xref: /linux/drivers/gpu/drm/amd/amdgpu/amdgpu.h (revision 063246641d4a9e9de84a2466fbad50112faf88dc)
1 /*
2  * Copyright 2008 Advanced Micro Devices, Inc.
3  * Copyright 2008 Red Hat Inc.
4  * Copyright 2009 Jerome Glisse.
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a
7  * copy of this software and associated documentation files (the "Software"),
8  * to deal in the Software without restriction, including without limitation
9  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10  * and/or sell copies of the Software, and to permit persons to whom the
11  * Software is furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
19  * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
20  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22  * OTHER DEALINGS IN THE SOFTWARE.
23  *
24  * Authors: Dave Airlie
25  *          Alex Deucher
26  *          Jerome Glisse
27  */
28 #ifndef __AMDGPU_H__
29 #define __AMDGPU_H__
30 
31 #include <linux/atomic.h>
32 #include <linux/wait.h>
33 #include <linux/list.h>
34 #include <linux/kref.h>
35 #include <linux/rbtree.h>
36 #include <linux/hashtable.h>
37 #include <linux/dma-fence.h>
38 
39 #include <ttm/ttm_bo_api.h>
40 #include <ttm/ttm_bo_driver.h>
41 #include <ttm/ttm_placement.h>
42 #include <ttm/ttm_module.h>
43 #include <ttm/ttm_execbuf_util.h>
44 
45 #include <drm/drmP.h>
46 #include <drm/drm_gem.h>
47 #include <drm/amdgpu_drm.h>
48 
49 #include "amd_shared.h"
50 #include "amdgpu_mode.h"
51 #include "amdgpu_ih.h"
52 #include "amdgpu_irq.h"
53 #include "amdgpu_ucode.h"
54 #include "amdgpu_ttm.h"
55 #include "amdgpu_psp.h"
56 #include "amdgpu_gds.h"
57 #include "amdgpu_sync.h"
58 #include "amdgpu_ring.h"
59 #include "amdgpu_vm.h"
60 #include "amd_powerplay.h"
61 #include "amdgpu_dpm.h"
62 #include "amdgpu_acp.h"
63 #include "amdgpu_uvd.h"
64 #include "amdgpu_vce.h"
65 
66 #include "gpu_scheduler.h"
67 #include "amdgpu_virt.h"
68 
69 /*
70  * Modules parameters.
71  */
72 extern int amdgpu_modeset;
73 extern int amdgpu_vram_limit;
74 extern int amdgpu_gart_size;
75 extern int amdgpu_moverate;
76 extern int amdgpu_benchmarking;
77 extern int amdgpu_testing;
78 extern int amdgpu_audio;
79 extern int amdgpu_disp_priority;
80 extern int amdgpu_hw_i2c;
81 extern int amdgpu_pcie_gen2;
82 extern int amdgpu_msi;
83 extern int amdgpu_lockup_timeout;
84 extern int amdgpu_dpm;
85 extern int amdgpu_fw_load_type;
86 extern int amdgpu_aspm;
87 extern int amdgpu_runtime_pm;
88 extern unsigned amdgpu_ip_block_mask;
89 extern int amdgpu_bapm;
90 extern int amdgpu_deep_color;
91 extern int amdgpu_vm_size;
92 extern int amdgpu_vm_block_size;
93 extern int amdgpu_vm_fault_stop;
94 extern int amdgpu_vm_debug;
95 extern int amdgpu_sched_jobs;
96 extern int amdgpu_sched_hw_submission;
97 extern int amdgpu_no_evict;
98 extern int amdgpu_direct_gma_size;
99 extern unsigned amdgpu_pcie_gen_cap;
100 extern unsigned amdgpu_pcie_lane_cap;
101 extern unsigned amdgpu_cg_mask;
102 extern unsigned amdgpu_pg_mask;
103 extern char *amdgpu_disable_cu;
104 extern char *amdgpu_virtual_display;
105 extern unsigned amdgpu_pp_feature_mask;
106 extern int amdgpu_vram_page_split;
107 extern int amdgpu_ngg;
108 extern int amdgpu_prim_buf_per_se;
109 extern int amdgpu_pos_buf_per_se;
110 extern int amdgpu_cntl_sb_buf_per_se;
111 extern int amdgpu_param_buf_per_se;
112 
113 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
114 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */
115 #define AMDGPU_FENCE_JIFFIES_TIMEOUT		(HZ / 2)
116 /* AMDGPU_IB_POOL_SIZE must be a power of 2 */
117 #define AMDGPU_IB_POOL_SIZE			16
118 #define AMDGPU_DEBUGFS_MAX_COMPONENTS		32
119 #define AMDGPUFB_CONN_LIMIT			4
120 #define AMDGPU_BIOS_NUM_SCRATCH			16
121 
122 /* max number of IP instances */
123 #define AMDGPU_MAX_SDMA_INSTANCES		2
124 
125 /* hard reset data */
126 #define AMDGPU_ASIC_RESET_DATA                  0x39d5e86b
127 
128 /* reset flags */
129 #define AMDGPU_RESET_GFX			(1 << 0)
130 #define AMDGPU_RESET_COMPUTE			(1 << 1)
131 #define AMDGPU_RESET_DMA			(1 << 2)
132 #define AMDGPU_RESET_CP				(1 << 3)
133 #define AMDGPU_RESET_GRBM			(1 << 4)
134 #define AMDGPU_RESET_DMA1			(1 << 5)
135 #define AMDGPU_RESET_RLC			(1 << 6)
136 #define AMDGPU_RESET_SEM			(1 << 7)
137 #define AMDGPU_RESET_IH				(1 << 8)
138 #define AMDGPU_RESET_VMC			(1 << 9)
139 #define AMDGPU_RESET_MC				(1 << 10)
140 #define AMDGPU_RESET_DISPLAY			(1 << 11)
141 #define AMDGPU_RESET_UVD			(1 << 12)
142 #define AMDGPU_RESET_VCE			(1 << 13)
143 #define AMDGPU_RESET_VCE1			(1 << 14)
144 
145 /* GFX current status */
146 #define AMDGPU_GFX_NORMAL_MODE			0x00000000L
147 #define AMDGPU_GFX_SAFE_MODE			0x00000001L
148 #define AMDGPU_GFX_PG_DISABLED_MODE		0x00000002L
149 #define AMDGPU_GFX_CG_DISABLED_MODE		0x00000004L
150 #define AMDGPU_GFX_LBPW_DISABLED_MODE		0x00000008L
151 
152 /* max cursor sizes (in pixels) */
153 #define CIK_CURSOR_WIDTH 128
154 #define CIK_CURSOR_HEIGHT 128
155 
156 struct amdgpu_device;
157 struct amdgpu_ib;
158 struct amdgpu_cs_parser;
159 struct amdgpu_job;
160 struct amdgpu_irq_src;
161 struct amdgpu_fpriv;
162 
163 enum amdgpu_cp_irq {
164 	AMDGPU_CP_IRQ_GFX_EOP = 0,
165 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP,
166 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP,
167 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP,
168 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP,
169 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE0_EOP,
170 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE1_EOP,
171 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE2_EOP,
172 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE3_EOP,
173 
174 	AMDGPU_CP_IRQ_LAST
175 };
176 
177 enum amdgpu_sdma_irq {
178 	AMDGPU_SDMA_IRQ_TRAP0 = 0,
179 	AMDGPU_SDMA_IRQ_TRAP1,
180 
181 	AMDGPU_SDMA_IRQ_LAST
182 };
183 
184 enum amdgpu_thermal_irq {
185 	AMDGPU_THERMAL_IRQ_LOW_TO_HIGH = 0,
186 	AMDGPU_THERMAL_IRQ_HIGH_TO_LOW,
187 
188 	AMDGPU_THERMAL_IRQ_LAST
189 };
190 
191 enum amdgpu_kiq_irq {
192 	AMDGPU_CP_KIQ_IRQ_DRIVER0 = 0,
193 	AMDGPU_CP_KIQ_IRQ_LAST
194 };
195 
196 int amdgpu_set_clockgating_state(struct amdgpu_device *adev,
197 				  enum amd_ip_block_type block_type,
198 				  enum amd_clockgating_state state);
199 int amdgpu_set_powergating_state(struct amdgpu_device *adev,
200 				  enum amd_ip_block_type block_type,
201 				  enum amd_powergating_state state);
202 void amdgpu_get_clockgating_state(struct amdgpu_device *adev, u32 *flags);
203 int amdgpu_wait_for_idle(struct amdgpu_device *adev,
204 			 enum amd_ip_block_type block_type);
205 bool amdgpu_is_idle(struct amdgpu_device *adev,
206 		    enum amd_ip_block_type block_type);
207 
208 #define AMDGPU_MAX_IP_NUM 16
209 
210 struct amdgpu_ip_block_status {
211 	bool valid;
212 	bool sw;
213 	bool hw;
214 	bool late_initialized;
215 	bool hang;
216 };
217 
218 struct amdgpu_ip_block_version {
219 	const enum amd_ip_block_type type;
220 	const u32 major;
221 	const u32 minor;
222 	const u32 rev;
223 	const struct amd_ip_funcs *funcs;
224 };
225 
226 struct amdgpu_ip_block {
227 	struct amdgpu_ip_block_status status;
228 	const struct amdgpu_ip_block_version *version;
229 };
230 
231 int amdgpu_ip_block_version_cmp(struct amdgpu_device *adev,
232 				enum amd_ip_block_type type,
233 				u32 major, u32 minor);
234 
235 struct amdgpu_ip_block * amdgpu_get_ip_block(struct amdgpu_device *adev,
236 					     enum amd_ip_block_type type);
237 
238 int amdgpu_ip_block_add(struct amdgpu_device *adev,
239 			const struct amdgpu_ip_block_version *ip_block_version);
240 
241 /* provided by hw blocks that can move/clear data.  e.g., gfx or sdma */
242 struct amdgpu_buffer_funcs {
243 	/* maximum bytes in a single operation */
244 	uint32_t	copy_max_bytes;
245 
246 	/* number of dw to reserve per operation */
247 	unsigned	copy_num_dw;
248 
249 	/* used for buffer migration */
250 	void (*emit_copy_buffer)(struct amdgpu_ib *ib,
251 				 /* src addr in bytes */
252 				 uint64_t src_offset,
253 				 /* dst addr in bytes */
254 				 uint64_t dst_offset,
255 				 /* number of byte to transfer */
256 				 uint32_t byte_count);
257 
258 	/* maximum bytes in a single operation */
259 	uint32_t	fill_max_bytes;
260 
261 	/* number of dw to reserve per operation */
262 	unsigned	fill_num_dw;
263 
264 	/* used for buffer clearing */
265 	void (*emit_fill_buffer)(struct amdgpu_ib *ib,
266 				 /* value to write to memory */
267 				 uint32_t src_data,
268 				 /* dst addr in bytes */
269 				 uint64_t dst_offset,
270 				 /* number of byte to fill */
271 				 uint32_t byte_count);
272 };
273 
274 /* provided by hw blocks that can write ptes, e.g., sdma */
275 struct amdgpu_vm_pte_funcs {
276 	/* copy pte entries from GART */
277 	void (*copy_pte)(struct amdgpu_ib *ib,
278 			 uint64_t pe, uint64_t src,
279 			 unsigned count);
280 	/* write pte one entry at a time with addr mapping */
281 	void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe,
282 			  uint64_t value, unsigned count,
283 			  uint32_t incr);
284 	/* for linear pte/pde updates without addr mapping */
285 	void (*set_pte_pde)(struct amdgpu_ib *ib,
286 			    uint64_t pe,
287 			    uint64_t addr, unsigned count,
288 			    uint32_t incr, uint64_t flags);
289 };
290 
291 /* provided by the gmc block */
292 struct amdgpu_gart_funcs {
293 	/* flush the vm tlb via mmio */
294 	void (*flush_gpu_tlb)(struct amdgpu_device *adev,
295 			      uint32_t vmid);
296 	/* write pte/pde updates using the cpu */
297 	int (*set_pte_pde)(struct amdgpu_device *adev,
298 			   void *cpu_pt_addr, /* cpu addr of page table */
299 			   uint32_t gpu_page_idx, /* pte/pde to update */
300 			   uint64_t addr, /* addr to write into pte/pde */
301 			   uint64_t flags); /* access flags */
302 	/* enable/disable PRT support */
303 	void (*set_prt)(struct amdgpu_device *adev, bool enable);
304 	/* set pte flags based per asic */
305 	uint64_t (*get_vm_pte_flags)(struct amdgpu_device *adev,
306 				     uint32_t flags);
307 	/* adjust mc addr in fb for APU case */
308 	u64 (*adjust_mc_addr)(struct amdgpu_device *adev, u64 addr);
309 	uint32_t (*get_invalidate_req)(unsigned int vm_id);
310 };
311 
312 /* provided by the ih block */
313 struct amdgpu_ih_funcs {
314 	/* ring read/write ptr handling, called from interrupt context */
315 	u32 (*get_wptr)(struct amdgpu_device *adev);
316 	void (*decode_iv)(struct amdgpu_device *adev,
317 			  struct amdgpu_iv_entry *entry);
318 	void (*set_rptr)(struct amdgpu_device *adev);
319 };
320 
321 /*
322  * BIOS.
323  */
324 bool amdgpu_get_bios(struct amdgpu_device *adev);
325 bool amdgpu_read_bios(struct amdgpu_device *adev);
326 
327 /*
328  * Dummy page
329  */
330 struct amdgpu_dummy_page {
331 	struct page	*page;
332 	dma_addr_t	addr;
333 };
334 int amdgpu_dummy_page_init(struct amdgpu_device *adev);
335 void amdgpu_dummy_page_fini(struct amdgpu_device *adev);
336 
337 
338 /*
339  * Clocks
340  */
341 
342 #define AMDGPU_MAX_PPLL 3
343 
344 struct amdgpu_clock {
345 	struct amdgpu_pll ppll[AMDGPU_MAX_PPLL];
346 	struct amdgpu_pll spll;
347 	struct amdgpu_pll mpll;
348 	/* 10 Khz units */
349 	uint32_t default_mclk;
350 	uint32_t default_sclk;
351 	uint32_t default_dispclk;
352 	uint32_t current_dispclk;
353 	uint32_t dp_extclk;
354 	uint32_t max_pixel_clock;
355 };
356 
357 /*
358  * BO.
359  */
360 struct amdgpu_bo_list_entry {
361 	struct amdgpu_bo		*robj;
362 	struct ttm_validate_buffer	tv;
363 	struct amdgpu_bo_va		*bo_va;
364 	uint32_t			priority;
365 	struct page			**user_pages;
366 	int				user_invalidated;
367 };
368 
369 struct amdgpu_bo_va_mapping {
370 	struct list_head		list;
371 	struct rb_node			rb;
372 	uint64_t			start;
373 	uint64_t			last;
374 	uint64_t			__subtree_last;
375 	uint64_t			offset;
376 	uint64_t			flags;
377 };
378 
379 /* bo virtual addresses in a specific vm */
380 struct amdgpu_bo_va {
381 	/* protected by bo being reserved */
382 	struct list_head		bo_list;
383 	struct dma_fence	        *last_pt_update;
384 	unsigned			ref_count;
385 
386 	/* protected by vm mutex and spinlock */
387 	struct list_head		vm_status;
388 
389 	/* mappings for this bo_va */
390 	struct list_head		invalids;
391 	struct list_head		valids;
392 
393 	/* constant after initialization */
394 	struct amdgpu_vm		*vm;
395 	struct amdgpu_bo		*bo;
396 };
397 
398 #define AMDGPU_GEM_DOMAIN_MAX		0x3
399 
400 struct amdgpu_bo {
401 	/* Protected by tbo.reserved */
402 	u32				prefered_domains;
403 	u32				allowed_domains;
404 	struct ttm_place		placements[AMDGPU_GEM_DOMAIN_MAX + 1];
405 	struct ttm_placement		placement;
406 	struct ttm_buffer_object	tbo;
407 	struct ttm_bo_kmap_obj		kmap;
408 	u64				flags;
409 	unsigned			pin_count;
410 	void				*kptr;
411 	u64				tiling_flags;
412 	u64				metadata_flags;
413 	void				*metadata;
414 	u32				metadata_size;
415 	unsigned			prime_shared_count;
416 	/* list of all virtual address to which this bo
417 	 * is associated to
418 	 */
419 	struct list_head		va;
420 	/* Constant after initialization */
421 	struct drm_gem_object		gem_base;
422 	struct amdgpu_bo		*parent;
423 	struct amdgpu_bo		*shadow;
424 
425 	struct ttm_bo_kmap_obj		dma_buf_vmap;
426 	struct amdgpu_mn		*mn;
427 	struct list_head		mn_list;
428 	struct list_head		shadow_list;
429 };
430 #define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, gem_base)
431 
432 void amdgpu_gem_object_free(struct drm_gem_object *obj);
433 int amdgpu_gem_object_open(struct drm_gem_object *obj,
434 				struct drm_file *file_priv);
435 void amdgpu_gem_object_close(struct drm_gem_object *obj,
436 				struct drm_file *file_priv);
437 unsigned long amdgpu_gem_timeout(uint64_t timeout_ns);
438 struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj);
439 struct drm_gem_object *
440 amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
441 				 struct dma_buf_attachment *attach,
442 				 struct sg_table *sg);
443 struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev,
444 					struct drm_gem_object *gobj,
445 					int flags);
446 int amdgpu_gem_prime_pin(struct drm_gem_object *obj);
447 void amdgpu_gem_prime_unpin(struct drm_gem_object *obj);
448 struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *);
449 void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj);
450 void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr);
451 int amdgpu_gem_debugfs_init(struct amdgpu_device *adev);
452 
453 /* sub-allocation manager, it has to be protected by another lock.
454  * By conception this is an helper for other part of the driver
455  * like the indirect buffer or semaphore, which both have their
456  * locking.
457  *
458  * Principe is simple, we keep a list of sub allocation in offset
459  * order (first entry has offset == 0, last entry has the highest
460  * offset).
461  *
462  * When allocating new object we first check if there is room at
463  * the end total_size - (last_object_offset + last_object_size) >=
464  * alloc_size. If so we allocate new object there.
465  *
466  * When there is not enough room at the end, we start waiting for
467  * each sub object until we reach object_offset+object_size >=
468  * alloc_size, this object then become the sub object we return.
469  *
470  * Alignment can't be bigger than page size.
471  *
472  * Hole are not considered for allocation to keep things simple.
473  * Assumption is that there won't be hole (all object on same
474  * alignment).
475  */
476 
477 #define AMDGPU_SA_NUM_FENCE_LISTS	32
478 
479 struct amdgpu_sa_manager {
480 	wait_queue_head_t	wq;
481 	struct amdgpu_bo	*bo;
482 	struct list_head	*hole;
483 	struct list_head	flist[AMDGPU_SA_NUM_FENCE_LISTS];
484 	struct list_head	olist;
485 	unsigned		size;
486 	uint64_t		gpu_addr;
487 	void			*cpu_ptr;
488 	uint32_t		domain;
489 	uint32_t		align;
490 };
491 
492 /* sub-allocation buffer */
493 struct amdgpu_sa_bo {
494 	struct list_head		olist;
495 	struct list_head		flist;
496 	struct amdgpu_sa_manager	*manager;
497 	unsigned			soffset;
498 	unsigned			eoffset;
499 	struct dma_fence	        *fence;
500 };
501 
502 /*
503  * GEM objects.
504  */
505 void amdgpu_gem_force_release(struct amdgpu_device *adev);
506 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
507 				int alignment, u32 initial_domain,
508 				u64 flags, bool kernel,
509 				struct drm_gem_object **obj);
510 
511 int amdgpu_mode_dumb_create(struct drm_file *file_priv,
512 			    struct drm_device *dev,
513 			    struct drm_mode_create_dumb *args);
514 int amdgpu_mode_dumb_mmap(struct drm_file *filp,
515 			  struct drm_device *dev,
516 			  uint32_t handle, uint64_t *offset_p);
517 int amdgpu_fence_slab_init(void);
518 void amdgpu_fence_slab_fini(void);
519 
520 /*
521  * GART structures, functions & helpers
522  */
523 struct amdgpu_mc;
524 
525 #define AMDGPU_GPU_PAGE_SIZE 4096
526 #define AMDGPU_GPU_PAGE_MASK (AMDGPU_GPU_PAGE_SIZE - 1)
527 #define AMDGPU_GPU_PAGE_SHIFT 12
528 #define AMDGPU_GPU_PAGE_ALIGN(a) (((a) + AMDGPU_GPU_PAGE_MASK) & ~AMDGPU_GPU_PAGE_MASK)
529 
530 struct amdgpu_gart {
531 	dma_addr_t			table_addr;
532 	struct amdgpu_bo		*robj;
533 	void				*ptr;
534 	unsigned			num_gpu_pages;
535 	unsigned			num_cpu_pages;
536 	unsigned			table_size;
537 #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
538 	struct page			**pages;
539 #endif
540 	bool				ready;
541 
542 	/* Asic default pte flags */
543 	uint64_t			gart_pte_flags;
544 
545 	const struct amdgpu_gart_funcs *gart_funcs;
546 };
547 
548 int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev);
549 void amdgpu_gart_table_ram_free(struct amdgpu_device *adev);
550 int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev);
551 void amdgpu_gart_table_vram_free(struct amdgpu_device *adev);
552 int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev);
553 void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev);
554 int amdgpu_gart_init(struct amdgpu_device *adev);
555 void amdgpu_gart_fini(struct amdgpu_device *adev);
556 void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
557 			int pages);
558 int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
559 		     int pages, struct page **pagelist,
560 		     dma_addr_t *dma_addr, uint64_t flags);
561 int amdgpu_ttm_recover_gart(struct amdgpu_device *adev);
562 
563 /*
564  * VMHUB structures, functions & helpers
565  */
566 struct amdgpu_vmhub {
567 	uint32_t	ctx0_ptb_addr_lo32;
568 	uint32_t	ctx0_ptb_addr_hi32;
569 	uint32_t	vm_inv_eng0_req;
570 	uint32_t	vm_inv_eng0_ack;
571 	uint32_t	vm_context0_cntl;
572 	uint32_t	vm_l2_pro_fault_status;
573 	uint32_t	vm_l2_pro_fault_cntl;
574 };
575 
576 /*
577  * GPU MC structures, functions & helpers
578  */
579 struct amdgpu_mc {
580 	resource_size_t		aper_size;
581 	resource_size_t		aper_base;
582 	resource_size_t		agp_base;
583 	/* for some chips with <= 32MB we need to lie
584 	 * about vram size near mc fb location */
585 	u64			mc_vram_size;
586 	u64			visible_vram_size;
587 	u64			gtt_size;
588 	u64			gtt_start;
589 	u64			gtt_end;
590 	u64			vram_start;
591 	u64			vram_end;
592 	unsigned		vram_width;
593 	u64			real_vram_size;
594 	int			vram_mtrr;
595 	u64                     gtt_base_align;
596 	u64                     mc_mask;
597 	const struct firmware   *fw;	/* MC firmware */
598 	uint32_t                fw_version;
599 	struct amdgpu_irq_src	vm_fault;
600 	uint32_t		vram_type;
601 	uint32_t                srbm_soft_reset;
602 	struct amdgpu_mode_mc_save save;
603 	bool			prt_warning;
604 	/* apertures */
605 	u64					shared_aperture_start;
606 	u64					shared_aperture_end;
607 	u64					private_aperture_start;
608 	u64					private_aperture_end;
609 	/* protects concurrent invalidation */
610 	spinlock_t		invalidate_lock;
611 };
612 
613 /*
614  * GPU doorbell structures, functions & helpers
615  */
616 typedef enum _AMDGPU_DOORBELL_ASSIGNMENT
617 {
618 	AMDGPU_DOORBELL_KIQ                     = 0x000,
619 	AMDGPU_DOORBELL_HIQ                     = 0x001,
620 	AMDGPU_DOORBELL_DIQ                     = 0x002,
621 	AMDGPU_DOORBELL_MEC_RING0               = 0x010,
622 	AMDGPU_DOORBELL_MEC_RING1               = 0x011,
623 	AMDGPU_DOORBELL_MEC_RING2               = 0x012,
624 	AMDGPU_DOORBELL_MEC_RING3               = 0x013,
625 	AMDGPU_DOORBELL_MEC_RING4               = 0x014,
626 	AMDGPU_DOORBELL_MEC_RING5               = 0x015,
627 	AMDGPU_DOORBELL_MEC_RING6               = 0x016,
628 	AMDGPU_DOORBELL_MEC_RING7               = 0x017,
629 	AMDGPU_DOORBELL_GFX_RING0               = 0x020,
630 	AMDGPU_DOORBELL_sDMA_ENGINE0            = 0x1E0,
631 	AMDGPU_DOORBELL_sDMA_ENGINE1            = 0x1E1,
632 	AMDGPU_DOORBELL_IH                      = 0x1E8,
633 	AMDGPU_DOORBELL_MAX_ASSIGNMENT          = 0x3FF,
634 	AMDGPU_DOORBELL_INVALID                 = 0xFFFF
635 } AMDGPU_DOORBELL_ASSIGNMENT;
636 
637 struct amdgpu_doorbell {
638 	/* doorbell mmio */
639 	resource_size_t		base;
640 	resource_size_t		size;
641 	u32 __iomem		*ptr;
642 	u32			num_doorbells;	/* Number of doorbells actually reserved for amdgpu. */
643 };
644 
645 /*
646  * 64bit doorbell, offset are in QWORD, occupy 2KB doorbell space
647  */
648 typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
649 {
650 	/*
651 	 * All compute related doorbells: kiq, hiq, diq, traditional compute queue, user queue, should locate in
652 	 * a continues range so that programming CP_MEC_DOORBELL_RANGE_LOWER/UPPER can cover this range.
653 	 *  Compute related doorbells are allocated from 0x00 to 0x8a
654 	 */
655 
656 
657 	/* kernel scheduling */
658 	AMDGPU_DOORBELL64_KIQ                     = 0x00,
659 
660 	/* HSA interface queue and debug queue */
661 	AMDGPU_DOORBELL64_HIQ                     = 0x01,
662 	AMDGPU_DOORBELL64_DIQ                     = 0x02,
663 
664 	/* Compute engines */
665 	AMDGPU_DOORBELL64_MEC_RING0               = 0x03,
666 	AMDGPU_DOORBELL64_MEC_RING1               = 0x04,
667 	AMDGPU_DOORBELL64_MEC_RING2               = 0x05,
668 	AMDGPU_DOORBELL64_MEC_RING3               = 0x06,
669 	AMDGPU_DOORBELL64_MEC_RING4               = 0x07,
670 	AMDGPU_DOORBELL64_MEC_RING5               = 0x08,
671 	AMDGPU_DOORBELL64_MEC_RING6               = 0x09,
672 	AMDGPU_DOORBELL64_MEC_RING7               = 0x0a,
673 
674 	/* User queue doorbell range (128 doorbells) */
675 	AMDGPU_DOORBELL64_USERQUEUE_START         = 0x0b,
676 	AMDGPU_DOORBELL64_USERQUEUE_END           = 0x8a,
677 
678 	/* Graphics engine */
679 	AMDGPU_DOORBELL64_GFX_RING0               = 0x8b,
680 
681 	/*
682 	 * Other graphics doorbells can be allocated here: from 0x8c to 0xef
683 	 * Graphics voltage island aperture 1
684 	 * default non-graphics QWORD index is 0xF0 - 0xFF inclusive
685 	 */
686 
687 	/* sDMA engines */
688 	AMDGPU_DOORBELL64_sDMA_ENGINE0            = 0xF0,
689 	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0     = 0xF1,
690 	AMDGPU_DOORBELL64_sDMA_ENGINE1            = 0xF2,
691 	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1     = 0xF3,
692 
693 	/* Interrupt handler */
694 	AMDGPU_DOORBELL64_IH                      = 0xF4,  /* For legacy interrupt ring buffer */
695 	AMDGPU_DOORBELL64_IH_RING1                = 0xF5,  /* For page migration request log */
696 	AMDGPU_DOORBELL64_IH_RING2                = 0xF6,  /* For page migration translation/invalidation log */
697 
698 	/* VCN engine use 32 bits doorbell  */
699 	AMDGPU_DOORBELL64_VCN0_1                  = 0xF8, /* lower 32 bits for VNC0 and upper 32 bits for VNC1 */
700 	AMDGPU_DOORBELL64_VCN2_3                  = 0xF9,
701 	AMDGPU_DOORBELL64_VCN4_5                  = 0xFA,
702 	AMDGPU_DOORBELL64_VCN6_7                  = 0xFB,
703 
704 	/* overlap the doorbell assignment with VCN as they are  mutually exclusive
705 	 * VCE engine's doorbell is 32 bit and two VCE ring share one QWORD
706 	 */
707 	AMDGPU_DOORBELL64_RING0_1                 = 0xF8,
708 	AMDGPU_DOORBELL64_RING2_3                 = 0xF9,
709 	AMDGPU_DOORBELL64_RING4_5                 = 0xFA,
710 	AMDGPU_DOORBELL64_RING6_7                 = 0xFB,
711 
712 	AMDGPU_DOORBELL64_UVD_RING0_1             = 0xFC,
713 	AMDGPU_DOORBELL64_UVD_RING2_3             = 0xFD,
714 	AMDGPU_DOORBELL64_UVD_RING4_5             = 0xFE,
715 	AMDGPU_DOORBELL64_UVD_RING6_7             = 0xFF,
716 
717 	AMDGPU_DOORBELL64_MAX_ASSIGNMENT          = 0xFF,
718 	AMDGPU_DOORBELL64_INVALID                 = 0xFFFF
719 } AMDGPU_DOORBELL64_ASSIGNMENT;
720 
721 
722 void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev,
723 				phys_addr_t *aperture_base,
724 				size_t *aperture_size,
725 				size_t *start_offset);
726 
727 /*
728  * IRQS.
729  */
730 
731 struct amdgpu_flip_work {
732 	struct delayed_work		flip_work;
733 	struct work_struct		unpin_work;
734 	struct amdgpu_device		*adev;
735 	int				crtc_id;
736 	u32				target_vblank;
737 	uint64_t			base;
738 	struct drm_pending_vblank_event *event;
739 	struct amdgpu_bo		*old_abo;
740 	struct dma_fence		*excl;
741 	unsigned			shared_count;
742 	struct dma_fence		**shared;
743 	struct dma_fence_cb		cb;
744 	bool				async;
745 };
746 
747 
748 /*
749  * CP & rings.
750  */
751 
752 struct amdgpu_ib {
753 	struct amdgpu_sa_bo		*sa_bo;
754 	uint32_t			length_dw;
755 	uint64_t			gpu_addr;
756 	uint32_t			*ptr;
757 	uint32_t			flags;
758 };
759 
760 extern const struct amd_sched_backend_ops amdgpu_sched_ops;
761 
762 int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs,
763 		     struct amdgpu_job **job, struct amdgpu_vm *vm);
764 int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size,
765 			     struct amdgpu_job **job);
766 
767 void amdgpu_job_free_resources(struct amdgpu_job *job);
768 void amdgpu_job_free(struct amdgpu_job *job);
769 int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
770 		      struct amd_sched_entity *entity, void *owner,
771 		      struct dma_fence **f);
772 
773 /*
774  * context related structures
775  */
776 
777 struct amdgpu_ctx_ring {
778 	uint64_t		sequence;
779 	struct dma_fence	**fences;
780 	struct amd_sched_entity	entity;
781 };
782 
783 struct amdgpu_ctx {
784 	struct kref		refcount;
785 	struct amdgpu_device    *adev;
786 	unsigned		reset_counter;
787 	spinlock_t		ring_lock;
788 	struct dma_fence	**fences;
789 	struct amdgpu_ctx_ring	rings[AMDGPU_MAX_RINGS];
790 	bool preamble_presented;
791 };
792 
793 struct amdgpu_ctx_mgr {
794 	struct amdgpu_device	*adev;
795 	struct mutex		lock;
796 	/* protected by lock */
797 	struct idr		ctx_handles;
798 };
799 
800 struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id);
801 int amdgpu_ctx_put(struct amdgpu_ctx *ctx);
802 
803 uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
804 			      struct dma_fence *fence);
805 struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
806 				   struct amdgpu_ring *ring, uint64_t seq);
807 
808 int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
809 		     struct drm_file *filp);
810 
811 void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr);
812 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
813 
814 /*
815  * file private structure
816  */
817 
818 struct amdgpu_fpriv {
819 	struct amdgpu_vm	vm;
820 	struct amdgpu_bo_va	*prt_va;
821 	struct mutex		bo_list_lock;
822 	struct idr		bo_list_handles;
823 	struct amdgpu_ctx_mgr	ctx_mgr;
824 };
825 
826 /*
827  * residency list
828  */
829 
830 struct amdgpu_bo_list {
831 	struct mutex lock;
832 	struct amdgpu_bo *gds_obj;
833 	struct amdgpu_bo *gws_obj;
834 	struct amdgpu_bo *oa_obj;
835 	unsigned first_userptr;
836 	unsigned num_entries;
837 	struct amdgpu_bo_list_entry *array;
838 };
839 
840 struct amdgpu_bo_list *
841 amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id);
842 void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list,
843 			     struct list_head *validated);
844 void amdgpu_bo_list_put(struct amdgpu_bo_list *list);
845 void amdgpu_bo_list_free(struct amdgpu_bo_list *list);
846 
847 /*
848  * GFX stuff
849  */
850 #include "clearstate_defs.h"
851 
852 struct amdgpu_rlc_funcs {
853 	void (*enter_safe_mode)(struct amdgpu_device *adev);
854 	void (*exit_safe_mode)(struct amdgpu_device *adev);
855 };
856 
857 struct amdgpu_rlc {
858 	/* for power gating */
859 	struct amdgpu_bo	*save_restore_obj;
860 	uint64_t		save_restore_gpu_addr;
861 	volatile uint32_t	*sr_ptr;
862 	const u32               *reg_list;
863 	u32                     reg_list_size;
864 	/* for clear state */
865 	struct amdgpu_bo	*clear_state_obj;
866 	uint64_t		clear_state_gpu_addr;
867 	volatile uint32_t	*cs_ptr;
868 	const struct cs_section_def   *cs_data;
869 	u32                     clear_state_size;
870 	/* for cp tables */
871 	struct amdgpu_bo	*cp_table_obj;
872 	uint64_t		cp_table_gpu_addr;
873 	volatile uint32_t	*cp_table_ptr;
874 	u32                     cp_table_size;
875 
876 	/* safe mode for updating CG/PG state */
877 	bool in_safe_mode;
878 	const struct amdgpu_rlc_funcs *funcs;
879 
880 	/* for firmware data */
881 	u32 save_and_restore_offset;
882 	u32 clear_state_descriptor_offset;
883 	u32 avail_scratch_ram_locations;
884 	u32 reg_restore_list_size;
885 	u32 reg_list_format_start;
886 	u32 reg_list_format_separate_start;
887 	u32 starting_offsets_start;
888 	u32 reg_list_format_size_bytes;
889 	u32 reg_list_size_bytes;
890 
891 	u32 *register_list_format;
892 	u32 *register_restore;
893 };
894 
895 struct amdgpu_mec {
896 	struct amdgpu_bo	*hpd_eop_obj;
897 	u64			hpd_eop_gpu_addr;
898 	struct amdgpu_bo	*mec_fw_obj;
899 	u64			mec_fw_gpu_addr;
900 	u32 num_pipe;
901 	u32 num_mec;
902 	u32 num_queue;
903 	void			*mqd_backup[AMDGPU_MAX_COMPUTE_RINGS + 1];
904 };
905 
906 struct amdgpu_kiq {
907 	u64			eop_gpu_addr;
908 	struct amdgpu_bo	*eop_obj;
909 	struct amdgpu_ring	ring;
910 	struct amdgpu_irq_src	irq;
911 };
912 
913 /*
914  * GPU scratch registers structures, functions & helpers
915  */
916 struct amdgpu_scratch {
917 	unsigned		num_reg;
918 	uint32_t                reg_base;
919 	uint32_t		free_mask;
920 };
921 
922 /*
923  * GFX configurations
924  */
925 #define AMDGPU_GFX_MAX_SE 4
926 #define AMDGPU_GFX_MAX_SH_PER_SE 2
927 
928 struct amdgpu_rb_config {
929 	uint32_t rb_backend_disable;
930 	uint32_t user_rb_backend_disable;
931 	uint32_t raster_config;
932 	uint32_t raster_config_1;
933 };
934 
935 struct gb_addr_config {
936 	uint16_t pipe_interleave_size;
937 	uint8_t num_pipes;
938 	uint8_t max_compress_frags;
939 	uint8_t num_banks;
940 	uint8_t num_se;
941 	uint8_t num_rb_per_se;
942 };
943 
944 struct amdgpu_gfx_config {
945 	unsigned max_shader_engines;
946 	unsigned max_tile_pipes;
947 	unsigned max_cu_per_sh;
948 	unsigned max_sh_per_se;
949 	unsigned max_backends_per_se;
950 	unsigned max_texture_channel_caches;
951 	unsigned max_gprs;
952 	unsigned max_gs_threads;
953 	unsigned max_hw_contexts;
954 	unsigned sc_prim_fifo_size_frontend;
955 	unsigned sc_prim_fifo_size_backend;
956 	unsigned sc_hiz_tile_fifo_size;
957 	unsigned sc_earlyz_tile_fifo_size;
958 
959 	unsigned num_tile_pipes;
960 	unsigned backend_enable_mask;
961 	unsigned mem_max_burst_length_bytes;
962 	unsigned mem_row_size_in_kb;
963 	unsigned shader_engine_tile_size;
964 	unsigned num_gpus;
965 	unsigned multi_gpu_tile_size;
966 	unsigned mc_arb_ramcfg;
967 	unsigned gb_addr_config;
968 	unsigned num_rbs;
969 
970 	uint32_t tile_mode_array[32];
971 	uint32_t macrotile_mode_array[16];
972 
973 	struct gb_addr_config gb_addr_config_fields;
974 	struct amdgpu_rb_config rb_config[AMDGPU_GFX_MAX_SE][AMDGPU_GFX_MAX_SH_PER_SE];
975 
976 	/* gfx configure feature */
977 	uint32_t double_offchip_lds_buf;
978 };
979 
980 struct amdgpu_cu_info {
981 	uint32_t number; /* total active CU number */
982 	uint32_t ao_cu_mask;
983 	uint32_t bitmap[4][4];
984 };
985 
986 struct amdgpu_gfx_funcs {
987 	/* get the gpu clock counter */
988 	uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev);
989 	void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance);
990 	void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields);
991 	void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t thread, uint32_t start, uint32_t size, uint32_t *dst);
992 	void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t start, uint32_t size, uint32_t *dst);
993 };
994 
995 struct amdgpu_ngg_buf {
996 	struct amdgpu_bo	*bo;
997 	uint64_t		gpu_addr;
998 	uint32_t		size;
999 	uint32_t		bo_size;
1000 };
1001 
1002 enum {
1003 	PRIM = 0,
1004 	POS,
1005 	CNTL,
1006 	PARAM,
1007 	NGG_BUF_MAX
1008 };
1009 
1010 struct amdgpu_ngg {
1011 	struct amdgpu_ngg_buf	buf[NGG_BUF_MAX];
1012 	uint32_t		gds_reserve_addr;
1013 	uint32_t		gds_reserve_size;
1014 	bool			init;
1015 };
1016 
1017 struct amdgpu_gfx {
1018 	struct mutex			gpu_clock_mutex;
1019 	struct amdgpu_gfx_config	config;
1020 	struct amdgpu_rlc		rlc;
1021 	struct amdgpu_mec		mec;
1022 	struct amdgpu_kiq		kiq;
1023 	struct amdgpu_scratch		scratch;
1024 	const struct firmware		*me_fw;	/* ME firmware */
1025 	uint32_t			me_fw_version;
1026 	const struct firmware		*pfp_fw; /* PFP firmware */
1027 	uint32_t			pfp_fw_version;
1028 	const struct firmware		*ce_fw;	/* CE firmware */
1029 	uint32_t			ce_fw_version;
1030 	const struct firmware		*rlc_fw; /* RLC firmware */
1031 	uint32_t			rlc_fw_version;
1032 	const struct firmware		*mec_fw; /* MEC firmware */
1033 	uint32_t			mec_fw_version;
1034 	const struct firmware		*mec2_fw; /* MEC2 firmware */
1035 	uint32_t			mec2_fw_version;
1036 	uint32_t			me_feature_version;
1037 	uint32_t			ce_feature_version;
1038 	uint32_t			pfp_feature_version;
1039 	uint32_t			rlc_feature_version;
1040 	uint32_t			mec_feature_version;
1041 	uint32_t			mec2_feature_version;
1042 	struct amdgpu_ring		gfx_ring[AMDGPU_MAX_GFX_RINGS];
1043 	unsigned			num_gfx_rings;
1044 	struct amdgpu_ring		compute_ring[AMDGPU_MAX_COMPUTE_RINGS];
1045 	unsigned			num_compute_rings;
1046 	struct amdgpu_irq_src		eop_irq;
1047 	struct amdgpu_irq_src		priv_reg_irq;
1048 	struct amdgpu_irq_src		priv_inst_irq;
1049 	/* gfx status */
1050 	uint32_t			gfx_current_status;
1051 	/* ce ram size*/
1052 	unsigned			ce_ram_size;
1053 	struct amdgpu_cu_info		cu_info;
1054 	const struct amdgpu_gfx_funcs	*funcs;
1055 
1056 	/* reset mask */
1057 	uint32_t                        grbm_soft_reset;
1058 	uint32_t                        srbm_soft_reset;
1059 	bool                            in_reset;
1060 	/* NGG */
1061 	struct amdgpu_ngg		ngg;
1062 };
1063 
1064 int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
1065 		  unsigned size, struct amdgpu_ib *ib);
1066 void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
1067 		    struct dma_fence *f);
1068 int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
1069 		       struct amdgpu_ib *ibs, struct amdgpu_job *job,
1070 		       struct dma_fence **f);
1071 int amdgpu_ib_pool_init(struct amdgpu_device *adev);
1072 void amdgpu_ib_pool_fini(struct amdgpu_device *adev);
1073 int amdgpu_ib_ring_tests(struct amdgpu_device *adev);
1074 
1075 /*
1076  * CS.
1077  */
1078 struct amdgpu_cs_chunk {
1079 	uint32_t		chunk_id;
1080 	uint32_t		length_dw;
1081 	void			*kdata;
1082 };
1083 
1084 struct amdgpu_cs_parser {
1085 	struct amdgpu_device	*adev;
1086 	struct drm_file		*filp;
1087 	struct amdgpu_ctx	*ctx;
1088 
1089 	/* chunks */
1090 	unsigned		nchunks;
1091 	struct amdgpu_cs_chunk	*chunks;
1092 
1093 	/* scheduler job object */
1094 	struct amdgpu_job	*job;
1095 
1096 	/* buffer objects */
1097 	struct ww_acquire_ctx		ticket;
1098 	struct amdgpu_bo_list		*bo_list;
1099 	struct amdgpu_bo_list_entry	vm_pd;
1100 	struct list_head		validated;
1101 	struct dma_fence		*fence;
1102 	uint64_t			bytes_moved_threshold;
1103 	uint64_t			bytes_moved;
1104 	struct amdgpu_bo_list_entry	*evictable;
1105 
1106 	/* user fence */
1107 	struct amdgpu_bo_list_entry	uf_entry;
1108 };
1109 
1110 #define AMDGPU_PREAMBLE_IB_PRESENT          (1 << 0) /* bit set means command submit involves a preamble IB */
1111 #define AMDGPU_PREAMBLE_IB_PRESENT_FIRST    (1 << 1) /* bit set means preamble IB is first presented in belonging context */
1112 #define AMDGPU_HAVE_CTX_SWITCH              (1 << 2) /* bit set means context switch occured */
1113 #define AMDGPU_VM_DOMAIN                    (1 << 3) /* bit set means in virtual memory context */
1114 
1115 struct amdgpu_job {
1116 	struct amd_sched_job    base;
1117 	struct amdgpu_device	*adev;
1118 	struct amdgpu_vm	*vm;
1119 	struct amdgpu_ring	*ring;
1120 	struct amdgpu_sync	sync;
1121 	struct amdgpu_ib	*ibs;
1122 	struct dma_fence	*fence; /* the hw fence */
1123 	uint32_t		preamble_status;
1124 	uint32_t		num_ibs;
1125 	void			*owner;
1126 	uint64_t		fence_ctx; /* the fence_context this job uses */
1127 	bool                    vm_needs_flush;
1128 	unsigned		vm_id;
1129 	uint64_t		vm_pd_addr;
1130 	uint32_t		gds_base, gds_size;
1131 	uint32_t		gws_base, gws_size;
1132 	uint32_t		oa_base, oa_size;
1133 
1134 	/* user fence handling */
1135 	uint64_t		uf_addr;
1136 	uint64_t		uf_sequence;
1137 
1138 };
1139 #define to_amdgpu_job(sched_job)		\
1140 		container_of((sched_job), struct amdgpu_job, base)
1141 
1142 static inline u32 amdgpu_get_ib_value(struct amdgpu_cs_parser *p,
1143 				      uint32_t ib_idx, int idx)
1144 {
1145 	return p->job->ibs[ib_idx].ptr[idx];
1146 }
1147 
1148 static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,
1149 				       uint32_t ib_idx, int idx,
1150 				       uint32_t value)
1151 {
1152 	p->job->ibs[ib_idx].ptr[idx] = value;
1153 }
1154 
1155 /*
1156  * Writeback
1157  */
1158 #define AMDGPU_MAX_WB 1024	/* Reserve at most 1024 WB slots for amdgpu-owned rings. */
1159 
1160 struct amdgpu_wb {
1161 	struct amdgpu_bo	*wb_obj;
1162 	volatile uint32_t	*wb;
1163 	uint64_t		gpu_addr;
1164 	u32			num_wb;	/* Number of wb slots actually reserved for amdgpu. */
1165 	unsigned long		used[DIV_ROUND_UP(AMDGPU_MAX_WB, BITS_PER_LONG)];
1166 };
1167 
1168 int amdgpu_wb_get(struct amdgpu_device *adev, u32 *wb);
1169 void amdgpu_wb_free(struct amdgpu_device *adev, u32 wb);
1170 int amdgpu_wb_get_64bit(struct amdgpu_device *adev, u32 *wb);
1171 void amdgpu_wb_free_64bit(struct amdgpu_device *adev, u32 wb);
1172 
1173 void amdgpu_get_pcie_info(struct amdgpu_device *adev);
1174 
1175 /*
1176  * SDMA
1177  */
1178 struct amdgpu_sdma_instance {
1179 	/* SDMA firmware */
1180 	const struct firmware	*fw;
1181 	uint32_t		fw_version;
1182 	uint32_t		feature_version;
1183 
1184 	struct amdgpu_ring	ring;
1185 	bool			burst_nop;
1186 };
1187 
1188 struct amdgpu_sdma {
1189 	struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES];
1190 #ifdef CONFIG_DRM_AMDGPU_SI
1191 	//SI DMA has a difference trap irq number for the second engine
1192 	struct amdgpu_irq_src	trap_irq_1;
1193 #endif
1194 	struct amdgpu_irq_src	trap_irq;
1195 	struct amdgpu_irq_src	illegal_inst_irq;
1196 	int			num_instances;
1197 	uint32_t                    srbm_soft_reset;
1198 };
1199 
1200 /*
1201  * Firmware
1202  */
1203 enum amdgpu_firmware_load_type {
1204 	AMDGPU_FW_LOAD_DIRECT = 0,
1205 	AMDGPU_FW_LOAD_SMU,
1206 	AMDGPU_FW_LOAD_PSP,
1207 };
1208 
1209 struct amdgpu_firmware {
1210 	struct amdgpu_firmware_info ucode[AMDGPU_UCODE_ID_MAXIMUM];
1211 	enum amdgpu_firmware_load_type load_type;
1212 	struct amdgpu_bo *fw_buf;
1213 	unsigned int fw_size;
1214 	unsigned int max_ucodes;
1215 	/* firmwares are loaded by psp instead of smu from vega10 */
1216 	const struct amdgpu_psp_funcs *funcs;
1217 	struct amdgpu_bo *rbuf;
1218 	struct mutex mutex;
1219 };
1220 
1221 /*
1222  * Benchmarking
1223  */
1224 void amdgpu_benchmark(struct amdgpu_device *adev, int test_number);
1225 
1226 
1227 /*
1228  * Testing
1229  */
1230 void amdgpu_test_moves(struct amdgpu_device *adev);
1231 
1232 /*
1233  * MMU Notifier
1234  */
1235 #if defined(CONFIG_MMU_NOTIFIER)
1236 int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
1237 void amdgpu_mn_unregister(struct amdgpu_bo *bo);
1238 #else
1239 static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
1240 {
1241 	return -ENODEV;
1242 }
1243 static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {}
1244 #endif
1245 
1246 /*
1247  * Debugfs
1248  */
1249 struct amdgpu_debugfs {
1250 	const struct drm_info_list	*files;
1251 	unsigned		num_files;
1252 };
1253 
1254 int amdgpu_debugfs_add_files(struct amdgpu_device *adev,
1255 			     const struct drm_info_list *files,
1256 			     unsigned nfiles);
1257 int amdgpu_debugfs_fence_init(struct amdgpu_device *adev);
1258 
1259 #if defined(CONFIG_DEBUG_FS)
1260 int amdgpu_debugfs_init(struct drm_minor *minor);
1261 #endif
1262 
1263 int amdgpu_debugfs_firmware_init(struct amdgpu_device *adev);
1264 
1265 /*
1266  * amdgpu smumgr functions
1267  */
1268 struct amdgpu_smumgr_funcs {
1269 	int (*check_fw_load_finish)(struct amdgpu_device *adev, uint32_t fwtype);
1270 	int (*request_smu_load_fw)(struct amdgpu_device *adev);
1271 	int (*request_smu_specific_fw)(struct amdgpu_device *adev, uint32_t fwtype);
1272 };
1273 
1274 /*
1275  * amdgpu smumgr
1276  */
1277 struct amdgpu_smumgr {
1278 	struct amdgpu_bo *toc_buf;
1279 	struct amdgpu_bo *smu_buf;
1280 	/* asic priv smu data */
1281 	void *priv;
1282 	spinlock_t smu_lock;
1283 	/* smumgr functions */
1284 	const struct amdgpu_smumgr_funcs *smumgr_funcs;
1285 	/* ucode loading complete flag */
1286 	uint32_t fw_flags;
1287 };
1288 
1289 /*
1290  * ASIC specific register table accessible by UMD
1291  */
1292 struct amdgpu_allowed_register_entry {
1293 	uint32_t reg_offset;
1294 	bool untouched;
1295 	bool grbm_indexed;
1296 };
1297 
1298 /*
1299  * ASIC specific functions.
1300  */
1301 struct amdgpu_asic_funcs {
1302 	bool (*read_disabled_bios)(struct amdgpu_device *adev);
1303 	bool (*read_bios_from_rom)(struct amdgpu_device *adev,
1304 				   u8 *bios, u32 length_bytes);
1305 	int (*read_register)(struct amdgpu_device *adev, u32 se_num,
1306 			     u32 sh_num, u32 reg_offset, u32 *value);
1307 	void (*set_vga_state)(struct amdgpu_device *adev, bool state);
1308 	int (*reset)(struct amdgpu_device *adev);
1309 	/* get the reference clock */
1310 	u32 (*get_xclk)(struct amdgpu_device *adev);
1311 	/* MM block clocks */
1312 	int (*set_uvd_clocks)(struct amdgpu_device *adev, u32 vclk, u32 dclk);
1313 	int (*set_vce_clocks)(struct amdgpu_device *adev, u32 evclk, u32 ecclk);
1314 	/* static power management */
1315 	int (*get_pcie_lanes)(struct amdgpu_device *adev);
1316 	void (*set_pcie_lanes)(struct amdgpu_device *adev, int lanes);
1317 	/* get config memsize register */
1318 	u32 (*get_config_memsize)(struct amdgpu_device *adev);
1319 };
1320 
1321 /*
1322  * IOCTL.
1323  */
1324 int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
1325 			    struct drm_file *filp);
1326 int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data,
1327 				struct drm_file *filp);
1328 
1329 int amdgpu_gem_info_ioctl(struct drm_device *dev, void *data,
1330 			  struct drm_file *filp);
1331 int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
1332 			struct drm_file *filp);
1333 int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data,
1334 			  struct drm_file *filp);
1335 int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
1336 			      struct drm_file *filp);
1337 int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
1338 			  struct drm_file *filp);
1339 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
1340 			struct drm_file *filp);
1341 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
1342 int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
1343 int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
1344 				struct drm_file *filp);
1345 
1346 int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
1347 				struct drm_file *filp);
1348 
1349 /* VRAM scratch page for HDP bug, default vram page */
1350 struct amdgpu_vram_scratch {
1351 	struct amdgpu_bo		*robj;
1352 	volatile uint32_t		*ptr;
1353 	u64				gpu_addr;
1354 };
1355 
1356 /*
1357  * ACPI
1358  */
1359 struct amdgpu_atif_notification_cfg {
1360 	bool enabled;
1361 	int command_code;
1362 };
1363 
1364 struct amdgpu_atif_notifications {
1365 	bool display_switch;
1366 	bool expansion_mode_change;
1367 	bool thermal_state;
1368 	bool forced_power_state;
1369 	bool system_power_state;
1370 	bool display_conf_change;
1371 	bool px_gfx_switch;
1372 	bool brightness_change;
1373 	bool dgpu_display_event;
1374 };
1375 
1376 struct amdgpu_atif_functions {
1377 	bool system_params;
1378 	bool sbios_requests;
1379 	bool select_active_disp;
1380 	bool lid_state;
1381 	bool get_tv_standard;
1382 	bool set_tv_standard;
1383 	bool get_panel_expansion_mode;
1384 	bool set_panel_expansion_mode;
1385 	bool temperature_change;
1386 	bool graphics_device_types;
1387 };
1388 
1389 struct amdgpu_atif {
1390 	struct amdgpu_atif_notifications notifications;
1391 	struct amdgpu_atif_functions functions;
1392 	struct amdgpu_atif_notification_cfg notification_cfg;
1393 	struct amdgpu_encoder *encoder_for_bl;
1394 };
1395 
1396 struct amdgpu_atcs_functions {
1397 	bool get_ext_state;
1398 	bool pcie_perf_req;
1399 	bool pcie_dev_rdy;
1400 	bool pcie_bus_width;
1401 };
1402 
1403 struct amdgpu_atcs {
1404 	struct amdgpu_atcs_functions functions;
1405 };
1406 
1407 /*
1408  * CGS
1409  */
1410 struct cgs_device *amdgpu_cgs_create_device(struct amdgpu_device *adev);
1411 void amdgpu_cgs_destroy_device(struct cgs_device *cgs_device);
1412 
1413 /*
1414  * Core structure, functions and helpers.
1415  */
1416 typedef uint32_t (*amdgpu_rreg_t)(struct amdgpu_device*, uint32_t);
1417 typedef void (*amdgpu_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
1418 
1419 typedef uint32_t (*amdgpu_block_rreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
1420 typedef void (*amdgpu_block_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t, uint32_t);
1421 
1422 struct amdgpu_device {
1423 	struct device			*dev;
1424 	struct drm_device		*ddev;
1425 	struct pci_dev			*pdev;
1426 
1427 #ifdef CONFIG_DRM_AMD_ACP
1428 	struct amdgpu_acp		acp;
1429 #endif
1430 
1431 	/* ASIC */
1432 	enum amd_asic_type		asic_type;
1433 	uint32_t			family;
1434 	uint32_t			rev_id;
1435 	uint32_t			external_rev_id;
1436 	unsigned long			flags;
1437 	int				usec_timeout;
1438 	const struct amdgpu_asic_funcs	*asic_funcs;
1439 	bool				shutdown;
1440 	bool				need_dma32;
1441 	bool				accel_working;
1442 	struct work_struct		reset_work;
1443 	struct notifier_block		acpi_nb;
1444 	struct amdgpu_i2c_chan		*i2c_bus[AMDGPU_MAX_I2C_BUS];
1445 	struct amdgpu_debugfs		debugfs[AMDGPU_DEBUGFS_MAX_COMPONENTS];
1446 	unsigned			debugfs_count;
1447 #if defined(CONFIG_DEBUG_FS)
1448 	struct dentry			*debugfs_regs[AMDGPU_DEBUGFS_MAX_COMPONENTS];
1449 #endif
1450 	struct amdgpu_atif		atif;
1451 	struct amdgpu_atcs		atcs;
1452 	struct mutex			srbm_mutex;
1453 	/* GRBM index mutex. Protects concurrent access to GRBM index */
1454 	struct mutex                    grbm_idx_mutex;
1455 	struct dev_pm_domain		vga_pm_domain;
1456 	bool				have_disp_power_ref;
1457 
1458 	/* BIOS */
1459 	bool				is_atom_fw;
1460 	uint8_t				*bios;
1461 	uint32_t			bios_size;
1462 	struct amdgpu_bo		*stollen_vga_memory;
1463 	uint32_t			bios_scratch_reg_offset;
1464 	uint32_t			bios_scratch[AMDGPU_BIOS_NUM_SCRATCH];
1465 
1466 	/* Register/doorbell mmio */
1467 	resource_size_t			rmmio_base;
1468 	resource_size_t			rmmio_size;
1469 	void __iomem			*rmmio;
1470 	/* protects concurrent MM_INDEX/DATA based register access */
1471 	spinlock_t mmio_idx_lock;
1472 	/* protects concurrent SMC based register access */
1473 	spinlock_t smc_idx_lock;
1474 	amdgpu_rreg_t			smc_rreg;
1475 	amdgpu_wreg_t			smc_wreg;
1476 	/* protects concurrent PCIE register access */
1477 	spinlock_t pcie_idx_lock;
1478 	amdgpu_rreg_t			pcie_rreg;
1479 	amdgpu_wreg_t			pcie_wreg;
1480 	amdgpu_rreg_t			pciep_rreg;
1481 	amdgpu_wreg_t			pciep_wreg;
1482 	/* protects concurrent UVD register access */
1483 	spinlock_t uvd_ctx_idx_lock;
1484 	amdgpu_rreg_t			uvd_ctx_rreg;
1485 	amdgpu_wreg_t			uvd_ctx_wreg;
1486 	/* protects concurrent DIDT register access */
1487 	spinlock_t didt_idx_lock;
1488 	amdgpu_rreg_t			didt_rreg;
1489 	amdgpu_wreg_t			didt_wreg;
1490 	/* protects concurrent gc_cac register access */
1491 	spinlock_t gc_cac_idx_lock;
1492 	amdgpu_rreg_t			gc_cac_rreg;
1493 	amdgpu_wreg_t			gc_cac_wreg;
1494 	/* protects concurrent ENDPOINT (audio) register access */
1495 	spinlock_t audio_endpt_idx_lock;
1496 	amdgpu_block_rreg_t		audio_endpt_rreg;
1497 	amdgpu_block_wreg_t		audio_endpt_wreg;
1498 	void __iomem                    *rio_mem;
1499 	resource_size_t			rio_mem_size;
1500 	struct amdgpu_doorbell		doorbell;
1501 
1502 	/* clock/pll info */
1503 	struct amdgpu_clock            clock;
1504 
1505 	/* MC */
1506 	struct amdgpu_mc		mc;
1507 	struct amdgpu_gart		gart;
1508 	struct amdgpu_dummy_page	dummy_page;
1509 	struct amdgpu_vm_manager	vm_manager;
1510 	struct amdgpu_vmhub             vmhub[AMDGPU_MAX_VMHUBS];
1511 
1512 	/* memory management */
1513 	struct amdgpu_mman		mman;
1514 	struct amdgpu_vram_scratch	vram_scratch;
1515 	struct amdgpu_wb		wb;
1516 	atomic64_t			vram_usage;
1517 	atomic64_t			vram_vis_usage;
1518 	atomic64_t			gtt_usage;
1519 	atomic64_t			num_bytes_moved;
1520 	atomic64_t			num_evictions;
1521 	atomic_t			gpu_reset_counter;
1522 
1523 	/* data for buffer migration throttling */
1524 	struct {
1525 		spinlock_t		lock;
1526 		s64			last_update_us;
1527 		s64			accum_us; /* accumulated microseconds */
1528 		u32			log2_max_MBps;
1529 	} mm_stats;
1530 
1531 	/* display */
1532 	bool				enable_virtual_display;
1533 	struct amdgpu_mode_info		mode_info;
1534 	struct work_struct		hotplug_work;
1535 	struct amdgpu_irq_src		crtc_irq;
1536 	struct amdgpu_irq_src		pageflip_irq;
1537 	struct amdgpu_irq_src		hpd_irq;
1538 
1539 	/* rings */
1540 	u64				fence_context;
1541 	unsigned			num_rings;
1542 	struct amdgpu_ring		*rings[AMDGPU_MAX_RINGS];
1543 	bool				ib_pool_ready;
1544 	struct amdgpu_sa_manager	ring_tmp_bo;
1545 
1546 	/* interrupts */
1547 	struct amdgpu_irq		irq;
1548 
1549 	/* powerplay */
1550 	struct amd_powerplay		powerplay;
1551 	bool				pp_enabled;
1552 	bool				pp_force_state_enabled;
1553 
1554 	/* dpm */
1555 	struct amdgpu_pm		pm;
1556 	u32				cg_flags;
1557 	u32				pg_flags;
1558 
1559 	/* amdgpu smumgr */
1560 	struct amdgpu_smumgr smu;
1561 
1562 	/* gfx */
1563 	struct amdgpu_gfx		gfx;
1564 
1565 	/* sdma */
1566 	struct amdgpu_sdma		sdma;
1567 
1568 	/* uvd */
1569 	struct amdgpu_uvd		uvd;
1570 
1571 	/* vce */
1572 	struct amdgpu_vce		vce;
1573 
1574 	/* firmwares */
1575 	struct amdgpu_firmware		firmware;
1576 
1577 	/* PSP */
1578 	struct psp_context		psp;
1579 
1580 	/* GDS */
1581 	struct amdgpu_gds		gds;
1582 
1583 	struct amdgpu_ip_block          ip_blocks[AMDGPU_MAX_IP_NUM];
1584 	int				num_ip_blocks;
1585 	struct mutex	mn_lock;
1586 	DECLARE_HASHTABLE(mn_hash, 7);
1587 
1588 	/* tracking pinned memory */
1589 	u64 vram_pin_size;
1590 	u64 invisible_pin_size;
1591 	u64 gart_pin_size;
1592 
1593 	/* amdkfd interface */
1594 	struct kfd_dev          *kfd;
1595 
1596 	struct amdgpu_virt	virt;
1597 
1598 	/* link all shadow bo */
1599 	struct list_head                shadow_list;
1600 	struct mutex                    shadow_list_lock;
1601 	/* link all gtt */
1602 	spinlock_t			gtt_list_lock;
1603 	struct list_head                gtt_list;
1604 
1605 	/* record hw reset is performed */
1606 	bool has_hw_reset;
1607 
1608 };
1609 
1610 static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev)
1611 {
1612 	return container_of(bdev, struct amdgpu_device, mman.bdev);
1613 }
1614 
1615 bool amdgpu_device_is_px(struct drm_device *dev);
1616 int amdgpu_device_init(struct amdgpu_device *adev,
1617 		       struct drm_device *ddev,
1618 		       struct pci_dev *pdev,
1619 		       uint32_t flags);
1620 void amdgpu_device_fini(struct amdgpu_device *adev);
1621 int amdgpu_gpu_wait_for_idle(struct amdgpu_device *adev);
1622 
1623 uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg,
1624 			uint32_t acc_flags);
1625 void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v,
1626 		    uint32_t acc_flags);
1627 u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg);
1628 void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v);
1629 
1630 u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index);
1631 void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v);
1632 u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index);
1633 void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v);
1634 
1635 /*
1636  * Registers read & write functions.
1637  */
1638 
1639 #define AMDGPU_REGS_IDX       (1<<0)
1640 #define AMDGPU_REGS_NO_KIQ    (1<<1)
1641 
1642 #define RREG32_NO_KIQ(reg) amdgpu_mm_rreg(adev, (reg), AMDGPU_REGS_NO_KIQ)
1643 #define WREG32_NO_KIQ(reg, v) amdgpu_mm_wreg(adev, (reg), (v), AMDGPU_REGS_NO_KIQ)
1644 
1645 #define RREG32(reg) amdgpu_mm_rreg(adev, (reg), 0)
1646 #define RREG32_IDX(reg) amdgpu_mm_rreg(adev, (reg), AMDGPU_REGS_IDX)
1647 #define DREG32(reg) printk(KERN_INFO "REGISTER: " #reg " : 0x%08X\n", amdgpu_mm_rreg(adev, (reg), 0))
1648 #define WREG32(reg, v) amdgpu_mm_wreg(adev, (reg), (v), 0)
1649 #define WREG32_IDX(reg, v) amdgpu_mm_wreg(adev, (reg), (v), AMDGPU_REGS_IDX)
1650 #define REG_SET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
1651 #define REG_GET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
1652 #define RREG32_PCIE(reg) adev->pcie_rreg(adev, (reg))
1653 #define WREG32_PCIE(reg, v) adev->pcie_wreg(adev, (reg), (v))
1654 #define RREG32_PCIE_PORT(reg) adev->pciep_rreg(adev, (reg))
1655 #define WREG32_PCIE_PORT(reg, v) adev->pciep_wreg(adev, (reg), (v))
1656 #define RREG32_SMC(reg) adev->smc_rreg(adev, (reg))
1657 #define WREG32_SMC(reg, v) adev->smc_wreg(adev, (reg), (v))
1658 #define RREG32_UVD_CTX(reg) adev->uvd_ctx_rreg(adev, (reg))
1659 #define WREG32_UVD_CTX(reg, v) adev->uvd_ctx_wreg(adev, (reg), (v))
1660 #define RREG32_DIDT(reg) adev->didt_rreg(adev, (reg))
1661 #define WREG32_DIDT(reg, v) adev->didt_wreg(adev, (reg), (v))
1662 #define RREG32_GC_CAC(reg) adev->gc_cac_rreg(adev, (reg))
1663 #define WREG32_GC_CAC(reg, v) adev->gc_cac_wreg(adev, (reg), (v))
1664 #define RREG32_AUDIO_ENDPT(block, reg) adev->audio_endpt_rreg(adev, (block), (reg))
1665 #define WREG32_AUDIO_ENDPT(block, reg, v) adev->audio_endpt_wreg(adev, (block), (reg), (v))
1666 #define WREG32_P(reg, val, mask)				\
1667 	do {							\
1668 		uint32_t tmp_ = RREG32(reg);			\
1669 		tmp_ &= (mask);					\
1670 		tmp_ |= ((val) & ~(mask));			\
1671 		WREG32(reg, tmp_);				\
1672 	} while (0)
1673 #define WREG32_AND(reg, and) WREG32_P(reg, 0, and)
1674 #define WREG32_OR(reg, or) WREG32_P(reg, or, ~(or))
1675 #define WREG32_PLL_P(reg, val, mask)				\
1676 	do {							\
1677 		uint32_t tmp_ = RREG32_PLL(reg);		\
1678 		tmp_ &= (mask);					\
1679 		tmp_ |= ((val) & ~(mask));			\
1680 		WREG32_PLL(reg, tmp_);				\
1681 	} while (0)
1682 #define DREG32_SYS(sqf, adev, reg) seq_printf((sqf), #reg " : 0x%08X\n", amdgpu_mm_rreg((adev), (reg), false))
1683 #define RREG32_IO(reg) amdgpu_io_rreg(adev, (reg))
1684 #define WREG32_IO(reg, v) amdgpu_io_wreg(adev, (reg), (v))
1685 
1686 #define RDOORBELL32(index) amdgpu_mm_rdoorbell(adev, (index))
1687 #define WDOORBELL32(index, v) amdgpu_mm_wdoorbell(adev, (index), (v))
1688 #define RDOORBELL64(index) amdgpu_mm_rdoorbell64(adev, (index))
1689 #define WDOORBELL64(index, v) amdgpu_mm_wdoorbell64(adev, (index), (v))
1690 
1691 #define REG_FIELD_SHIFT(reg, field) reg##__##field##__SHIFT
1692 #define REG_FIELD_MASK(reg, field) reg##__##field##_MASK
1693 
1694 #define REG_SET_FIELD(orig_val, reg, field, field_val)			\
1695 	(((orig_val) & ~REG_FIELD_MASK(reg, field)) |			\
1696 	 (REG_FIELD_MASK(reg, field) & ((field_val) << REG_FIELD_SHIFT(reg, field))))
1697 
1698 #define REG_GET_FIELD(value, reg, field)				\
1699 	(((value) & REG_FIELD_MASK(reg, field)) >> REG_FIELD_SHIFT(reg, field))
1700 
1701 #define WREG32_FIELD(reg, field, val)	\
1702 	WREG32(mm##reg, (RREG32(mm##reg) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
1703 
1704 #define WREG32_FIELD_OFFSET(reg, offset, field, val)	\
1705 	WREG32(mm##reg + offset, (RREG32(mm##reg + offset) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
1706 
1707 #define WREG32_FIELD15(ip, idx, reg, field, val)	\
1708 	WREG32(SOC15_REG_OFFSET(ip, idx, mm##reg), (RREG32(SOC15_REG_OFFSET(ip, idx, mm##reg)) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
1709 
1710 /*
1711  * BIOS helpers.
1712  */
1713 #define RBIOS8(i) (adev->bios[i])
1714 #define RBIOS16(i) (RBIOS8(i) | (RBIOS8((i)+1) << 8))
1715 #define RBIOS32(i) ((RBIOS16(i)) | (RBIOS16((i)+2) << 16))
1716 
1717 /*
1718  * RING helpers.
1719  */
1720 static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v)
1721 {
1722 	if (ring->count_dw <= 0)
1723 		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
1724 	ring->ring[ring->wptr++ & ring->buf_mask] = v;
1725 	ring->wptr &= ring->ptr_mask;
1726 	ring->count_dw--;
1727 }
1728 
1729 static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, void *src, int count_dw)
1730 {
1731 	unsigned occupied, chunk1, chunk2;
1732 	void *dst;
1733 
1734 	if (ring->count_dw < count_dw) {
1735 		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
1736 	} else {
1737 		occupied = ring->wptr & ring->buf_mask;
1738 		dst = (void *)&ring->ring[occupied];
1739 		chunk1 = ring->buf_mask + 1 - occupied;
1740 		chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1;
1741 		chunk2 = count_dw - chunk1;
1742 		chunk1 <<= 2;
1743 		chunk2 <<= 2;
1744 
1745 		if (chunk1)
1746 			memcpy(dst, src, chunk1);
1747 
1748 		if (chunk2) {
1749 			src += chunk1;
1750 			dst = (void *)ring->ring;
1751 			memcpy(dst, src, chunk2);
1752 		}
1753 
1754 		ring->wptr += count_dw;
1755 		ring->wptr &= ring->ptr_mask;
1756 		ring->count_dw -= count_dw;
1757 	}
1758 }
1759 
1760 static inline struct amdgpu_sdma_instance *
1761 amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
1762 {
1763 	struct amdgpu_device *adev = ring->adev;
1764 	int i;
1765 
1766 	for (i = 0; i < adev->sdma.num_instances; i++)
1767 		if (&adev->sdma.instance[i].ring == ring)
1768 			break;
1769 
1770 	if (i < AMDGPU_MAX_SDMA_INSTANCES)
1771 		return &adev->sdma.instance[i];
1772 	else
1773 		return NULL;
1774 }
1775 
1776 /*
1777  * ASICs macro.
1778  */
1779 #define amdgpu_asic_set_vga_state(adev, state) (adev)->asic_funcs->set_vga_state((adev), (state))
1780 #define amdgpu_asic_reset(adev) (adev)->asic_funcs->reset((adev))
1781 #define amdgpu_asic_get_xclk(adev) (adev)->asic_funcs->get_xclk((adev))
1782 #define amdgpu_asic_set_uvd_clocks(adev, v, d) (adev)->asic_funcs->set_uvd_clocks((adev), (v), (d))
1783 #define amdgpu_asic_set_vce_clocks(adev, ev, ec) (adev)->asic_funcs->set_vce_clocks((adev), (ev), (ec))
1784 #define amdgpu_get_pcie_lanes(adev) (adev)->asic_funcs->get_pcie_lanes((adev))
1785 #define amdgpu_set_pcie_lanes(adev, l) (adev)->asic_funcs->set_pcie_lanes((adev), (l))
1786 #define amdgpu_asic_get_gpu_clock_counter(adev) (adev)->asic_funcs->get_gpu_clock_counter((adev))
1787 #define amdgpu_asic_read_disabled_bios(adev) (adev)->asic_funcs->read_disabled_bios((adev))
1788 #define amdgpu_asic_read_bios_from_rom(adev, b, l) (adev)->asic_funcs->read_bios_from_rom((adev), (b), (l))
1789 #define amdgpu_asic_read_register(adev, se, sh, offset, v)((adev)->asic_funcs->read_register((adev), (se), (sh), (offset), (v)))
1790 #define amdgpu_asic_get_config_memsize(adev) (adev)->asic_funcs->get_config_memsize((adev))
1791 #define amdgpu_gart_flush_gpu_tlb(adev, vmid) (adev)->gart.gart_funcs->flush_gpu_tlb((adev), (vmid))
1792 #define amdgpu_gart_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gart.gart_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags))
1793 #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count)))
1794 #define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr)))
1795 #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))
1796 #define amdgpu_vm_get_pte_flags(adev, flags) (adev)->gart.gart_funcs->get_vm_pte_flags((adev),(flags))
1797 #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib)))
1798 #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r))
1799 #define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t))
1800 #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r))
1801 #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r))
1802 #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r))
1803 #define amdgpu_ring_emit_ib(r, ib, vm_id, c) (r)->funcs->emit_ib((r), (ib), (vm_id), (c))
1804 #define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r))
1805 #define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr))
1806 #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags))
1807 #define amdgpu_ring_emit_gds_switch(r, v, db, ds, wb, ws, ab, as) (r)->funcs->emit_gds_switch((r), (v), (db), (ds), (wb), (ws), (ab), (as))
1808 #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r))
1809 #define amdgpu_ring_emit_hdp_invalidate(r) (r)->funcs->emit_hdp_invalidate((r))
1810 #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r))
1811 #define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d))
1812 #define amdgpu_ring_emit_rreg(r, d) (r)->funcs->emit_rreg((r), (d))
1813 #define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v))
1814 #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
1815 #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
1816 #define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
1817 #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev))
1818 #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv))
1819 #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev))
1820 #define amdgpu_display_set_vga_render_state(adev, r) (adev)->mode_info.funcs->set_vga_render_state((adev), (r))
1821 #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc))
1822 #define amdgpu_display_vblank_wait(adev, crtc) (adev)->mode_info.funcs->vblank_wait((adev), (crtc))
1823 #define amdgpu_display_backlight_set_level(adev, e, l) (adev)->mode_info.funcs->backlight_set_level((e), (l))
1824 #define amdgpu_display_backlight_get_level(adev, e) (adev)->mode_info.funcs->backlight_get_level((e))
1825 #define amdgpu_display_hpd_sense(adev, h) (adev)->mode_info.funcs->hpd_sense((adev), (h))
1826 #define amdgpu_display_hpd_set_polarity(adev, h) (adev)->mode_info.funcs->hpd_set_polarity((adev), (h))
1827 #define amdgpu_display_hpd_get_gpio_reg(adev) (adev)->mode_info.funcs->hpd_get_gpio_reg((adev))
1828 #define amdgpu_display_bandwidth_update(adev) (adev)->mode_info.funcs->bandwidth_update((adev))
1829 #define amdgpu_display_page_flip(adev, crtc, base, async) (adev)->mode_info.funcs->page_flip((adev), (crtc), (base), (async))
1830 #define amdgpu_display_page_flip_get_scanoutpos(adev, crtc, vbl, pos) (adev)->mode_info.funcs->page_flip_get_scanoutpos((adev), (crtc), (vbl), (pos))
1831 #define amdgpu_display_add_encoder(adev, e, s, c) (adev)->mode_info.funcs->add_encoder((adev), (e), (s), (c))
1832 #define amdgpu_display_add_connector(adev, ci, sd, ct, ib, coi, h, r) (adev)->mode_info.funcs->add_connector((adev), (ci), (sd), (ct), (ib), (coi), (h), (r))
1833 #define amdgpu_display_stop_mc_access(adev, s) (adev)->mode_info.funcs->stop_mc_access((adev), (s))
1834 #define amdgpu_display_resume_mc_access(adev, s) (adev)->mode_info.funcs->resume_mc_access((adev), (s))
1835 #define amdgpu_emit_copy_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_copy_buffer((ib),  (s), (d), (b))
1836 #define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b))
1837 #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev))
1838 #define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance))
1839 #define amdgpu_gds_switch(adev, r, v, d, w, a) (adev)->gds.funcs->patch_gds_switch((r), (v), (d), (w), (a))
1840 #define amdgpu_psp_check_fw_loading_status(adev, i) (adev)->firmware.funcs->check_fw_loading_status((adev), (i))
1841 
1842 /* Common functions */
1843 int amdgpu_gpu_reset(struct amdgpu_device *adev);
1844 bool amdgpu_need_backup(struct amdgpu_device *adev);
1845 void amdgpu_pci_config_reset(struct amdgpu_device *adev);
1846 bool amdgpu_need_post(struct amdgpu_device *adev);
1847 void amdgpu_update_display_priority(struct amdgpu_device *adev);
1848 
1849 int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data);
1850 int amdgpu_cs_get_ring(struct amdgpu_device *adev, u32 ip_type,
1851 		       u32 ip_instance, u32 ring,
1852 		       struct amdgpu_ring **out_ring);
1853 void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes);
1854 void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain);
1855 bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
1856 int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
1857 int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
1858 				     uint32_t flags);
1859 bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
1860 struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
1861 bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
1862 				  unsigned long end);
1863 bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
1864 				       int *last_invalidated);
1865 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
1866 uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
1867 				 struct ttm_mem_reg *mem);
1868 void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base);
1869 void amdgpu_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc);
1870 void amdgpu_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size);
1871 int amdgpu_ttm_init(struct amdgpu_device *adev);
1872 void amdgpu_ttm_fini(struct amdgpu_device *adev);
1873 void amdgpu_program_register_sequence(struct amdgpu_device *adev,
1874 					     const u32 *registers,
1875 					     const u32 array_size);
1876 
1877 bool amdgpu_device_is_px(struct drm_device *dev);
1878 /* atpx handler */
1879 #if defined(CONFIG_VGA_SWITCHEROO)
1880 void amdgpu_register_atpx_handler(void);
1881 void amdgpu_unregister_atpx_handler(void);
1882 bool amdgpu_has_atpx_dgpu_power_cntl(void);
1883 bool amdgpu_is_atpx_hybrid(void);
1884 bool amdgpu_atpx_dgpu_req_power_for_displays(void);
1885 bool amdgpu_has_atpx(void);
1886 #else
1887 static inline void amdgpu_register_atpx_handler(void) {}
1888 static inline void amdgpu_unregister_atpx_handler(void) {}
1889 static inline bool amdgpu_has_atpx_dgpu_power_cntl(void) { return false; }
1890 static inline bool amdgpu_is_atpx_hybrid(void) { return false; }
1891 static inline bool amdgpu_atpx_dgpu_req_power_for_displays(void) { return false; }
1892 static inline bool amdgpu_has_atpx(void) { return false; }
1893 #endif
1894 
1895 /*
1896  * KMS
1897  */
1898 extern const struct drm_ioctl_desc amdgpu_ioctls_kms[];
1899 extern const int amdgpu_max_kms_ioctl;
1900 
1901 int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags);
1902 void amdgpu_driver_unload_kms(struct drm_device *dev);
1903 void amdgpu_driver_lastclose_kms(struct drm_device *dev);
1904 int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv);
1905 void amdgpu_driver_postclose_kms(struct drm_device *dev,
1906 				 struct drm_file *file_priv);
1907 int amdgpu_suspend(struct amdgpu_device *adev);
1908 int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon);
1909 int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon);
1910 u32 amdgpu_get_vblank_counter_kms(struct drm_device *dev, unsigned int pipe);
1911 int amdgpu_enable_vblank_kms(struct drm_device *dev, unsigned int pipe);
1912 void amdgpu_disable_vblank_kms(struct drm_device *dev, unsigned int pipe);
1913 int amdgpu_get_vblank_timestamp_kms(struct drm_device *dev, unsigned int pipe,
1914 				    int *max_error,
1915 				    struct timeval *vblank_time,
1916 				    unsigned flags);
1917 long amdgpu_kms_compat_ioctl(struct file *filp, unsigned int cmd,
1918 			     unsigned long arg);
1919 
1920 /*
1921  * functions used by amdgpu_encoder.c
1922  */
1923 struct amdgpu_afmt_acr {
1924 	u32 clock;
1925 
1926 	int n_32khz;
1927 	int cts_32khz;
1928 
1929 	int n_44_1khz;
1930 	int cts_44_1khz;
1931 
1932 	int n_48khz;
1933 	int cts_48khz;
1934 
1935 };
1936 
1937 struct amdgpu_afmt_acr amdgpu_afmt_acr(uint32_t clock);
1938 
1939 /* amdgpu_acpi.c */
1940 #if defined(CONFIG_ACPI)
1941 int amdgpu_acpi_init(struct amdgpu_device *adev);
1942 void amdgpu_acpi_fini(struct amdgpu_device *adev);
1943 bool amdgpu_acpi_is_pcie_performance_request_supported(struct amdgpu_device *adev);
1944 int amdgpu_acpi_pcie_performance_request(struct amdgpu_device *adev,
1945 						u8 perf_req, bool advertise);
1946 int amdgpu_acpi_pcie_notify_device_ready(struct amdgpu_device *adev);
1947 #else
1948 static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; }
1949 static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { }
1950 #endif
1951 
1952 struct amdgpu_bo_va_mapping *
1953 amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
1954 		       uint64_t addr, struct amdgpu_bo **bo);
1955 int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser);
1956 
1957 #include "amdgpu_object.h"
1958 #endif
1959