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
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
|
From 7d78c8e4a3fe6f5735c815db7d285a803c62d0df Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@amd.com>
Date: Fri, 18 Mar 2016 21:00:35 +0100
Subject: [PATCH 0461/1110] drm/amdgpu: change parameter passing in the VM code
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Make it more flexible by passing src and page addresses
directly instead of the structures they contain.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Kalyan Alle <kalyan.alle@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 100 ++++++++++++++++++---------------
1 file changed, 54 insertions(+), 46 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index ac4e092..4db8a2b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -356,8 +356,8 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
* amdgpu_vm_update_pages - helper to call the right asic function
*
* @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: GTT hw access flags
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
* @ib: indirect buffer to fill with commands
* @pe: addr of the page entry
* @addr: dst addr to write into pe
@@ -369,8 +369,8 @@ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
* to setup the page table using the DMA.
*/
static void amdgpu_vm_update_pages(struct amdgpu_device *adev,
- struct amdgpu_gart *gtt,
- uint32_t gtt_flags,
+ uint64_t src,
+ dma_addr_t *pages_addr,
struct amdgpu_ib *ib,
uint64_t pe, uint64_t addr,
unsigned count, uint32_t incr,
@@ -378,12 +378,11 @@ static void amdgpu_vm_update_pages(struct amdgpu_device *adev,
{
trace_amdgpu_vm_set_page(pe, addr, count, incr, flags);
- if ((gtt == &adev->gart) && (flags == gtt_flags)) {
- uint64_t src = gtt->table_addr + (addr >> 12) * 8;
+ if (src) {
+ src += (addr >> 12) * 8;
amdgpu_vm_copy_pte(adev, ib, pe, src, count);
- } else if (gtt) {
- dma_addr_t *pages_addr = gtt->pages_addr;
+ } else if (pages_addr) {
amdgpu_vm_write_pte(adev, ib, pages_addr, pe, addr,
count, incr, flags);
@@ -432,7 +431,7 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev,
r = amdgpu_job_alloc_with_ib(adev, 64, &job);
if (r)
goto error;
- amdgpu_vm_update_pages(adev, NULL, 0, &job->ibs[0], addr, 0, entries,
+ amdgpu_vm_update_pages(adev, 0, NULL, &job->ibs[0], addr, 0, entries,
0, 0);
amdgpu_ring_pad_ib(ring, &job->ibs[0]);
@@ -538,7 +537,7 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
((last_pt + incr * count) != pt)) {
if (count) {
- amdgpu_vm_update_pages(adev, NULL, 0, ib,
+ amdgpu_vm_update_pages(adev, 0, NULL, ib,
last_pde, last_pt,
count, incr,
AMDGPU_PTE_VALID);
@@ -553,7 +552,7 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
}
if (count)
- amdgpu_vm_update_pages(adev, NULL, 0, ib, last_pde, last_pt,
+ amdgpu_vm_update_pages(adev, 0, NULL, ib, last_pde, last_pt,
count, incr, AMDGPU_PTE_VALID);
if (ib->length_dw != 0) {
@@ -584,8 +583,8 @@ error_free:
* amdgpu_vm_frag_ptes - add fragment information to PTEs
*
* @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: GTT hw mapping flags
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
* @ib: IB for the update
* @pe_start: first PTE to handle
* @pe_end: last PTE to handle
@@ -595,8 +594,8 @@ error_free:
* Global and local mutex must be locked!
*/
static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
- struct amdgpu_gart *gtt,
- uint32_t gtt_flags,
+ uint64_t src,
+ dma_addr_t *pages_addr,
struct amdgpu_ib *ib,
uint64_t pe_start, uint64_t pe_end,
uint64_t addr, uint32_t flags)
@@ -634,10 +633,11 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
return;
/* system pages are non continuously */
- if (gtt || !(flags & AMDGPU_PTE_VALID) || (frag_start >= frag_end)) {
+ if (src || pages_addr || !(flags & AMDGPU_PTE_VALID) ||
+ (frag_start >= frag_end)) {
count = (pe_end - pe_start) / 8;
- amdgpu_vm_update_pages(adev, gtt, gtt_flags, ib, pe_start,
+ amdgpu_vm_update_pages(adev, src, pages_addr, ib, pe_start,
addr, count, AMDGPU_GPU_PAGE_SIZE,
flags);
return;
@@ -646,21 +646,21 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
/* handle the 4K area at the beginning */
if (pe_start != frag_start) {
count = (frag_start - pe_start) / 8;
- amdgpu_vm_update_pages(adev, NULL, 0, ib, pe_start, addr,
+ amdgpu_vm_update_pages(adev, 0, NULL, ib, pe_start, addr,
count, AMDGPU_GPU_PAGE_SIZE, flags);
addr += AMDGPU_GPU_PAGE_SIZE * count;
}
/* handle the area in the middle */
count = (frag_end - frag_start) / 8;
- amdgpu_vm_update_pages(adev, NULL, 0, ib, frag_start, addr, count,
+ amdgpu_vm_update_pages(adev, 0, NULL, ib, frag_start, addr, count,
AMDGPU_GPU_PAGE_SIZE, flags | frag_flags);
/* handle the 4K area at the end */
if (frag_end != pe_end) {
addr += AMDGPU_GPU_PAGE_SIZE * count;
count = (pe_end - frag_end) / 8;
- amdgpu_vm_update_pages(adev, NULL, 0, ib, frag_end, addr,
+ amdgpu_vm_update_pages(adev, 0, NULL, ib, frag_end, addr,
count, AMDGPU_GPU_PAGE_SIZE, flags);
}
}
@@ -669,8 +669,8 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
* amdgpu_vm_update_ptes - make sure that page tables are valid
*
* @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: GTT hw mapping flags
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
* @vm: requested vm
* @start: start of GPU address range
* @end: end of GPU address range
@@ -680,8 +680,8 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_device *adev,
* Update the page tables in the range @start - @end.
*/
static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
- struct amdgpu_gart *gtt,
- uint32_t gtt_flags,
+ uint64_t src,
+ dma_addr_t *pages_addr,
struct amdgpu_vm *vm,
struct amdgpu_ib *ib,
uint64_t start, uint64_t end,
@@ -712,7 +712,7 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
if (last_pe_end != pe_start) {
- amdgpu_vm_frag_ptes(adev, gtt, gtt_flags, ib,
+ amdgpu_vm_frag_ptes(adev, src, pages_addr, ib,
last_pe_start, last_pe_end,
last_dst, flags);
@@ -727,17 +727,16 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
dst += nptes * AMDGPU_GPU_PAGE_SIZE;
}
- amdgpu_vm_frag_ptes(adev, gtt, gtt_flags, ib,
- last_pe_start, last_pe_end,
- last_dst, flags);
+ amdgpu_vm_frag_ptes(adev, src, pages_addr, ib, last_pe_start,
+ last_pe_end, last_dst, flags);
}
/**
* amdgpu_vm_bo_update_mapping - update a mapping in the vm page table
*
* @adev: amdgpu_device pointer
- * @gtt: GART instance to use for mapping
- * @gtt_flags: flags as they are used for GTT
+ * @src: address where to copy page table entries from
+ * @pages_addr: DMA addresses to use for mapping
* @vm: requested vm
* @start: start of mapped range
* @last: last mapped entry
@@ -749,8 +748,8 @@ static void amdgpu_vm_update_ptes(struct amdgpu_device *adev,
* Returns 0 for success, -EINVAL for failure.
*/
static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
- struct amdgpu_gart *gtt,
- uint32_t gtt_flags,
+ uint64_t src,
+ dma_addr_t *pages_addr,
struct amdgpu_vm *vm,
uint64_t start, uint64_t last,
uint32_t flags, uint64_t addr,
@@ -781,11 +780,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
/* padding, etc. */
ndw = 64;
- if ((gtt == &adev->gart) && (flags == gtt_flags)) {
+ if (src) {
/* only copy commands needed */
ndw += ncmds * 7;
- } else if (gtt) {
+ } else if (pages_addr) {
/* header for write data commands */
ndw += ncmds * 4;
@@ -815,8 +814,8 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
if (r)
goto error_free;
- amdgpu_vm_update_ptes(adev, gtt, gtt_flags, vm, ib, start, last + 1,
- addr, flags);
+ amdgpu_vm_update_ptes(adev, src, pages_addr, vm, ib, start,
+ last + 1, addr, flags);
amdgpu_ring_pad_ib(ring, ib);
WARN_ON(ib->length_dw > ndw);
@@ -858,12 +857,13 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
uint32_t gtt_flags,
struct amdgpu_vm *vm,
struct amdgpu_bo_va_mapping *mapping,
- uint64_t addr, struct fence **fence)
+ uint32_t flags, uint64_t addr,
+ struct fence **fence)
{
const uint64_t max_size = 64ULL * 1024ULL * 1024ULL / AMDGPU_GPU_PAGE_SIZE;
- uint64_t start = mapping->it.start;
- uint32_t flags = gtt_flags;
+ uint64_t src = 0, start = mapping->it.start;
+ dma_addr_t *pages_addr = NULL;
int r;
/* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here
@@ -876,10 +876,17 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
trace_amdgpu_vm_bo_update(mapping);
+ if (gtt) {
+ if (flags == gtt_flags)
+ src = adev->gart.table_addr + (addr >> 12) * 8;
+ else
+ pages_addr = >t->pages_addr[addr >> 12];
+ addr = 0;
+ }
addr += mapping->offset;
- if (!gtt || ((gtt == &adev->gart) && (flags == gtt_flags)))
- return amdgpu_vm_bo_update_mapping(adev, gtt, gtt_flags, vm,
+ if (!gtt || src)
+ return amdgpu_vm_bo_update_mapping(adev, src, pages_addr, vm,
start, mapping->it.last,
flags, addr, fence);
@@ -887,7 +894,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev,
uint64_t last;
last = min((uint64_t)mapping->it.last, start + max_size - 1);
- r = amdgpu_vm_bo_update_mapping(adev, gtt, gtt_flags, vm,
+ r = amdgpu_vm_bo_update_mapping(adev, src, pages_addr, vm,
start, last, flags, addr,
fence);
if (r)
@@ -919,7 +926,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
struct amdgpu_vm *vm = bo_va->vm;
struct amdgpu_bo_va_mapping *mapping;
struct amdgpu_gart *gtt = NULL;
- uint32_t flags;
+ uint32_t gtt_flags, flags;
uint64_t addr;
int r;
@@ -942,6 +949,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
}
flags = amdgpu_ttm_tt_pte_flags(adev, bo_va->bo->tbo.ttm, mem);
+ gtt_flags = (adev == bo_va->bo->adev) ? flags : 0;
spin_lock(&vm->status_lock);
if (!list_empty(&bo_va->vm_status))
@@ -949,8 +957,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
spin_unlock(&vm->status_lock);
list_for_each_entry(mapping, &bo_va->invalids, list) {
- r = amdgpu_vm_bo_split_mapping(adev, gtt, flags, vm, mapping, addr,
- &bo_va->last_pt_update);
+ r = amdgpu_vm_bo_split_mapping(adev, gtt, gtt_flags, vm, mapping,
+ flags, addr, &bo_va->last_pt_update);
if (r)
return r;
}
@@ -996,7 +1004,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
list_del(&mapping->list);
r = amdgpu_vm_bo_split_mapping(adev, NULL, 0, vm, mapping,
- 0, NULL);
+ 0, 0, NULL);
kfree(mapping);
if (r)
return r;
--
2.7.4
|
