intel_ringbuffer.c 36.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
/*
 * Copyright © 2008-2010 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 *
 * Authors:
 *    Eric Anholt <eric@anholt.net>
 *    Zou Nan hai <nanhai.zou@intel.com>
 *    Xiang Hai hao<haihao.xiang@intel.com>
 *
 */

#include "drmP.h"
#include "drm.h"
#include "i915_drv.h"
33
#include "i915_drm.h"
34
#include "i915_trace.h"
35
#include "intel_drv.h"
36

37 38 39 40 41 42 43 44 45 46
/*
 * 965+ support PIPE_CONTROL commands, which provide finer grained control
 * over cache flushing.
 */
struct pipe_control {
	struct drm_i915_gem_object *obj;
	volatile u32 *cpu_page;
	u32 gtt_offset;
};

47 48 49 50 51 52 53 54
static inline int ring_space(struct intel_ring_buffer *ring)
{
	int space = (ring->head & HEAD_ADDR) - (ring->tail + 8);
	if (space < 0)
		space += ring->size;
	return space;
}

55
static int
56
render_ring_flush(struct intel_ring_buffer *ring,
57 58
		  u32	invalidate_domains,
		  u32	flush_domains)
59
{
60
	struct drm_device *dev = ring->dev;
61
	u32 cmd;
62
	int ret;
63

64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
	/*
	 * read/write caches:
	 *
	 * I915_GEM_DOMAIN_RENDER is always invalidated, but is
	 * only flushed if MI_NO_WRITE_FLUSH is unset.  On 965, it is
	 * also flushed at 2d versus 3d pipeline switches.
	 *
	 * read-only caches:
	 *
	 * I915_GEM_DOMAIN_SAMPLER is flushed on pre-965 if
	 * MI_READ_FLUSH is set, and is always flushed on 965.
	 *
	 * I915_GEM_DOMAIN_COMMAND may not exist?
	 *
	 * I915_GEM_DOMAIN_INSTRUCTION, which exists on 965, is
	 * invalidated when MI_EXE_FLUSH is set.
	 *
	 * I915_GEM_DOMAIN_VERTEX, which exists on 965, is
	 * invalidated with every MI_FLUSH.
	 *
	 * TLBs:
	 *
	 * On 965, TLBs associated with I915_GEM_DOMAIN_COMMAND
	 * and I915_GEM_DOMAIN_CPU in are invalidated at PTE write and
	 * I915_GEM_DOMAIN_RENDER and I915_GEM_DOMAIN_SAMPLER
	 * are flushed at any MI_FLUSH.
	 */

	cmd = MI_FLUSH | MI_NO_WRITE_FLUSH;
	if ((invalidate_domains|flush_domains) &
	    I915_GEM_DOMAIN_RENDER)
		cmd &= ~MI_NO_WRITE_FLUSH;
	if (INTEL_INFO(dev)->gen < 4) {
97
		/*
98 99
		 * On the 965, the sampler cache always gets flushed
		 * and this bit is reserved.
100
		 */
101 102 103 104 105
		if (invalidate_domains & I915_GEM_DOMAIN_SAMPLER)
			cmd |= MI_READ_FLUSH;
	}
	if (invalidate_domains & I915_GEM_DOMAIN_INSTRUCTION)
		cmd |= MI_EXE_FLUSH;
106

107 108 109
	if (invalidate_domains & I915_GEM_DOMAIN_COMMAND &&
	    (IS_G4X(dev) || IS_GEN5(dev)))
		cmd |= MI_INVALIDATE_ISP;
110

111 112 113
	ret = intel_ring_begin(ring, 2);
	if (ret)
		return ret;
114

115 116 117
	intel_ring_emit(ring, cmd);
	intel_ring_emit(ring, MI_NOOP);
	intel_ring_advance(ring);
118 119

	return 0;
120 121
}

122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233
/**
 * Emits a PIPE_CONTROL with a non-zero post-sync operation, for
 * implementing two workarounds on gen6.  From section 1.4.7.1
 * "PIPE_CONTROL" of the Sandy Bridge PRM volume 2 part 1:
 *
 * [DevSNB-C+{W/A}] Before any depth stall flush (including those
 * produced by non-pipelined state commands), software needs to first
 * send a PIPE_CONTROL with no bits set except Post-Sync Operation !=
 * 0.
 *
 * [Dev-SNB{W/A}]: Before a PIPE_CONTROL with Write Cache Flush Enable
 * =1, a PIPE_CONTROL with any non-zero post-sync-op is required.
 *
 * And the workaround for these two requires this workaround first:
 *
 * [Dev-SNB{W/A}]: Pipe-control with CS-stall bit set must be sent
 * BEFORE the pipe-control with a post-sync op and no write-cache
 * flushes.
 *
 * And this last workaround is tricky because of the requirements on
 * that bit.  From section 1.4.7.2.3 "Stall" of the Sandy Bridge PRM
 * volume 2 part 1:
 *
 *     "1 of the following must also be set:
 *      - Render Target Cache Flush Enable ([12] of DW1)
 *      - Depth Cache Flush Enable ([0] of DW1)
 *      - Stall at Pixel Scoreboard ([1] of DW1)
 *      - Depth Stall ([13] of DW1)
 *      - Post-Sync Operation ([13] of DW1)
 *      - Notify Enable ([8] of DW1)"
 *
 * The cache flushes require the workaround flush that triggered this
 * one, so we can't use it.  Depth stall would trigger the same.
 * Post-sync nonzero is what triggered this second workaround, so we
 * can't use that one either.  Notify enable is IRQs, which aren't
 * really our business.  That leaves only stall at scoreboard.
 */
static int
intel_emit_post_sync_nonzero_flush(struct intel_ring_buffer *ring)
{
	struct pipe_control *pc = ring->private;
	u32 scratch_addr = pc->gtt_offset + 128;
	int ret;


	ret = intel_ring_begin(ring, 6);
	if (ret)
		return ret;

	intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(5));
	intel_ring_emit(ring, PIPE_CONTROL_CS_STALL |
			PIPE_CONTROL_STALL_AT_SCOREBOARD);
	intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT); /* address */
	intel_ring_emit(ring, 0); /* low dword */
	intel_ring_emit(ring, 0); /* high dword */
	intel_ring_emit(ring, MI_NOOP);
	intel_ring_advance(ring);

	ret = intel_ring_begin(ring, 6);
	if (ret)
		return ret;

	intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(5));
	intel_ring_emit(ring, PIPE_CONTROL_QW_WRITE);
	intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT); /* address */
	intel_ring_emit(ring, 0);
	intel_ring_emit(ring, 0);
	intel_ring_emit(ring, MI_NOOP);
	intel_ring_advance(ring);

	return 0;
}

static int
gen6_render_ring_flush(struct intel_ring_buffer *ring,
                         u32 invalidate_domains, u32 flush_domains)
{
	u32 flags = 0;
	struct pipe_control *pc = ring->private;
	u32 scratch_addr = pc->gtt_offset + 128;
	int ret;

	/* Force SNB workarounds for PIPE_CONTROL flushes */
	intel_emit_post_sync_nonzero_flush(ring);

	/* Just flush everything.  Experiments have shown that reducing the
	 * number of bits based on the write domains has little performance
	 * impact.
	 */
	flags |= PIPE_CONTROL_RENDER_TARGET_CACHE_FLUSH;
	flags |= PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE;
	flags |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
	flags |= PIPE_CONTROL_DEPTH_CACHE_FLUSH;
	flags |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
	flags |= PIPE_CONTROL_CONST_CACHE_INVALIDATE;
	flags |= PIPE_CONTROL_STATE_CACHE_INVALIDATE;

	ret = intel_ring_begin(ring, 6);
	if (ret)
		return ret;

	intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(5));
	intel_ring_emit(ring, flags);
	intel_ring_emit(ring, scratch_addr | PIPE_CONTROL_GLOBAL_GTT);
	intel_ring_emit(ring, 0); /* lower dword */
	intel_ring_emit(ring, 0); /* uppwer dword */
	intel_ring_emit(ring, MI_NOOP);
	intel_ring_advance(ring);

	return 0;
}

234
static void ring_write_tail(struct intel_ring_buffer *ring,
235
			    u32 value)
236
{
237
	drm_i915_private_t *dev_priv = ring->dev->dev_private;
238
	I915_WRITE_TAIL(ring, value);
239 240
}

241
u32 intel_ring_get_active_head(struct intel_ring_buffer *ring)
242
{
243 244
	drm_i915_private_t *dev_priv = ring->dev->dev_private;
	u32 acthd_reg = INTEL_INFO(ring->dev)->gen >= 4 ?
D
Daniel Vetter 已提交
245
			RING_ACTHD(ring->mmio_base) : ACTHD;
246 247 248 249

	return I915_READ(acthd_reg);
}

250
static int init_ring_common(struct intel_ring_buffer *ring)
251
{
252
	drm_i915_private_t *dev_priv = ring->dev->dev_private;
253
	struct drm_i915_gem_object *obj = ring->obj;
254 255 256
	u32 head;

	/* Stop the ring if it's running. */
257
	I915_WRITE_CTL(ring, 0);
258
	I915_WRITE_HEAD(ring, 0);
259
	ring->write_tail(ring, 0);
260 261

	/* Initialize the ring. */
262
	I915_WRITE_START(ring, obj->gtt_offset);
263
	head = I915_READ_HEAD(ring) & HEAD_ADDR;
264 265 266

	/* G45 ring initialization fails to reset head to zero */
	if (head != 0) {
267 268 269 270 271 272 273
		DRM_DEBUG_KMS("%s head not reset to zero "
			      "ctl %08x head %08x tail %08x start %08x\n",
			      ring->name,
			      I915_READ_CTL(ring),
			      I915_READ_HEAD(ring),
			      I915_READ_TAIL(ring),
			      I915_READ_START(ring));
274

275
		I915_WRITE_HEAD(ring, 0);
276

277 278 279 280 281 282 283 284 285
		if (I915_READ_HEAD(ring) & HEAD_ADDR) {
			DRM_ERROR("failed to set %s head to zero "
				  "ctl %08x head %08x tail %08x start %08x\n",
				  ring->name,
				  I915_READ_CTL(ring),
				  I915_READ_HEAD(ring),
				  I915_READ_TAIL(ring),
				  I915_READ_START(ring));
		}
286 287
	}

288
	I915_WRITE_CTL(ring,
289
			((ring->size - PAGE_SIZE) & RING_NR_PAGES)
290
			| RING_VALID);
291 292

	/* If the head is still not zero, the ring is dead */
293 294 295
	if (wait_for((I915_READ_CTL(ring) & RING_VALID) != 0 &&
		     I915_READ_START(ring) == obj->gtt_offset &&
		     (I915_READ_HEAD(ring) & HEAD_ADDR) == 0, 50)) {
296 297 298 299 300 301 302 303
		DRM_ERROR("%s initialization failed "
				"ctl %08x head %08x tail %08x start %08x\n",
				ring->name,
				I915_READ_CTL(ring),
				I915_READ_HEAD(ring),
				I915_READ_TAIL(ring),
				I915_READ_START(ring));
		return -EIO;
304 305
	}

306 307
	if (!drm_core_check_feature(ring->dev, DRIVER_MODESET))
		i915_kernel_lost_context(ring->dev);
308
	else {
309
		ring->head = I915_READ_HEAD(ring);
310
		ring->tail = I915_READ_TAIL(ring) & TAIL_ADDR;
311
		ring->space = ring_space(ring);
312
	}
313

314 315 316
	return 0;
}

317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336
static int
init_pipe_control(struct intel_ring_buffer *ring)
{
	struct pipe_control *pc;
	struct drm_i915_gem_object *obj;
	int ret;

	if (ring->private)
		return 0;

	pc = kmalloc(sizeof(*pc), GFP_KERNEL);
	if (!pc)
		return -ENOMEM;

	obj = i915_gem_alloc_object(ring->dev, 4096);
	if (obj == NULL) {
		DRM_ERROR("Failed to allocate seqno page\n");
		ret = -ENOMEM;
		goto err;
	}
337 338

	i915_gem_object_set_cache_level(obj, I915_CACHE_LLC);
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379

	ret = i915_gem_object_pin(obj, 4096, true);
	if (ret)
		goto err_unref;

	pc->gtt_offset = obj->gtt_offset;
	pc->cpu_page =  kmap(obj->pages[0]);
	if (pc->cpu_page == NULL)
		goto err_unpin;

	pc->obj = obj;
	ring->private = pc;
	return 0;

err_unpin:
	i915_gem_object_unpin(obj);
err_unref:
	drm_gem_object_unreference(&obj->base);
err:
	kfree(pc);
	return ret;
}

static void
cleanup_pipe_control(struct intel_ring_buffer *ring)
{
	struct pipe_control *pc = ring->private;
	struct drm_i915_gem_object *obj;

	if (!ring->private)
		return;

	obj = pc->obj;
	kunmap(obj->pages[0]);
	i915_gem_object_unpin(obj);
	drm_gem_object_unreference(&obj->base);

	kfree(pc);
	ring->private = NULL;
}

380
static int init_render_ring(struct intel_ring_buffer *ring)
381
{
382
	struct drm_device *dev = ring->dev;
383
	struct drm_i915_private *dev_priv = dev->dev_private;
384
	int ret = init_ring_common(ring);
385

386
	if (INTEL_INFO(dev)->gen > 3) {
387
		int mode = VS_TIMER_DISPATCH << 16 | VS_TIMER_DISPATCH;
388
		I915_WRITE(MI_MODE, mode);
389 390 391 392
		if (IS_GEN7(dev))
			I915_WRITE(GFX_MODE_GEN7,
				   GFX_MODE_DISABLE(GFX_TLB_INVALIDATE_ALWAYS) |
				   GFX_MODE_ENABLE(GFX_REPLAY_MODE));
393
	}
394

395
	if (INTEL_INFO(dev)->gen >= 5) {
396 397 398 399 400
		ret = init_pipe_control(ring);
		if (ret)
			return ret;
	}

401 402 403 404 405
	if (INTEL_INFO(dev)->gen >= 6) {
		I915_WRITE(INSTPM,
			   INSTPM_FORCE_ORDERING << 16 | INSTPM_FORCE_ORDERING);
	}

406 407 408
	return ret;
}

409 410 411 412 413 414 415 416
static void render_ring_cleanup(struct intel_ring_buffer *ring)
{
	if (!ring->private)
		return;

	cleanup_pipe_control(ring);
}

417
static void
418 419 420
update_mboxes(struct intel_ring_buffer *ring,
	    u32 seqno,
	    u32 mmio_offset)
421
{
422 423 424 425
	intel_ring_emit(ring, MI_SEMAPHORE_MBOX |
			      MI_SEMAPHORE_GLOBAL_GTT |
			      MI_SEMAPHORE_REGISTER |
			      MI_SEMAPHORE_UPDATE);
426
	intel_ring_emit(ring, seqno);
427
	intel_ring_emit(ring, mmio_offset);
428 429
}

430 431 432 433 434 435 436 437 438
/**
 * gen6_add_request - Update the semaphore mailbox registers
 * 
 * @ring - ring that is adding a request
 * @seqno - return seqno stuck into the ring
 *
 * Update the mailbox registers in the *other* rings with the current seqno.
 * This acts like a signal in the canonical semaphore.
 */
439 440
static int
gen6_add_request(struct intel_ring_buffer *ring,
441
		 u32 *seqno)
442
{
443 444
	u32 mbox1_reg;
	u32 mbox2_reg;
445 446 447 448 449 450
	int ret;

	ret = intel_ring_begin(ring, 10);
	if (ret)
		return ret;

451 452
	mbox1_reg = ring->signal_mbox[0];
	mbox2_reg = ring->signal_mbox[1];
453

454
	*seqno = i915_gem_next_request_seqno(ring);
455 456 457

	update_mboxes(ring, *seqno, mbox1_reg);
	update_mboxes(ring, *seqno, mbox2_reg);
458 459
	intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
	intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
460
	intel_ring_emit(ring, *seqno);
461 462 463 464 465 466
	intel_ring_emit(ring, MI_USER_INTERRUPT);
	intel_ring_advance(ring);

	return 0;
}

467 468 469 470 471 472 473 474
/**
 * intel_ring_sync - sync the waiter to the signaller on seqno
 *
 * @waiter - ring that is waiting
 * @signaller - ring which has, or will signal
 * @seqno - seqno which the waiter will block on
 */
static int
475 476 477
gen6_ring_sync(struct intel_ring_buffer *waiter,
	       struct intel_ring_buffer *signaller,
	       u32 seqno)
478 479
{
	int ret;
480 481 482
	u32 dw1 = MI_SEMAPHORE_MBOX |
		  MI_SEMAPHORE_COMPARE |
		  MI_SEMAPHORE_REGISTER;
483

484 485 486 487 488 489
	/* Throughout all of the GEM code, seqno passed implies our current
	 * seqno is >= the last seqno executed. However for hardware the
	 * comparison is strictly greater than.
	 */
	seqno -= 1;

490 491 492
	WARN_ON(signaller->semaphore_register[waiter->id] ==
		MI_SEMAPHORE_SYNC_INVALID);

493
	ret = intel_ring_begin(waiter, 4);
494 495 496
	if (ret)
		return ret;

497 498
	intel_ring_emit(waiter,
			dw1 | signaller->semaphore_register[waiter->id]);
499 500 501 502
	intel_ring_emit(waiter, seqno);
	intel_ring_emit(waiter, 0);
	intel_ring_emit(waiter, MI_NOOP);
	intel_ring_advance(waiter);
503 504 505 506

	return 0;
}

507 508
#define PIPE_CONTROL_FLUSH(ring__, addr__)					\
do {									\
509 510
	intel_ring_emit(ring__, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |		\
		 PIPE_CONTROL_DEPTH_STALL);				\
511 512 513 514 515 516 517 518 519
	intel_ring_emit(ring__, (addr__) | PIPE_CONTROL_GLOBAL_GTT);			\
	intel_ring_emit(ring__, 0);							\
	intel_ring_emit(ring__, 0);							\
} while (0)

static int
pc_render_add_request(struct intel_ring_buffer *ring,
		      u32 *result)
{
520
	u32 seqno = i915_gem_next_request_seqno(ring);
521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536
	struct pipe_control *pc = ring->private;
	u32 scratch_addr = pc->gtt_offset + 128;
	int ret;

	/* For Ironlake, MI_USER_INTERRUPT was deprecated and apparently
	 * incoherent with writes to memory, i.e. completely fubar,
	 * so we need to use PIPE_NOTIFY instead.
	 *
	 * However, we also need to workaround the qword write
	 * incoherence by flushing the 6 PIPE_NOTIFY buffers out to
	 * memory before requesting an interrupt.
	 */
	ret = intel_ring_begin(ring, 32);
	if (ret)
		return ret;

537
	intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |
538 539
			PIPE_CONTROL_WRITE_FLUSH |
			PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE);
540 541 542 543 544 545 546 547 548 549 550 551 552 553
	intel_ring_emit(ring, pc->gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
	intel_ring_emit(ring, seqno);
	intel_ring_emit(ring, 0);
	PIPE_CONTROL_FLUSH(ring, scratch_addr);
	scratch_addr += 128; /* write to separate cachelines */
	PIPE_CONTROL_FLUSH(ring, scratch_addr);
	scratch_addr += 128;
	PIPE_CONTROL_FLUSH(ring, scratch_addr);
	scratch_addr += 128;
	PIPE_CONTROL_FLUSH(ring, scratch_addr);
	scratch_addr += 128;
	PIPE_CONTROL_FLUSH(ring, scratch_addr);
	scratch_addr += 128;
	PIPE_CONTROL_FLUSH(ring, scratch_addr);
554

555
	intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE |
556 557
			PIPE_CONTROL_WRITE_FLUSH |
			PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE |
558 559 560 561 562 563 564 565 566 567
			PIPE_CONTROL_NOTIFY);
	intel_ring_emit(ring, pc->gtt_offset | PIPE_CONTROL_GLOBAL_GTT);
	intel_ring_emit(ring, seqno);
	intel_ring_emit(ring, 0);
	intel_ring_advance(ring);

	*result = seqno;
	return 0;
}

568 569 570 571 572 573 574 575
static u32
gen6_ring_get_seqno(struct intel_ring_buffer *ring)
{
	struct drm_device *dev = ring->dev;

	/* Workaround to force correct ordering between irq and seqno writes on
	 * ivb (and maybe also on snb) by reading from a CS register (like
	 * ACTHD) before reading the status page. */
576
	if (IS_GEN6(dev) || IS_GEN7(dev))
577 578 579 580
		intel_ring_get_active_head(ring);
	return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
}

581
static u32
582
ring_get_seqno(struct intel_ring_buffer *ring)
583
{
584 585 586
	return intel_read_status_page(ring, I915_GEM_HWS_INDEX);
}

587 588 589 590 591 592 593
static u32
pc_render_get_seqno(struct intel_ring_buffer *ring)
{
	struct pipe_control *pc = ring->private;
	return pc->cpu_page[0];
}

594 595 596 597 598 599 600 601 602 603
static bool
gen5_ring_get_irq(struct intel_ring_buffer *ring)
{
	struct drm_device *dev = ring->dev;
	drm_i915_private_t *dev_priv = dev->dev_private;

	if (!dev->irq_enabled)
		return false;

	spin_lock(&ring->irq_lock);
604 605 606 607 608
	if (ring->irq_refcount++ == 0) {
		dev_priv->gt_irq_mask &= ~ring->irq_enable_mask;
		I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
		POSTING_READ(GTIMR);
	}
609 610 611 612 613 614 615 616 617 618 619 620
	spin_unlock(&ring->irq_lock);

	return true;
}

static void
gen5_ring_put_irq(struct intel_ring_buffer *ring)
{
	struct drm_device *dev = ring->dev;
	drm_i915_private_t *dev_priv = dev->dev_private;

	spin_lock(&ring->irq_lock);
621 622 623 624 625
	if (--ring->irq_refcount == 0) {
		dev_priv->gt_irq_mask |= ring->irq_enable_mask;
		I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
		POSTING_READ(GTIMR);
	}
626 627 628
	spin_unlock(&ring->irq_lock);
}

629
static bool
630
i9xx_ring_get_irq(struct intel_ring_buffer *ring)
631
{
632
	struct drm_device *dev = ring->dev;
633
	drm_i915_private_t *dev_priv = dev->dev_private;
634

635 636 637
	if (!dev->irq_enabled)
		return false;

638
	spin_lock(&ring->irq_lock);
639 640 641 642 643
	if (ring->irq_refcount++ == 0) {
		dev_priv->irq_mask &= ~ring->irq_enable_mask;
		I915_WRITE(IMR, dev_priv->irq_mask);
		POSTING_READ(IMR);
	}
644
	spin_unlock(&ring->irq_lock);
645 646

	return true;
647 648
}

649
static void
650
i9xx_ring_put_irq(struct intel_ring_buffer *ring)
651
{
652
	struct drm_device *dev = ring->dev;
653
	drm_i915_private_t *dev_priv = dev->dev_private;
654

655
	spin_lock(&ring->irq_lock);
656 657 658 659 660
	if (--ring->irq_refcount == 0) {
		dev_priv->irq_mask |= ring->irq_enable_mask;
		I915_WRITE(IMR, dev_priv->irq_mask);
		POSTING_READ(IMR);
	}
661
	spin_unlock(&ring->irq_lock);
662 663
}

664
void intel_ring_setup_status_page(struct intel_ring_buffer *ring)
665
{
666
	struct drm_device *dev = ring->dev;
667
	drm_i915_private_t *dev_priv = ring->dev->dev_private;
668 669 670 671 672 673 674
	u32 mmio = 0;

	/* The ring status page addresses are no longer next to the rest of
	 * the ring registers as of gen7.
	 */
	if (IS_GEN7(dev)) {
		switch (ring->id) {
675
		case RCS:
676 677
			mmio = RENDER_HWS_PGA_GEN7;
			break;
678
		case BCS:
679 680
			mmio = BLT_HWS_PGA_GEN7;
			break;
681
		case VCS:
682 683 684 685 686 687 688 689 690
			mmio = BSD_HWS_PGA_GEN7;
			break;
		}
	} else if (IS_GEN6(ring->dev)) {
		mmio = RING_HWS_PGA_GEN6(ring->mmio_base);
	} else {
		mmio = RING_HWS_PGA(ring->mmio_base);
	}

691 692
	I915_WRITE(mmio, (u32)ring->status_page.gfx_addr);
	POSTING_READ(mmio);
693 694
}

695
static int
696 697 698
bsd_ring_flush(struct intel_ring_buffer *ring,
	       u32     invalidate_domains,
	       u32     flush_domains)
699
{
700 701 702 703 704 705 706 707 708 709
	int ret;

	ret = intel_ring_begin(ring, 2);
	if (ret)
		return ret;

	intel_ring_emit(ring, MI_FLUSH);
	intel_ring_emit(ring, MI_NOOP);
	intel_ring_advance(ring);
	return 0;
710 711
}

712
static int
713
i9xx_add_request(struct intel_ring_buffer *ring,
714
		 u32 *result)
715 716
{
	u32 seqno;
717 718 719 720 721
	int ret;

	ret = intel_ring_begin(ring, 4);
	if (ret)
		return ret;
722

723
	seqno = i915_gem_next_request_seqno(ring);
724

725 726 727 728 729
	intel_ring_emit(ring, MI_STORE_DWORD_INDEX);
	intel_ring_emit(ring, I915_GEM_HWS_INDEX << MI_STORE_DWORD_INDEX_SHIFT);
	intel_ring_emit(ring, seqno);
	intel_ring_emit(ring, MI_USER_INTERRUPT);
	intel_ring_advance(ring);
730

731 732
	*result = seqno;
	return 0;
733 734
}

735
static bool
736
gen6_ring_get_irq(struct intel_ring_buffer *ring)
737 738
{
	struct drm_device *dev = ring->dev;
739
	drm_i915_private_t *dev_priv = dev->dev_private;
740 741 742 743

	if (!dev->irq_enabled)
	       return false;

744 745 746
	/* It looks like we need to prevent the gt from suspending while waiting
	 * for an notifiy irq, otherwise irqs seem to get lost on at least the
	 * blt/bsd rings on ivb. */
747
	gen6_gt_force_wake_get(dev_priv);
748

749
	spin_lock(&ring->irq_lock);
750
	if (ring->irq_refcount++ == 0) {
D
Daniel Vetter 已提交
751
		I915_WRITE_IMR(ring, ~ring->irq_enable_mask);
752 753 754
		dev_priv->gt_irq_mask &= ~ring->irq_enable_mask;
		I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
		POSTING_READ(GTIMR);
755
	}
756
	spin_unlock(&ring->irq_lock);
757 758 759 760 761

	return true;
}

static void
762
gen6_ring_put_irq(struct intel_ring_buffer *ring)
763 764
{
	struct drm_device *dev = ring->dev;
765
	drm_i915_private_t *dev_priv = dev->dev_private;
766

767
	spin_lock(&ring->irq_lock);
768
	if (--ring->irq_refcount == 0) {
D
Daniel Vetter 已提交
769
		I915_WRITE_IMR(ring, ~0);
770 771 772
		dev_priv->gt_irq_mask |= ring->irq_enable_mask;
		I915_WRITE(GTIMR, dev_priv->gt_irq_mask);
		POSTING_READ(GTIMR);
773
	}
774
	spin_unlock(&ring->irq_lock);
775

776
	gen6_gt_force_wake_put(dev_priv);
777 778 779
}

static int
780
i965_dispatch_execbuffer(struct intel_ring_buffer *ring, u32 offset, u32 length)
781
{
782
	int ret;
783

784 785 786 787
	ret = intel_ring_begin(ring, 2);
	if (ret)
		return ret;

788
	intel_ring_emit(ring,
789
			MI_BATCH_BUFFER_START | (2 << 6) |
790
			MI_BATCH_NON_SECURE_I965);
791
	intel_ring_emit(ring, offset);
792 793
	intel_ring_advance(ring);

794 795 796
	return 0;
}

797
static int
798
i830_dispatch_execbuffer(struct intel_ring_buffer *ring,
799
				u32 offset, u32 len)
800
{
801
	int ret;
802

803 804 805
	ret = intel_ring_begin(ring, 4);
	if (ret)
		return ret;
806

807 808 809 810 811
	intel_ring_emit(ring, MI_BATCH_BUFFER);
	intel_ring_emit(ring, offset | MI_BATCH_NON_SECURE);
	intel_ring_emit(ring, offset + len - 8);
	intel_ring_emit(ring, 0);
	intel_ring_advance(ring);
812

813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
	return 0;
}

static int
i915_dispatch_execbuffer(struct intel_ring_buffer *ring,
				u32 offset, u32 len)
{
	int ret;

	ret = intel_ring_begin(ring, 2);
	if (ret)
		return ret;

	intel_ring_emit(ring, MI_BATCH_BUFFER_START | (2 << 6));
	intel_ring_emit(ring, offset | MI_BATCH_NON_SECURE);
828
	intel_ring_advance(ring);
829 830 831 832

	return 0;
}

833
static void cleanup_status_page(struct intel_ring_buffer *ring)
834
{
835
	drm_i915_private_t *dev_priv = ring->dev->dev_private;
836
	struct drm_i915_gem_object *obj;
837

838 839
	obj = ring->status_page.obj;
	if (obj == NULL)
840 841
		return;

842
	kunmap(obj->pages[0]);
843
	i915_gem_object_unpin(obj);
844
	drm_gem_object_unreference(&obj->base);
845
	ring->status_page.obj = NULL;
846 847 848 849

	memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
}

850
static int init_status_page(struct intel_ring_buffer *ring)
851
{
852
	struct drm_device *dev = ring->dev;
853
	drm_i915_private_t *dev_priv = dev->dev_private;
854
	struct drm_i915_gem_object *obj;
855 856 857 858 859 860 861 862
	int ret;

	obj = i915_gem_alloc_object(dev, 4096);
	if (obj == NULL) {
		DRM_ERROR("Failed to allocate status page\n");
		ret = -ENOMEM;
		goto err;
	}
863 864

	i915_gem_object_set_cache_level(obj, I915_CACHE_LLC);
865

866
	ret = i915_gem_object_pin(obj, 4096, true);
867 868 869 870
	if (ret != 0) {
		goto err_unref;
	}

871 872
	ring->status_page.gfx_addr = obj->gtt_offset;
	ring->status_page.page_addr = kmap(obj->pages[0]);
873
	if (ring->status_page.page_addr == NULL) {
874 875 876
		memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map));
		goto err_unpin;
	}
877 878
	ring->status_page.obj = obj;
	memset(ring->status_page.page_addr, 0, PAGE_SIZE);
879

880
	intel_ring_setup_status_page(ring);
881 882
	DRM_DEBUG_DRIVER("%s hws offset: 0x%08x\n",
			ring->name, ring->status_page.gfx_addr);
883 884 885 886 887 888

	return 0;

err_unpin:
	i915_gem_object_unpin(obj);
err_unref:
889
	drm_gem_object_unreference(&obj->base);
890
err:
891
	return ret;
892 893
}

894
int intel_init_ring_buffer(struct drm_device *dev,
895
			   struct intel_ring_buffer *ring)
896
{
897
	struct drm_i915_gem_object *obj;
898 899
	int ret;

900
	ring->dev = dev;
901 902
	INIT_LIST_HEAD(&ring->active_list);
	INIT_LIST_HEAD(&ring->request_list);
903
	INIT_LIST_HEAD(&ring->gpu_write_list);
904
	ring->size = 32 * PAGE_SIZE;
905

906
	init_waitqueue_head(&ring->irq_queue);
907
	spin_lock_init(&ring->irq_lock);
908

909
	if (I915_NEED_GFX_HWS(dev)) {
910
		ret = init_status_page(ring);
911 912 913
		if (ret)
			return ret;
	}
914

915
	obj = i915_gem_alloc_object(dev, ring->size);
916 917
	if (obj == NULL) {
		DRM_ERROR("Failed to allocate ringbuffer\n");
918
		ret = -ENOMEM;
919
		goto err_hws;
920 921
	}

922
	ring->obj = obj;
923

924
	ret = i915_gem_object_pin(obj, PAGE_SIZE, true);
925 926
	if (ret)
		goto err_unref;
927

928
	ring->map.size = ring->size;
929
	ring->map.offset = dev->agp->base + obj->gtt_offset;
930 931 932 933 934 935 936
	ring->map.type = 0;
	ring->map.flags = 0;
	ring->map.mtrr = 0;

	drm_core_ioremap_wc(&ring->map, dev);
	if (ring->map.handle == NULL) {
		DRM_ERROR("Failed to map ringbuffer.\n");
937
		ret = -EINVAL;
938
		goto err_unpin;
939 940
	}

941
	ring->virtual_start = ring->map.handle;
942
	ret = ring->init(ring);
943 944
	if (ret)
		goto err_unmap;
945

946 947 948 949 950
	/* Workaround an erratum on the i830 which causes a hang if
	 * the TAIL pointer points to within the last 2 cachelines
	 * of the buffer.
	 */
	ring->effective_size = ring->size;
951
	if (IS_I830(ring->dev) || IS_845G(ring->dev))
952 953
		ring->effective_size -= 128;

954
	return 0;
955 956 957 958 959 960

err_unmap:
	drm_core_ioremapfree(&ring->map, dev);
err_unpin:
	i915_gem_object_unpin(obj);
err_unref:
961 962
	drm_gem_object_unreference(&obj->base);
	ring->obj = NULL;
963
err_hws:
964
	cleanup_status_page(ring);
965
	return ret;
966 967
}

968
void intel_cleanup_ring_buffer(struct intel_ring_buffer *ring)
969
{
970 971 972
	struct drm_i915_private *dev_priv;
	int ret;

973
	if (ring->obj == NULL)
974 975
		return;

976 977
	/* Disable the ring buffer. The ring must be idle at this point */
	dev_priv = ring->dev->dev_private;
978
	ret = intel_wait_ring_idle(ring);
979 980 981 982
	if (ret)
		DRM_ERROR("failed to quiesce %s whilst cleaning up: %d\n",
			  ring->name, ret);

983 984
	I915_WRITE_CTL(ring, 0);

985
	drm_core_ioremapfree(&ring->map, ring->dev);
986

987 988 989
	i915_gem_object_unpin(ring->obj);
	drm_gem_object_unreference(&ring->obj->base);
	ring->obj = NULL;
990

Z
Zou Nan hai 已提交
991 992 993
	if (ring->cleanup)
		ring->cleanup(ring);

994
	cleanup_status_page(ring);
995 996
}

997
static int intel_wrap_ring_buffer(struct intel_ring_buffer *ring)
998
{
999
	unsigned int *virt;
1000
	int rem = ring->size - ring->tail;
1001

1002
	if (ring->space < rem) {
1003
		int ret = intel_wait_ring_buffer(ring, rem);
1004 1005 1006 1007
		if (ret)
			return ret;
	}

1008
	virt = (unsigned int *)(ring->virtual_start + ring->tail);
1009 1010
	rem /= 8;
	while (rem--) {
1011
		*virt++ = MI_NOOP;
1012 1013
		*virt++ = MI_NOOP;
	}
1014

1015
	ring->tail = 0;
1016
	ring->space = ring_space(ring);
1017 1018 1019 1020

	return 0;
}

1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097
static int intel_ring_wait_seqno(struct intel_ring_buffer *ring, u32 seqno)
{
	struct drm_i915_private *dev_priv = ring->dev->dev_private;
	bool was_interruptible;
	int ret;

	/* XXX As we have not yet audited all the paths to check that
	 * they are ready for ERESTARTSYS from intel_ring_begin, do not
	 * allow us to be interruptible by a signal.
	 */
	was_interruptible = dev_priv->mm.interruptible;
	dev_priv->mm.interruptible = false;

	ret = i915_wait_request(ring, seqno, true);

	dev_priv->mm.interruptible = was_interruptible;

	return ret;
}

static int intel_ring_wait_request(struct intel_ring_buffer *ring, int n)
{
	struct drm_i915_gem_request *request;
	u32 seqno = 0;
	int ret;

	i915_gem_retire_requests_ring(ring);

	if (ring->last_retired_head != -1) {
		ring->head = ring->last_retired_head;
		ring->last_retired_head = -1;
		ring->space = ring_space(ring);
		if (ring->space >= n)
			return 0;
	}

	list_for_each_entry(request, &ring->request_list, list) {
		int space;

		if (request->tail == -1)
			continue;

		space = request->tail - (ring->tail + 8);
		if (space < 0)
			space += ring->size;
		if (space >= n) {
			seqno = request->seqno;
			break;
		}

		/* Consume this request in case we need more space than
		 * is available and so need to prevent a race between
		 * updating last_retired_head and direct reads of
		 * I915_RING_HEAD. It also provides a nice sanity check.
		 */
		request->tail = -1;
	}

	if (seqno == 0)
		return -ENOSPC;

	ret = intel_ring_wait_seqno(ring, seqno);
	if (ret)
		return ret;

	if (WARN_ON(ring->last_retired_head == -1))
		return -ENOSPC;

	ring->head = ring->last_retired_head;
	ring->last_retired_head = -1;
	ring->space = ring_space(ring);
	if (WARN_ON(ring->space < n))
		return -ENOSPC;

	return 0;
}

1098
int intel_wait_ring_buffer(struct intel_ring_buffer *ring, int n)
1099
{
1100
	struct drm_device *dev = ring->dev;
1101
	struct drm_i915_private *dev_priv = dev->dev_private;
1102
	unsigned long end;
1103
	int ret;
1104

1105 1106 1107 1108
	ret = intel_ring_wait_request(ring, n);
	if (ret != -ENOSPC)
		return ret;

C
Chris Wilson 已提交
1109
	trace_i915_ring_wait_begin(ring);
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119
	if (drm_core_check_feature(dev, DRIVER_GEM))
		/* With GEM the hangcheck timer should kick us out of the loop,
		 * leaving it early runs the risk of corrupting GEM state (due
		 * to running on almost untested codepaths). But on resume
		 * timers don't work yet, so prevent a complete hang in that
		 * case by choosing an insanely large timeout. */
		end = jiffies + 60 * HZ;
	else
		end = jiffies + 3 * HZ;

1120
	do {
1121 1122
		ring->head = I915_READ_HEAD(ring);
		ring->space = ring_space(ring);
1123
		if (ring->space >= n) {
C
Chris Wilson 已提交
1124
			trace_i915_ring_wait_end(ring);
1125 1126 1127 1128 1129 1130 1131 1132
			return 0;
		}

		if (dev->primary->master) {
			struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
			if (master_priv->sarea_priv)
				master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
		}
1133

1134
		msleep(1);
1135 1136
		if (atomic_read(&dev_priv->mm.wedged))
			return -EAGAIN;
1137
	} while (!time_after(jiffies, end));
C
Chris Wilson 已提交
1138
	trace_i915_ring_wait_end(ring);
1139 1140
	return -EBUSY;
}
1141

1142 1143
int intel_ring_begin(struct intel_ring_buffer *ring,
		     int num_dwords)
1144
{
1145
	struct drm_i915_private *dev_priv = ring->dev->dev_private;
1146
	int n = 4*num_dwords;
1147
	int ret;
1148

1149 1150 1151
	if (unlikely(atomic_read(&dev_priv->mm.wedged)))
		return -EIO;

1152
	if (unlikely(ring->tail + n > ring->effective_size)) {
1153 1154 1155 1156
		ret = intel_wrap_ring_buffer(ring);
		if (unlikely(ret))
			return ret;
	}
1157

1158 1159 1160 1161 1162
	if (unlikely(ring->space < n)) {
		ret = intel_wait_ring_buffer(ring, n);
		if (unlikely(ret))
			return ret;
	}
1163 1164

	ring->space -= n;
1165
	return 0;
1166
}
1167

1168
void intel_ring_advance(struct intel_ring_buffer *ring)
1169
{
1170
	ring->tail &= ring->size - 1;
1171
	ring->write_tail(ring, ring->tail);
1172
}
1173

1174

1175
static void gen6_bsd_ring_write_tail(struct intel_ring_buffer *ring,
1176
				     u32 value)
1177
{
1178
	drm_i915_private_t *dev_priv = ring->dev->dev_private;
1179 1180

       /* Every tail move must follow the sequence below */
1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
	I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL,
		GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_MODIFY_MASK |
		GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_DISABLE);
	I915_WRITE(GEN6_BSD_RNCID, 0x0);

	if (wait_for((I915_READ(GEN6_BSD_SLEEP_PSMI_CONTROL) &
		GEN6_BSD_SLEEP_PSMI_CONTROL_IDLE_INDICATOR) == 0,
		50))
	DRM_ERROR("timed out waiting for IDLE Indicator\n");

	I915_WRITE_TAIL(ring, value);
	I915_WRITE(GEN6_BSD_SLEEP_PSMI_CONTROL,
		GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_MODIFY_MASK |
		GEN6_BSD_SLEEP_PSMI_CONTROL_RC_ILDL_MESSAGE_ENABLE);
1195 1196
}

1197
static int gen6_ring_flush(struct intel_ring_buffer *ring,
1198
			   u32 invalidate, u32 flush)
1199
{
1200
	uint32_t cmd;
1201 1202 1203 1204 1205 1206
	int ret;

	ret = intel_ring_begin(ring, 4);
	if (ret)
		return ret;

1207 1208 1209 1210
	cmd = MI_FLUSH_DW;
	if (invalidate & I915_GEM_GPU_DOMAINS)
		cmd |= MI_INVALIDATE_TLB | MI_INVALIDATE_BSD;
	intel_ring_emit(ring, cmd);
1211 1212
	intel_ring_emit(ring, 0);
	intel_ring_emit(ring, 0);
1213
	intel_ring_emit(ring, MI_NOOP);
1214 1215
	intel_ring_advance(ring);
	return 0;
1216 1217 1218
}

static int
1219
gen6_ring_dispatch_execbuffer(struct intel_ring_buffer *ring,
1220
			      u32 offset, u32 len)
1221
{
1222
	int ret;
1223

1224 1225 1226
	ret = intel_ring_begin(ring, 2);
	if (ret)
		return ret;
1227

1228 1229 1230 1231
	intel_ring_emit(ring, MI_BATCH_BUFFER_START | MI_BATCH_NON_SECURE_I965);
	/* bit0-7 is the length on GEN6+ */
	intel_ring_emit(ring, offset);
	intel_ring_advance(ring);
1232

1233
	return 0;
1234 1235
}

1236 1237
/* Blitter support (SandyBridge+) */

1238
static int blt_ring_flush(struct intel_ring_buffer *ring,
1239
			  u32 invalidate, u32 flush)
Z
Zou Nan hai 已提交
1240
{
1241
	uint32_t cmd;
1242 1243
	int ret;

1244
	ret = intel_ring_begin(ring, 4);
1245 1246 1247
	if (ret)
		return ret;

1248 1249 1250 1251
	cmd = MI_FLUSH_DW;
	if (invalidate & I915_GEM_DOMAIN_RENDER)
		cmd |= MI_INVALIDATE_TLB;
	intel_ring_emit(ring, cmd);
1252 1253
	intel_ring_emit(ring, 0);
	intel_ring_emit(ring, 0);
1254
	intel_ring_emit(ring, MI_NOOP);
1255 1256
	intel_ring_advance(ring);
	return 0;
Z
Zou Nan hai 已提交
1257 1258
}

1259 1260 1261
int intel_init_render_ring_buffer(struct drm_device *dev)
{
	drm_i915_private_t *dev_priv = dev->dev_private;
1262
	struct intel_ring_buffer *ring = &dev_priv->ring[RCS];
1263

1264 1265 1266 1267
	ring->name = "render ring";
	ring->id = RCS;
	ring->mmio_base = RENDER_RING_BASE;

1268 1269
	if (INTEL_INFO(dev)->gen >= 6) {
		ring->add_request = gen6_add_request;
1270
		ring->flush = gen6_render_ring_flush;
1271 1272
		ring->irq_get = gen6_ring_get_irq;
		ring->irq_put = gen6_ring_put_irq;
D
Daniel Vetter 已提交
1273
		ring->irq_enable_mask = GT_USER_INTERRUPT;
1274
		ring->get_seqno = gen6_ring_get_seqno;
1275
		ring->sync_to = gen6_ring_sync;
1276 1277 1278 1279 1280
		ring->semaphore_register[0] = MI_SEMAPHORE_SYNC_INVALID;
		ring->semaphore_register[1] = MI_SEMAPHORE_SYNC_RV;
		ring->semaphore_register[2] = MI_SEMAPHORE_SYNC_RB;
		ring->signal_mbox[0] = GEN6_VRSYNC;
		ring->signal_mbox[1] = GEN6_BRSYNC;
1281 1282
	} else if (IS_GEN5(dev)) {
		ring->add_request = pc_render_add_request;
1283
		ring->flush = render_ring_flush;
1284
		ring->get_seqno = pc_render_get_seqno;
1285 1286
		ring->irq_get = gen5_ring_get_irq;
		ring->irq_put = gen5_ring_put_irq;
1287
		ring->irq_enable_mask = GT_USER_INTERRUPT | GT_PIPE_NOTIFY;
1288
	} else {
1289
		ring->add_request = i9xx_add_request;
1290 1291
		ring->flush = render_ring_flush;
		ring->get_seqno = ring_get_seqno;
1292 1293 1294
		ring->irq_get = i9xx_ring_get_irq;
		ring->irq_put = i9xx_ring_put_irq;
		ring->irq_enable_mask = I915_USER_INTERRUPT;
1295
	}
1296
	ring->write_tail = ring_write_tail;
1297 1298 1299 1300 1301 1302 1303 1304
	if (INTEL_INFO(dev)->gen >= 6)
		ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
	else if (INTEL_INFO(dev)->gen >= 4)
		ring->dispatch_execbuffer = i965_dispatch_execbuffer;
	else if (IS_I830(dev) || IS_845G(dev))
		ring->dispatch_execbuffer = i830_dispatch_execbuffer;
	else
		ring->dispatch_execbuffer = i915_dispatch_execbuffer;
1305 1306 1307
	ring->init = init_render_ring;
	ring->cleanup = render_ring_cleanup;

1308 1309

	if (!I915_NEED_GFX_HWS(dev)) {
1310 1311
		ring->status_page.page_addr = dev_priv->status_page_dmah->vaddr;
		memset(ring->status_page.page_addr, 0, PAGE_SIZE);
1312 1313
	}

1314
	return intel_init_ring_buffer(dev, ring);
1315 1316
}

1317 1318 1319 1320 1321
int intel_render_ring_init_dri(struct drm_device *dev, u64 start, u32 size)
{
	drm_i915_private_t *dev_priv = dev->dev_private;
	struct intel_ring_buffer *ring = &dev_priv->ring[RCS];

1322 1323 1324 1325
	ring->name = "render ring";
	ring->id = RCS;
	ring->mmio_base = RENDER_RING_BASE;

1326
	if (INTEL_INFO(dev)->gen >= 6) {
1327 1328
		/* non-kms not supported on gen6+ */
		return -ENODEV;
1329
	}
1330 1331 1332 1333 1334 1335 1336 1337 1338 1339

	/* Note: gem is not supported on gen5/ilk without kms (the corresponding
	 * gem_init ioctl returns with -ENODEV). Hence we do not need to set up
	 * the special gen5 functions. */
	ring->add_request = i9xx_add_request;
	ring->flush = render_ring_flush;
	ring->get_seqno = ring_get_seqno;
	ring->irq_get = i9xx_ring_get_irq;
	ring->irq_put = i9xx_ring_put_irq;
	ring->irq_enable_mask = I915_USER_INTERRUPT;
1340
	ring->write_tail = ring_write_tail;
1341 1342 1343 1344 1345 1346
	if (INTEL_INFO(dev)->gen >= 4)
		ring->dispatch_execbuffer = i965_dispatch_execbuffer;
	else if (IS_I830(dev) || IS_845G(dev))
		ring->dispatch_execbuffer = i830_dispatch_execbuffer;
	else
		ring->dispatch_execbuffer = i915_dispatch_execbuffer;
1347 1348
	ring->init = init_render_ring;
	ring->cleanup = render_ring_cleanup;
1349

1350 1351 1352
	if (!I915_NEED_GFX_HWS(dev))
		ring->status_page.page_addr = dev_priv->status_page_dmah->vaddr;

1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
	ring->dev = dev;
	INIT_LIST_HEAD(&ring->active_list);
	INIT_LIST_HEAD(&ring->request_list);
	INIT_LIST_HEAD(&ring->gpu_write_list);

	ring->size = size;
	ring->effective_size = ring->size;
	if (IS_I830(ring->dev))
		ring->effective_size -= 128;

	ring->map.offset = start;
	ring->map.size = size;
	ring->map.type = 0;
	ring->map.flags = 0;
	ring->map.mtrr = 0;

	drm_core_ioremap_wc(&ring->map, dev);
	if (ring->map.handle == NULL) {
		DRM_ERROR("can not ioremap virtual address for"
			  " ring buffer\n");
		return -ENOMEM;
	}

	ring->virtual_start = (void __force __iomem *)ring->map.handle;
	return 0;
}

1380 1381 1382
int intel_init_bsd_ring_buffer(struct drm_device *dev)
{
	drm_i915_private_t *dev_priv = dev->dev_private;
1383
	struct intel_ring_buffer *ring = &dev_priv->ring[VCS];
1384

1385 1386 1387
	ring->name = "bsd ring";
	ring->id = VCS;

1388
	ring->write_tail = ring_write_tail;
1389 1390
	if (IS_GEN6(dev) || IS_GEN7(dev)) {
		ring->mmio_base = GEN6_BSD_RING_BASE;
1391 1392 1393
		/* gen6 bsd needs a special wa for tail updates */
		if (IS_GEN6(dev))
			ring->write_tail = gen6_bsd_ring_write_tail;
1394 1395 1396 1397 1398 1399 1400
		ring->flush = gen6_ring_flush;
		ring->add_request = gen6_add_request;
		ring->get_seqno = gen6_ring_get_seqno;
		ring->irq_enable_mask = GEN6_BSD_USER_INTERRUPT;
		ring->irq_get = gen6_ring_get_irq;
		ring->irq_put = gen6_ring_put_irq;
		ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
1401
		ring->sync_to = gen6_ring_sync;
1402 1403 1404 1405 1406 1407 1408 1409
		ring->semaphore_register[0] = MI_SEMAPHORE_SYNC_VR;
		ring->semaphore_register[1] = MI_SEMAPHORE_SYNC_INVALID;
		ring->semaphore_register[2] = MI_SEMAPHORE_SYNC_VB;
		ring->signal_mbox[0] = GEN6_RVSYNC;
		ring->signal_mbox[1] = GEN6_BVSYNC;
	} else {
		ring->mmio_base = BSD_RING_BASE;
		ring->flush = bsd_ring_flush;
1410
		ring->add_request = i9xx_add_request;
1411
		ring->get_seqno = ring_get_seqno;
1412
		if (IS_GEN5(dev)) {
1413
			ring->irq_enable_mask = GT_BSD_USER_INTERRUPT;
1414 1415 1416
			ring->irq_get = gen5_ring_get_irq;
			ring->irq_put = gen5_ring_put_irq;
		} else {
1417
			ring->irq_enable_mask = I915_BSD_USER_INTERRUPT;
1418 1419 1420
			ring->irq_get = i9xx_ring_get_irq;
			ring->irq_put = i9xx_ring_put_irq;
		}
1421
		ring->dispatch_execbuffer = i965_dispatch_execbuffer;
1422 1423 1424
	}
	ring->init = init_ring_common;

1425

1426
	return intel_init_ring_buffer(dev, ring);
1427
}
1428 1429 1430 1431

int intel_init_blt_ring_buffer(struct drm_device *dev)
{
	drm_i915_private_t *dev_priv = dev->dev_private;
1432
	struct intel_ring_buffer *ring = &dev_priv->ring[BCS];
1433

1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
	ring->name = "blitter ring";
	ring->id = BCS;

	ring->mmio_base = BLT_RING_BASE;
	ring->write_tail = ring_write_tail;
	ring->flush = blt_ring_flush;
	ring->add_request = gen6_add_request;
	ring->get_seqno = gen6_ring_get_seqno;
	ring->irq_enable_mask = GEN6_BLITTER_USER_INTERRUPT;
	ring->irq_get = gen6_ring_get_irq;
	ring->irq_put = gen6_ring_put_irq;
	ring->dispatch_execbuffer = gen6_ring_dispatch_execbuffer;
1446
	ring->sync_to = gen6_ring_sync;
1447 1448 1449 1450 1451 1452
	ring->semaphore_register[0] = MI_SEMAPHORE_SYNC_BR;
	ring->semaphore_register[1] = MI_SEMAPHORE_SYNC_BV;
	ring->semaphore_register[2] = MI_SEMAPHORE_SYNC_INVALID;
	ring->signal_mbox[0] = GEN6_RBSYNC;
	ring->signal_mbox[1] = GEN6_VBSYNC;
	ring->init = init_ring_common;
1453

1454
	return intel_init_ring_buffer(dev, ring);
1455
}