blob: fa586dbaa3c40b91f1281f888ef09f13704b1ee7 (
plain) (
blame)
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
|
/*
* Copyright (c) 2016-2017, NVIDIA CORPORATION. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#ifndef PAGE_ALLOCATOR_PRIV_H
#define PAGE_ALLOCATOR_PRIV_H
#include <linux/list.h>
#include <linux/rbtree.h>
#include <nvgpu/allocator.h>
#include <nvgpu/kmem.h>
struct nvgpu_allocator;
/*
* This allocator implements the ability to do SLAB style allocation since the
* GPU has two page sizes available - 4k and 64k/128k. When the default
* granularity is the large page size (64k/128k) small allocations become very
* space inefficient. This is most notable in PDE and PTE blocks which are 4k
* in size.
*
* Thus we need the ability to suballocate in 64k pages. The way we do this for
* the GPU is as follows. We have several buckets for sub-64K allocations:
*
* B0 - 4k
* B1 - 8k
* B3 - 16k
* B4 - 32k
* B5 - 64k (for when large pages are 128k)
*
* When an allocation comes in for less than the large page size (from now on
* assumed to be 64k) the allocation is satisfied by one of the buckets.
*/
struct page_alloc_slab {
struct list_head empty;
struct list_head partial;
struct list_head full;
int nr_empty;
int nr_partial;
int nr_full;
u32 slab_size;
};
enum slab_page_state {
SP_EMPTY,
SP_PARTIAL,
SP_FULL,
SP_NONE
};
struct page_alloc_slab_page {
unsigned long bitmap;
u64 page_addr;
u32 slab_size;
u32 nr_objects;
u32 nr_objects_alloced;
enum slab_page_state state;
struct page_alloc_slab *owner;
struct list_head list_entry;
};
struct page_alloc_chunk {
struct list_head list_entry;
u64 base;
u64 length;
};
/*
* Struct to handle internal management of page allocation. It holds a list
* of the chunks of pages that make up the overall allocation - much like a
* scatter gather table.
*/
struct nvgpu_page_alloc {
struct list_head alloc_chunks;
int nr_chunks;
u64 length;
/*
* Only useful for the RB tree - since the alloc may have discontiguous
* pages the base is essentially irrelevant except for the fact that it
* is guarenteed to be unique.
*/
u64 base;
struct rb_node tree_entry;
/*
* Set if this is a slab alloc. Points back to the slab page that owns
* this particular allocation. nr_chunks will always be 1 if this is
* set.
*/
struct page_alloc_slab_page *slab_page;
};
struct nvgpu_page_allocator {
struct nvgpu_allocator *owner; /* Owner of this allocator. */
/*
* Use a buddy allocator to manage the allocation of the underlying
* pages. This lets us abstract the discontiguous allocation handling
* out of the annoyingly complicated buddy allocator.
*/
struct nvgpu_allocator source_allocator;
/*
* Page params.
*/
u64 base;
u64 length;
u64 page_size;
u32 page_shift;
struct rb_root allocs; /* Outstanding allocations. */
struct page_alloc_slab *slabs;
int nr_slabs;
struct nvgpu_kmem_cache *alloc_cache;
struct nvgpu_kmem_cache *chunk_cache;
struct nvgpu_kmem_cache *slab_page_cache;
u64 flags;
/*
* Stat tracking.
*/
u64 nr_allocs;
u64 nr_frees;
u64 nr_fixed_allocs;
u64 nr_fixed_frees;
u64 nr_slab_allocs;
u64 nr_slab_frees;
u64 pages_alloced;
u64 pages_freed;
};
static inline struct nvgpu_page_allocator *page_allocator(
struct nvgpu_allocator *a)
{
return (struct nvgpu_page_allocator *)(a)->priv;
}
static inline struct nvgpu_allocator *palloc_owner(
struct nvgpu_page_allocator *a)
{
return a->owner;
}
#endif
|
