diff options
author | Arto Merilainen <amerilainen@nvidia.com> | 2014-03-19 03:38:25 -0400 |
---|---|---|
committer | Dan Willemsen <dwillemsen@nvidia.com> | 2015-03-18 15:08:53 -0400 |
commit | a9785995d5f22aaeb659285f8aeb64d8b56982e0 (patch) | |
tree | cc75f75bcf43db316a002a7a240b81f299bf6d7f /drivers/gpu/nvgpu/gk20a/mm_gk20a.h | |
parent | 61efaf843c22b85424036ec98015121c08f5f16c (diff) |
gpu: nvgpu: Add NVIDIA GPU Driver
This patch moves the NVIDIA GPU driver to a new location.
Bug 1482562
Change-Id: I24293810b9d0f1504fd9be00135e21dad656ccb6
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/383722
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Diffstat (limited to 'drivers/gpu/nvgpu/gk20a/mm_gk20a.h')
-rw-r--r-- | drivers/gpu/nvgpu/gk20a/mm_gk20a.h | 464 |
1 files changed, 464 insertions, 0 deletions
diff --git a/drivers/gpu/nvgpu/gk20a/mm_gk20a.h b/drivers/gpu/nvgpu/gk20a/mm_gk20a.h new file mode 100644 index 00000000..23d15c23 --- /dev/null +++ b/drivers/gpu/nvgpu/gk20a/mm_gk20a.h | |||
@@ -0,0 +1,464 @@ | |||
1 | /* | ||
2 | * drivers/video/tegra/host/gk20a/mm_gk20a.h | ||
3 | * | ||
4 | * GK20A memory management | ||
5 | * | ||
6 | * Copyright (c) 2011-2014, NVIDIA CORPORATION. All rights reserved. | ||
7 | * | ||
8 | * This program is free software; you can redistribute it and/or modify it | ||
9 | * under the terms and conditions of the GNU General Public License, | ||
10 | * version 2, as published by the Free Software Foundation. | ||
11 | * | ||
12 | * This program is distributed in the hope it will be useful, but WITHOUT | ||
13 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | ||
14 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | ||
15 | * more details. | ||
16 | * | ||
17 | * You should have received a copy of the GNU General Public License along with | ||
18 | * this program; if not, write to the Free Software Foundation, Inc., | ||
19 | * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA. | ||
20 | */ | ||
21 | #ifndef __MM_GK20A_H__ | ||
22 | #define __MM_GK20A_H__ | ||
23 | |||
24 | #include <linux/scatterlist.h> | ||
25 | #include <linux/dma-attrs.h> | ||
26 | #include <linux/iommu.h> | ||
27 | #include <asm/dma-iommu.h> | ||
28 | #include "gk20a_allocator.h" | ||
29 | |||
30 | /* This "address bit" in the gmmu ptes (and other gk20a accesses) | ||
31 | * signals the address as presented should be translated by the SMMU. | ||
32 | * Without this bit present gk20a accesses are *not* translated. | ||
33 | */ | ||
34 | /* Hack, get this from manuals somehow... */ | ||
35 | #define NV_MC_SMMU_VADDR_TRANSLATION_BIT 34 | ||
36 | #define NV_MC_SMMU_VADDR_TRANSLATE(x) (x | \ | ||
37 | (1ULL << NV_MC_SMMU_VADDR_TRANSLATION_BIT)) | ||
38 | |||
39 | /* For now keep the size relatively small-ish compared to the full | ||
40 | * 40b va. 32GB for now. It consists of two 16GB spaces. */ | ||
41 | #define NV_GMMU_VA_RANGE 35ULL | ||
42 | #define NV_GMMU_VA_IS_UPPER(x) ((x) >= ((u64)0x1 << (NV_GMMU_VA_RANGE-1))) | ||
43 | |||
44 | struct mem_desc { | ||
45 | struct dma_buf *ref; | ||
46 | struct sg_table *sgt; | ||
47 | u32 size; | ||
48 | }; | ||
49 | |||
50 | struct mem_desc_sub { | ||
51 | u32 offset; | ||
52 | u32 size; | ||
53 | }; | ||
54 | |||
55 | struct gpfifo_desc { | ||
56 | size_t size; | ||
57 | u32 entry_num; | ||
58 | |||
59 | u32 get; | ||
60 | u32 put; | ||
61 | |||
62 | bool wrap; | ||
63 | |||
64 | u64 iova; | ||
65 | struct gpfifo *cpu_va; | ||
66 | u64 gpu_va; | ||
67 | }; | ||
68 | |||
69 | struct mmu_desc { | ||
70 | void *cpuva; | ||
71 | u64 iova; | ||
72 | size_t size; | ||
73 | }; | ||
74 | |||
75 | struct inst_desc { | ||
76 | u64 iova; | ||
77 | void *cpuva; | ||
78 | phys_addr_t cpu_pa; | ||
79 | size_t size; | ||
80 | }; | ||
81 | |||
82 | struct surface_mem_desc { | ||
83 | u64 iova; | ||
84 | void *cpuva; | ||
85 | struct sg_table *sgt; | ||
86 | size_t size; | ||
87 | }; | ||
88 | |||
89 | struct userd_desc { | ||
90 | struct sg_table *sgt; | ||
91 | u64 iova; | ||
92 | void *cpuva; | ||
93 | size_t size; | ||
94 | u64 gpu_va; | ||
95 | }; | ||
96 | |||
97 | struct runlist_mem_desc { | ||
98 | u64 iova; | ||
99 | void *cpuva; | ||
100 | size_t size; | ||
101 | }; | ||
102 | |||
103 | struct patch_desc { | ||
104 | struct page **pages; | ||
105 | u64 iova; | ||
106 | size_t size; | ||
107 | void *cpu_va; | ||
108 | u64 gpu_va; | ||
109 | u32 data_count; | ||
110 | }; | ||
111 | |||
112 | struct pmu_mem_desc { | ||
113 | void *cpuva; | ||
114 | u64 iova; | ||
115 | u64 pmu_va; | ||
116 | size_t size; | ||
117 | }; | ||
118 | |||
119 | struct priv_cmd_queue_mem_desc { | ||
120 | u64 base_iova; | ||
121 | u32 *base_cpuva; | ||
122 | size_t size; | ||
123 | }; | ||
124 | |||
125 | struct zcull_ctx_desc { | ||
126 | struct mem_desc mem; | ||
127 | u64 gpu_va; | ||
128 | u32 ctx_attr; | ||
129 | u32 ctx_sw_mode; | ||
130 | }; | ||
131 | |||
132 | struct pm_ctx_desc { | ||
133 | struct mem_desc mem; | ||
134 | u64 gpu_va; | ||
135 | u32 ctx_attr; | ||
136 | u32 ctx_sw_mode; | ||
137 | }; | ||
138 | |||
139 | struct gr_ctx_buffer_desc; | ||
140 | struct platform_device; | ||
141 | struct gr_ctx_buffer_desc { | ||
142 | void (*destroy)(struct platform_device *, struct gr_ctx_buffer_desc *); | ||
143 | struct sg_table *sgt; | ||
144 | struct page **pages; | ||
145 | size_t size; | ||
146 | u64 iova; | ||
147 | struct dma_attrs attrs; | ||
148 | void *priv; | ||
149 | }; | ||
150 | |||
151 | struct gr_ctx_desc { | ||
152 | struct page **pages; | ||
153 | u64 iova; | ||
154 | size_t size; | ||
155 | u64 gpu_va; | ||
156 | }; | ||
157 | |||
158 | struct compbit_store_desc { | ||
159 | struct pages **pages; | ||
160 | size_t size; | ||
161 | u64 base_iova; | ||
162 | }; | ||
163 | |||
164 | struct page_table_gk20a { | ||
165 | /* backing for */ | ||
166 | /* Either a *page or a *mem_handle */ | ||
167 | void *ref; | ||
168 | /* track mapping cnt on this page table */ | ||
169 | u32 ref_cnt; | ||
170 | struct sg_table *sgt; | ||
171 | size_t size; | ||
172 | }; | ||
173 | |||
174 | #ifndef _NVHOST_MEM_MGR_H | ||
175 | enum gk20a_mem_rw_flag { | ||
176 | gk20a_mem_flag_none = 0, | ||
177 | gk20a_mem_flag_read_only = 1, | ||
178 | gk20a_mem_flag_write_only = 2, | ||
179 | }; | ||
180 | #endif | ||
181 | |||
182 | enum gmmu_pgsz_gk20a { | ||
183 | gmmu_page_size_small = 0, | ||
184 | gmmu_page_size_big = 1, | ||
185 | gmmu_nr_page_sizes = 2 | ||
186 | }; | ||
187 | |||
188 | |||
189 | struct page_directory_gk20a { | ||
190 | /* backing for */ | ||
191 | u32 num_pdes; | ||
192 | void *kv; | ||
193 | /* Either a *page or a *mem_handle */ | ||
194 | void *ref; | ||
195 | struct sg_table *sgt; | ||
196 | size_t size; | ||
197 | struct page_table_gk20a *ptes[gmmu_nr_page_sizes]; | ||
198 | }; | ||
199 | |||
200 | struct priv_cmd_queue { | ||
201 | struct priv_cmd_queue_mem_desc mem; | ||
202 | u64 base_gpuva; /* gpu_va base */ | ||
203 | u16 size; /* num of entries in words */ | ||
204 | u16 put; /* put for priv cmd queue */ | ||
205 | u16 get; /* get for priv cmd queue */ | ||
206 | struct list_head free; /* list of pre-allocated free entries */ | ||
207 | struct list_head head; /* list of used entries */ | ||
208 | }; | ||
209 | |||
210 | struct priv_cmd_entry { | ||
211 | u32 *ptr; | ||
212 | u64 gva; | ||
213 | u16 get; /* start of entry in queue */ | ||
214 | u16 size; /* in words */ | ||
215 | u32 gp_get; /* gp_get when submitting last priv cmd */ | ||
216 | u32 gp_put; /* gp_put when submitting last priv cmd */ | ||
217 | u32 gp_wrap; /* wrap when submitting last priv cmd */ | ||
218 | bool pre_alloc; /* prealloc entry, free to free list */ | ||
219 | struct list_head list; /* node for lists */ | ||
220 | }; | ||
221 | |||
222 | struct mapped_buffer_node { | ||
223 | struct vm_gk20a *vm; | ||
224 | struct rb_node node; | ||
225 | struct list_head unmap_list; | ||
226 | struct list_head va_buffers_list; | ||
227 | struct vm_reserved_va_node *va_node; | ||
228 | u64 addr; | ||
229 | u64 size; | ||
230 | struct dma_buf *dmabuf; | ||
231 | struct sg_table *sgt; | ||
232 | struct kref ref; | ||
233 | u32 user_mapped; | ||
234 | bool own_mem_ref; | ||
235 | u32 pgsz_idx; | ||
236 | u32 ctag_offset; | ||
237 | u32 ctag_lines; | ||
238 | u32 flags; | ||
239 | u32 kind; | ||
240 | bool va_allocated; | ||
241 | }; | ||
242 | |||
243 | struct vm_reserved_va_node { | ||
244 | struct list_head reserved_va_list; | ||
245 | struct list_head va_buffers_list; | ||
246 | u32 pgsz_idx; | ||
247 | u64 vaddr_start; | ||
248 | u64 size; | ||
249 | bool sparse; | ||
250 | }; | ||
251 | |||
252 | struct vm_gk20a { | ||
253 | struct mm_gk20a *mm; | ||
254 | struct gk20a_as_share *as_share; /* as_share this represents */ | ||
255 | |||
256 | u64 va_start; | ||
257 | u64 va_limit; | ||
258 | |||
259 | int num_user_mapped_buffers; | ||
260 | |||
261 | bool big_pages; /* enable large page support */ | ||
262 | bool enable_ctag; | ||
263 | bool tlb_dirty; | ||
264 | bool mapped; | ||
265 | |||
266 | struct kref ref; | ||
267 | |||
268 | struct mutex update_gmmu_lock; | ||
269 | |||
270 | struct page_directory_gk20a pdes; | ||
271 | |||
272 | struct gk20a_allocator vma[gmmu_nr_page_sizes]; | ||
273 | struct rb_root mapped_buffers; | ||
274 | |||
275 | struct list_head reserved_va_list; | ||
276 | |||
277 | dma_addr_t zero_page_iova; | ||
278 | void *zero_page_cpuva; | ||
279 | struct sg_table *zero_page_sgt; | ||
280 | }; | ||
281 | |||
282 | struct gk20a; | ||
283 | struct channel_gk20a; | ||
284 | |||
285 | int gk20a_init_mm_support(struct gk20a *g); | ||
286 | int gk20a_init_mm_setup_sw(struct gk20a *g); | ||
287 | int gk20a_init_bar1_vm(struct mm_gk20a *mm); | ||
288 | int gk20a_init_pmu_vm(struct mm_gk20a *mm); | ||
289 | |||
290 | void gk20a_mm_fb_flush(struct gk20a *g); | ||
291 | void gk20a_mm_l2_flush(struct gk20a *g, bool invalidate); | ||
292 | void gk20a_mm_l2_invalidate(struct gk20a *g); | ||
293 | |||
294 | struct mm_gk20a { | ||
295 | struct gk20a *g; | ||
296 | |||
297 | u32 compression_page_size; | ||
298 | u32 big_page_size; | ||
299 | u32 pde_stride; | ||
300 | u32 pde_stride_shift; | ||
301 | |||
302 | struct { | ||
303 | u32 order; | ||
304 | u32 num_ptes; | ||
305 | } page_table_sizing[gmmu_nr_page_sizes]; | ||
306 | |||
307 | |||
308 | struct { | ||
309 | u64 size; | ||
310 | } channel; | ||
311 | |||
312 | struct { | ||
313 | u32 aperture_size; | ||
314 | struct vm_gk20a vm; | ||
315 | struct inst_desc inst_block; | ||
316 | } bar1; | ||
317 | |||
318 | struct { | ||
319 | u32 aperture_size; | ||
320 | struct vm_gk20a vm; | ||
321 | struct inst_desc inst_block; | ||
322 | } pmu; | ||
323 | |||
324 | struct mutex tlb_lock; | ||
325 | struct mutex l2_op_lock; | ||
326 | |||
327 | void (*remove_support)(struct mm_gk20a *mm); | ||
328 | bool sw_ready; | ||
329 | #ifdef CONFIG_DEBUG_FS | ||
330 | u32 ltc_enabled; | ||
331 | u32 ltc_enabled_debug; | ||
332 | #endif | ||
333 | }; | ||
334 | |||
335 | int gk20a_mm_init(struct mm_gk20a *mm); | ||
336 | |||
337 | #define gk20a_from_mm(mm) ((mm)->g) | ||
338 | #define gk20a_from_vm(vm) ((vm)->mm->g) | ||
339 | |||
340 | #define dev_from_vm(vm) dev_from_gk20a(vm->mm->g) | ||
341 | |||
342 | #define DEFAULT_ALLOC_ALIGNMENT (4*1024) | ||
343 | |||
344 | static inline int bar1_aperture_size_mb_gk20a(void) | ||
345 | { | ||
346 | return 128; /*TBD read this from fuses?*/ | ||
347 | } | ||
348 | /* max address bits */ | ||
349 | static inline int max_physaddr_bits_gk20a(void) | ||
350 | { | ||
351 | return 40;/*"old" sys physaddr, meaningful? */ | ||
352 | } | ||
353 | static inline int max_vid_physaddr_bits_gk20a(void) | ||
354 | { | ||
355 | /* "vid phys" is asid/smmu phys?, | ||
356 | * i.e. is this the real sys physaddr? */ | ||
357 | return 37; | ||
358 | } | ||
359 | static inline int max_vaddr_bits_gk20a(void) | ||
360 | { | ||
361 | return 40; /* chopped for area? */ | ||
362 | } | ||
363 | |||
364 | #if 0 /*related to addr bits above, concern below TBD on which is accurate */ | ||
365 | #define bar1_instance_block_shift_gk20a() (max_physaddr_bits_gk20a() -\ | ||
366 | bus_bar1_block_ptr_s()) | ||
367 | #else | ||
368 | #define bar1_instance_block_shift_gk20a() bus_bar1_block_ptr_shift_v() | ||
369 | #endif | ||
370 | |||
371 | void gk20a_mm_dump_vm(struct vm_gk20a *vm, | ||
372 | u64 va_begin, u64 va_end, char *label); | ||
373 | |||
374 | int gk20a_mm_suspend(struct gk20a *g); | ||
375 | |||
376 | phys_addr_t gk20a_get_phys_from_iova(struct device *d, | ||
377 | u64 dma_addr); | ||
378 | |||
379 | int gk20a_get_sgtable(struct device *d, struct sg_table **sgt, | ||
380 | void *cpuva, u64 iova, | ||
381 | size_t size); | ||
382 | |||
383 | int gk20a_get_sgtable_from_pages(struct device *d, struct sg_table **sgt, | ||
384 | struct page **pages, u64 iova, | ||
385 | size_t size); | ||
386 | |||
387 | void gk20a_free_sgtable(struct sg_table **sgt); | ||
388 | |||
389 | u64 gk20a_mm_iova_addr(struct scatterlist *sgl); | ||
390 | |||
391 | void gk20a_mm_ltc_isr(struct gk20a *g); | ||
392 | |||
393 | bool gk20a_mm_mmu_debug_mode_enabled(struct gk20a *g); | ||
394 | |||
395 | int gk20a_mm_mmu_vpr_info_fetch(struct gk20a *g); | ||
396 | |||
397 | u64 gk20a_gmmu_map(struct vm_gk20a *vm, | ||
398 | struct sg_table **sgt, | ||
399 | u64 size, | ||
400 | u32 flags, | ||
401 | int rw_flag); | ||
402 | |||
403 | void gk20a_gmmu_unmap(struct vm_gk20a *vm, | ||
404 | u64 vaddr, | ||
405 | u64 size, | ||
406 | int rw_flag); | ||
407 | |||
408 | struct sg_table *gk20a_mm_pin(struct device *dev, struct dma_buf *dmabuf); | ||
409 | void gk20a_mm_unpin(struct device *dev, struct dma_buf *dmabuf, | ||
410 | struct sg_table *sgt); | ||
411 | |||
412 | u64 gk20a_vm_map(struct vm_gk20a *vm, | ||
413 | struct dma_buf *dmabuf, | ||
414 | u64 offset_align, | ||
415 | u32 flags /*NVHOST_AS_MAP_BUFFER_FLAGS_*/, | ||
416 | int kind, | ||
417 | struct sg_table **sgt, | ||
418 | bool user_mapped, | ||
419 | int rw_flag); | ||
420 | |||
421 | /* unmap handle from kernel */ | ||
422 | void gk20a_vm_unmap(struct vm_gk20a *vm, u64 offset); | ||
423 | |||
424 | /* get reference to all currently mapped buffers */ | ||
425 | int gk20a_vm_get_buffers(struct vm_gk20a *vm, | ||
426 | struct mapped_buffer_node ***mapped_buffers, | ||
427 | int *num_buffers); | ||
428 | |||
429 | /* put references on the given buffers */ | ||
430 | void gk20a_vm_put_buffers(struct vm_gk20a *vm, | ||
431 | struct mapped_buffer_node **mapped_buffers, | ||
432 | int num_buffers); | ||
433 | |||
434 | /* invalidate tlbs for the vm area */ | ||
435 | void gk20a_mm_tlb_invalidate(struct vm_gk20a *vm); | ||
436 | |||
437 | /* find buffer corresponding to va */ | ||
438 | int gk20a_vm_find_buffer(struct vm_gk20a *vm, u64 gpu_va, | ||
439 | struct dma_buf **dmabuf, | ||
440 | u64 *offset); | ||
441 | |||
442 | void gk20a_vm_get(struct vm_gk20a *vm); | ||
443 | void gk20a_vm_put(struct vm_gk20a *vm); | ||
444 | |||
445 | /* vm-as interface */ | ||
446 | struct nvhost_as_alloc_space_args; | ||
447 | struct nvhost_as_free_space_args; | ||
448 | int gk20a_vm_alloc_share(struct gk20a_as_share *as_share); | ||
449 | int gk20a_vm_release_share(struct gk20a_as_share *as_share); | ||
450 | int gk20a_vm_alloc_space(struct gk20a_as_share *as_share, | ||
451 | struct nvhost_as_alloc_space_args *args); | ||
452 | int gk20a_vm_free_space(struct gk20a_as_share *as_share, | ||
453 | struct nvhost_as_free_space_args *args); | ||
454 | int gk20a_vm_bind_channel(struct gk20a_as_share *as_share, | ||
455 | struct channel_gk20a *ch); | ||
456 | int gk20a_vm_map_buffer(struct gk20a_as_share *as_share, | ||
457 | int dmabuf_fd, | ||
458 | u64 *offset_align, | ||
459 | u32 flags, /*NVHOST_AS_MAP_BUFFER_FLAGS_*/ | ||
460 | int kind); | ||
461 | int gk20a_vm_unmap_buffer(struct gk20a_as_share *, u64 offset); | ||
462 | |||
463 | int gk20a_dmabuf_alloc_drvdata(struct dma_buf *dmabuf, struct device *dev); | ||
464 | #endif /*_MM_GK20A_H_ */ | ||