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
|
From 60d88de85ee6c3e91253719333071f4c15d864a1 Mon Sep 17 00:00:00 2001
From: Eric Huang <JinhuiEric.Huang@amd.com>
Date: Wed, 5 Sep 2018 11:46:14 -0400
Subject: [PATCH 5715/5725] drm/amdkfd: change system memory overcommit limit
It is to improve system limit by:
1. replacing userptrlimit with a total memory limit that
conunts TTM memory usage and userptr usage.
2. counting acc size for all BOs.
KFD-387
Change-Id: I30a356c5cb7517a7e7ad5425de4e79cfc3109594
Signed-off-by: Eric Huang <JinHuiEric.Huang@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 106 +++++++++++++----------
1 file changed, 59 insertions(+), 47 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index 8cd9f0d..15a64e8 100755
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -41,9 +41,9 @@
/* Impose limit on how much memory KFD can use */
static struct {
uint64_t max_system_mem_limit;
- uint64_t max_userptr_mem_limit;
+ uint64_t max_ttm_mem_limit;
int64_t system_mem_used;
- int64_t userptr_mem_used;
+ int64_t ttm_mem_used;
spinlock_t mem_limit_lock;
} kfd_mem_limit;
@@ -85,8 +85,8 @@ static bool check_if_add_bo_to_vm(struct amdgpu_vm *avm,
}
/* Set memory usage limits. Current, limits are
- * System (kernel) memory - 3/8th System RAM
- * Userptr memory - 3/4th System RAM
+ * System (TTM + userptr) memory - 3/4th System RAM
+ * TTM memory - 3/8th System RAM
*/
void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
{
@@ -98,48 +98,54 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
mem *= si.mem_unit;
spin_lock_init(&kfd_mem_limit.mem_limit_lock);
- kfd_mem_limit.max_system_mem_limit = (mem >> 1) - (mem >> 3);
- kfd_mem_limit.max_userptr_mem_limit = mem - (mem >> 2);
- pr_debug("Kernel memory limit %lluM, userptr limit %lluM\n",
+ kfd_mem_limit.max_system_mem_limit = (mem >> 1) + (mem >> 2);
+ kfd_mem_limit.max_ttm_mem_limit = (mem >> 1) - (mem >> 3);
+ pr_debug("Kernel memory limit %lluM, TTM limit %lluM\n",
(kfd_mem_limit.max_system_mem_limit >> 20),
- (kfd_mem_limit.max_userptr_mem_limit >> 20));
+ (kfd_mem_limit.max_ttm_mem_limit >> 20));
}
static int amdgpu_amdkfd_reserve_system_mem_limit(struct amdgpu_device *adev,
- uint64_t size, u32 domain)
+ uint64_t size, u32 domain, bool sg)
{
- size_t acc_size;
+ size_t acc_size, system_mem_needed, ttm_mem_needed;
int ret = 0;
acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size,
sizeof(struct amdgpu_bo));
spin_lock(&kfd_mem_limit.mem_limit_lock);
+
if (domain == AMDGPU_GEM_DOMAIN_GTT) {
- if (kfd_mem_limit.system_mem_used + (acc_size + size) >
- kfd_mem_limit.max_system_mem_limit) {
- ret = -ENOMEM;
- goto err_no_mem;
- }
- kfd_mem_limit.system_mem_used += (acc_size + size);
- } else if (domain == AMDGPU_GEM_DOMAIN_CPU) {
- if ((kfd_mem_limit.system_mem_used + acc_size >
- kfd_mem_limit.max_system_mem_limit) ||
- (kfd_mem_limit.userptr_mem_used + (size + acc_size) >
- kfd_mem_limit.max_userptr_mem_limit)) {
- ret = -ENOMEM;
- goto err_no_mem;
- }
- kfd_mem_limit.system_mem_used += acc_size;
- kfd_mem_limit.userptr_mem_used += size;
+ /* TTM GTT memory */
+ system_mem_needed = acc_size + size;
+ ttm_mem_needed = acc_size + size;
+ } else if (domain == AMDGPU_GEM_DOMAIN_CPU && !sg) {
+ /* Userptr */
+ system_mem_needed = acc_size + size;
+ ttm_mem_needed = acc_size;
+ } else {
+ /* VRAM and SG */
+ system_mem_needed = acc_size;
+ ttm_mem_needed = acc_size;
}
-err_no_mem:
+
+ if ((kfd_mem_limit.system_mem_used + system_mem_needed >
+ kfd_mem_limit.max_system_mem_limit) ||
+ (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
+ kfd_mem_limit.max_ttm_mem_limit))
+ ret = -ENOMEM;
+ else {
+ kfd_mem_limit.system_mem_used += system_mem_needed;
+ kfd_mem_limit.ttm_mem_used += ttm_mem_needed;
+ }
+
spin_unlock(&kfd_mem_limit.mem_limit_lock);
return ret;
}
static void unreserve_system_mem_limit(struct amdgpu_device *adev,
- uint64_t size, u32 domain)
+ uint64_t size, u32 domain, bool sg)
{
size_t acc_size;
@@ -149,14 +155,18 @@ static void unreserve_system_mem_limit(struct amdgpu_device *adev,
spin_lock(&kfd_mem_limit.mem_limit_lock);
if (domain == AMDGPU_GEM_DOMAIN_GTT) {
kfd_mem_limit.system_mem_used -= (acc_size + size);
- } else if (domain == AMDGPU_GEM_DOMAIN_CPU) {
+ kfd_mem_limit.ttm_mem_used -= (acc_size + size);
+ } else if (domain == AMDGPU_GEM_DOMAIN_CPU && !sg) {
+ kfd_mem_limit.system_mem_used -= (acc_size + size);
+ kfd_mem_limit.ttm_mem_used -= acc_size;
+ } else {
kfd_mem_limit.system_mem_used -= acc_size;
- kfd_mem_limit.userptr_mem_used -= size;
+ kfd_mem_limit.ttm_mem_used -= acc_size;
}
WARN_ONCE(kfd_mem_limit.system_mem_used < 0,
"kfd system memory accounting unbalanced");
- WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0,
- "kfd userptr memory accounting unbalanced");
+ WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0,
+ "kfd TTM memory accounting unbalanced");
spin_unlock(&kfd_mem_limit.mem_limit_lock);
}
@@ -166,17 +176,22 @@ void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo)
spin_lock(&kfd_mem_limit.mem_limit_lock);
if (bo->flags & AMDGPU_AMDKFD_USERPTR_BO) {
- kfd_mem_limit.system_mem_used -= bo->tbo.acc_size;
- kfd_mem_limit.userptr_mem_used -= amdgpu_bo_size(bo);
- } else if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT &&
- !bo->tbo.sg) {
kfd_mem_limit.system_mem_used -=
(bo->tbo.acc_size + amdgpu_bo_size(bo));
+ kfd_mem_limit.ttm_mem_used -= bo->tbo.acc_size;
+ } else if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT) {
+ kfd_mem_limit.system_mem_used -=
+ (bo->tbo.acc_size + amdgpu_bo_size(bo));
+ kfd_mem_limit.ttm_mem_used -=
+ (bo->tbo.acc_size + amdgpu_bo_size(bo));
+ } else {
+ kfd_mem_limit.system_mem_used -= bo->tbo.acc_size;
+ kfd_mem_limit.ttm_mem_used -= bo->tbo.acc_size;
}
WARN_ONCE(kfd_mem_limit.system_mem_used < 0,
"kfd system memory accounting unbalanced");
- WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0,
- "kfd userptr memory accounting unbalanced");
+ WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0,
+ "kfd TTM memory accounting unbalanced");
spin_unlock(&kfd_mem_limit.mem_limit_lock);
}
@@ -1285,13 +1300,11 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
amdgpu_sync_create(&(*mem)->sync);
- if (!sg) {
- ret = amdgpu_amdkfd_reserve_system_mem_limit(adev, size,
- alloc_domain);
- if (ret) {
- pr_debug("Insufficient system memory\n");
- goto err_reserve_limit;
- }
+ ret = amdgpu_amdkfd_reserve_system_mem_limit(adev, size,
+ alloc_domain, !!sg);
+ if (ret) {
+ pr_debug("Insufficient system memory\n");
+ goto err_reserve_limit;
}
pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s\n",
@@ -1346,8 +1359,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
/* Don't unreserve system mem limit twice */
goto err_reserve_limit;
err_bo_create:
- if (!sg)
- unreserve_system_mem_limit(adev, size, alloc_domain);
+ unreserve_system_mem_limit(adev, size, alloc_domain, !!sg);
err_reserve_limit:
kfree(*mem);
err:
--
2.7.4
|
