xref: /openbmc/linux/drivers/gpu/drm/amd/amdgpu/amdgpu.h (revision a36954f5)
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_DEFAULT_GTT_SIZE_MB		3072ULL /* 3GB by default */
114 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
115 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */
116 #define AMDGPU_FENCE_JIFFIES_TIMEOUT		(HZ / 2)
117 /* AMDGPU_IB_POOL_SIZE must be a power of 2 */
118 #define AMDGPU_IB_POOL_SIZE			16
119 #define AMDGPU_DEBUGFS_MAX_COMPONENTS		32
120 #define AMDGPUFB_CONN_LIMIT			4
121 #define AMDGPU_BIOS_NUM_SCRATCH			16
122 
123 /* max number of IP instances */
124 #define AMDGPU_MAX_SDMA_INSTANCES		2
125 
126 /* hard reset data */
127 #define AMDGPU_ASIC_RESET_DATA                  0x39d5e86b
128 
129 /* reset flags */
130 #define AMDGPU_RESET_GFX			(1 << 0)
131 #define AMDGPU_RESET_COMPUTE			(1 << 1)
132 #define AMDGPU_RESET_DMA			(1 << 2)
133 #define AMDGPU_RESET_CP				(1 << 3)
134 #define AMDGPU_RESET_GRBM			(1 << 4)
135 #define AMDGPU_RESET_DMA1			(1 << 5)
136 #define AMDGPU_RESET_RLC			(1 << 6)
137 #define AMDGPU_RESET_SEM			(1 << 7)
138 #define AMDGPU_RESET_IH				(1 << 8)
139 #define AMDGPU_RESET_VMC			(1 << 9)
140 #define AMDGPU_RESET_MC				(1 << 10)
141 #define AMDGPU_RESET_DISPLAY			(1 << 11)
142 #define AMDGPU_RESET_UVD			(1 << 12)
143 #define AMDGPU_RESET_VCE			(1 << 13)
144 #define AMDGPU_RESET_VCE1			(1 << 14)
145 
146 /* GFX current status */
147 #define AMDGPU_GFX_NORMAL_MODE			0x00000000L
148 #define AMDGPU_GFX_SAFE_MODE			0x00000001L
149 #define AMDGPU_GFX_PG_DISABLED_MODE		0x00000002L
150 #define AMDGPU_GFX_CG_DISABLED_MODE		0x00000004L
151 #define AMDGPU_GFX_LBPW_DISABLED_MODE		0x00000008L
152 
153 /* max cursor sizes (in pixels) */
154 #define CIK_CURSOR_WIDTH 128
155 #define CIK_CURSOR_HEIGHT 128
156 
157 struct amdgpu_device;
158 struct amdgpu_ib;
159 struct amdgpu_cs_parser;
160 struct amdgpu_job;
161 struct amdgpu_irq_src;
162 struct amdgpu_fpriv;
163 
164 enum amdgpu_cp_irq {
165 	AMDGPU_CP_IRQ_GFX_EOP = 0,
166 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP,
167 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP,
168 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP,
169 	AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP,
170 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE0_EOP,
171 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE1_EOP,
172 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE2_EOP,
173 	AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE3_EOP,
174 
175 	AMDGPU_CP_IRQ_LAST
176 };
177 
178 enum amdgpu_sdma_irq {
179 	AMDGPU_SDMA_IRQ_TRAP0 = 0,
180 	AMDGPU_SDMA_IRQ_TRAP1,
181 
182 	AMDGPU_SDMA_IRQ_LAST
183 };
184 
185 enum amdgpu_thermal_irq {
186 	AMDGPU_THERMAL_IRQ_LOW_TO_HIGH = 0,
187 	AMDGPU_THERMAL_IRQ_HIGH_TO_LOW,
188 
189 	AMDGPU_THERMAL_IRQ_LAST
190 };
191 
192 enum amdgpu_kiq_irq {
193 	AMDGPU_CP_KIQ_IRQ_DRIVER0 = 0,
194 	AMDGPU_CP_KIQ_IRQ_LAST
195 };
196 
197 int amdgpu_set_clockgating_state(struct amdgpu_device *adev,
198 				  enum amd_ip_block_type block_type,
199 				  enum amd_clockgating_state state);
200 int amdgpu_set_powergating_state(struct amdgpu_device *adev,
201 				  enum amd_ip_block_type block_type,
202 				  enum amd_powergating_state state);
203 void amdgpu_get_clockgating_state(struct amdgpu_device *adev, u32 *flags);
204 int amdgpu_wait_for_idle(struct amdgpu_device *adev,
205 			 enum amd_ip_block_type block_type);
206 bool amdgpu_is_idle(struct amdgpu_device *adev,
207 		    enum amd_ip_block_type block_type);
208 
209 #define AMDGPU_MAX_IP_NUM 16
210 
211 struct amdgpu_ip_block_status {
212 	bool valid;
213 	bool sw;
214 	bool hw;
215 	bool late_initialized;
216 	bool hang;
217 };
218 
219 struct amdgpu_ip_block_version {
220 	const enum amd_ip_block_type type;
221 	const u32 major;
222 	const u32 minor;
223 	const u32 rev;
224 	const struct amd_ip_funcs *funcs;
225 };
226 
227 struct amdgpu_ip_block {
228 	struct amdgpu_ip_block_status status;
229 	const struct amdgpu_ip_block_version *version;
230 };
231 
232 int amdgpu_ip_block_version_cmp(struct amdgpu_device *adev,
233 				enum amd_ip_block_type type,
234 				u32 major, u32 minor);
235 
236 struct amdgpu_ip_block * amdgpu_get_ip_block(struct amdgpu_device *adev,
237 					     enum amd_ip_block_type type);
238 
239 int amdgpu_ip_block_add(struct amdgpu_device *adev,
240 			const struct amdgpu_ip_block_version *ip_block_version);
241 
242 /* provided by hw blocks that can move/clear data.  e.g., gfx or sdma */
243 struct amdgpu_buffer_funcs {
244 	/* maximum bytes in a single operation */
245 	uint32_t	copy_max_bytes;
246 
247 	/* number of dw to reserve per operation */
248 	unsigned	copy_num_dw;
249 
250 	/* used for buffer migration */
251 	void (*emit_copy_buffer)(struct amdgpu_ib *ib,
252 				 /* src addr in bytes */
253 				 uint64_t src_offset,
254 				 /* dst addr in bytes */
255 				 uint64_t dst_offset,
256 				 /* number of byte to transfer */
257 				 uint32_t byte_count);
258 
259 	/* maximum bytes in a single operation */
260 	uint32_t	fill_max_bytes;
261 
262 	/* number of dw to reserve per operation */
263 	unsigned	fill_num_dw;
264 
265 	/* used for buffer clearing */
266 	void (*emit_fill_buffer)(struct amdgpu_ib *ib,
267 				 /* value to write to memory */
268 				 uint32_t src_data,
269 				 /* dst addr in bytes */
270 				 uint64_t dst_offset,
271 				 /* number of byte to fill */
272 				 uint32_t byte_count);
273 };
274 
275 /* provided by hw blocks that can write ptes, e.g., sdma */
276 struct amdgpu_vm_pte_funcs {
277 	/* copy pte entries from GART */
278 	void (*copy_pte)(struct amdgpu_ib *ib,
279 			 uint64_t pe, uint64_t src,
280 			 unsigned count);
281 	/* write pte one entry at a time with addr mapping */
282 	void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe,
283 			  uint64_t value, unsigned count,
284 			  uint32_t incr);
285 	/* for linear pte/pde updates without addr mapping */
286 	void (*set_pte_pde)(struct amdgpu_ib *ib,
287 			    uint64_t pe,
288 			    uint64_t addr, unsigned count,
289 			    uint32_t incr, uint64_t flags);
290 };
291 
292 /* provided by the gmc block */
293 struct amdgpu_gart_funcs {
294 	/* flush the vm tlb via mmio */
295 	void (*flush_gpu_tlb)(struct amdgpu_device *adev,
296 			      uint32_t vmid);
297 	/* write pte/pde updates using the cpu */
298 	int (*set_pte_pde)(struct amdgpu_device *adev,
299 			   void *cpu_pt_addr, /* cpu addr of page table */
300 			   uint32_t gpu_page_idx, /* pte/pde to update */
301 			   uint64_t addr, /* addr to write into pte/pde */
302 			   uint64_t flags); /* access flags */
303 	/* enable/disable PRT support */
304 	void (*set_prt)(struct amdgpu_device *adev, bool enable);
305 	/* set pte flags based per asic */
306 	uint64_t (*get_vm_pte_flags)(struct amdgpu_device *adev,
307 				     uint32_t flags);
308 	/* adjust mc addr in fb for APU case */
309 	u64 (*adjust_mc_addr)(struct amdgpu_device *adev, u64 addr);
310 	uint32_t (*get_invalidate_req)(unsigned int vm_id);
311 };
312 
313 /* provided by the ih block */
314 struct amdgpu_ih_funcs {
315 	/* ring read/write ptr handling, called from interrupt context */
316 	u32 (*get_wptr)(struct amdgpu_device *adev);
317 	void (*decode_iv)(struct amdgpu_device *adev,
318 			  struct amdgpu_iv_entry *entry);
319 	void (*set_rptr)(struct amdgpu_device *adev);
320 };
321 
322 /*
323  * BIOS.
324  */
325 bool amdgpu_get_bios(struct amdgpu_device *adev);
326 bool amdgpu_read_bios(struct amdgpu_device *adev);
327 
328 /*
329  * Dummy page
330  */
331 struct amdgpu_dummy_page {
332 	struct page	*page;
333 	dma_addr_t	addr;
334 };
335 int amdgpu_dummy_page_init(struct amdgpu_device *adev);
336 void amdgpu_dummy_page_fini(struct amdgpu_device *adev);
337 
338 
339 /*
340  * Clocks
341  */
342 
343 #define AMDGPU_MAX_PPLL 3
344 
345 struct amdgpu_clock {
346 	struct amdgpu_pll ppll[AMDGPU_MAX_PPLL];
347 	struct amdgpu_pll spll;
348 	struct amdgpu_pll mpll;
349 	/* 10 Khz units */
350 	uint32_t default_mclk;
351 	uint32_t default_sclk;
352 	uint32_t default_dispclk;
353 	uint32_t current_dispclk;
354 	uint32_t dp_extclk;
355 	uint32_t max_pixel_clock;
356 };
357 
358 /*
359  * BO.
360  */
361 struct amdgpu_bo_list_entry {
362 	struct amdgpu_bo		*robj;
363 	struct ttm_validate_buffer	tv;
364 	struct amdgpu_bo_va		*bo_va;
365 	uint32_t			priority;
366 	struct page			**user_pages;
367 	int				user_invalidated;
368 };
369 
370 struct amdgpu_bo_va_mapping {
371 	struct list_head		list;
372 	struct rb_node			rb;
373 	uint64_t			start;
374 	uint64_t			last;
375 	uint64_t			__subtree_last;
376 	uint64_t			offset;
377 	uint64_t			flags;
378 };
379 
380 /* bo virtual addresses in a specific vm */
381 struct amdgpu_bo_va {
382 	/* protected by bo being reserved */
383 	struct list_head		bo_list;
384 	struct dma_fence	        *last_pt_update;
385 	unsigned			ref_count;
386 
387 	/* protected by vm mutex and spinlock */
388 	struct list_head		vm_status;
389 
390 	/* mappings for this bo_va */
391 	struct list_head		invalids;
392 	struct list_head		valids;
393 
394 	/* constant after initialization */
395 	struct amdgpu_vm		*vm;
396 	struct amdgpu_bo		*bo;
397 };
398 
399 #define AMDGPU_GEM_DOMAIN_MAX		0x3
400 
401 struct amdgpu_bo {
402 	/* Protected by tbo.reserved */
403 	u32				prefered_domains;
404 	u32				allowed_domains;
405 	struct ttm_place		placements[AMDGPU_GEM_DOMAIN_MAX + 1];
406 	struct ttm_placement		placement;
407 	struct ttm_buffer_object	tbo;
408 	struct ttm_bo_kmap_obj		kmap;
409 	u64				flags;
410 	unsigned			pin_count;
411 	void				*kptr;
412 	u64				tiling_flags;
413 	u64				metadata_flags;
414 	void				*metadata;
415 	u32				metadata_size;
416 	unsigned			prime_shared_count;
417 	/* list of all virtual address to which this bo
418 	 * is associated to
419 	 */
420 	struct list_head		va;
421 	/* Constant after initialization */
422 	struct drm_gem_object		gem_base;
423 	struct amdgpu_bo		*parent;
424 	struct amdgpu_bo		*shadow;
425 
426 	struct ttm_bo_kmap_obj		dma_buf_vmap;
427 	struct amdgpu_mn		*mn;
428 	struct list_head		mn_list;
429 	struct list_head		shadow_list;
430 };
431 #define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, gem_base)
432 
433 void amdgpu_gem_object_free(struct drm_gem_object *obj);
434 int amdgpu_gem_object_open(struct drm_gem_object *obj,
435 				struct drm_file *file_priv);
436 void amdgpu_gem_object_close(struct drm_gem_object *obj,
437 				struct drm_file *file_priv);
438 unsigned long amdgpu_gem_timeout(uint64_t timeout_ns);
439 struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj);
440 struct drm_gem_object *
441 amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
442 				 struct dma_buf_attachment *attach,
443 				 struct sg_table *sg);
444 struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev,
445 					struct drm_gem_object *gobj,
446 					int flags);
447 int amdgpu_gem_prime_pin(struct drm_gem_object *obj);
448 void amdgpu_gem_prime_unpin(struct drm_gem_object *obj);
449 struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *);
450 void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj);
451 void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr);
452 int amdgpu_gem_debugfs_init(struct amdgpu_device *adev);
453 
454 /* sub-allocation manager, it has to be protected by another lock.
455  * By conception this is an helper for other part of the driver
456  * like the indirect buffer or semaphore, which both have their
457  * locking.
458  *
459  * Principe is simple, we keep a list of sub allocation in offset
460  * order (first entry has offset == 0, last entry has the highest
461  * offset).
462  *
463  * When allocating new object we first check if there is room at
464  * the end total_size - (last_object_offset + last_object_size) >=
465  * alloc_size. If so we allocate new object there.
466  *
467  * When there is not enough room at the end, we start waiting for
468  * each sub object until we reach object_offset+object_size >=
469  * alloc_size, this object then become the sub object we return.
470  *
471  * Alignment can't be bigger than page size.
472  *
473  * Hole are not considered for allocation to keep things simple.
474  * Assumption is that there won't be hole (all object on same
475  * alignment).
476  */
477 
478 #define AMDGPU_SA_NUM_FENCE_LISTS	32
479 
480 struct amdgpu_sa_manager {
481 	wait_queue_head_t	wq;
482 	struct amdgpu_bo	*bo;
483 	struct list_head	*hole;
484 	struct list_head	flist[AMDGPU_SA_NUM_FENCE_LISTS];
485 	struct list_head	olist;
486 	unsigned		size;
487 	uint64_t		gpu_addr;
488 	void			*cpu_ptr;
489 	uint32_t		domain;
490 	uint32_t		align;
491 };
492 
493 /* sub-allocation buffer */
494 struct amdgpu_sa_bo {
495 	struct list_head		olist;
496 	struct list_head		flist;
497 	struct amdgpu_sa_manager	*manager;
498 	unsigned			soffset;
499 	unsigned			eoffset;
500 	struct dma_fence	        *fence;
501 };
502 
503 /*
504  * GEM objects.
505  */
506 void amdgpu_gem_force_release(struct amdgpu_device *adev);
507 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
508 				int alignment, u32 initial_domain,
509 				u64 flags, bool kernel,
510 				struct drm_gem_object **obj);
511 
512 int amdgpu_mode_dumb_create(struct drm_file *file_priv,
513 			    struct drm_device *dev,
514 			    struct drm_mode_create_dumb *args);
515 int amdgpu_mode_dumb_mmap(struct drm_file *filp,
516 			  struct drm_device *dev,
517 			  uint32_t handle, uint64_t *offset_p);
518 int amdgpu_fence_slab_init(void);
519 void amdgpu_fence_slab_fini(void);
520 
521 /*
522  * GART structures, functions & helpers
523  */
524 struct amdgpu_mc;
525 
526 #define AMDGPU_GPU_PAGE_SIZE 4096
527 #define AMDGPU_GPU_PAGE_MASK (AMDGPU_GPU_PAGE_SIZE - 1)
528 #define AMDGPU_GPU_PAGE_SHIFT 12
529 #define AMDGPU_GPU_PAGE_ALIGN(a) (((a) + AMDGPU_GPU_PAGE_MASK) & ~AMDGPU_GPU_PAGE_MASK)
530 
531 struct amdgpu_gart {
532 	dma_addr_t			table_addr;
533 	struct amdgpu_bo		*robj;
534 	void				*ptr;
535 	unsigned			num_gpu_pages;
536 	unsigned			num_cpu_pages;
537 	unsigned			table_size;
538 #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS
539 	struct page			**pages;
540 #endif
541 	bool				ready;
542 
543 	/* Asic default pte flags */
544 	uint64_t			gart_pte_flags;
545 
546 	const struct amdgpu_gart_funcs *gart_funcs;
547 };
548 
549 int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev);
550 void amdgpu_gart_table_ram_free(struct amdgpu_device *adev);
551 int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev);
552 void amdgpu_gart_table_vram_free(struct amdgpu_device *adev);
553 int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev);
554 void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev);
555 int amdgpu_gart_init(struct amdgpu_device *adev);
556 void amdgpu_gart_fini(struct amdgpu_device *adev);
557 void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
558 			int pages);
559 int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
560 		     int pages, struct page **pagelist,
561 		     dma_addr_t *dma_addr, uint64_t flags);
562 int amdgpu_ttm_recover_gart(struct amdgpu_device *adev);
563 
564 /*
565  * VMHUB structures, functions & helpers
566  */
567 struct amdgpu_vmhub {
568 	uint32_t	ctx0_ptb_addr_lo32;
569 	uint32_t	ctx0_ptb_addr_hi32;
570 	uint32_t	vm_inv_eng0_req;
571 	uint32_t	vm_inv_eng0_ack;
572 	uint32_t	vm_context0_cntl;
573 	uint32_t	vm_l2_pro_fault_status;
574 	uint32_t	vm_l2_pro_fault_cntl;
575 };
576 
577 /*
578  * GPU MC structures, functions & helpers
579  */
580 struct amdgpu_mc {
581 	resource_size_t		aper_size;
582 	resource_size_t		aper_base;
583 	resource_size_t		agp_base;
584 	/* for some chips with <= 32MB we need to lie
585 	 * about vram size near mc fb location */
586 	u64			mc_vram_size;
587 	u64			visible_vram_size;
588 	u64			gtt_size;
589 	u64			gtt_start;
590 	u64			gtt_end;
591 	u64			vram_start;
592 	u64			vram_end;
593 	unsigned		vram_width;
594 	u64			real_vram_size;
595 	int			vram_mtrr;
596 	u64                     gtt_base_align;
597 	u64                     mc_mask;
598 	const struct firmware   *fw;	/* MC firmware */
599 	uint32_t                fw_version;
600 	struct amdgpu_irq_src	vm_fault;
601 	uint32_t		vram_type;
602 	uint32_t                srbm_soft_reset;
603 	struct amdgpu_mode_mc_save save;
604 	bool			prt_warning;
605 	/* apertures */
606 	u64					shared_aperture_start;
607 	u64					shared_aperture_end;
608 	u64					private_aperture_start;
609 	u64					private_aperture_end;
610 	/* protects concurrent invalidation */
611 	spinlock_t		invalidate_lock;
612 };
613 
614 /*
615  * GPU doorbell structures, functions & helpers
616  */
617 typedef enum _AMDGPU_DOORBELL_ASSIGNMENT
618 {
619 	AMDGPU_DOORBELL_KIQ                     = 0x000,
620 	AMDGPU_DOORBELL_HIQ                     = 0x001,
621 	AMDGPU_DOORBELL_DIQ                     = 0x002,
622 	AMDGPU_DOORBELL_MEC_RING0               = 0x010,
623 	AMDGPU_DOORBELL_MEC_RING1               = 0x011,
624 	AMDGPU_DOORBELL_MEC_RING2               = 0x012,
625 	AMDGPU_DOORBELL_MEC_RING3               = 0x013,
626 	AMDGPU_DOORBELL_MEC_RING4               = 0x014,
627 	AMDGPU_DOORBELL_MEC_RING5               = 0x015,
628 	AMDGPU_DOORBELL_MEC_RING6               = 0x016,
629 	AMDGPU_DOORBELL_MEC_RING7               = 0x017,
630 	AMDGPU_DOORBELL_GFX_RING0               = 0x020,
631 	AMDGPU_DOORBELL_sDMA_ENGINE0            = 0x1E0,
632 	AMDGPU_DOORBELL_sDMA_ENGINE1            = 0x1E1,
633 	AMDGPU_DOORBELL_IH                      = 0x1E8,
634 	AMDGPU_DOORBELL_MAX_ASSIGNMENT          = 0x3FF,
635 	AMDGPU_DOORBELL_INVALID                 = 0xFFFF
636 } AMDGPU_DOORBELL_ASSIGNMENT;
637 
638 struct amdgpu_doorbell {
639 	/* doorbell mmio */
640 	resource_size_t		base;
641 	resource_size_t		size;
642 	u32 __iomem		*ptr;
643 	u32			num_doorbells;	/* Number of doorbells actually reserved for amdgpu. */
644 };
645 
646 /*
647  * 64bit doorbell, offset are in QWORD, occupy 2KB doorbell space
648  */
649 typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
650 {
651 	/*
652 	 * All compute related doorbells: kiq, hiq, diq, traditional compute queue, user queue, should locate in
653 	 * a continues range so that programming CP_MEC_DOORBELL_RANGE_LOWER/UPPER can cover this range.
654 	 *  Compute related doorbells are allocated from 0x00 to 0x8a
655 	 */
656 
657 
658 	/* kernel scheduling */
659 	AMDGPU_DOORBELL64_KIQ                     = 0x00,
660 
661 	/* HSA interface queue and debug queue */
662 	AMDGPU_DOORBELL64_HIQ                     = 0x01,
663 	AMDGPU_DOORBELL64_DIQ                     = 0x02,
664 
665 	/* Compute engines */
666 	AMDGPU_DOORBELL64_MEC_RING0               = 0x03,
667 	AMDGPU_DOORBELL64_MEC_RING1               = 0x04,
668 	AMDGPU_DOORBELL64_MEC_RING2               = 0x05,
669 	AMDGPU_DOORBELL64_MEC_RING3               = 0x06,
670 	AMDGPU_DOORBELL64_MEC_RING4               = 0x07,
671 	AMDGPU_DOORBELL64_MEC_RING5               = 0x08,
672 	AMDGPU_DOORBELL64_MEC_RING6               = 0x09,
673 	AMDGPU_DOORBELL64_MEC_RING7               = 0x0a,
674 
675 	/* User queue doorbell range (128 doorbells) */
676 	AMDGPU_DOORBELL64_USERQUEUE_START         = 0x0b,
677 	AMDGPU_DOORBELL64_USERQUEUE_END           = 0x8a,
678 
679 	/* Graphics engine */
680 	AMDGPU_DOORBELL64_GFX_RING0               = 0x8b,
681 
682 	/*
683 	 * Other graphics doorbells can be allocated here: from 0x8c to 0xef
684 	 * Graphics voltage island aperture 1
685 	 * default non-graphics QWORD index is 0xF0 - 0xFF inclusive
686 	 */
687 
688 	/* sDMA engines */
689 	AMDGPU_DOORBELL64_sDMA_ENGINE0            = 0xF0,
690 	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0     = 0xF1,
691 	AMDGPU_DOORBELL64_sDMA_ENGINE1            = 0xF2,
692 	AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1     = 0xF3,
693 
694 	/* Interrupt handler */
695 	AMDGPU_DOORBELL64_IH                      = 0xF4,  /* For legacy interrupt ring buffer */
696 	AMDGPU_DOORBELL64_IH_RING1                = 0xF5,  /* For page migration request log */
697 	AMDGPU_DOORBELL64_IH_RING2                = 0xF6,  /* For page migration translation/invalidation log */
698 
699 	/* VCN engine use 32 bits doorbell  */
700 	AMDGPU_DOORBELL64_VCN0_1                  = 0xF8, /* lower 32 bits for VNC0 and upper 32 bits for VNC1 */
701 	AMDGPU_DOORBELL64_VCN2_3                  = 0xF9,
702 	AMDGPU_DOORBELL64_VCN4_5                  = 0xFA,
703 	AMDGPU_DOORBELL64_VCN6_7                  = 0xFB,
704 
705 	/* overlap the doorbell assignment with VCN as they are  mutually exclusive
706 	 * VCE engine's doorbell is 32 bit and two VCE ring share one QWORD
707 	 */
708 	AMDGPU_DOORBELL64_RING0_1                 = 0xF8,
709 	AMDGPU_DOORBELL64_RING2_3                 = 0xF9,
710 	AMDGPU_DOORBELL64_RING4_5                 = 0xFA,
711 	AMDGPU_DOORBELL64_RING6_7                 = 0xFB,
712 
713 	AMDGPU_DOORBELL64_UVD_RING0_1             = 0xFC,
714 	AMDGPU_DOORBELL64_UVD_RING2_3             = 0xFD,
715 	AMDGPU_DOORBELL64_UVD_RING4_5             = 0xFE,
716 	AMDGPU_DOORBELL64_UVD_RING6_7             = 0xFF,
717 
718 	AMDGPU_DOORBELL64_MAX_ASSIGNMENT          = 0xFF,
719 	AMDGPU_DOORBELL64_INVALID                 = 0xFFFF
720 } AMDGPU_DOORBELL64_ASSIGNMENT;
721 
722 
723 void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev,
724 				phys_addr_t *aperture_base,
725 				size_t *aperture_size,
726 				size_t *start_offset);
727 
728 /*
729  * IRQS.
730  */
731 
732 struct amdgpu_flip_work {
733 	struct delayed_work		flip_work;
734 	struct work_struct		unpin_work;
735 	struct amdgpu_device		*adev;
736 	int				crtc_id;
737 	u32				target_vblank;
738 	uint64_t			base;
739 	struct drm_pending_vblank_event *event;
740 	struct amdgpu_bo		*old_abo;
741 	struct dma_fence		*excl;
742 	unsigned			shared_count;
743 	struct dma_fence		**shared;
744 	struct dma_fence_cb		cb;
745 	bool				async;
746 };
747 
748 
749 /*
750  * CP & rings.
751  */
752 
753 struct amdgpu_ib {
754 	struct amdgpu_sa_bo		*sa_bo;
755 	uint32_t			length_dw;
756 	uint64_t			gpu_addr;
757 	uint32_t			*ptr;
758 	uint32_t			flags;
759 };
760 
761 extern const struct amd_sched_backend_ops amdgpu_sched_ops;
762 
763 int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs,
764 		     struct amdgpu_job **job, struct amdgpu_vm *vm);
765 int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size,
766 			     struct amdgpu_job **job);
767 
768 void amdgpu_job_free_resources(struct amdgpu_job *job);
769 void amdgpu_job_free(struct amdgpu_job *job);
770 int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
771 		      struct amd_sched_entity *entity, void *owner,
772 		      struct dma_fence **f);
773 
774 /*
775  * context related structures
776  */
777 
778 struct amdgpu_ctx_ring {
779 	uint64_t		sequence;
780 	struct dma_fence	**fences;
781 	struct amd_sched_entity	entity;
782 };
783 
784 struct amdgpu_ctx {
785 	struct kref		refcount;
786 	struct amdgpu_device    *adev;
787 	unsigned		reset_counter;
788 	spinlock_t		ring_lock;
789 	struct dma_fence	**fences;
790 	struct amdgpu_ctx_ring	rings[AMDGPU_MAX_RINGS];
791 	bool preamble_presented;
792 };
793 
794 struct amdgpu_ctx_mgr {
795 	struct amdgpu_device	*adev;
796 	struct mutex		lock;
797 	/* protected by lock */
798 	struct idr		ctx_handles;
799 };
800 
801 struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id);
802 int amdgpu_ctx_put(struct amdgpu_ctx *ctx);
803 
804 uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
805 			      struct dma_fence *fence);
806 struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
807 				   struct amdgpu_ring *ring, uint64_t seq);
808 
809 int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
810 		     struct drm_file *filp);
811 
812 void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr);
813 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
814 
815 /*
816  * file private structure
817  */
818 
819 struct amdgpu_fpriv {
820 	struct amdgpu_vm	vm;
821 	struct amdgpu_bo_va	*prt_va;
822 	struct mutex		bo_list_lock;
823 	struct idr		bo_list_handles;
824 	struct amdgpu_ctx_mgr	ctx_mgr;
825 };
826 
827 /*
828  * residency list
829  */
830 
831 struct amdgpu_bo_list {
832 	struct mutex lock;
833 	struct amdgpu_bo *gds_obj;
834 	struct amdgpu_bo *gws_obj;
835 	struct amdgpu_bo *oa_obj;
836 	unsigned first_userptr;
837 	unsigned num_entries;
838 	struct amdgpu_bo_list_entry *array;
839 };
840 
841 struct amdgpu_bo_list *
842 amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id);
843 void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list,
844 			     struct list_head *validated);
845 void amdgpu_bo_list_put(struct amdgpu_bo_list *list);
846 void amdgpu_bo_list_free(struct amdgpu_bo_list *list);
847 
848 /*
849  * GFX stuff
850  */
851 #include "clearstate_defs.h"
852 
853 struct amdgpu_rlc_funcs {
854 	void (*enter_safe_mode)(struct amdgpu_device *adev);
855 	void (*exit_safe_mode)(struct amdgpu_device *adev);
856 };
857 
858 struct amdgpu_rlc {
859 	/* for power gating */
860 	struct amdgpu_bo	*save_restore_obj;
861 	uint64_t		save_restore_gpu_addr;
862 	volatile uint32_t	*sr_ptr;
863 	const u32               *reg_list;
864 	u32                     reg_list_size;
865 	/* for clear state */
866 	struct amdgpu_bo	*clear_state_obj;
867 	uint64_t		clear_state_gpu_addr;
868 	volatile uint32_t	*cs_ptr;
869 	const struct cs_section_def   *cs_data;
870 	u32                     clear_state_size;
871 	/* for cp tables */
872 	struct amdgpu_bo	*cp_table_obj;
873 	uint64_t		cp_table_gpu_addr;
874 	volatile uint32_t	*cp_table_ptr;
875 	u32                     cp_table_size;
876 
877 	/* safe mode for updating CG/PG state */
878 	bool in_safe_mode;
879 	const struct amdgpu_rlc_funcs *funcs;
880 
881 	/* for firmware data */
882 	u32 save_and_restore_offset;
883 	u32 clear_state_descriptor_offset;
884 	u32 avail_scratch_ram_locations;
885 	u32 reg_restore_list_size;
886 	u32 reg_list_format_start;
887 	u32 reg_list_format_separate_start;
888 	u32 starting_offsets_start;
889 	u32 reg_list_format_size_bytes;
890 	u32 reg_list_size_bytes;
891 
892 	u32 *register_list_format;
893 	u32 *register_restore;
894 };
895 
896 struct amdgpu_mec {
897 	struct amdgpu_bo	*hpd_eop_obj;
898 	u64			hpd_eop_gpu_addr;
899 	struct amdgpu_bo	*mec_fw_obj;
900 	u64			mec_fw_gpu_addr;
901 	u32 num_pipe;
902 	u32 num_mec;
903 	u32 num_queue;
904 	void			*mqd_backup[AMDGPU_MAX_COMPUTE_RINGS + 1];
905 };
906 
907 struct amdgpu_kiq {
908 	u64			eop_gpu_addr;
909 	struct amdgpu_bo	*eop_obj;
910 	struct amdgpu_ring	ring;
911 	struct amdgpu_irq_src	irq;
912 };
913 
914 /*
915  * GPU scratch registers structures, functions & helpers
916  */
917 struct amdgpu_scratch {
918 	unsigned		num_reg;
919 	uint32_t                reg_base;
920 	uint32_t		free_mask;
921 };
922 
923 /*
924  * GFX configurations
925  */
926 #define AMDGPU_GFX_MAX_SE 4
927 #define AMDGPU_GFX_MAX_SH_PER_SE 2
928 
929 struct amdgpu_rb_config {
930 	uint32_t rb_backend_disable;
931 	uint32_t user_rb_backend_disable;
932 	uint32_t raster_config;
933 	uint32_t raster_config_1;
934 };
935 
936 struct gb_addr_config {
937 	uint16_t pipe_interleave_size;
938 	uint8_t num_pipes;
939 	uint8_t max_compress_frags;
940 	uint8_t num_banks;
941 	uint8_t num_se;
942 	uint8_t num_rb_per_se;
943 };
944 
945 struct amdgpu_gfx_config {
946 	unsigned max_shader_engines;
947 	unsigned max_tile_pipes;
948 	unsigned max_cu_per_sh;
949 	unsigned max_sh_per_se;
950 	unsigned max_backends_per_se;
951 	unsigned max_texture_channel_caches;
952 	unsigned max_gprs;
953 	unsigned max_gs_threads;
954 	unsigned max_hw_contexts;
955 	unsigned sc_prim_fifo_size_frontend;
956 	unsigned sc_prim_fifo_size_backend;
957 	unsigned sc_hiz_tile_fifo_size;
958 	unsigned sc_earlyz_tile_fifo_size;
959 
960 	unsigned num_tile_pipes;
961 	unsigned backend_enable_mask;
962 	unsigned mem_max_burst_length_bytes;
963 	unsigned mem_row_size_in_kb;
964 	unsigned shader_engine_tile_size;
965 	unsigned num_gpus;
966 	unsigned multi_gpu_tile_size;
967 	unsigned mc_arb_ramcfg;
968 	unsigned gb_addr_config;
969 	unsigned num_rbs;
970 	unsigned gs_vgt_table_depth;
971 	unsigned gs_prim_buffer_depth;
972 
973 	uint32_t tile_mode_array[32];
974 	uint32_t macrotile_mode_array[16];
975 
976 	struct gb_addr_config gb_addr_config_fields;
977 	struct amdgpu_rb_config rb_config[AMDGPU_GFX_MAX_SE][AMDGPU_GFX_MAX_SH_PER_SE];
978 
979 	/* gfx configure feature */
980 	uint32_t double_offchip_lds_buf;
981 };
982 
983 struct amdgpu_cu_info {
984 	uint32_t number; /* total active CU number */
985 	uint32_t ao_cu_mask;
986 	uint32_t wave_front_size;
987 	uint32_t bitmap[4][4];
988 };
989 
990 struct amdgpu_gfx_funcs {
991 	/* get the gpu clock counter */
992 	uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev);
993 	void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance);
994 	void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields);
995 	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);
996 	void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t start, uint32_t size, uint32_t *dst);
997 };
998 
999 struct amdgpu_ngg_buf {
1000 	struct amdgpu_bo	*bo;
1001 	uint64_t		gpu_addr;
1002 	uint32_t		size;
1003 	uint32_t		bo_size;
1004 };
1005 
1006 enum {
1007 	NGG_PRIM = 0,
1008 	NGG_POS,
1009 	NGG_CNTL,
1010 	NGG_PARAM,
1011 	NGG_BUF_MAX
1012 };
1013 
1014 struct amdgpu_ngg {
1015 	struct amdgpu_ngg_buf	buf[NGG_BUF_MAX];
1016 	uint32_t		gds_reserve_addr;
1017 	uint32_t		gds_reserve_size;
1018 	bool			init;
1019 };
1020 
1021 struct amdgpu_gfx {
1022 	struct mutex			gpu_clock_mutex;
1023 	struct amdgpu_gfx_config	config;
1024 	struct amdgpu_rlc		rlc;
1025 	struct amdgpu_mec		mec;
1026 	struct amdgpu_kiq		kiq;
1027 	struct amdgpu_scratch		scratch;
1028 	const struct firmware		*me_fw;	/* ME firmware */
1029 	uint32_t			me_fw_version;
1030 	const struct firmware		*pfp_fw; /* PFP firmware */
1031 	uint32_t			pfp_fw_version;
1032 	const struct firmware		*ce_fw;	/* CE firmware */
1033 	uint32_t			ce_fw_version;
1034 	const struct firmware		*rlc_fw; /* RLC firmware */
1035 	uint32_t			rlc_fw_version;
1036 	const struct firmware		*mec_fw; /* MEC firmware */
1037 	uint32_t			mec_fw_version;
1038 	const struct firmware		*mec2_fw; /* MEC2 firmware */
1039 	uint32_t			mec2_fw_version;
1040 	uint32_t			me_feature_version;
1041 	uint32_t			ce_feature_version;
1042 	uint32_t			pfp_feature_version;
1043 	uint32_t			rlc_feature_version;
1044 	uint32_t			mec_feature_version;
1045 	uint32_t			mec2_feature_version;
1046 	struct amdgpu_ring		gfx_ring[AMDGPU_MAX_GFX_RINGS];
1047 	unsigned			num_gfx_rings;
1048 	struct amdgpu_ring		compute_ring[AMDGPU_MAX_COMPUTE_RINGS];
1049 	unsigned			num_compute_rings;
1050 	struct amdgpu_irq_src		eop_irq;
1051 	struct amdgpu_irq_src		priv_reg_irq;
1052 	struct amdgpu_irq_src		priv_inst_irq;
1053 	/* gfx status */
1054 	uint32_t			gfx_current_status;
1055 	/* ce ram size*/
1056 	unsigned			ce_ram_size;
1057 	struct amdgpu_cu_info		cu_info;
1058 	const struct amdgpu_gfx_funcs	*funcs;
1059 
1060 	/* reset mask */
1061 	uint32_t                        grbm_soft_reset;
1062 	uint32_t                        srbm_soft_reset;
1063 	bool                            in_reset;
1064 	/* NGG */
1065 	struct amdgpu_ngg		ngg;
1066 };
1067 
1068 int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
1069 		  unsigned size, struct amdgpu_ib *ib);
1070 void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
1071 		    struct dma_fence *f);
1072 int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
1073 		       struct amdgpu_ib *ibs, struct amdgpu_job *job,
1074 		       struct dma_fence **f);
1075 int amdgpu_ib_pool_init(struct amdgpu_device *adev);
1076 void amdgpu_ib_pool_fini(struct amdgpu_device *adev);
1077 int amdgpu_ib_ring_tests(struct amdgpu_device *adev);
1078 
1079 /*
1080  * CS.
1081  */
1082 struct amdgpu_cs_chunk {
1083 	uint32_t		chunk_id;
1084 	uint32_t		length_dw;
1085 	void			*kdata;
1086 };
1087 
1088 struct amdgpu_cs_parser {
1089 	struct amdgpu_device	*adev;
1090 	struct drm_file		*filp;
1091 	struct amdgpu_ctx	*ctx;
1092 
1093 	/* chunks */
1094 	unsigned		nchunks;
1095 	struct amdgpu_cs_chunk	*chunks;
1096 
1097 	/* scheduler job object */
1098 	struct amdgpu_job	*job;
1099 
1100 	/* buffer objects */
1101 	struct ww_acquire_ctx		ticket;
1102 	struct amdgpu_bo_list		*bo_list;
1103 	struct amdgpu_bo_list_entry	vm_pd;
1104 	struct list_head		validated;
1105 	struct dma_fence		*fence;
1106 	uint64_t			bytes_moved_threshold;
1107 	uint64_t			bytes_moved;
1108 	struct amdgpu_bo_list_entry	*evictable;
1109 
1110 	/* user fence */
1111 	struct amdgpu_bo_list_entry	uf_entry;
1112 };
1113 
1114 #define AMDGPU_PREAMBLE_IB_PRESENT          (1 << 0) /* bit set means command submit involves a preamble IB */
1115 #define AMDGPU_PREAMBLE_IB_PRESENT_FIRST    (1 << 1) /* bit set means preamble IB is first presented in belonging context */
1116 #define AMDGPU_HAVE_CTX_SWITCH              (1 << 2) /* bit set means context switch occured */
1117 #define AMDGPU_VM_DOMAIN                    (1 << 3) /* bit set means in virtual memory context */
1118 
1119 struct amdgpu_job {
1120 	struct amd_sched_job    base;
1121 	struct amdgpu_device	*adev;
1122 	struct amdgpu_vm	*vm;
1123 	struct amdgpu_ring	*ring;
1124 	struct amdgpu_sync	sync;
1125 	struct amdgpu_ib	*ibs;
1126 	struct dma_fence	*fence; /* the hw fence */
1127 	uint32_t		preamble_status;
1128 	uint32_t		num_ibs;
1129 	void			*owner;
1130 	uint64_t		fence_ctx; /* the fence_context this job uses */
1131 	bool                    vm_needs_flush;
1132 	bool			need_pipeline_sync;
1133 	unsigned		vm_id;
1134 	uint64_t		vm_pd_addr;
1135 	uint32_t		gds_base, gds_size;
1136 	uint32_t		gws_base, gws_size;
1137 	uint32_t		oa_base, oa_size;
1138 
1139 	/* user fence handling */
1140 	uint64_t		uf_addr;
1141 	uint64_t		uf_sequence;
1142 
1143 };
1144 #define to_amdgpu_job(sched_job)		\
1145 		container_of((sched_job), struct amdgpu_job, base)
1146 
1147 static inline u32 amdgpu_get_ib_value(struct amdgpu_cs_parser *p,
1148 				      uint32_t ib_idx, int idx)
1149 {
1150 	return p->job->ibs[ib_idx].ptr[idx];
1151 }
1152 
1153 static inline void amdgpu_set_ib_value(struct amdgpu_cs_parser *p,
1154 				       uint32_t ib_idx, int idx,
1155 				       uint32_t value)
1156 {
1157 	p->job->ibs[ib_idx].ptr[idx] = value;
1158 }
1159 
1160 /*
1161  * Writeback
1162  */
1163 #define AMDGPU_MAX_WB 1024	/* Reserve at most 1024 WB slots for amdgpu-owned rings. */
1164 
1165 struct amdgpu_wb {
1166 	struct amdgpu_bo	*wb_obj;
1167 	volatile uint32_t	*wb;
1168 	uint64_t		gpu_addr;
1169 	u32			num_wb;	/* Number of wb slots actually reserved for amdgpu. */
1170 	unsigned long		used[DIV_ROUND_UP(AMDGPU_MAX_WB, BITS_PER_LONG)];
1171 };
1172 
1173 int amdgpu_wb_get(struct amdgpu_device *adev, u32 *wb);
1174 void amdgpu_wb_free(struct amdgpu_device *adev, u32 wb);
1175 int amdgpu_wb_get_64bit(struct amdgpu_device *adev, u32 *wb);
1176 void amdgpu_wb_free_64bit(struct amdgpu_device *adev, u32 wb);
1177 
1178 void amdgpu_get_pcie_info(struct amdgpu_device *adev);
1179 
1180 /*
1181  * SDMA
1182  */
1183 struct amdgpu_sdma_instance {
1184 	/* SDMA firmware */
1185 	const struct firmware	*fw;
1186 	uint32_t		fw_version;
1187 	uint32_t		feature_version;
1188 
1189 	struct amdgpu_ring	ring;
1190 	bool			burst_nop;
1191 };
1192 
1193 struct amdgpu_sdma {
1194 	struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES];
1195 #ifdef CONFIG_DRM_AMDGPU_SI
1196 	//SI DMA has a difference trap irq number for the second engine
1197 	struct amdgpu_irq_src	trap_irq_1;
1198 #endif
1199 	struct amdgpu_irq_src	trap_irq;
1200 	struct amdgpu_irq_src	illegal_inst_irq;
1201 	int			num_instances;
1202 	uint32_t                    srbm_soft_reset;
1203 };
1204 
1205 /*
1206  * Firmware
1207  */
1208 enum amdgpu_firmware_load_type {
1209 	AMDGPU_FW_LOAD_DIRECT = 0,
1210 	AMDGPU_FW_LOAD_SMU,
1211 	AMDGPU_FW_LOAD_PSP,
1212 };
1213 
1214 struct amdgpu_firmware {
1215 	struct amdgpu_firmware_info ucode[AMDGPU_UCODE_ID_MAXIMUM];
1216 	enum amdgpu_firmware_load_type load_type;
1217 	struct amdgpu_bo *fw_buf;
1218 	unsigned int fw_size;
1219 	unsigned int max_ucodes;
1220 	/* firmwares are loaded by psp instead of smu from vega10 */
1221 	const struct amdgpu_psp_funcs *funcs;
1222 	struct amdgpu_bo *rbuf;
1223 	struct mutex mutex;
1224 };
1225 
1226 /*
1227  * Benchmarking
1228  */
1229 void amdgpu_benchmark(struct amdgpu_device *adev, int test_number);
1230 
1231 
1232 /*
1233  * Testing
1234  */
1235 void amdgpu_test_moves(struct amdgpu_device *adev);
1236 
1237 /*
1238  * MMU Notifier
1239  */
1240 #if defined(CONFIG_MMU_NOTIFIER)
1241 int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
1242 void amdgpu_mn_unregister(struct amdgpu_bo *bo);
1243 #else
1244 static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
1245 {
1246 	return -ENODEV;
1247 }
1248 static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {}
1249 #endif
1250 
1251 /*
1252  * Debugfs
1253  */
1254 struct amdgpu_debugfs {
1255 	const struct drm_info_list	*files;
1256 	unsigned		num_files;
1257 };
1258 
1259 int amdgpu_debugfs_add_files(struct amdgpu_device *adev,
1260 			     const struct drm_info_list *files,
1261 			     unsigned nfiles);
1262 int amdgpu_debugfs_fence_init(struct amdgpu_device *adev);
1263 
1264 #if defined(CONFIG_DEBUG_FS)
1265 int amdgpu_debugfs_init(struct drm_minor *minor);
1266 #endif
1267 
1268 int amdgpu_debugfs_firmware_init(struct amdgpu_device *adev);
1269 
1270 /*
1271  * amdgpu smumgr functions
1272  */
1273 struct amdgpu_smumgr_funcs {
1274 	int (*check_fw_load_finish)(struct amdgpu_device *adev, uint32_t fwtype);
1275 	int (*request_smu_load_fw)(struct amdgpu_device *adev);
1276 	int (*request_smu_specific_fw)(struct amdgpu_device *adev, uint32_t fwtype);
1277 };
1278 
1279 /*
1280  * amdgpu smumgr
1281  */
1282 struct amdgpu_smumgr {
1283 	struct amdgpu_bo *toc_buf;
1284 	struct amdgpu_bo *smu_buf;
1285 	/* asic priv smu data */
1286 	void *priv;
1287 	spinlock_t smu_lock;
1288 	/* smumgr functions */
1289 	const struct amdgpu_smumgr_funcs *smumgr_funcs;
1290 	/* ucode loading complete flag */
1291 	uint32_t fw_flags;
1292 };
1293 
1294 /*
1295  * ASIC specific register table accessible by UMD
1296  */
1297 struct amdgpu_allowed_register_entry {
1298 	uint32_t reg_offset;
1299 	bool untouched;
1300 	bool grbm_indexed;
1301 };
1302 
1303 /*
1304  * ASIC specific functions.
1305  */
1306 struct amdgpu_asic_funcs {
1307 	bool (*read_disabled_bios)(struct amdgpu_device *adev);
1308 	bool (*read_bios_from_rom)(struct amdgpu_device *adev,
1309 				   u8 *bios, u32 length_bytes);
1310 	int (*read_register)(struct amdgpu_device *adev, u32 se_num,
1311 			     u32 sh_num, u32 reg_offset, u32 *value);
1312 	void (*set_vga_state)(struct amdgpu_device *adev, bool state);
1313 	int (*reset)(struct amdgpu_device *adev);
1314 	/* get the reference clock */
1315 	u32 (*get_xclk)(struct amdgpu_device *adev);
1316 	/* MM block clocks */
1317 	int (*set_uvd_clocks)(struct amdgpu_device *adev, u32 vclk, u32 dclk);
1318 	int (*set_vce_clocks)(struct amdgpu_device *adev, u32 evclk, u32 ecclk);
1319 	/* static power management */
1320 	int (*get_pcie_lanes)(struct amdgpu_device *adev);
1321 	void (*set_pcie_lanes)(struct amdgpu_device *adev, int lanes);
1322 	/* get config memsize register */
1323 	u32 (*get_config_memsize)(struct amdgpu_device *adev);
1324 };
1325 
1326 /*
1327  * IOCTL.
1328  */
1329 int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
1330 			    struct drm_file *filp);
1331 int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data,
1332 				struct drm_file *filp);
1333 
1334 int amdgpu_gem_info_ioctl(struct drm_device *dev, void *data,
1335 			  struct drm_file *filp);
1336 int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
1337 			struct drm_file *filp);
1338 int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data,
1339 			  struct drm_file *filp);
1340 int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
1341 			      struct drm_file *filp);
1342 int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
1343 			  struct drm_file *filp);
1344 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
1345 			struct drm_file *filp);
1346 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
1347 int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
1348 int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
1349 				struct drm_file *filp);
1350 
1351 int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
1352 				struct drm_file *filp);
1353 
1354 /* VRAM scratch page for HDP bug, default vram page */
1355 struct amdgpu_vram_scratch {
1356 	struct amdgpu_bo		*robj;
1357 	volatile uint32_t		*ptr;
1358 	u64				gpu_addr;
1359 };
1360 
1361 /*
1362  * ACPI
1363  */
1364 struct amdgpu_atif_notification_cfg {
1365 	bool enabled;
1366 	int command_code;
1367 };
1368 
1369 struct amdgpu_atif_notifications {
1370 	bool display_switch;
1371 	bool expansion_mode_change;
1372 	bool thermal_state;
1373 	bool forced_power_state;
1374 	bool system_power_state;
1375 	bool display_conf_change;
1376 	bool px_gfx_switch;
1377 	bool brightness_change;
1378 	bool dgpu_display_event;
1379 };
1380 
1381 struct amdgpu_atif_functions {
1382 	bool system_params;
1383 	bool sbios_requests;
1384 	bool select_active_disp;
1385 	bool lid_state;
1386 	bool get_tv_standard;
1387 	bool set_tv_standard;
1388 	bool get_panel_expansion_mode;
1389 	bool set_panel_expansion_mode;
1390 	bool temperature_change;
1391 	bool graphics_device_types;
1392 };
1393 
1394 struct amdgpu_atif {
1395 	struct amdgpu_atif_notifications notifications;
1396 	struct amdgpu_atif_functions functions;
1397 	struct amdgpu_atif_notification_cfg notification_cfg;
1398 	struct amdgpu_encoder *encoder_for_bl;
1399 };
1400 
1401 struct amdgpu_atcs_functions {
1402 	bool get_ext_state;
1403 	bool pcie_perf_req;
1404 	bool pcie_dev_rdy;
1405 	bool pcie_bus_width;
1406 };
1407 
1408 struct amdgpu_atcs {
1409 	struct amdgpu_atcs_functions functions;
1410 };
1411 
1412 /*
1413  * CGS
1414  */
1415 struct cgs_device *amdgpu_cgs_create_device(struct amdgpu_device *adev);
1416 void amdgpu_cgs_destroy_device(struct cgs_device *cgs_device);
1417 
1418 /*
1419  * Core structure, functions and helpers.
1420  */
1421 typedef uint32_t (*amdgpu_rreg_t)(struct amdgpu_device*, uint32_t);
1422 typedef void (*amdgpu_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
1423 
1424 typedef uint32_t (*amdgpu_block_rreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
1425 typedef void (*amdgpu_block_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t, uint32_t);
1426 
1427 struct amdgpu_device {
1428 	struct device			*dev;
1429 	struct drm_device		*ddev;
1430 	struct pci_dev			*pdev;
1431 
1432 #ifdef CONFIG_DRM_AMD_ACP
1433 	struct amdgpu_acp		acp;
1434 #endif
1435 
1436 	/* ASIC */
1437 	enum amd_asic_type		asic_type;
1438 	uint32_t			family;
1439 	uint32_t			rev_id;
1440 	uint32_t			external_rev_id;
1441 	unsigned long			flags;
1442 	int				usec_timeout;
1443 	const struct amdgpu_asic_funcs	*asic_funcs;
1444 	bool				shutdown;
1445 	bool				need_dma32;
1446 	bool				accel_working;
1447 	struct work_struct		reset_work;
1448 	struct notifier_block		acpi_nb;
1449 	struct amdgpu_i2c_chan		*i2c_bus[AMDGPU_MAX_I2C_BUS];
1450 	struct amdgpu_debugfs		debugfs[AMDGPU_DEBUGFS_MAX_COMPONENTS];
1451 	unsigned			debugfs_count;
1452 #if defined(CONFIG_DEBUG_FS)
1453 	struct dentry			*debugfs_regs[AMDGPU_DEBUGFS_MAX_COMPONENTS];
1454 #endif
1455 	struct amdgpu_atif		atif;
1456 	struct amdgpu_atcs		atcs;
1457 	struct mutex			srbm_mutex;
1458 	/* GRBM index mutex. Protects concurrent access to GRBM index */
1459 	struct mutex                    grbm_idx_mutex;
1460 	struct dev_pm_domain		vga_pm_domain;
1461 	bool				have_disp_power_ref;
1462 
1463 	/* BIOS */
1464 	bool				is_atom_fw;
1465 	uint8_t				*bios;
1466 	uint32_t			bios_size;
1467 	struct amdgpu_bo		*stollen_vga_memory;
1468 	uint32_t			bios_scratch_reg_offset;
1469 	uint32_t			bios_scratch[AMDGPU_BIOS_NUM_SCRATCH];
1470 
1471 	/* Register/doorbell mmio */
1472 	resource_size_t			rmmio_base;
1473 	resource_size_t			rmmio_size;
1474 	void __iomem			*rmmio;
1475 	/* protects concurrent MM_INDEX/DATA based register access */
1476 	spinlock_t mmio_idx_lock;
1477 	/* protects concurrent SMC based register access */
1478 	spinlock_t smc_idx_lock;
1479 	amdgpu_rreg_t			smc_rreg;
1480 	amdgpu_wreg_t			smc_wreg;
1481 	/* protects concurrent PCIE register access */
1482 	spinlock_t pcie_idx_lock;
1483 	amdgpu_rreg_t			pcie_rreg;
1484 	amdgpu_wreg_t			pcie_wreg;
1485 	amdgpu_rreg_t			pciep_rreg;
1486 	amdgpu_wreg_t			pciep_wreg;
1487 	/* protects concurrent UVD register access */
1488 	spinlock_t uvd_ctx_idx_lock;
1489 	amdgpu_rreg_t			uvd_ctx_rreg;
1490 	amdgpu_wreg_t			uvd_ctx_wreg;
1491 	/* protects concurrent DIDT register access */
1492 	spinlock_t didt_idx_lock;
1493 	amdgpu_rreg_t			didt_rreg;
1494 	amdgpu_wreg_t			didt_wreg;
1495 	/* protects concurrent gc_cac register access */
1496 	spinlock_t gc_cac_idx_lock;
1497 	amdgpu_rreg_t			gc_cac_rreg;
1498 	amdgpu_wreg_t			gc_cac_wreg;
1499 	/* protects concurrent ENDPOINT (audio) register access */
1500 	spinlock_t audio_endpt_idx_lock;
1501 	amdgpu_block_rreg_t		audio_endpt_rreg;
1502 	amdgpu_block_wreg_t		audio_endpt_wreg;
1503 	void __iomem                    *rio_mem;
1504 	resource_size_t			rio_mem_size;
1505 	struct amdgpu_doorbell		doorbell;
1506 
1507 	/* clock/pll info */
1508 	struct amdgpu_clock            clock;
1509 
1510 	/* MC */
1511 	struct amdgpu_mc		mc;
1512 	struct amdgpu_gart		gart;
1513 	struct amdgpu_dummy_page	dummy_page;
1514 	struct amdgpu_vm_manager	vm_manager;
1515 	struct amdgpu_vmhub             vmhub[AMDGPU_MAX_VMHUBS];
1516 
1517 	/* memory management */
1518 	struct amdgpu_mman		mman;
1519 	struct amdgpu_vram_scratch	vram_scratch;
1520 	struct amdgpu_wb		wb;
1521 	atomic64_t			vram_usage;
1522 	atomic64_t			vram_vis_usage;
1523 	atomic64_t			gtt_usage;
1524 	atomic64_t			num_bytes_moved;
1525 	atomic64_t			num_evictions;
1526 	atomic_t			gpu_reset_counter;
1527 
1528 	/* data for buffer migration throttling */
1529 	struct {
1530 		spinlock_t		lock;
1531 		s64			last_update_us;
1532 		s64			accum_us; /* accumulated microseconds */
1533 		u32			log2_max_MBps;
1534 	} mm_stats;
1535 
1536 	/* display */
1537 	bool				enable_virtual_display;
1538 	struct amdgpu_mode_info		mode_info;
1539 	struct work_struct		hotplug_work;
1540 	struct amdgpu_irq_src		crtc_irq;
1541 	struct amdgpu_irq_src		pageflip_irq;
1542 	struct amdgpu_irq_src		hpd_irq;
1543 
1544 	/* rings */
1545 	u64				fence_context;
1546 	unsigned			num_rings;
1547 	struct amdgpu_ring		*rings[AMDGPU_MAX_RINGS];
1548 	bool				ib_pool_ready;
1549 	struct amdgpu_sa_manager	ring_tmp_bo;
1550 
1551 	/* interrupts */
1552 	struct amdgpu_irq		irq;
1553 
1554 	/* powerplay */
1555 	struct amd_powerplay		powerplay;
1556 	bool				pp_enabled;
1557 	bool				pp_force_state_enabled;
1558 
1559 	/* dpm */
1560 	struct amdgpu_pm		pm;
1561 	u32				cg_flags;
1562 	u32				pg_flags;
1563 
1564 	/* amdgpu smumgr */
1565 	struct amdgpu_smumgr smu;
1566 
1567 	/* gfx */
1568 	struct amdgpu_gfx		gfx;
1569 
1570 	/* sdma */
1571 	struct amdgpu_sdma		sdma;
1572 
1573 	/* uvd */
1574 	struct amdgpu_uvd		uvd;
1575 
1576 	/* vce */
1577 	struct amdgpu_vce		vce;
1578 
1579 	/* firmwares */
1580 	struct amdgpu_firmware		firmware;
1581 
1582 	/* PSP */
1583 	struct psp_context		psp;
1584 
1585 	/* GDS */
1586 	struct amdgpu_gds		gds;
1587 
1588 	struct amdgpu_ip_block          ip_blocks[AMDGPU_MAX_IP_NUM];
1589 	int				num_ip_blocks;
1590 	struct mutex	mn_lock;
1591 	DECLARE_HASHTABLE(mn_hash, 7);
1592 
1593 	/* tracking pinned memory */
1594 	u64 vram_pin_size;
1595 	u64 invisible_pin_size;
1596 	u64 gart_pin_size;
1597 
1598 	/* amdkfd interface */
1599 	struct kfd_dev          *kfd;
1600 
1601 	struct amdgpu_virt	virt;
1602 
1603 	/* link all shadow bo */
1604 	struct list_head                shadow_list;
1605 	struct mutex                    shadow_list_lock;
1606 	/* link all gtt */
1607 	spinlock_t			gtt_list_lock;
1608 	struct list_head                gtt_list;
1609 
1610 	/* record hw reset is performed */
1611 	bool has_hw_reset;
1612 
1613 };
1614 
1615 static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev)
1616 {
1617 	return container_of(bdev, struct amdgpu_device, mman.bdev);
1618 }
1619 
1620 bool amdgpu_device_is_px(struct drm_device *dev);
1621 int amdgpu_device_init(struct amdgpu_device *adev,
1622 		       struct drm_device *ddev,
1623 		       struct pci_dev *pdev,
1624 		       uint32_t flags);
1625 void amdgpu_device_fini(struct amdgpu_device *adev);
1626 int amdgpu_gpu_wait_for_idle(struct amdgpu_device *adev);
1627 
1628 uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg,
1629 			uint32_t acc_flags);
1630 void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v,
1631 		    uint32_t acc_flags);
1632 u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg);
1633 void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v);
1634 
1635 u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index);
1636 void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v);
1637 u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index);
1638 void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v);
1639 
1640 /*
1641  * Registers read & write functions.
1642  */
1643 
1644 #define AMDGPU_REGS_IDX       (1<<0)
1645 #define AMDGPU_REGS_NO_KIQ    (1<<1)
1646 
1647 #define RREG32_NO_KIQ(reg) amdgpu_mm_rreg(adev, (reg), AMDGPU_REGS_NO_KIQ)
1648 #define WREG32_NO_KIQ(reg, v) amdgpu_mm_wreg(adev, (reg), (v), AMDGPU_REGS_NO_KIQ)
1649 
1650 #define RREG32(reg) amdgpu_mm_rreg(adev, (reg), 0)
1651 #define RREG32_IDX(reg) amdgpu_mm_rreg(adev, (reg), AMDGPU_REGS_IDX)
1652 #define DREG32(reg) printk(KERN_INFO "REGISTER: " #reg " : 0x%08X\n", amdgpu_mm_rreg(adev, (reg), 0))
1653 #define WREG32(reg, v) amdgpu_mm_wreg(adev, (reg), (v), 0)
1654 #define WREG32_IDX(reg, v) amdgpu_mm_wreg(adev, (reg), (v), AMDGPU_REGS_IDX)
1655 #define REG_SET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
1656 #define REG_GET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
1657 #define RREG32_PCIE(reg) adev->pcie_rreg(adev, (reg))
1658 #define WREG32_PCIE(reg, v) adev->pcie_wreg(adev, (reg), (v))
1659 #define RREG32_PCIE_PORT(reg) adev->pciep_rreg(adev, (reg))
1660 #define WREG32_PCIE_PORT(reg, v) adev->pciep_wreg(adev, (reg), (v))
1661 #define RREG32_SMC(reg) adev->smc_rreg(adev, (reg))
1662 #define WREG32_SMC(reg, v) adev->smc_wreg(adev, (reg), (v))
1663 #define RREG32_UVD_CTX(reg) adev->uvd_ctx_rreg(adev, (reg))
1664 #define WREG32_UVD_CTX(reg, v) adev->uvd_ctx_wreg(adev, (reg), (v))
1665 #define RREG32_DIDT(reg) adev->didt_rreg(adev, (reg))
1666 #define WREG32_DIDT(reg, v) adev->didt_wreg(adev, (reg), (v))
1667 #define RREG32_GC_CAC(reg) adev->gc_cac_rreg(adev, (reg))
1668 #define WREG32_GC_CAC(reg, v) adev->gc_cac_wreg(adev, (reg), (v))
1669 #define RREG32_AUDIO_ENDPT(block, reg) adev->audio_endpt_rreg(adev, (block), (reg))
1670 #define WREG32_AUDIO_ENDPT(block, reg, v) adev->audio_endpt_wreg(adev, (block), (reg), (v))
1671 #define WREG32_P(reg, val, mask)				\
1672 	do {							\
1673 		uint32_t tmp_ = RREG32(reg);			\
1674 		tmp_ &= (mask);					\
1675 		tmp_ |= ((val) & ~(mask));			\
1676 		WREG32(reg, tmp_);				\
1677 	} while (0)
1678 #define WREG32_AND(reg, and) WREG32_P(reg, 0, and)
1679 #define WREG32_OR(reg, or) WREG32_P(reg, or, ~(or))
1680 #define WREG32_PLL_P(reg, val, mask)				\
1681 	do {							\
1682 		uint32_t tmp_ = RREG32_PLL(reg);		\
1683 		tmp_ &= (mask);					\
1684 		tmp_ |= ((val) & ~(mask));			\
1685 		WREG32_PLL(reg, tmp_);				\
1686 	} while (0)
1687 #define DREG32_SYS(sqf, adev, reg) seq_printf((sqf), #reg " : 0x%08X\n", amdgpu_mm_rreg((adev), (reg), false))
1688 #define RREG32_IO(reg) amdgpu_io_rreg(adev, (reg))
1689 #define WREG32_IO(reg, v) amdgpu_io_wreg(adev, (reg), (v))
1690 
1691 #define RDOORBELL32(index) amdgpu_mm_rdoorbell(adev, (index))
1692 #define WDOORBELL32(index, v) amdgpu_mm_wdoorbell(adev, (index), (v))
1693 #define RDOORBELL64(index) amdgpu_mm_rdoorbell64(adev, (index))
1694 #define WDOORBELL64(index, v) amdgpu_mm_wdoorbell64(adev, (index), (v))
1695 
1696 #define REG_FIELD_SHIFT(reg, field) reg##__##field##__SHIFT
1697 #define REG_FIELD_MASK(reg, field) reg##__##field##_MASK
1698 
1699 #define REG_SET_FIELD(orig_val, reg, field, field_val)			\
1700 	(((orig_val) & ~REG_FIELD_MASK(reg, field)) |			\
1701 	 (REG_FIELD_MASK(reg, field) & ((field_val) << REG_FIELD_SHIFT(reg, field))))
1702 
1703 #define REG_GET_FIELD(value, reg, field)				\
1704 	(((value) & REG_FIELD_MASK(reg, field)) >> REG_FIELD_SHIFT(reg, field))
1705 
1706 #define WREG32_FIELD(reg, field, val)	\
1707 	WREG32(mm##reg, (RREG32(mm##reg) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
1708 
1709 #define WREG32_FIELD_OFFSET(reg, offset, field, val)	\
1710 	WREG32(mm##reg + offset, (RREG32(mm##reg + offset) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
1711 
1712 /*
1713  * BIOS helpers.
1714  */
1715 #define RBIOS8(i) (adev->bios[i])
1716 #define RBIOS16(i) (RBIOS8(i) | (RBIOS8((i)+1) << 8))
1717 #define RBIOS32(i) ((RBIOS16(i)) | (RBIOS16((i)+2) << 16))
1718 
1719 /*
1720  * RING helpers.
1721  */
1722 static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v)
1723 {
1724 	if (ring->count_dw <= 0)
1725 		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
1726 	ring->ring[ring->wptr++ & ring->buf_mask] = v;
1727 	ring->wptr &= ring->ptr_mask;
1728 	ring->count_dw--;
1729 }
1730 
1731 static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring, void *src, int count_dw)
1732 {
1733 	unsigned occupied, chunk1, chunk2;
1734 	void *dst;
1735 
1736 	if (ring->count_dw < count_dw) {
1737 		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
1738 	} else {
1739 		occupied = ring->wptr & ring->buf_mask;
1740 		dst = (void *)&ring->ring[occupied];
1741 		chunk1 = ring->buf_mask + 1 - occupied;
1742 		chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1;
1743 		chunk2 = count_dw - chunk1;
1744 		chunk1 <<= 2;
1745 		chunk2 <<= 2;
1746 
1747 		if (chunk1)
1748 			memcpy(dst, src, chunk1);
1749 
1750 		if (chunk2) {
1751 			src += chunk1;
1752 			dst = (void *)ring->ring;
1753 			memcpy(dst, src, chunk2);
1754 		}
1755 
1756 		ring->wptr += count_dw;
1757 		ring->wptr &= ring->ptr_mask;
1758 		ring->count_dw -= count_dw;
1759 	}
1760 }
1761 
1762 static inline struct amdgpu_sdma_instance *
1763 amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
1764 {
1765 	struct amdgpu_device *adev = ring->adev;
1766 	int i;
1767 
1768 	for (i = 0; i < adev->sdma.num_instances; i++)
1769 		if (&adev->sdma.instance[i].ring == ring)
1770 			break;
1771 
1772 	if (i < AMDGPU_MAX_SDMA_INSTANCES)
1773 		return &adev->sdma.instance[i];
1774 	else
1775 		return NULL;
1776 }
1777 
1778 /*
1779  * ASICs macro.
1780  */
1781 #define amdgpu_asic_set_vga_state(adev, state) (adev)->asic_funcs->set_vga_state((adev), (state))
1782 #define amdgpu_asic_reset(adev) (adev)->asic_funcs->reset((adev))
1783 #define amdgpu_asic_get_xclk(adev) (adev)->asic_funcs->get_xclk((adev))
1784 #define amdgpu_asic_set_uvd_clocks(adev, v, d) (adev)->asic_funcs->set_uvd_clocks((adev), (v), (d))
1785 #define amdgpu_asic_set_vce_clocks(adev, ev, ec) (adev)->asic_funcs->set_vce_clocks((adev), (ev), (ec))
1786 #define amdgpu_get_pcie_lanes(adev) (adev)->asic_funcs->get_pcie_lanes((adev))
1787 #define amdgpu_set_pcie_lanes(adev, l) (adev)->asic_funcs->set_pcie_lanes((adev), (l))
1788 #define amdgpu_asic_get_gpu_clock_counter(adev) (adev)->asic_funcs->get_gpu_clock_counter((adev))
1789 #define amdgpu_asic_read_disabled_bios(adev) (adev)->asic_funcs->read_disabled_bios((adev))
1790 #define amdgpu_asic_read_bios_from_rom(adev, b, l) (adev)->asic_funcs->read_bios_from_rom((adev), (b), (l))
1791 #define amdgpu_asic_read_register(adev, se, sh, offset, v)((adev)->asic_funcs->read_register((adev), (se), (sh), (offset), (v)))
1792 #define amdgpu_asic_get_config_memsize(adev) (adev)->asic_funcs->get_config_memsize((adev))
1793 #define amdgpu_gart_flush_gpu_tlb(adev, vmid) (adev)->gart.gart_funcs->flush_gpu_tlb((adev), (vmid))
1794 #define amdgpu_gart_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gart.gart_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags))
1795 #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count)))
1796 #define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr)))
1797 #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)))
1798 #define amdgpu_vm_get_pte_flags(adev, flags) (adev)->gart.gart_funcs->get_vm_pte_flags((adev),(flags))
1799 #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib)))
1800 #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r))
1801 #define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t))
1802 #define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r))
1803 #define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r))
1804 #define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r))
1805 #define amdgpu_ring_emit_ib(r, ib, vm_id, c) (r)->funcs->emit_ib((r), (ib), (vm_id), (c))
1806 #define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r))
1807 #define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr))
1808 #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags))
1809 #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))
1810 #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r))
1811 #define amdgpu_ring_emit_hdp_invalidate(r) (r)->funcs->emit_hdp_invalidate((r))
1812 #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r))
1813 #define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d))
1814 #define amdgpu_ring_emit_rreg(r, d) (r)->funcs->emit_rreg((r), (d))
1815 #define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v))
1816 #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
1817 #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
1818 #define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
1819 #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev))
1820 #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv))
1821 #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev))
1822 #define amdgpu_display_set_vga_render_state(adev, r) (adev)->mode_info.funcs->set_vga_render_state((adev), (r))
1823 #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc))
1824 #define amdgpu_display_vblank_wait(adev, crtc) (adev)->mode_info.funcs->vblank_wait((adev), (crtc))
1825 #define amdgpu_display_backlight_set_level(adev, e, l) (adev)->mode_info.funcs->backlight_set_level((e), (l))
1826 #define amdgpu_display_backlight_get_level(adev, e) (adev)->mode_info.funcs->backlight_get_level((e))
1827 #define amdgpu_display_hpd_sense(adev, h) (adev)->mode_info.funcs->hpd_sense((adev), (h))
1828 #define amdgpu_display_hpd_set_polarity(adev, h) (adev)->mode_info.funcs->hpd_set_polarity((adev), (h))
1829 #define amdgpu_display_hpd_get_gpio_reg(adev) (adev)->mode_info.funcs->hpd_get_gpio_reg((adev))
1830 #define amdgpu_display_bandwidth_update(adev) (adev)->mode_info.funcs->bandwidth_update((adev))
1831 #define amdgpu_display_page_flip(adev, crtc, base, async) (adev)->mode_info.funcs->page_flip((adev), (crtc), (base), (async))
1832 #define amdgpu_display_page_flip_get_scanoutpos(adev, crtc, vbl, pos) (adev)->mode_info.funcs->page_flip_get_scanoutpos((adev), (crtc), (vbl), (pos))
1833 #define amdgpu_display_add_encoder(adev, e, s, c) (adev)->mode_info.funcs->add_encoder((adev), (e), (s), (c))
1834 #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))
1835 #define amdgpu_display_stop_mc_access(adev, s) (adev)->mode_info.funcs->stop_mc_access((adev), (s))
1836 #define amdgpu_display_resume_mc_access(adev, s) (adev)->mode_info.funcs->resume_mc_access((adev), (s))
1837 #define amdgpu_emit_copy_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_copy_buffer((ib),  (s), (d), (b))
1838 #define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b))
1839 #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev))
1840 #define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance))
1841 #define amdgpu_gds_switch(adev, r, v, d, w, a) (adev)->gds.funcs->patch_gds_switch((r), (v), (d), (w), (a))
1842 #define amdgpu_psp_check_fw_loading_status(adev, i) (adev)->firmware.funcs->check_fw_loading_status((adev), (i))
1843 
1844 /* Common functions */
1845 int amdgpu_gpu_reset(struct amdgpu_device *adev);
1846 bool amdgpu_need_backup(struct amdgpu_device *adev);
1847 void amdgpu_pci_config_reset(struct amdgpu_device *adev);
1848 bool amdgpu_need_post(struct amdgpu_device *adev);
1849 void amdgpu_update_display_priority(struct amdgpu_device *adev);
1850 
1851 int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data);
1852 int amdgpu_cs_get_ring(struct amdgpu_device *adev, u32 ip_type,
1853 		       u32 ip_instance, u32 ring,
1854 		       struct amdgpu_ring **out_ring);
1855 void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes);
1856 void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain);
1857 bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
1858 int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages);
1859 int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
1860 				     uint32_t flags);
1861 bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm);
1862 struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm);
1863 bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
1864 				  unsigned long end);
1865 bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm,
1866 				       int *last_invalidated);
1867 bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm);
1868 uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm,
1869 				 struct ttm_mem_reg *mem);
1870 void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base);
1871 void amdgpu_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc);
1872 void amdgpu_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size);
1873 int amdgpu_ttm_init(struct amdgpu_device *adev);
1874 void amdgpu_ttm_fini(struct amdgpu_device *adev);
1875 void amdgpu_program_register_sequence(struct amdgpu_device *adev,
1876 					     const u32 *registers,
1877 					     const u32 array_size);
1878 
1879 bool amdgpu_device_is_px(struct drm_device *dev);
1880 /* atpx handler */
1881 #if defined(CONFIG_VGA_SWITCHEROO)
1882 void amdgpu_register_atpx_handler(void);
1883 void amdgpu_unregister_atpx_handler(void);
1884 bool amdgpu_has_atpx_dgpu_power_cntl(void);
1885 bool amdgpu_is_atpx_hybrid(void);
1886 bool amdgpu_atpx_dgpu_req_power_for_displays(void);
1887 bool amdgpu_has_atpx(void);
1888 #else
1889 static inline void amdgpu_register_atpx_handler(void) {}
1890 static inline void amdgpu_unregister_atpx_handler(void) {}
1891 static inline bool amdgpu_has_atpx_dgpu_power_cntl(void) { return false; }
1892 static inline bool amdgpu_is_atpx_hybrid(void) { return false; }
1893 static inline bool amdgpu_atpx_dgpu_req_power_for_displays(void) { return false; }
1894 static inline bool amdgpu_has_atpx(void) { return false; }
1895 #endif
1896 
1897 /*
1898  * KMS
1899  */
1900 extern const struct drm_ioctl_desc amdgpu_ioctls_kms[];
1901 extern const int amdgpu_max_kms_ioctl;
1902 
1903 int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags);
1904 void amdgpu_driver_unload_kms(struct drm_device *dev);
1905 void amdgpu_driver_lastclose_kms(struct drm_device *dev);
1906 int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv);
1907 void amdgpu_driver_postclose_kms(struct drm_device *dev,
1908 				 struct drm_file *file_priv);
1909 int amdgpu_suspend(struct amdgpu_device *adev);
1910 int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon);
1911 int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon);
1912 u32 amdgpu_get_vblank_counter_kms(struct drm_device *dev, unsigned int pipe);
1913 int amdgpu_enable_vblank_kms(struct drm_device *dev, unsigned int pipe);
1914 void amdgpu_disable_vblank_kms(struct drm_device *dev, unsigned int pipe);
1915 int amdgpu_get_vblank_timestamp_kms(struct drm_device *dev, unsigned int pipe,
1916 				    int *max_error,
1917 				    struct timeval *vblank_time,
1918 				    unsigned flags);
1919 long amdgpu_kms_compat_ioctl(struct file *filp, unsigned int cmd,
1920 			     unsigned long arg);
1921 
1922 /*
1923  * functions used by amdgpu_encoder.c
1924  */
1925 struct amdgpu_afmt_acr {
1926 	u32 clock;
1927 
1928 	int n_32khz;
1929 	int cts_32khz;
1930 
1931 	int n_44_1khz;
1932 	int cts_44_1khz;
1933 
1934 	int n_48khz;
1935 	int cts_48khz;
1936 
1937 };
1938 
1939 struct amdgpu_afmt_acr amdgpu_afmt_acr(uint32_t clock);
1940 
1941 /* amdgpu_acpi.c */
1942 #if defined(CONFIG_ACPI)
1943 int amdgpu_acpi_init(struct amdgpu_device *adev);
1944 void amdgpu_acpi_fini(struct amdgpu_device *adev);
1945 bool amdgpu_acpi_is_pcie_performance_request_supported(struct amdgpu_device *adev);
1946 int amdgpu_acpi_pcie_performance_request(struct amdgpu_device *adev,
1947 						u8 perf_req, bool advertise);
1948 int amdgpu_acpi_pcie_notify_device_ready(struct amdgpu_device *adev);
1949 #else
1950 static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; }
1951 static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { }
1952 #endif
1953 
1954 struct amdgpu_bo_va_mapping *
1955 amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
1956 		       uint64_t addr, struct amdgpu_bo **bo);
1957 int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser);
1958 
1959 #include "amdgpu_object.h"
1960 #endif
1961