1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
|
From 7fc11959018f8ba2d92025679a72339da18a74ad Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@amd.com>
Date: Thu, 30 Jul 2015 11:53:42 +0200
Subject: [PATCH 0412/1050] drm/amdgpu: stop using addr to check for BO move v3
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
It is theoretically possible that a swapped out BO gets the
same GTT address, but different backing pages while being swapped in.
Instead just use another VA state to note updated areas.
Ported from not upstream yet radeon commit with the same name.
v2: fix some bugs in the original implementation found in the radeon code.
v3: squash in VCE/UVD fix
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 15 +++++---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 11 +++++-
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 67 ++++++++++++++++++++--------------
3 files changed, 59 insertions(+), 34 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 98b4760..5b8e1ae 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -539,14 +539,16 @@ struct amdgpu_bo_va_mapping {
struct amdgpu_bo_va {
/* protected by bo being reserved */
struct list_head bo_list;
- uint64_t addr;
struct amdgpu_fence *last_pt_update;
unsigned ref_count;
- /* protected by vm mutex */
- struct list_head mappings;
+ /* protected by vm mutex and spinlock */
struct list_head vm_status;
+ /* mappings for this bo_va */
+ struct list_head invalids;
+ struct list_head valids;
+
/* constant after initialization */
struct amdgpu_vm *vm;
struct amdgpu_bo *bo;
@@ -964,13 +966,16 @@ struct amdgpu_vm {
struct rb_root va;
- /* protecting invalidated and freed */
+ /* protecting invalidated */
spinlock_t status_lock;
/* BOs moved, but not yet updated in the PT */
struct list_head invalidated;
- /* BOs freed, but not yet updated in the PT */
+ /* BOs cleared in the PT because of a move */
+ struct list_head cleared;
+
+ /* BO mappings freed, but not yet updated in the PT */
struct list_head freed;
/* contains the page directory */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 40e85bf..fe81b46 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -1009,7 +1009,16 @@ amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
if (!reloc->bo_va)
continue;
- list_for_each_entry(mapping, &reloc->bo_va->mappings, list) {
+ list_for_each_entry(mapping, &reloc->bo_va->valids, list) {
+ if (mapping->it.start > addr ||
+ addr > mapping->it.last)
+ continue;
+
+ *bo = reloc->bo_va->bo;
+ return mapping;
+ }
+
+ list_for_each_entry(mapping, &reloc->bo_va->invalids, list) {
if (mapping->it.start > addr ||
addr > mapping->it.last)
continue;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 5624d44..f216632 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -951,21 +951,24 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
addr = 0;
}
- if (addr == bo_va->addr)
- return 0;
-
flags = amdgpu_ttm_tt_pte_flags(adev, bo_va->bo->tbo.ttm, mem);
- list_for_each_entry(mapping, &bo_va->mappings, list) {
+ spin_lock(&vm->status_lock);
+ if (!list_empty(&bo_va->vm_status))
+ list_splice_init(&bo_va->valids, &bo_va->invalids);
+ spin_unlock(&vm->status_lock);
+
+ list_for_each_entry(mapping, &bo_va->invalids, list) {
r = amdgpu_vm_bo_update_mapping(adev, vm, mapping, addr,
flags, &bo_va->last_pt_update);
if (r)
return r;
}
- bo_va->addr = addr;
spin_lock(&vm->status_lock);
list_del_init(&bo_va->vm_status);
+ if (!mem)
+ list_add(&bo_va->vm_status, &vm->cleared);
spin_unlock(&vm->status_lock);
return 0;
@@ -1065,10 +1068,10 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
}
bo_va->vm = vm;
bo_va->bo = bo;
- bo_va->addr = 0;
bo_va->ref_count = 1;
INIT_LIST_HEAD(&bo_va->bo_list);
- INIT_LIST_HEAD(&bo_va->mappings);
+ INIT_LIST_HEAD(&bo_va->valids);
+ INIT_LIST_HEAD(&bo_va->invalids);
INIT_LIST_HEAD(&bo_va->vm_status);
mutex_lock(&vm->mutex);
@@ -1157,12 +1160,10 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
mapping->offset = offset;
mapping->flags = flags;
- list_add(&mapping->list, &bo_va->mappings);
+ list_add(&mapping->list, &bo_va->invalids);
interval_tree_insert(&mapping->it, &vm->va);
trace_amdgpu_vm_bo_map(bo_va, mapping);
- bo_va->addr = 0;
-
/* Make sure the page tables are allocated */
saddr >>= amdgpu_vm_block_size;
eaddr >>= amdgpu_vm_block_size;
@@ -1243,17 +1244,27 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
{
struct amdgpu_bo_va_mapping *mapping;
struct amdgpu_vm *vm = bo_va->vm;
+ bool valid = true;
saddr /= AMDGPU_GPU_PAGE_SIZE;
- list_for_each_entry(mapping, &bo_va->mappings, list) {
+ list_for_each_entry(mapping, &bo_va->valids, list) {
if (mapping->it.start == saddr)
break;
}
- if (&mapping->list == &bo_va->mappings) {
- amdgpu_bo_unreserve(bo_va->bo);
- return -ENOENT;
+ if (&mapping->list == &bo_va->valids) {
+ valid = false;
+
+ list_for_each_entry(mapping, &bo_va->invalids, list) {
+ if (mapping->it.start == saddr)
+ break;
+ }
+
+ if (&mapping->list == &bo_va->invalids) {
+ amdgpu_bo_unreserve(bo_va->bo);
+ return -ENOENT;
+ }
}
mutex_lock(&vm->mutex);
@@ -1261,12 +1272,10 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
interval_tree_remove(&mapping->it, &vm->va);
trace_amdgpu_vm_bo_unmap(bo_va, mapping);
- if (bo_va->addr) {
- /* clear the old address */
+ if (valid)
list_add(&mapping->list, &vm->freed);
- } else {
+ else
kfree(mapping);
- }
mutex_unlock(&vm->mutex);
amdgpu_bo_unreserve(bo_va->bo);
@@ -1297,15 +1306,18 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,
list_del(&bo_va->vm_status);
spin_unlock(&vm->status_lock);
- list_for_each_entry_safe(mapping, next, &bo_va->mappings, list) {
+ list_for_each_entry_safe(mapping, next, &bo_va->valids, list) {
list_del(&mapping->list);
interval_tree_remove(&mapping->it, &vm->va);
trace_amdgpu_vm_bo_unmap(bo_va, mapping);
- if (bo_va->addr)
- list_add(&mapping->list, &vm->freed);
- else
- kfree(mapping);
+ list_add(&mapping->list, &vm->freed);
+ }
+ list_for_each_entry_safe(mapping, next, &bo_va->invalids, list) {
+ list_del(&mapping->list);
+ interval_tree_remove(&mapping->it, &vm->va);
+ kfree(mapping);
}
+
amdgpu_fence_unref(&bo_va->last_pt_update);
kfree(bo_va);
@@ -1327,12 +1339,10 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
struct amdgpu_bo_va *bo_va;
list_for_each_entry(bo_va, &bo->va, bo_list) {
- if (bo_va->addr) {
- spin_lock(&bo_va->vm->status_lock);
- list_del(&bo_va->vm_status);
+ spin_lock(&bo_va->vm->status_lock);
+ if (list_empty(&bo_va->vm_status))
list_add(&bo_va->vm_status, &bo_va->vm->invalidated);
- spin_unlock(&bo_va->vm->status_lock);
- }
+ spin_unlock(&bo_va->vm->status_lock);
}
}
@@ -1360,6 +1370,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
vm->va = RB_ROOT;
spin_lock_init(&vm->status_lock);
INIT_LIST_HEAD(&vm->invalidated);
+ INIT_LIST_HEAD(&vm->cleared);
INIT_LIST_HEAD(&vm->freed);
pd_size = amdgpu_vm_directory_size(adev);
--
1.9.1
|
