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
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
|
// SPDX-License-Identifier: GPL-2.0
/* Copyright (c) 2019 HiSilicon Limited. */
#include <linux/dma-mapping.h>
#include <linux/module.h>
#include <linux/slab.h>
#include "qm.h"
#define HISI_ACC_SGL_SGE_NR_MIN 1
#define HISI_ACC_SGL_NR_MAX 256
#define HISI_ACC_SGL_ALIGN_SIZE 64
#define HISI_ACC_MEM_BLOCK_NR 5
struct acc_hw_sge {
dma_addr_t buf;
void *page_ctrl;
__le32 len;
__le32 pad;
__le32 pad0;
__le32 pad1;
};
/* use default sgl head size 64B */
struct hisi_acc_hw_sgl {
dma_addr_t next_dma;
__le16 entry_sum_in_chain;
__le16 entry_sum_in_sgl;
__le16 entry_length_in_sgl;
__le16 pad0;
__le64 pad1[5];
struct hisi_acc_hw_sgl *next;
struct acc_hw_sge sge_entries[];
} __aligned(1);
struct hisi_acc_sgl_pool {
struct mem_block {
struct hisi_acc_hw_sgl *sgl;
dma_addr_t sgl_dma;
size_t size;
} mem_block[HISI_ACC_MEM_BLOCK_NR];
u32 sgl_num_per_block;
u32 block_num;
u32 count;
u32 sge_nr;
size_t sgl_size;
};
/**
* hisi_acc_create_sgl_pool() - Create a hw sgl pool.
* @dev: The device which hw sgl pool belongs to.
* @count: Count of hisi_acc_hw_sgl in pool.
* @sge_nr: The count of sge in hw_sgl
*
* This function creates a hw sgl pool, after this user can get hw sgl memory
* from it.
*/
struct hisi_acc_sgl_pool *hisi_acc_create_sgl_pool(struct device *dev,
u32 count, u32 sge_nr)
{
u32 sgl_size, block_size, sgl_num_per_block, block_num, remain_sgl = 0;
struct hisi_acc_sgl_pool *pool;
struct mem_block *block;
u32 i, j;
if (!dev || !count || !sge_nr || sge_nr > HISI_ACC_SGL_SGE_NR_MAX)
return ERR_PTR(-EINVAL);
sgl_size = sizeof(struct acc_hw_sge) * sge_nr +
sizeof(struct hisi_acc_hw_sgl);
block_size = PAGE_SIZE * (1 << (MAX_ORDER - 1));
sgl_num_per_block = block_size / sgl_size;
block_num = count / sgl_num_per_block;
remain_sgl = count % sgl_num_per_block;
if ((!remain_sgl && block_num > HISI_ACC_MEM_BLOCK_NR) ||
(remain_sgl > 0 && block_num > HISI_ACC_MEM_BLOCK_NR - 1))
return ERR_PTR(-EINVAL);
pool = kzalloc(sizeof(*pool), GFP_KERNEL);
if (!pool)
return ERR_PTR(-ENOMEM);
block = pool->mem_block;
for (i = 0; i < block_num; i++) {
block[i].sgl = dma_alloc_coherent(dev, block_size,
&block[i].sgl_dma,
GFP_KERNEL);
if (!block[i].sgl)
goto err_free_mem;
block[i].size = block_size;
}
if (remain_sgl > 0) {
block[i].sgl = dma_alloc_coherent(dev, remain_sgl * sgl_size,
&block[i].sgl_dma,
GFP_KERNEL);
if (!block[i].sgl)
goto err_free_mem;
block[i].size = remain_sgl * sgl_size;
}
pool->sgl_num_per_block = sgl_num_per_block;
pool->block_num = remain_sgl ? block_num + 1 : block_num;
pool->count = count;
pool->sgl_size = sgl_size;
pool->sge_nr = sge_nr;
return pool;
err_free_mem:
for (j = 0; j < i; j++) {
dma_free_coherent(dev, block_size, block[j].sgl,
block[j].sgl_dma);
memset(block + j, 0, sizeof(*block));
}
kfree(pool);
return ERR_PTR(-ENOMEM);
}
EXPORT_SYMBOL_GPL(hisi_acc_create_sgl_pool);
/**
* hisi_acc_free_sgl_pool() - Free a hw sgl pool.
* @dev: The device which hw sgl pool belongs to.
* @pool: Pointer of pool.
*
* This function frees memory of a hw sgl pool.
*/
void hisi_acc_free_sgl_pool(struct device *dev, struct hisi_acc_sgl_pool *pool)
{
struct mem_block *block;
int i;
if (!dev || !pool)
return;
block = pool->mem_block;
for (i = 0; i < pool->block_num; i++)
dma_free_coherent(dev, block[i].size, block[i].sgl,
block[i].sgl_dma);
kfree(pool);
}
EXPORT_SYMBOL_GPL(hisi_acc_free_sgl_pool);
static struct hisi_acc_hw_sgl *acc_get_sgl(struct hisi_acc_sgl_pool *pool,
u32 index, dma_addr_t *hw_sgl_dma)
{
struct mem_block *block;
u32 block_index, offset;
if (!pool || !hw_sgl_dma || index >= pool->count)
return ERR_PTR(-EINVAL);
block = pool->mem_block;
block_index = index / pool->sgl_num_per_block;
offset = index % pool->sgl_num_per_block;
*hw_sgl_dma = block[block_index].sgl_dma + pool->sgl_size * offset;
return (void *)block[block_index].sgl + pool->sgl_size * offset;
}
static void sg_map_to_hw_sg(struct scatterlist *sgl,
struct acc_hw_sge *hw_sge)
{
hw_sge->buf = sg_dma_address(sgl);
hw_sge->len = cpu_to_le32(sg_dma_len(sgl));
}
static void inc_hw_sgl_sge(struct hisi_acc_hw_sgl *hw_sgl)
{
u16 var = le16_to_cpu(hw_sgl->entry_sum_in_sgl);
var++;
hw_sgl->entry_sum_in_sgl = cpu_to_le16(var);
}
static void update_hw_sgl_sum_sge(struct hisi_acc_hw_sgl *hw_sgl, u16 sum)
{
hw_sgl->entry_sum_in_chain = cpu_to_le16(sum);
}
/**
* hisi_acc_sg_buf_map_to_hw_sgl - Map a scatterlist to a hw sgl.
* @dev: The device which hw sgl belongs to.
* @sgl: Scatterlist which will be mapped to hw sgl.
* @pool: Pool which hw sgl memory will be allocated in.
* @index: Index of hisi_acc_hw_sgl in pool.
* @hw_sgl_dma: The dma address of allocated hw sgl.
*
* This function builds hw sgl according input sgl, user can use hw_sgl_dma
* as src/dst in its BD. Only support single hw sgl currently.
*/
struct hisi_acc_hw_sgl *
hisi_acc_sg_buf_map_to_hw_sgl(struct device *dev,
struct scatterlist *sgl,
struct hisi_acc_sgl_pool *pool,
u32 index, dma_addr_t *hw_sgl_dma)
{
struct hisi_acc_hw_sgl *curr_hw_sgl;
dma_addr_t curr_sgl_dma = 0;
struct acc_hw_sge *curr_hw_sge;
struct scatterlist *sg;
int i, sg_n, sg_n_mapped;
if (!dev || !sgl || !pool || !hw_sgl_dma)
return ERR_PTR(-EINVAL);
sg_n = sg_nents(sgl);
sg_n_mapped = dma_map_sg(dev, sgl, sg_n, DMA_BIDIRECTIONAL);
if (!sg_n_mapped)
return ERR_PTR(-EINVAL);
if (sg_n_mapped > pool->sge_nr) {
dma_unmap_sg(dev, sgl, sg_n, DMA_BIDIRECTIONAL);
return ERR_PTR(-EINVAL);
}
curr_hw_sgl = acc_get_sgl(pool, index, &curr_sgl_dma);
if (IS_ERR(curr_hw_sgl)) {
dma_unmap_sg(dev, sgl, sg_n, DMA_BIDIRECTIONAL);
return ERR_PTR(-ENOMEM);
}
curr_hw_sgl->entry_length_in_sgl = cpu_to_le16(pool->sge_nr);
curr_hw_sge = curr_hw_sgl->sge_entries;
for_each_sg(sgl, sg, sg_n_mapped, i) {
sg_map_to_hw_sg(sg, curr_hw_sge);
inc_hw_sgl_sge(curr_hw_sgl);
curr_hw_sge++;
}
update_hw_sgl_sum_sge(curr_hw_sgl, pool->sge_nr);
*hw_sgl_dma = curr_sgl_dma;
return curr_hw_sgl;
}
EXPORT_SYMBOL_GPL(hisi_acc_sg_buf_map_to_hw_sgl);
/**
* hisi_acc_sg_buf_unmap() - Unmap allocated hw sgl.
* @dev: The device which hw sgl belongs to.
* @sgl: Related scatterlist.
* @hw_sgl: Virtual address of hw sgl.
* @hw_sgl_dma: DMA address of hw sgl.
* @pool: Pool which hw sgl is allocated in.
*
* This function unmaps allocated hw sgl.
*/
void hisi_acc_sg_buf_unmap(struct device *dev, struct scatterlist *sgl,
struct hisi_acc_hw_sgl *hw_sgl)
{
if (!dev || !sgl || !hw_sgl)
return;
dma_unmap_sg(dev, sgl, sg_nents(sgl), DMA_BIDIRECTIONAL);
hw_sgl->entry_sum_in_chain = 0;
hw_sgl->entry_sum_in_sgl = 0;
hw_sgl->entry_length_in_sgl = 0;
}
EXPORT_SYMBOL_GPL(hisi_acc_sg_buf_unmap);
|