Rev 4392 | Rev 4560 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
3263 | Serge | 1 | /* |
2 | * Copyright © 2008,2010 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 |
||
25 | * Chris Wilson |
||
26 | * |
||
27 | */ |
||
28 | |||
29 | #include |
||
30 | #include |
||
31 | #include "i915_drv.h" |
||
32 | #include "i915_trace.h" |
||
33 | #include "intel_drv.h" |
||
34 | //#include |
||
35 | |||
36 | |||
37 | static unsigned long |
||
38 | copy_to_user(void __user *to, const void *from, unsigned long n) |
||
39 | { |
||
40 | memcpy(to, from, n); |
||
41 | return 0; |
||
42 | } |
||
43 | |||
44 | static unsigned long |
||
45 | copy_from_user(void *to, const void __user *from, unsigned long n) |
||
46 | { |
||
47 | memcpy(to, from, n); |
||
48 | return 0; |
||
49 | } |
||
50 | |||
51 | struct eb_objects { |
||
3480 | Serge | 52 | struct list_head objects; |
3263 | Serge | 53 | int and; |
3480 | Serge | 54 | union { |
55 | struct drm_i915_gem_object *lut[0]; |
||
3263 | Serge | 56 | struct hlist_head buckets[0]; |
3480 | Serge | 57 | }; |
3263 | Serge | 58 | }; |
59 | |||
60 | static struct eb_objects * |
||
3480 | Serge | 61 | eb_create(struct drm_i915_gem_execbuffer2 *args) |
3263 | Serge | 62 | { |
3480 | Serge | 63 | struct eb_objects *eb = NULL; |
64 | |||
65 | if (args->flags & I915_EXEC_HANDLE_LUT) { |
||
66 | int size = args->buffer_count; |
||
67 | size *= sizeof(struct drm_i915_gem_object *); |
||
68 | size += sizeof(struct eb_objects); |
||
69 | eb = kmalloc(size, GFP_TEMPORARY | __GFP_NOWARN | __GFP_NORETRY); |
||
70 | } |
||
71 | |||
72 | if (eb == NULL) { |
||
73 | int size = args->buffer_count; |
||
3263 | Serge | 74 | int count = PAGE_SIZE / sizeof(struct hlist_head) / 2; |
3480 | Serge | 75 | BUILD_BUG_ON_NOT_POWER_OF_2(PAGE_SIZE / sizeof(struct hlist_head)); |
76 | while (count > 2*size) |
||
3263 | Serge | 77 | count >>= 1; |
78 | eb = kzalloc(count*sizeof(struct hlist_head) + |
||
79 | sizeof(struct eb_objects), |
||
3480 | Serge | 80 | GFP_TEMPORARY); |
3263 | Serge | 81 | if (eb == NULL) |
82 | return eb; |
||
83 | |||
84 | eb->and = count - 1; |
||
3480 | Serge | 85 | } else |
86 | eb->and = -args->buffer_count; |
||
87 | |||
88 | INIT_LIST_HEAD(&eb->objects); |
||
3263 | Serge | 89 | return eb; |
90 | } |
||
91 | |||
92 | static void |
||
93 | eb_reset(struct eb_objects *eb) |
||
94 | { |
||
3480 | Serge | 95 | if (eb->and >= 0) |
3263 | Serge | 96 | memset(eb->buckets, 0, (eb->and+1)*sizeof(struct hlist_head)); |
97 | } |
||
98 | |||
3480 | Serge | 99 | static int |
100 | eb_lookup_objects(struct eb_objects *eb, |
||
101 | struct drm_i915_gem_exec_object2 *exec, |
||
102 | const struct drm_i915_gem_execbuffer2 *args, |
||
103 | struct drm_file *file) |
||
3263 | Serge | 104 | { |
3480 | Serge | 105 | int i; |
106 | |||
107 | spin_lock(&file->table_lock); |
||
108 | for (i = 0; i < args->buffer_count; i++) { |
||
109 | struct drm_i915_gem_object *obj; |
||
110 | |||
111 | obj = to_intel_bo(idr_find(&file->object_idr, exec[i].handle)); |
||
112 | if (obj == NULL) { |
||
113 | spin_unlock(&file->table_lock); |
||
114 | DRM_DEBUG("Invalid object handle %d at index %d\n", |
||
115 | exec[i].handle, i); |
||
116 | return -ENOENT; |
||
117 | } |
||
118 | |||
119 | if (!list_empty(&obj->exec_list)) { |
||
120 | spin_unlock(&file->table_lock); |
||
121 | DRM_DEBUG("Object %p [handle %d, index %d] appears more than once in object list\n", |
||
122 | obj, exec[i].handle, i); |
||
123 | return -EINVAL; |
||
124 | } |
||
125 | |||
126 | drm_gem_object_reference(&obj->base); |
||
127 | list_add_tail(&obj->exec_list, &eb->objects); |
||
128 | |||
129 | obj->exec_entry = &exec[i]; |
||
130 | if (eb->and < 0) { |
||
131 | eb->lut[i] = obj; |
||
132 | } else { |
||
133 | uint32_t handle = args->flags & I915_EXEC_HANDLE_LUT ? i : exec[i].handle; |
||
134 | obj->exec_handle = handle; |
||
3263 | Serge | 135 | hlist_add_head(&obj->exec_node, |
3480 | Serge | 136 | &eb->buckets[handle & eb->and]); |
137 | } |
||
138 | } |
||
139 | spin_unlock(&file->table_lock); |
||
140 | |||
141 | return 0; |
||
3263 | Serge | 142 | } |
143 | |||
144 | static struct drm_i915_gem_object * |
||
145 | eb_get_object(struct eb_objects *eb, unsigned long handle) |
||
146 | { |
||
3480 | Serge | 147 | if (eb->and < 0) { |
148 | if (handle >= -eb->and) |
||
149 | return NULL; |
||
150 | return eb->lut[handle]; |
||
151 | } else { |
||
3263 | Serge | 152 | struct hlist_head *head; |
153 | struct hlist_node *node; |
||
154 | |||
155 | head = &eb->buckets[handle & eb->and]; |
||
156 | hlist_for_each(node, head) { |
||
3480 | Serge | 157 | struct drm_i915_gem_object *obj; |
158 | |||
3263 | Serge | 159 | obj = hlist_entry(node, struct drm_i915_gem_object, exec_node); |
160 | if (obj->exec_handle == handle) |
||
161 | return obj; |
||
162 | } |
||
163 | return NULL; |
||
3480 | Serge | 164 | } |
3263 | Serge | 165 | } |
166 | |||
167 | static void |
||
168 | eb_destroy(struct eb_objects *eb) |
||
169 | { |
||
3480 | Serge | 170 | while (!list_empty(&eb->objects)) { |
171 | struct drm_i915_gem_object *obj; |
||
172 | |||
173 | obj = list_first_entry(&eb->objects, |
||
174 | struct drm_i915_gem_object, |
||
175 | exec_list); |
||
176 | list_del_init(&obj->exec_list); |
||
177 | drm_gem_object_unreference(&obj->base); |
||
178 | } |
||
3263 | Serge | 179 | kfree(eb); |
180 | } |
||
181 | |||
182 | static inline int use_cpu_reloc(struct drm_i915_gem_object *obj) |
||
183 | { |
||
184 | return (obj->base.write_domain == I915_GEM_DOMAIN_CPU || |
||
185 | !obj->map_and_fenceable || |
||
186 | obj->cache_level != I915_CACHE_NONE); |
||
187 | } |
||
188 | |||
189 | static int |
||
4371 | Serge | 190 | relocate_entry_cpu(struct drm_i915_gem_object *obj, |
191 | struct drm_i915_gem_relocation_entry *reloc) |
||
192 | { |
||
4539 | Serge | 193 | struct drm_device *dev = obj->base.dev; |
194 | struct drm_i915_private *dev_priv = dev->dev_private; |
||
4371 | Serge | 195 | uint32_t page_offset = offset_in_page(reloc->offset); |
196 | char *vaddr; |
||
197 | int ret = -EINVAL; |
||
198 | |||
199 | ret = i915_gem_object_set_to_cpu_domain(obj, 1); |
||
200 | if (ret) |
||
201 | return ret; |
||
202 | |||
4539 | Serge | 203 | vaddr = dev_priv->gtt.mappable+4096; |
204 | MapPage(vaddr,(addr_t)i915_gem_object_get_page(obj,reloc->offset >> PAGE_SHIFT), PG_SW); |
||
4371 | Serge | 205 | *(uint32_t *)(vaddr + page_offset) = reloc->delta; |
206 | |||
207 | return 0; |
||
208 | } |
||
209 | |||
210 | static int |
||
211 | relocate_entry_gtt(struct drm_i915_gem_object *obj, |
||
212 | struct drm_i915_gem_relocation_entry *reloc) |
||
213 | { |
||
214 | struct drm_device *dev = obj->base.dev; |
||
215 | struct drm_i915_private *dev_priv = dev->dev_private; |
||
216 | uint32_t __iomem *reloc_entry; |
||
217 | void __iomem *reloc_page; |
||
218 | int ret = -EINVAL; |
||
219 | |||
220 | ret = i915_gem_object_set_to_gtt_domain(obj, true); |
||
221 | if (ret) |
||
222 | return ret; |
||
223 | |||
224 | ret = i915_gem_object_put_fence(obj); |
||
225 | if (ret) |
||
226 | return ret; |
||
227 | |||
228 | /* Map the page containing the relocation we're going to perform. */ |
||
229 | reloc->offset += i915_gem_obj_ggtt_offset(obj); |
||
4539 | Serge | 230 | MapPage(dev_priv->gtt.mappable,dev_priv->gtt.mappable_base + |
231 | (reloc->offset & PAGE_MASK), PG_SW); |
||
232 | reloc_page = dev_priv->gtt.mappable; |
||
4371 | Serge | 233 | reloc_entry = (uint32_t __iomem *) |
234 | (reloc_page + offset_in_page(reloc->offset)); |
||
235 | iowrite32(reloc->delta, reloc_entry); |
||
236 | |||
237 | return 0; |
||
238 | } |
||
239 | |||
240 | static int |
||
3263 | Serge | 241 | i915_gem_execbuffer_relocate_entry(struct drm_i915_gem_object *obj, |
242 | struct eb_objects *eb, |
||
4104 | Serge | 243 | struct drm_i915_gem_relocation_entry *reloc, |
244 | struct i915_address_space *vm) |
||
3263 | Serge | 245 | { |
246 | struct drm_device *dev = obj->base.dev; |
||
247 | struct drm_gem_object *target_obj; |
||
248 | struct drm_i915_gem_object *target_i915_obj; |
||
249 | uint32_t target_offset; |
||
250 | int ret = -EINVAL; |
||
251 | |||
252 | /* we've already hold a reference to all valid objects */ |
||
253 | target_obj = &eb_get_object(eb, reloc->target_handle)->base; |
||
254 | if (unlikely(target_obj == NULL)) |
||
255 | return -ENOENT; |
||
256 | |||
257 | target_i915_obj = to_intel_bo(target_obj); |
||
4104 | Serge | 258 | target_offset = i915_gem_obj_ggtt_offset(target_i915_obj); |
3263 | Serge | 259 | |
260 | /* Sandybridge PPGTT errata: We need a global gtt mapping for MI and |
||
261 | * pipe_control writes because the gpu doesn't properly redirect them |
||
262 | * through the ppgtt for non_secure batchbuffers. */ |
||
263 | if (unlikely(IS_GEN6(dev) && |
||
264 | reloc->write_domain == I915_GEM_DOMAIN_INSTRUCTION && |
||
265 | !target_i915_obj->has_global_gtt_mapping)) { |
||
266 | i915_gem_gtt_bind_object(target_i915_obj, |
||
267 | target_i915_obj->cache_level); |
||
268 | } |
||
269 | |||
270 | /* Validate that the target is in a valid r/w GPU domain */ |
||
271 | if (unlikely(reloc->write_domain & (reloc->write_domain - 1))) { |
||
272 | DRM_DEBUG("reloc with multiple write domains: " |
||
273 | "obj %p target %d offset %d " |
||
274 | "read %08x write %08x", |
||
275 | obj, reloc->target_handle, |
||
276 | (int) reloc->offset, |
||
277 | reloc->read_domains, |
||
278 | reloc->write_domain); |
||
279 | return ret; |
||
280 | } |
||
281 | if (unlikely((reloc->write_domain | reloc->read_domains) |
||
282 | & ~I915_GEM_GPU_DOMAINS)) { |
||
283 | DRM_DEBUG("reloc with read/write non-GPU domains: " |
||
284 | "obj %p target %d offset %d " |
||
285 | "read %08x write %08x", |
||
286 | obj, reloc->target_handle, |
||
287 | (int) reloc->offset, |
||
288 | reloc->read_domains, |
||
289 | reloc->write_domain); |
||
290 | return ret; |
||
291 | } |
||
292 | |||
293 | target_obj->pending_read_domains |= reloc->read_domains; |
||
294 | target_obj->pending_write_domain |= reloc->write_domain; |
||
295 | |||
296 | /* If the relocation already has the right value in it, no |
||
297 | * more work needs to be done. |
||
298 | */ |
||
299 | if (target_offset == reloc->presumed_offset) |
||
300 | return 0; |
||
301 | |||
302 | /* Check that the relocation address is valid... */ |
||
303 | if (unlikely(reloc->offset > obj->base.size - 4)) { |
||
304 | DRM_DEBUG("Relocation beyond object bounds: " |
||
305 | "obj %p target %d offset %d size %d.\n", |
||
306 | obj, reloc->target_handle, |
||
307 | (int) reloc->offset, |
||
308 | (int) obj->base.size); |
||
309 | return ret; |
||
310 | } |
||
311 | if (unlikely(reloc->offset & 3)) { |
||
312 | DRM_DEBUG("Relocation not 4-byte aligned: " |
||
313 | "obj %p target %d offset %d.\n", |
||
314 | obj, reloc->target_handle, |
||
315 | (int) reloc->offset); |
||
316 | return ret; |
||
317 | } |
||
318 | |||
319 | /* We can't wait for rendering with pagefaults disabled */ |
||
320 | |||
321 | reloc->delta += target_offset; |
||
4371 | Serge | 322 | if (use_cpu_reloc(obj)) |
323 | ret = relocate_entry_cpu(obj, reloc); |
||
324 | else |
||
325 | ret = relocate_entry_gtt(obj, reloc); |
||
3263 | Serge | 326 | |
327 | if (ret) |
||
328 | return ret; |
||
329 | |||
330 | /* and update the user's relocation entry */ |
||
331 | reloc->presumed_offset = target_offset; |
||
332 | |||
333 | return 0; |
||
334 | } |
||
335 | |||
336 | static int |
||
337 | i915_gem_execbuffer_relocate_object(struct drm_i915_gem_object *obj, |
||
4104 | Serge | 338 | struct eb_objects *eb, |
339 | struct i915_address_space *vm) |
||
3263 | Serge | 340 | { |
341 | #define N_RELOC(x) ((x) / sizeof(struct drm_i915_gem_relocation_entry)) |
||
3266 | Serge | 342 | struct drm_i915_gem_relocation_entry stack_reloc[N_RELOC(64)]; |
3263 | Serge | 343 | struct drm_i915_gem_relocation_entry __user *user_relocs; |
344 | struct drm_i915_gem_exec_object2 *entry = obj->exec_entry; |
||
345 | int remain, ret; |
||
346 | |||
4539 | Serge | 347 | user_relocs = to_user_ptr(entry->relocs_ptr); |
3263 | Serge | 348 | |
349 | remain = entry->relocation_count; |
||
350 | while (remain) { |
||
351 | struct drm_i915_gem_relocation_entry *r = stack_reloc; |
||
352 | int count = remain; |
||
353 | if (count > ARRAY_SIZE(stack_reloc)) |
||
354 | count = ARRAY_SIZE(stack_reloc); |
||
355 | remain -= count; |
||
356 | |||
357 | memcpy(r, user_relocs, count*sizeof(r[0])); |
||
358 | |||
359 | do { |
||
360 | u64 offset = r->presumed_offset; |
||
361 | |||
4104 | Serge | 362 | ret = i915_gem_execbuffer_relocate_entry(obj, eb, r, |
363 | vm); |
||
3263 | Serge | 364 | if (ret) |
365 | return ret; |
||
366 | |||
4392 | Serge | 367 | if (r->presumed_offset != offset) |
368 | { |
||
3263 | Serge | 369 | memcpy(&user_relocs->presumed_offset, |
370 | &r->presumed_offset, |
||
371 | sizeof(r->presumed_offset)); |
||
4392 | Serge | 372 | } |
3263 | Serge | 373 | |
374 | user_relocs++; |
||
375 | r++; |
||
376 | } while (--count); |
||
377 | } |
||
378 | |||
379 | return 0; |
||
380 | #undef N_RELOC |
||
381 | } |
||
382 | |||
383 | static int |
||
384 | i915_gem_execbuffer_relocate_object_slow(struct drm_i915_gem_object *obj, |
||
385 | struct eb_objects *eb, |
||
4104 | Serge | 386 | struct drm_i915_gem_relocation_entry *relocs, |
387 | struct i915_address_space *vm) |
||
3263 | Serge | 388 | { |
389 | const struct drm_i915_gem_exec_object2 *entry = obj->exec_entry; |
||
390 | int i, ret; |
||
391 | |||
392 | for (i = 0; i < entry->relocation_count; i++) { |
||
4104 | Serge | 393 | ret = i915_gem_execbuffer_relocate_entry(obj, eb, &relocs[i], |
394 | vm); |
||
3263 | Serge | 395 | if (ret) |
396 | return ret; |
||
397 | } |
||
398 | |||
399 | return 0; |
||
400 | } |
||
401 | |||
402 | static int |
||
4104 | Serge | 403 | i915_gem_execbuffer_relocate(struct eb_objects *eb, |
404 | struct i915_address_space *vm) |
||
3263 | Serge | 405 | { |
406 | struct drm_i915_gem_object *obj; |
||
407 | int ret = 0; |
||
408 | |||
409 | /* This is the fast path and we cannot handle a pagefault whilst |
||
410 | * holding the struct mutex lest the user pass in the relocations |
||
411 | * contained within a mmaped bo. For in such a case we, the page |
||
412 | * fault handler would call i915_gem_fault() and we would try to |
||
413 | * acquire the struct mutex again. Obviously this is bad and so |
||
414 | * lockdep complains vehemently. |
||
415 | */ |
||
4104 | Serge | 416 | // pagefault_disable(); |
3480 | Serge | 417 | list_for_each_entry(obj, &eb->objects, exec_list) { |
4104 | Serge | 418 | ret = i915_gem_execbuffer_relocate_object(obj, eb, vm); |
3263 | Serge | 419 | if (ret) |
420 | break; |
||
421 | } |
||
422 | // pagefault_enable(); |
||
423 | |||
424 | return ret; |
||
425 | } |
||
426 | |||
427 | #define __EXEC_OBJECT_HAS_PIN (1<<31) |
||
428 | #define __EXEC_OBJECT_HAS_FENCE (1<<30) |
||
429 | |||
430 | static int |
||
431 | need_reloc_mappable(struct drm_i915_gem_object *obj) |
||
432 | { |
||
433 | struct drm_i915_gem_exec_object2 *entry = obj->exec_entry; |
||
434 | return entry->relocation_count && !use_cpu_reloc(obj); |
||
435 | } |
||
436 | |||
437 | static int |
||
438 | i915_gem_execbuffer_reserve_object(struct drm_i915_gem_object *obj, |
||
3480 | Serge | 439 | struct intel_ring_buffer *ring, |
4104 | Serge | 440 | struct i915_address_space *vm, |
3480 | Serge | 441 | bool *need_reloc) |
3263 | Serge | 442 | { |
443 | struct drm_i915_private *dev_priv = obj->base.dev->dev_private; |
||
444 | struct drm_i915_gem_exec_object2 *entry = obj->exec_entry; |
||
445 | bool has_fenced_gpu_access = INTEL_INFO(ring->dev)->gen < 4; |
||
446 | bool need_fence, need_mappable; |
||
447 | int ret; |
||
448 | |||
449 | need_fence = |
||
450 | has_fenced_gpu_access && |
||
451 | entry->flags & EXEC_OBJECT_NEEDS_FENCE && |
||
452 | obj->tiling_mode != I915_TILING_NONE; |
||
453 | need_mappable = need_fence || need_reloc_mappable(obj); |
||
454 | |||
4104 | Serge | 455 | ret = i915_gem_object_pin(obj, vm, entry->alignment, need_mappable, |
456 | false); |
||
3263 | Serge | 457 | if (ret) |
458 | return ret; |
||
459 | |||
460 | entry->flags |= __EXEC_OBJECT_HAS_PIN; |
||
461 | |||
462 | if (has_fenced_gpu_access) { |
||
463 | if (entry->flags & EXEC_OBJECT_NEEDS_FENCE) { |
||
464 | ret = i915_gem_object_get_fence(obj); |
||
465 | if (ret) |
||
466 | return ret; |
||
467 | |||
468 | if (i915_gem_object_pin_fence(obj)) |
||
469 | entry->flags |= __EXEC_OBJECT_HAS_FENCE; |
||
470 | |||
471 | obj->pending_fenced_gpu_access = true; |
||
472 | } |
||
473 | } |
||
474 | |||
475 | /* Ensure ppgtt mapping exists if needed */ |
||
476 | if (dev_priv->mm.aliasing_ppgtt && !obj->has_aliasing_ppgtt_mapping) { |
||
477 | i915_ppgtt_bind_object(dev_priv->mm.aliasing_ppgtt, |
||
478 | obj, obj->cache_level); |
||
479 | |||
480 | obj->has_aliasing_ppgtt_mapping = 1; |
||
481 | } |
||
482 | |||
4104 | Serge | 483 | if (entry->offset != i915_gem_obj_offset(obj, vm)) { |
484 | entry->offset = i915_gem_obj_offset(obj, vm); |
||
3480 | Serge | 485 | *need_reloc = true; |
486 | } |
||
3266 | Serge | 487 | |
3480 | Serge | 488 | if (entry->flags & EXEC_OBJECT_WRITE) { |
489 | obj->base.pending_read_domains = I915_GEM_DOMAIN_RENDER; |
||
490 | obj->base.pending_write_domain = I915_GEM_DOMAIN_RENDER; |
||
491 | } |
||
492 | |||
493 | if (entry->flags & EXEC_OBJECT_NEEDS_GTT && |
||
494 | !obj->has_global_gtt_mapping) |
||
495 | i915_gem_gtt_bind_object(obj, obj->cache_level); |
||
496 | |||
3263 | Serge | 497 | return 0; |
498 | } |
||
499 | |||
500 | static void |
||
501 | i915_gem_execbuffer_unreserve_object(struct drm_i915_gem_object *obj) |
||
502 | { |
||
503 | struct drm_i915_gem_exec_object2 *entry; |
||
504 | |||
4104 | Serge | 505 | if (!i915_gem_obj_bound_any(obj)) |
3263 | Serge | 506 | return; |
507 | |||
508 | entry = obj->exec_entry; |
||
509 | |||
510 | if (entry->flags & __EXEC_OBJECT_HAS_FENCE) |
||
511 | i915_gem_object_unpin_fence(obj); |
||
512 | |||
513 | if (entry->flags & __EXEC_OBJECT_HAS_PIN) |
||
514 | i915_gem_object_unpin(obj); |
||
515 | |||
516 | entry->flags &= ~(__EXEC_OBJECT_HAS_FENCE | __EXEC_OBJECT_HAS_PIN); |
||
517 | } |
||
518 | |||
519 | static int |
||
520 | i915_gem_execbuffer_reserve(struct intel_ring_buffer *ring, |
||
3480 | Serge | 521 | struct list_head *objects, |
4104 | Serge | 522 | struct i915_address_space *vm, |
3480 | Serge | 523 | bool *need_relocs) |
3263 | Serge | 524 | { |
525 | struct drm_i915_gem_object *obj; |
||
526 | struct list_head ordered_objects; |
||
527 | bool has_fenced_gpu_access = INTEL_INFO(ring->dev)->gen < 4; |
||
528 | int retry; |
||
529 | |||
530 | INIT_LIST_HEAD(&ordered_objects); |
||
531 | while (!list_empty(objects)) { |
||
532 | struct drm_i915_gem_exec_object2 *entry; |
||
533 | bool need_fence, need_mappable; |
||
534 | |||
535 | obj = list_first_entry(objects, |
||
536 | struct drm_i915_gem_object, |
||
537 | exec_list); |
||
538 | entry = obj->exec_entry; |
||
539 | |||
540 | need_fence = |
||
541 | has_fenced_gpu_access && |
||
542 | entry->flags & EXEC_OBJECT_NEEDS_FENCE && |
||
543 | obj->tiling_mode != I915_TILING_NONE; |
||
544 | need_mappable = need_fence || need_reloc_mappable(obj); |
||
545 | |||
546 | if (need_mappable) |
||
547 | list_move(&obj->exec_list, &ordered_objects); |
||
548 | else |
||
549 | list_move_tail(&obj->exec_list, &ordered_objects); |
||
550 | |||
3480 | Serge | 551 | obj->base.pending_read_domains = I915_GEM_GPU_DOMAINS & ~I915_GEM_DOMAIN_COMMAND; |
3263 | Serge | 552 | obj->base.pending_write_domain = 0; |
553 | obj->pending_fenced_gpu_access = false; |
||
554 | } |
||
555 | list_splice(&ordered_objects, objects); |
||
556 | |||
557 | /* Attempt to pin all of the buffers into the GTT. |
||
558 | * This is done in 3 phases: |
||
559 | * |
||
560 | * 1a. Unbind all objects that do not match the GTT constraints for |
||
561 | * the execbuffer (fenceable, mappable, alignment etc). |
||
562 | * 1b. Increment pin count for already bound objects. |
||
563 | * 2. Bind new objects. |
||
564 | * 3. Decrement pin count. |
||
565 | * |
||
566 | * This avoid unnecessary unbinding of later objects in order to make |
||
567 | * room for the earlier objects *unless* we need to defragment. |
||
568 | */ |
||
569 | retry = 0; |
||
570 | do { |
||
571 | int ret = 0; |
||
572 | |||
573 | /* Unbind any ill-fitting objects or pin. */ |
||
574 | list_for_each_entry(obj, objects, exec_list) { |
||
575 | struct drm_i915_gem_exec_object2 *entry = obj->exec_entry; |
||
576 | bool need_fence, need_mappable; |
||
4104 | Serge | 577 | u32 obj_offset; |
3263 | Serge | 578 | |
4104 | Serge | 579 | if (!i915_gem_obj_bound(obj, vm)) |
3263 | Serge | 580 | continue; |
581 | |||
4104 | Serge | 582 | obj_offset = i915_gem_obj_offset(obj, vm); |
3263 | Serge | 583 | need_fence = |
584 | has_fenced_gpu_access && |
||
585 | entry->flags & EXEC_OBJECT_NEEDS_FENCE && |
||
586 | obj->tiling_mode != I915_TILING_NONE; |
||
587 | need_mappable = need_fence || need_reloc_mappable(obj); |
||
588 | |||
4246 | Serge | 589 | WARN_ON((need_mappable || need_fence) && |
590 | !i915_is_ggtt(vm)); |
||
4104 | Serge | 591 | |
592 | if ((entry->alignment && |
||
593 | obj_offset & (entry->alignment - 1)) || |
||
3263 | Serge | 594 | (need_mappable && !obj->map_and_fenceable)) |
4104 | Serge | 595 | ret = i915_vma_unbind(i915_gem_obj_to_vma(obj, vm)); |
3263 | Serge | 596 | else |
4104 | Serge | 597 | ret = i915_gem_execbuffer_reserve_object(obj, ring, vm, need_relocs); |
3263 | Serge | 598 | if (ret) |
599 | goto err; |
||
600 | } |
||
601 | |||
602 | /* Bind fresh objects */ |
||
603 | list_for_each_entry(obj, objects, exec_list) { |
||
4104 | Serge | 604 | if (i915_gem_obj_bound(obj, vm)) |
3263 | Serge | 605 | continue; |
606 | |||
4104 | Serge | 607 | ret = i915_gem_execbuffer_reserve_object(obj, ring, vm, need_relocs); |
3263 | Serge | 608 | if (ret) |
609 | goto err; |
||
610 | } |
||
611 | |||
612 | err: /* Decrement pin count for bound objects */ |
||
613 | list_for_each_entry(obj, objects, exec_list) |
||
614 | i915_gem_execbuffer_unreserve_object(obj); |
||
615 | |||
616 | if (ret != -ENOSPC || retry++) |
||
617 | return ret; |
||
618 | |||
619 | // ret = i915_gem_evict_everything(ring->dev); |
||
620 | if (ret) |
||
621 | return ret; |
||
622 | } while (1); |
||
623 | } |
||
624 | |||
625 | static int |
||
626 | i915_gem_execbuffer_relocate_slow(struct drm_device *dev, |
||
3480 | Serge | 627 | struct drm_i915_gem_execbuffer2 *args, |
3263 | Serge | 628 | struct drm_file *file, |
629 | struct intel_ring_buffer *ring, |
||
630 | struct eb_objects *eb, |
||
4104 | Serge | 631 | struct drm_i915_gem_exec_object2 *exec, |
632 | struct i915_address_space *vm) |
||
3263 | Serge | 633 | { |
634 | struct drm_i915_gem_relocation_entry *reloc; |
||
635 | struct drm_i915_gem_object *obj; |
||
3480 | Serge | 636 | bool need_relocs; |
3263 | Serge | 637 | int *reloc_offset; |
638 | int i, total, ret; |
||
3480 | Serge | 639 | int count = args->buffer_count; |
3263 | Serge | 640 | |
641 | /* We may process another execbuffer during the unlock... */ |
||
3480 | Serge | 642 | while (!list_empty(&eb->objects)) { |
643 | obj = list_first_entry(&eb->objects, |
||
3263 | Serge | 644 | struct drm_i915_gem_object, |
645 | exec_list); |
||
646 | list_del_init(&obj->exec_list); |
||
647 | drm_gem_object_unreference(&obj->base); |
||
648 | } |
||
649 | |||
650 | mutex_unlock(&dev->struct_mutex); |
||
651 | |||
652 | total = 0; |
||
653 | for (i = 0; i < count; i++) |
||
654 | total += exec[i].relocation_count; |
||
655 | |||
656 | reloc_offset = malloc(count * sizeof(*reloc_offset)); |
||
657 | reloc = malloc(total * sizeof(*reloc)); |
||
658 | if (reloc == NULL || reloc_offset == NULL) { |
||
3266 | Serge | 659 | kfree(reloc); |
660 | kfree(reloc_offset); |
||
3263 | Serge | 661 | mutex_lock(&dev->struct_mutex); |
662 | return -ENOMEM; |
||
663 | } |
||
664 | |||
665 | total = 0; |
||
666 | for (i = 0; i < count; i++) { |
||
667 | struct drm_i915_gem_relocation_entry __user *user_relocs; |
||
668 | u64 invalid_offset = (u64)-1; |
||
669 | int j; |
||
670 | |||
4539 | Serge | 671 | user_relocs = to_user_ptr(exec[i].relocs_ptr); |
3263 | Serge | 672 | |
673 | if (copy_from_user(reloc+total, user_relocs, |
||
674 | exec[i].relocation_count * sizeof(*reloc))) { |
||
675 | ret = -EFAULT; |
||
676 | mutex_lock(&dev->struct_mutex); |
||
677 | goto err; |
||
678 | } |
||
679 | |||
680 | /* As we do not update the known relocation offsets after |
||
681 | * relocating (due to the complexities in lock handling), |
||
682 | * we need to mark them as invalid now so that we force the |
||
683 | * relocation processing next time. Just in case the target |
||
684 | * object is evicted and then rebound into its old |
||
685 | * presumed_offset before the next execbuffer - if that |
||
686 | * happened we would make the mistake of assuming that the |
||
687 | * relocations were valid. |
||
688 | */ |
||
689 | for (j = 0; j < exec[i].relocation_count; j++) { |
||
690 | if (copy_to_user(&user_relocs[j].presumed_offset, |
||
691 | &invalid_offset, |
||
692 | sizeof(invalid_offset))) { |
||
693 | ret = -EFAULT; |
||
694 | mutex_lock(&dev->struct_mutex); |
||
695 | goto err; |
||
696 | } |
||
697 | } |
||
698 | |||
699 | reloc_offset[i] = total; |
||
700 | total += exec[i].relocation_count; |
||
701 | } |
||
702 | |||
703 | ret = i915_mutex_lock_interruptible(dev); |
||
704 | if (ret) { |
||
705 | mutex_lock(&dev->struct_mutex); |
||
706 | goto err; |
||
707 | } |
||
708 | |||
709 | /* reacquire the objects */ |
||
710 | eb_reset(eb); |
||
3480 | Serge | 711 | ret = eb_lookup_objects(eb, exec, args, file); |
712 | if (ret) |
||
3263 | Serge | 713 | goto err; |
714 | |||
3480 | Serge | 715 | need_relocs = (args->flags & I915_EXEC_NO_RELOC) == 0; |
4104 | Serge | 716 | ret = i915_gem_execbuffer_reserve(ring, &eb->objects, vm, &need_relocs); |
3263 | Serge | 717 | if (ret) |
718 | goto err; |
||
719 | |||
3480 | Serge | 720 | list_for_each_entry(obj, &eb->objects, exec_list) { |
3263 | Serge | 721 | int offset = obj->exec_entry - exec; |
722 | ret = i915_gem_execbuffer_relocate_object_slow(obj, eb, |
||
4104 | Serge | 723 | reloc + reloc_offset[offset], |
724 | vm); |
||
3263 | Serge | 725 | if (ret) |
726 | goto err; |
||
727 | } |
||
728 | |||
729 | /* Leave the user relocations as are, this is the painfully slow path, |
||
730 | * and we want to avoid the complication of dropping the lock whilst |
||
731 | * having buffers reserved in the aperture and so causing spurious |
||
732 | * ENOSPC for random operations. |
||
733 | */ |
||
734 | |||
735 | err: |
||
3266 | Serge | 736 | kfree(reloc); |
737 | kfree(reloc_offset); |
||
3263 | Serge | 738 | return ret; |
739 | } |
||
740 | |||
741 | static int |
||
742 | i915_gem_execbuffer_move_to_gpu(struct intel_ring_buffer *ring, |
||
743 | struct list_head *objects) |
||
744 | { |
||
745 | struct drm_i915_gem_object *obj; |
||
746 | uint32_t flush_domains = 0; |
||
4104 | Serge | 747 | bool flush_chipset = false; |
3263 | Serge | 748 | int ret; |
749 | |||
750 | list_for_each_entry(obj, objects, exec_list) { |
||
751 | ret = i915_gem_object_sync(obj, ring); |
||
752 | if (ret) |
||
753 | return ret; |
||
754 | |||
755 | if (obj->base.write_domain & I915_GEM_DOMAIN_CPU) |
||
4104 | Serge | 756 | flush_chipset |= i915_gem_clflush_object(obj, false); |
3263 | Serge | 757 | |
758 | flush_domains |= obj->base.write_domain; |
||
759 | } |
||
760 | |||
4104 | Serge | 761 | if (flush_chipset) |
3263 | Serge | 762 | i915_gem_chipset_flush(ring->dev); |
763 | |||
764 | if (flush_domains & I915_GEM_DOMAIN_GTT) |
||
765 | wmb(); |
||
766 | |||
767 | /* Unconditionally invalidate gpu caches and ensure that we do flush |
||
768 | * any residual writes from the previous batch. |
||
769 | */ |
||
770 | return intel_ring_invalidate_all_caches(ring); |
||
771 | } |
||
772 | |||
773 | static bool |
||
774 | i915_gem_check_execbuffer(struct drm_i915_gem_execbuffer2 *exec) |
||
775 | { |
||
3480 | Serge | 776 | if (exec->flags & __I915_EXEC_UNKNOWN_FLAGS) |
777 | return false; |
||
778 | |||
3263 | Serge | 779 | return ((exec->batch_start_offset | exec->batch_len) & 0x7) == 0; |
780 | } |
||
781 | |||
782 | static int |
||
783 | validate_exec_list(struct drm_i915_gem_exec_object2 *exec, |
||
784 | int count) |
||
785 | { |
||
786 | int i; |
||
3480 | Serge | 787 | int relocs_total = 0; |
788 | int relocs_max = INT_MAX / sizeof(struct drm_i915_gem_relocation_entry); |
||
3263 | Serge | 789 | |
790 | for (i = 0; i < count; i++) { |
||
3746 | Serge | 791 | char __user *ptr = to_user_ptr(exec[i].relocs_ptr); |
3263 | Serge | 792 | int length; /* limited by fault_in_pages_readable() */ |
793 | |||
3480 | Serge | 794 | if (exec[i].flags & __EXEC_OBJECT_UNKNOWN_FLAGS) |
3263 | Serge | 795 | return -EINVAL; |
796 | |||
3480 | Serge | 797 | /* First check for malicious input causing overflow in |
798 | * the worst case where we need to allocate the entire |
||
799 | * relocation tree as a single array. |
||
800 | */ |
||
801 | if (exec[i].relocation_count > relocs_max - relocs_total) |
||
802 | return -EINVAL; |
||
803 | relocs_total += exec[i].relocation_count; |
||
804 | |||
3263 | Serge | 805 | length = exec[i].relocation_count * |
806 | sizeof(struct drm_i915_gem_relocation_entry); |
||
3746 | Serge | 807 | /* |
808 | * We must check that the entire relocation array is safe |
||
809 | * to read, but since we may need to update the presumed |
||
810 | * offsets during execution, check for full write access. |
||
811 | */ |
||
3263 | Serge | 812 | |
813 | } |
||
814 | |||
815 | return 0; |
||
816 | } |
||
817 | |||
818 | static void |
||
819 | i915_gem_execbuffer_move_to_active(struct list_head *objects, |
||
4104 | Serge | 820 | struct i915_address_space *vm, |
3263 | Serge | 821 | struct intel_ring_buffer *ring) |
822 | { |
||
823 | struct drm_i915_gem_object *obj; |
||
824 | |||
825 | list_for_each_entry(obj, objects, exec_list) { |
||
826 | u32 old_read = obj->base.read_domains; |
||
827 | u32 old_write = obj->base.write_domain; |
||
828 | |||
3480 | Serge | 829 | obj->base.write_domain = obj->base.pending_write_domain; |
830 | if (obj->base.write_domain == 0) |
||
831 | obj->base.pending_read_domains |= obj->base.read_domains; |
||
3263 | Serge | 832 | obj->base.read_domains = obj->base.pending_read_domains; |
833 | obj->fenced_gpu_access = obj->pending_fenced_gpu_access; |
||
834 | |||
4104 | Serge | 835 | /* FIXME: This lookup gets fixed later <-- danvet */ |
836 | list_move_tail(&i915_gem_obj_to_vma(obj, vm)->mm_list, &vm->active_list); |
||
3263 | Serge | 837 | i915_gem_object_move_to_active(obj, ring); |
838 | if (obj->base.write_domain) { |
||
839 | obj->dirty = 1; |
||
840 | obj->last_write_seqno = intel_ring_get_seqno(ring); |
||
841 | if (obj->pin_count) /* check for potential scanout */ |
||
4104 | Serge | 842 | intel_mark_fb_busy(obj, ring); |
3263 | Serge | 843 | } |
844 | |||
845 | trace_i915_gem_object_change_domain(obj, old_read, old_write); |
||
846 | } |
||
847 | } |
||
848 | |||
849 | static void |
||
850 | i915_gem_execbuffer_retire_commands(struct drm_device *dev, |
||
851 | struct drm_file *file, |
||
4104 | Serge | 852 | struct intel_ring_buffer *ring, |
853 | struct drm_i915_gem_object *obj) |
||
3263 | Serge | 854 | { |
855 | /* Unconditionally force add_request to emit a full flush. */ |
||
856 | ring->gpu_caches_dirty = true; |
||
857 | |||
858 | /* Add a breadcrumb for the completion of the batch buffer */ |
||
4104 | Serge | 859 | (void)__i915_add_request(ring, file, obj, NULL); |
3263 | Serge | 860 | } |
861 | |||
862 | static int |
||
863 | i915_reset_gen7_sol_offsets(struct drm_device *dev, |
||
864 | struct intel_ring_buffer *ring) |
||
865 | { |
||
866 | drm_i915_private_t *dev_priv = dev->dev_private; |
||
867 | int ret, i; |
||
868 | |||
869 | if (!IS_GEN7(dev) || ring != &dev_priv->ring[RCS]) |
||
870 | return 0; |
||
871 | |||
872 | ret = intel_ring_begin(ring, 4 * 3); |
||
873 | if (ret) |
||
874 | return ret; |
||
875 | |||
876 | for (i = 0; i < 4; i++) { |
||
877 | intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1)); |
||
878 | intel_ring_emit(ring, GEN7_SO_WRITE_OFFSET(i)); |
||
879 | intel_ring_emit(ring, 0); |
||
880 | } |
||
881 | |||
882 | intel_ring_advance(ring); |
||
883 | |||
884 | return 0; |
||
885 | } |
||
886 | |||
887 | static int |
||
888 | i915_gem_do_execbuffer(struct drm_device *dev, void *data, |
||
889 | struct drm_file *file, |
||
890 | struct drm_i915_gem_execbuffer2 *args, |
||
4104 | Serge | 891 | struct drm_i915_gem_exec_object2 *exec, |
892 | struct i915_address_space *vm) |
||
3263 | Serge | 893 | { |
894 | drm_i915_private_t *dev_priv = dev->dev_private; |
||
895 | struct eb_objects *eb; |
||
896 | struct drm_i915_gem_object *batch_obj; |
||
897 | struct drm_clip_rect *cliprects = NULL; |
||
898 | struct intel_ring_buffer *ring; |
||
899 | u32 ctx_id = i915_execbuffer2_get_context_id(*args); |
||
900 | u32 exec_start, exec_len; |
||
3480 | Serge | 901 | u32 mask, flags; |
3263 | Serge | 902 | int ret, mode, i; |
3480 | Serge | 903 | bool need_relocs; |
3263 | Serge | 904 | |
3480 | Serge | 905 | if (!i915_gem_check_execbuffer(args)) |
3263 | Serge | 906 | return -EINVAL; |
907 | |||
908 | ret = validate_exec_list(exec, args->buffer_count); |
||
909 | if (ret) |
||
910 | return ret; |
||
911 | |||
912 | flags = 0; |
||
913 | if (args->flags & I915_EXEC_SECURE) { |
||
914 | |||
915 | flags |= I915_DISPATCH_SECURE; |
||
916 | } |
||
917 | if (args->flags & I915_EXEC_IS_PINNED) |
||
918 | flags |= I915_DISPATCH_PINNED; |
||
919 | |||
920 | switch (args->flags & I915_EXEC_RING_MASK) { |
||
921 | case I915_EXEC_DEFAULT: |
||
922 | case I915_EXEC_RENDER: |
||
923 | ring = &dev_priv->ring[RCS]; |
||
924 | break; |
||
925 | case I915_EXEC_BSD: |
||
926 | ring = &dev_priv->ring[VCS]; |
||
4104 | Serge | 927 | if (ctx_id != DEFAULT_CONTEXT_ID) { |
3263 | Serge | 928 | DRM_DEBUG("Ring %s doesn't support contexts\n", |
929 | ring->name); |
||
930 | return -EPERM; |
||
931 | } |
||
932 | break; |
||
933 | case I915_EXEC_BLT: |
||
934 | ring = &dev_priv->ring[BCS]; |
||
4104 | Serge | 935 | if (ctx_id != DEFAULT_CONTEXT_ID) { |
3263 | Serge | 936 | DRM_DEBUG("Ring %s doesn't support contexts\n", |
937 | ring->name); |
||
938 | return -EPERM; |
||
939 | } |
||
940 | break; |
||
4104 | Serge | 941 | case I915_EXEC_VEBOX: |
942 | ring = &dev_priv->ring[VECS]; |
||
943 | if (ctx_id != DEFAULT_CONTEXT_ID) { |
||
944 | DRM_DEBUG("Ring %s doesn't support contexts\n", |
||
945 | ring->name); |
||
946 | return -EPERM; |
||
947 | } |
||
948 | break; |
||
949 | |||
3263 | Serge | 950 | default: |
951 | DRM_DEBUG("execbuf with unknown ring: %d\n", |
||
952 | (int)(args->flags & I915_EXEC_RING_MASK)); |
||
953 | return -EINVAL; |
||
954 | } |
||
955 | if (!intel_ring_initialized(ring)) { |
||
956 | DRM_DEBUG("execbuf with invalid ring: %d\n", |
||
957 | (int)(args->flags & I915_EXEC_RING_MASK)); |
||
958 | return -EINVAL; |
||
959 | } |
||
960 | |||
961 | mode = args->flags & I915_EXEC_CONSTANTS_MASK; |
||
962 | mask = I915_EXEC_CONSTANTS_MASK; |
||
963 | switch (mode) { |
||
964 | case I915_EXEC_CONSTANTS_REL_GENERAL: |
||
965 | case I915_EXEC_CONSTANTS_ABSOLUTE: |
||
966 | case I915_EXEC_CONSTANTS_REL_SURFACE: |
||
967 | if (ring == &dev_priv->ring[RCS] && |
||
968 | mode != dev_priv->relative_constants_mode) { |
||
969 | if (INTEL_INFO(dev)->gen < 4) |
||
970 | return -EINVAL; |
||
971 | |||
972 | if (INTEL_INFO(dev)->gen > 5 && |
||
973 | mode == I915_EXEC_CONSTANTS_REL_SURFACE) |
||
974 | return -EINVAL; |
||
975 | |||
976 | /* The HW changed the meaning on this bit on gen6 */ |
||
977 | if (INTEL_INFO(dev)->gen >= 6) |
||
978 | mask &= ~I915_EXEC_CONSTANTS_REL_SURFACE; |
||
979 | } |
||
980 | break; |
||
981 | default: |
||
982 | DRM_DEBUG("execbuf with unknown constants: %d\n", mode); |
||
983 | return -EINVAL; |
||
984 | } |
||
985 | |||
986 | if (args->buffer_count < 1) { |
||
987 | DRM_DEBUG("execbuf with %d buffers\n", args->buffer_count); |
||
988 | return -EINVAL; |
||
989 | } |
||
990 | |||
991 | if (args->num_cliprects != 0) { |
||
992 | if (ring != &dev_priv->ring[RCS]) { |
||
993 | DRM_DEBUG("clip rectangles are only valid with the render ring\n"); |
||
994 | return -EINVAL; |
||
995 | } |
||
996 | |||
997 | if (INTEL_INFO(dev)->gen >= 5) { |
||
998 | DRM_DEBUG("clip rectangles are only valid on pre-gen5\n"); |
||
999 | return -EINVAL; |
||
1000 | } |
||
1001 | |||
1002 | if (args->num_cliprects > UINT_MAX / sizeof(*cliprects)) { |
||
1003 | DRM_DEBUG("execbuf with %u cliprects\n", |
||
1004 | args->num_cliprects); |
||
1005 | return -EINVAL; |
||
1006 | } |
||
1007 | |||
1008 | cliprects = kmalloc(args->num_cliprects * sizeof(*cliprects), |
||
1009 | GFP_KERNEL); |
||
1010 | if (cliprects == NULL) { |
||
1011 | ret = -ENOMEM; |
||
1012 | goto pre_mutex_err; |
||
1013 | } |
||
1014 | |||
1015 | if (copy_from_user(cliprects, |
||
3746 | Serge | 1016 | to_user_ptr(args->cliprects_ptr), |
3263 | Serge | 1017 | sizeof(*cliprects)*args->num_cliprects)) { |
1018 | ret = -EFAULT; |
||
1019 | goto pre_mutex_err; |
||
1020 | } |
||
1021 | } |
||
1022 | |||
1023 | ret = i915_mutex_lock_interruptible(dev); |
||
1024 | if (ret) |
||
1025 | goto pre_mutex_err; |
||
1026 | |||
4104 | Serge | 1027 | if (dev_priv->ums.mm_suspended) { |
3263 | Serge | 1028 | mutex_unlock(&dev->struct_mutex); |
1029 | ret = -EBUSY; |
||
1030 | goto pre_mutex_err; |
||
1031 | } |
||
1032 | |||
3480 | Serge | 1033 | eb = eb_create(args); |
3263 | Serge | 1034 | if (eb == NULL) { |
1035 | mutex_unlock(&dev->struct_mutex); |
||
1036 | ret = -ENOMEM; |
||
1037 | goto pre_mutex_err; |
||
1038 | } |
||
1039 | |||
1040 | /* Look up object handles */ |
||
3480 | Serge | 1041 | ret = eb_lookup_objects(eb, exec, args, file); |
1042 | if (ret) |
||
3263 | Serge | 1043 | goto err; |
1044 | |||
1045 | /* take note of the batch buffer before we might reorder the lists */ |
||
3480 | Serge | 1046 | batch_obj = list_entry(eb->objects.prev, |
3263 | Serge | 1047 | struct drm_i915_gem_object, |
1048 | exec_list); |
||
1049 | |||
1050 | /* Move the objects en-masse into the GTT, evicting if necessary. */ |
||
3480 | Serge | 1051 | need_relocs = (args->flags & I915_EXEC_NO_RELOC) == 0; |
4104 | Serge | 1052 | ret = i915_gem_execbuffer_reserve(ring, &eb->objects, vm, &need_relocs); |
3263 | Serge | 1053 | if (ret) |
1054 | goto err; |
||
1055 | |||
1056 | /* The objects are in their final locations, apply the relocations. */ |
||
3480 | Serge | 1057 | if (need_relocs) |
4104 | Serge | 1058 | ret = i915_gem_execbuffer_relocate(eb, vm); |
3263 | Serge | 1059 | if (ret) { |
1060 | if (ret == -EFAULT) { |
||
3480 | Serge | 1061 | ret = i915_gem_execbuffer_relocate_slow(dev, args, file, ring, |
4104 | Serge | 1062 | eb, exec, vm); |
3263 | Serge | 1063 | BUG_ON(!mutex_is_locked(&dev->struct_mutex)); |
1064 | } |
||
1065 | if (ret) |
||
1066 | goto err; |
||
1067 | } |
||
1068 | |||
1069 | /* Set the pending read domains for the batch buffer to COMMAND */ |
||
1070 | if (batch_obj->base.pending_write_domain) { |
||
1071 | DRM_DEBUG("Attempting to use self-modifying batch buffer\n"); |
||
1072 | ret = -EINVAL; |
||
1073 | goto err; |
||
1074 | } |
||
1075 | batch_obj->base.pending_read_domains |= I915_GEM_DOMAIN_COMMAND; |
||
1076 | |||
1077 | /* snb/ivb/vlv conflate the "batch in ppgtt" bit with the "non-secure |
||
1078 | * batch" bit. Hence we need to pin secure batches into the global gtt. |
||
1079 | * hsw should have this fixed, but let's be paranoid and do it |
||
1080 | * unconditionally for now. */ |
||
1081 | if (flags & I915_DISPATCH_SECURE && !batch_obj->has_global_gtt_mapping) |
||
1082 | i915_gem_gtt_bind_object(batch_obj, batch_obj->cache_level); |
||
1083 | |||
3480 | Serge | 1084 | ret = i915_gem_execbuffer_move_to_gpu(ring, &eb->objects); |
3263 | Serge | 1085 | if (ret) |
1086 | goto err; |
||
1087 | |||
1088 | ret = i915_switch_context(ring, file, ctx_id); |
||
1089 | if (ret) |
||
1090 | goto err; |
||
1091 | |||
1092 | if (ring == &dev_priv->ring[RCS] && |
||
1093 | mode != dev_priv->relative_constants_mode) { |
||
1094 | ret = intel_ring_begin(ring, 4); |
||
1095 | if (ret) |
||
1096 | goto err; |
||
1097 | |||
1098 | intel_ring_emit(ring, MI_NOOP); |
||
1099 | intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1)); |
||
1100 | intel_ring_emit(ring, INSTPM); |
||
1101 | intel_ring_emit(ring, mask << 16 | mode); |
||
1102 | intel_ring_advance(ring); |
||
1103 | |||
1104 | dev_priv->relative_constants_mode = mode; |
||
1105 | } |
||
1106 | |||
1107 | if (args->flags & I915_EXEC_GEN7_SOL_RESET) { |
||
1108 | ret = i915_reset_gen7_sol_offsets(dev, ring); |
||
1109 | if (ret) |
||
1110 | goto err; |
||
1111 | } |
||
1112 | |||
4104 | Serge | 1113 | exec_start = i915_gem_obj_offset(batch_obj, vm) + |
1114 | args->batch_start_offset; |
||
3263 | Serge | 1115 | exec_len = args->batch_len; |
1116 | if (cliprects) { |
||
4246 | Serge | 1117 | for (i = 0; i < args->num_cliprects; i++) { |
1118 | ret = i915_emit_box(dev, &cliprects[i], |
||
1119 | args->DR1, args->DR4); |
||
1120 | if (ret) |
||
1121 | goto err; |
||
3263 | Serge | 1122 | |
4246 | Serge | 1123 | ret = ring->dispatch_execbuffer(ring, |
1124 | exec_start, exec_len, |
||
1125 | flags); |
||
1126 | if (ret) |
||
1127 | goto err; |
||
1128 | } |
||
3263 | Serge | 1129 | } else { |
1130 | ret = ring->dispatch_execbuffer(ring, |
||
1131 | exec_start, exec_len, |
||
1132 | flags); |
||
1133 | if (ret) |
||
1134 | goto err; |
||
1135 | } |
||
1136 | |||
3266 | Serge | 1137 | trace_i915_gem_ring_dispatch(ring, intel_ring_get_seqno(ring), flags); |
3263 | Serge | 1138 | |
4104 | Serge | 1139 | i915_gem_execbuffer_move_to_active(&eb->objects, vm, ring); |
1140 | i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj); |
||
3266 | Serge | 1141 | |
3263 | Serge | 1142 | err: |
1143 | eb_destroy(eb); |
||
1144 | |||
1145 | mutex_unlock(&dev->struct_mutex); |
||
1146 | |||
1147 | pre_mutex_err: |
||
4104 | Serge | 1148 | kfree(cliprects); |
3263 | Serge | 1149 | return ret; |
1150 | } |
||
1151 | |||
4246 | Serge | 1152 | #if 0 |
1153 | /* |
||
1154 | * Legacy execbuffer just creates an exec2 list from the original exec object |
||
1155 | * list array and passes it to the real function. |
||
1156 | */ |
||
1157 | int |
||
1158 | i915_gem_execbuffer(struct drm_device *dev, void *data, |
||
1159 | struct drm_file *file) |
||
1160 | { |
||
1161 | struct drm_i915_private *dev_priv = dev->dev_private; |
||
1162 | struct drm_i915_gem_execbuffer *args = data; |
||
1163 | struct drm_i915_gem_execbuffer2 exec2; |
||
1164 | struct drm_i915_gem_exec_object *exec_list = NULL; |
||
1165 | struct drm_i915_gem_exec_object2 *exec2_list = NULL; |
||
1166 | int ret, i; |
||
3480 | Serge | 1167 | |
4246 | Serge | 1168 | if (args->buffer_count < 1) { |
1169 | DRM_DEBUG("execbuf with %d buffers\n", args->buffer_count); |
||
1170 | return -EINVAL; |
||
1171 | } |
||
3480 | Serge | 1172 | |
4246 | Serge | 1173 | /* Copy in the exec list from userland */ |
1174 | exec_list = drm_malloc_ab(sizeof(*exec_list), args->buffer_count); |
||
1175 | exec2_list = drm_malloc_ab(sizeof(*exec2_list), args->buffer_count); |
||
1176 | if (exec_list == NULL || exec2_list == NULL) { |
||
1177 | DRM_DEBUG("Failed to allocate exec list for %d buffers\n", |
||
1178 | args->buffer_count); |
||
1179 | drm_free_large(exec_list); |
||
1180 | drm_free_large(exec2_list); |
||
1181 | return -ENOMEM; |
||
1182 | } |
||
1183 | ret = copy_from_user(exec_list, |
||
1184 | to_user_ptr(args->buffers_ptr), |
||
1185 | sizeof(*exec_list) * args->buffer_count); |
||
1186 | if (ret != 0) { |
||
1187 | DRM_DEBUG("copy %d exec entries failed %d\n", |
||
1188 | args->buffer_count, ret); |
||
1189 | drm_free_large(exec_list); |
||
1190 | drm_free_large(exec2_list); |
||
1191 | return -EFAULT; |
||
1192 | } |
||
1193 | |||
1194 | for (i = 0; i < args->buffer_count; i++) { |
||
1195 | exec2_list[i].handle = exec_list[i].handle; |
||
1196 | exec2_list[i].relocation_count = exec_list[i].relocation_count; |
||
1197 | exec2_list[i].relocs_ptr = exec_list[i].relocs_ptr; |
||
1198 | exec2_list[i].alignment = exec_list[i].alignment; |
||
1199 | exec2_list[i].offset = exec_list[i].offset; |
||
1200 | if (INTEL_INFO(dev)->gen < 4) |
||
1201 | exec2_list[i].flags = EXEC_OBJECT_NEEDS_FENCE; |
||
1202 | else |
||
1203 | exec2_list[i].flags = 0; |
||
1204 | } |
||
1205 | |||
1206 | exec2.buffers_ptr = args->buffers_ptr; |
||
1207 | exec2.buffer_count = args->buffer_count; |
||
1208 | exec2.batch_start_offset = args->batch_start_offset; |
||
1209 | exec2.batch_len = args->batch_len; |
||
1210 | exec2.DR1 = args->DR1; |
||
1211 | exec2.DR4 = args->DR4; |
||
1212 | exec2.num_cliprects = args->num_cliprects; |
||
1213 | exec2.cliprects_ptr = args->cliprects_ptr; |
||
1214 | exec2.flags = I915_EXEC_RENDER; |
||
1215 | i915_execbuffer2_set_context_id(exec2, 0); |
||
1216 | |||
1217 | ret = i915_gem_do_execbuffer(dev, data, file, &exec2, exec2_list, |
||
1218 | &dev_priv->gtt.base); |
||
1219 | if (!ret) { |
||
1220 | /* Copy the new buffer offsets back to the user's exec list. */ |
||
1221 | for (i = 0; i < args->buffer_count; i++) |
||
1222 | exec_list[i].offset = exec2_list[i].offset; |
||
1223 | /* ... and back out to userspace */ |
||
1224 | ret = copy_to_user(to_user_ptr(args->buffers_ptr), |
||
1225 | exec_list, |
||
1226 | sizeof(*exec_list) * args->buffer_count); |
||
1227 | if (ret) { |
||
1228 | ret = -EFAULT; |
||
1229 | DRM_DEBUG("failed to copy %d exec entries " |
||
1230 | "back to user (%d)\n", |
||
1231 | args->buffer_count, ret); |
||
1232 | } |
||
1233 | } |
||
1234 | |||
1235 | drm_free_large(exec_list); |
||
1236 | drm_free_large(exec2_list); |
||
1237 | return ret; |
||
1238 | } |
||
1239 | #endif |
||
1240 | |||
3263 | Serge | 1241 | int |
1242 | i915_gem_execbuffer2(struct drm_device *dev, void *data, |
||
1243 | struct drm_file *file) |
||
1244 | { |
||
4104 | Serge | 1245 | struct drm_i915_private *dev_priv = dev->dev_private; |
3263 | Serge | 1246 | struct drm_i915_gem_execbuffer2 *args = data; |
1247 | struct drm_i915_gem_exec_object2 *exec2_list = NULL; |
||
1248 | int ret; |
||
1249 | |||
1250 | if (args->buffer_count < 1 || |
||
1251 | args->buffer_count > UINT_MAX / sizeof(*exec2_list)) { |
||
1252 | DRM_DEBUG("execbuf2 with %d buffers\n", args->buffer_count); |
||
1253 | return -EINVAL; |
||
1254 | } |
||
1255 | |||
3480 | Serge | 1256 | exec2_list = kmalloc(sizeof(*exec2_list)*args->buffer_count, |
1257 | GFP_TEMPORARY | __GFP_NOWARN | __GFP_NORETRY); |
||
3263 | Serge | 1258 | if (exec2_list == NULL) { |
1259 | DRM_DEBUG("Failed to allocate exec list for %d buffers\n", |
||
1260 | args->buffer_count); |
||
1261 | return -ENOMEM; |
||
1262 | } |
||
1263 | ret = copy_from_user(exec2_list, |
||
4539 | Serge | 1264 | to_user_ptr(args->buffers_ptr), |
3263 | Serge | 1265 | sizeof(*exec2_list) * args->buffer_count); |
1266 | if (ret != 0) { |
||
1267 | DRM_DEBUG("copy %d exec entries failed %d\n", |
||
1268 | args->buffer_count, ret); |
||
3266 | Serge | 1269 | kfree(exec2_list); |
1270 | FAIL(); |
||
3263 | Serge | 1271 | return -EFAULT; |
1272 | } |
||
1273 | |||
4104 | Serge | 1274 | ret = i915_gem_do_execbuffer(dev, data, file, args, exec2_list, |
1275 | &dev_priv->gtt.base); |
||
3263 | Serge | 1276 | if (!ret) { |
1277 | /* Copy the new buffer offsets back to the user's exec list. */ |
||
4539 | Serge | 1278 | ret = copy_to_user(to_user_ptr(args->buffers_ptr), |
3263 | Serge | 1279 | exec2_list, |
1280 | sizeof(*exec2_list) * args->buffer_count); |
||
1281 | if (ret) { |
||
1282 | ret = -EFAULT; |
||
1283 | DRM_DEBUG("failed to copy %d exec entries " |
||
1284 | "back to user (%d)\n", |
||
1285 | args->buffer_count, ret); |
||
1286 | } |
||
1287 | } |
||
1288 | |||
3266 | Serge | 1289 | kfree(exec2_list); |
3263 | Serge | 1290 | return ret; |
1291 | }>>>>>>><>>>>-->>>>>>>30) |