059 | tmp = save & ~(0x3f | RADEON_PLL_WR_EN); | |
2060 | WREG32(RADEON_CLOCK_CNTL_INDEX, tmp); | |
2061 | tmp = RREG32(RADEON_CLOCK_CNTL_DATA); | |
2062 | WREG32(RADEON_CLOCK_CNTL_INDEX, save); | |
2063 | } | |
2064 | } | |
2065 | ||
2066 | uint32_t r100_pll_rreg(struct radeon_device *rdev, uint32_t reg) | |
2067 | { | |
2068 | uint32_t data; | |
2069 | ||
2070 | WREG8(RADEON_CLOCK_CNTL_INDEX, reg & 0x3f); | |
2071 | r100_pll_errata_after_index(rdev); | |
2072 | data = RREG32(RADEON_CLOCK_CNTL_DATA); | |
2073 | r100_pll_errata_after_data(rdev); | |
2074 | return data; | |
2075 | } | |
2076 | ||
2077 | void r100_pll_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v) | |
2078 | { | |
2079 | WREG8(RADEON_CLOCK_CNTL_INDEX, ((reg & 0x3f) | RADEON_PLL_WR_EN)); | |
2080 | r100_pll_errata_after_index(rdev); | |
2081 | WREG32(RADEON_CLOCK_CNTL_DATA, v); | |
2082 | r100_pll_errata_after_data(rdev); | |
2083 | } | |
2084 | ||
d4550907 | 2085 | void r100_set_safe_registers(struct radeon_device *rdev) |
068a117c | 2086 | { |
551ebd83 DA |
2087 | if (ASIC_IS_RN50(rdev)) { |
2088 | rdev->config.r100.reg_safe_bm = rn50_reg_safe_bm; | |
2089 | rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(rn50_reg_safe_bm); | |
2090 | } else if (rdev->family < CHIP_R200) { | |
2091 | rdev->config.r100.reg_safe_bm = r100_reg_safe_bm; | |
2092 | rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(r100_reg_safe_bm); | |
2093 | } else { | |
d4550907 | 2094 | r200_set_safe_registers(rdev); |
551ebd83 | 2095 | } |
068a117c JG |
2096 | } |
2097 | ||
771fe6b9 JG |
2098 | /* |
2099 | * Debugfs info | |
2100 | */ | |
2101 | #if defined(CONFIG_DEBUG_FS) | |
2102 | static int r100_debugfs_rbbm_info(struct seq_file *m, void *data) | |
2103 | { | |
2104 | struct drm_info_node *node = (struct drm_info_node *) m->private; | |
2105 | struct drm_device *dev = node->minor->dev; | |
2106 | struct radeon_device *rdev = dev->dev_private; | |
2107 | uint32_t reg, value; | |
2108 | unsigned i; | |
2109 | ||
2110 | seq_printf(m, "RBBM_STATUS 0x%08x\n", RREG32(RADEON_RBBM_STATUS)); | |
2111 | seq_printf(m, "RBBM_CMDFIFO_STAT 0x%08x\n", RREG32(0xE7C)); | |
2112 | seq_printf(m, "CP_STAT 0x%08x\n", RREG32(RADEON_CP_STAT)); | |
2113 | for (i = 0; i < 64; i++) { | |
2114 | WREG32(RADEON_RBBM_CMDFIFO_ADDR, i | 0x100); | |
2115 | reg = (RREG32(RADEON_RBBM_CMDFIFO_DATA) - 1) >> 2; | |
2116 | WREG32(RADEON_RBBM_CMDFIFO_ADDR, i); | |
2117 | value = RREG32(RADEON_RBBM_CMDFIFO_DATA); | |
2118 | seq_printf(m, "[0x%03X] 0x%04X=0x%08X\n", i, reg, value); | |
2119 | } | |
2120 | return 0; | |
2121 | } | |
2122 | ||
2123 | static int r100_debugfs_cp_ring_info(struct seq_file *m, void *data) | |
2124 | { | |
2125 | struct drm_info_node *node = (struct drm_info_node *) m->private; | |
2126 | struct drm_device *dev = node->minor->dev; | |
2127 | struct radeon_device *rdev = dev->dev_private; | |
2128 | uint32_t rdp, wdp; | |
2129 | unsigned count, i, j; | |
2130 | ||
2131 | radeon_ring_free_size(rdev); | |
2132 | rdp = RREG32(RADEON_CP_RB_RPTR); | |
2133 | wdp = RREG32(RADEON_CP_RB_WPTR); | |
2134 | count = (rdp + rdev->cp.ring_size - wdp) & rdev->cp.ptr_mask; | |
2135 | seq_printf(m, "CP_STAT 0x%08x\n", RREG32(RADEON_CP_STAT)); | |
2136 | seq_printf(m, "CP_RB_WPTR 0x%08x\n", wdp); | |
2137 | seq_printf(m, "CP_RB_RPTR 0x%08x\n", rdp); | |
2138 | seq_printf(m, "%u free dwords in ring\n", rdev->cp.ring_free_dw); | |
2139 | seq_printf(m, "%u dwords in ring\n", count); | |
2140 | for (j = 0; j <= count; j++) { | |
2141 | i = (rdp + j) & rdev->cp.ptr_mask; | |
2142 | seq_printf(m, "r[%04d]=0x%08x\n", i, rdev->cp.ring[i]); | |
2143 | } | |
2144 | return 0; | |
2145 | } | |
2146 | ||
2147 | ||
2148 | static int r100_debugfs_cp_csq_fifo(struct seq_file *m, void *data) | |
2149 | { | |
2150 | struct drm_info_node *node = (struct drm_info_node *) m->private; | |
2151 | struct drm_device *dev = node->minor->dev; | |
2152 | struct radeon_device *rdev = dev->dev_private; | |
2153 | uint32_t csq_stat, csq2_stat, tmp; | |
2154 | unsigned r_rptr, r_wptr, ib1_rptr, ib1_wptr, ib2_rptr, ib2_wptr; | |
2155 | unsigned i; | |
2156 | ||
2157 | seq_printf(m, "CP_STAT 0x%08x\n", RREG32(RADEON_CP_STAT)); | |
2158 | seq_printf(m, "CP_CSQ_MODE 0x%08x\n", RREG32(RADEON_CP_CSQ_MODE)); | |
2159 | csq_stat = RREG32(RADEON_CP_CSQ_STAT); | |
2160 | csq2_stat = RREG32(RADEON_CP_CSQ2_STAT); | |
2161 | r_rptr = (csq_stat >> 0) & 0x3ff; | |
2162 | r_wptr = (csq_stat >> 10) & 0x3ff; | |
2163 | ib1_rptr = (csq_stat >> 20) & 0x3ff; | |
2164 | ib1_wptr = (csq2_stat >> 0) & 0x3ff; | |
2165 | ib2_rptr = (csq2_stat >> 10) & 0x3ff; | |
2166 | ib2_wptr = (csq2_stat >> 20) & 0x3ff; | |
2167 | seq_printf(m, "CP_CSQ_STAT 0x%08x\n", csq_stat); | |
2168 | seq_printf(m, "CP_CSQ2_STAT 0x%08x\n", csq2_stat); | |
2169 | seq_printf(m, "Ring rptr %u\n", r_rptr); | |
2170 | seq_printf(m, "Ring wptr %u\n", r_wptr); | |
2171 | seq_printf(m, "Indirect1 rptr %u\n", ib1_rptr); | |
2172 | seq_printf(m, "Indirect1 wptr %u\n", ib1_wptr); | |
2173 | seq_printf(m, "Indirect2 rptr %u\n", ib2_rptr); | |
2174 | seq_printf(m, "Indirect2 wptr %u\n", ib2_wptr); | |
2175 | /* FIXME: 0, 128, 640 depends on fifo setup see cp_init_kms | |
2176 | * 128 = indirect1_start * 8 & 640 = indirect2_start * 8 */ | |
2177 | seq_printf(m, "Ring fifo:\n"); | |
2178 | for (i = 0; i < 256; i++) { | |
2179 | WREG32(RADEON_CP_CSQ_ADDR, i << 2); | |
2180 | tmp = RREG32(RADEON_CP_CSQ_DATA); | |
2181 | seq_printf(m, "rfifo[%04d]=0x%08X\n", i, tmp); | |
2182 | } | |
2183 | seq_printf(m, "Indirect1 fifo:\n"); | |
2184 | for (i = 256; i <= 512; i++) { | |
2185 | WREG32(RADEON_CP_CSQ_ADDR, i << 2); | |
2186 | tmp = RREG32(RADEON_CP_CSQ_DATA); | |
2187 | seq_printf(m, "ib1fifo[%04d]=0x%08X\n", i, tmp); | |
2188 | } | |
2189 | seq_printf(m, "Indirect2 fifo:\n"); | |
2190 | for (i = 640; i < ib1_wptr; i++) { | |
2191 | WREG32(RADEON_CP_CSQ_ADDR, i << 2); | |
2192 | tmp = RREG32(RADEON_CP_CSQ_DATA); | |
2193 | seq_printf(m, "ib2fifo[%04d]=0x%08X\n", i, tmp); | |
2194 | } | |
2195 | return 0; | |
2196 | } | |
2197 | ||
2198 | static int r100_debugfs_mc_info(struct seq_file *m, void *data) | |
2199 | { | |
2200 | struct drm_info_node *node = (struct drm_info_node *) m->private; | |
2201 | struct drm_device *dev = node->minor->dev; | |
2202 | struct radeon_device *rdev = dev->dev_private; | |
2203 | uint32_t tmp; | |
2204 | ||
2205 | tmp = RREG32(RADEON_CONFIG_MEMSIZE); | |
2206 | seq_printf(m, "CONFIG_MEMSIZE 0x%08x\n", tmp); | |
2207 | tmp = RREG32(RADEON_MC_FB_LOCATION); | |
2208 | seq_printf(m, "MC_FB_LOCATION 0x%08x\n", tmp); | |
2209 | tmp = RREG32(RADEON_BUS_CNTL); | |
2210 | seq_printf(m, "BUS_CNTL 0x%08x\n", tmp); | |
2211 | tmp = RREG32(RADEON_MC_AGP_LOCATION); | |
2212 | seq_printf(m, "MC_AGP_LOCATION 0x%08x\n", tmp); | |
2213 | tmp = RREG32(RADEON_AGP_BASE); | |
2214 | seq_printf(m, "AGP_BASE 0x%08x\n", tmp); | |
2215 | tmp = RREG32(RADEON_HOST_PATH_CNTL); | |
2216 | seq_printf(m, "HOST_PATH_CNTL 0x%08x\n", tmp); | |
2217 | tmp = RREG32(0x01D0); | |
2218 | seq_printf(m, "AIC_CTRL 0x%08x\n", tmp); | |
2219 | tmp = RREG32(RADEON_AIC_LO_ADDR); | |
2220 | seq_printf(m, "AIC_LO_ADDR 0x%08x\n", tmp); | |
2221 | tmp = RREG32(RADEON_AIC_HI_ADDR); | |
2222 | seq_printf(m, "AIC_HI_ADDR 0x%08x\n", tmp); | |
2223 | tmp = RREG32(0x01E4); | |
2224 | seq_printf(m, "AIC_TLB_ADDR 0x%08x\n", tmp); | |
2225 | return 0; | |
2226 | } | |
2227 | ||
2228 | static struct drm_info_list r100_debugfs_rbbm_list[] = { | |
2229 | {"r100_rbbm_info", r100_debugfs_rbbm_info, 0, NULL}, | |
2230 | }; | |
2231 | ||
2232 | static struct drm_info_list r100_debugfs_cp_list[] = { | |
2233 | {"r100_cp_ring_info", r100_debugfs_cp_ring_info, 0, NULL}, | |
2234 | {"r100_cp_csq_fifo", r100_debugfs_cp_csq_fifo, 0, NULL}, | |
2235 | }; | |
2236 | ||
2237 | static struct drm_info_list r100_debugfs_mc_info_list[] = { | |
2238 | {"r100_mc_info", r100_debugfs_mc_info, 0, NULL}, | |
2239 | }; | |
2240 | #endif | |
2241 | ||
2242 | int r100_debugfs_rbbm_init(struct radeon_device *rdev) | |
2243 | { | |
2244 | #if defined(CONFIG_DEBUG_FS) | |
2245 | return radeon_debugfs_add_files(rdev, r100_debugfs_rbbm_list, 1); | |
2246 | #else | |
2247 | return 0; | |
2248 | #endif | |
2249 | } | |
2250 | ||
2251 | int r100_debugfs_cp_init(struct radeon_device *rdev) | |
2252 | { | |
2253 | #if defined(CONFIG_DEBUG_FS) | |
2254 | return radeon_debugfs_add_files(rdev, r100_debugfs_cp_list, 2); | |
2255 | #else | |
2256 | return 0; | |
2257 | #endif | |
2258 | } | |
2259 | ||
2260 | int r100_debugfs_mc_info_init(struct radeon_device *rdev) | |
2261 | { | |
2262 | #if defined(CONFIG_DEBUG_FS) | |
2263 | return radeon_debugfs_add_files(rdev, r100_debugfs_mc_info_list, 1); | |
2264 | #else | |
2265 | return 0; | |
2266 | #endif | |
2267 | } | |
e024e110 DA |
2268 | |
2269 | int r100_set_surface_reg(struct radeon_device *rdev, int reg, | |
2270 | uint32_t tiling_flags, uint32_t pitch, | |
2271 | uint32_t offset, uint32_t obj_size) | |
2272 | { | |
2273 | int surf_index = reg * 16; | |
2274 | int flags = 0; | |
2275 | ||
2276 | /* r100/r200 divide by 16 */ | |
2277 | if (rdev->family < CHIP_R300) | |
2278 | flags = pitch / 16; | |
2279 | else | |
2280 | flags = pitch / 8; | |
2281 | ||
2282 | if (rdev->family <= CHIP_RS200) { | |
2283 | if ((tiling_flags & (RADEON_TILING_MACRO|RADEON_TILING_MICRO)) | |
2284 | == (RADEON_TILING_MACRO|RADEON_TILING_MICRO)) | |
2285 | flags |= RADEON_SURF_TILE_COLOR_BOTH; | |
2286 | if (tiling_flags & RADEON_TILING_MACRO) | |
2287 | flags |= RADEON_SURF_TILE_COLOR_MACRO; | |
2288 | } else if (rdev->family <= CHIP_RV280) { | |
2289 | if (tiling_flags & (RADEON_TILING_MACRO)) | |
2290 | flags |= R200_SURF_TILE_COLOR_MACRO; | |
2291 | if (tiling_flags & RADEON_TILING_MICRO) | |
2292 | flags |= R200_SURF_TILE_COLOR_MICRO; | |
2293 | } else { | |
2294 | if (tiling_flags & RADEON_TILING_MACRO) | |
2295 | flags |= R300_SURF_TILE_MACRO; | |
2296 | if (tiling_flags & RADEON_TILING_MICRO) | |
2297 | flags |= R300_SURF_TILE_MICRO; | |
2298 | } | |
2299 | ||
c88f9f0c MD |
2300 | if (tiling_flags & RADEON_TILING_SWAP_16BIT) |
2301 | flags |= RADEON_SURF_AP0_SWP_16BPP | RADEON_SURF_AP1_SWP_16BPP; | |
2302 | if (tiling_flags & RADEON_TILING_SWAP_32BIT) | |
2303 | flags |= RADEON_SURF_AP0_SWP_32BPP | RADEON_SURF_AP1_SWP_32BPP; | |
2304 | ||
e024e110 DA |
2305 | DRM_DEBUG("writing surface %d %d %x %x\n", reg, flags, offset, offset+obj_size-1); |
2306 | WREG32(RADEON_SURFACE0_INFO + surf_index, flags); | |
2307 | WREG32(RADEON_SURFACE0_LOWER_BOUND + surf_index, offset); | |
2308 | WREG32(RADEON_SURFACE0_UPPER_BOUND + surf_index, offset + obj_size - 1); | |
2309 | return 0; | |
2310 | } | |
2311 | ||
2312 | void r100_clear_surface_reg(struct radeon_device *rdev, int reg) | |
2313 | { | |
2314 | int surf_index = reg * 16; | |
2315 | WREG32(RADEON_SURFACE0_INFO + surf_index, 0); | |
2316 | } | |
c93bb85b JG |
2317 | |
2318 | void r100_bandwidth_update(struct radeon_device *rdev) | |
2319 | { | |
2320 | fixed20_12 trcd_ff, trp_ff, tras_ff, trbs_ff, tcas_ff; | |
2321 | fixed20_12 sclk_ff, mclk_ff, sclk_eff_ff, sclk_delay_ff; | |
2322 | fixed20_12 peak_disp_bw, mem_bw, pix_clk, pix_clk2, temp_ff, crit_point_ff; | |
2323 | uint32_t temp, data, mem_trcd, mem_trp, mem_tras; | |
2324 | fixed20_12 memtcas_ff[8] = { | |
2325 | fixed_init(1), | |
2326 | fixed_init(2), | |
2327 | fixed_init(3), | |
2328 | fixed_init(0), | |
2329 | fixed_init_half(1), | |
2330 | fixed_init_half(2), | |
2331 | fixed_init(0), | |
2332 | }; | |
2333 | fixed20_12 memtcas_rs480_ff[8] = { | |
2334 | fixed_init(0), | |
2335 | fixed_init(1), | |
2336 | fixed_init(2), | |
2337 | fixed_init(3), | |
2338 | fixed_init(0), | |
2339 | fixed_init_half(1), | |
2340 | fixed_init_half(2), | |
2341 | fixed_init_half(3), | |
2342 | }; | |
2343 | fixed20_12 memtcas2_ff[8] = { | |
2344 | fixed_init(0), | |
2345 | fixed_init(1), | |
2346 | fixed_init(2), | |
2347 | fixed_init(3), | |
2348 | fixed_init(4), | |
2349 | fixed_init(5), | |
2350 | fixed_init(6), | |
2351 | fixed_init(7), | |
2352 | }; | |
2353 | fixed20_12 memtrbs[8] = { | |
2354 | fixed_init(1), | |
2355 | fixed_init_half(1), | |
2356 | fixed_init(2), | |
2357 | fixed_init_half(2), | |
2358 | fixed_init(3), | |
2359 | fixed_init_half(3), | |
2360 | fixed_init(4), | |
2361 | fixed_init_half(4) | |
2362 | }; | |
2363 | fixed20_12 memtrbs_r4xx[8] = { | |
2364 | fixed_init(4), | |
2365 | fixed_init(5), | |
2366 | fixed_init(6), | |
2367 | fixed_init(7), | |
2368 | fixed_init(8), | |
2369 | fixed_init(9), | |
2370 | fixed_init(10), | |
2371 | fixed_init(11) | |
2372 | }; | |
2373 | fixed20_12 min_mem_eff; | |
2374 | fixed20_12 mc_latency_sclk, mc_latency_mclk, k1; | |
2375 | fixed20_12 cur_latency_mclk, cur_latency_sclk; | |
2376 | fixed20_12 disp_latency, disp_latency_overhead, disp_drain_rate, | |
2377 | disp_drain_rate2, read_return_rate; | |
2378 | fixed20_12 time_disp1_drop_priority; | |
2379 | int c; | |
2380 | int cur_size = 16; /* in octawords */ | |
2381 | int critical_point = 0, critical_point2; | |
2382 | /* uint32_t read_return_rate, time_disp1_drop_priority; */ | |
2383 | int stop_req, max_stop_req; | |
2384 | struct drm_display_mode *mode1 = NULL; | |
2385 | struct drm_display_mode *mode2 = NULL; | |
2386 | uint32_t pixel_bytes1 = 0; | |
2387 | uint32_t pixel_bytes2 = 0; | |
2388 | ||
2389 | if (rdev->mode_info.crtcs[0]->base.enabled) { | |
2390 | mode1 = &rdev->mode_info.crtcs[0]->base.mode; | |
2391 | pixel_bytes1 = rdev->mode_info.crtcs[0]->base.fb->bits_per_pixel / 8; | |
2392 | } | |
dfee5614 DA |
2393 | if (!(rdev->flags & RADEON_SINGLE_CRTC)) { |
2394 | if (rdev->mode_info.crtcs[1]->base.enabled) { | |
2395 | mode2 = &rdev->mode_info.crtcs[1]->base.mode; | |
2396 | pixel_bytes2 = rdev->mode_info.crtcs[1]->base.fb->bits_per_pixel / 8; | |
2397 | } | |
c93bb85b JG |
2398 | } |
2399 | ||
2400 | min_mem_eff.full = rfixed_const_8(0); | |
2401 | /* get modes */ | |
2402 | if ((rdev->disp_priority == 2) && ASIC_IS_R300(rdev)) { | |
2403 | uint32_t mc_init_misc_lat_timer = RREG32(R300_MC_INIT_MISC_LAT_TIMER); | |
2404 | mc_init_misc_lat_timer &= ~(R300_MC_DISP1R_INIT_LAT_MASK << R300_MC_DISP1R_INIT_LAT_SHIFT); | |
2405 | mc_init_misc_lat_timer &= ~(R300_MC_DISP0R_INIT_LAT_MASK << R300_MC_DISP0R_INIT_LAT_SHIFT); | |
2406 | /* check crtc enables */ | |
2407 | if (mode2) | |
2408 | mc_init_misc_lat_timer |= (1 << R300_MC_DISP1R_INIT_LAT_SHIFT); | |
2409 | if (mode1) | |
2410 | mc_init_misc_lat_timer |= (1 << R300_MC_DISP0R_INIT_LAT_SHIFT); | |
2411 | WREG32(R300_MC_INIT_MISC_LAT_TIMER, mc_init_misc_lat_timer); | |
2412 | } | |
2413 | ||
2414 | /* | |
2415 | * determine is there is enough bw for current mode | |
2416 | */ | |
2417 | mclk_ff.full = rfixed_const(rdev->clock.default_mclk); | |
2418 | temp_ff.full = rfixed_const(100); | |
2419 | mclk_ff.full = rfixed_div(mclk_ff, temp_ff); | |
2420 | sclk_ff.full = rfixed_const(rdev->clock.default_sclk); | |
2421 | sclk_ff.full = rfixed_div(sclk_ff, temp_ff); | |
2422 | ||
2423 | temp = (rdev->mc.vram_width / 8) * (rdev->mc.vram_is_ddr ? 2 : 1); | |
2424 | temp_ff.full = rfixed_const(temp); | |
2425 | mem_bw.full = rfixed_mul(mclk_ff, temp_ff); | |
2426 | ||
2427 | pix_clk.full = 0; | |
2428 | pix_clk2.full = 0; | |
2429 | peak_disp_bw.full = 0; | |
2430 | if (mode1) { | |
2431 | temp_ff.full = rfixed_const(1000); | |
2432 | pix_clk.full = rfixed_const(mode1->clock); /* convert to fixed point */ | |
2433 | pix_clk.full = rfixed_div(pix_clk, temp_ff); | |
2434 | temp_ff.full = rfixed_const(pixel_bytes1); | |
2435 | peak_disp_bw.full += rfixed_mul(pix_clk, temp_ff); | |
2436 | } | |
2437 | if (mode2) { | |
2438 | temp_ff.full = rfixed_const(1000); | |
2439 | pix_clk2.full = rfixed_const(mode2->clock); /* convert to fixed point */ | |
2440 | pix_clk2.full = rfixed_div(pix_clk2, temp_ff); | |
2441 | temp_ff.full = rfixed_const(pixel_bytes2); | |
2442 | peak_disp_bw.full += rfixed_mul(pix_clk2, temp_ff); | |
2443 | } | |
2444 | ||
2445 | mem_bw.full = rfixed_mul(mem_bw, min_mem_eff); | |
2446 | if (peak_disp_bw.full >= mem_bw.full) { | |
2447 | DRM_ERROR("You may not have enough display bandwidth for current mode\n" | |
2448 | "If you have flickering problem, try to lower resolution, refresh rate, or color depth\n"); | |
2449 | } | |
2450 | ||
2451 | /* Get values from the EXT_MEM_CNTL register...converting its contents. */ | |
2452 | temp = RREG32(RADEON_MEM_TIMING_CNTL); | |
2453 | if ((rdev->family == CHIP_RV100) || (rdev->flags & RADEON_IS_IGP)) { /* RV100, M6, IGPs */ | |
2454 | mem_trcd = ((temp >> 2) & 0x3) + 1; | |
2455 | mem_trp = ((temp & 0x3)) + 1; | |
2456 | mem_tras = ((temp & 0x70) >> 4) + 1; | |
2457 | } else if (rdev->family == CHIP_R300 || | |
2458 | rdev->family == CHIP_R350) { /* r300, r350 */ | |
2459 | mem_trcd = (temp & 0x7) + 1; | |
2460 | mem_trp = ((temp >> 8) & 0x7) + 1; | |
2461 | mem_tras = ((temp >> 11) & 0xf) + 4; | |
2462 | } else if (rdev->family == CHIP_RV350 || | |
2463 | rdev->family <= CHIP_RV380) { | |
2464 | /* rv3x0 */ | |
2465 | mem_trcd = (temp & 0x7) + 3; | |
2466 | mem_trp = ((temp >> 8) & 0x7) + 3; | |
2467 | mem_tras = ((temp >> 11) & 0xf) + 6; | |
2468 | } else if (rdev->family == CHIP_R420 || | |
2469 | rdev->family == CHIP_R423 || | |
2470 | rdev->family == CHIP_RV410) { | |
2471 | /* r4xx */ | |
2472 | mem_trcd = (temp & 0xf) + 3; | |
2473 | if (mem_trcd > 15) | |
2474 | mem_trcd = 15; | |
2475 | mem_trp = ((temp >> 8) & 0xf) + 3; | |
2476 | if (mem_trp > 15) | |
2477 | mem_trp = 15; | |
2478 | mem_tras = ((temp >> 12) & 0x1f) + 6; | |
2479 | if (mem_tras > 31) | |
2480 | mem_tras = 31; | |
2481 | } else { /* RV200, R200 */ | |
2482 | mem_trcd = (temp & 0x7) + 1; | |
2483 | mem_trp = ((temp >> 8) & 0x7) + 1; | |
2484 | mem_tras = ((temp >> 12) & 0xf) + 4; | |
2485 | } | |
2486 | /* convert to FF */ | |
2487 | trcd_ff.full = rfixed_const(mem_trcd); | |
2488 | trp_ff.full = rfixed_const(mem_trp); | |
2489 | tras_ff.full = rfixed_const(mem_tras); | |
2490 | ||
2491 | /* Get values from the MEM_SDRAM_MODE_REG register...converting its */ | |
2492 | temp = RREG32(RADEON_MEM_SDRAM_MODE_REG); | |
2493 | data = (temp & (7 << 20)) >> 20; | |
2494 | if ((rdev->family == CHIP_RV100) || rdev->flags & RADEON_IS_IGP) { | |
2495 | if (rdev->family == CHIP_RS480) /* don't think rs400 */ | |
2496 | tcas_ff = memtcas_rs480_ff[data]; | |
2497 | else | |
2498 | tcas_ff = memtcas_ff[data]; | |
2499 | } else | |
2500 | tcas_ff = memtcas2_ff[data]; | |
2501 | ||
2502 | if (rdev->family == CHIP_RS400 || | |
2503 | rdev->family == CHIP_RS480) { | |
2504 | /* extra cas latency stored in bits 23-25 0-4 clocks */ | |
2505 | data = (temp >> 23) & 0x7; | |
2506 | if (data < 5) | |
2507 | tcas_ff.full += rfixed_const(data); | |
2508 | } | |
2509 | ||
2510 | if (ASIC_IS_R300(rdev) && !(rdev->flags & RADEON_IS_IGP)) { | |
2511 | /* on the R300, Tcas is included in Trbs. | |
2512 | */ | |
2513 | temp = RREG32(RADEON_MEM_CNTL); | |
2514 | data = (R300_MEM_NUM_CHANNELS_MASK & temp); | |
2515 | if (data == 1) { | |
2516 | if (R300_MEM_USE_CD_CH_ONLY & temp) { | |
2517 | temp = RREG32(R300_MC_IND_INDEX); | |
2518 | temp &= ~R300_MC_IND_ADDR_MASK; | |
2519 | temp |= R300_MC_READ_CNTL_CD_mcind; | |
2520 | WREG32(R300_MC_IND_INDEX, temp); | |
2521 | temp = RREG32(R300_MC_IND_DATA); | |
2522 | data = (R300_MEM_RBS_POSITION_C_MASK & temp); | |
2523 | } else { | |
2524 | temp = RREG32(R300_MC_READ_CNTL_AB); | |
2525 | data = (R300_MEM_RBS_POSITION_A_MASK & temp); | |
2526 | } | |
2527 | } else { | |
2528 | temp = RREG32(R300_MC_READ_CNTL_AB); | |
2529 | data = (R300_MEM_RBS_POSITION_A_MASK & temp); | |
2530 | } | |
2531 | if (rdev->family == CHIP_RV410 || | |
2532 | rdev->family == CHIP_R420 || | |
2533 | rdev->family == CHIP_R423) | |
2534 | trbs_ff = memtrbs_r4xx[data]; | |
2535 | else | |
2536 | trbs_ff = memtrbs[data]; | |
2537 | tcas_ff.full += trbs_ff.full; | |
2538 | } | |
2539 | ||
2540 | sclk_eff_ff.full = sclk_ff.full; | |
2541 | ||
2542 | if (rdev->flags & RADEON_IS_AGP) { | |
2543 | fixed20_12 agpmode_ff; | |
2544 | agpmode_ff.full = rfixed_const(radeon_agpmode); | |
2545 | temp_ff.full = rfixed_const_666(16); | |
2546 | sclk_eff_ff.full -= rfixed_mul(agpmode_ff, temp_ff); | |
2547 | } | |
2548 | /* TODO PCIE lanes may affect this - agpmode == 16?? */ | |
2549 | ||
2550 | if (ASIC_IS_R300(rdev)) { | |
2551 | sclk_delay_ff.full = rfixed_const(250); | |
2552 | } else { | |
2553 | if ((rdev->family == CHIP_RV100) || | |
2554 | rdev->flags & RADEON_IS_IGP) { | |
2555 | if (rdev->mc.vram_is_ddr) | |
2556 | sclk_delay_ff.full = rfixed_const(41); | |
2557 | else | |
2558 | sclk_delay_ff.full = rfixed_const(33); | |
2559 | } else { | |
2560 | if (rdev->mc.vram_width == 128) | |
2561 | sclk_delay_ff.full = rfixed_const(57); | |
2562 | else | |
2563 | sclk_delay_ff.full = rfixed_const(41); | |
2564 | } | |
2565 | } | |
2566 | ||
2567 | mc_latency_sclk.full = rfixed_div(sclk_delay_ff, sclk_eff_ff); | |
2568 | ||
2569 | if (rdev->mc.vram_is_ddr) { | |
2570 | if (rdev->mc.vram_width == 32) { | |
2571 | k1.full = rfixed_const(40); | |
2572 | c = 3; | |
2573 | } else { | |
2574 | k1.full = rfixed_const(20); | |
2575 | c = 1; | |
2576 | } | |
2577 | } else { | |
2578 | k1.full = rfixed_const(40); | |
2579 | c = 3; | |
2580 | } | |
2581 | ||
2582 | temp_ff.full = rfixed_const(2); | |
2583 | mc_latency_mclk.full = rfixed_mul(trcd_ff, temp_ff); | |
2584 | temp_ff.full = rfixed_const(c); | |
2585 | mc_latency_mclk.full += rfixed_mul(tcas_ff, temp_ff); | |
2586 | temp_ff.full = rfixed_const(4); | |
2587 | mc_latency_mclk.full += rfixed_mul(tras_ff, temp_ff); | |
2588 | mc_latency_mclk.full += rfixed_mul(trp_ff, temp_ff); | |
2589 | mc_latency_mclk.full += k1.full; | |
2590 | ||
2591 | mc_latency_mclk.full = rfixed_div(mc_latency_mclk, mclk_ff); | |
2592 | mc_latency_mclk.full += rfixed_div(temp_ff, sclk_eff_ff); | |
2593 | ||
2594 | /* | |
2595 | HW cursor time assuming worst case of full size colour cursor. | |
2596 | */ | |
2597 | temp_ff.full = rfixed_const((2 * (cur_size - (rdev->mc.vram_is_ddr + 1)))); | |
2598 | temp_ff.full += trcd_ff.full; | |
2599 | if (temp_ff.full < tras_ff.full) | |
2600 | temp_ff.full = tras_ff.full; | |
2601 | cur_latency_mclk.full = rfixed_div(temp_ff, mclk_ff); | |
2602 | ||
2603 | temp_ff.full = rfixed_const(cur_size); | |
2604 | cur_latency_sclk.full = rfixed_div(temp_ff, sclk_eff_ff); | |
2605 | /* | |
2606 | Find the total latency for the display data. | |
2607 | */ | |
b5fc9010 | 2608 | disp_latency_overhead.full = rfixed_const(8); |
c93bb85b JG |
2609 | disp_latency_overhead.full = rfixed_div(disp_latency_overhead, sclk_ff); |
2610 | mc_latency_mclk.full += disp_latency_overhead.full + cur_latency_mclk.full; | |
2611 | mc_latency_sclk.full += disp_latency_overhead.full + cur_latency_sclk.full; | |
2612 | ||
2613 | if (mc_latency_mclk.full > mc_latency_sclk.full) | |
2614 | disp_latency.full = mc_latency_mclk.full; | |
2615 | else | |
2616 | disp_latency.full = mc_latency_sclk.full; | |
2617 | ||
2618 | /* setup Max GRPH_STOP_REQ default value */ | |
2619 | if (ASIC_IS_RV100(rdev)) | |
2620 | max_stop_req = 0x5c; | |
2621 | else | |
2622 | max_stop_req = 0x7c; | |
2623 | ||
2624 | if (mode1) { | |
2625 | /* CRTC1 | |
2626 | Set GRPH_BUFFER_CNTL register using h/w defined optimal values. | |
2627 | GRPH_STOP_REQ <= MIN[ 0x7C, (CRTC_H_DISP + 1) * (bit depth) / 0x10 ] | |
2628 | */ | |
2629 | stop_req = mode1->hdisplay * pixel_bytes1 / 16; | |
2630 | ||
2631 | if (stop_req > max_stop_req) | |
2632 | stop_req = max_stop_req; | |
2633 | ||
2634 | /* | |
2635 | Find the drain rate of the display buffer. | |
2636 | */ | |
2637 | temp_ff.full = rfixed_const((16/pixel_bytes1)); | |
2638 | disp_drain_rate.full = rfixed_div(pix_clk, temp_ff); | |
2639 | ||
2640 | /* | |
2641 | Find the critical point of the display buffer. | |
2642 | */ | |
2643 | crit_point_ff.full = rfixed_mul(disp_drain_rate, disp_latency); | |
2644 | crit_point_ff.full += rfixed_const_half(0); | |
2645 | ||
2646 | critical_point = rfixed_trunc(crit_point_ff); | |
2647 | ||
2648 | if (rdev->disp_priority == 2) { | |
2649 | critical_point = 0; | |
2650 | } | |
2651 | ||
2652 | /* | |
2653 | The critical point should never be above max_stop_req-4. Setting | |
2654 | GRPH_CRITICAL_CNTL = 0 will thus force high priority all the time. | |
2655 | */ | |
2656 | if (max_stop_req - critical_point < 4) | |
2657 | critical_point = 0; | |
2658 | ||
2659 | if (critical_point == 0 && mode2 && rdev->family == CHIP_R300) { | |
2660 | /* some R300 cards have problem with this set to 0, when CRTC2 is enabled.*/ | |
2661 | critical_point = 0x10; | |
2662 | } | |
2663 | ||
2664 | temp = RREG32(RADEON_GRPH_BUFFER_CNTL); | |
2665 | temp &= ~(RADEON_GRPH_STOP_REQ_MASK); | |
2666 | temp |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); | |
2667 | temp &= ~(RADEON_GRPH_START_REQ_MASK); | |
2668 | if ((rdev->family == CHIP_R350) && | |
2669 | (stop_req > 0x15)) { | |
2670 | stop_req -= 0x10; | |
2671 | } | |
2672 | temp |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); | |
2673 | temp |= RADEON_GRPH_BUFFER_SIZE; | |
2674 | temp &= ~(RADEON_GRPH_CRITICAL_CNTL | | |
2675 | RADEON_GRPH_CRITICAL_AT_SOF | | |
2676 | RADEON_GRPH_STOP_CNTL); | |
2677 | /* | |
2678 | Write the result into the register. | |
2679 | */ | |
2680 | WREG32(RADEON_GRPH_BUFFER_CNTL, ((temp & ~RADEON_GRPH_CRITICAL_POINT_MASK) | | |
2681 | (critical_point << RADEON_GRPH_CRITICAL_POINT_SHIFT))); | |
2682 | ||
2683 | #if 0 | |
2684 | if ((rdev->family == CHIP_RS400) || | |
2685 | (rdev->family == CHIP_RS480)) { | |
2686 | /* attempt to program RS400 disp regs correctly ??? */ | |
2687 | temp = RREG32(RS400_DISP1_REG_CNTL); | |
2688 | temp &= ~(RS400_DISP1_START_REQ_LEVEL_MASK | | |
2689 | RS400_DISP1_STOP_REQ_LEVEL_MASK); | |
2690 | WREG32(RS400_DISP1_REQ_CNTL1, (temp | | |
2691 | (critical_point << RS400_DISP1_START_REQ_LEVEL_SHIFT) | | |
2692 | (critical_point << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); | |
2693 | temp = RREG32(RS400_DMIF_MEM_CNTL1); | |
2694 | temp &= ~(RS400_DISP1_CRITICAL_POINT_START_MASK | | |
2695 | RS400_DISP1_CRITICAL_POINT_STOP_MASK); | |
2696 | WREG32(RS400_DMIF_MEM_CNTL1, (temp | | |
2697 | (critical_point << RS400_DISP1_CRITICAL_POINT_START_SHIFT) | | |
2698 | (critical_point << RS400_DISP1_CRITICAL_POINT_STOP_SHIFT))); | |
2699 | } | |
2700 | #endif | |
2701 | ||
2702 | DRM_DEBUG("GRPH_BUFFER_CNTL from to %x\n", | |
2703 | /* (unsigned int)info->SavedReg->grph_buffer_cntl, */ | |
2704 | (unsigned int)RREG32(RADEON_GRPH_BUFFER_CNTL)); | |
2705 | } | |
2706 | ||
2707 | if (mode2) { | |
2708 | u32 grph2_cntl; | |
2709 | stop_req = mode2->hdisplay * pixel_bytes2 / 16; | |
2710 | ||
2711 | if (stop_req > max_stop_req) | |
2712 | stop_req = max_stop_req; | |
2713 | ||
2714 | /* | |
2715 | Find the drain rate of the display buffer. | |
2716 | */ | |
2717 | temp_ff.full = rfixed_const((16/pixel_bytes2)); | |
2718 | disp_drain_rate2.full = rfixed_div(pix_clk2, temp_ff); | |
2719 | ||
2720 | grph2_cntl = RREG32(RADEON_GRPH2_BUFFER_CNTL); | |
2721 | grph2_cntl &= ~(RADEON_GRPH_STOP_REQ_MASK); | |
2722 | grph2_cntl |= (stop_req << RADEON_GRPH_STOP_REQ_SHIFT); | |
2723 | grph2_cntl &= ~(RADEON_GRPH_START_REQ_MASK); | |
2724 | if ((rdev->family == CHIP_R350) && | |
2725 | (stop_req > 0x15)) { | |
2726 | stop_req -= 0x10; | |
2727 | } | |
2728 | grph2_cntl |= (stop_req << RADEON_GRPH_START_REQ_SHIFT); | |
2729 | grph2_cntl |= RADEON_GRPH_BUFFER_SIZE; | |
2730 | grph2_cntl &= ~(RADEON_GRPH_CRITICAL_CNTL | | |
2731 | RADEON_GRPH_CRITICAL_AT_SOF | | |
2732 | RADEON_GRPH_STOP_CNTL); | |
2733 | ||
2734 | if ((rdev->family == CHIP_RS100) || | |
2735 | (rdev->family == CHIP_RS200)) | |
2736 | critical_point2 = 0; | |
2737 | else { | |
2738 | temp = (rdev->mc.vram_width * rdev->mc.vram_is_ddr + 1)/128; | |
2739 | temp_ff.full = rfixed_const(temp); | |
2740 | temp_ff.full = rfixed_mul(mclk_ff, temp_ff); | |
2741 | if (sclk_ff.full < temp_ff.full) | |
2742 | temp_ff.full = sclk_ff.full; | |
2743 | ||
2744 | read_return_rate.full = temp_ff.full; | |
2745 | ||
2746 | if (mode1) { | |
2747 | temp_ff.full = read_return_rate.full - disp_drain_rate.full; | |
2748 | time_disp1_drop_priority.full = rfixed_div(crit_point_ff, temp_ff); | |
2749 | } else { | |
2750 | time_disp1_drop_priority.full = 0; | |
2751 | } | |
2752 | crit_point_ff.full = disp_latency.full + time_disp1_drop_priority.full + disp_latency.full; | |
2753 | crit_point_ff.full = rfixed_mul(crit_point_ff, disp_drain_rate2); | |
2754 | crit_point_ff.full += rfixed_const_half(0); | |
2755 | ||
2756 | critical_point2 = rfixed_trunc(crit_point_ff); | |
2757 | ||
2758 | if (rdev->disp_priority == 2) { | |
2759 | critical_point2 = 0; | |
2760 | } | |
2761 | ||
2762 | if (max_stop_req - critical_point2 < 4) | |
2763 | critical_point2 = 0; | |
2764 | ||
2765 | } | |
2766 | ||
2767 | if (critical_point2 == 0 && rdev->family == CHIP_R300) { | |
2768 | /* some R300 cards have problem with this set to 0 */ | |
2769 | critical_point2 = 0x10; | |
2770 | } | |
2771 | ||
2772 | WREG32(RADEON_GRPH2_BUFFER_CNTL, ((grph2_cntl & ~RADEON_GRPH_CRITICAL_POINT_MASK) | | |
2773 | (critical_point2 << RADEON_GRPH_CRITICAL_POINT_SHIFT))); | |
2774 | ||
2775 | if ((rdev->family == CHIP_RS400) || | |
2776 | (rdev->family == CHIP_RS480)) { | |
2777 | #if 0 | |
2778 | /* attempt to program RS400 disp2 regs correctly ??? */ | |
2779 | temp = RREG32(RS400_DISP2_REQ_CNTL1); | |
2780 | temp &= ~(RS400_DISP2_START_REQ_LEVEL_MASK | | |
2781 | RS400_DISP2_STOP_REQ_LEVEL_MASK); | |
2782 | WREG32(RS400_DISP2_REQ_CNTL1, (temp | | |
2783 | (critical_point2 << RS400_DISP1_START_REQ_LEVEL_SHIFT) | | |
2784 | (critical_point2 << RS400_DISP1_STOP_REQ_LEVEL_SHIFT))); | |
2785 | temp = RREG32(RS400_DISP2_REQ_CNTL2); | |
2786 | temp &= ~(RS400_DISP2_CRITICAL_POINT_START_MASK | | |
2787 | RS400_DISP2_CRITICAL_POINT_STOP_MASK); | |
2788 | WREG32(RS400_DISP2_REQ_CNTL2, (temp | | |
2789 | (critical_point2 << RS400_DISP2_CRITICAL_POINT_START_SHIFT) | | |
2790 | (critical_point2 << RS400_DISP2_CRITICAL_POINT_STOP_SHIFT))); | |
2791 | #endif | |
2792 | WREG32(RS400_DISP2_REQ_CNTL1, 0x105DC1CC); | |
2793 | WREG32(RS400_DISP2_REQ_CNTL2, 0x2749D000); | |
2794 | WREG32(RS400_DMIF_MEM_CNTL1, 0x29CA71DC); | |
2795 | WREG32(RS400_DISP1_REQ_CNTL1, 0x28FBC3AC); | |
2796 | } | |
2797 | ||
2798 | DRM_DEBUG("GRPH2_BUFFER_CNTL from to %x\n", | |
2799 | (unsigned int)RREG32(RADEON_GRPH2_BUFFER_CNTL)); | |
2800 | } | |
2801 | } | |
551ebd83 DA |
2802 | |
2803 | static inline void r100_cs_track_texture_print(struct r100_cs_track_texture *t) | |
2804 | { | |
2805 | DRM_ERROR("pitch %d\n", t->pitch); | |
ceb776bc | 2806 | DRM_ERROR("use_pitch %d\n", t->use_pitch); |
551ebd83 | 2807 | DRM_ERROR("width %d\n", t->width); |
ceb776bc | 2808 | DRM_ERROR("width_11 %d\n", t->width_11); |
551ebd83 | 2809 | DRM_ERROR("height %d\n", t->height); |
ceb776bc | 2810 | DRM_ERROR("height_11 %d\n", t->height_11); |
551ebd83 DA |
2811 | DRM_ERROR("num levels %d\n", t->num_levels); |
2812 | DRM_ERROR("depth %d\n", t->txdepth); | |
2813 | DRM_ERROR("bpp %d\n", t->cpp); | |
2814 | DRM_ERROR("coordinate type %d\n", t->tex_coord_type); | |
2815 | DRM_ERROR("width round to power of 2 %d\n", t->roundup_w); | |
2816 | DRM_ERROR("height round to power of 2 %d\n", t->roundup_h); | |
d785d78b | 2817 | DRM_ERROR("compress format %d\n", t->compress_format); |
551ebd83 DA |
2818 | } |
2819 | ||
2820 | static int r100_cs_track_cube(struct radeon_device *rdev, | |
2821 | struct r100_cs_track *track, unsigned idx) | |
2822 | { | |
2823 | unsigned face, w, h; | |
4c788679 | 2824 | struct radeon_bo *cube_robj; |
551ebd83 DA |
2825 | unsigned long size; |
2826 | ||
2827 | for (face = 0; face < 5; face++) { | |
2828 | cube_robj = track->textures[idx].cube_info[face].robj; | |
2829 | w = track->textures[idx].cube_info[face].width; | |
2830 | h = track->textures[idx].cube_info[face].height; | |
2831 | ||
2832 | size = w * h; | |
2833 | size *= track->textures[idx].cpp; | |
2834 | ||
2835 | size += track->textures[idx].cube_info[face].offset; | |
2836 | ||
4c788679 | 2837 | if (size > radeon_bo_size(cube_robj)) { |
551ebd83 | 2838 | DRM_ERROR("Cube texture offset greater than object size %lu %lu\n", |
4c788679 | 2839 | size, radeon_bo_size(cube_robj)); |
551ebd83 DA |
2840 | r100_cs_track_texture_print(&track->textures[idx]); |
2841 | return -1; | |
2842 | } | |
2843 | } | |
2844 | return 0; | |
2845 | } | |
2846 | ||
d785d78b DA |
2847 | static int r100_track_compress_size(int compress_format, int w, int h) |
2848 | { | |
2849 | int block_width, block_height, block_bytes; | |
2850 | int wblocks, hblocks; | |
2851 | int min_wblocks; | |
2852 | int sz; | |
2853 | ||
2854 | block_width = 4; | |
2855 | block_height = 4; | |
2856 | ||
2857 | switch (compress_format) { | |
2858 | case R100_TRACK_COMP_DXT1: | |
2859 | block_bytes = 8; | |
2860 | min_wblocks = 4; | |
2861 | break; | |
2862 | default: | |
2863 | case R100_TRACK_COMP_DXT35: | |
2864 | block_bytes = 16; | |
2865 | min_wblocks = 2; | |
2866 | break; | |
2867 | } | |
2868 | ||
2869 | hblocks = (h + block_height - 1) / block_height; | |
2870 | wblocks = (w + block_width - 1) / block_width; | |
2871 | if (wblocks < min_wblocks) | |
2872 | wblocks = min_wblocks; | |
2873 | sz = wblocks * hblocks * block_bytes; | |
2874 | return sz; | |
2875 | } | |
2876 | ||
551ebd83 DA |
2877 | static int r100_cs_track_texture_check(struct radeon_device *rdev, |
2878 | struct r100_cs_track *track) | |
2879 | { | |
4c788679 | 2880 | struct radeon_bo *robj; |
551ebd83 DA |
2881 | unsigned long size; |
2882 | unsigned u, i, w, h; | |
2883 | int ret; | |
2884 | ||
2885 | for (u = 0; u < track->num_texture; u++) { | |
2886 | if (!track->textures[u].enabled) | |
2887 | continue; | |
2888 | robj = track->textures[u].robj; | |
2889 | if (robj == NULL) { | |
2890 | DRM_ERROR("No texture bound to unit %u\n", u); | |
2891 | return -EINVAL; | |
2892 | } | |
2893 | size = 0; | |
2894 | for (i = 0; i <= track->textures[u].num_levels; i++) { | |
2895 | if (track->textures[u].use_pitch) { | |
2896 | if (rdev->family < CHIP_R300) | |
2897 | w = (track->textures[u].pitch / track->textures[u].cpp) / (1 << i); | |
2898 | else | |
2899 | w = track->textures[u].pitch / (1 << i); | |
2900 | } else { | |
ceb776bc | 2901 | w = track->textures[u].width; |
551ebd83 DA |
2902 | if (rdev->family >= CHIP_RV515) |
2903 | w |= track->textures[u].width_11; | |
ceb776bc | 2904 | w = w / (1 << i); |
551ebd83 DA |
2905 | if (track->textures[u].roundup_w) |
2906 | w = roundup_pow_of_two(w); | |
2907 | } | |
ceb776bc | 2908 | h = track->textures[u].height; |
551ebd83 DA |
2909 | if (rdev->family >= CHIP_RV515) |
2910 | h |= track->textures[u].height_11; | |
ceb776bc | 2911 | h = h / (1 << i); |
551ebd83 DA |
2912 | if (track->textures[u].roundup_h) |
2913 | h = roundup_pow_of_two(h); | |
d785d78b DA |
2914 | if (track->textures[u].compress_format) { |
2915 | ||
2916 | size += r100_track_compress_size(track->textures[u].compress_format, w, h); | |
2917 | /* compressed textures are block based */ | |
2918 | } else | |
2919 | size += w * h; | |
551ebd83 DA |
2920 | } |
2921 | size *= track->textures[u].cpp; | |
d785d78b | 2922 | |
551ebd83 DA |
2923 | switch (track->textures[u].tex_coord_type) { |
2924 | case 0: | |
2925 | break; | |
2926 | case 1: | |
2927 | size *= (1 << track->textures[u].txdepth); | |
2928 | break; | |
2929 | case 2: | |
2930 | if (track->separate_cube) { | |
2931 | ret = r100_cs_track_cube(rdev, track, u); | |
2932 | if (ret) | |
2933 | return ret; | |
2934 | } else | |
2935 | size *= 6; | |
2936 | break; | |
2937 | default: | |
2938 | DRM_ERROR("Invalid texture coordinate type %u for unit " | |
2939 | "%u\n", track->textures[u].tex_coord_type, u); | |
2940 | return -EINVAL; | |
2941 | } | |
4c788679 | 2942 | if (size > radeon_bo_size(robj)) { |
551ebd83 | 2943 | DRM_ERROR("Texture of unit %u needs %lu bytes but is " |
4c788679 | 2944 | "%lu\n", u, size, radeon_bo_size(robj)); |
551ebd83 DA |
2945 | r100_cs_track_texture_print(&track->textures[u]); |
2946 | return -EINVAL; | |
2947 | } | |
2948 | } | |
2949 | return 0; | |
2950 | } | |
2951 | ||
2952 | int r100_cs_track_check(struct radeon_device *rdev, struct r100_cs_track *track) | |
2953 | { | |
2954 | unsigned i; | |
2955 | unsigned long size; | |
2956 | unsigned prim_walk; | |
2957 | unsigned nverts; | |
2958 | ||
2959 | for (i = 0; i < track->num_cb; i++) { | |
2960 | if (track->cb[i].robj == NULL) { | |
46c64d4b MO |
2961 | if (!(track->fastfill || track->color_channel_mask || |
2962 | track->blend_read_enable)) { | |
2963 | continue; | |
2964 | } | |
551ebd83 DA |
2965 | DRM_ERROR("[drm] No buffer for color buffer %d !\n", i); |
2966 | return -EINVAL; | |
2967 | } | |
2968 | size = track->cb[i].pitch * track->cb[i].cpp * track->maxy; | |
2969 | size += track->cb[i].offset; | |
4c788679 | 2970 | if (size > radeon_bo_size(track->cb[i].robj)) { |
551ebd83 DA |
2971 | DRM_ERROR("[drm] Buffer too small for color buffer %d " |
2972 | "(need %lu have %lu) !\n", i, size, | |
4c788679 | 2973 | radeon_bo_size(track->cb[i].robj)); |
551ebd83 DA |
2974 | DRM_ERROR("[drm] color buffer %d (%u %u %u %u)\n", |
2975 | i, track->cb[i].pitch, track->cb[i].cpp, | |
2976 | track->cb[i].offset, track->maxy); | |
2977 | return -EINVAL; | |
2978 | } | |
2979 | } | |
2980 | if (track->z_enabled) { | |
2981 | if (track->zb.robj == NULL) { | |
2982 | DRM_ERROR("[drm] No buffer for z buffer !\n"); | |
2983 | return -EINVAL; | |
2984 | } | |
2985 | size = track->zb.pitch * track->zb.cpp * track->maxy; | |
2986 | size += track->zb.offset; | |
4c788679 | 2987 | if (size > radeon_bo_size(track->zb.robj)) { |
551ebd83 DA |
2988 | DRM_ERROR("[drm] Buffer too small for z buffer " |
2989 | "(need %lu have %lu) !\n", size, | |
4c788679 | 2990 | radeon_bo_size(track->zb.robj)); |
551ebd83 DA |
2991 | DRM_ERROR("[drm] zbuffer (%u %u %u %u)\n", |
2992 | track->zb.pitch, track->zb.cpp, | |
2993 | track->zb.offset, track->maxy); | |
2994 | return -EINVAL; | |
2995 | } | |
2996 | } | |
2997 | prim_walk = (track->vap_vf_cntl >> 4) & 0x3; | |
2998 | nverts = (track->vap_vf_cntl >> 16) & 0xFFFF; | |
2999 | switch (prim_walk) { | |
3000 | case 1: | |
3001 | for (i = 0; i < track->num_arrays; i++) { | |
3002 | size = track->arrays[i].esize * track->max_indx * 4; | |
3003 | if (track->arrays[i].robj == NULL) { | |
3004 | DRM_ERROR("(PW %u) Vertex array %u no buffer " | |
3005 | "bound\n", prim_walk, i); | |
3006 | return -EINVAL; | |
3007 | } | |
4c788679 JG |
3008 | if (size > radeon_bo_size(track->arrays[i].robj)) { |
3009 | dev_err(rdev->dev, "(PW %u) Vertex array %u " | |
3010 | "need %lu dwords have %lu dwords\n", | |
3011 | prim_walk, i, size >> 2, | |
3012 | radeon_bo_size(track->arrays[i].robj) | |
3013 | >> 2); | |
551ebd83 DA |
3014 | DRM_ERROR("Max indices %u\n", track->max_indx); |
3015 | return -EINVAL; | |
3016 | } | |
3017 | } | |
3018 | break; | |
3019 | case 2: | |
3020 | for (i = 0; i < track->num_arrays; i++) { | |
3021 | size = track->arrays[i].esize * (nverts - 1) * 4; | |
3022 | if (track->arrays[i].robj == NULL) { | |
3023 | DRM_ERROR("(PW %u) Vertex array %u no buffer " | |
3024 | "bound\n", prim_walk, i); | |
3025 | return -EINVAL; | |
3026 | } | |
4c788679 JG |
3027 | if (size > radeon_bo_size(track->arrays[i].robj)) { |
3028 | dev_err(rdev->dev, "(PW %u) Vertex array %u " | |
3029 | "need %lu dwords have %lu dwords\n", | |
3030 | prim_walk, i, size >> 2, | |
3031 | radeon_bo_size(track->arrays[i].robj) | |
3032 | >> 2); | |
551ebd83 DA |
3033 | return -EINVAL; |
3034 | } | |
3035 | } | |
3036 | break; | |
3037 | case 3: | |
3038 | size = track->vtx_size * nverts; | |
3039 | if (size != track->immd_dwords) { | |
3040 | DRM_ERROR("IMMD draw %u dwors but needs %lu dwords\n", | |
3041 | track->immd_dwords, size); | |
3042 | DRM_ERROR("VAP_VF_CNTL.NUM_VERTICES %u, VTX_SIZE %u\n", | |
3043 | nverts, track->vtx_size); | |
3044 | return -EINVAL; | |
3045 | } | |
3046 | break; | |
3047 | default: | |
3048 | DRM_ERROR("[drm] Invalid primitive walk %d for VAP_VF_CNTL\n", | |
3049 | prim_walk); | |
3050 | return -EINVAL; | |
3051 | } | |
3052 | return r100_cs_track_texture_check(rdev, track); | |
3053 | } | |
3054 | ||
3055 | void r100_cs_track_clear(struct radeon_device *rdev, struct r100_cs_track *track) | |
3056 | { | |
3057 | unsigned i, face; | |
3058 | ||
3059 | if (rdev->family < CHIP_R300) { | |
3060 | track->num_cb = 1; | |
3061 | if (rdev->family <= CHIP_RS200) | |
3062 | track->num_texture = 3; | |
3063 | else | |
3064 | track->num_texture = 6; | |
3065 | track->maxy = 2048; | |
3066 | track->separate_cube = 1; | |
3067 | } else { | |
3068 | track->num_cb = 4; | |
3069 | track->num_texture = 16; | |
3070 | track->maxy = 4096; | |
3071 | track->separate_cube = 0; | |
3072 | } | |
3073 | ||
3074 | for (i = 0; i < track->num_cb; i++) { | |
3075 | track->cb[i].robj = NULL; | |
3076 | track->cb[i].pitch = 8192; | |
3077 | track->cb[i].cpp = 16; | |
3078 | track->cb[i].offset = 0; | |
3079 | } | |
3080 | track->z_enabled = true; | |
3081 | track->zb.robj = NULL; | |
3082 | track->zb.pitch = 8192; | |
3083 | track->zb.cpp = 4; | |
3084 | track->zb.offset = 0; | |
3085 | track->vtx_size = 0x7F; | |
3086 | track->immd_dwords = 0xFFFFFFFFUL; | |
3087 | track->num_arrays = 11; | |
3088 | track->max_indx = 0x00FFFFFFUL; | |
3089 | for (i = 0; i < track->num_arrays; i++) { | |
3090 | track->arrays[i].robj = NULL; | |
3091 | track->arrays[i].esize = 0x7F; | |
3092 | } | |
3093 | for (i = 0; i < track->num_texture; i++) { | |
d785d78b | 3094 | track->textures[i].compress_format = R100_TRACK_COMP_NONE; |
551ebd83 DA |
3095 | track->textures[i].pitch = 16536; |
3096 | track->textures[i].width = 16536; | |
3097 | track->textures[i].height = 16536; | |
3098 | track->textures[i].width_11 = 1 << 11; | |
3099 | track->textures[i].height_11 = 1 << 11; | |
3100 | track->textures[i].num_levels = 12; | |
3101 | if (rdev->family <= CHIP_RS200) { | |
3102 | track->textures[i].tex_coord_type = 0; | |
3103 | track->textures[i].txdepth = 0; | |
3104 | } else { | |
3105 | track->textures[i].txdepth = 16; | |
3106 | track->textures[i].tex_coord_type = 1; | |
3107 | } | |
3108 | track->textures[i].cpp = 64; | |
3109 | track->textures[i].robj = NULL; | |
3110 | /* CS IB emission code makes sure texture unit are disabled */ | |
3111 | track->textures[i].enabled = false; | |
3112 | track->textures[i].roundup_w = true; | |
3113 | track->textures[i].roundup_h = true; | |
3114 | if (track->separate_cube) | |
3115 | for (face = 0; face < 5; face++) { | |
3116 | track->textures[i].cube_info[face].robj = NULL; | |
3117 | track->textures[i].cube_info[face].width = 16536; | |
3118 | track->textures[i].cube_info[face].height = 16536; | |
3119 | track->textures[i].cube_info[face].offset = 0; | |
3120 | } | |
3121 | } | |
3122 | } | |
3ce0a23d JG |
3123 | |
3124 | int r100_ring_test(struct radeon_device *rdev) | |
3125 | { | |
3126 | uint32_t scratch; | |
3127 | uint32_t tmp = 0; | |
3128 | unsigned i; | |
3129 | int r; | |
3130 | ||
3131 | r = radeon_scratch_get(rdev, &scratch); | |
3132 | if (r) { | |
3133 | DRM_ERROR("radeon: cp failed to get scratch reg (%d).\n", r); | |
3134 | return r; | |
3135 | } | |
3136 | WREG32(scratch, 0xCAFEDEAD); | |
3137 | r = radeon_ring_lock(rdev, 2); | |
3138 | if (r) { | |
3139 | DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); | |
3140 | radeon_scratch_free(rdev, scratch); | |
3141 | return r; | |
3142 | } | |
3143 | radeon_ring_write(rdev, PACKET0(scratch, 0)); | |
3144 | radeon_ring_write(rdev, 0xDEADBEEF); | |
3145 | radeon_ring_unlock_commit(rdev); | |
3146 | for (i = 0; i < rdev->usec_timeout; i++) { | |
3147 | tmp = RREG32(scratch); | |
3148 | if (tmp == 0xDEADBEEF) { | |
3149 | break; | |
3150 | } | |
3151 | DRM_UDELAY(1); | |
3152 | } | |
3153 | if (i < rdev->usec_timeout) { | |
3154 | DRM_INFO("ring test succeeded in %d usecs\n", i); | |
3155 | } else { | |
3156 | DRM_ERROR("radeon: ring test failed (sracth(0x%04X)=0x%08X)\n", | |
3157 | scratch, tmp); | |
3158 | r = -EINVAL; | |
3159 | } | |
3160 | radeon_scratch_free(rdev, scratch); | |
3161 | return r; | |
3162 | } | |
3163 | ||
3164 | void r100_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) | |
3165 | { | |
3166 | radeon_ring_write(rdev, PACKET0(RADEON_CP_IB_BASE, 1)); | |
3167 | radeon_ring_write(rdev, ib->gpu_addr); | |
3168 | radeon_ring_write(rdev, ib->length_dw); | |
3169 | } | |
3170 | ||
3171 | int r100_ib_test(struct radeon_device *rdev) | |
3172 | { | |
3173 | struct radeon_ib *ib; | |
3174 | uint32_t scratch; | |
3175 | uint32_t tmp = 0; | |
3176 | unsigned i; | |
3177 | int r; | |
3178 | ||
3179 | r = radeon_scratch_get(rdev, &scratch); | |
3180 | if (r) { | |
3181 | DRM_ERROR("radeon: failed to get scratch reg (%d).\n", r); | |
3182 | return r; | |
3183 | } | |
3184 | WREG32(scratch, 0xCAFEDEAD); | |
3185 | r = radeon_ib_get(rdev, &ib); | |
3186 | if (r) { | |
3187 | return r; | |
3188 | } | |
3189 | ib->ptr[0] = PACKET0(scratch, 0); | |
3190 | ib->ptr[1] = 0xDEADBEEF; | |
3191 | ib->ptr[2] = PACKET2(0); | |
3192 | ib->ptr[3] = PACKET2(0); | |
3193 | ib->ptr[4] = PACKET2(0); | |
3194 | ib->ptr[5] = PACKET2(0); | |
3195 | ib->ptr[6] = PACKET2(0); | |
3196 | ib->ptr[7] = PACKET2(0); | |
3197 | ib->length_dw = 8; | |
3198 | r = radeon_ib_schedule(rdev, ib); | |
3199 | if (r) { | |
3200 | radeon_scratch_free(rdev, scratch); | |
3201 | radeon_ib_free(rdev, &ib); | |
3202 | return r; | |
3203 | } | |
3204 | r = radeon_fence_wait(ib->fence, false); | |
3205 | if (r) { | |
3206 | return r; | |
3207 | } | |
3208 | for (i = 0; i < rdev->usec_timeout; i++) { | |
3209 | tmp = RREG32(scratch); | |
3210 | if (tmp == 0xDEADBEEF) { | |
3211 | break; | |
3212 | } | |
3213 | DRM_UDELAY(1); | |
3214 | } | |
3215 | if (i < rdev->usec_timeout) { | |
3216 | DRM_INFO("ib test succeeded in %u usecs\n", i); | |
3217 | } else { | |
3218 | DRM_ERROR("radeon: ib test failed (sracth(0x%04X)=0x%08X)\n", | |
3219 | scratch, tmp); | |
3220 | r = -EINVAL; | |
3221 | } | |
3222 | radeon_scratch_free(rdev, scratch); | |
3223 | radeon_ib_free(rdev, &ib); | |
3224 | return r; | |
3225 | } | |
9f022ddf JG |
3226 | |
3227 | void r100_ib_fini(struct radeon_device *rdev) | |
3228 | { | |
3229 | radeon_ib_pool_fini(rdev); | |
3230 | } | |
3231 | ||
3232 | int r100_ib_init(struct radeon_device *rdev) | |
3233 | { | |
3234 | int r; | |
3235 | ||
3236 | r = radeon_ib_pool_init(rdev); | |
3237 | if (r) { | |
3238 | dev_err(rdev->dev, "failled initializing IB pool (%d).\n", r); | |
3239 | r100_ib_fini(rdev); | |
3240 | return r; | |
3241 | } | |
3242 | r = r100_ib_test(rdev); | |
3243 | if (r) { | |
3244 | dev_err(rdev->dev, "failled testing IB (%d).\n", r); | |
3245 | r100_ib_fini(rdev); | |
3246 | return r; | |
3247 | } | |
3248 | return 0; | |
3249 | } | |
3250 | ||
3251 | void r100_mc_stop(struct radeon_device *rdev, struct r100_mc_save *save) | |
3252 | { | |
3253 | /* Shutdown CP we shouldn't need to do that but better be safe than | |
3254 | * sorry | |
3255 | */ | |
3256 | rdev->cp.ready = false; | |
3257 | WREG32(R_000740_CP_CSQ_CNTL, 0); | |
3258 | ||
3259 | /* Save few CRTC registers */ | |
ca6ffc64 | 3260 | save->GENMO_WT = RREG8(R_0003C2_GENMO_WT); |
9f022ddf JG |
3261 | save->CRTC_EXT_CNTL = RREG32(R_000054_CRTC_EXT_CNTL); |
3262 | save->CRTC_GEN_CNTL = RREG32(R_000050_CRTC_GEN_CNTL); | |
3263 | save->CUR_OFFSET = RREG32(R_000260_CUR_OFFSET); | |
3264 | if (!(rdev->flags & RADEON_SINGLE_CRTC)) { | |
3265 | save->CRTC2_GEN_CNTL = RREG32(R_0003F8_CRTC2_GEN_CNTL); | |
3266 | save->CUR2_OFFSET = RREG32(R_000360_CUR2_OFFSET); | |
3267 | } | |
3268 | ||
3269 | /* Disable VGA aperture access */ | |
ca6ffc64 | 3270 | WREG8(R_0003C2_GENMO_WT, C_0003C2_VGA_RAM_EN & save->GENMO_WT); |
9f022ddf JG |
3271 | /* Disable cursor, overlay, crtc */ |
3272 | WREG32(R_000260_CUR_OFFSET, save->CUR_OFFSET | S_000260_CUR_LOCK(1)); | |
3273 | WREG32(R_000054_CRTC_EXT_CNTL, save->CRTC_EXT_CNTL | | |
3274 | S_000054_CRTC_DISPLAY_DIS(1)); | |
3275 | WREG32(R_000050_CRTC_GEN_CNTL, | |
3276 | (C_000050_CRTC_CUR_EN & save->CRTC_GEN_CNTL) | | |
3277 | S_000050_CRTC_DISP_REQ_EN_B(1)); | |
3278 | WREG32(R_000420_OV0_SCALE_CNTL, | |
3279 | C_000420_OV0_OVERLAY_EN & RREG32(R_000420_OV0_SCALE_CNTL)); | |
3280 | WREG32(R_000260_CUR_OFFSET, C_000260_CUR_LOCK & save->CUR_OFFSET); | |
3281 | if (!(rdev->flags & RADEON_SINGLE_CRTC)) { | |
3282 | WREG32(R_000360_CUR2_OFFSET, save->CUR2_OFFSET | | |
3283 | S_000360_CUR2_LOCK(1)); | |
3284 | WREG32(R_0003F8_CRTC2_GEN_CNTL, | |
3285 | (C_0003F8_CRTC2_CUR_EN & save->CRTC2_GEN_CNTL) | | |
3286 | S_0003F8_CRTC2_DISPLAY_DIS(1) | | |
3287 | S_0003F8_CRTC2_DISP_REQ_EN_B(1)); | |
3288 | WREG32(R_000360_CUR2_OFFSET, | |
3289 | C_000360_CUR2_LOCK & save->CUR2_OFFSET); | |
3290 | } | |
3291 | } | |
3292 | ||
3293 | void r100_mc_resume(struct radeon_device *rdev, struct r100_mc_save *save) | |
3294 | { | |
3295 | /* Update base address for crtc */ | |
d594e46a | 3296 | WREG32(R_00023C_DISPLAY_BASE_ADDR, rdev->mc.vram_start); |
9f022ddf | 3297 | if (!(rdev->flags & RADEON_SINGLE_CRTC)) { |
d594e46a | 3298 | WREG32(R_00033C_CRTC2_DISPLAY_BASE_ADDR, rdev->mc.vram_start); |
9f022ddf JG |
3299 | } |
3300 | /* Restore CRTC registers */ | |
ca6ffc64 | 3301 | WREG8(R_0003C2_GENMO_WT, save->GENMO_WT); |
9f022ddf JG |
3302 | WREG32(R_000054_CRTC_EXT_CNTL, save->CRTC_EXT_CNTL); |
3303 | WREG32(R_000050_CRTC_GEN_CNTL, save->CRTC_GEN_CNTL); | |
3304 | if (!(rdev->flags & RADEON_SINGLE_CRTC)) { | |
3305 | WREG32(R_0003F8_CRTC2_GEN_CNTL, save->CRTC2_GEN_CNTL); | |
3306 | } | |
3307 | } | |
ca6ffc64 JG |
3308 | |
3309 | void r100_vga_render_disable(struct radeon_device *rdev) | |
3310 | { | |
d4550907 | 3311 | u32 tmp; |
ca6ffc64 | 3312 | |
d4550907 | 3313 | tmp = RREG8(R_0003C2_GENMO_WT); |
ca6ffc64 JG |
3314 | WREG8(R_0003C2_GENMO_WT, C_0003C2_VGA_RAM_EN & tmp); |
3315 | } | |
d4550907 JG |
3316 | |
3317 | static void r100_debugfs(struct radeon_device *rdev) | |
3318 | { | |
3319 | int r; | |
3320 | ||
3321 | r = r100_debugfs_mc_info_init(rdev); | |
3322 | if (r) | |
3323 | dev_warn(rdev->dev, "Failed to create r100_mc debugfs file.\n"); | |
3324 | } | |
3325 | ||
3326 | static void r100_mc_program(struct radeon_device *rdev) | |
3327 | { | |
3328 | struct r100_mc_save save; | |
3329 | ||
3330 | /* Stops all mc clients */ | |
3331 | r100_mc_stop(rdev, &save); | |
3332 | if (rdev->flags & RADEON_IS_AGP) { | |
3333 | WREG32(R_00014C_MC_AGP_LOCATION, | |
3334 | S_00014C_MC_AGP_START(rdev->mc.gtt_start >> 16) | | |
3335 | S_00014C_MC_AGP_TOP(rdev->mc.gtt_end >> 16)); | |
3336 | WREG32(R_000170_AGP_BASE, lower_32_bits(rdev->mc.agp_base)); | |
3337 | if (rdev->family > CHIP_RV200) | |
3338 | WREG32(R_00015C_AGP_BASE_2, | |
3339 | upper_32_bits(rdev->mc.agp_base) & 0xff); | |
3340 | } else { | |
3341 | WREG32(R_00014C_MC_AGP_LOCATION, 0x0FFFFFFF); | |
3342 | WREG32(R_000170_AGP_BASE, 0); | |
3343 | if (rdev->family > CHIP_RV200) | |
3344 | WREG32(R_00015C_AGP_BASE_2, 0); | |
3345 | } | |
3346 | /* Wait for mc idle */ | |
3347 | if (r100_mc_wait_for_idle(rdev)) | |
3348 | dev_warn(rdev->dev, "Wait for MC idle timeout.\n"); | |
3349 | /* Program MC, should be a 32bits limited address space */ | |
3350 | WREG32(R_000148_MC_FB_LOCATION, | |
3351 | S_000148_MC_FB_START(rdev->mc.vram_start >> 16) | | |
3352 | S_000148_MC_FB_TOP(rdev->mc.vram_end >> 16)); | |
3353 | r100_mc_resume(rdev, &save); | |
3354 | } | |
3355 | ||
3356 | void r100_clock_startup(struct radeon_device *rdev) | |
3357 | { | |
3358 | u32 tmp; | |
3359 | ||
3360 | if (radeon_dynclks != -1 && radeon_dynclks) | |
3361 | radeon_legacy_set_clock_gating(rdev, 1); | |
3362 | /* We need to force on some of the block */ | |
3363 | tmp = RREG32_PLL(R_00000D_SCLK_CNTL); | |
3364 | tmp |= S_00000D_FORCE_CP(1) | S_00000D_FORCE_VIP(1); | |
3365 | if ((rdev->family == CHIP_RV250) || (rdev->family == CHIP_RV280)) | |
3366 | tmp |= S_00000D_FORCE_DISP1(1) | S_00000D_FORCE_DISP2(1); | |
3367 | WREG32_PLL(R_00000D_SCLK_CNTL, tmp); | |
3368 | } | |
3369 | ||
3370 | static int r100_startup(struct radeon_device *rdev) | |
3371 | { | |
3372 | int r; | |
3373 | ||
92cde00c AD |
3374 | /* set common regs */ |
3375 | r100_set_common_regs(rdev); | |
3376 | /* program mc */ | |
d4550907 JG |
3377 | r100_mc_program(rdev); |
3378 | /* Resume clock */ | |
3379 | r100_clock_startup(rdev); | |
3380 | /* Initialize GPU configuration (# pipes, ...) */ | |
3381 | r100_gpu_init(rdev); | |
3382 | /* Initialize GART (initialize after TTM so we can allocate | |
3383 | * memory through TTM but finalize after TTM) */ | |
17e15b0c | 3384 | r100_enable_bm(rdev); |
d4550907 JG |
3385 | if (rdev->flags & RADEON_IS_PCI) { |
3386 | r = r100_pci_gart_enable(rdev); | |
3387 | if (r) | |
3388 | return r; | |
3389 | } | |
3390 | /* Enable IRQ */ | |
d4550907 | 3391 | r100_irq_set(rdev); |
cafe6609 | 3392 | rdev->config.r100.hdp_cntl = RREG32(RADEON_HOST_PATH_CNTL); |
d4550907 JG |
3393 | /* 1M ring buffer */ |
3394 | r = r100_cp_init(rdev, 1024 * 1024); | |
3395 | if (r) { | |
3396 | dev_err(rdev->dev, "failled initializing CP (%d).\n", r); | |
3397 | return r; | |
3398 | } | |
3399 | r = r100_wb_init(rdev); | |
3400 | if (r) | |
3401 | dev_err(rdev->dev, "failled initializing WB (%d).\n", r); | |
3402 | r = r100_ib_init(rdev); | |
3403 | if (r) { | |
3404 | dev_err(rdev->dev, "failled initializing IB (%d).\n", r); | |
3405 | return r; | |
3406 | } | |
3407 | return 0; | |
3408 | } | |
3409 | ||
3410 | int r100_resume(struct radeon_device *rdev) | |
3411 | { | |
3412 | /* Make sur GART are not working */ | |
3413 | if (rdev->flags & RADEON_IS_PCI) | |
3414 | r100_pci_gart_disable(rdev); | |
3415 | /* Resume clock before doing reset */ | |
3416 | r100_clock_startup(rdev); | |
3417 | /* Reset gpu before posting otherwise ATOM will enter infinite loop */ | |
3418 | if (radeon_gpu_reset(rdev)) { | |
3419 | dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n", | |
3420 | RREG32(R_000E40_RBBM_STATUS), | |
3421 | RREG32(R_0007C0_CP_STAT)); | |
3422 | } | |
3423 | /* post */ | |
3424 | radeon_combios_asic_init(rdev->ddev); | |
3425 | /* Resume clock after posting */ | |
3426 | r100_clock_startup(rdev); | |
550e2d92 DA |
3427 | /* Initialize surface registers */ |
3428 | radeon_surface_init(rdev); | |
d4550907 JG |
3429 | return r100_startup(rdev); |
3430 | } | |
3431 | ||
3432 | int r100_suspend(struct radeon_device *rdev) | |
3433 | { | |
3434 | r100_cp_disable(rdev); | |
3435 | r100_wb_disable(rdev); | |
3436 | r100_irq_disable(rdev); | |
3437 | if (rdev->flags & RADEON_IS_PCI) | |
3438 | r100_pci_gart_disable(rdev); | |
3439 | return 0; | |
3440 | } | |
3441 | ||
3442 | void r100_fini(struct radeon_device *rdev) | |
3443 | { | |
d4550907 JG |
3444 | r100_cp_fini(rdev); |
3445 | r100_wb_fini(rdev); | |
3446 | r100_ib_fini(rdev); | |
3447 | radeon_gem_fini(rdev); | |
3448 | if (rdev->flags & RADEON_IS_PCI) | |
3449 | r100_pci_gart_fini(rdev); | |
d0269ed8 | 3450 | radeon_agp_fini(rdev); |
d4550907 JG |
3451 | radeon_irq_kms_fini(rdev); |
3452 | radeon_fence_driver_fini(rdev); | |
4c788679 | 3453 | radeon_bo_fini(rdev); |
d4550907 JG |
3454 | radeon_atombios_fini(rdev); |
3455 | kfree(rdev->bios); | |
3456 | rdev->bios = NULL; | |
3457 | } | |
3458 | ||
d4550907 JG |
3459 | int r100_init(struct radeon_device *rdev) |
3460 | { | |
3461 | int r; | |
3462 | ||
d4550907 JG |
3463 | /* Register debugfs file specific to this group of asics */ |
3464 | r100_debugfs(rdev); | |
3465 | /* Disable VGA */ | |
3466 | r100_vga_render_disable(rdev); | |
3467 | /* Initialize scratch registers */ | |
3468 | radeon_scratch_init(rdev); | |
3469 | /* Initialize surface registers */ | |
3470 | radeon_surface_init(rdev); | |
3471 | /* TODO: disable VGA need to use VGA request */ | |
3472 | /* BIOS*/ | |
3473 | if (!radeon_get_bios(rdev)) { | |
3474 | if (ASIC_IS_AVIVO(rdev)) | |
3475 | return -EINVAL; | |
3476 | } | |
3477 | if (rdev->is_atom_bios) { | |
3478 | dev_err(rdev->dev, "Expecting combios for RS400/RS480 GPU\n"); | |
3479 | return -EINVAL; | |
3480 | } else { | |
3481 | r = radeon_combios_init(rdev); | |
3482 | if (r) | |
3483 | return r; | |
3484 | } | |
3485 | /* Reset gpu before posting otherwise ATOM will enter infinite loop */ | |
3486 | if (radeon_gpu_reset(rdev)) { | |
3487 | dev_warn(rdev->dev, | |
3488 | "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n", | |
3489 | RREG32(R_000E40_RBBM_STATUS), | |
3490 | RREG32(R_0007C0_CP_STAT)); | |
3491 | } | |
3492 | /* check if cards are posted or not */ | |
72542d77 DA |
3493 | if (radeon_boot_test_post_card(rdev) == false) |
3494 | return -EINVAL; | |
d4550907 JG |
3495 | /* Set asic errata */ |
3496 | r100_errata(rdev); | |
3497 | /* Initialize clocks */ | |
3498 | radeon_get_clock_info(rdev->ddev); | |
6234077d RM |
3499 | /* Initialize power management */ |
3500 | radeon_pm_init(rdev); | |
d594e46a JG |
3501 | /* initialize AGP */ |
3502 | if (rdev->flags & RADEON_IS_AGP) { | |
3503 | r = radeon_agp_init(rdev); | |
3504 | if (r) { | |
3505 | radeon_agp_disable(rdev); | |
3506 | } | |
3507 | } | |
3508 | /* initialize VRAM */ | |
3509 | r100_mc_init(rdev); | |
d4550907 JG |
3510 | /* Fence driver */ |
3511 | r = radeon_fence_driver_init(rdev); | |
3512 | if (r) | |
3513 | return r; | |
3514 | r = radeon_irq_kms_init(rdev); | |
3515 | if (r) | |
3516 | return r; | |
3517 | /* Memory manager */ | |
4c788679 | 3518 | r = radeon_bo_init(rdev); |
d4550907 JG |
3519 | if (r) |
3520 | return r; | |
3521 | if (rdev->flags & RADEON_IS_PCI) { | |
3522 | r = r100_pci_gart_init(rdev); | |
3523 | if (r) | |
3524 | return r; | |
3525 | } | |
3526 | r100_set_safe_registers(rdev); | |
3527 | rdev->accel_working = true; | |
3528 | r = r100_startup(rdev); | |
3529 | if (r) { | |
3530 | /* Somethings want wront with the accel init stop accel */ | |
3531 | dev_err(rdev->dev, "Disabling GPU acceleration\n"); | |
d4550907 JG |
3532 | r100_cp_fini(rdev); |
3533 | r100_wb_fini(rdev); | |
3534 | r100_ib_fini(rdev); | |
655efd3d | 3535 | radeon_irq_kms_fini(rdev); |
d4550907 JG |
3536 | if (rdev->flags & RADEON_IS_PCI) |
3537 | r100_pci_gart_fini(rdev); | |
d4550907 JG |
3538 | rdev->accel_working = false; |
3539 | } | |
3540 | return 0; | |
3541 | } |