1 | /* |
2 | * Copyright 2011 Christian König. |
3 | * All Rights Reserved. |
4 | * |
5 | * Permission is hereby granted, free of charge, to any person obtaining a |
6 | * copy of this software and associated documentation files (the |
7 | * "Software"), to deal in the Software without restriction, including |
8 | * without limitation the rights to use, copy, modify, merge, publish, |
9 | * distribute, sub license, and/or sell copies of the Software, and to |
10 | * permit persons to whom the Software is furnished to do so, subject to |
11 | * the following conditions: |
12 | * |
13 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
14 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
15 | * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL |
16 | * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, |
17 | * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR |
18 | * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE |
19 | * USE OR OTHER DEALINGS IN THE SOFTWARE. |
20 | * |
21 | * The above copyright notice and this permission notice (including the |
22 | * next paragraph) shall be included in all copies or substantial portions |
23 | * of the Software. |
24 | * |
25 | */ |
26 | /* |
27 | * Authors: |
28 | * Christian König <deathsimple@vodafone.de> |
29 | */ |
30 | #include <drm/drmP.h> |
31 | #include "radeon.h" |
32 | #include "radeon_trace.h" |
33 | |
34 | int radeon_semaphore_create(struct radeon_device *rdev, |
35 | struct radeon_semaphore **semaphore) |
36 | { |
37 | uint32_t *cpu_addr; |
38 | int i, r; |
39 | |
40 | *semaphore = kmalloc(sizeof(struct radeon_semaphore), GFP_KERNEL); |
41 | if (*semaphore == NULL) { |
42 | return -ENOMEM; |
43 | } |
44 | r = radeon_sa_bo_new(rdev, &rdev->ring_tmp_bo, &(*semaphore)->sa_bo, |
45 | 8 * RADEON_NUM_SYNCS, 8); |
46 | if (r) { |
47 | kfree(*semaphore); |
48 | *semaphore = NULL; |
49 | return r; |
50 | } |
51 | (*semaphore)->waiters = 0; |
52 | (*semaphore)->gpu_addr = radeon_sa_bo_gpu_addr((*semaphore)->sa_bo); |
53 | |
54 | cpu_addr = radeon_sa_bo_cpu_addr((*semaphore)->sa_bo); |
55 | for (i = 0; i < RADEON_NUM_SYNCS; ++i) |
56 | cpu_addr[i] = 0; |
57 | |
58 | for (i = 0; i < RADEON_NUM_RINGS; ++i) |
59 | (*semaphore)->sync_to[i] = NULL; |
60 | |
61 | return 0; |
62 | } |
63 | |
64 | bool radeon_semaphore_emit_signal(struct radeon_device *rdev, int ridx, |
65 | struct radeon_semaphore *semaphore) |
66 | { |
67 | struct radeon_ring *ring = &rdev->ring[ridx]; |
68 | |
69 | trace_radeon_semaphore_signale(ridx, semaphore); |
70 | |
71 | if (radeon_semaphore_ring_emit(rdev, ridx, ring, semaphore, false)) { |
72 | --semaphore->waiters; |
73 | |
74 | /* for debugging lockup only, used by sysfs debug files */ |
75 | ring->last_semaphore_signal_addr = semaphore->gpu_addr; |
76 | return true; |
77 | } |
78 | return false; |
79 | } |
80 | |
81 | bool radeon_semaphore_emit_wait(struct radeon_device *rdev, int ridx, |
82 | struct radeon_semaphore *semaphore) |
83 | { |
84 | struct radeon_ring *ring = &rdev->ring[ridx]; |
85 | |
86 | trace_radeon_semaphore_wait(ridx, semaphore); |
87 | |
88 | if (radeon_semaphore_ring_emit(rdev, ridx, ring, semaphore, true)) { |
89 | ++semaphore->waiters; |
90 | |
91 | /* for debugging lockup only, used by sysfs debug files */ |
92 | ring->last_semaphore_wait_addr = semaphore->gpu_addr; |
93 | return true; |
94 | } |
95 | return false; |
96 | } |
97 | |
98 | /** |
99 | * radeon_semaphore_sync_to - use the semaphore to sync to a fence |
100 | * |
101 | * @semaphore: semaphore object to add fence to |
102 | * @fence: fence to sync to |
103 | * |
104 | * Sync to the fence using this semaphore object |
105 | */ |
106 | void radeon_semaphore_sync_to(struct radeon_semaphore *semaphore, |
107 | struct radeon_fence *fence) |
108 | { |
109 | struct radeon_fence *other; |
110 | |
111 | if (!fence) |
112 | return; |
113 | |
114 | other = semaphore->sync_to[fence->ring]; |
115 | semaphore->sync_to[fence->ring] = radeon_fence_later(fence, other); |
116 | } |
117 | |
118 | /** |
119 | * radeon_semaphore_sync_rings - sync ring to all registered fences |
120 | * |
121 | * @rdev: radeon_device pointer |
122 | * @semaphore: semaphore object to use for sync |
123 | * @ring: ring that needs sync |
124 | * |
125 | * Ensure that all registered fences are signaled before letting |
126 | * the ring continue. The caller must hold the ring lock. |
127 | */ |
128 | int radeon_semaphore_sync_rings(struct radeon_device *rdev, |
129 | struct radeon_semaphore *semaphore, |
130 | int ring) |
131 | { |
132 | unsigned count = 0; |
133 | int i, r; |
134 | |
135 | for (i = 0; i < RADEON_NUM_RINGS; ++i) { |
136 | struct radeon_fence *fence = semaphore->sync_to[i]; |
137 | |
138 | /* check if we really need to sync */ |
139 | if (!radeon_fence_need_sync(fence, ring)) |
140 | continue; |
141 | |
142 | /* prevent GPU deadlocks */ |
143 | if (!rdev->ring[i].ready) { |
144 | dev_err(rdev->dev, "Syncing to a disabled ring!" ); |
145 | return -EINVAL; |
146 | } |
147 | |
148 | if (++count > RADEON_NUM_SYNCS) { |
149 | /* not enough room, wait manually */ |
150 | r = radeon_fence_wait(fence, false); |
151 | if (r) |
152 | return r; |
153 | continue; |
154 | } |
155 | |
156 | /* allocate enough space for sync command */ |
157 | r = radeon_ring_alloc(rdev, &rdev->ring[i], 16); |
158 | if (r) { |
159 | return r; |
160 | } |
161 | |
162 | /* emit the signal semaphore */ |
163 | if (!radeon_semaphore_emit_signal(rdev, i, semaphore)) { |
164 | /* signaling wasn't successful wait manually */ |
165 | radeon_ring_undo(&rdev->ring[i]); |
166 | r = radeon_fence_wait(fence, false); |
167 | if (r) |
168 | return r; |
169 | continue; |
170 | } |
171 | |
172 | /* we assume caller has already allocated space on waiters ring */ |
173 | if (!radeon_semaphore_emit_wait(rdev, ring, semaphore)) { |
174 | /* waiting wasn't successful wait manually */ |
175 | radeon_ring_undo(&rdev->ring[i]); |
176 | r = radeon_fence_wait(fence, false); |
177 | if (r) |
178 | return r; |
179 | continue; |
180 | } |
181 | |
182 | radeon_ring_commit(rdev, &rdev->ring[i]); |
183 | radeon_fence_note_sync(fence, ring); |
184 | |
185 | semaphore->gpu_addr += 8; |
186 | } |
187 | |
188 | return 0; |
189 | } |
190 | |
191 | void radeon_semaphore_free(struct radeon_device *rdev, |
192 | struct radeon_semaphore **semaphore, |
193 | struct radeon_fence *fence) |
194 | { |
195 | if (semaphore == NULL || *semaphore == NULL) { |
196 | return; |
197 | } |
198 | if ((*semaphore)->waiters > 0) { |
199 | dev_err(rdev->dev, "semaphore %p has more waiters than signalers," |
200 | " hardware lockup imminent!\n" , *semaphore); |
201 | } |
202 | radeon_sa_bo_free(rdev, &(*semaphore)->sa_bo, fence); |
203 | kfree(*semaphore); |
204 | *semaphore = NULL; |
205 | } |
206 | |