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 - 4096;
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)
142 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
143 tmp |= RADEON_PCIE_TX_GART_UNMAPPED_ACCESS_DISCARD;
144 WREG32_PCIE(RADEON_PCIE_TX_GART_CNTL, tmp & ~RADEON_PCIE_TX_GART_EN);
145 if (rdev->gart.table.vram.robj) {
146 radeon_object_kunmap(rdev->gart.table.vram.robj);
147 radeon_object_unpin(rdev->gart.table.vram.robj);
151 void rv370_pcie_gart_fini(struct radeon_device *rdev)
153 rv370_pcie_gart_disable(rdev);
154 radeon_gart_table_vram_free(rdev);
155 radeon_gart_fini(rdev);
158 void r300_fence_ring_emit(struct radeon_device *rdev,
159 struct radeon_fence *fence)
161 /* Who ever call radeon_fence_emit should call ring_lock and ask
162 * for enough space (today caller are ib schedule and buffer move) */
163 /* Write SC register so SC & US assert idle */
164 radeon_ring_write(rdev, PACKET0(0x43E0, 0));
165 radeon_ring_write(rdev, 0);
166 radeon_ring_write(rdev, PACKET0(0x43E4, 0));
167 radeon_ring_write(rdev, 0);
169 radeon_ring_write(rdev, PACKET0(0x4E4C, 0));
170 radeon_ring_write(rdev, (2 << 0));
171 radeon_ring_write(rdev, PACKET0(0x4F18, 0));
172 radeon_ring_write(rdev, (1 << 0));
173 /* Wait until IDLE & CLEAN */
174 radeon_ring_write(rdev, PACKET0(0x1720, 0));
175 radeon_ring_write(rdev, (1 << 17) | (1 << 16) | (1 << 9));
176 /* Emit fence sequence & fire IRQ */
177 radeon_ring_write(rdev, PACKET0(rdev->fence_drv.scratch_reg, 0));
178 radeon_ring_write(rdev, fence->seq);
179 radeon_ring_write(rdev, PACKET0(RADEON_GEN_INT_STATUS, 0));
180 radeon_ring_write(rdev, RADEON_SW_INT_FIRE);
183 int r300_copy_dma(struct radeon_device *rdev,
187 struct radeon_fence *fence)
194 /* radeon pitch is /64 */
195 size = num_pages << PAGE_SHIFT;
196 num_loops = DIV_ROUND_UP(size, 0x1FFFFF);
197 r = radeon_ring_lock(rdev, num_loops * 4 + 64);
199 DRM_ERROR("radeon: moving bo (%d).\n", r);
202 /* Must wait for 2D idle & clean before DMA or hangs might happen */
203 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0 ));
204 radeon_ring_write(rdev, (1 << 16));
205 for (i = 0; i < num_loops; i++) {
207 if (cur_size > 0x1FFFFF) {
211 radeon_ring_write(rdev, PACKET0(0x720, 2));
212 radeon_ring_write(rdev, src_offset);
213 radeon_ring_write(rdev, dst_offset);
214 radeon_ring_write(rdev, cur_size | (1 << 31) | (1 << 30));
215 src_offset += cur_size;
216 dst_offset += cur_size;
218 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
219 radeon_ring_write(rdev, RADEON_WAIT_DMA_GUI_IDLE);
221 r = radeon_fence_emit(rdev, fence);
223 radeon_ring_unlock_commit(rdev);
227 void r300_ring_start(struct radeon_device *rdev)
229 unsigned gb_tile_config;
232 /* Sub pixel 1/12 so we can have 4K rendering according to doc */
233 gb_tile_config = (R300_ENABLE_TILING | R300_TILE_SIZE_16);
234 switch(rdev->num_gb_pipes) {
236 gb_tile_config |= R300_PIPE_COUNT_R300;
239 gb_tile_config |= R300_PIPE_COUNT_R420_3P;
242 gb_tile_config |= R300_PIPE_COUNT_R420;
246 gb_tile_config |= R300_PIPE_COUNT_RV350;
250 r = radeon_ring_lock(rdev, 64);
254 radeon_ring_write(rdev, PACKET0(RADEON_ISYNC_CNTL, 0));
255 radeon_ring_write(rdev,
256 RADEON_ISYNC_ANY2D_IDLE3D |
257 RADEON_ISYNC_ANY3D_IDLE2D |
258 RADEON_ISYNC_WAIT_IDLEGUI |
259 RADEON_ISYNC_CPSCRATCH_IDLEGUI);
260 radeon_ring_write(rdev, PACKET0(R300_GB_TILE_CONFIG, 0));
261 radeon_ring_write(rdev, gb_tile_config);
262 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
263 radeon_ring_write(rdev,
264 RADEON_WAIT_2D_IDLECLEAN |
265 RADEON_WAIT_3D_IDLECLEAN);
266 radeon_ring_write(rdev, PACKET0(0x170C, 0));
267 radeon_ring_write(rdev, 1 << 31);
268 radeon_ring_write(rdev, PACKET0(R300_GB_SELECT, 0));
269 radeon_ring_write(rdev, 0);
270 radeon_ring_write(rdev, PACKET0(R300_GB_ENABLE, 0));
271 radeon_ring_write(rdev, 0);
272 radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
273 radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
274 radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
275 radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
276 radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
277 radeon_ring_write(rdev,
278 RADEON_WAIT_2D_IDLECLEAN |
279 RADEON_WAIT_3D_IDLECLEAN);
280 radeon_ring_write(rdev, PACKET0(R300_GB_AA_CONFIG, 0));
281 radeon_ring_write(rdev, 0);
282 radeon_ring_write(rdev, PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
283 radeon_ring_write(rdev, R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE);
284 radeon_ring_write(rdev, PACKET0(R300_RB3D_ZCACHE_CTLSTAT, 0));
285 radeon_ring_write(rdev, R300_ZC_FLUSH | R300_ZC_FREE);
286 radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS0, 0));
287 radeon_ring_write(rdev,
288 ((6 << R300_MS_X0_SHIFT) |
289 (6 << R300_MS_Y0_SHIFT) |
290 (6 << R300_MS_X1_SHIFT) |
291 (6 << R300_MS_Y1_SHIFT) |
292 (6 << R300_MS_X2_SHIFT) |
293 (6 << R300_MS_Y2_SHIFT) |
294 (6 << R300_MSBD0_Y_SHIFT) |
295 (6 << R300_MSBD0_X_SHIFT)));
296 radeon_ring_write(rdev, PACKET0(R300_GB_MSPOS1, 0));
297 radeon_ring_write(rdev,
298 ((6 << R300_MS_X3_SHIFT) |
299 (6 << R300_MS_Y3_SHIFT) |
300 (6 << R300_MS_X4_SHIFT) |
301 (6 << R300_MS_Y4_SHIFT) |
302 (6 << R300_MS_X5_SHIFT) |
303 (6 << R300_MS_Y5_SHIFT) |
304 (6 << R300_MSBD1_SHIFT)));
305 radeon_ring_write(rdev, PACKET0(R300_GA_ENHANCE, 0));
306 radeon_ring_write(rdev, R300_GA_DEADLOCK_CNTL | R300_GA_FASTSYNC_CNTL);
307 radeon_ring_write(rdev, PACKET0(R300_GA_POLY_MODE, 0));
308 radeon_ring_write(rdev,
309 R300_FRONT_PTYPE_TRIANGE | R300_BACK_PTYPE_TRIANGE);
310 radeon_ring_write(rdev, PACKET0(R300_GA_ROUND_MODE, 0));
311 radeon_ring_write(rdev,
312 R300_GEOMETRY_ROUND_NEAREST |
313 R300_COLOR_ROUND_NEAREST);
314 radeon_ring_unlock_commit(rdev);
317 void r300_errata(struct radeon_device *rdev)
319 rdev->pll_errata = 0;
321 if (rdev->family == CHIP_R300 &&
322 (RREG32(RADEON_CONFIG_CNTL) & RADEON_CFG_ATI_REV_ID_MASK) == RADEON_CFG_ATI_REV_A11) {
323 rdev->pll_errata |= CHIP_ERRATA_R300_CG;
327 int r300_mc_wait_for_idle(struct radeon_device *rdev)
332 for (i = 0; i < rdev->usec_timeout; i++) {
334 tmp = RREG32(0x0150);
335 if (tmp & (1 << 4)) {
343 void r300_gpu_init(struct radeon_device *rdev)
345 uint32_t gb_tile_config, tmp;
347 r100_hdp_reset(rdev);
348 /* FIXME: rv380 one pipes ? */
349 if ((rdev->family == CHIP_R300) || (rdev->family == CHIP_R350)) {
351 rdev->num_gb_pipes = 2;
353 /* rv350,rv370,rv380 */
354 rdev->num_gb_pipes = 1;
356 rdev->num_z_pipes = 1;
357 gb_tile_config = (R300_ENABLE_TILING | R300_TILE_SIZE_16);
358 switch (rdev->num_gb_pipes) {
360 gb_tile_config |= R300_PIPE_COUNT_R300;
363 gb_tile_config |= R300_PIPE_COUNT_R420_3P;
366 gb_tile_config |= R300_PIPE_COUNT_R420;
370 gb_tile_config |= R300_PIPE_COUNT_RV350;
373 WREG32(R300_GB_TILE_CONFIG, gb_tile_config);
375 if (r100_gui_wait_for_idle(rdev)) {
376 printk(KERN_WARNING "Failed to wait GUI idle while "
377 "programming pipes. Bad things might happen.\n");
380 tmp = RREG32(0x170C);
381 WREG32(0x170C, tmp | (1 << 31));
383 WREG32(R300_RB2D_DSTCACHE_MODE,
384 R300_DC_AUTOFLUSH_ENABLE |
385 R300_DC_DC_DISABLE_IGNORE_PE);
387 if (r100_gui_wait_for_idle(rdev)) {
388 printk(KERN_WARNING "Failed to wait GUI idle while "
389 "programming pipes. Bad things might happen.\n");
391 if (r300_mc_wait_for_idle(rdev)) {
392 printk(KERN_WARNING "Failed to wait MC idle while "
393 "programming pipes. Bad things might happen.\n");
395 DRM_INFO("radeon: %d quad pipes, %d Z pipes initialized.\n",
396 rdev->num_gb_pipes, rdev->num_z_pipes);
399 int r300_ga_reset(struct radeon_device *rdev)
405 reinit_cp = rdev->cp.ready;
406 rdev->cp.ready = false;
407 for (i = 0; i < rdev->usec_timeout; i++) {
408 WREG32(RADEON_CP_CSQ_MODE, 0);
409 WREG32(RADEON_CP_CSQ_CNTL, 0);
410 WREG32(RADEON_RBBM_SOFT_RESET, 0x32005);
411 (void)RREG32(RADEON_RBBM_SOFT_RESET);
413 WREG32(RADEON_RBBM_SOFT_RESET, 0);
414 /* Wait to prevent race in RBBM_STATUS */
416 tmp = RREG32(RADEON_RBBM_STATUS);
417 if (tmp & ((1 << 20) | (1 << 26))) {
418 DRM_ERROR("VAP & CP still busy (RBBM_STATUS=0x%08X)", tmp);
419 /* GA still busy soft reset it */
420 WREG32(0x429C, 0x200);
421 WREG32(R300_VAP_PVS_STATE_FLUSH_REG, 0);
426 /* Wait to prevent race in RBBM_STATUS */
428 tmp = RREG32(RADEON_RBBM_STATUS);
429 if (!(tmp & ((1 << 20) | (1 << 26)))) {
433 for (i = 0; i < rdev->usec_timeout; i++) {
434 tmp = RREG32(RADEON_RBBM_STATUS);
435 if (!(tmp & ((1 << 20) | (1 << 26)))) {
436 DRM_INFO("GA reset succeed (RBBM_STATUS=0x%08X)\n",
439 return r100_cp_init(rdev, rdev->cp.ring_size);
445 tmp = RREG32(RADEON_RBBM_STATUS);
446 DRM_ERROR("Failed to reset GA ! (RBBM_STATUS=0x%08X)\n", tmp);
450 int r300_gpu_reset(struct radeon_device *rdev)
454 /* reset order likely matter */
455 status = RREG32(RADEON_RBBM_STATUS);
457 r100_hdp_reset(rdev);
459 if (status & ((1 << 17) | (1 << 18) | (1 << 27))) {
460 r100_rb2d_reset(rdev);
463 if (status & ((1 << 20) | (1 << 26))) {
467 status = RREG32(RADEON_RBBM_STATUS);
468 if (status & (1 << 16)) {
471 /* Check if GPU is idle */
472 status = RREG32(RADEON_RBBM_STATUS);
473 if (status & (1 << 31)) {
474 DRM_ERROR("Failed to reset GPU (RBBM_STATUS=0x%08X)\n", status);
477 DRM_INFO("GPU reset succeed (RBBM_STATUS=0x%08X)\n", status);
483 * r300,r350,rv350,rv380 VRAM info
485 void r300_vram_info(struct radeon_device *rdev)
489 /* DDR for all card after R300 & IGP */
490 rdev->mc.vram_is_ddr = true;
491 tmp = RREG32(RADEON_MEM_CNTL);
492 if (tmp & R300_MEM_NUM_CHANNELS_MASK) {
493 rdev->mc.vram_width = 128;
495 rdev->mc.vram_width = 64;
498 r100_vram_init_sizes(rdev);
501 void rv370_set_pcie_lanes(struct radeon_device *rdev, int lanes)
503 uint32_t link_width_cntl, mask;
505 if (rdev->flags & RADEON_IS_IGP)
508 if (!(rdev->flags & RADEON_IS_PCIE))
511 /* FIXME wait for idle */
515 mask = RADEON_PCIE_LC_LINK_WIDTH_X0;
518 mask = RADEON_PCIE_LC_LINK_WIDTH_X1;
521 mask = RADEON_PCIE_LC_LINK_WIDTH_X2;
524 mask = RADEON_PCIE_LC_LINK_WIDTH_X4;
527 mask = RADEON_PCIE_LC_LINK_WIDTH_X8;
530 mask = RADEON_PCIE_LC_LINK_WIDTH_X12;
534 mask = RADEON_PCIE_LC_LINK_WIDTH_X16;
538 link_width_cntl = RREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
540 if ((link_width_cntl & RADEON_PCIE_LC_LINK_WIDTH_RD_MASK) ==
541 (mask << RADEON_PCIE_LC_LINK_WIDTH_RD_SHIFT))
544 link_width_cntl &= ~(RADEON_PCIE_LC_LINK_WIDTH_MASK |
545 RADEON_PCIE_LC_RECONFIG_NOW |
546 RADEON_PCIE_LC_RECONFIG_LATER |
547 RADEON_PCIE_LC_SHORT_RECONFIG_EN);
548 link_width_cntl |= mask;
549 WREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl);
550 WREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL, (link_width_cntl |
551 RADEON_PCIE_LC_RECONFIG_NOW));
553 /* wait for lane set to complete */
554 link_width_cntl = RREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
555 while (link_width_cntl == 0xffffffff)
556 link_width_cntl = RREG32_PCIE(RADEON_PCIE_LC_LINK_WIDTH_CNTL);
560 #if defined(CONFIG_DEBUG_FS)
561 static int rv370_debugfs_pcie_gart_info(struct seq_file *m, void *data)
563 struct drm_info_node *node = (struct drm_info_node *) m->private;
564 struct drm_device *dev = node->minor->dev;
565 struct radeon_device *rdev = dev->dev_private;
568 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_CNTL);
569 seq_printf(m, "PCIE_TX_GART_CNTL 0x%08x\n", tmp);
570 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_BASE);
571 seq_printf(m, "PCIE_TX_GART_BASE 0x%08x\n", tmp);
572 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_START_LO);
573 seq_printf(m, "PCIE_TX_GART_START_LO 0x%08x\n", tmp);
574 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_START_HI);
575 seq_printf(m, "PCIE_TX_GART_START_HI 0x%08x\n", tmp);
576 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_END_LO);
577 seq_printf(m, "PCIE_TX_GART_END_LO 0x%08x\n", tmp);
578 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_END_HI);
579 seq_printf(m, "PCIE_TX_GART_END_HI 0x%08x\n", tmp);
580 tmp = RREG32_PCIE(RADEON_PCIE_TX_GART_ERROR);
581 seq_printf(m, "PCIE_TX_GART_ERROR 0x%08x\n", tmp);
585 static struct drm_info_list rv370_pcie_gart_info_list[] = {
586 {"rv370_pcie_gart_info", rv370_debugfs_pcie_gart_info, 0, NULL},
590 static int rv370_debugfs_pcie_gart_info_init(struct radeon_device *rdev)
592 #if defined(CONFIG_DEBUG_FS)
593 return radeon_debugfs_add_files(rdev, rv370_pcie_gart_info_list, 1);
599 static int r300_packet0_check(struct radeon_cs_parser *p,
600 struct radeon_cs_packet *pkt,
601 unsigned idx, unsigned reg)
603 struct radeon_cs_reloc *reloc;
604 struct r100_cs_track *track;
605 volatile uint32_t *ib;
606 uint32_t tmp, tile_flags = 0;
612 track = (struct r100_cs_track *)p->track;
613 idx_value = radeon_get_ib_value(p, idx);
616 case AVIVO_D1MODE_VLINE_START_END:
617 case RADEON_CRTC_GUI_TRIG_VLINE:
618 r = r100_cs_packet_parse_vline(p);
620 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
622 r100_cs_dump_packet(p, pkt);
626 case RADEON_DST_PITCH_OFFSET:
627 case RADEON_SRC_PITCH_OFFSET:
628 r = r100_reloc_pitch_offset(p, pkt, idx, reg);
632 case R300_RB3D_COLOROFFSET0:
633 case R300_RB3D_COLOROFFSET1:
634 case R300_RB3D_COLOROFFSET2:
635 case R300_RB3D_COLOROFFSET3:
636 i = (reg - R300_RB3D_COLOROFFSET0) >> 2;
637 r = r100_cs_packet_next_reloc(p, &reloc);
639 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
641 r100_cs_dump_packet(p, pkt);
644 track->cb[i].robj = reloc->robj;
645 track->cb[i].offset = idx_value;
646 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
648 case R300_ZB_DEPTHOFFSET:
649 r = r100_cs_packet_next_reloc(p, &reloc);
651 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
653 r100_cs_dump_packet(p, pkt);
656 track->zb.robj = reloc->robj;
657 track->zb.offset = idx_value;
658 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
660 case R300_TX_OFFSET_0:
661 case R300_TX_OFFSET_0+4:
662 case R300_TX_OFFSET_0+8:
663 case R300_TX_OFFSET_0+12:
664 case R300_TX_OFFSET_0+16:
665 case R300_TX_OFFSET_0+20:
666 case R300_TX_OFFSET_0+24:
667 case R300_TX_OFFSET_0+28:
668 case R300_TX_OFFSET_0+32:
669 case R300_TX_OFFSET_0+36:
670 case R300_TX_OFFSET_0+40:
671 case R300_TX_OFFSET_0+44:
672 case R300_TX_OFFSET_0+48:
673 case R300_TX_OFFSET_0+52:
674 case R300_TX_OFFSET_0+56:
675 case R300_TX_OFFSET_0+60:
676 i = (reg - R300_TX_OFFSET_0) >> 2;
677 r = r100_cs_packet_next_reloc(p, &reloc);
679 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
681 r100_cs_dump_packet(p, pkt);
684 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
685 track->textures[i].robj = reloc->robj;
687 /* Tracked registers */
690 track->vap_vf_cntl = idx_value;
694 track->vtx_size = idx_value & 0x7F;
697 /* VAP_VF_MAX_VTX_INDX */
698 track->max_indx = idx_value & 0x00FFFFFFUL;
702 track->maxy = ((idx_value >> 13) & 0x1FFF) + 1;
703 if (p->rdev->family < CHIP_RV515) {
709 track->num_cb = ((idx_value >> 5) & 0x3) + 1;
715 /* RB3D_COLORPITCH0 */
716 /* RB3D_COLORPITCH1 */
717 /* RB3D_COLORPITCH2 */
718 /* RB3D_COLORPITCH3 */
719 r = r100_cs_packet_next_reloc(p, &reloc);
721 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
723 r100_cs_dump_packet(p, pkt);
727 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
728 tile_flags |= R300_COLOR_TILE_ENABLE;
729 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
730 tile_flags |= R300_COLOR_MICROTILE_ENABLE;
732 tmp = idx_value & ~(0x7 << 16);
736 i = (reg - 0x4E38) >> 2;
737 track->cb[i].pitch = idx_value & 0x3FFE;
738 switch (((idx_value >> 21) & 0xF)) {
742 track->cb[i].cpp = 1;
748 track->cb[i].cpp = 2;
751 track->cb[i].cpp = 4;
754 track->cb[i].cpp = 8;
757 track->cb[i].cpp = 16;
760 DRM_ERROR("Invalid color buffer format (%d) !\n",
761 ((idx_value >> 21) & 0xF));
768 track->z_enabled = true;
770 track->z_enabled = false;
775 switch ((idx_value & 0xF)) {
784 DRM_ERROR("Invalid z buffer format (%d) !\n",
791 r = r100_cs_packet_next_reloc(p, &reloc);
793 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
795 r100_cs_dump_packet(p, pkt);
799 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
800 tile_flags |= R300_DEPTHMACROTILE_ENABLE;
801 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
802 tile_flags |= R300_DEPTHMICROTILE_TILED;;
804 tmp = idx_value & ~(0x7 << 16);
808 track->zb.pitch = idx_value & 0x3FFC;
811 for (i = 0; i < 16; i++) {
814 enabled = !!(idx_value & (1 << i));
815 track->textures[i].enabled = enabled;
834 /* TX_FORMAT1_[0-15] */
835 i = (reg - 0x44C0) >> 2;
836 tmp = (idx_value >> 25) & 0x3;
837 track->textures[i].tex_coord_type = tmp;
838 switch ((idx_value & 0x1F)) {
839 case R300_TX_FORMAT_X8:
840 case R300_TX_FORMAT_Y4X4:
841 case R300_TX_FORMAT_Z3Y3X2:
842 track->textures[i].cpp = 1;
844 case R300_TX_FORMAT_X16:
845 case R300_TX_FORMAT_Y8X8:
846 case R300_TX_FORMAT_Z5Y6X5:
847 case R300_TX_FORMAT_Z6Y5X5:
848 case R300_TX_FORMAT_W4Z4Y4X4:
849 case R300_TX_FORMAT_W1Z5Y5X5:
850 case R300_TX_FORMAT_DXT1:
851 case R300_TX_FORMAT_D3DMFT_CxV8U8:
852 case R300_TX_FORMAT_B8G8_B8G8:
853 case R300_TX_FORMAT_G8R8_G8B8:
854 track->textures[i].cpp = 2;
856 case R300_TX_FORMAT_Y16X16:
857 case R300_TX_FORMAT_Z11Y11X10:
858 case R300_TX_FORMAT_Z10Y11X11:
859 case R300_TX_FORMAT_W8Z8Y8X8:
860 case R300_TX_FORMAT_W2Z10Y10X10:
862 case R300_TX_FORMAT_FL_I32:
864 case R300_TX_FORMAT_DXT3:
865 case R300_TX_FORMAT_DXT5:
866 track->textures[i].cpp = 4;
868 case R300_TX_FORMAT_W16Z16Y16X16:
869 case R300_TX_FORMAT_FL_R16G16B16A16:
870 case R300_TX_FORMAT_FL_I32A32:
871 track->textures[i].cpp = 8;
873 case R300_TX_FORMAT_FL_R32G32B32A32:
874 track->textures[i].cpp = 16;
877 DRM_ERROR("Invalid texture format %u\n",
899 /* TX_FILTER0_[0-15] */
900 i = (reg - 0x4400) >> 2;
901 tmp = idx_value & 0x7;
902 if (tmp == 2 || tmp == 4 || tmp == 6) {
903 track->textures[i].roundup_w = false;
905 tmp = (idx_value >> 3) & 0x7;
906 if (tmp == 2 || tmp == 4 || tmp == 6) {
907 track->textures[i].roundup_h = false;
926 /* TX_FORMAT2_[0-15] */
927 i = (reg - 0x4500) >> 2;
928 tmp = idx_value & 0x3FFF;
929 track->textures[i].pitch = tmp + 1;
930 if (p->rdev->family >= CHIP_RV515) {
931 tmp = ((idx_value >> 15) & 1) << 11;
932 track->textures[i].width_11 = tmp;
933 tmp = ((idx_value >> 16) & 1) << 11;
934 track->textures[i].height_11 = tmp;
953 /* TX_FORMAT0_[0-15] */
954 i = (reg - 0x4480) >> 2;
955 tmp = idx_value & 0x7FF;
956 track->textures[i].width = tmp + 1;
957 tmp = (idx_value >> 11) & 0x7FF;
958 track->textures[i].height = tmp + 1;
959 tmp = (idx_value >> 26) & 0xF;
960 track->textures[i].num_levels = tmp;
961 tmp = idx_value & (1 << 31);
962 track->textures[i].use_pitch = !!tmp;
963 tmp = (idx_value >> 22) & 0xF;
964 track->textures[i].txdepth = tmp;
966 case R300_ZB_ZPASS_ADDR:
967 r = r100_cs_packet_next_reloc(p, &reloc);
969 DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
971 r100_cs_dump_packet(p, pkt);
974 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
977 /* valid register only on RV530 */
978 if (p->rdev->family == CHIP_RV530)
980 /* fallthrough do not move */
982 printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
989 static int r300_packet3_check(struct radeon_cs_parser *p,
990 struct radeon_cs_packet *pkt)
992 struct radeon_cs_reloc *reloc;
993 struct r100_cs_track *track;
994 volatile uint32_t *ib;
1000 track = (struct r100_cs_track *)p->track;
1001 switch(pkt->opcode) {
1002 case PACKET3_3D_LOAD_VBPNTR:
1003 r = r100_packet3_load_vbpntr(p, pkt, idx);
1007 case PACKET3_INDX_BUFFER:
1008 r = r100_cs_packet_next_reloc(p, &reloc);
1010 DRM_ERROR("No reloc for packet3 %d\n", pkt->opcode);
1011 r100_cs_dump_packet(p, pkt);
1014 ib[idx+1] = radeon_get_ib_value(p, idx + 1) + ((u32)reloc->lobj.gpu_offset);
1015 r = r100_cs_track_check_pkt3_indx_buffer(p, pkt, reloc->robj);
1021 case PACKET3_3D_DRAW_IMMD:
1022 /* Number of dwords is vtx_size * (num_vertices - 1)
1023 * PRIM_WALK must be equal to 3 vertex data in embedded
1025 if (((radeon_get_ib_value(p, idx + 1) >> 4) & 0x3) != 3) {
1026 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1029 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1030 track->immd_dwords = pkt->count - 1;
1031 r = r100_cs_track_check(p->rdev, track);
1036 case PACKET3_3D_DRAW_IMMD_2:
1037 /* Number of dwords is vtx_size * (num_vertices - 1)
1038 * PRIM_WALK must be equal to 3 vertex data in embedded
1040 if (((radeon_get_ib_value(p, idx) >> 4) & 0x3) != 3) {
1041 DRM_ERROR("PRIM_WALK must be 3 for IMMD draw\n");
1044 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1045 track->immd_dwords = pkt->count;
1046 r = r100_cs_track_check(p->rdev, track);
1051 case PACKET3_3D_DRAW_VBUF:
1052 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1053 r = r100_cs_track_check(p->rdev, track);
1058 case PACKET3_3D_DRAW_VBUF_2:
1059 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1060 r = r100_cs_track_check(p->rdev, track);
1065 case PACKET3_3D_DRAW_INDX:
1066 track->vap_vf_cntl = radeon_get_ib_value(p, idx + 1);
1067 r = r100_cs_track_check(p->rdev, track);
1072 case PACKET3_3D_DRAW_INDX_2:
1073 track->vap_vf_cntl = radeon_get_ib_value(p, idx);
1074 r = r100_cs_track_check(p->rdev, track);
1082 DRM_ERROR("Packet3 opcode %x not supported\n", pkt->opcode);
1088 int r300_cs_parse(struct radeon_cs_parser *p)
1090 struct radeon_cs_packet pkt;
1091 struct r100_cs_track *track;
1094 track = kzalloc(sizeof(*track), GFP_KERNEL);
1095 r100_cs_track_clear(p->rdev, track);
1098 r = r100_cs_packet_parse(p, &pkt, p->idx);
1102 p->idx += pkt.count + 2;
1105 r = r100_cs_parse_packet0(p, &pkt,
1106 p->rdev->config.r300.reg_safe_bm,
1107 p->rdev->config.r300.reg_safe_bm_size,
1108 &r300_packet0_check);
1113 r = r300_packet3_check(p, &pkt);
1116 DRM_ERROR("Unknown packet type %d !\n", pkt.type);
1122 } while (p->idx < p->chunks[p->chunk_ib_idx].length_dw);
1126 void r300_set_reg_safe(struct radeon_device *rdev)
1128 rdev->config.r300.reg_safe_bm = r300_reg_safe_bm;
1129 rdev->config.r300.reg_safe_bm_size = ARRAY_SIZE(r300_reg_safe_bm);
1132 void r300_mc_program(struct radeon_device *rdev)
1134 struct r100_mc_save save;
1137 r = r100_debugfs_mc_info_init(rdev);
1139 dev_err(rdev->dev, "Failed to create r100_mc debugfs file.\n");
1142 /* Stops all mc clients */
1143 r100_mc_stop(rdev, &save);
1144 if (rdev->flags & RADEON_IS_AGP) {
1145 WREG32(R_00014C_MC_AGP_LOCATION,
1146 S_00014C_MC_AGP_START(rdev->mc.gtt_start >> 16) |
1147 S_00014C_MC_AGP_TOP(rdev->mc.gtt_end >> 16));
1148 WREG32(R_000170_AGP_BASE, lower_32_bits(rdev->mc.agp_base));
1149 WREG32(R_00015C_AGP_BASE_2,
1150 upper_32_bits(rdev->mc.agp_base) & 0xff);
1152 WREG32(R_00014C_MC_AGP_LOCATION, 0x0FFFFFFF);
1153 WREG32(R_000170_AGP_BASE, 0);
1154 WREG32(R_00015C_AGP_BASE_2, 0);
1156 /* Wait for mc idle */
1157 if (r300_mc_wait_for_idle(rdev))
1158 DRM_INFO("Failed to wait MC idle before programming MC.\n");
1159 /* Program MC, should be a 32bits limited address space */
1160 WREG32(R_000148_MC_FB_LOCATION,
1161 S_000148_MC_FB_START(rdev->mc.vram_start >> 16) |
1162 S_000148_MC_FB_TOP(rdev->mc.vram_end >> 16));
1163 r100_mc_resume(rdev, &save);
1166 void r300_clock_startup(struct radeon_device *rdev)
1170 if (radeon_dynclks != -1 && radeon_dynclks)
1171 radeon_legacy_set_clock_gating(rdev, 1);
1172 /* We need to force on some of the block */
1173 tmp = RREG32_PLL(R_00000D_SCLK_CNTL);
1174 tmp |= S_00000D_FORCE_CP(1) | S_00000D_FORCE_VIP(1);
1175 if ((rdev->family == CHIP_RV350) || (rdev->family == CHIP_RV380))
1176 tmp |= S_00000D_FORCE_VAP(1);
1177 WREG32_PLL(R_00000D_SCLK_CNTL, tmp);
1180 static int r300_startup(struct radeon_device *rdev)
1184 r300_mc_program(rdev);
1186 r300_clock_startup(rdev);
1187 /* Initialize GPU configuration (# pipes, ...) */
1188 r300_gpu_init(rdev);
1189 /* Initialize GART (initialize after TTM so we can allocate
1190 * memory through TTM but finalize after TTM) */
1191 if (rdev->flags & RADEON_IS_PCIE) {
1192 r = rv370_pcie_gart_enable(rdev);
1196 if (rdev->flags & RADEON_IS_PCI) {
1197 r = r100_pci_gart_enable(rdev);
1202 rdev->irq.sw_int = true;
1204 /* 1M ring buffer */
1205 r = r100_cp_init(rdev, 1024 * 1024);
1207 dev_err(rdev->dev, "failled initializing CP (%d).\n", r);
1210 r = r100_wb_init(rdev);
1212 dev_err(rdev->dev, "failled initializing WB (%d).\n", r);
1213 r = r100_ib_init(rdev);
1215 dev_err(rdev->dev, "failled initializing IB (%d).\n", r);
1221 int r300_resume(struct radeon_device *rdev)
1223 /* Make sur GART are not working */
1224 if (rdev->flags & RADEON_IS_PCIE)
1225 rv370_pcie_gart_disable(rdev);
1226 if (rdev->flags & RADEON_IS_PCI)
1227 r100_pci_gart_disable(rdev);
1228 /* Resume clock before doing reset */
1229 r300_clock_startup(rdev);
1230 /* Reset gpu before posting otherwise ATOM will enter infinite loop */
1231 if (radeon_gpu_reset(rdev)) {
1232 dev_warn(rdev->dev, "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
1233 RREG32(R_000E40_RBBM_STATUS),
1234 RREG32(R_0007C0_CP_STAT));
1237 radeon_combios_asic_init(rdev->ddev);
1238 /* Resume clock after posting */
1239 r300_clock_startup(rdev);
1240 return r300_startup(rdev);
1243 int r300_suspend(struct radeon_device *rdev)
1245 r100_cp_disable(rdev);
1246 r100_wb_disable(rdev);
1247 r100_irq_disable(rdev);
1248 if (rdev->flags & RADEON_IS_PCIE)
1249 rv370_pcie_gart_disable(rdev);
1250 if (rdev->flags & RADEON_IS_PCI)
1251 r100_pci_gart_disable(rdev);
1255 void r300_fini(struct radeon_device *rdev)
1261 radeon_gem_fini(rdev);
1262 if (rdev->flags & RADEON_IS_PCIE)
1263 rv370_pcie_gart_fini(rdev);
1264 if (rdev->flags & RADEON_IS_PCI)
1265 r100_pci_gart_fini(rdev);
1266 radeon_irq_kms_fini(rdev);
1267 radeon_fence_driver_fini(rdev);
1268 radeon_object_fini(rdev);
1269 radeon_atombios_fini(rdev);
1274 int r300_init(struct radeon_device *rdev)
1279 r100_vga_render_disable(rdev);
1280 /* Initialize scratch registers */
1281 radeon_scratch_init(rdev);
1282 /* Initialize surface registers */
1283 radeon_surface_init(rdev);
1284 /* TODO: disable VGA need to use VGA request */
1286 if (!radeon_get_bios(rdev)) {
1287 if (ASIC_IS_AVIVO(rdev))
1290 if (rdev->is_atom_bios) {
1291 dev_err(rdev->dev, "Expecting combios for RS400/RS480 GPU\n");
1294 r = radeon_combios_init(rdev);
1298 /* Reset gpu before posting otherwise ATOM will enter infinite loop */
1299 if (radeon_gpu_reset(rdev)) {
1301 "GPU reset failed ! (0xE40=0x%08X, 0x7C0=0x%08X)\n",
1302 RREG32(R_000E40_RBBM_STATUS),
1303 RREG32(R_0007C0_CP_STAT));
1305 /* check if cards are posted or not */
1306 if (!radeon_card_posted(rdev) && rdev->bios) {
1307 DRM_INFO("GPU not posted. posting now...\n");
1308 radeon_combios_asic_init(rdev->ddev);
1310 /* Set asic errata */
1312 /* Initialize clocks */
1313 radeon_get_clock_info(rdev->ddev);
1314 /* Get vram informations */
1315 r300_vram_info(rdev);
1316 /* Initialize memory controller (also test AGP) */
1317 r = r420_mc_init(rdev);
1321 r = radeon_fence_driver_init(rdev);
1324 r = radeon_irq_kms_init(rdev);
1327 /* Memory manager */
1328 r = radeon_object_init(rdev);
1331 if (rdev->flags & RADEON_IS_PCIE) {
1332 r = rv370_pcie_gart_init(rdev);
1336 if (rdev->flags & RADEON_IS_PCI) {
1337 r = r100_pci_gart_init(rdev);
1341 r300_set_reg_safe(rdev);
1342 rdev->accel_working = true;
1343 r = r300_startup(rdev);
1345 /* Somethings want wront with the accel init stop accel */
1346 dev_err(rdev->dev, "Disabling GPU acceleration\n");
1351 if (rdev->flags & RADEON_IS_PCIE)
1352 rv370_pcie_gart_fini(rdev);
1353 if (rdev->flags & RADEON_IS_PCI)
1354 r100_pci_gart_fini(rdev);
1355 radeon_irq_kms_fini(rdev);
1356 rdev->accel_working = false;