drivers/gpu/drm/i915/i915_gem_tiling.c

   1 /*
   2  * Copyright © 2008 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21  * IN THE SOFTWARE.
  22  *
  23  * Authors:
  24  *    Eric Anholt <eric@anholt.net>
  25  *
  26  */
  27
  28 #include <linux/string.h>
  29 #include <linux/bitops.h>
  30 #include <drm/drmP.h>
  31 #include <drm/i915_drm.h>
  32 #include "i915_drv.h"
  33
  34 /**
  35  * DOC: buffer object tiling
  36  *
  37  * i915_gem_set_tiling() and i915_gem_get_tiling() is the userspace interface to
  38  * declare fence register requirements.
  39  *
  40  * In principle GEM doesn't care at all about the internal data layout of an
  41  * object, and hence it also doesn't care about tiling or swizzling. There's two
  42  * exceptions:
  43  *
  44  * - For X and Y tiling the hardware provides detilers for CPU access, so called
  45  *   fences. Since there's only a limited amount of them the kernel must manage
  46  *   these, and therefore userspace must tell the kernel the object tiling if it
  47  *   wants to use fences for detiling.
  48  * - On gen3 and gen4 platforms have a swizzling pattern for tiled objects which
  49  *   depends upon the physical page frame number. When swapping such objects the
  50  *   page frame number might change and the kernel must be able to fix this up
  51  *   and hence now the tiling. Note that on a subset of platforms with
  52  *   asymmetric memory channel population the swizzling pattern changes in an
  53  *   unknown way, and for those the kernel simply forbids swapping completely.
  54  *
  55  * Since neither of this applies for new tiling layouts on modern platforms like
  56  * W, Ys and Yf tiling GEM only allows object tiling to be set to X or Y tiled.
  57  * Anything else can be handled in userspace entirely without the kernel's
  58  * invovlement.
  59  */
  60
  61 /* Check pitch constriants for all chips & tiling formats */
  62 static bool
  63 i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode)
  64 {
  65         int tile_width;
  66
  67         /* Linear is always fine */
  68         if (tiling_mode == I915_TILING_NONE)
  69                 return true;
  70
  71         if (IS_GEN2(dev) ||
  72             (tiling_mode == I915_TILING_Y && HAS_128_BYTE_Y_TILING(dev)))
  73                 tile_width = 128;
  74         else
  75                 tile_width = 512;
  76
  77         /* check maximum stride & object size */
  78         /* i965+ stores the end address of the gtt mapping in the fence
  79          * reg, so dont bother to check the size */
  80         if (INTEL_INFO(dev)->gen >= 7) {
  81                 if (stride / 128 > GEN7_FENCE_MAX_PITCH_VAL)
  82                         return false;
  83         } else if (INTEL_INFO(dev)->gen >= 4) {
  84                 if (stride / 128 > I965_FENCE_MAX_PITCH_VAL)
  85                         return false;
  86         } else {
  87                 if (stride > 8192)
  88                         return false;
  89
  90                 if (IS_GEN3(dev)) {
  91                         if (size > I830_FENCE_MAX_SIZE_VAL << 20)
  92                                 return false;
  93                 } else {
  94                         if (size > I830_FENCE_MAX_SIZE_VAL << 19)
  95                                 return false;
  96                 }
  97         }
  98
  99         if (stride < tile_width)
 100                 return false;
 101
 102         /* 965+ just needs multiples of tile width */
 103         if (INTEL_INFO(dev)->gen >= 4) {
 104                 if (stride & (tile_width - 1))
 105                         return false;
 106                 return true;
 107         }
 108
 109         /* Pre-965 needs power of two tile widths */
 110         if (stride & (stride - 1))
 111                 return false;
 112
 113         return true;
 114 }
 115
 116 /* Is the current GTT allocation valid for the change in tiling? */
 117 static bool
 118 i915_gem_object_fence_ok(struct drm_i915_gem_object *obj, int tiling_mode)
 119 {
 120         u32 size;
 121
 122         if (tiling_mode == I915_TILING_NONE)
 123                 return true;
 124
 125         if (INTEL_INFO(obj->base.dev)->gen >= 4)
 126                 return true;
 127
 128         if (IS_GEN3(obj->base.dev)) {
 129                 if (i915_gem_obj_ggtt_offset(obj) & ~I915_FENCE_START_MASK)
 130                         return false;
 131         } else {
 132                 if (i915_gem_obj_ggtt_offset(obj) & ~I830_FENCE_START_MASK)
 133                         return false;
 134         }
 135
 136         size = i915_gem_get_gtt_size(obj->base.dev, obj->base.size, tiling_mode);
 137         if (i915_gem_obj_ggtt_size(obj) != size)
 138                 return false;
 139
 140         if (i915_gem_obj_ggtt_offset(obj) & (size - 1))
 141                 return false;
 142
 143         return true;
 144 }
 145
 146 /**
 147  * i915_gem_set_tiling - IOCTL handler to set tiling mode
 148  * @dev: DRM device
 149  * @data: data pointer for the ioctl
 150  * @file: DRM file for the ioctl call
 151  *
 152  * Sets the tiling mode of an object, returning the required swizzling of
 153  * bit 6 of addresses in the object.
 154  *
 155  * Called by the user via ioctl.
 156  *
 157  * Returns:
 158  * Zero on success, negative errno on failure.
 159  */
 160 int
 161 i915_gem_set_tiling(struct drm_device *dev, void *data,
 162                    struct drm_file *file)
 163 {
 164         struct drm_i915_gem_set_tiling *args = data;
 165         struct drm_i915_private *dev_priv = to_i915(dev);
 166         struct drm_i915_gem_object *obj;
 167         int ret = 0;
 168
 169         obj = to_intel_bo(drm_gem_object_lookup(file, args->handle));
 170         if (&obj->base == NULL)
 171                 return -ENOENT;
 172
 173         if (!i915_tiling_ok(dev,
 174                             args->stride, obj->base.size, args->tiling_mode)) {
 175                 drm_gem_object_unreference_unlocked(&obj->base);
 176                 return -EINVAL;
 177         }
 178
 179         intel_runtime_pm_get(dev_priv);
 180
 181         mutex_lock(&dev->struct_mutex);
 182         if (obj->pin_display || obj->framebuffer_references) {
 183                 ret = -EBUSY;
 184                 goto err;
 185         }
 186
 187         if (args->tiling_mode == I915_TILING_NONE) {
 188                 args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
 189                 args->stride = 0;
 190         } else {
 191                 if (args->tiling_mode == I915_TILING_X)
 192                         args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x;
 193                 else
 194                         args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y;
 195
 196                 /* Hide bit 17 swizzling from the user.  This prevents old Mesa
 197                  * from aborting the application on sw fallbacks to bit 17,
 198                  * and we use the pread/pwrite bit17 paths to swizzle for it.
 199                  * If there was a user that was relying on the swizzle
 200                  * information for drm_intel_bo_map()ed reads/writes this would
 201                  * break it, but we don't have any of those.
 202                  */
 203                 if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_17)
 204                         args->swizzle_mode = I915_BIT_6_SWIZZLE_9;
 205                 if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_10_17)
 206                         args->swizzle_mode = I915_BIT_6_SWIZZLE_9_10;
 207
 208                 /* If we can't handle the swizzling, make it untiled. */
 209                 if (args->swizzle_mode == I915_BIT_6_SWIZZLE_UNKNOWN) {
 210                         args->tiling_mode = I915_TILING_NONE;
 211                         args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
 212                         args->stride = 0;
 213                 }
 214         }
 215
 216         if (args->tiling_mode != obj->tiling_mode ||
 217             args->stride != obj->stride) {
 218                 /* We need to rebind the object if its current allocation
 219                  * no longer meets the alignment restrictions for its new
 220                  * tiling mode. Otherwise we can just leave it alone, but
 221                  * need to ensure that any fence register is updated before
 222                  * the next fenced (either through the GTT or by the BLT unit
 223                  * on older GPUs) access.
 224                  *
 225                  * After updating the tiling parameters, we then flag whether
 226                  * we need to update an associated fence register. Note this
 227                  * has to also include the unfenced register the GPU uses
 228                  * whilst executing a fenced command for an untiled object.
 229                  */
 230                 if (obj->map_and_fenceable &&
 231                     !i915_gem_object_fence_ok(obj, args->tiling_mode))
 232                         ret = i915_vma_unbind(i915_gem_obj_to_ggtt(obj));
 233
 234                 if (ret == 0) {
 235                         if (obj->pages &&
 236                             obj->madv == I915_MADV_WILLNEED &&
 237                             dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES) {
 238                                 if (args->tiling_mode == I915_TILING_NONE)
 239                                         i915_gem_object_unpin_pages(obj);
 240                                 if (obj->tiling_mode == I915_TILING_NONE)
 241                                         i915_gem_object_pin_pages(obj);
 242                         }
 243
 244                         obj->fence_dirty =
 245                                 obj->last_fenced_req ||
 246                                 obj->fence_reg != I915_FENCE_REG_NONE;
 247
 248                         obj->tiling_mode = args->tiling_mode;
 249                         obj->stride = args->stride;
 250
 251                         /* Force the fence to be reacquired for GTT access */
 252                         i915_gem_release_mmap(obj);
 253                 }
 254         }
 255         /* we have to maintain this existing ABI... */
 256         args->stride = obj->stride;
 257         args->tiling_mode = obj->tiling_mode;
 258
 259         /* Try to preallocate memory required to save swizzling on put-pages */
 260         if (i915_gem_object_needs_bit17_swizzle(obj)) {
 261                 if (obj->bit_17 == NULL) {
 262                         obj->bit_17 = kcalloc(BITS_TO_LONGS(obj->base.size >> PAGE_SHIFT),
 263                                               sizeof(long), GFP_KERNEL);
 264                 }
 265         } else {
 266                 kfree(obj->bit_17);
 267                 obj->bit_17 = NULL;
 268         }
 269
 270 err:
 271         drm_gem_object_unreference(&obj->base);
 272         mutex_unlock(&dev->struct_mutex);
 273
 274         intel_runtime_pm_put(dev_priv);
 275
 276         return ret;
 277 }
 278
 279 /**
 280  * i915_gem_get_tiling - IOCTL handler to get tiling mode
 281  * @dev: DRM device
 282  * @data: data pointer for the ioctl
 283  * @file: DRM file for the ioctl call
 284  *
 285  * Returns the current tiling mode and required bit 6 swizzling for the object.
 286  *
 287  * Called by the user via ioctl.
 288  *
 289  * Returns:
 290  * Zero on success, negative errno on failure.
 291  */
 292 int
 293 i915_gem_get_tiling(struct drm_device *dev, void *data,
 294                    struct drm_file *file)
 295 {
 296         struct drm_i915_gem_get_tiling *args = data;
 297         struct drm_i915_private *dev_priv = to_i915(dev);
 298         struct drm_i915_gem_object *obj;
 299
 300         obj = to_intel_bo(drm_gem_object_lookup(file, args->handle));
 301         if (&obj->base == NULL)
 302                 return -ENOENT;
 303
 304         mutex_lock(&dev->struct_mutex);
 305
 306         args->tiling_mode = obj->tiling_mode;
 307         switch (obj->tiling_mode) {
 308         case I915_TILING_X:
 309                 args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x;
 310                 break;
 311         case I915_TILING_Y:
 312                 args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y;
 313                 break;
 314         case I915_TILING_NONE:
 315                 args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
 316                 break;
 317         default:
 318                 DRM_ERROR("unknown tiling mode\n");
 319         }
 320
 321         /* Hide bit 17 from the user -- see comment in i915_gem_set_tiling */
 322         if (dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES)
 323                 args->phys_swizzle_mode = I915_BIT_6_SWIZZLE_UNKNOWN;
 324         else
 325                 args->phys_swizzle_mode = args->swizzle_mode;
 326         if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_17)
 327                 args->swizzle_mode = I915_BIT_6_SWIZZLE_9;
 328         if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_10_17)
 329                 args->swizzle_mode = I915_BIT_6_SWIZZLE_9_10;
 330
 331         drm_gem_object_unreference(&obj->base);
 332         mutex_unlock(&dev->struct_mutex);
 333
 334         return 0;
 335 }