Rev 6084 | Only display areas with differences | Regard whitespace | Details | Blame | Last modification | View Log | RSS feed
Rev 6084 | Rev 6937 | ||
---|---|---|---|
1 | /* |
1 | /* |
2 | * Copyright © 2014 Intel Corporation |
2 | * Copyright © 2014 Intel Corporation |
3 | * |
3 | * |
4 | * Permission is hereby granted, free of charge, to any person obtaining a |
4 | * Permission is hereby granted, free of charge, to any person obtaining a |
5 | * copy of this software and associated documentation files (the "Software"), |
5 | * copy of this software and associated documentation files (the "Software"), |
6 | * to deal in the Software without restriction, including without limitation |
6 | * to deal in the Software without restriction, including without limitation |
7 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
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 |
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: |
9 | * Software is furnished to do so, subject to the following conditions: |
10 | * |
10 | * |
11 | * The above copyright notice and this permission notice (including the next |
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 |
12 | * paragraph) shall be included in all copies or substantial portions of the |
13 | * Software. |
13 | * Software. |
14 | * |
14 | * |
15 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
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, |
16 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
17 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
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 |
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 |
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 |
20 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS |
21 | * IN THE SOFTWARE. |
21 | * IN THE SOFTWARE. |
22 | * |
22 | * |
23 | * Authors: |
23 | * Authors: |
24 | * Mika Kuoppala |
24 | * Mika Kuoppala |
25 | * |
25 | * |
26 | */ |
26 | */ |
27 | 27 | ||
28 | #include "i915_drv.h" |
28 | #include "i915_drv.h" |
29 | #include "intel_renderstate.h" |
29 | #include "intel_renderstate.h" |
30 | 30 | ||
31 | static const struct intel_renderstate_rodata * |
31 | static const struct intel_renderstate_rodata * |
32 | render_state_get_rodata(struct drm_device *dev, const int gen) |
32 | render_state_get_rodata(struct drm_device *dev, const int gen) |
33 | { |
33 | { |
34 | switch (gen) { |
34 | switch (gen) { |
35 | case 6: |
35 | case 6: |
36 | return &gen6_null_state; |
36 | return &gen6_null_state; |
37 | case 7: |
37 | case 7: |
38 | return &gen7_null_state; |
38 | return &gen7_null_state; |
39 | case 8: |
39 | case 8: |
40 | return &gen8_null_state; |
40 | return &gen8_null_state; |
41 | case 9: |
41 | case 9: |
42 | return &gen9_null_state; |
42 | return &gen9_null_state; |
43 | } |
43 | } |
44 | 44 | ||
45 | return NULL; |
45 | return NULL; |
46 | } |
46 | } |
47 | 47 | ||
48 | static int render_state_init(struct render_state *so, struct drm_device *dev) |
48 | static int render_state_init(struct render_state *so, struct drm_device *dev) |
49 | { |
49 | { |
50 | int ret; |
50 | int ret; |
51 | 51 | ||
52 | so->gen = INTEL_INFO(dev)->gen; |
52 | so->gen = INTEL_INFO(dev)->gen; |
53 | so->rodata = render_state_get_rodata(dev, so->gen); |
53 | so->rodata = render_state_get_rodata(dev, so->gen); |
54 | if (so->rodata == NULL) |
54 | if (so->rodata == NULL) |
55 | return 0; |
55 | return 0; |
56 | 56 | ||
57 | if (so->rodata->batch_items * 4 > 4096) |
57 | if (so->rodata->batch_items * 4 > 4096) |
58 | return -EINVAL; |
58 | return -EINVAL; |
59 | 59 | ||
60 | so->obj = i915_gem_alloc_object(dev, 4096); |
60 | so->obj = i915_gem_alloc_object(dev, 4096); |
61 | if (so->obj == NULL) |
61 | if (so->obj == NULL) |
62 | return -ENOMEM; |
62 | return -ENOMEM; |
63 | 63 | ||
64 | ret = i915_gem_obj_ggtt_pin(so->obj, 4096, 0); |
64 | ret = i915_gem_obj_ggtt_pin(so->obj, 4096, 0); |
65 | if (ret) |
65 | if (ret) |
66 | goto free_gem; |
66 | goto free_gem; |
67 | 67 | ||
68 | so->ggtt_offset = i915_gem_obj_ggtt_offset(so->obj); |
68 | so->ggtt_offset = i915_gem_obj_ggtt_offset(so->obj); |
69 | return 0; |
69 | return 0; |
70 | 70 | ||
71 | free_gem: |
71 | free_gem: |
72 | drm_gem_object_unreference(&so->obj->base); |
72 | drm_gem_object_unreference(&so->obj->base); |
73 | return ret; |
73 | return ret; |
74 | } |
74 | } |
75 | 75 | ||
76 | /* |
76 | /* |
77 | * Macro to add commands to auxiliary batch. |
77 | * Macro to add commands to auxiliary batch. |
78 | * This macro only checks for page overflow before inserting the commands, |
78 | * This macro only checks for page overflow before inserting the commands, |
79 | * this is sufficient as the null state generator makes the final batch |
79 | * this is sufficient as the null state generator makes the final batch |
80 | * with two passes to build command and state separately. At this point |
80 | * with two passes to build command and state separately. At this point |
81 | * the size of both are known and it compacts them by relocating the state |
81 | * the size of both are known and it compacts them by relocating the state |
82 | * right after the commands taking care of aligment so we should sufficient |
82 | * right after the commands taking care of aligment so we should sufficient |
83 | * space below them for adding new commands. |
83 | * space below them for adding new commands. |
84 | */ |
84 | */ |
85 | #define OUT_BATCH(batch, i, val) \ |
85 | #define OUT_BATCH(batch, i, val) \ |
86 | do { \ |
86 | do { \ |
87 | if (WARN_ON((i) >= PAGE_SIZE / sizeof(u32))) { \ |
87 | if (WARN_ON((i) >= PAGE_SIZE / sizeof(u32))) { \ |
88 | ret = -ENOSPC; \ |
88 | ret = -ENOSPC; \ |
89 | goto err_out; \ |
89 | goto err_out; \ |
90 | } \ |
90 | } \ |
91 | (batch)[(i)++] = (val); \ |
91 | (batch)[(i)++] = (val); \ |
92 | } while(0) |
92 | } while(0) |
93 | 93 | ||
94 | static int render_state_setup(struct render_state *so) |
94 | static int render_state_setup(struct render_state *so) |
95 | { |
95 | { |
96 | const struct intel_renderstate_rodata *rodata = so->rodata; |
96 | const struct intel_renderstate_rodata *rodata = so->rodata; |
97 | unsigned int i = 0, reloc_index = 0; |
97 | unsigned int i = 0, reloc_index = 0; |
98 | struct page *page; |
98 | struct page *page; |
99 | u32 *d; |
99 | u32 *d; |
100 | int ret; |
100 | int ret; |
101 | 101 | ||
102 | ret = i915_gem_object_set_to_cpu_domain(so->obj, true); |
102 | ret = i915_gem_object_set_to_cpu_domain(so->obj, true); |
103 | if (ret) |
103 | if (ret) |
104 | return ret; |
104 | return ret; |
105 | 105 | ||
106 | page = sg_page(so->obj->pages->sgl); |
106 | page = i915_gem_object_get_dirty_page(so->obj, 0); |
107 | d = kmap(page); |
107 | d = kmap(page); |
108 | 108 | ||
109 | while (i < rodata->batch_items) { |
109 | while (i < rodata->batch_items) { |
110 | u32 s = rodata->batch[i]; |
110 | u32 s = rodata->batch[i]; |
111 | 111 | ||
112 | if (i * 4 == rodata->reloc[reloc_index]) { |
112 | if (i * 4 == rodata->reloc[reloc_index]) { |
113 | u64 r = s + so->ggtt_offset; |
113 | u64 r = s + so->ggtt_offset; |
114 | s = lower_32_bits(r); |
114 | s = lower_32_bits(r); |
115 | if (so->gen >= 8) { |
115 | if (so->gen >= 8) { |
116 | if (i + 1 >= rodata->batch_items || |
116 | if (i + 1 >= rodata->batch_items || |
117 | rodata->batch[i + 1] != 0) { |
117 | rodata->batch[i + 1] != 0) { |
118 | ret = -EINVAL; |
118 | ret = -EINVAL; |
119 | goto err_out; |
119 | goto err_out; |
120 | } |
120 | } |
121 | 121 | ||
122 | d[i++] = s; |
122 | d[i++] = s; |
123 | s = upper_32_bits(r); |
123 | s = upper_32_bits(r); |
124 | } |
124 | } |
125 | 125 | ||
126 | reloc_index++; |
126 | reloc_index++; |
127 | } |
127 | } |
128 | 128 | ||
129 | d[i++] = s; |
129 | d[i++] = s; |
130 | } |
130 | } |
131 | 131 | ||
132 | while (i % CACHELINE_DWORDS) |
132 | while (i % CACHELINE_DWORDS) |
133 | OUT_BATCH(d, i, MI_NOOP); |
133 | OUT_BATCH(d, i, MI_NOOP); |
134 | 134 | ||
135 | so->aux_batch_offset = i * sizeof(u32); |
135 | so->aux_batch_offset = i * sizeof(u32); |
136 | 136 | ||
137 | OUT_BATCH(d, i, MI_BATCH_BUFFER_END); |
137 | OUT_BATCH(d, i, MI_BATCH_BUFFER_END); |
138 | so->aux_batch_size = (i * sizeof(u32)) - so->aux_batch_offset; |
138 | so->aux_batch_size = (i * sizeof(u32)) - so->aux_batch_offset; |
139 | 139 | ||
140 | /* |
140 | /* |
141 | * Since we are sending length, we need to strictly conform to |
141 | * Since we are sending length, we need to strictly conform to |
142 | * all requirements. For Gen2 this must be a multiple of 8. |
142 | * all requirements. For Gen2 this must be a multiple of 8. |
143 | */ |
143 | */ |
144 | so->aux_batch_size = ALIGN(so->aux_batch_size, 8); |
144 | so->aux_batch_size = ALIGN(so->aux_batch_size, 8); |
145 | 145 | ||
146 | kunmap(page); |
146 | kunmap(page); |
147 | 147 | ||
148 | ret = i915_gem_object_set_to_gtt_domain(so->obj, false); |
148 | ret = i915_gem_object_set_to_gtt_domain(so->obj, false); |
149 | if (ret) |
149 | if (ret) |
150 | return ret; |
150 | return ret; |
151 | 151 | ||
152 | if (rodata->reloc[reloc_index] != -1) { |
152 | if (rodata->reloc[reloc_index] != -1) { |
153 | DRM_ERROR("only %d relocs resolved\n", reloc_index); |
153 | DRM_ERROR("only %d relocs resolved\n", reloc_index); |
154 | return -EINVAL; |
154 | return -EINVAL; |
155 | } |
155 | } |
156 | 156 | ||
157 | return 0; |
157 | return 0; |
158 | 158 | ||
159 | err_out: |
159 | err_out: |
160 | kunmap(page); |
160 | kunmap(page); |
161 | return ret; |
161 | return ret; |
162 | } |
162 | } |
163 | 163 | ||
164 | #undef OUT_BATCH |
164 | #undef OUT_BATCH |
165 | 165 | ||
166 | void i915_gem_render_state_fini(struct render_state *so) |
166 | void i915_gem_render_state_fini(struct render_state *so) |
167 | { |
167 | { |
168 | i915_gem_object_ggtt_unpin(so->obj); |
168 | i915_gem_object_ggtt_unpin(so->obj); |
169 | drm_gem_object_unreference(&so->obj->base); |
169 | drm_gem_object_unreference(&so->obj->base); |
170 | } |
170 | } |
171 | 171 | ||
172 | int i915_gem_render_state_prepare(struct intel_engine_cs *ring, |
172 | int i915_gem_render_state_prepare(struct intel_engine_cs *ring, |
173 | struct render_state *so) |
173 | struct render_state *so) |
174 | { |
174 | { |
175 | int ret; |
175 | int ret; |
176 | 176 | ||
177 | if (WARN_ON(ring->id != RCS)) |
177 | if (WARN_ON(ring->id != RCS)) |
178 | return -ENOENT; |
178 | return -ENOENT; |
179 | 179 | ||
180 | ret = render_state_init(so, ring->dev); |
180 | ret = render_state_init(so, ring->dev); |
181 | if (ret) |
181 | if (ret) |
182 | return ret; |
182 | return ret; |
183 | 183 | ||
184 | if (so->rodata == NULL) |
184 | if (so->rodata == NULL) |
185 | return 0; |
185 | return 0; |
186 | 186 | ||
187 | ret = render_state_setup(so); |
187 | ret = render_state_setup(so); |
188 | if (ret) { |
188 | if (ret) { |
189 | i915_gem_render_state_fini(so); |
189 | i915_gem_render_state_fini(so); |
190 | return ret; |
190 | return ret; |
191 | } |
191 | } |
192 | 192 | ||
193 | return 0; |
193 | return 0; |
194 | } |
194 | } |
195 | 195 | ||
196 | int i915_gem_render_state_init(struct drm_i915_gem_request *req) |
196 | int i915_gem_render_state_init(struct drm_i915_gem_request *req) |
197 | { |
197 | { |
198 | struct render_state so; |
198 | struct render_state so; |
199 | int ret; |
199 | int ret; |
200 | 200 | ||
201 | ret = i915_gem_render_state_prepare(req->ring, &so); |
201 | ret = i915_gem_render_state_prepare(req->ring, &so); |
202 | if (ret) |
202 | if (ret) |
203 | return ret; |
203 | return ret; |
204 | 204 | ||
205 | if (so.rodata == NULL) |
205 | if (so.rodata == NULL) |
206 | return 0; |
206 | return 0; |
207 | 207 | ||
208 | ret = req->ring->dispatch_execbuffer(req, so.ggtt_offset, |
208 | ret = req->ring->dispatch_execbuffer(req, so.ggtt_offset, |
209 | so.rodata->batch_items * 4, |
209 | so.rodata->batch_items * 4, |
210 | I915_DISPATCH_SECURE); |
210 | I915_DISPATCH_SECURE); |
211 | if (ret) |
211 | if (ret) |
212 | goto out; |
212 | goto out; |
213 | 213 | ||
214 | if (so.aux_batch_size > 8) { |
214 | if (so.aux_batch_size > 8) { |
215 | ret = req->ring->dispatch_execbuffer(req, |
215 | ret = req->ring->dispatch_execbuffer(req, |
216 | (so.ggtt_offset + |
216 | (so.ggtt_offset + |
217 | so.aux_batch_offset), |
217 | so.aux_batch_offset), |
218 | so.aux_batch_size, |
218 | so.aux_batch_size, |
219 | I915_DISPATCH_SECURE); |
219 | I915_DISPATCH_SECURE); |
220 | if (ret) |
220 | if (ret) |
221 | goto out; |
221 | goto out; |
222 | } |
222 | } |
223 | 223 | ||
224 | i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), req); |
224 | i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), req); |
225 | 225 | ||
226 | out: |
226 | out: |
227 | i915_gem_render_state_fini(&so); |
227 | i915_gem_render_state_fini(&so); |
228 | return ret; |
228 | return ret; |
229 | }> |
229 | }> |