2 * Copyright 2008 Advanced Micro Devices, Inc.
3 * Copyright 2008 Red Hat Inc.
4 * Copyright 2009 Jerome Glisse.
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
24 * Authors: Dave Airlie
28 #include <linux/seq_file.h>
31 #include "radeon_reg.h"
33 #include "radeon_drm.h"
34 #include "r100_track.h"
37 #include "r300_reg_safe.h"
39 /* This files gather functions specifics to: r300,r350,rv350,rv370,rv380 */
42 * rv370,rv380 PCIE GART
44 static int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev);
46 void rv370_pcie_gart_tlb_flush(struct radeon_device *rdev)
51 /* Workaround HW bug do flush 2 times */
52 for (i = 0; i < 2; i++) {
53 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
54 WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp | RADEON_PCIE_TX_GART_INVALIDATE_TLB);
55 (void)RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
56 WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp);
61 int rv370_pcie_gart_set_page(struct radeon_device *rdev, int i, uint64_t addr)
63 void __iomem *ptr = (void *)rdev->gart.table.vram.ptr;
65 if (i < 0 || i > rdev->gart.num_gpu_pages) {
68 addr = (lower_32_bits(addr) >> 8) |
69 ((upper_32_bits(addr) & 0xff) << 24) |
71 /* on x86 we want this to be CPU endian, on powerpc
72 * on powerpc without HW swappers, it'll get swapped on way
73 * into VRAM - so no need for cpu_to_le32 on VRAM tables */
74 writel(addr, ((void __iomem *)ptr) + (i * 4));
78 int rv370_pcie_gart_init(struct radeon_device *rdev)
82 if (rdev->gart.table.vram.robj) {
83 WARN(1, "RV370 PCIE GART already initialized.\n");
86 /* Initialize common gart structure */
87 r = radeon_gart_init(rdev);
90 r = rv370_debugfs_pcie_gart_info_init(rdev);
92 DRM_ERROR("Failed to register debugfs file for PCIE gart !\n");
93 rdev->gart.table_size = rdev->gart.num_gpu_pages * 4;
94 rdev->asic->gart_tlb_flush = &rv370_pcie_gart_tlb_flush;
95 rdev->asic->gart_set_page = &rv370_pcie_gart_set_page;
96 return radeon_gart_table_vram_alloc(rdev);
99 int rv370_pcie_gart_enable(struct radeon_device *rdev)
105 if (rdev->gart.table.vram.robj == NULL) {
106 dev_err(rdev->dev, "No VRAM object for PCIE GART.\n");
109 r = radeon_gart_table_vram_pin(rdev);
112 /* discard memory request outside of configured range */
113 tmp = RADEON_PCIE_TX_GART_UNMAPPED_ACCESS_DISCARD;
114 WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp);
115 WREG32_PCIE(RADEON_PCIE_TX_GART_START_LO, rdev->mc.gtt_location);
116 tmp = rdev->mc.gtt_location + rdev->mc.gtt_size - RADEON_GPU_PAGE_SIZE;
117 WREG32_PCIE(RADEON_PCIE_TX_GART_END_LO, tmp);
118 WREG32_PCIE(RADEON_PCIE_TX_GART_START_HI, 0);
119 WREG32_PCIE(RADEON_PCIE_TX_GART_END_HI, 0);
120 table_addr = rdev->gart.table_addr;
121 WREG32_PCIE(RADEON_PCIE_TX_GART_BASE, table_addr);
122 /* FIXME: setup default page */
123 WREG32_PCIE(RADEON_PCIE_TX_DISCARD_RD_ADDR_LO, rdev->mc.vram_location);
124 WREG32_PCIE(RADEON_PCIE_TX_DISCARD_RD_ADDR_HI, 0);
126 WREG32_PCIE(0x18, 0);
127 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
128 tmp |= RADEON_PCIE_TX_GART_EN;
129 tmp |= RADEON_PCIE_TX_GART_UNMAPPED_ACCESS_DISCARD;
130 WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp);
131 rv370_pcie_gart_tlb_flush(rdev);
132 DRM_INFO("PCIE GART of %uM enabled (table at 0x%08X).\n",
133 (unsigned)(rdev->mc.gtt_size >> 20), table_addr);
134 rdev->gart.ready = true;
138 void rv370_pcie_gart_disable(struct radeon_device *rdev)
143 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
144 tmp |= RADEON_PCIE_TX_GART_UNMAPPED_ACCESS_DISCARD;
145 WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp & ~RADEON_PCIE_TX_GART_EN);
146 if (rdev->gart.table.vram.robj) {
147 r = radeon_bo_reserve(rdev->gart.table.vram.robj, false);
148 if (likely(r == 0)) {
149 radeon_bo_kunmap(rdev->gart.table.vram.robj);
150 radeon_bo_unpin(rdev->gart.table.vram.robj);
151 radeon_bo_unreserve(rdev->gart.table.vram.robj);
156 void rv370_pcie_gart_fini(struct radeon_device *rdev)
158 rv370_pcie_gart_disable(rdev);
159 radeon_gart_table_vram_free(rdev);
160 radeon_gart_fini(rdev);
163 void r300_fence_ring_emit(struct radeon_device *rdev,
164 struct radeon_fence *fence)
166 /* Who ever call radeon_fence_emit should call ring_lock and ask
167 * for enough space (today caller are ib schedule and buffer move) */
168 /* Write SC register so SC & US assert idle */
169 radeon_ring_write(rdev, PACKET0(0x43E0, 0));
170 radeon_ring_write(rdev, 0);
171 radeon_ring_write(rdev, PACKET0(0x43E4, 0));
172 radeon_ring_write(rdev, 0);
174 radeon_ring_write(rdev, PACKET0(0x4E4C, 0));
175 radeon_ring_write(rdev, (2 << 0));
176 radeon_ring_write(rdev, PACKET0(0x4F18, 0));
177 radeon_ring_write(rdev, (1 << 0));
178 /* Wait until IDLE & CLEAN */
179 radeon_ring_write(rdev, PACKET0(0x1720, 0));
180 radeon_ring_write(rdev, (1 << 17) | (1 << 16) | (1 << 9));
181 /* Emit fence sequence & fire IRQ */
182 radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0));
183 radeon_ring_write(rdev, fence->seq);
184 radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0));
185 radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
188 int r300_copy_dma(struct radeon_device *rdev,
192 struct radeon_fence *fence)
199 /* radeon pitch is /64 */
200 size = num_pages << PAGE_SHIFT;
201 num_loops = DIV_ROUND_UP(size, 0x1FFFFF);
202 r = radeon_ring_lock(rdev, num_loops * 4 + 64);
204 DRM_ERROR("radeon: moving bo (%d).\n", r);
207 /* Must wait for 2D idle & clean before DMA or hangs might happen */
208 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0 ));
209 radeon_ring_write(rdev, (1 << 16));
210 for (i = 0; i < num_loops; i++) {
212 if (cur_size > 0x1FFFFF) {
216 radeon_ring_write(rdev, PACKET0(0x720, 2));
217 radeon_ring_write(rdev, src_offset);
218 radeon_ring_write(rdev, dst_offset);
219 radeon_ring_write(rdev, cur_size | (1 << 31) | (1 << 30));
220 src_offset += cur_size;
221 dst_offset += cur_size;
223 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
224 radeon_ring_write(rdev, RADEON_WAIT_DMA_GUI_IDLE);
226 r = radeon_fence_emit(rdev, fence);
228 radeon_ring_unlock_commit(rdev);
232 void r300_ring_start(struct radeon_device *rdev)
234 unsigned gb_tile_config;
237 /* Sub pixel 1/12 so we can have 4K rendering according to doc */
238 gb_tile_config = (R300_ENABLE_TILING | R300_TILE_SIZE_16);
239 switch(rdev->num_gb_pipes) {
241 gb_tile_config |= R300_PIPE_COUNT_R300;
244 gb_tile_config |= R300_PIPE_COUNT_R420_3P;
247 gb_tile_config |= R300_PIPE_COUNT_R420;
251 gb_tile_config |= R300_PIPE_COUNT_RV350;
255 r = radeon_ring_lock(rdev, 64);
259 radeon_ring_write(rdev, PACKET0(RADEON_ISYNC_CNTL, 0));
260 radeon_ring_write(rdev,
261 RADEON_ISYNC_ANY2D_IDLE3D |
262 RADEON_ISYNC_ANY3D_IDLE2D |
263 RADEON_ISYNC_WAIT_IDLEGUI |
264 RADEON_ISYNC_CPSCRATCH_IDLEGUI);
265 radeon_ring_write(rdev, PACKET0(R300_GB_TILE_CONFIG, 0));
266 radeon_ring_write(rdev, gb_tile_config);
267 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
268 radeon_ring_write(rdev,
269 RADEON_WAIT_2D_IDLECLEAN |
270 RADEON_WAIT_3D_IDLECLEAN);
271 radeon_ring_write(rdev, PACKET0(0x170C, 0));
272 radeon_ring_write(rdev, 1 << 31);
273 radeon_ring_write(rdev, PACKET0(R300_GB_SELECT, 0));
274 radeon_ring_write(rdev, 0);
275 radeon_ring_write(rdev, PACKET0(R300_GB_ENABLE, 0));
276 radeon_ring_write(rdev, 0);
277 radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
278 radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
279 radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
280 radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
281 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
282 radeon_ring_write(rdev,
283 RADEON_WAIT_2D_IDLECLEAN |
284 RADEON_WAIT_3D_IDLECLEAN);
285 radeon_ring_write(rdev, PACKET0(R300_GB_AA_CONFIG, 0));
286 radeon_ring_write(rdev, 0);
287 radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
288 radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
289 radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
290 radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
291 radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS0, 0));
292 radeon_ring_write(rdev,
293 ((6 << R300_MS_X0_SHIFT) |
294 (6 << R300_MS_Y0_SHIFT) |
295 (6 << R300_MS_X1_SHIFT) |
296 (6 << R300_MS_Y1_SHIFT) |
297 (6 << R300_MS_X2_SHIFT) |
298 (6 << R300_MS_Y2_SHIFT) |
299 (6 << R300_MSBD0_Y_SHIFT) |
300 (6 << R300_MSBD0_X_SHIFT)));
301 radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS1, 0));
302 radeon_ring_write(rdev,
303 ((6 << R300_MS_X3_SHIFT) |
304 (6 << R300_MS_Y3_SHIFT) |
305 (6 << R300_MS_X4_SHIFT) |
306 (6 << R300_MS_Y4_SHIFT) |
307 (6 << R300_MS_X5_SHIFT) |
308 (6 << R300_MS_Y5_SHIFT) |
309 (6 << R300_MSBD1_SHIFT)));
310 radeon_ring_write(rdev, PACKET0(R300_GA_ENHANCE, 0));
311 radeon_ring_write(rdev, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL);
312 radeon_ring_write(rdev, PACKET0(R300_GA_POLY_MODE, 0));
313 radeon_ring_write(rdev,
314 R300_FRONT_PTYPE_TRIANGE | R300_BACK_PTYPE_TRIANGE);
315 radeon_ring_write(rdev, PACKET0(R300_GA_ROUND_MODE, 0));
316 radeon_ring_write(rdev,
317 R300_GEOMETRY_ROUND_NEAREST |
318 R300_COLOR_ROUND_NEAREST);
319 radeon_ring_unlock_commit(rdev);
322 void r300_errata(struct radeon_device *rdev)
324 rdev->pll_errata = 0;
326 if (rdev->family == CHIP_R300 &&
327 (RREG32(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) == RADEON_CFG_ATI_REV_A11) {
328 rdev->pll_errata |= CHIP_ERRATA_R300_CG;
332 int r300_mc_wait_for_idle(struct radeon_device *rdev)
337 for (i = 0; i < rdev->usec_timeout; i++) {
339 tmp = RREG32(0x0150);
340 if (tmp & (1 << 4)) {
348 void r300_gpu_init(struct radeon_device *rdev)
350 uint32_t gb_tile_config, tmp;
352 r100_hdp_reset(rdev);
353 /* FIXME: rv380 one pipes ? */
354 if ((rdev->family == CHIP_R300) || (rdev->family == CHIP_R350)) {
356 rdev->num_gb_pipes = 2;
358 /* rv350,rv370,rv380 */
359 rdev->num_gb_pipes = 1;
361 rdev->num_z_pipes = 1;
362 gb_tile_config = (R300_ENABLE_TILING | R300_TILE_SIZE_16);
363 switch (rdev->num_gb_pipes) {
365 gb_tile_config |= R300_PIPE_COUNT_R300;
368 gb_tile_config |= R300_PIPE_COUNT_R420_3P;
371 gb_tile_config |= R300_PIPE_COUNT_R420;
375 gb_tile_config |= R300_PIPE_COUNT_RV350;
378 WREG32(R300_GB_TILE_CONFIG, gb_tile_config);
380 if (r100_gui_wait_for_idle(rdev)) {
381 printk(KERN_WARNING "Failed to wait GUI idle while "
382 "programming pipes. Bad things might happen.\n");
385 tmp = RREG32(0x170C);
386 WREG32(0x170C, tmp | (1 << 31));
388 WREG32(R300_RB2D_DSTCACHE_MODE,
389 R300_DC_AUTOFLUSH_ENABLE |
390 R300_DC_DC_DISABLE_IGNORE_PE);
392 if (r100_gui_wait_for_idle(rdev)) {
393 printk(KERN_WARNING "Failed to wait GUI idle while "
394 "programming pipes. Bad things might happen.\n");
396 if (r300_mc_wait_for_idle(rdev)) {
397 printk(KERN_WARNING "Failed to wait MC idle while "
398 "programming pipes. Bad things might happen.\n");
400 DRM_INFO("radeon: %d quad pipes, %d Z pipes initialized.\n",
401 rdev->num_gb_pipes, rdev->num_z_pipes);
404 int r300_ga_reset(struct radeon_device *rdev)
410 reinit_cp = rdev->cp.ready;
411 rdev->cp.ready = false;
412 for (i = 0; i < rdev->usec_timeout; i++) {
413 WREG32(RADEON_CP_CSQ_MODE, 0);
414 WREG32(RADEON_CP_CSQ_CNTL, 0);
415 WREG32(RADEON_RBBM_SOFT_RESET, 0x32005);
416 (void)RREG32(RADEON_RBBM_SOFT_RESET);
418 WREG32(RADEON_RBBM_SOFT_RESET, 0);
419 /* Wait to prevent race in RBBM_STATUS */
421 tmp = RREG32(RADEON_RBBM_STATUS);
422 if (tmp & ((1 << 20) | (1 << 26))) {
423 DRM_ERROR("VAP & CP still busy (RBBM_STATUS=0x%08X)", tmp);
424 /* GA still busy soft reset it */
425 WREG32(0x429C, 0x200);
426 WREG32(R300_VAP_PVS_STATE_FLUSH_REG, 0);
431 /* Wait to prevent race in RBBM_STATUS */
433 tmp = RREG32(RADEON_RBBM_STATUS);
434 if (!(tmp & ((1 << 20) | (1 << 26)))) {
438 for (i = 0; i < rdev->usec_timeout; i++) {
439 tmp = RREG32(RADEON_RBBM_STATUS);
440 if (!(tmp & ((1 << 20) | (1 << 26)))) {
441 DRM_INFO("GA reset succeed (RBBM_STATUS=0x%08X)\n",
444 return r100_cp_init(rdev, rdev->cp.ring_size);
450 tmp = RREG32(RADEON_RBBM_STATUS);
451 DRM_ERROR("Failed to reset GA ! (RBBM_STATUS=0x%08X)\n", tmp);
455 int r300_gpu_reset(struct radeon_device *rdev)
459 /* reset order likely matter */
460 status = RREG32(RADEON_RBBM_STATUS);
462 r100_hdp_reset(rdev);
464 if (status & ((1 << 17) | (1 << 18) | (1 << 27))) {
465 r100_rb2d_reset(rdev);
468 if (status & ((1 << 20) | (1 << 26))) {
472 status = RREG32(RADEON_RBBM_STATUS);
473 if (status & (1 << 16)) {
476 /* Check if GPU is idle */
477 status = RREG32(RADEON_RBBM_STATUS);
478 if (status & (1 << 31)) {
479 DRM_ERROR("Failed to reset GPU (RBBM_STATUS=0x%08X)\n", status);
482 DRM_INFO("GPU reset succeed (RBBM_STATUS=0x%08X)\n", status);
488 * r300,r350,rv350,rv380 VRAM info
490 void r300_vram_info(struct radeon_device *rdev)
494 /* DDR for all card after R300 & IGP */
495 rdev->mc.vram_is_ddr = true;
496 tmp = RREG32(RADEON_MEM_CNTL);
497 if (tmp & R300_MEM_NUM_CHANNELS_MASK) {
498 rdev->mc.vram_width = 128;
500 rdev->mc.vram_width = 64;
503 r100_vram_init_sizes(rdev);
506 void rv370_set_pcie_lanes(struct radeon_device *rdev, int lanes)
508 uint32_t link_width_cntl, mask;
510 if (rdev->flags & RADEON_IS_IGP)
513 if (!(rdev->flags & RADEON_IS_PCIE))
516 /* FIXME wait for idle */
520 mask = RADEON_PCIE_LC_LINK_WIDTH_X0;
523 mask = RADEON_PCIE_LC_LINK_WIDTH_X1;
526 mask = RADEON_PCIE_LC_LINK_WIDTH_X2;
529 mask = RADEON_PCIE_LC_LINK_WIDTH_X4;
532 mask = RADEON_PCIE_LC_LINK_WIDTH_X8;
535 mask = RADEON_PCIE_LC_LINK_WIDTH_X12;
539 mask = RADEON_PCIE_LC_LINK_WIDTH_X16;
543 link_width_cntl = RREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
545 if ((link_width_cntl & RADEON_PCIE_LC_LINK_WIDTH_RD_MASK) ==
546 (mask << RADEON_PCIE_LC_LINK_WIDTH_RD_SHIFT))
549 link_width_cntl &= ~(RADEON_PCIE_LC_LINK_WIDTH_MASK |
550 RADEON_PCIE_LC_RECONFIG_NOW |
551 RADEON_PCIE_LC_RECONFIG_LATER |
552 RADEON_PCIE_LC_SHORT_RECONFIG_EN);
553 link_width_cntl |= mask;
554 WREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl);
555 WREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL, (link_width_cntl |
556 RADEON_PCIE_LC_RECONFIG_NOW));
558 /* wait for lane set to complete */
559 link_width_cntl = RREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
560 while (link_width_cntl == 0xffffffff)
561 link_width_cntl = RREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
565 #if defined(CONFIG_DEBUG_FS)
566 static int rv370_debugfs_pcie_gart_info(struct seq_file *m, void *data)
568 struct drm_info_node *node = (struct drm_info_node *) m->private;
569 struct drm_device *dev = node->minor->dev;
570 struct radeon_device *rdev = dev->dev_private;
573 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
574 seq_printf(m, "PCIE_TX_GART_CNTL 0x%08x\n", tmp);
575 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_BASE);
576 seq_printf(m, "PCIE_TX_GART_BASE 0x%08x\n", tmp);
577 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_START_LO);
578 seq_printf(m, "PCIE_TX_GART_START_LO 0x%08x\n", tmp);
579 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_START_HI);
580 seq_printf(m, "PCIE_TX_GART_START_HI 0x%08x\n", tmp);
581 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_END_LO);
582 seq_printf(m, "PCIE_TX_GART_END_LO 0x%08x\n", tmp);
583 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_END_HI);
584 seq_printf(m, "PCIE_TX_GART_END_HI 0x%08x\n", tmp);
585 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_ERROR);
586 seq_printf(m, "PCIE_TX_GART_ERROR 0x%08x\n", tmp);
590 static struct drm_info_list rv370_pcie_gart_info_list[] = {
591 {"rv370_pcie_gart_info", rv370_debugfs_pcie_gart_info, 0, NULL},
595 static int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
597 #if defined(CONFIG_DEBUG_FS)
598 return radeon_debugfs_add_files(rdev, rv370_pcie_gart_info_list, 1);
604 static int r300_packet0_check(struct radeon_cs_parser *p,
605 struct radeon_cs_packet *pkt,
606 unsigned idx, unsigned reg)
608 struct radeon_cs_reloc *reloc;
609 struct r100_cs_track *track;
610 volatile uint32_t *ib;
611 uint32_t tmp, tile_flags = 0;
617 track = (struct r100_cs_track *)p->track;
618 idx_value = radeon_get_ib_value(p, idx);
621 case AVIVO_D1MODE_VLINE_START_END:
622 case RADEON_CRTC_GUI_TRIG_VLINE:
623 r = r100_cs_packet_parse_vline(p);
625 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
627 r100_cs_dump_packet(p, pkt);
631 case RADEON_DST_PITCH_OFFSET:
632 case RADEON_SRC_PITCH_OFFSET:
633 r = r100_reloc_pitch_offset(p, pkt, idx, reg);
637 case R300_RB3D_COLOROFFSET0:
638 case R300_RB3D_COLOROFFSET1:
639 case R300_RB3D_COLOROFFSET2:
640 case R300_RB3D_COLOROFFSET3:
641 i = (reg - R300_RB3D_COLOROFFSET0) >> 2;
642 r = r100_cs_packet_next_reloc(p, &reloc);
644 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
646 r100_cs_dump_packet(p, pkt);
649 track->cb[i].robj = reloc->robj;
650 track->cb[i].offset = idx_value;
651 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
653 case R300_ZB_DEPTHOFFSET:
654 r = r100_cs_packet_next_reloc(p, &reloc);
656 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
658 r100_cs_dump_packet(p, pkt);
661 track->zb.robj = reloc->robj;
662 track->zb.offset = idx_value;
663 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
665 case R300_TX_OFFSET_0:
666 case R300_TX_OFFSET_0+4:
667 case R300_TX_OFFSET_0+8:
668 case R300_TX_OFFSET_0+12:
669 case R300_TX_OFFSET_0+16:
670 case R300_TX_OFFSET_0+20:
671 case R300_TX_OFFSET_0+24:
672 case R300_TX_OFFSET_0+28:
673 case R300_TX_OFFSET_0+32:
674 case R300_TX_OFFSET_0+36:
675 case R300_TX_OFFSET_0+40:
676 case R300_TX_OFFSET_0+44:
677 case R300_TX_OFFSET_0+48:
678 case R300_TX_OFFSET_0+52:
679 case R300_TX_OFFSET_0+56:
680 case R300_TX_OFFSET_0+60:
681 i = (reg - R300_TX_OFFSET_0) >> 2;
682 r = r100_cs_packet_next_reloc(p, &reloc);
684 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
686 r100_cs_dump_packet(p, pkt);
690 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
691 tile_flags |= R300_TXO_MACRO_TILE;
692 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
693 tile_flags |= R300_TXO_MICRO_TILE;
695 tmp = idx_value + ((u32)reloc->lobj.gpu_offset);
698 track->textures[i].robj = reloc->robj;
700 /* Tracked registers */
703 track->vap_vf_cntl = idx_value;
707 track->vtx_size = idx_value & 0x7F;
710 /* VAP_VF_MAX_VTX_INDX */
711 track->max_indx = idx_value & 0x00FFFFFFUL;
715 track->maxy = ((idx_value >> 13) & 0x1FFF) + 1;
716 if (p->rdev->family < CHIP_RV515) {
722 track->num_cb = ((idx_value >> 5) & 0x3) + 1;
728 /* RB3D_COLORPITCH0 */
729 /* RB3D_COLORPITCH1 */
730 /* RB3D_COLORPITCH2 */
731 /* RB3D_COLORPITCH3 */
732 r = r100_cs_packet_next_reloc(p, &reloc);
734 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
736 r100_cs_dump_packet(p, pkt);
740 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
741 tile_flags |= R300_COLOR_TILE_ENABLE;
742 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
743 tile_flags |= R300_COLOR_MICROTILE_ENABLE;
745 tmp = idx_value & ~(0x7 << 16);
749 i = (reg - 0x4E38) >> 2;
750 track->cb[i].pitch = idx_value & 0x3FFE;
751 switch (((idx_value >> 21) & 0xF)) {
755 track->cb[i].cpp = 1;
761 track->cb[i].cpp = 2;
764 track->cb[i].cpp = 4;
767 track->cb[i].cpp = 8;
770 track->cb[i].cpp = 16;
773 DRM_ERROR("Invalid color buffer format (%d) !\n",
774 ((idx_value >> 21) & 0xF));
781 track->z_enabled = true;
783 track->z_enabled = false;
788 switch ((idx_value & 0xF)) {
797 DRM_ERROR("Invalid z buffer format (%d) !\n",
804 r = r100_cs_packet_next_reloc(p, &reloc);
806 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
808 r100_cs_dump_packet(p, pkt);
812 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
813 tile_flags |= R300_DEPTHMACROTILE_ENABLE;
814 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
815 tile_flags |= R300_DEPTHMICROTILE_TILED;;
817 tmp = idx_value & ~(0x7 << 16);
821 track->zb.pitch = idx_value & 0x3FFC;
824 for (i = 0; i < 16; i++) {
827 enabled = !!(idx_value & (1 << i));
828 track->textures[i].enabled = enabled;
847 /* TX_FORMAT1_[0-15] */
848 i = (reg - 0x44C0) >> 2;
849 tmp = (idx_value >> 25) & 0x3;
850 track->textures[i].tex_coord_type = tmp;
851 switch ((idx_value & 0x1F)) {
852 case R300_TX_FORMAT_X8:
853 case R300_TX_FORMAT_Y4X4:
854 case R300_TX_FORMAT_Z3Y3X2:
855 track->textures[i].cpp = 1;
857 case R300_TX_FORMAT_X16:
858 case R300_TX_FORMAT_Y8X8:
859 case R300_TX_FORMAT_Z5Y6X5:
860 case R300_TX_FORMAT_Z6Y5X5:
861 case R300_TX_FORMAT_W4Z4Y4X4:
862 case R300_TX_FORMAT_W1Z5Y5X5:
863 case R300_TX_FORMAT_D3DMFT_CxV8U8:
864 case R300_TX_FORMAT_B8G8_B8G8:
865 case R300_TX_FORMAT_G8R8_G8B8:
866 track->textures[i].cpp = 2;
868 case R300_TX_FORMAT_Y16X16:
869 case R300_TX_FORMAT_Z11Y11X10:
870 case R300_TX_FORMAT_Z10Y11X11:
871 case R300_TX_FORMAT_W8Z8Y8X8:
872 case R300_TX_FORMAT_W2Z10Y10X10:
874 case R300_TX_FORMAT_FL_I32:
876 track->textures[i].cpp = 4;
878 case R300_TX_FORMAT_W16Z16Y16X16:
879 case R300_TX_FORMAT_FL_R16G16B16A16:
880 case R300_TX_FORMAT_FL_I32A32:
881 track->textures[i].cpp = 8;
883 case R300_TX_FORMAT_FL_R32G32B32A32:
884 track->textures[i].cpp = 16;
886 case R300_TX_FORMAT_DXT1:
887 track->textures[i].cpp = 1;
888 track->textures[i].compress_format = R100_TRACK_COMP_DXT1;
890 case R300_TX_FORMAT_DXT3:
891 case R300_TX_FORMAT_DXT5:
892 track->textures[i].cpp = 1;
893 track->textures[i].compress_format = R100_TRACK_COMP_DXT35;
896 DRM_ERROR("Invalid texture format %u\n",
918 /* TX_FILTER0_[0-15] */
919 i = (reg - 0x4400) >> 2;
920 tmp = idx_value & 0x7;
921 if (tmp == 2 || tmp == 4 || tmp == 6) {
922 track->textures[i].roundup_w = false;
924 tmp = (idx_value >> 3) & 0x7;
925 if (tmp == 2 || tmp == 4 || tmp == 6) {
926 track->textures[i].roundup_h = false;
945 /* TX_FORMAT2_[0-15] */
946 i = (reg - 0x4500) >> 2;
947 tmp = idx_value & 0x3FFF;
948 track->textures[i].pitch = tmp + 1;
949 if (p->rdev->family >= CHIP_RV515) {
950 tmp = ((idx_value >> 15) & 1) << 11;
951 track->textures[i].width_11 = tmp;
952 tmp = ((idx_value >> 16) & 1) << 11;
953 track->textures[i].height_11 = tmp;
972 /* TX_FORMAT0_[0-15] */
973 i = (reg - 0x4480) >> 2;
974 tmp = idx_value & 0x7FF;
975 track->textures[i].width = tmp + 1;
976 tmp = (idx_value >> 11) & 0x7FF;
977 track->textures[i].height = tmp + 1;
978 tmp = (idx_value >> 26) & 0xF;
979 track->textures[i].num_levels = tmp;
980 tmp = idx_value & (1 << 31);
981 track->textures[i].use_pitch = !!tmp;
982 tmp = (idx_value >> 22) & 0xF;
983 track->textures[i].txdepth = tmp;
985 case R300_ZB_ZPASS_ADDR:
986 r = r100_cs_packet_next_reloc(p, &reloc);
988 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
990 r100_cs_dump_packet(p, pkt);
993 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
996 /* valid register only on RV530 */
997 if (p->rdev->family == CHIP_RV530)
999 /* fallthrough do not move */
1001 printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
1008 static int r300_packet3_check(struct radeon_cs_parser *p,
1009 struct radeon_cs_packet *pkt)
1011 struct radeon_cs_reloc *reloc;
1012 struct r100_cs_track *track;
1013 volatile uint32_t *ib;
1019 track = (struct r100_cs_track *)p->track;
1020 switch(pkt->opcode) {
1021 case PACKET3_3D_LOAD_VBPNTR:
1022 r = r100_packet3_load_vbpntr(p, pkt, idx);
1026 case PACKET3_INDX_BUFFER:
1027 r = r100_cs_packet_next_reloc(p, &reloc);
1029 DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode);
1030 r100_cs_dump_packet(p, pkt);
1033 ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
1034 r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj);
1040 case PACKET3_3D_DRAW_IMMD:
1041 /* Number of dwords is vtx_size * (num_vertices - 1)
1042 * PRIM_WALK must be equal to 3 vertex data in embedded
1044 if (((radeon_get_ib_value(p, idx + 1) >> 4) & 0x3) != 3) {
1045 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1048 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1049 track->immd_dwords = pkt->count - 1;
1050 r = r100_cs_track_check(p->rdev, track);
1055 case PACKET3_3D_DRAW_IMMD_2:
1056 /* Number of dwords is vtx_size * (num_vertices - 1)
1057 * PRIM_WALK must be equal to 3 vertex data in embedded
1059 if (((radeon_get_ib_value(p, idx) >> 4) & 0x3) != 3) {
1060 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1063 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1064 track->immd_dwords = pkt->count;
1065 r = r100_cs_track_check(p->rdev, track);
1070 case PACKET3_3D_DRAW_VBUF:
1071 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1072 r = r100_cs_track_check(p->rdev, track);
1077 case PACKET3_3D_DRAW_VBUF_2:
1078 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1079 r = r100_cs_track_check(p->rdev, track);
1084 case PACKET3_3D_DRAW_INDX:
1085 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1086 r = r100_cs_track_check(p->rdev, track);
1091 case PACKET3_3D_DRAW_INDX_2:
1092 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1093 r = r100_cs_track_check(p->rdev, track);
1101 DRM_ERROR("Packet3 opcode %x not supported\n", pkt->opcode);
1107 int r300_cs_parse(struct radeon_cs_parser *p)
1109 struct radeon_cs_packet pkt;
1110 struct r100_cs_track *track;
1113 track = kzalloc(sizeof(*track), GFP_KERNEL);
1114 r100_cs_track_clear(p->rdev, track);
1117 r = r100_cs_packet_parse(p, &pkt, p->idx);
1121 p->idx += pkt.count + 2;
1124 r = r100_cs_parse_packet0(p, &pkt,
1125 p->rdev->config.r300.reg_safe_bm,
1126 p->rdev->config.r300.reg_safe_bm_size,
1127 &r300_packet0_check);
1132 r = r300_packet3_check(p, &pkt);
1135 DRM_ERROR("Unknown packet type %d !\n", pkt.type);
1141 } while (p->idx < p->chunks[p->chunk_ib_idx].length_dw);
1145 void r300_set_reg_safe(struct radeon_device *rdev)
1147 rdev->config.r300.reg_safe_bm = r300_reg_safe_bm;
1148 rdev->config.r300.reg_safe_bm_size = ARRAY_SIZE(r300_reg_safe_bm);
1151 void r300_mc_program(struct radeon_device *rdev)
1153 struct r100_mc_save save;
1156 r = r100_debugfs_mc_info_init(rdev);
1158 dev_err(rdev->dev, "Failed to create r100_mc debugfs file.\n");
1161 /* Stops all mc clients */
1162 r100_mc_stop(rdev, &save);
1163 if (rdev->flags & RADEON_IS_AGP) {
1164 WREG32(R_00014C_MC_AGP_LOCATION,
1165 S_00014C_MC_AGP_START(rdev->mc.gtt_start >> 16) |
1166 S_00014C_MC_AGP_TOP(rdev->mc.gtt_end >> 16));
1167 WREG32(R_000170_AGP_BASE, lower_32_bits(rdev->mc.agp_base));
1168 WREG32(R_00015C_AGP_BASE_2,
1169 upper_32_bits(rdev->mc.agp_base) & 0xff);
1171 WREG32(R_00014C_MC_AGP_LOCATION, 0x0FFFFFFF);
1172 WREG32(R_000170_AGP_BASE, 0);
1173 WREG32(R_00015C_AGP_BASE_2, 0);
1175 /* Wait for mc idle */
1176 if (r300_mc_wait_for_idle(rdev))
1177 DRM_INFO("Failed to wait MC idle before programming MC.\n");
1178 /* Program MC, should be a 32bits limited address space */
1179 WREG32(R_000148_MC_FB_LOCATION,
1180 S_000148_MC_FB_START(rdev->mc.vram_start >> 16) |
1181 S_000148_MC_FB_TOP(rdev->mc.vram_end >> 16));
1182 r100_mc_resume(rdev, &save);
1185 void r300_clock_startup(struct radeon_device *rdev)
1189 if (radeon_dynclks != -1 && radeon_dynclks)
1190 radeon_legacy_set_clock_gating(rdev, 1);
1191 /* We need to force on some of the block */
1192 tmp = RREG32_PLL(R_00000D_SCLK_CNTL);
1193 tmp |= S_00000D_FORCE_CP(1) | S_00000D_FORCE_VIP(1);
1194 if ((rdev->family == CHIP_RV350) || (rdev->family == CHIP_RV380))
1195 tmp |= S_00000D_FORCE_VAP(1);
1196 WREG32_PLL(R_00000D_SCLK_CNTL, tmp);
1199 static int r300_startup(struct radeon_device *rdev)
1203 /* set common regs */
1204 r100_set_common_regs(rdev);
1206 r300_mc_program(rdev);
1208 r300_clock_startup(rdev);
1209 /* Initialize GPU configuration (# pipes, ...) */
1210 r300_gpu_init(rdev);
1211 /* Initialize GART (initialize after TTM so we can allocate
1212 * memory through TTM but finalize after TTM) */
1213 if (rdev->flags & RADEON_IS_PCIE) {
1214 r = rv370_pcie_gart_enable(rdev);
1219 if (rdev->family == CHIP_R300 ||
1220 rdev->family == CHIP_R350 ||
1221 rdev->family == CHIP_RV350)
1222 r100_enable_bm(rdev);
1224 if (rdev->flags & RADEON_IS_PCI) {
1225 r = r100_pci_gart_enable(rdev);
1231 /* 1M ring buffer */
1232 r = r100_cp_init(rdev, 1024 * 1024);
1234 dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
1237 r = r100_wb_init(rdev);
1239 dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
1240 r = r100_ib_init(rdev);
1242 dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
1248 int r300_resume(struct radeon_device *rdev)
1250 /* Make sur GART are not working */
1251 if (rdev->flags & RADEON_IS_PCIE)
1252 rv370_pcie_gart_disable(rdev);
1253 if (rdev->flags & RADEON_IS_PCI)
1254 r100_pci_gart_disable(rdev);
1255 /* Resume clock before doing reset */
1256 r300_clock_startup(rdev);
1257 /* Reset gpu before posting otherwise ATOM will enter infinite loop */
1258 if (radeon_gpu_reset(rdev)) {
1259 dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
1260 RREG32(R_000E40_RBBM_STATUS),
1261 RREG32(R_0007C0_CP_STAT));
1264 radeon_combios_asic_init(rdev->ddev);
1265 /* Resume clock after posting */
1266 r300_clock_startup(rdev);
1267 /* Initialize surface registers */
1268 radeon_surface_init(rdev);
1269 return r300_startup(rdev);
1272 int r300_suspend(struct radeon_device *rdev)
1274 r100_cp_disable(rdev);
1275 r100_wb_disable(rdev);
1276 r100_irq_disable(rdev);
1277 if (rdev->flags & RADEON_IS_PCIE)
1278 rv370_pcie_gart_disable(rdev);
1279 if (rdev->flags & RADEON_IS_PCI)
1280 r100_pci_gart_disable(rdev);
1284 void r300_fini(struct radeon_device *rdev)
1290 radeon_gem_fini(rdev);
1291 if (rdev->flags & RADEON_IS_PCIE)
1292 rv370_pcie_gart_fini(rdev);
1293 if (rdev->flags & RADEON_IS_PCI)
1294 r100_pci_gart_fini(rdev);
1295 radeon_irq_kms_fini(rdev);
1296 radeon_fence_driver_fini(rdev);
1297 radeon_bo_fini(rdev);
1298 radeon_atombios_fini(rdev);
1303 int r300_init(struct radeon_device *rdev)
1308 r100_vga_render_disable(rdev);
1309 /* Initialize scratch registers */
1310 radeon_scratch_init(rdev);
1311 /* Initialize surface registers */
1312 radeon_surface_init(rdev);
1313 /* TODO: disable VGA need to use VGA request */
1315 if (!radeon_get_bios(rdev)) {
1316 if (ASIC_IS_AVIVO(rdev))
1319 if (rdev->is_atom_bios) {
1320 dev_err(rdev->dev, "Expecting combios for RS400/RS480 GPU\n");
1323 r = radeon_combios_init(rdev);
1327 /* Reset gpu before posting otherwise ATOM will enter infinite loop */
1328 if (radeon_gpu_reset(rdev)) {
1330 "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
1331 RREG32(R_000E40_RBBM_STATUS),
1332 RREG32(R_0007C0_CP_STAT));
1334 /* check if cards are posted or not */
1335 if (radeon_boot_test_post_card(rdev) == false)
1337 /* Set asic errata */
1339 /* Initialize clocks */
1340 radeon_get_clock_info(rdev->ddev);
1341 /* Initialize power management */
1342 radeon_pm_init(rdev);
1343 /* Get vram informations */
1344 r300_vram_info(rdev);
1345 /* Initialize memory controller (also test AGP) */
1346 r = r420_mc_init(rdev);
1350 r = radeon_fence_driver_init(rdev);
1353 r = radeon_irq_kms_init(rdev);
1356 /* Memory manager */
1357 r = radeon_bo_init(rdev);
1360 if (rdev->flags & RADEON_IS_PCIE) {
1361 r = rv370_pcie_gart_init(rdev);
1365 if (rdev->flags & RADEON_IS_PCI) {
1366 r = r100_pci_gart_init(rdev);
1370 r300_set_reg_safe(rdev);
1371 rdev->accel_working = true;
1372 r = r300_startup(rdev);
1374 /* Somethings want wront with the accel init stop accel */
1375 dev_err(rdev->dev, "Disabling GPU acceleration\n");
1380 if (rdev->flags & RADEON_IS_PCIE)
1381 rv370_pcie_gart_fini(rdev);
1382 if (rdev->flags & RADEON_IS_PCI)
1383 r100_pci_gart_fini(rdev);
1384 radeon_irq_kms_fini(rdev);
1385 rdev->accel_working = false;