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
|
/*******************************************************************************
Copyright (c) 2016 NVIDIA Corporation
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to
deal in the Software without restriction, including without limitation the
rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
sell copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
DEALINGS IN THE SOFTWARE.
*******************************************************************************/
#include "uvm_tlb_batch.h"
#include "uvm_hal.h"
void uvm_tlb_batch_begin(uvm_page_tree_t *tree, uvm_tlb_batch_t *batch)
{
memset(batch, 0, sizeof(*batch));
batch->tree = tree;
}
static NvU32 smallest_page_size(NvU32 page_sizes)
{
UVM_ASSERT(page_sizes != 0);
return 1u << __ffs(page_sizes);
}
static NvU32 biggest_page_size(NvU32 page_sizes)
{
UVM_ASSERT(page_sizes != 0);
return 1u << __fls(page_sizes);
}
static void tlb_batch_flush_invalidate_per_va(uvm_tlb_batch_t *batch, uvm_push_t *push)
{
uvm_page_tree_t *tree = batch->tree;
uvm_gpu_phys_address_t pdb_addr = uvm_page_tree_pdb(tree)->addr;
uvm_membar_t membar = UVM_MEMBAR_NONE;
NvU32 i;
for (i = 0; i < batch->count; ++i) {
uvm_tlb_batch_range_t *entry = &batch->ranges[i];
NvU32 min_page_size = smallest_page_size(entry->page_sizes);
NvU32 max_page_size = biggest_page_size(entry->page_sizes);
// Use the depth of the max page size as it's the broadest
NvU32 depth = tree->hal->page_table_depth(max_page_size);
UVM_ASSERT(hweight32(entry->page_sizes) > 0);
// Do the required membar only after the last invalidate
if (i == batch->count - 1)
membar = batch->membar;
// Use the min page size for the targeted VA invalidate as each page
// needs to be invalidated separately.
tree->gpu->parent->host_hal->tlb_invalidate_va(push,
pdb_addr,
depth,
entry->start,
entry->size,
min_page_size,
membar);
}
}
static void tlb_batch_flush_invalidate_all(uvm_tlb_batch_t *batch, uvm_push_t *push)
{
uvm_page_tree_t *tree = batch->tree;
uvm_gpu_t *gpu = tree->gpu;
NvU32 page_table_depth = tree->hal->page_table_depth(batch->biggest_page_size);
gpu->parent->host_hal->tlb_invalidate_all(push, uvm_page_tree_pdb(tree)->addr, page_table_depth, batch->membar);
}
static bool tlb_batch_should_invalidate_all(uvm_tlb_batch_t *batch)
{
if (!batch->tree->gpu->parent->tlb_batch.va_invalidate_supported)
return true;
if (batch->count > UVM_TLB_BATCH_MAX_ENTRIES)
return true;
if (batch->tree->gpu->parent->tlb_batch.va_range_invalidate_supported)
return batch->total_ranges > batch->tree->gpu->parent->tlb_batch.max_ranges;
return batch->total_pages > batch->tree->gpu->parent->tlb_batch.max_pages;
}
void uvm_tlb_batch_end(uvm_tlb_batch_t *batch, uvm_push_t *push, uvm_membar_t tlb_membar)
{
if (batch->count == 0)
return;
batch->membar = uvm_membar_max(tlb_membar, batch->membar);
if (tlb_batch_should_invalidate_all(batch))
tlb_batch_flush_invalidate_all(batch, push);
else
tlb_batch_flush_invalidate_per_va(batch, push);
}
void uvm_tlb_batch_invalidate(uvm_tlb_batch_t *batch, NvU64 start, NvU64 size, NvU32 page_sizes, uvm_membar_t tlb_membar)
{
uvm_tlb_batch_range_t *new_entry;
batch->membar = uvm_membar_max(tlb_membar, batch->membar);
++batch->count;
if (batch->tree->gpu->parent->tlb_batch.va_range_invalidate_supported)
batch->total_ranges++;
else
batch->total_pages += uvm_div_pow2_64(size, smallest_page_size(page_sizes));
batch->biggest_page_size = max(batch->biggest_page_size, biggest_page_size(page_sizes));
if (tlb_batch_should_invalidate_all(batch))
return;
new_entry = &batch->ranges[batch->count - 1];
new_entry->start = start;
new_entry->size = size;
new_entry->page_sizes = page_sizes;
}
|
