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
|
From 6a4fd5e216c832b860ff80d5eacef0693fb8ad2f Mon Sep 17 00:00:00 2001
From: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Date: Tue, 10 Oct 2017 16:50:16 -0400
Subject: [PATCH 1999/4131] drm/amdgpu: Refactor amdgpu_cs_ib_vm_chunk and
amdgpu_cs_ib_fill.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This enables old fence waiting before reservation lock is aquired
which in turn is part of a bigger solution to deadlock happening
when gpu reset with VRAM recovery accures during intensive rendering.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 112 ++++++++++++++++++---------------
1 file changed, 61 insertions(+), 51 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 7c6f574..5133b7a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -886,15 +886,60 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,
struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
struct amdgpu_vm *vm = &fpriv->vm;
struct amdgpu_ring *ring = p->job->ring;
- int i, r;
+ int i, j, r;
+
+ for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) {
+
+ struct amdgpu_cs_chunk *chunk;
+ struct amdgpu_ib *ib;
+ struct drm_amdgpu_cs_chunk_ib *chunk_ib;
+
+ chunk = &p->chunks[i];
+ ib = &p->job->ibs[j];
+ chunk_ib = (struct drm_amdgpu_cs_chunk_ib *)chunk->kdata;
+
+ if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB)
+ continue;
+
+ if (p->job->ring->funcs->parse_cs) {
+ struct amdgpu_bo_va_mapping *m;
+ struct amdgpu_bo *aobj = NULL;
+ uint64_t offset;
+ uint8_t *kptr;
+
+ r = amdgpu_cs_find_mapping(p, chunk_ib->va_start,
+ &aobj, &m);
+ if (r) {
+ DRM_ERROR("IB va_start is invalid\n");
+ return r;
+ }
- /* Only for UVD/VCE VM emulation */
- if (ring->funcs->parse_cs) {
- for (i = 0; i < p->job->num_ibs; i++) {
- r = amdgpu_ring_parse_cs(ring, p, i);
+ if ((chunk_ib->va_start + chunk_ib->ib_bytes) >
+ (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
+ DRM_ERROR("IB va_start+ib_bytes is invalid\n");
+ return -EINVAL;
+ }
+
+ /* the IB should be reserved at this point */
+ r = amdgpu_bo_kmap(aobj, (void **)&kptr);
+ if (r) {
+ return r;
+ }
+
+ offset = m->start * AMDGPU_GPU_PAGE_SIZE;
+ kptr += chunk_ib->va_start - offset;
+
+ memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
+ amdgpu_bo_kunmap(aobj);
+
+ /* Only for UVD/VCE VM emulation */
+ r = amdgpu_ring_parse_cs(ring, p, j);
if (r)
return r;
+
}
+
+ j++;
}
if (p->job->vm) {
@@ -960,54 +1005,18 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev,
parser->job->ring = ring;
- if (ring->funcs->parse_cs) {
- struct amdgpu_bo_va_mapping *m;
- struct amdgpu_bo *aobj = NULL;
- uint64_t offset;
- uint8_t *kptr;
-
- r = amdgpu_cs_find_mapping(parser, chunk_ib->va_start,
- &aobj, &m);
- if (r) {
- DRM_ERROR("IB va_start is invalid\n");
- return r;
- }
-
- if ((chunk_ib->va_start + chunk_ib->ib_bytes) >
- (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
- DRM_ERROR("IB va_start+ib_bytes is invalid\n");
- return -EINVAL;
- }
-
- /* the IB should be reserved at this point */
- r = amdgpu_bo_kmap(aobj, (void **)&kptr);
- if (r) {
- return r;
- }
-
- offset = m->start * AMDGPU_GPU_PAGE_SIZE;
- kptr += chunk_ib->va_start - offset;
-
- r = amdgpu_ib_get(adev, vm, chunk_ib->ib_bytes, ib);
- if (r) {
- DRM_ERROR("Failed to get ib !\n");
- return r;
- }
-
- memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
- amdgpu_bo_kunmap(aobj);
- } else {
- r = amdgpu_ib_get(adev, vm, 0, ib);
- if (r) {
- DRM_ERROR("Failed to get ib !\n");
- return r;
- }
-
+ r = amdgpu_ib_get(adev, vm,
+ ring->funcs->parse_cs ? chunk_ib->ib_bytes : 0,
+ ib);
+ if (r) {
+ DRM_ERROR("Failed to get ib !\n");
+ return r;
}
ib->gpu_addr = chunk_ib->va_start;
ib->length_dw = chunk_ib->ib_bytes / 4;
ib->flags = chunk_ib->flags;
+
j++;
}
@@ -1261,6 +1270,10 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
goto out;
}
+ r = amdgpu_cs_ib_fill(adev, &parser);
+ if (r)
+ goto out;
+
r = amdgpu_cs_parser_bos(&parser, data);
if (r) {
if (r == -ENOMEM)
@@ -1271,9 +1284,6 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
}
reserved_buffers = true;
- r = amdgpu_cs_ib_fill(adev, &parser);
- if (r)
- goto out;
r = amdgpu_cs_dependencies(adev, &parser);
if (r) {
--
2.7.4
|
