Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
1
fork

Configure Feed

Select the types of activity you want to include in your feed.

at v5.2 178 lines 5.2 kB view raw
1/* 2 * Copyright © 2016 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 */ 24 25#include "i915_drv.h" 26#include "intel_frontbuffer.h" 27#include "i915_gem_clflush.h" 28 29static DEFINE_SPINLOCK(clflush_lock); 30 31struct clflush { 32 struct dma_fence dma; /* Must be first for dma_fence_free() */ 33 struct i915_sw_fence wait; 34 struct work_struct work; 35 struct drm_i915_gem_object *obj; 36}; 37 38static const char *i915_clflush_get_driver_name(struct dma_fence *fence) 39{ 40 return DRIVER_NAME; 41} 42 43static const char *i915_clflush_get_timeline_name(struct dma_fence *fence) 44{ 45 return "clflush"; 46} 47 48static void i915_clflush_release(struct dma_fence *fence) 49{ 50 struct clflush *clflush = container_of(fence, typeof(*clflush), dma); 51 52 i915_sw_fence_fini(&clflush->wait); 53 54 BUILD_BUG_ON(offsetof(typeof(*clflush), dma)); 55 dma_fence_free(&clflush->dma); 56} 57 58static const struct dma_fence_ops i915_clflush_ops = { 59 .get_driver_name = i915_clflush_get_driver_name, 60 .get_timeline_name = i915_clflush_get_timeline_name, 61 .release = i915_clflush_release, 62}; 63 64static void __i915_do_clflush(struct drm_i915_gem_object *obj) 65{ 66 GEM_BUG_ON(!i915_gem_object_has_pages(obj)); 67 drm_clflush_sg(obj->mm.pages); 68 intel_fb_obj_flush(obj, ORIGIN_CPU); 69} 70 71static void i915_clflush_work(struct work_struct *work) 72{ 73 struct clflush *clflush = container_of(work, typeof(*clflush), work); 74 struct drm_i915_gem_object *obj = clflush->obj; 75 76 if (i915_gem_object_pin_pages(obj)) { 77 DRM_ERROR("Failed to acquire obj->pages for clflushing\n"); 78 goto out; 79 } 80 81 __i915_do_clflush(obj); 82 83 i915_gem_object_unpin_pages(obj); 84 85out: 86 i915_gem_object_put(obj); 87 88 dma_fence_signal(&clflush->dma); 89 dma_fence_put(&clflush->dma); 90} 91 92static int __i915_sw_fence_call 93i915_clflush_notify(struct i915_sw_fence *fence, 94 enum i915_sw_fence_notify state) 95{ 96 struct clflush *clflush = container_of(fence, typeof(*clflush), wait); 97 98 switch (state) { 99 case FENCE_COMPLETE: 100 schedule_work(&clflush->work); 101 break; 102 103 case FENCE_FREE: 104 dma_fence_put(&clflush->dma); 105 break; 106 } 107 108 return NOTIFY_DONE; 109} 110 111bool i915_gem_clflush_object(struct drm_i915_gem_object *obj, 112 unsigned int flags) 113{ 114 struct clflush *clflush; 115 116 /* 117 * Stolen memory is always coherent with the GPU as it is explicitly 118 * marked as wc by the system, or the system is cache-coherent. 119 * Similarly, we only access struct pages through the CPU cache, so 120 * anything not backed by physical memory we consider to be always 121 * coherent and not need clflushing. 122 */ 123 if (!i915_gem_object_has_struct_page(obj)) { 124 obj->cache_dirty = false; 125 return false; 126 } 127 128 /* If the GPU is snooping the contents of the CPU cache, 129 * we do not need to manually clear the CPU cache lines. However, 130 * the caches are only snooped when the render cache is 131 * flushed/invalidated. As we always have to emit invalidations 132 * and flushes when moving into and out of the RENDER domain, correct 133 * snooping behaviour occurs naturally as the result of our domain 134 * tracking. 135 */ 136 if (!(flags & I915_CLFLUSH_FORCE) && 137 obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ) 138 return false; 139 140 trace_i915_gem_object_clflush(obj); 141 142 clflush = NULL; 143 if (!(flags & I915_CLFLUSH_SYNC)) 144 clflush = kmalloc(sizeof(*clflush), GFP_KERNEL); 145 if (clflush) { 146 GEM_BUG_ON(!obj->cache_dirty); 147 148 dma_fence_init(&clflush->dma, 149 &i915_clflush_ops, 150 &clflush_lock, 151 to_i915(obj->base.dev)->mm.unordered_timeline, 152 0); 153 i915_sw_fence_init(&clflush->wait, i915_clflush_notify); 154 155 clflush->obj = i915_gem_object_get(obj); 156 INIT_WORK(&clflush->work, i915_clflush_work); 157 158 dma_fence_get(&clflush->dma); 159 160 i915_sw_fence_await_reservation(&clflush->wait, 161 obj->resv, NULL, 162 true, I915_FENCE_TIMEOUT, 163 I915_FENCE_GFP); 164 165 reservation_object_lock(obj->resv, NULL); 166 reservation_object_add_excl_fence(obj->resv, &clflush->dma); 167 reservation_object_unlock(obj->resv); 168 169 i915_sw_fence_commit(&clflush->wait); 170 } else if (obj->mm.pages) { 171 __i915_do_clflush(obj); 172 } else { 173 GEM_BUG_ON(obj->write_domain != I915_GEM_DOMAIN_CPU); 174 } 175 176 obj->cache_dirty = false; 177 return true; 178}