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
|
/*
* Virtualized GPU Memory Management
*
* Copyright (c) 2015-2017, 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.
*/
#include "vgpu/vgpu.h"
#include "vgpu_mm_gp10b.h"
#include "gk20a/mm_gk20a.h"
#include <nvgpu/bug.h>
static int vgpu_gp10b_init_mm_setup_hw(struct gk20a *g)
{
g->mm.bypass_smmu = true;
g->mm.disable_bigpage = true;
return 0;
}
static inline int add_mem_desc(struct tegra_vgpu_mem_desc *mem_desc,
u64 addr, u64 size, size_t *oob_size)
{
if (*oob_size < sizeof(*mem_desc))
return -ENOMEM;
mem_desc->addr = addr;
mem_desc->length = size;
*oob_size -= sizeof(*mem_desc);
return 0;
}
static u64 vgpu_gp10b_locked_gmmu_map(struct vm_gk20a *vm,
u64 map_offset,
struct nvgpu_sgt *sgt,
u64 buffer_offset,
u64 size,
int pgsz_idx,
u8 kind_v,
u32 ctag_offset,
u32 flags,
int rw_flag,
bool clear_ctags,
bool sparse,
bool priv,
struct vm_gk20a_mapping_batch *batch,
enum nvgpu_aperture aperture)
{
int err = 0;
struct gk20a *g = gk20a_from_vm(vm);
struct tegra_vgpu_cmd_msg msg;
struct tegra_vgpu_as_map_ex_params *p = &msg.params.as_map_ex;
struct tegra_vgpu_mem_desc *mem_desc;
u32 page_size = vm->gmmu_page_sizes[pgsz_idx];
u64 buffer_size = PAGE_ALIGN(size);
u64 space_to_skip = buffer_offset;
u32 mem_desc_count = 0, i;
void *handle = NULL;
size_t oob_size;
u8 prot;
void *sgl;
gk20a_dbg_fn("");
/* FIXME: add support for sparse mappings */
if (WARN_ON(!sgt) || WARN_ON(!g->mm.bypass_smmu))
return 0;
if (space_to_skip & (page_size - 1))
return 0;
memset(&msg, 0, sizeof(msg));
/* Allocate (or validate when map_offset != 0) the virtual address. */
if (!map_offset) {
map_offset = __nvgpu_vm_alloc_va(vm, size, pgsz_idx);
if (!map_offset) {
nvgpu_err(g, "failed to allocate va space");
err = -ENOMEM;
goto fail;
}
}
handle = tegra_gr_comm_oob_get_ptr(TEGRA_GR_COMM_CTX_CLIENT,
tegra_gr_comm_get_server_vmid(),
TEGRA_VGPU_QUEUE_CMD,
(void **)&mem_desc, &oob_size);
if (!handle) {
err = -EINVAL;
goto fail;
}
sgl = sgt->sgl;
while (sgl) {
u64 phys_addr;
u64 chunk_length;
/*
* Cut out sgl ents for space_to_skip.
*/
if (space_to_skip &&
space_to_skip >= nvgpu_sgt_get_length(sgt, sgl)) {
space_to_skip -= nvgpu_sgt_get_length(sgt, sgl);
sgl = nvgpu_sgt_get_next(sgt, sgl);
continue;
}
phys_addr = nvgpu_sgt_get_phys(sgt, sgl) + space_to_skip;
chunk_length = min(size,
nvgpu_sgt_get_length(sgt, sgl) - space_to_skip);
if (add_mem_desc(&mem_desc[mem_desc_count++], phys_addr,
chunk_length, &oob_size)) {
err = -ENOMEM;
goto fail;
}
space_to_skip = 0;
size -= chunk_length;
sgl = nvgpu_sgt_get_next(sgt, sgl);
if (size == 0)
break;
}
if (rw_flag == gk20a_mem_flag_read_only)
prot = TEGRA_VGPU_MAP_PROT_READ_ONLY;
else if (rw_flag == gk20a_mem_flag_write_only)
prot = TEGRA_VGPU_MAP_PROT_WRITE_ONLY;
else
prot = TEGRA_VGPU_MAP_PROT_NONE;
if (pgsz_idx == gmmu_page_size_kernel) {
if (page_size == vm->gmmu_page_sizes[gmmu_page_size_small]) {
pgsz_idx = gmmu_page_size_small;
} else if (page_size ==
vm->gmmu_page_sizes[gmmu_page_size_big]) {
pgsz_idx = gmmu_page_size_big;
} else {
nvgpu_err(g, "invalid kernel page size %d",
page_size);
goto fail;
}
}
msg.cmd = TEGRA_VGPU_CMD_AS_MAP_EX;
msg.handle = vgpu_get_handle(g);
p->handle = vm->handle;
p->gpu_va = map_offset;
p->size = buffer_size;
p->mem_desc_count = mem_desc_count;
p->pgsz_idx = pgsz_idx;
p->iova = 0;
p->kind = kind_v;
p->cacheable =
(flags & NVGPU_MAP_BUFFER_FLAGS_CACHEABLE_TRUE) ? 1 : 0;
p->prot = prot;
p->ctag_offset = ctag_offset;
p->clear_ctags = clear_ctags;
err = vgpu_comm_sendrecv(&msg, sizeof(msg), sizeof(msg));
if (err || msg.ret)
goto fail;
/* TLB invalidate handled on server side */
tegra_gr_comm_oob_put_ptr(handle);
return map_offset;
fail:
if (handle)
tegra_gr_comm_oob_put_ptr(handle);
nvgpu_err(g, "Failed: err=%d, msg.ret=%d", err, msg.ret);
nvgpu_err(g,
" Map: %-5s GPU virt %#-12llx +%#-9llx "
"phys offset: %#-4llx; pgsz: %3dkb perm=%-2s | "
"kind=%#02x APT=%-6s",
vm->name, map_offset, buffer_size, buffer_offset,
vm->gmmu_page_sizes[pgsz_idx] >> 10,
nvgpu_gmmu_perm_str(rw_flag),
kind_v, "SYSMEM");
for (i = 0; i < mem_desc_count; i++)
nvgpu_err(g, " > 0x%010llx + 0x%llx",
mem_desc[i].addr, mem_desc[i].length);
return 0;
}
void vgpu_gp10b_init_mm_ops(struct gpu_ops *gops)
{
gk20a_dbg_fn("");
gops->mm.gmmu_map = vgpu_gp10b_locked_gmmu_map;
gops->mm.init_mm_setup_hw = vgpu_gp10b_init_mm_setup_hw;
/* FIXME: add support for sparse mappings */
gops->mm.support_sparse = NULL;
}
|
