2 * Copyright 2014 Advanced Micro Devices, Inc.
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
24 #ifndef __AMDGPU_GFX_H__
25 #define __AMDGPU_GFX_H__
30 #include "clearstate_defs.h"
31 #include "amdgpu_ring.h"
32 #include "amdgpu_rlc.h"
34 #include "amdgpu_ras.h"
36 /* GFX current status */
37 #define AMDGPU_GFX_NORMAL_MODE 0x00000000L
38 #define AMDGPU_GFX_SAFE_MODE 0x00000001L
39 #define AMDGPU_GFX_PG_DISABLED_MODE 0x00000002L
40 #define AMDGPU_GFX_CG_DISABLED_MODE 0x00000004L
41 #define AMDGPU_GFX_LBPW_DISABLED_MODE 0x00000008L
43 #define AMDGPU_MAX_GFX_QUEUES KGD_MAX_QUEUES
44 #define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES
46 enum amdgpu_gfx_pipe_priority {
47 AMDGPU_GFX_PIPE_PRIO_NORMAL = AMDGPU_RING_PRIO_1,
48 AMDGPU_GFX_PIPE_PRIO_HIGH = AMDGPU_RING_PRIO_2
51 #define AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM 0
52 #define AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM 15
55 struct amdgpu_bo *hpd_eop_obj;
57 struct amdgpu_bo *mec_fw_obj;
61 u32 num_queue_per_pipe;
62 void *mqd_backup[AMDGPU_MAX_COMPUTE_RINGS + 1];
64 /* These are the resources for which amdgpu takes ownership */
65 DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
68 enum amdgpu_unmap_queues_action {
71 DISABLE_PROCESS_QUEUES,
72 PREEMPT_QUEUES_NO_UNMAP,
75 struct kiq_pm4_funcs {
76 /* Support ASIC-specific kiq pm4 packets*/
77 void (*kiq_set_resources)(struct amdgpu_ring *kiq_ring,
79 void (*kiq_map_queues)(struct amdgpu_ring *kiq_ring,
80 struct amdgpu_ring *ring);
81 void (*kiq_unmap_queues)(struct amdgpu_ring *kiq_ring,
82 struct amdgpu_ring *ring,
83 enum amdgpu_unmap_queues_action action,
84 u64 gpu_addr, u64 seq);
85 void (*kiq_query_status)(struct amdgpu_ring *kiq_ring,
86 struct amdgpu_ring *ring,
89 void (*kiq_invalidate_tlbs)(struct amdgpu_ring *kiq_ring,
90 uint16_t pasid, uint32_t flush_type,
93 int set_resources_size;
95 int unmap_queues_size;
96 int query_status_size;
97 int invalidate_tlbs_size;
102 struct amdgpu_bo *eop_obj;
103 spinlock_t ring_lock;
104 struct amdgpu_ring ring;
105 struct amdgpu_irq_src irq;
106 const struct kiq_pm4_funcs *pmf;
110 * GPU scratch registers structures, functions & helpers
112 struct amdgpu_scratch {
121 #define AMDGPU_GFX_MAX_SE 4
122 #define AMDGPU_GFX_MAX_SH_PER_SE 2
124 struct amdgpu_rb_config {
125 uint32_t rb_backend_disable;
126 uint32_t user_rb_backend_disable;
127 uint32_t raster_config;
128 uint32_t raster_config_1;
131 struct gb_addr_config {
132 uint16_t pipe_interleave_size;
134 uint8_t max_compress_frags;
137 uint8_t num_rb_per_se;
141 struct amdgpu_gfx_config {
142 unsigned max_shader_engines;
143 unsigned max_tile_pipes;
144 unsigned max_cu_per_sh;
145 unsigned max_sh_per_se;
146 unsigned max_backends_per_se;
147 unsigned max_texture_channel_caches;
149 unsigned max_gs_threads;
150 unsigned max_hw_contexts;
151 unsigned sc_prim_fifo_size_frontend;
152 unsigned sc_prim_fifo_size_backend;
153 unsigned sc_hiz_tile_fifo_size;
154 unsigned sc_earlyz_tile_fifo_size;
156 unsigned num_tile_pipes;
157 unsigned backend_enable_mask;
158 unsigned mem_max_burst_length_bytes;
159 unsigned mem_row_size_in_kb;
160 unsigned shader_engine_tile_size;
162 unsigned multi_gpu_tile_size;
163 unsigned mc_arb_ramcfg;
166 unsigned gb_addr_config;
168 unsigned gs_vgt_table_depth;
169 unsigned gs_prim_buffer_depth;
171 uint32_t tile_mode_array[32];
172 uint32_t macrotile_mode_array[16];
174 struct gb_addr_config gb_addr_config_fields;
175 struct amdgpu_rb_config rb_config[AMDGPU_GFX_MAX_SE][AMDGPU_GFX_MAX_SH_PER_SE];
177 /* gfx configure feature */
178 uint32_t double_offchip_lds_buf;
179 /* cached value of DB_DEBUG2 */
181 /* gfx10 specific config */
182 uint32_t num_sc_per_sh;
183 uint32_t num_packer_per_sc;
184 uint32_t pa_sc_tile_steering_override;
185 uint64_t tcc_disabled_mask;
186 uint32_t gc_num_tcp_per_sa;
187 uint32_t gc_num_sdp_interface;
188 uint32_t gc_num_tcps;
189 uint32_t gc_num_tcp_per_wpg;
190 uint32_t gc_tcp_l1_size;
191 uint32_t gc_num_sqc_per_wgp;
192 uint32_t gc_l1_instruction_cache_size_per_sqc;
193 uint32_t gc_l1_data_cache_size_per_sqc;
194 uint32_t gc_gl1c_per_sa;
195 uint32_t gc_gl1c_size_per_instance;
196 uint32_t gc_gl2c_per_gpu;
199 struct amdgpu_cu_info {
200 uint32_t simd_per_cu;
201 uint32_t max_waves_per_simd;
202 uint32_t wave_front_size;
203 uint32_t max_scratch_slots_per_cu;
206 /* total active CU number */
209 uint32_t ao_cu_bitmap[4][4];
210 uint32_t bitmap[4][4];
213 struct amdgpu_gfx_ras {
214 struct amdgpu_ras_block_object ras_block;
215 void (*enable_watchdog_timer)(struct amdgpu_device *adev);
216 bool (*query_utcl2_poison_status)(struct amdgpu_device *adev);
219 struct amdgpu_gfx_funcs {
220 /* get the gpu clock counter */
221 uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev);
222 void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num,
223 u32 sh_num, u32 instance);
224 void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd,
225 uint32_t wave, uint32_t *dst, int *no_fields);
226 void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t simd,
227 uint32_t wave, uint32_t thread, uint32_t start,
228 uint32_t size, uint32_t *dst);
229 void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd,
230 uint32_t wave, uint32_t start, uint32_t size,
232 void (*select_me_pipe_q)(struct amdgpu_device *adev, u32 me, u32 pipe,
233 u32 queue, u32 vmid);
234 void (*init_spm_golden)(struct amdgpu_device *adev);
235 void (*update_perfmon_mgcg)(struct amdgpu_device *adev, bool enable);
239 struct work_struct work;
244 struct amdgpu_bo *pfp_fw_obj;
245 uint64_t pfp_fw_gpu_addr;
246 uint32_t *pfp_fw_ptr;
250 struct amdgpu_bo *ce_fw_obj;
251 uint64_t ce_fw_gpu_addr;
256 struct amdgpu_bo *me_fw_obj;
257 uint64_t me_fw_gpu_addr;
260 uint32_t num_pipe_per_me;
261 uint32_t num_queue_per_pipe;
262 void *mqd_backup[AMDGPU_MAX_GFX_RINGS];
264 /* These are the resources for which amdgpu takes ownership */
265 DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
269 struct mutex gpu_clock_mutex;
270 struct amdgpu_gfx_config config;
271 struct amdgpu_rlc rlc;
272 struct amdgpu_pfp pfp;
275 struct amdgpu_mec mec;
276 struct amdgpu_kiq kiq;
277 struct amdgpu_scratch scratch;
278 const struct firmware *me_fw; /* ME firmware */
279 uint32_t me_fw_version;
280 const struct firmware *pfp_fw; /* PFP firmware */
281 uint32_t pfp_fw_version;
282 const struct firmware *ce_fw; /* CE firmware */
283 uint32_t ce_fw_version;
284 const struct firmware *rlc_fw; /* RLC firmware */
285 uint32_t rlc_fw_version;
286 const struct firmware *mec_fw; /* MEC firmware */
287 uint32_t mec_fw_version;
288 const struct firmware *mec2_fw; /* MEC2 firmware */
289 uint32_t mec2_fw_version;
290 uint32_t me_feature_version;
291 uint32_t ce_feature_version;
292 uint32_t pfp_feature_version;
293 uint32_t rlc_feature_version;
294 uint32_t rlc_srlc_fw_version;
295 uint32_t rlc_srlc_feature_version;
296 uint32_t rlc_srlg_fw_version;
297 uint32_t rlc_srlg_feature_version;
298 uint32_t rlc_srls_fw_version;
299 uint32_t rlc_srls_feature_version;
300 uint32_t mec_feature_version;
301 uint32_t mec2_feature_version;
302 bool mec_fw_write_wait;
303 bool me_fw_write_wait;
304 bool cp_fw_write_wait;
305 struct amdgpu_ring gfx_ring[AMDGPU_MAX_GFX_RINGS];
306 unsigned num_gfx_rings;
307 struct amdgpu_ring compute_ring[AMDGPU_MAX_COMPUTE_RINGS];
308 unsigned num_compute_rings;
309 struct amdgpu_irq_src eop_irq;
310 struct amdgpu_irq_src priv_reg_irq;
311 struct amdgpu_irq_src priv_inst_irq;
312 struct amdgpu_irq_src cp_ecc_error_irq;
313 struct amdgpu_irq_src sq_irq;
314 struct sq_work sq_work;
317 uint32_t gfx_current_status;
319 unsigned ce_ram_size;
320 struct amdgpu_cu_info cu_info;
321 const struct amdgpu_gfx_funcs *funcs;
324 uint32_t grbm_soft_reset;
325 uint32_t srbm_soft_reset;
328 bool gfx_off_state; /* true: enabled, false: disabled */
329 struct mutex gfx_off_mutex;
330 uint32_t gfx_off_req_count; /* default 1, enable gfx off: dec 1, disable gfx off: add 1 */
331 struct delayed_work gfx_off_delay_work;
333 /* pipe reservation */
334 struct mutex pipe_reserve_mutex;
335 DECLARE_BITMAP (pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
338 struct ras_common_if *ras_if;
339 struct amdgpu_gfx_ras *ras;
342 #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev))
343 #define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance))
344 #define amdgpu_gfx_select_me_pipe_q(adev, me, pipe, q, vmid) (adev)->gfx.funcs->select_me_pipe_q((adev), (me), (pipe), (q), (vmid))
345 #define amdgpu_gfx_init_spm_golden(adev) (adev)->gfx.funcs->init_spm_golden((adev))
348 * amdgpu_gfx_create_bitmask - create a bitmask
350 * @bit_width: length of the mask
352 * create a variable length bit mask.
353 * Returns the bitmask.
355 static inline u32 amdgpu_gfx_create_bitmask(u32 bit_width)
357 return (u32)((1ULL << bit_width) - 1);
360 int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg);
361 void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg);
363 void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se,
366 int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
367 struct amdgpu_ring *ring,
368 struct amdgpu_irq_src *irq);
370 void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring);
372 void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev);
373 int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
376 int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
378 void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev);
379 int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev);
380 int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev);
382 void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev);
383 void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev);
385 int amdgpu_gfx_mec_queue_to_bit(struct amdgpu_device *adev, int mec,
386 int pipe, int queue);
387 void amdgpu_queue_mask_bit_to_mec_queue(struct amdgpu_device *adev, int bit,
388 int *mec, int *pipe, int *queue);
389 bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, int mec,
390 int pipe, int queue);
391 bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev,
392 struct amdgpu_ring *ring);
393 int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev, int me,
394 int pipe, int queue);
395 void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit,
396 int *me, int *pipe, int *queue);
397 bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev, int me,
398 int pipe, int queue);
399 void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable);
400 int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value);
401 int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
402 int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev,
404 struct amdgpu_iv_entry *entry);
405 int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev,
406 struct amdgpu_irq_src *source,
407 struct amdgpu_iv_entry *entry);
408 uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg);
409 void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v);
410 int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev);