diff options
Diffstat (limited to 'drivers/gpu/nvgpu/gk20a/gk20a.h')
-rw-r--r-- | drivers/gpu/nvgpu/gk20a/gk20a.h | 1474 |
1 files changed, 1474 insertions, 0 deletions
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.h b/drivers/gpu/nvgpu/gk20a/gk20a.h new file mode 100644 index 00000000..09597bb2 --- /dev/null +++ b/drivers/gpu/nvgpu/gk20a/gk20a.h | |||
@@ -0,0 +1,1474 @@ | |||
1 | /* | ||
2 | * Copyright (c) 2011-2017, NVIDIA CORPORATION. All rights reserved. | ||
3 | * | ||
4 | * GK20A Graphics | ||
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 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | ||
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING | ||
21 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER | ||
22 | * DEALINGS IN THE SOFTWARE. | ||
23 | */ | ||
24 | #ifndef GK20A_H | ||
25 | #define GK20A_H | ||
26 | |||
27 | struct gk20a; | ||
28 | struct fifo_gk20a; | ||
29 | struct channel_gk20a; | ||
30 | struct gr_gk20a; | ||
31 | struct sim_gk20a; | ||
32 | struct gk20a_ctxsw_ucode_segments; | ||
33 | struct gk20a_fecs_trace; | ||
34 | struct gk20a_ctxsw_trace; | ||
35 | struct acr_desc; | ||
36 | struct nvgpu_mem_alloc_tracker; | ||
37 | struct dbg_profiler_object_data; | ||
38 | struct ecc_gk20a; | ||
39 | struct gk20a_debug_output; | ||
40 | struct nvgpu_clk_pll_debug_data; | ||
41 | struct nvgpu_nvhost_dev; | ||
42 | struct nvgpu_cpu_time_correlation_sample; | ||
43 | struct nvgpu_mem_sgt; | ||
44 | struct nvgpu_warpstate; | ||
45 | |||
46 | #include <nvgpu/lock.h> | ||
47 | #include <nvgpu/thread.h> | ||
48 | #include <nvgpu/io.h> | ||
49 | |||
50 | #include <nvgpu/mm.h> | ||
51 | #include <nvgpu/as.h> | ||
52 | #include <nvgpu/log.h> | ||
53 | #include <nvgpu/pramin.h> | ||
54 | #include <nvgpu/acr/nvgpu_acr.h> | ||
55 | #include <nvgpu/kref.h> | ||
56 | #include <nvgpu/falcon.h> | ||
57 | #include <nvgpu/pmu.h> | ||
58 | #include <nvgpu/atomic.h> | ||
59 | #include <nvgpu/barrier.h> | ||
60 | #include <nvgpu/rwsem.h> | ||
61 | #include <nvgpu/clk_arb.h> | ||
62 | |||
63 | #include "clk_gk20a.h" | ||
64 | #include "ce2_gk20a.h" | ||
65 | #include "fifo_gk20a.h" | ||
66 | #include "tsg_gk20a.h" | ||
67 | #include "gr_gk20a.h" | ||
68 | #include "sim_gk20a.h" | ||
69 | #include "pmu_gk20a.h" | ||
70 | #include "priv_ring_gk20a.h" | ||
71 | #include "therm_gk20a.h" | ||
72 | #ifdef CONFIG_ARCH_TEGRA_18x_SOC | ||
73 | #include "clk/clk.h" | ||
74 | #include "perf/perf.h" | ||
75 | #include "pmgr/pmgr.h" | ||
76 | #include "therm/thrm.h" | ||
77 | #endif | ||
78 | #include "ecc_gk20a.h" | ||
79 | |||
80 | /* PTIMER_REF_FREQ_HZ corresponds to a period of 32 nanoseconds. | ||
81 | 32 ns is the resolution of ptimer. */ | ||
82 | #define PTIMER_REF_FREQ_HZ 31250000 | ||
83 | |||
84 | #ifdef CONFIG_DEBUG_FS | ||
85 | struct railgate_stats { | ||
86 | unsigned long last_rail_gate_start; | ||
87 | unsigned long last_rail_gate_complete; | ||
88 | unsigned long last_rail_ungate_start; | ||
89 | unsigned long last_rail_ungate_complete; | ||
90 | unsigned long total_rail_gate_time_ms; | ||
91 | unsigned long total_rail_ungate_time_ms; | ||
92 | unsigned long railgating_cycle_count; | ||
93 | }; | ||
94 | #endif | ||
95 | |||
96 | enum gk20a_cbc_op { | ||
97 | gk20a_cbc_op_clear, | ||
98 | gk20a_cbc_op_clean, | ||
99 | gk20a_cbc_op_invalidate, | ||
100 | }; | ||
101 | |||
102 | #define MC_INTR_UNIT_DISABLE false | ||
103 | #define MC_INTR_UNIT_ENABLE true | ||
104 | |||
105 | #define GPU_LIT_NUM_GPCS 0 | ||
106 | #define GPU_LIT_NUM_PES_PER_GPC 1 | ||
107 | #define GPU_LIT_NUM_ZCULL_BANKS 2 | ||
108 | #define GPU_LIT_NUM_TPC_PER_GPC 3 | ||
109 | #define GPU_LIT_NUM_SM_PER_TPC 4 | ||
110 | #define GPU_LIT_NUM_FBPS 5 | ||
111 | #define GPU_LIT_GPC_BASE 6 | ||
112 | #define GPU_LIT_GPC_STRIDE 7 | ||
113 | #define GPU_LIT_GPC_SHARED_BASE 8 | ||
114 | #define GPU_LIT_TPC_IN_GPC_BASE 9 | ||
115 | #define GPU_LIT_TPC_IN_GPC_STRIDE 10 | ||
116 | #define GPU_LIT_TPC_IN_GPC_SHARED_BASE 11 | ||
117 | #define GPU_LIT_PPC_IN_GPC_BASE 12 | ||
118 | #define GPU_LIT_PPC_IN_GPC_STRIDE 13 | ||
119 | #define GPU_LIT_PPC_IN_GPC_SHARED_BASE 14 | ||
120 | #define GPU_LIT_ROP_BASE 15 | ||
121 | #define GPU_LIT_ROP_STRIDE 16 | ||
122 | #define GPU_LIT_ROP_SHARED_BASE 17 | ||
123 | #define GPU_LIT_HOST_NUM_ENGINES 18 | ||
124 | #define GPU_LIT_HOST_NUM_PBDMA 19 | ||
125 | #define GPU_LIT_LTC_STRIDE 20 | ||
126 | #define GPU_LIT_LTS_STRIDE 21 | ||
127 | #define GPU_LIT_NUM_FBPAS 22 | ||
128 | #define GPU_LIT_FBPA_STRIDE 23 | ||
129 | #define GPU_LIT_FBPA_BASE 24 | ||
130 | #define GPU_LIT_FBPA_SHARED_BASE 25 | ||
131 | #define GPU_LIT_SM_PRI_STRIDE 26 | ||
132 | #define GPU_LIT_SMPC_PRI_BASE 27 | ||
133 | #define GPU_LIT_SMPC_PRI_SHARED_BASE 28 | ||
134 | #define GPU_LIT_SMPC_PRI_UNIQUE_BASE 29 | ||
135 | #define GPU_LIT_SMPC_PRI_STRIDE 30 | ||
136 | #define GPU_LIT_TWOD_CLASS 31 | ||
137 | #define GPU_LIT_THREED_CLASS 32 | ||
138 | #define GPU_LIT_COMPUTE_CLASS 33 | ||
139 | #define GPU_LIT_GPFIFO_CLASS 34 | ||
140 | #define GPU_LIT_I2M_CLASS 35 | ||
141 | #define GPU_LIT_DMA_COPY_CLASS 36 | ||
142 | |||
143 | #define nvgpu_get_litter_value(g, v) (g)->ops.get_litter_value((g), v) | ||
144 | |||
145 | enum nvgpu_unit; | ||
146 | |||
147 | enum nvgpu_flush_op; | ||
148 | /* | ||
149 | * gpu_ops should only contain function pointers! Non-function pointer members | ||
150 | * should go in struct gk20a or be implemented with the boolean flag API defined | ||
151 | * in nvgpu/enabled.h | ||
152 | */ | ||
153 | struct gpu_ops { | ||
154 | struct { | ||
155 | int (*determine_L2_size_bytes)(struct gk20a *gk20a); | ||
156 | int (*init_comptags)(struct gk20a *g, struct gr_gk20a *gr); | ||
157 | int (*cbc_ctrl)(struct gk20a *g, enum gk20a_cbc_op op, | ||
158 | u32 min, u32 max); | ||
159 | void (*set_zbc_color_entry)(struct gk20a *g, | ||
160 | struct zbc_entry *color_val, | ||
161 | u32 index); | ||
162 | void (*set_zbc_depth_entry)(struct gk20a *g, | ||
163 | struct zbc_entry *depth_val, | ||
164 | u32 index); | ||
165 | void (*set_zbc_s_entry)(struct gk20a *g, | ||
166 | struct zbc_entry *s_val, | ||
167 | u32 index); | ||
168 | void (*init_cbc)(struct gk20a *g, struct gr_gk20a *gr); | ||
169 | void (*set_enabled)(struct gk20a *g, bool enabled); | ||
170 | void (*init_fs_state)(struct gk20a *g); | ||
171 | void (*isr)(struct gk20a *g); | ||
172 | u32 (*cbc_fix_config)(struct gk20a *g, int base); | ||
173 | void (*flush)(struct gk20a *g); | ||
174 | } ltc; | ||
175 | struct { | ||
176 | void (*isr_stall)(struct gk20a *g, u32 inst_id, u32 pri_base); | ||
177 | int (*isr_nonstall)(struct gk20a *g, u32 inst_id, u32 pri_base); | ||
178 | u32 (*get_num_pce)(struct gk20a *g); | ||
179 | } ce2; | ||
180 | struct { | ||
181 | u32 (*get_patch_slots)(struct gk20a *g); | ||
182 | int (*init_fs_state)(struct gk20a *g); | ||
183 | int (*init_preemption_state)(struct gk20a *g); | ||
184 | void (*access_smpc_reg)(struct gk20a *g, u32 quad, u32 offset); | ||
185 | void (*bundle_cb_defaults)(struct gk20a *g); | ||
186 | void (*cb_size_default)(struct gk20a *g); | ||
187 | int (*calc_global_ctx_buffer_size)(struct gk20a *g); | ||
188 | void (*commit_global_attrib_cb)(struct gk20a *g, | ||
189 | struct channel_ctx_gk20a *ch_ctx, | ||
190 | u64 addr, bool patch); | ||
191 | void (*commit_global_bundle_cb)(struct gk20a *g, | ||
192 | struct channel_ctx_gk20a *ch_ctx, | ||
193 | u64 addr, u64 size, bool patch); | ||
194 | int (*commit_global_cb_manager)(struct gk20a *g, | ||
195 | struct channel_gk20a *ch, | ||
196 | bool patch); | ||
197 | void (*commit_global_pagepool)(struct gk20a *g, | ||
198 | struct channel_ctx_gk20a *ch_ctx, | ||
199 | u64 addr, u32 size, bool patch); | ||
200 | void (*init_gpc_mmu)(struct gk20a *g); | ||
201 | int (*handle_sw_method)(struct gk20a *g, u32 addr, | ||
202 | u32 class_num, u32 offset, u32 data); | ||
203 | void (*set_alpha_circular_buffer_size)(struct gk20a *g, | ||
204 | u32 data); | ||
205 | void (*set_circular_buffer_size)(struct gk20a *g, u32 data); | ||
206 | void (*set_bes_crop_debug3)(struct gk20a *g, u32 data); | ||
207 | void (*enable_hww_exceptions)(struct gk20a *g); | ||
208 | bool (*is_valid_class)(struct gk20a *g, u32 class_num); | ||
209 | bool (*is_valid_gfx_class)(struct gk20a *g, u32 class_num); | ||
210 | bool (*is_valid_compute_class)(struct gk20a *g, u32 class_num); | ||
211 | void (*get_sm_dsm_perf_regs)(struct gk20a *g, | ||
212 | u32 *num_sm_dsm_perf_regs, | ||
213 | u32 **sm_dsm_perf_regs, | ||
214 | u32 *perf_register_stride); | ||
215 | void (*get_sm_dsm_perf_ctrl_regs)(struct gk20a *g, | ||
216 | u32 *num_sm_dsm_perf_regs, | ||
217 | u32 **sm_dsm_perf_regs, | ||
218 | u32 *perf_register_stride); | ||
219 | void (*get_ovr_perf_regs)(struct gk20a *g, | ||
220 | u32 *num_ovr_perf_regs, | ||
221 | u32 **ovr_perf_regsr); | ||
222 | void (*set_hww_esr_report_mask)(struct gk20a *g); | ||
223 | int (*setup_alpha_beta_tables)(struct gk20a *g, | ||
224 | struct gr_gk20a *gr); | ||
225 | int (*falcon_load_ucode)(struct gk20a *g, | ||
226 | u64 addr_base, | ||
227 | struct gk20a_ctxsw_ucode_segments *segments, | ||
228 | u32 reg_offset); | ||
229 | int (*load_ctxsw_ucode)(struct gk20a *g); | ||
230 | u32 (*get_gpc_tpc_mask)(struct gk20a *g, u32 gpc_index); | ||
231 | void (*set_gpc_tpc_mask)(struct gk20a *g, u32 gpc_index); | ||
232 | void (*free_channel_ctx)(struct channel_gk20a *c, bool is_tsg); | ||
233 | int (*alloc_obj_ctx)(struct channel_gk20a *c, | ||
234 | u32 class_num, u32 flags); | ||
235 | int (*bind_ctxsw_zcull)(struct gk20a *g, struct gr_gk20a *gr, | ||
236 | struct channel_gk20a *c, u64 zcull_va, | ||
237 | u32 mode); | ||
238 | int (*get_zcull_info)(struct gk20a *g, struct gr_gk20a *gr, | ||
239 | struct gr_zcull_info *zcull_params); | ||
240 | int (*decode_egpc_addr)(struct gk20a *g, | ||
241 | u32 addr, int *addr_type, | ||
242 | u32 *gpc_num, u32 *tpc_num, u32 *broadcast_flags); | ||
243 | void (*egpc_etpc_priv_addr_table)(struct gk20a *g, u32 addr, | ||
244 | u32 gpc, u32 broadcast_flags, u32 *priv_addr_table, | ||
245 | u32 *priv_addr_table_index); | ||
246 | bool (*is_tpc_addr)(struct gk20a *g, u32 addr); | ||
247 | bool (*is_egpc_addr)(struct gk20a *g, u32 addr); | ||
248 | bool (*is_etpc_addr)(struct gk20a *g, u32 addr); | ||
249 | void (*get_egpc_etpc_num)(struct gk20a *g, u32 addr, | ||
250 | u32 *gpc_num, u32 *tpc_num); | ||
251 | u32 (*get_tpc_num)(struct gk20a *g, u32 addr); | ||
252 | u32 (*get_egpc_base)(struct gk20a *g); | ||
253 | bool (*is_ltcs_ltss_addr)(struct gk20a *g, u32 addr); | ||
254 | bool (*is_ltcn_ltss_addr)(struct gk20a *g, u32 addr); | ||
255 | bool (*get_lts_in_ltc_shared_base)(void); | ||
256 | void (*split_lts_broadcast_addr)(struct gk20a *g, u32 addr, | ||
257 | u32 *priv_addr_table, | ||
258 | u32 *priv_addr_table_index); | ||
259 | void (*split_ltc_broadcast_addr)(struct gk20a *g, u32 addr, | ||
260 | u32 *priv_addr_table, | ||
261 | u32 *priv_addr_table_index); | ||
262 | void (*detect_sm_arch)(struct gk20a *g); | ||
263 | int (*add_zbc_color)(struct gk20a *g, struct gr_gk20a *gr, | ||
264 | struct zbc_entry *color_val, u32 index); | ||
265 | int (*add_zbc_depth)(struct gk20a *g, struct gr_gk20a *gr, | ||
266 | struct zbc_entry *depth_val, u32 index); | ||
267 | int (*add_zbc_s)(struct gk20a *g, struct gr_gk20a *gr, | ||
268 | struct zbc_entry *s_val, u32 index); | ||
269 | int (*zbc_set_table)(struct gk20a *g, struct gr_gk20a *gr, | ||
270 | struct zbc_entry *zbc_val); | ||
271 | int (*zbc_query_table)(struct gk20a *g, struct gr_gk20a *gr, | ||
272 | struct zbc_query_params *query_params); | ||
273 | int (*zbc_s_query_table)(struct gk20a *g, struct gr_gk20a *gr, | ||
274 | struct zbc_query_params *query_params); | ||
275 | int (*load_zbc_s_default_tbl)(struct gk20a *g, | ||
276 | struct gr_gk20a *gr); | ||
277 | int (*load_zbc_s_tbl)(struct gk20a *g, | ||
278 | struct gr_gk20a *gr); | ||
279 | void (*pmu_save_zbc)(struct gk20a *g, u32 entries); | ||
280 | int (*add_zbc)(struct gk20a *g, struct gr_gk20a *gr, | ||
281 | struct zbc_entry *zbc_val); | ||
282 | bool (*add_zbc_type_s)(struct gk20a *g, struct gr_gk20a *gr, | ||
283 | struct zbc_entry *zbc_val, int *ret_val); | ||
284 | u32 (*pagepool_default_size)(struct gk20a *g); | ||
285 | int (*init_ctx_state)(struct gk20a *g); | ||
286 | int (*alloc_gr_ctx)(struct gk20a *g, | ||
287 | struct gr_ctx_desc **__gr_ctx, struct vm_gk20a *vm, | ||
288 | u32 class, u32 padding); | ||
289 | void (*free_gr_ctx)(struct gk20a *g, | ||
290 | struct vm_gk20a *vm, | ||
291 | struct gr_ctx_desc *gr_ctx); | ||
292 | void (*update_ctxsw_preemption_mode)(struct gk20a *g, | ||
293 | struct channel_ctx_gk20a *ch_ctx, | ||
294 | struct nvgpu_mem *mem); | ||
295 | int (*update_smpc_ctxsw_mode)(struct gk20a *g, | ||
296 | struct channel_gk20a *c, | ||
297 | bool enable); | ||
298 | int (*update_hwpm_ctxsw_mode)(struct gk20a *g, | ||
299 | struct channel_gk20a *c, | ||
300 | bool enable); | ||
301 | int (*dump_gr_regs)(struct gk20a *g, | ||
302 | struct gk20a_debug_output *o); | ||
303 | int (*update_pc_sampling)(struct channel_gk20a *ch, | ||
304 | bool enable); | ||
305 | u32 (*get_max_fbps_count)(struct gk20a *g); | ||
306 | u32 (*get_fbp_en_mask)(struct gk20a *g); | ||
307 | u32 (*get_max_ltc_per_fbp)(struct gk20a *g); | ||
308 | u32 (*get_max_lts_per_ltc)(struct gk20a *g); | ||
309 | u32* (*get_rop_l2_en_mask)(struct gk20a *g); | ||
310 | void (*init_sm_dsm_reg_info)(void); | ||
311 | void (*init_ovr_sm_dsm_perf)(void); | ||
312 | int (*wait_empty)(struct gk20a *g, unsigned long duration_ms, | ||
313 | u32 expect_delay); | ||
314 | void (*init_cyclestats)(struct gk20a *g); | ||
315 | void (*enable_cde_in_fecs)(struct gk20a *g, | ||
316 | struct nvgpu_mem *mem); | ||
317 | int (*set_sm_debug_mode)(struct gk20a *g, struct channel_gk20a *ch, | ||
318 | u64 sms, bool enable); | ||
319 | void (*bpt_reg_info)(struct gk20a *g, | ||
320 | struct nvgpu_warpstate *w_state); | ||
321 | void (*get_access_map)(struct gk20a *g, | ||
322 | u32 **whitelist, int *num_entries); | ||
323 | int (*handle_fecs_error)(struct gk20a *g, | ||
324 | struct channel_gk20a *ch, | ||
325 | struct gr_gk20a_isr_data *isr_data); | ||
326 | int (*pre_process_sm_exception)(struct gk20a *g, | ||
327 | u32 gpc, u32 tpc, u32 sm, u32 global_esr, u32 warp_esr, | ||
328 | bool sm_debugger_attached, | ||
329 | struct channel_gk20a *fault_ch, | ||
330 | bool *early_exit, bool *ignore_debugger); | ||
331 | u32 (*get_sm_hww_warp_esr)(struct gk20a *g, | ||
332 | u32 gpc, u32 tpc, u32 sm); | ||
333 | u32 (*get_sm_hww_global_esr)(struct gk20a *g, | ||
334 | u32 gpc, u32 tpc, u32 sm); | ||
335 | u32 (*get_sm_no_lock_down_hww_global_esr_mask)(struct gk20a *g); | ||
336 | int (*lock_down_sm)(struct gk20a *g, u32 gpc, u32 tpc, u32 sm, | ||
337 | u32 global_esr_mask, bool check_errors); | ||
338 | int (*wait_for_sm_lock_down)(struct gk20a *g, u32 gpc, u32 tpc, | ||
339 | u32 sm, u32 global_esr_mask, bool check_errors); | ||
340 | void (*clear_sm_hww)(struct gk20a *g, u32 gpc, u32 tpc, u32 sm, | ||
341 | u32 global_esr); | ||
342 | void (*get_esr_sm_sel)(struct gk20a *g, u32 gpc, u32 tpc, | ||
343 | u32 *esr_sm_sel); | ||
344 | int (*handle_tpc_sm_ecc_exception)(struct gk20a *g, | ||
345 | u32 gpc, u32 tpc, | ||
346 | bool *post_event, struct channel_gk20a *fault_ch, | ||
347 | u32 *hww_global_esr); | ||
348 | int (*handle_sm_exception)(struct gk20a *g, | ||
349 | u32 gpc, u32 tpc, u32 sm, | ||
350 | bool *post_event, struct channel_gk20a *fault_ch, | ||
351 | u32 *hww_global_esr); | ||
352 | int (*handle_gcc_exception)(struct gk20a *g, u32 gpc, u32 tpc, | ||
353 | bool *post_event, struct channel_gk20a *fault_ch, | ||
354 | u32 *hww_global_esr); | ||
355 | int (*handle_tex_exception)(struct gk20a *g, u32 gpc, u32 tpc, | ||
356 | bool *post_event); | ||
357 | int (*handle_tpc_mpc_exception)(struct gk20a *g, | ||
358 | u32 gpc, u32 tpc, bool *post_event); | ||
359 | int (*handle_gpc_gpccs_exception)(struct gk20a *g, u32 gpc, | ||
360 | u32 gpc_exception); | ||
361 | int (*handle_gpc_gpcmmu_exception)(struct gk20a *g, u32 gpc, | ||
362 | u32 gpc_exception); | ||
363 | void (*enable_gpc_exceptions)(struct gk20a *g); | ||
364 | void (*enable_exceptions)(struct gk20a *g); | ||
365 | void (*create_gr_sysfs)(struct gk20a *g); | ||
366 | u32 (*get_lrf_tex_ltc_dram_override)(struct gk20a *g); | ||
367 | int (*record_sm_error_state)(struct gk20a *g, | ||
368 | u32 gpc, u32 tpc); | ||
369 | int (*update_sm_error_state)(struct gk20a *g, | ||
370 | struct channel_gk20a *ch, u32 sm_id, | ||
371 | struct nvgpu_gr_sm_error_state *sm_error_state); | ||
372 | int (*clear_sm_error_state)(struct gk20a *g, | ||
373 | struct channel_gk20a *ch, u32 sm_id); | ||
374 | int (*suspend_contexts)(struct gk20a *g, | ||
375 | struct dbg_session_gk20a *dbg_s, | ||
376 | int *ctx_resident_ch_fd); | ||
377 | int (*resume_contexts)(struct gk20a *g, | ||
378 | struct dbg_session_gk20a *dbg_s, | ||
379 | int *ctx_resident_ch_fd); | ||
380 | int (*set_preemption_mode)(struct channel_gk20a *ch, | ||
381 | u32 graphics_preempt_mode, | ||
382 | u32 compute_preempt_mode); | ||
383 | int (*get_preemption_mode_flags)(struct gk20a *g, | ||
384 | struct nvgpu_preemption_modes_rec *preemption_modes_rec); | ||
385 | int (*set_ctxsw_preemption_mode)(struct gk20a *g, | ||
386 | struct gr_ctx_desc *gr_ctx, | ||
387 | struct vm_gk20a *vm, u32 class, | ||
388 | u32 graphics_preempt_mode, | ||
389 | u32 compute_preempt_mode); | ||
390 | int (*set_boosted_ctx)(struct channel_gk20a *ch, bool boost); | ||
391 | void (*update_boosted_ctx)(struct gk20a *g, | ||
392 | struct nvgpu_mem *mem, | ||
393 | struct gr_ctx_desc *gr_ctx); | ||
394 | void (*init_sm_id_table)(struct gk20a *g); | ||
395 | int (*load_smid_config)(struct gk20a *g); | ||
396 | void (*program_sm_id_numbering)(struct gk20a *g, | ||
397 | u32 gpc, u32 tpc, u32 smid); | ||
398 | void (*program_active_tpc_counts)(struct gk20a *g, u32 gpc); | ||
399 | int (*setup_rop_mapping)(struct gk20a *g, struct gr_gk20a *gr); | ||
400 | int (*init_sw_veid_bundle)(struct gk20a *g); | ||
401 | void (*program_zcull_mapping)(struct gk20a *g, | ||
402 | u32 zcull_alloc_num, u32 *zcull_map_tiles); | ||
403 | int (*commit_global_timeslice)(struct gk20a *g, | ||
404 | struct channel_gk20a *c); | ||
405 | int (*commit_inst)(struct channel_gk20a *c, u64 gpu_va); | ||
406 | void (*write_zcull_ptr)(struct gk20a *g, | ||
407 | struct nvgpu_mem *mem, u64 gpu_va); | ||
408 | void (*write_pm_ptr)(struct gk20a *g, | ||
409 | struct nvgpu_mem *mem, u64 gpu_va); | ||
410 | void (*set_preemption_buffer_va)(struct gk20a *g, | ||
411 | struct nvgpu_mem *mem, u64 gpu_va); | ||
412 | void (*init_elcg_mode)(struct gk20a *g, u32 mode, u32 engine); | ||
413 | void (*load_tpc_mask)(struct gk20a *g); | ||
414 | int (*inval_icache)(struct gk20a *g, struct channel_gk20a *ch); | ||
415 | int (*trigger_suspend)(struct gk20a *g); | ||
416 | int (*wait_for_pause)(struct gk20a *g, struct nvgpu_warpstate *w_state); | ||
417 | int (*resume_from_pause)(struct gk20a *g); | ||
418 | int (*clear_sm_errors)(struct gk20a *g); | ||
419 | u32 (*tpc_enabled_exceptions)(struct gk20a *g); | ||
420 | int (*set_czf_bypass)(struct gk20a *g, | ||
421 | struct channel_gk20a *ch); | ||
422 | void (*init_czf_bypass)(struct gk20a *g); | ||
423 | bool (*sm_debugger_attached)(struct gk20a *g); | ||
424 | void (*suspend_single_sm)(struct gk20a *g, | ||
425 | u32 gpc, u32 tpc, u32 sm, | ||
426 | u32 global_esr_mask, bool check_errors); | ||
427 | void (*suspend_all_sms)(struct gk20a *g, | ||
428 | u32 global_esr_mask, bool check_errors); | ||
429 | void (*resume_single_sm)(struct gk20a *g, | ||
430 | u32 gpc, u32 tpc, u32 sm); | ||
431 | void (*resume_all_sms)(struct gk20a *g); | ||
432 | void (*disable_rd_coalesce)(struct gk20a *g); | ||
433 | void (*init_ctxsw_hdr_data)(struct gk20a *g, | ||
434 | struct nvgpu_mem *mem); | ||
435 | |||
436 | } gr; | ||
437 | struct { | ||
438 | void (*init_hw)(struct gk20a *g); | ||
439 | void (*init_cbc)(struct gk20a *g, struct gr_gk20a *gr); | ||
440 | void (*init_fs_state)(struct gk20a *g); | ||
441 | void (*reset)(struct gk20a *g); | ||
442 | void (*init_uncompressed_kind_map)(struct gk20a *g); | ||
443 | void (*init_kind_attr)(struct gk20a *g); | ||
444 | void (*set_mmu_page_size)(struct gk20a *g); | ||
445 | bool (*set_use_full_comp_tag_line)(struct gk20a *g); | ||
446 | unsigned int (*compression_page_size)(struct gk20a *g); | ||
447 | unsigned int (*compressible_page_size)(struct gk20a *g); | ||
448 | void (*dump_vpr_wpr_info)(struct gk20a *g); | ||
449 | int (*vpr_info_fetch)(struct gk20a *g); | ||
450 | void (*read_wpr_info)(struct gk20a *g, | ||
451 | struct wpr_carveout_info *inf); | ||
452 | bool (*is_debug_mode_enabled)(struct gk20a *g); | ||
453 | void (*set_debug_mode)(struct gk20a *g, bool enable); | ||
454 | void (*tlb_invalidate)(struct gk20a *g, struct nvgpu_mem *pdb); | ||
455 | void (*hub_isr)(struct gk20a *g); | ||
456 | int (*mem_unlock)(struct gk20a *g); | ||
457 | } fb; | ||
458 | struct { | ||
459 | void (*slcg_bus_load_gating_prod)(struct gk20a *g, bool prod); | ||
460 | void (*slcg_ce2_load_gating_prod)(struct gk20a *g, bool prod); | ||
461 | void (*slcg_chiplet_load_gating_prod)(struct gk20a *g, bool prod); | ||
462 | void (*slcg_ctxsw_firmware_load_gating_prod)(struct gk20a *g, bool prod); | ||
463 | void (*slcg_fb_load_gating_prod)(struct gk20a *g, bool prod); | ||
464 | void (*slcg_fifo_load_gating_prod)(struct gk20a *g, bool prod); | ||
465 | void (*slcg_gr_load_gating_prod)(struct gk20a *g, bool prod); | ||
466 | void (*slcg_ltc_load_gating_prod)(struct gk20a *g, bool prod); | ||
467 | void (*slcg_perf_load_gating_prod)(struct gk20a *g, bool prod); | ||
468 | void (*slcg_priring_load_gating_prod)(struct gk20a *g, bool prod); | ||
469 | void (*slcg_pmu_load_gating_prod)(struct gk20a *g, bool prod); | ||
470 | void (*slcg_therm_load_gating_prod)(struct gk20a *g, bool prod); | ||
471 | void (*slcg_xbar_load_gating_prod)(struct gk20a *g, bool prod); | ||
472 | void (*blcg_bus_load_gating_prod)(struct gk20a *g, bool prod); | ||
473 | void (*blcg_ce_load_gating_prod)(struct gk20a *g, bool prod); | ||
474 | void (*blcg_ctxsw_firmware_load_gating_prod)(struct gk20a *g, bool prod); | ||
475 | void (*blcg_fb_load_gating_prod)(struct gk20a *g, bool prod); | ||
476 | void (*blcg_fifo_load_gating_prod)(struct gk20a *g, bool prod); | ||
477 | void (*blcg_gr_load_gating_prod)(struct gk20a *g, bool prod); | ||
478 | void (*blcg_ltc_load_gating_prod)(struct gk20a *g, bool prod); | ||
479 | void (*blcg_pwr_csb_load_gating_prod)(struct gk20a *g, bool prod); | ||
480 | void (*blcg_pmu_load_gating_prod)(struct gk20a *g, bool prod); | ||
481 | void (*blcg_xbar_load_gating_prod)(struct gk20a *g, bool prod); | ||
482 | void (*pg_gr_load_gating_prod)(struct gk20a *g, bool prod); | ||
483 | } clock_gating; | ||
484 | struct { | ||
485 | int (*init_fifo_setup_hw)(struct gk20a *g); | ||
486 | void (*bind_channel)(struct channel_gk20a *ch_gk20a); | ||
487 | void (*unbind_channel)(struct channel_gk20a *ch_gk20a); | ||
488 | void (*disable_channel)(struct channel_gk20a *ch); | ||
489 | void (*enable_channel)(struct channel_gk20a *ch); | ||
490 | int (*alloc_inst)(struct gk20a *g, struct channel_gk20a *ch); | ||
491 | void (*free_inst)(struct gk20a *g, struct channel_gk20a *ch); | ||
492 | int (*setup_ramfc)(struct channel_gk20a *c, u64 gpfifo_base, | ||
493 | u32 gpfifo_entries, | ||
494 | unsigned long acquire_timeout, | ||
495 | u32 flags); | ||
496 | int (*resetup_ramfc)(struct channel_gk20a *c); | ||
497 | int (*preempt_channel)(struct gk20a *g, u32 chid); | ||
498 | int (*preempt_tsg)(struct gk20a *g, u32 tsgid); | ||
499 | int (*enable_tsg)(struct tsg_gk20a *tsg); | ||
500 | int (*disable_tsg)(struct tsg_gk20a *tsg); | ||
501 | int (*tsg_verify_channel_status)(struct channel_gk20a *ch); | ||
502 | void (*tsg_verify_status_ctx_reload)(struct channel_gk20a *ch); | ||
503 | void (*tsg_verify_status_faulted)(struct channel_gk20a *ch); | ||
504 | int (*reschedule_runlist)(struct gk20a *g, u32 runlist_id); | ||
505 | int (*update_runlist)(struct gk20a *g, u32 runlist_id, | ||
506 | u32 chid, bool add, | ||
507 | bool wait_for_finish); | ||
508 | void (*trigger_mmu_fault)(struct gk20a *g, | ||
509 | unsigned long engine_ids); | ||
510 | void (*get_mmu_fault_info)(struct gk20a *g, u32 mmu_fault_id, | ||
511 | struct mmu_fault_info *mmfault); | ||
512 | void (*apply_pb_timeout)(struct gk20a *g); | ||
513 | int (*wait_engine_idle)(struct gk20a *g); | ||
514 | u32 (*get_num_fifos)(struct gk20a *g); | ||
515 | u32 (*get_pbdma_signature)(struct gk20a *g); | ||
516 | int (*set_runlist_interleave)(struct gk20a *g, u32 id, | ||
517 | bool is_tsg, u32 runlist_id, | ||
518 | u32 new_level); | ||
519 | int (*channel_set_timeslice)(struct channel_gk20a *ch, | ||
520 | u32 timeslice); | ||
521 | int (*tsg_set_timeslice)(struct tsg_gk20a *tsg, u32 timeslice); | ||
522 | u32 (*default_timeslice_us)(struct gk20a *); | ||
523 | int (*force_reset_ch)(struct channel_gk20a *ch, | ||
524 | u32 err_code, bool verbose); | ||
525 | int (*engine_enum_from_type)(struct gk20a *g, u32 engine_type, | ||
526 | u32 *inst_id); | ||
527 | void (*device_info_data_parse)(struct gk20a *g, | ||
528 | u32 table_entry, u32 *inst_id, | ||
529 | u32 *pri_base, u32 *fault_id); | ||
530 | u32 (*device_info_fault_id)(u32 table_entry); | ||
531 | int (*tsg_bind_channel)(struct tsg_gk20a *tsg, | ||
532 | struct channel_gk20a *ch); | ||
533 | int (*tsg_unbind_channel)(struct channel_gk20a *ch); | ||
534 | int (*tsg_open)(struct tsg_gk20a *tsg); | ||
535 | u32 (*eng_runlist_base_size)(void); | ||
536 | int (*init_engine_info)(struct fifo_gk20a *f); | ||
537 | u32 (*runlist_entry_size)(void); | ||
538 | void (*get_tsg_runlist_entry)(struct tsg_gk20a *tsg, | ||
539 | u32 *runlist); | ||
540 | void (*get_ch_runlist_entry)(struct channel_gk20a *ch, | ||
541 | u32 *runlist); | ||
542 | u32 (*userd_gp_get)(struct gk20a *g, struct channel_gk20a *ch); | ||
543 | void (*userd_gp_put)(struct gk20a *g, struct channel_gk20a *ch); | ||
544 | u64 (*userd_pb_get)(struct gk20a *g, struct channel_gk20a *ch); | ||
545 | void (*free_channel_ctx_header)(struct channel_gk20a *ch); | ||
546 | bool (*is_fault_engine_subid_gpc)(struct gk20a *g, | ||
547 | u32 engine_subid); | ||
548 | void (*dump_pbdma_status)(struct gk20a *g, | ||
549 | struct gk20a_debug_output *o); | ||
550 | void (*dump_eng_status)(struct gk20a *g, | ||
551 | struct gk20a_debug_output *o); | ||
552 | void (*dump_channel_status_ramfc)(struct gk20a *g, | ||
553 | struct gk20a_debug_output *o, u32 chid, | ||
554 | struct ch_state *ch_state); | ||
555 | u32 (*intr_0_error_mask)(struct gk20a *g); | ||
556 | int (*is_preempt_pending)(struct gk20a *g, u32 id, | ||
557 | unsigned int id_type, unsigned int timeout_rc_type); | ||
558 | int (*preempt_ch_tsg)(struct gk20a *g, u32 id, | ||
559 | unsigned int id_type, unsigned int timeout_rc_type); | ||
560 | void (*init_pbdma_intr_descs)(struct fifo_gk20a *f); | ||
561 | int (*reset_enable_hw)(struct gk20a *g); | ||
562 | int (*setup_userd)(struct channel_gk20a *c); | ||
563 | u32 (*pbdma_acquire_val)(u64 timeout); | ||
564 | void (*teardown_ch_tsg)(struct gk20a *g, u32 act_eng_bitmask, | ||
565 | u32 id, unsigned int id_type, unsigned int rc_type, | ||
566 | struct mmu_fault_info *mmfault); | ||
567 | bool (*handle_sched_error)(struct gk20a *g); | ||
568 | bool (*handle_ctxsw_timeout)(struct gk20a *g, u32 fifo_intr); | ||
569 | unsigned int (*handle_pbdma_intr_0)(struct gk20a *g, | ||
570 | u32 pbdma_id, u32 pbdma_intr_0, | ||
571 | u32 *handled, u32 *error_notifier); | ||
572 | unsigned int (*handle_pbdma_intr_1)(struct gk20a *g, | ||
573 | u32 pbdma_id, u32 pbdma_intr_1, | ||
574 | u32 *handled, u32 *error_notifier); | ||
575 | void (*init_eng_method_buffers)(struct gk20a *g, | ||
576 | struct tsg_gk20a *tsg); | ||
577 | void (*deinit_eng_method_buffers)(struct gk20a *g, | ||
578 | struct tsg_gk20a *tsg); | ||
579 | u32 (*get_preempt_timeout)(struct gk20a *g); | ||
580 | #ifdef CONFIG_TEGRA_GK20A_NVHOST | ||
581 | int (*alloc_syncpt_buf)(struct channel_gk20a *c, | ||
582 | u32 syncpt_id, struct nvgpu_mem *syncpt_buf); | ||
583 | void (*free_syncpt_buf)(struct channel_gk20a *c, | ||
584 | struct nvgpu_mem *syncpt_buf); | ||
585 | void (*add_syncpt_wait_cmd)(struct gk20a *g, | ||
586 | struct priv_cmd_entry *cmd, u32 off, | ||
587 | u32 id, u32 thresh, u64 gpu_va); | ||
588 | u32 (*get_syncpt_wait_cmd_size)(void); | ||
589 | void (*add_syncpt_incr_cmd)(struct gk20a *g, | ||
590 | bool wfi_cmd, struct priv_cmd_entry *cmd, | ||
591 | u32 id, u64 gpu_va); | ||
592 | u32 (*get_syncpt_incr_cmd_size)(bool wfi_cmd); | ||
593 | #endif | ||
594 | } fifo; | ||
595 | struct pmu_v { | ||
596 | u32 (*get_pmu_cmdline_args_size)(struct nvgpu_pmu *pmu); | ||
597 | void (*set_pmu_cmdline_args_cpu_freq)(struct nvgpu_pmu *pmu, | ||
598 | u32 freq); | ||
599 | void (*set_pmu_cmdline_args_trace_size)(struct nvgpu_pmu *pmu, | ||
600 | u32 size); | ||
601 | void (*set_pmu_cmdline_args_trace_dma_base)( | ||
602 | struct nvgpu_pmu *pmu); | ||
603 | void (*set_pmu_cmdline_args_trace_dma_idx)( | ||
604 | struct nvgpu_pmu *pmu, u32 idx); | ||
605 | void * (*get_pmu_cmdline_args_ptr)(struct nvgpu_pmu *pmu); | ||
606 | u32 (*get_pmu_allocation_struct_size)(struct nvgpu_pmu *pmu); | ||
607 | void (*set_pmu_allocation_ptr)(struct nvgpu_pmu *pmu, | ||
608 | void **pmu_alloc_ptr, void *assign_ptr); | ||
609 | void (*pmu_allocation_set_dmem_size)(struct nvgpu_pmu *pmu, | ||
610 | void *pmu_alloc_ptr, u16 size); | ||
611 | u16 (*pmu_allocation_get_dmem_size)(struct nvgpu_pmu *pmu, | ||
612 | void *pmu_alloc_ptr); | ||
613 | u32 (*pmu_allocation_get_dmem_offset)(struct nvgpu_pmu *pmu, | ||
614 | void *pmu_alloc_ptr); | ||
615 | u32 * (*pmu_allocation_get_dmem_offset_addr)( | ||
616 | struct nvgpu_pmu *pmu, void *pmu_alloc_ptr); | ||
617 | void (*pmu_allocation_set_dmem_offset)(struct nvgpu_pmu *pmu, | ||
618 | void *pmu_alloc_ptr, u32 offset); | ||
619 | void * (*pmu_allocation_get_fb_addr)( | ||
620 | struct nvgpu_pmu *pmu, void *pmu_alloc_ptr); | ||
621 | u32 (*pmu_allocation_get_fb_size)( | ||
622 | struct nvgpu_pmu *pmu, void *pmu_alloc_ptr); | ||
623 | void (*get_pmu_init_msg_pmu_queue_params)( | ||
624 | struct pmu_queue *queue, u32 id, | ||
625 | void *pmu_init_msg); | ||
626 | void *(*get_pmu_msg_pmu_init_msg_ptr)( | ||
627 | struct pmu_init_msg *init); | ||
628 | u16 (*get_pmu_init_msg_pmu_sw_mg_off)( | ||
629 | union pmu_init_msg_pmu *init_msg); | ||
630 | u16 (*get_pmu_init_msg_pmu_sw_mg_size)( | ||
631 | union pmu_init_msg_pmu *init_msg); | ||
632 | u32 (*get_pmu_perfmon_cmd_start_size)(void); | ||
633 | int (*get_perfmon_cmd_start_offsetofvar)( | ||
634 | enum pmu_perfmon_cmd_start_fields field); | ||
635 | void (*perfmon_start_set_cmd_type)(struct pmu_perfmon_cmd *pc, | ||
636 | u8 value); | ||
637 | void (*perfmon_start_set_group_id)(struct pmu_perfmon_cmd *pc, | ||
638 | u8 value); | ||
639 | void (*perfmon_start_set_state_id)(struct pmu_perfmon_cmd *pc, | ||
640 | u8 value); | ||
641 | void (*perfmon_start_set_flags)(struct pmu_perfmon_cmd *pc, | ||
642 | u8 value); | ||
643 | u8 (*perfmon_start_get_flags)(struct pmu_perfmon_cmd *pc); | ||
644 | u32 (*get_pmu_perfmon_cmd_init_size)(void); | ||
645 | int (*get_perfmon_cmd_init_offsetofvar)( | ||
646 | enum pmu_perfmon_cmd_start_fields field); | ||
647 | void (*perfmon_cmd_init_set_sample_buffer)( | ||
648 | struct pmu_perfmon_cmd *pc, u16 value); | ||
649 | void (*perfmon_cmd_init_set_dec_cnt)( | ||
650 | struct pmu_perfmon_cmd *pc, u8 value); | ||
651 | void (*perfmon_cmd_init_set_base_cnt_id)( | ||
652 | struct pmu_perfmon_cmd *pc, u8 value); | ||
653 | void (*perfmon_cmd_init_set_samp_period_us)( | ||
654 | struct pmu_perfmon_cmd *pc, u32 value); | ||
655 | void (*perfmon_cmd_init_set_num_cnt)(struct pmu_perfmon_cmd *pc, | ||
656 | u8 value); | ||
657 | void (*perfmon_cmd_init_set_mov_avg)(struct pmu_perfmon_cmd *pc, | ||
658 | u8 value); | ||
659 | void *(*get_pmu_seq_in_a_ptr)( | ||
660 | struct pmu_sequence *seq); | ||
661 | void *(*get_pmu_seq_out_a_ptr)( | ||
662 | struct pmu_sequence *seq); | ||
663 | void (*set_pmu_cmdline_args_secure_mode)(struct nvgpu_pmu *pmu, | ||
664 | u32 val); | ||
665 | u32 (*get_perfmon_cntr_sz)(struct nvgpu_pmu *pmu); | ||
666 | void * (*get_perfmon_cntr_ptr)(struct nvgpu_pmu *pmu); | ||
667 | void (*set_perfmon_cntr_ut)(struct nvgpu_pmu *pmu, u16 ut); | ||
668 | void (*set_perfmon_cntr_lt)(struct nvgpu_pmu *pmu, u16 lt); | ||
669 | void (*set_perfmon_cntr_valid)(struct nvgpu_pmu *pmu, u8 val); | ||
670 | void (*set_perfmon_cntr_index)(struct nvgpu_pmu *pmu, u8 val); | ||
671 | void (*set_perfmon_cntr_group_id)(struct nvgpu_pmu *pmu, | ||
672 | u8 gid); | ||
673 | |||
674 | u8 (*pg_cmd_eng_buf_load_size)(struct pmu_pg_cmd *pg); | ||
675 | void (*pg_cmd_eng_buf_load_set_cmd_type)(struct pmu_pg_cmd *pg, | ||
676 | u8 value); | ||
677 | void (*pg_cmd_eng_buf_load_set_engine_id)(struct pmu_pg_cmd *pg, | ||
678 | u8 value); | ||
679 | void (*pg_cmd_eng_buf_load_set_buf_idx)(struct pmu_pg_cmd *pg, | ||
680 | u8 value); | ||
681 | void (*pg_cmd_eng_buf_load_set_pad)(struct pmu_pg_cmd *pg, | ||
682 | u8 value); | ||
683 | void (*pg_cmd_eng_buf_load_set_buf_size)(struct pmu_pg_cmd *pg, | ||
684 | u16 value); | ||
685 | void (*pg_cmd_eng_buf_load_set_dma_base)(struct pmu_pg_cmd *pg, | ||
686 | u32 value); | ||
687 | void (*pg_cmd_eng_buf_load_set_dma_offset)(struct pmu_pg_cmd *pg, | ||
688 | u8 value); | ||
689 | void (*pg_cmd_eng_buf_load_set_dma_idx)(struct pmu_pg_cmd *pg, | ||
690 | u8 value); | ||
691 | } pmu_ver; | ||
692 | struct { | ||
693 | int (*get_netlist_name)(struct gk20a *g, int index, char *name); | ||
694 | bool (*is_fw_defined)(void); | ||
695 | } gr_ctx; | ||
696 | #ifdef CONFIG_GK20A_CTXSW_TRACE | ||
697 | /* | ||
698 | * Currently only supported on Linux due to the extremely tight | ||
699 | * integration with Linux device driver structure (in particular | ||
700 | * mmap). | ||
701 | */ | ||
702 | struct { | ||
703 | int (*init)(struct gk20a *g); | ||
704 | int (*max_entries)(struct gk20a *, | ||
705 | struct nvgpu_ctxsw_trace_filter *filter); | ||
706 | int (*flush)(struct gk20a *g); | ||
707 | int (*poll)(struct gk20a *g); | ||
708 | int (*enable)(struct gk20a *g); | ||
709 | int (*disable)(struct gk20a *g); | ||
710 | bool (*is_enabled)(struct gk20a *g); | ||
711 | int (*reset)(struct gk20a *g); | ||
712 | int (*bind_channel)(struct gk20a *g, struct channel_gk20a *ch); | ||
713 | int (*unbind_channel)(struct gk20a *g, | ||
714 | struct channel_gk20a *ch); | ||
715 | int (*deinit)(struct gk20a *g); | ||
716 | int (*alloc_user_buffer)(struct gk20a *g, | ||
717 | void **buf, size_t *size); | ||
718 | int (*free_user_buffer)(struct gk20a *g); | ||
719 | int (*mmap_user_buffer)(struct gk20a *g, | ||
720 | struct vm_area_struct *vma); | ||
721 | int (*set_filter)(struct gk20a *g, | ||
722 | struct nvgpu_ctxsw_trace_filter *filter); | ||
723 | } fecs_trace; | ||
724 | #endif | ||
725 | struct { | ||
726 | bool (*support_sparse)(struct gk20a *g); | ||
727 | u64 (*gmmu_map)(struct vm_gk20a *vm, | ||
728 | u64 map_offset, | ||
729 | struct nvgpu_sgt *sgt, | ||
730 | u64 buffer_offset, | ||
731 | u64 size, | ||
732 | int pgsz_idx, | ||
733 | u8 kind_v, | ||
734 | u32 ctag_offset, | ||
735 | u32 flags, | ||
736 | int rw_flag, | ||
737 | bool clear_ctags, | ||
738 | bool sparse, | ||
739 | bool priv, | ||
740 | struct vm_gk20a_mapping_batch *batch, | ||
741 | enum nvgpu_aperture aperture); | ||
742 | void (*gmmu_unmap)(struct vm_gk20a *vm, | ||
743 | u64 vaddr, | ||
744 | u64 size, | ||
745 | int pgsz_idx, | ||
746 | bool va_allocated, | ||
747 | int rw_flag, | ||
748 | bool sparse, | ||
749 | struct vm_gk20a_mapping_batch *batch); | ||
750 | int (*vm_bind_channel)(struct gk20a_as_share *as_share, | ||
751 | struct channel_gk20a *ch); | ||
752 | int (*fb_flush)(struct gk20a *g); | ||
753 | void (*l2_invalidate)(struct gk20a *g); | ||
754 | void (*l2_flush)(struct gk20a *g, bool invalidate); | ||
755 | void (*cbc_clean)(struct gk20a *g); | ||
756 | void (*set_big_page_size)(struct gk20a *g, | ||
757 | struct nvgpu_mem *mem, int size); | ||
758 | u32 (*get_big_page_sizes)(void); | ||
759 | u32 (*get_default_big_page_size)(void); | ||
760 | u32 (*get_iommu_bit)(struct gk20a *g); | ||
761 | int (*init_mm_setup_hw)(struct gk20a *g); | ||
762 | bool (*is_bar1_supported)(struct gk20a *g); | ||
763 | int (*init_bar2_vm)(struct gk20a *g); | ||
764 | int (*init_bar2_mm_hw_setup)(struct gk20a *g); | ||
765 | void (*remove_bar2_vm)(struct gk20a *g); | ||
766 | const struct gk20a_mmu_level * | ||
767 | (*get_mmu_levels)(struct gk20a *g, u32 big_page_size); | ||
768 | void (*init_pdb)(struct gk20a *g, struct nvgpu_mem *inst_block, | ||
769 | struct vm_gk20a *vm); | ||
770 | u64 (*gpu_phys_addr)(struct gk20a *g, | ||
771 | struct nvgpu_gmmu_attrs *attrs, u64 phys); | ||
772 | size_t (*get_vidmem_size)(struct gk20a *g); | ||
773 | int (*alloc_inst_block)(struct gk20a *g, | ||
774 | struct nvgpu_mem *inst_block); | ||
775 | void (*init_inst_block)(struct nvgpu_mem *inst_block, | ||
776 | struct vm_gk20a *vm, u32 big_page_size); | ||
777 | bool (*mmu_fault_pending)(struct gk20a *g); | ||
778 | void (*fault_info_mem_destroy)(struct gk20a *g); | ||
779 | u32 (*get_kind_invalid)(void); | ||
780 | u32 (*get_kind_pitch)(void); | ||
781 | u32 (*get_flush_retries)(struct gk20a *g, | ||
782 | enum nvgpu_flush_op op); | ||
783 | } mm; | ||
784 | /* | ||
785 | * This function is called to allocate secure memory (memory | ||
786 | * that the CPU cannot see). The function should fill the | ||
787 | * context buffer descriptor (especially fields destroy, sgt, | ||
788 | * size). | ||
789 | */ | ||
790 | int (*secure_alloc)(struct gk20a *g, | ||
791 | struct gr_ctx_buffer_desc *desc, | ||
792 | size_t size); | ||
793 | struct { | ||
794 | u32 (*enter)(struct gk20a *g, struct nvgpu_mem *mem, | ||
795 | struct nvgpu_sgt *sgt, void *sgl, u32 w); | ||
796 | void (*exit)(struct gk20a *g, struct nvgpu_mem *mem, | ||
797 | void *sgl); | ||
798 | u32 (*data032_r)(u32 i); | ||
799 | } pramin; | ||
800 | struct { | ||
801 | int (*init_therm_setup_hw)(struct gk20a *g); | ||
802 | int (*elcg_init_idle_filters)(struct gk20a *g); | ||
803 | #ifdef CONFIG_DEBUG_FS | ||
804 | void (*therm_debugfs_init)(struct gk20a *g); | ||
805 | #endif | ||
806 | int (*get_internal_sensor_curr_temp)(struct gk20a *g, u32 *temp_f24_8); | ||
807 | void (*get_internal_sensor_limits)(s32 *max_24_8, | ||
808 | s32 *min_24_8); | ||
809 | u32 (*configure_therm_alert)(struct gk20a *g, s32 curr_warn_temp); | ||
810 | } therm; | ||
811 | struct { | ||
812 | bool (*is_pmu_supported)(struct gk20a *g); | ||
813 | int (*prepare_ucode)(struct gk20a *g); | ||
814 | int (*pmu_setup_hw_and_bootstrap)(struct gk20a *g); | ||
815 | int (*pmu_nsbootstrap)(struct nvgpu_pmu *pmu); | ||
816 | int (*pmu_setup_elpg)(struct gk20a *g); | ||
817 | u32 (*pmu_get_queue_head)(u32 i); | ||
818 | u32 (*pmu_get_queue_head_size)(void); | ||
819 | u32 (*pmu_get_queue_tail_size)(void); | ||
820 | u32 (*pmu_get_queue_tail)(u32 i); | ||
821 | int (*pmu_queue_head)(struct nvgpu_pmu *pmu, | ||
822 | struct pmu_queue *queue, u32 *head, bool set); | ||
823 | int (*pmu_queue_tail)(struct nvgpu_pmu *pmu, | ||
824 | struct pmu_queue *queue, u32 *tail, bool set); | ||
825 | void (*pmu_msgq_tail)(struct nvgpu_pmu *pmu, | ||
826 | u32 *tail, bool set); | ||
827 | u32 (*pmu_mutex_size)(void); | ||
828 | int (*pmu_mutex_acquire)(struct nvgpu_pmu *pmu, | ||
829 | u32 id, u32 *token); | ||
830 | int (*pmu_mutex_release)(struct nvgpu_pmu *pmu, | ||
831 | u32 id, u32 *token); | ||
832 | int (*init_wpr_region)(struct gk20a *g); | ||
833 | int (*load_lsfalcon_ucode)(struct gk20a *g, u32 falconidmask); | ||
834 | void (*write_dmatrfbase)(struct gk20a *g, u32 addr); | ||
835 | void (*pmu_elpg_statistics)(struct gk20a *g, u32 pg_engine_id, | ||
836 | struct pmu_pg_stats_data *pg_stat_data); | ||
837 | int (*pmu_pg_init_param)(struct gk20a *g, u32 pg_engine_id); | ||
838 | int (*pmu_pg_set_sub_feature_mask)(struct gk20a *g, | ||
839 | u32 pg_engine_id); | ||
840 | u32 (*pmu_pg_supported_engines_list)(struct gk20a *g); | ||
841 | u32 (*pmu_pg_engines_feature_list)(struct gk20a *g, | ||
842 | u32 pg_engine_id); | ||
843 | bool (*pmu_is_lpwr_feature_supported)(struct gk20a *g, | ||
844 | u32 feature_id); | ||
845 | int (*pmu_lpwr_enable_pg)(struct gk20a *g, bool pstate_lock); | ||
846 | int (*pmu_lpwr_disable_pg)(struct gk20a *g, bool pstate_lock); | ||
847 | u32 (*pmu_pg_param_post_init)(struct gk20a *g); | ||
848 | void (*dump_secure_fuses)(struct gk20a *g); | ||
849 | int (*reset_engine)(struct gk20a *g, bool do_reset); | ||
850 | bool (*is_engine_in_reset)(struct gk20a *g); | ||
851 | int (*falcon_wait_for_halt)(struct gk20a *g, | ||
852 | unsigned int timeout); | ||
853 | int (*falcon_clear_halt_interrupt_status)(struct gk20a *g, | ||
854 | unsigned int timeout); | ||
855 | int (*init_falcon_setup_hw)(struct gk20a *g, | ||
856 | void *desc, u32 bl_sz); | ||
857 | bool (*is_lazy_bootstrap)(u32 falcon_id); | ||
858 | bool (*is_priv_load)(u32 falcon_id); | ||
859 | void (*get_wpr)(struct gk20a *g, struct wpr_carveout_info *inf); | ||
860 | int (*alloc_blob_space)(struct gk20a *g, | ||
861 | size_t size, struct nvgpu_mem *mem); | ||
862 | int (*pmu_populate_loader_cfg)(struct gk20a *g, | ||
863 | void *lsfm, u32 *p_bl_gen_desc_size); | ||
864 | int (*flcn_populate_bl_dmem_desc)(struct gk20a *g, | ||
865 | void *lsfm, u32 *p_bl_gen_desc_size, u32 falconid); | ||
866 | void (*handle_ext_irq)(struct gk20a *g, u32 intr); | ||
867 | void (*set_irqmask)(struct gk20a *g); | ||
868 | } pmu; | ||
869 | struct { | ||
870 | int (*init_debugfs)(struct gk20a *g); | ||
871 | void (*disable_slowboot)(struct gk20a *g); | ||
872 | int (*init_clk_support)(struct gk20a *g); | ||
873 | int (*suspend_clk_support)(struct gk20a *g); | ||
874 | u32 (*get_crystal_clk_hz)(struct gk20a *g); | ||
875 | unsigned long (*measure_freq)(struct gk20a *g, u32 api_domain); | ||
876 | unsigned long (*get_rate)(struct gk20a *g, u32 api_domain); | ||
877 | int (*set_rate)(struct gk20a *g, u32 api_domain, unsigned long rate); | ||
878 | unsigned long (*get_fmax_at_vmin_safe)(struct clk_gk20a *clk); | ||
879 | u32 (*get_ref_clock_rate)(struct gk20a *g); | ||
880 | int (*predict_mv_at_hz_cur_tfloor)(struct clk_gk20a *clk, | ||
881 | unsigned long rate); | ||
882 | unsigned long (*get_maxrate)(struct gk20a *g, u32 api_domain); | ||
883 | int (*prepare_enable)(struct clk_gk20a *clk); | ||
884 | void (*disable_unprepare)(struct clk_gk20a *clk); | ||
885 | int (*get_voltage)(struct clk_gk20a *clk, u64 *val); | ||
886 | int (*get_gpcclk_clock_counter)(struct clk_gk20a *clk, u64 *val); | ||
887 | int (*pll_reg_write)(struct gk20a *g, u32 reg, u32 val); | ||
888 | int (*get_pll_debug_data)(struct gk20a *g, | ||
889 | struct nvgpu_clk_pll_debug_data *d); | ||
890 | int (*mclk_init)(struct gk20a *g); | ||
891 | void (*mclk_deinit)(struct gk20a *g); | ||
892 | int (*mclk_change)(struct gk20a *g, u16 val); | ||
893 | } clk; | ||
894 | struct { | ||
895 | u32 (*get_arbiter_clk_domains)(struct gk20a *g); | ||
896 | int (*get_arbiter_clk_range)(struct gk20a *g, u32 api_domain, | ||
897 | u16 *min_mhz, u16 *max_mhz); | ||
898 | int (*get_arbiter_clk_default)(struct gk20a *g, u32 api_domain, | ||
899 | u16 *default_mhz); | ||
900 | /* This function is inherently unsafe to call while | ||
901 | * arbiter is running arbiter must be blocked | ||
902 | * before calling this function */ | ||
903 | int (*get_current_pstate)(struct gk20a *g); | ||
904 | } clk_arb; | ||
905 | struct { | ||
906 | int (*handle_pmu_perf_event)(struct gk20a *g, void *pmu_msg); | ||
907 | } perf; | ||
908 | struct { | ||
909 | const struct regop_offset_range* ( | ||
910 | *get_global_whitelist_ranges)(void); | ||
911 | int (*get_global_whitelist_ranges_count)(void); | ||
912 | const struct regop_offset_range* ( | ||
913 | *get_context_whitelist_ranges)(void); | ||
914 | int (*get_context_whitelist_ranges_count)(void); | ||
915 | const u32* (*get_runcontrol_whitelist)(void); | ||
916 | int (*get_runcontrol_whitelist_count)(void); | ||
917 | const struct regop_offset_range* ( | ||
918 | *get_runcontrol_whitelist_ranges)(void); | ||
919 | int (*get_runcontrol_whitelist_ranges_count)(void); | ||
920 | const u32* (*get_qctl_whitelist)(void); | ||
921 | int (*get_qctl_whitelist_count)(void); | ||
922 | const struct regop_offset_range* ( | ||
923 | *get_qctl_whitelist_ranges)(void); | ||
924 | int (*get_qctl_whitelist_ranges_count)(void); | ||
925 | int (*apply_smpc_war)(struct dbg_session_gk20a *dbg_s); | ||
926 | } regops; | ||
927 | struct { | ||
928 | void (*intr_enable)(struct gk20a *g); | ||
929 | void (*intr_unit_config)(struct gk20a *g, | ||
930 | bool enable, bool is_stalling, u32 unit); | ||
931 | void (*isr_stall)(struct gk20a *g); | ||
932 | bool (*is_intr_hub_pending)(struct gk20a *g, u32 mc_intr); | ||
933 | u32 (*intr_stall)(struct gk20a *g); | ||
934 | void (*intr_stall_pause)(struct gk20a *g); | ||
935 | void (*intr_stall_resume)(struct gk20a *g); | ||
936 | u32 (*intr_nonstall)(struct gk20a *g); | ||
937 | void (*intr_nonstall_pause)(struct gk20a *g); | ||
938 | void (*intr_nonstall_resume)(struct gk20a *g); | ||
939 | void (*enable)(struct gk20a *g, u32 units); | ||
940 | void (*disable)(struct gk20a *g, u32 units); | ||
941 | void (*reset)(struct gk20a *g, u32 units); | ||
942 | u32 (*boot_0)(struct gk20a *g, u32 *arch, u32 *impl, u32 *rev); | ||
943 | bool (*is_intr1_pending)(struct gk20a *g, enum nvgpu_unit unit, u32 mc_intr_1); | ||
944 | } mc; | ||
945 | struct { | ||
946 | void (*show_dump)(struct gk20a *g, | ||
947 | struct gk20a_debug_output *o); | ||
948 | } debug; | ||
949 | struct { | ||
950 | int (*exec_reg_ops)(struct dbg_session_gk20a *dbg_s, | ||
951 | struct nvgpu_dbg_gpu_reg_op *ops, | ||
952 | u64 num_ops); | ||
953 | int (*dbg_set_powergate)(struct dbg_session_gk20a *dbg_s, | ||
954 | bool disable_powergate); | ||
955 | bool (*check_and_set_global_reservation)( | ||
956 | struct dbg_session_gk20a *dbg_s, | ||
957 | struct dbg_profiler_object_data *prof_obj); | ||
958 | bool (*check_and_set_context_reservation)( | ||
959 | struct dbg_session_gk20a *dbg_s, | ||
960 | struct dbg_profiler_object_data *prof_obj); | ||
961 | void (*release_profiler_reservation)( | ||
962 | struct dbg_session_gk20a *dbg_s, | ||
963 | struct dbg_profiler_object_data *prof_obj); | ||
964 | int (*perfbuffer_enable)(struct gk20a *g, u64 offset, u32 size); | ||
965 | int (*perfbuffer_disable)(struct gk20a *g); | ||
966 | } dbg_session_ops; | ||
967 | |||
968 | int (*get_litter_value)(struct gk20a *g, int value); | ||
969 | int (*chip_init_gpu_characteristics)(struct gk20a *g); | ||
970 | |||
971 | struct { | ||
972 | void (*init_hw)(struct gk20a *g); | ||
973 | void (*isr)(struct gk20a *g); | ||
974 | int (*read_ptimer)(struct gk20a *g, u64 *value); | ||
975 | int (*get_timestamps_zipper)(struct gk20a *g, | ||
976 | u32 source_id, u32 count, | ||
977 | struct nvgpu_cpu_time_correlation_sample *); | ||
978 | int (*bar1_bind)(struct gk20a *g, struct nvgpu_mem *bar1_inst); | ||
979 | } bus; | ||
980 | |||
981 | struct { | ||
982 | int (*init)(struct gk20a *g); | ||
983 | int (*preos_wait_for_halt)(struct gk20a *g); | ||
984 | void (*preos_reload_check)(struct gk20a *g); | ||
985 | } bios; | ||
986 | |||
987 | #if defined(CONFIG_GK20A_CYCLE_STATS) | ||
988 | struct { | ||
989 | int (*enable_snapshot)(struct channel_gk20a *ch, | ||
990 | struct gk20a_cs_snapshot_client *client); | ||
991 | void (*disable_snapshot)(struct gr_gk20a *gr); | ||
992 | int (*check_data_available)(struct channel_gk20a *ch, | ||
993 | u32 *pending, | ||
994 | bool *hw_overflow); | ||
995 | void (*set_handled_snapshots)(struct gk20a *g, u32 num); | ||
996 | u32 (*allocate_perfmon_ids)(struct gk20a_cs_snapshot *data, | ||
997 | u32 count); | ||
998 | u32 (*release_perfmon_ids)(struct gk20a_cs_snapshot *data, | ||
999 | u32 start, | ||
1000 | u32 count); | ||
1001 | int (*detach_snapshot)(struct channel_gk20a *ch, | ||
1002 | struct gk20a_cs_snapshot_client *client); | ||
1003 | } css; | ||
1004 | #endif | ||
1005 | struct { | ||
1006 | int (*get_speed)(struct gk20a *g, u32 *xve_link_speed); | ||
1007 | int (*set_speed)(struct gk20a *g, u32 xve_link_speed); | ||
1008 | void (*available_speeds)(struct gk20a *g, u32 *speed_mask); | ||
1009 | u32 (*xve_readl)(struct gk20a *g, u32 reg); | ||
1010 | void (*xve_writel)(struct gk20a *g, u32 reg, u32 val); | ||
1011 | void (*disable_aspm)(struct gk20a *g); | ||
1012 | void (*reset_gpu)(struct gk20a *g); | ||
1013 | #if defined(CONFIG_PCI_MSI) | ||
1014 | void (*rearm_msi)(struct gk20a *g); | ||
1015 | #endif | ||
1016 | void (*enable_shadow_rom)(struct gk20a *g); | ||
1017 | void (*disable_shadow_rom)(struct gk20a *g); | ||
1018 | u32 (*get_link_control_status)(struct gk20a *g); | ||
1019 | } xve; | ||
1020 | struct { | ||
1021 | void (*falcon_hal_sw_init)(struct nvgpu_falcon *flcn); | ||
1022 | } falcon; | ||
1023 | struct { | ||
1024 | void (*isr)(struct gk20a *g); | ||
1025 | } priv_ring; | ||
1026 | }; | ||
1027 | |||
1028 | struct nvgpu_bios_ucode { | ||
1029 | u8 *bootloader; | ||
1030 | u32 bootloader_phys_base; | ||
1031 | u32 bootloader_size; | ||
1032 | u8 *ucode; | ||
1033 | u32 phys_base; | ||
1034 | u32 size; | ||
1035 | u8 *dmem; | ||
1036 | u32 dmem_phys_base; | ||
1037 | u32 dmem_size; | ||
1038 | u32 code_entry_point; | ||
1039 | }; | ||
1040 | |||
1041 | struct nvgpu_bios { | ||
1042 | u32 vbios_version; | ||
1043 | u8 vbios_oem_version; | ||
1044 | |||
1045 | u8 *data; | ||
1046 | size_t size; | ||
1047 | |||
1048 | struct nvgpu_bios_ucode devinit; | ||
1049 | struct nvgpu_bios_ucode preos; | ||
1050 | |||
1051 | u8 *devinit_tables; | ||
1052 | u32 devinit_tables_size; | ||
1053 | u8 *bootscripts; | ||
1054 | u32 bootscripts_size; | ||
1055 | |||
1056 | u8 mem_strap_data_count; | ||
1057 | u16 mem_strap_xlat_tbl_ptr; | ||
1058 | |||
1059 | u32 condition_table_ptr; | ||
1060 | |||
1061 | u32 devinit_tables_phys_base; | ||
1062 | u32 devinit_script_phys_base; | ||
1063 | |||
1064 | struct bit_token *perf_token; | ||
1065 | struct bit_token *clock_token; | ||
1066 | struct bit_token *virt_token; | ||
1067 | u32 expansion_rom_offset; | ||
1068 | }; | ||
1069 | |||
1070 | struct nvgpu_gpu_params { | ||
1071 | /* GPU architecture ID */ | ||
1072 | u32 gpu_arch; | ||
1073 | /* GPU implementation ID */ | ||
1074 | u32 gpu_impl; | ||
1075 | /* GPU revision ID */ | ||
1076 | u32 gpu_rev; | ||
1077 | /* sm version */ | ||
1078 | u32 sm_arch_sm_version; | ||
1079 | /* sm instruction set */ | ||
1080 | u32 sm_arch_spa_version; | ||
1081 | u32 sm_arch_warp_count; | ||
1082 | }; | ||
1083 | |||
1084 | struct gk20a { | ||
1085 | void (*free)(struct gk20a *g); | ||
1086 | struct nvgpu_nvhost_dev *nvhost_dev; | ||
1087 | |||
1088 | /* | ||
1089 | * Used by <nvgpu/enabled.h>. Do not access directly! | ||
1090 | */ | ||
1091 | unsigned long *enabled_flags; | ||
1092 | |||
1093 | nvgpu_atomic_t usage_count; | ||
1094 | |||
1095 | struct nvgpu_ref refcount; | ||
1096 | |||
1097 | const char *name; | ||
1098 | |||
1099 | bool gpu_reset_done; | ||
1100 | bool power_on; | ||
1101 | bool suspended; | ||
1102 | |||
1103 | u32 log_mask; | ||
1104 | u32 log_trace; | ||
1105 | |||
1106 | struct nvgpu_gpu_params params; | ||
1107 | |||
1108 | /* | ||
1109 | * Guards access to hardware when usual gk20a_{busy,idle} are skipped | ||
1110 | * for submits and held for channel lifetime but dropped for an ongoing | ||
1111 | * gk20a_do_idle(). | ||
1112 | */ | ||
1113 | struct nvgpu_rwsem deterministic_busy; | ||
1114 | |||
1115 | struct nvgpu_falcon pmu_flcn; | ||
1116 | struct nvgpu_falcon sec2_flcn; | ||
1117 | struct nvgpu_falcon fecs_flcn; | ||
1118 | struct nvgpu_falcon gpccs_flcn; | ||
1119 | struct nvgpu_falcon nvdec_flcn; | ||
1120 | struct clk_gk20a clk; | ||
1121 | struct fifo_gk20a fifo; | ||
1122 | struct gr_gk20a gr; | ||
1123 | struct sim_gk20a sim; | ||
1124 | struct mm_gk20a mm; | ||
1125 | struct nvgpu_pmu pmu; | ||
1126 | struct acr_desc acr; | ||
1127 | struct ecc_gk20a ecc; | ||
1128 | #ifdef CONFIG_ARCH_TEGRA_18x_SOC | ||
1129 | struct clk_pmupstate clk_pmu; | ||
1130 | struct perf_pmupstate perf_pmu; | ||
1131 | struct pmgr_pmupstate pmgr_pmu; | ||
1132 | struct therm_pmupstate therm_pmu; | ||
1133 | #endif | ||
1134 | |||
1135 | #ifdef CONFIG_DEBUG_FS | ||
1136 | struct railgate_stats pstats; | ||
1137 | #endif | ||
1138 | u32 gr_idle_timeout_default; | ||
1139 | bool timeouts_enabled; | ||
1140 | unsigned int ch_wdt_timeout_ms; | ||
1141 | |||
1142 | struct nvgpu_mutex poweron_lock; | ||
1143 | struct nvgpu_mutex poweroff_lock; | ||
1144 | |||
1145 | /* Channel priorities */ | ||
1146 | u32 timeslice_low_priority_us; | ||
1147 | u32 timeslice_medium_priority_us; | ||
1148 | u32 timeslice_high_priority_us; | ||
1149 | u32 min_timeslice_us; | ||
1150 | u32 max_timeslice_us; | ||
1151 | bool runlist_interleave; | ||
1152 | |||
1153 | bool slcg_enabled; | ||
1154 | bool blcg_enabled; | ||
1155 | bool elcg_enabled; | ||
1156 | bool elpg_enabled; | ||
1157 | bool aelpg_enabled; | ||
1158 | bool can_elpg; | ||
1159 | bool mscg_enabled; | ||
1160 | bool forced_idle; | ||
1161 | bool forced_reset; | ||
1162 | bool allow_all; | ||
1163 | |||
1164 | u32 default_pri_timeout; | ||
1165 | |||
1166 | u32 ptimer_src_freq; | ||
1167 | |||
1168 | bool can_railgate; | ||
1169 | bool user_railgate_disabled; | ||
1170 | int railgate_delay; | ||
1171 | |||
1172 | unsigned int aggressive_sync_destroy_thresh; | ||
1173 | bool aggressive_sync_destroy; | ||
1174 | |||
1175 | bool has_syncpoints; | ||
1176 | /* Debugfs knob for forcing syncpt support off in runtime. */ | ||
1177 | u32 disable_syncpoints; | ||
1178 | |||
1179 | bool support_pmu; | ||
1180 | u32 bootstrap_owner; | ||
1181 | |||
1182 | bool is_virtual; | ||
1183 | |||
1184 | u32 emc3d_ratio; | ||
1185 | |||
1186 | struct nvgpu_spinlock ltc_enabled_lock; | ||
1187 | |||
1188 | struct gk20a_ctxsw_ucode_info ctxsw_ucode_info; | ||
1189 | |||
1190 | /* | ||
1191 | * A group of semaphore pools. One for each channel. | ||
1192 | */ | ||
1193 | struct nvgpu_semaphore_sea *sema_sea; | ||
1194 | |||
1195 | /* List of pending SW semaphore waits. */ | ||
1196 | struct nvgpu_list_node pending_sema_waits; | ||
1197 | struct nvgpu_raw_spinlock pending_sema_waits_lock; | ||
1198 | |||
1199 | /* held while manipulating # of debug/profiler sessions present */ | ||
1200 | /* also prevents debug sessions from attaching until released */ | ||
1201 | struct nvgpu_mutex dbg_sessions_lock; | ||
1202 | int dbg_powergating_disabled_refcount; /*refcount for pg disable */ | ||
1203 | int dbg_timeout_disabled_refcount; /*refcount for timeout disable */ | ||
1204 | |||
1205 | /* must have dbg_sessions_lock before use */ | ||
1206 | struct nvgpu_dbg_gpu_reg_op *dbg_regops_tmp_buf; | ||
1207 | u32 dbg_regops_tmp_buf_ops; | ||
1208 | |||
1209 | /* For perfbuf mapping */ | ||
1210 | struct { | ||
1211 | struct dbg_session_gk20a *owner; | ||
1212 | u64 offset; | ||
1213 | } perfbuf; | ||
1214 | |||
1215 | /* For profiler reservations */ | ||
1216 | struct nvgpu_list_node profiler_objects; | ||
1217 | bool global_profiler_reservation_held; | ||
1218 | int profiler_reservation_count; | ||
1219 | |||
1220 | void (*remove_support)(struct gk20a *); | ||
1221 | |||
1222 | u64 pg_ingating_time_us; | ||
1223 | u64 pg_ungating_time_us; | ||
1224 | u32 pg_gating_cnt; | ||
1225 | |||
1226 | struct nvgpu_spinlock mc_enable_lock; | ||
1227 | |||
1228 | struct gk20a_as as; | ||
1229 | |||
1230 | struct nvgpu_mutex client_lock; | ||
1231 | int client_refcount; /* open channels and ctrl nodes */ | ||
1232 | |||
1233 | struct gpu_ops ops; | ||
1234 | u32 mc_intr_mask_restore[4]; | ||
1235 | /*used for change of enum zbc update cmd id from ver 0 to ver1*/ | ||
1236 | u32 pmu_ver_cmd_id_zbc_table_update; | ||
1237 | u32 pmu_lsf_pmu_wpr_init_done; | ||
1238 | u32 pmu_lsf_loaded_falcon_id; | ||
1239 | |||
1240 | int irqs_enabled; | ||
1241 | int irq_stall; /* can be same as irq_nonstall in case of PCI */ | ||
1242 | int irq_nonstall; | ||
1243 | u32 max_ltc_count; | ||
1244 | u32 ltc_count; | ||
1245 | |||
1246 | struct gk20a_channel_worker { | ||
1247 | struct nvgpu_thread poll_task; | ||
1248 | nvgpu_atomic_t put; | ||
1249 | struct nvgpu_cond wq; | ||
1250 | struct nvgpu_list_node items; | ||
1251 | struct nvgpu_spinlock items_lock; | ||
1252 | struct nvgpu_mutex start_lock; | ||
1253 | } channel_worker; | ||
1254 | |||
1255 | struct gk20a_scale_profile *scale_profile; | ||
1256 | unsigned long last_freq; | ||
1257 | |||
1258 | struct gk20a_ctxsw_trace *ctxsw_trace; | ||
1259 | struct gk20a_fecs_trace *fecs_trace; | ||
1260 | |||
1261 | bool mmu_debug_ctrl; | ||
1262 | |||
1263 | u32 tpc_fs_mask_user; | ||
1264 | |||
1265 | struct nvgpu_bios bios; | ||
1266 | bool bios_is_init; | ||
1267 | |||
1268 | struct nvgpu_clk_arb *clk_arb; | ||
1269 | |||
1270 | struct gk20a_ce_app ce_app; | ||
1271 | |||
1272 | /* PCI device identifier */ | ||
1273 | u16 pci_vendor_id, pci_device_id; | ||
1274 | u16 pci_subsystem_vendor_id, pci_subsystem_device_id; | ||
1275 | u16 pci_class; | ||
1276 | u8 pci_revision; | ||
1277 | |||
1278 | /* PCIe power states. */ | ||
1279 | bool xve_l0s; | ||
1280 | bool xve_l1; | ||
1281 | |||
1282 | /* Current warning temp in sfxp24.8 */ | ||
1283 | s32 curr_warn_temp; | ||
1284 | |||
1285 | #if defined(CONFIG_PCI_MSI) | ||
1286 | /* Check if msi is enabled */ | ||
1287 | bool msi_enabled; | ||
1288 | #endif | ||
1289 | #ifdef CONFIG_NVGPU_TRACK_MEM_USAGE | ||
1290 | struct nvgpu_mem_alloc_tracker *vmallocs; | ||
1291 | struct nvgpu_mem_alloc_tracker *kmallocs; | ||
1292 | #endif | ||
1293 | |||
1294 | /* The minimum VBIOS version supported */ | ||
1295 | u32 vbios_min_version; | ||
1296 | |||
1297 | /* memory training sequence and mclk switch scripts */ | ||
1298 | u32 mem_config_idx; | ||
1299 | |||
1300 | u64 dma_memory_used; | ||
1301 | |||
1302 | #if defined(CONFIG_TEGRA_GK20A_NVHOST) && defined(CONFIG_TEGRA_19x_GPU) | ||
1303 | u64 syncpt_unit_base; | ||
1304 | size_t syncpt_unit_size; | ||
1305 | u32 syncpt_size; | ||
1306 | #endif | ||
1307 | struct nvgpu_mem syncpt_mem; | ||
1308 | |||
1309 | struct nvgpu_list_node boardobj_head; | ||
1310 | struct nvgpu_list_node boardobjgrp_head; | ||
1311 | }; | ||
1312 | |||
1313 | static inline unsigned long gk20a_get_gr_idle_timeout(struct gk20a *g) | ||
1314 | { | ||
1315 | return g->timeouts_enabled ? | ||
1316 | g->gr_idle_timeout_default : ULONG_MAX; | ||
1317 | } | ||
1318 | |||
1319 | #define MULTICHAR_TAG(a, b, c, d) (((a) << 24) | ((b) << 16) | ((c) << 8) | (d)) | ||
1320 | enum BAR0_DEBUG_OPERATION { | ||
1321 | BARO_ZERO_NOP = 0, | ||
1322 | OP_END = MULTICHAR_TAG('D', 'O', 'N', 'E'), | ||
1323 | BAR0_READ32 = MULTICHAR_TAG('0', 'R', '3', '2'), | ||
1324 | BAR0_WRITE32 = MULTICHAR_TAG('0', 'W', '3', '2'), | ||
1325 | }; | ||
1326 | |||
1327 | struct share_buffer_head { | ||
1328 | enum BAR0_DEBUG_OPERATION operation; | ||
1329 | /* size of the operation item */ | ||
1330 | u32 size; | ||
1331 | u32 completed; | ||
1332 | u32 failed; | ||
1333 | u64 context; | ||
1334 | u64 completion_callback; | ||
1335 | }; | ||
1336 | |||
1337 | struct gk20a_cyclestate_buffer_elem { | ||
1338 | struct share_buffer_head head; | ||
1339 | /* in */ | ||
1340 | u64 p_data; | ||
1341 | u64 p_done; | ||
1342 | u32 offset_bar0; | ||
1343 | u16 first_bit; | ||
1344 | u16 last_bit; | ||
1345 | /* out */ | ||
1346 | /* keep 64 bits to be consistent */ | ||
1347 | u64 data; | ||
1348 | }; | ||
1349 | |||
1350 | /* operations that will need to be executed on non stall workqueue */ | ||
1351 | enum gk20a_nonstall_ops { | ||
1352 | gk20a_nonstall_ops_wakeup_semaphore = BIT(0), /* wake up semaphore */ | ||
1353 | gk20a_nonstall_ops_post_events = BIT(1), | ||
1354 | }; | ||
1355 | |||
1356 | /* register accessors */ | ||
1357 | void __nvgpu_check_gpu_state(struct gk20a *g); | ||
1358 | void __gk20a_warn_on_no_regs(void); | ||
1359 | |||
1360 | /* convenience */ | ||
1361 | static inline struct gk20a *gk20a_from_as(struct gk20a_as *as) | ||
1362 | { | ||
1363 | return container_of(as, struct gk20a, as); | ||
1364 | } | ||
1365 | static inline struct gk20a *gk20a_from_pmu(struct nvgpu_pmu *pmu) | ||
1366 | { | ||
1367 | return container_of(pmu, struct gk20a, pmu); | ||
1368 | } | ||
1369 | |||
1370 | static inline u32 u64_hi32(u64 n) | ||
1371 | { | ||
1372 | return (u32)((n >> 32) & ~(u32)0); | ||
1373 | } | ||
1374 | |||
1375 | static inline u32 u64_lo32(u64 n) | ||
1376 | { | ||
1377 | return (u32)(n & ~(u32)0); | ||
1378 | } | ||
1379 | |||
1380 | static inline u64 hi32_lo32_to_u64(u32 hi, u32 lo) | ||
1381 | { | ||
1382 | return (((u64)hi) << 32) | (u64)lo; | ||
1383 | } | ||
1384 | |||
1385 | static inline u32 set_field(u32 val, u32 mask, u32 field) | ||
1386 | { | ||
1387 | return ((val & ~mask) | field); | ||
1388 | } | ||
1389 | |||
1390 | static inline u32 get_field(u32 reg, u32 mask) | ||
1391 | { | ||
1392 | return (reg & mask); | ||
1393 | } | ||
1394 | |||
1395 | /* invalidate channel lookup tlb */ | ||
1396 | static inline void gk20a_gr_flush_channel_tlb(struct gr_gk20a *gr) | ||
1397 | { | ||
1398 | nvgpu_spinlock_acquire(&gr->ch_tlb_lock); | ||
1399 | memset(gr->chid_tlb, 0, | ||
1400 | sizeof(struct gr_channel_map_tlb_entry) * | ||
1401 | GR_CHANNEL_MAP_TLB_SIZE); | ||
1402 | nvgpu_spinlock_release(&gr->ch_tlb_lock); | ||
1403 | } | ||
1404 | |||
1405 | /* classes that the device supports */ | ||
1406 | /* TBD: get these from an open-sourced SDK? */ | ||
1407 | enum { | ||
1408 | KEPLER_C = 0xA297, | ||
1409 | FERMI_TWOD_A = 0x902D, | ||
1410 | KEPLER_COMPUTE_A = 0xA0C0, | ||
1411 | KEPLER_INLINE_TO_MEMORY_A = 0xA040, | ||
1412 | KEPLER_DMA_COPY_A = 0xA0B5, | ||
1413 | KEPLER_CHANNEL_GPFIFO_C = 0xA26F, | ||
1414 | }; | ||
1415 | |||
1416 | #define GK20A_BAR0_IORESOURCE_MEM 0 | ||
1417 | #define GK20A_BAR1_IORESOURCE_MEM 1 | ||
1418 | #define GK20A_SIM_IORESOURCE_MEM 2 | ||
1419 | |||
1420 | void gk20a_busy_noresume(struct gk20a *g); | ||
1421 | void gk20a_idle_nosuspend(struct gk20a *g); | ||
1422 | int __must_check gk20a_busy(struct gk20a *g); | ||
1423 | void gk20a_idle(struct gk20a *g); | ||
1424 | int __gk20a_do_idle(struct gk20a *g, bool force_reset); | ||
1425 | int __gk20a_do_unidle(struct gk20a *g); | ||
1426 | |||
1427 | int gk20a_can_busy(struct gk20a *g); | ||
1428 | int gk20a_wait_for_idle(struct gk20a *g); | ||
1429 | |||
1430 | #define NVGPU_GPU_ARCHITECTURE_SHIFT 4 | ||
1431 | |||
1432 | /* constructs unique and compact GPUID from nvgpu_gpu_characteristics | ||
1433 | * arch/impl fields */ | ||
1434 | #define GK20A_GPUID(arch, impl) ((u32) ((arch) | (impl))) | ||
1435 | |||
1436 | #define GK20A_GPUID_GK20A 0x000000EA | ||
1437 | #define GK20A_GPUID_GM20B 0x0000012B | ||
1438 | #define GK20A_GPUID_GM20B_B 0x0000012E | ||
1439 | #define NVGPU_GPUID_GP10B 0x0000013B | ||
1440 | #define NVGPU_GPUID_GP104 0x00000134 | ||
1441 | #define NVGPU_GPUID_GP106 0x00000136 | ||
1442 | |||
1443 | int gk20a_init_gpu_characteristics(struct gk20a *g); | ||
1444 | |||
1445 | static inline u32 ptimer_scalingfactor10x(u32 ptimer_src_freq) | ||
1446 | { | ||
1447 | return (u32)(((u64)(PTIMER_REF_FREQ_HZ * 10)) / ptimer_src_freq); | ||
1448 | } | ||
1449 | static inline u32 scale_ptimer(u32 timeout , u32 scale10x) | ||
1450 | { | ||
1451 | if (((timeout*10) % scale10x) >= (scale10x/2)) | ||
1452 | return ((timeout * 10) / scale10x) + 1; | ||
1453 | else | ||
1454 | return (timeout * 10) / scale10x; | ||
1455 | } | ||
1456 | |||
1457 | int gk20a_prepare_poweroff(struct gk20a *g); | ||
1458 | int gk20a_finalize_poweron(struct gk20a *g); | ||
1459 | |||
1460 | void nvgpu_wait_for_deferred_interrupts(struct gk20a *g); | ||
1461 | |||
1462 | struct gk20a * __must_check gk20a_get(struct gk20a *g); | ||
1463 | void gk20a_put(struct gk20a *g); | ||
1464 | |||
1465 | static inline bool gk20a_platform_has_syncpoints(struct gk20a *g) | ||
1466 | { | ||
1467 | #ifdef CONFIG_TEGRA_GK20A_NVHOST | ||
1468 | return g->has_syncpoints && !g->disable_syncpoints; | ||
1469 | #else | ||
1470 | return false; | ||
1471 | #endif | ||
1472 | } | ||
1473 | |||
1474 | #endif /* GK20A_H */ | ||