i915_gem_tiling.c 13.0 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
/*
 * Copyright © 2008 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>
 *
 */

28 29 30 31
#include <linux/string.h>
#include <linux/bitops.h>
#include <drm/drmP.h>
#include <drm/i915_drm.h>
32 33
#include "i915_drv.h"

34 35
/**
 * DOC: buffer object tiling
36
 *
37 38
 * i915_gem_set_tiling_ioctl() and i915_gem_get_tiling_ioctl() is the userspace
 * interface to declare fence register requirements.
39
 *
40 41 42
 * In principle GEM doesn't care at all about the internal data layout of an
 * object, and hence it also doesn't care about tiling or swizzling. There's two
 * exceptions:
43
 *
44 45 46 47 48 49 50 51 52 53
 * - For X and Y tiling the hardware provides detilers for CPU access, so called
 *   fences. Since there's only a limited amount of them the kernel must manage
 *   these, and therefore userspace must tell the kernel the object tiling if it
 *   wants to use fences for detiling.
 * - On gen3 and gen4 platforms have a swizzling pattern for tiled objects which
 *   depends upon the physical page frame number. When swapping such objects the
 *   page frame number might change and the kernel must be able to fix this up
 *   and hence now the tiling. Note that on a subset of platforms with
 *   asymmetric memory channel population the swizzling pattern changes in an
 *   unknown way, and for those the kernel simply forbids swapping completely.
54
 *
55 56 57 58
 * Since neither of this applies for new tiling layouts on modern platforms like
 * W, Ys and Yf tiling GEM only allows object tiling to be set to X or Y tiled.
 * Anything else can be handled in userspace entirely without the kernel's
 * invovlement.
59 60
 */

61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
/**
 * i915_gem_fence_size - required global GTT size for a fence
 * @i915: i915 device
 * @size: object size
 * @tiling: tiling mode
 * @stride: tiling stride
 *
 * Return the required global GTT size for a fence (view of a tiled object),
 * taking into account potential fence register mapping.
 */
u32 i915_gem_fence_size(struct drm_i915_private *i915,
			u32 size, unsigned int tiling, unsigned int stride)
{
	u32 ggtt_size;

	GEM_BUG_ON(!size);

	if (tiling == I915_TILING_NONE)
		return size;

	GEM_BUG_ON(!stride);

	if (INTEL_GEN(i915) >= 4) {
		stride *= i915_gem_tile_height(tiling);
85
		GEM_BUG_ON(!IS_ALIGNED(stride, I965_FENCE_PAGE));
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
		return roundup(size, stride);
	}

	/* Previous chips need a power-of-two fence region when tiling */
	if (IS_GEN3(i915))
		ggtt_size = 1024*1024;
	else
		ggtt_size = 512*1024;

	while (ggtt_size < size)
		ggtt_size <<= 1;

	return ggtt_size;
}

/**
 * i915_gem_fence_alignment - required global GTT alignment for a fence
 * @i915: i915 device
 * @size: object size
 * @tiling: tiling mode
 * @stride: tiling stride
 *
 * Return the required global GTT alignment for a fence (a view of a tiled
 * object), taking into account potential fence register mapping.
 */
u32 i915_gem_fence_alignment(struct drm_i915_private *i915, u32 size,
			     unsigned int tiling, unsigned int stride)
{
	GEM_BUG_ON(!size);

	/*
	 * Minimum alignment is 4k (GTT page size), but might be greater
	 * if a fence register is needed for the object.
	 */
120 121 122 123 124
	if (tiling == I915_TILING_NONE)
		return I915_GTT_MIN_ALIGNMENT;

	if (INTEL_GEN(i915) >= 4)
		return I965_FENCE_PAGE;
125 126 127 128 129 130 131 132

	/*
	 * Previous chips need to be aligned to the size of the smallest
	 * fence register that can contain the object.
	 */
	return i915_gem_fence_size(i915, size, tiling, stride);
}

133
/* Check pitch constriants for all chips & tiling formats */
134
static bool
135 136
i915_tiling_ok(struct drm_i915_gem_object *obj,
	       unsigned int tiling, unsigned int stride)
137
{
138 139
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	unsigned int tile_width;
140 141

	/* Linear is always fine */
142
	if (tiling == I915_TILING_NONE)
143 144
		return true;

145
	if (tiling > I915_TILING_LAST)
146 147
		return false;

148
	/* check maximum stride & object size */
149 150
	/* i965+ stores the end address of the gtt mapping in the fence
	 * reg, so dont bother to check the size */
151
	if (INTEL_GEN(i915) >= 7) {
152 153
		if (stride / 128 > GEN7_FENCE_MAX_PITCH_VAL)
			return false;
154
	} else if (INTEL_GEN(i915) >= 4) {
155 156
		if (stride / 128 > I965_FENCE_MAX_PITCH_VAL)
			return false;
157
	} else {
158
		if (stride > 8192)
159
			return false;
160

161 162
		if (!is_power_of_2(stride))
			return false;
163 164
	}

165 166 167 168 169 170
	if (IS_GEN2(i915) ||
	    (tiling == I915_TILING_Y && HAS_128_BYTE_Y_TILING(i915)))
		tile_width = 128;
	else
		tile_width = 512;

171
	if (!stride || !IS_ALIGNED(stride, tile_width))
172 173
		return false;

174
	return true;
175 176
}

177 178
static bool i915_vma_fence_prepare(struct i915_vma *vma,
				   int tiling_mode, unsigned int stride)
179
{
180 181
	struct drm_i915_private *i915 = vma->vm->i915;
	u32 size, alignment;
182 183 184 185

	if (!i915_vma_is_map_and_fenceable(vma))
		return true;

186
	size = i915_gem_fence_size(i915, vma->size, tiling_mode, stride);
187 188 189
	if (vma->node.size < size)
		return false;

190
	alignment = i915_gem_fence_alignment(i915, vma->size, tiling_mode, stride);
191
	if (!IS_ALIGNED(vma->node.start, alignment))
192 193 194 195 196
		return false;

	return true;
}

197 198
/* Make the current GTT allocation valid for the change in tiling. */
static int
199 200
i915_gem_object_fence_prepare(struct drm_i915_gem_object *obj,
			      int tiling_mode, unsigned int stride)
201
{
202
	struct i915_vma *vma;
203
	int ret;
204 205

	if (tiling_mode == I915_TILING_NONE)
206
		return 0;
207

208
	list_for_each_entry(vma, &obj->vma_list, obj_link) {
209 210 211
		if (!i915_vma_is_ggtt(vma))
			break;

212
		if (i915_vma_fence_prepare(vma, tiling_mode, stride))
213
			continue;
214

215 216 217
		ret = i915_vma_unbind(vma);
		if (ret)
			return ret;
218 219
	}

220
	return 0;
221 222
}

223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
int
i915_gem_object_set_tiling(struct drm_i915_gem_object *obj,
			   unsigned int tiling, unsigned int stride)
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	struct i915_vma *vma;
	int err;

	/* Make sure we don't cross-contaminate obj->tiling_and_stride */
	BUILD_BUG_ON(I915_TILING_LAST & STRIDE_MASK);

	GEM_BUG_ON(!i915_tiling_ok(obj, tiling, stride));
	GEM_BUG_ON(!stride ^ (tiling == I915_TILING_NONE));
	lockdep_assert_held(&i915->drm.struct_mutex);

	if ((tiling | stride) == obj->tiling_and_stride)
		return 0;

241
	if (i915_gem_object_is_framebuffer(obj))
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
		return -EBUSY;

	/* We need to rebind the object if its current allocation
	 * no longer meets the alignment restrictions for its new
	 * tiling mode. Otherwise we can just leave it alone, but
	 * need to ensure that any fence register is updated before
	 * the next fenced (either through the GTT or by the BLT unit
	 * on older GPUs) access.
	 *
	 * After updating the tiling parameters, we then flag whether
	 * we need to update an associated fence register. Note this
	 * has to also include the unfenced register the GPU uses
	 * whilst executing a fenced command for an untiled object.
	 */

	err = i915_gem_object_fence_prepare(obj, tiling, stride);
	if (err)
		return err;

261 262 263 264 265 266
	i915_gem_object_lock(obj);
	if (i915_gem_object_is_framebuffer(obj)) {
		i915_gem_object_unlock(obj);
		return -EBUSY;
	}

267 268 269 270 271
	/* If the memory has unknown (i.e. varying) swizzling, we pin the
	 * pages to prevent them being swapped out and causing corruption
	 * due to the change in swizzling.
	 */
	mutex_lock(&obj->mm.lock);
272
	if (i915_gem_object_has_pages(obj) &&
273 274 275 276 277 278 279 280
	    obj->mm.madv == I915_MADV_WILLNEED &&
	    i915->quirks & QUIRK_PIN_SWIZZLED_PAGES) {
		if (tiling == I915_TILING_NONE) {
			GEM_BUG_ON(!obj->mm.quirked);
			__i915_gem_object_unpin_pages(obj);
			obj->mm.quirked = false;
		}
		if (!i915_gem_object_is_tiled(obj)) {
281
			GEM_BUG_ON(obj->mm.quirked);
282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302
			__i915_gem_object_pin_pages(obj);
			obj->mm.quirked = true;
		}
	}
	mutex_unlock(&obj->mm.lock);

	list_for_each_entry(vma, &obj->vma_list, obj_link) {
		if (!i915_vma_is_ggtt(vma))
			break;

		vma->fence_size =
			i915_gem_fence_size(i915, vma->size, tiling, stride);
		vma->fence_alignment =
			i915_gem_fence_alignment(i915,
						 vma->size, tiling, stride);

		if (vma->fence)
			vma->fence->dirty = true;
	}

	obj->tiling_and_stride = tiling | stride;
303
	i915_gem_object_unlock(obj);
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321

	/* Force the fence to be reacquired for GTT access */
	i915_gem_release_mmap(obj);

	/* Try to preallocate memory required to save swizzling on put-pages */
	if (i915_gem_object_needs_bit17_swizzle(obj)) {
		if (!obj->bit_17) {
			obj->bit_17 = kcalloc(BITS_TO_LONGS(obj->base.size >> PAGE_SHIFT),
					      sizeof(long), GFP_KERNEL);
		}
	} else {
		kfree(obj->bit_17);
		obj->bit_17 = NULL;
	}

	return 0;
}

322
/**
323
 * i915_gem_set_tiling_ioctl - IOCTL handler to set tiling mode
324 325 326 327
 * @dev: DRM device
 * @data: data pointer for the ioctl
 * @file: DRM file for the ioctl call
 *
328 329
 * Sets the tiling mode of an object, returning the required swizzling of
 * bit 6 of addresses in the object.
330 331 332 333 334
 *
 * Called by the user via ioctl.
 *
 * Returns:
 * Zero on success, negative errno on failure.
335 336
 */
int
337 338
i915_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
			  struct drm_file *file)
339 340
{
	struct drm_i915_gem_set_tiling *args = data;
341
	struct drm_i915_gem_object *obj;
342
	int err;
343

344 345
	obj = i915_gem_object_lookup(file, args->handle);
	if (!obj)
346
		return -ENOENT;
347

T
Tina Zhang 已提交
348 349 350 351 352 353 354 355 356
	/*
	 * The tiling mode of proxy objects is handled by its generator, and
	 * not allowed to be changed by userspace.
	 */
	if (i915_gem_object_is_proxy(obj)) {
		err = -ENXIO;
		goto err;
	}

357 358
	if (!i915_tiling_ok(obj, args->tiling_mode, args->stride)) {
		err = -EINVAL;
359
		goto err;
360 361
	}

362 363
	if (args->tiling_mode == I915_TILING_NONE) {
		args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
364
		args->stride = 0;
365 366
	} else {
		if (args->tiling_mode == I915_TILING_X)
367
			args->swizzle_mode = to_i915(dev)->mm.bit_6_swizzle_x;
368
		else
369
			args->swizzle_mode = to_i915(dev)->mm.bit_6_swizzle_y;
370 371 372 373 374 375 376 377 378 379 380 381 382

		/* Hide bit 17 swizzling from the user.  This prevents old Mesa
		 * from aborting the application on sw fallbacks to bit 17,
		 * and we use the pread/pwrite bit17 paths to swizzle for it.
		 * If there was a user that was relying on the swizzle
		 * information for drm_intel_bo_map()ed reads/writes this would
		 * break it, but we don't have any of those.
		 */
		if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_17)
			args->swizzle_mode = I915_BIT_6_SWIZZLE_9;
		if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_10_17)
			args->swizzle_mode = I915_BIT_6_SWIZZLE_9_10;

383 384 385 386
		/* If we can't handle the swizzling, make it untiled. */
		if (args->swizzle_mode == I915_BIT_6_SWIZZLE_UNKNOWN) {
			args->tiling_mode = I915_TILING_NONE;
			args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
387
			args->stride = 0;
388 389
		}
	}
390

391 392 393
	err = mutex_lock_interruptible(&dev->struct_mutex);
	if (err)
		goto err;
394

395 396 397 398
	err = i915_gem_object_set_tiling(obj, args->tiling_mode, args->stride);
	mutex_unlock(&dev->struct_mutex);

	/* We have to maintain this existing ABI... */
399 400
	args->stride = i915_gem_object_get_stride(obj);
	args->tiling_mode = i915_gem_object_get_tiling(obj);
401

402
err:
403
	i915_gem_object_put(obj);
404
	return err;
405 406 407
}

/**
408
 * i915_gem_get_tiling_ioctl - IOCTL handler to get tiling mode
409 410 411 412
 * @dev: DRM device
 * @data: data pointer for the ioctl
 * @file: DRM file for the ioctl call
 *
413
 * Returns the current tiling mode and required bit 6 swizzling for the object.
414 415 416 417 418
 *
 * Called by the user via ioctl.
 *
 * Returns:
 * Zero on success, negative errno on failure.
419 420
 */
int
421 422
i915_gem_get_tiling_ioctl(struct drm_device *dev, void *data,
			  struct drm_file *file)
423 424
{
	struct drm_i915_gem_get_tiling *args = data;
425
	struct drm_i915_private *dev_priv = to_i915(dev);
426
	struct drm_i915_gem_object *obj;
427 428 429 430 431 432 433 434 435 436 437 438
	int err = -ENOENT;

	rcu_read_lock();
	obj = i915_gem_object_lookup_rcu(file, args->handle);
	if (obj) {
		args->tiling_mode =
			READ_ONCE(obj->tiling_and_stride) & TILING_MASK;
		err = 0;
	}
	rcu_read_unlock();
	if (unlikely(err))
		return err;
439

440
	switch (args->tiling_mode) {
441 442 443 444 445 446
	case I915_TILING_X:
		args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x;
		break;
	case I915_TILING_Y:
		args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y;
		break;
447
	default:
448 449 450 451 452
	case I915_TILING_NONE:
		args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
		break;
	}

453
	/* Hide bit 17 from the user -- see comment in i915_gem_set_tiling */
454 455 456 457
	if (dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES)
		args->phys_swizzle_mode = I915_BIT_6_SWIZZLE_UNKNOWN;
	else
		args->phys_swizzle_mode = args->swizzle_mode;
458 459 460 461 462
	if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_17)
		args->swizzle_mode = I915_BIT_6_SWIZZLE_9;
	if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_10_17)
		args->swizzle_mode = I915_BIT_6_SWIZZLE_9_10;

463 464
	return 0;
}