xref: /linux/drivers/accel/ivpu/vpu_boot_api.h (revision 7f4f3b14e8079ecde096bd734af10e30d40c27b7)
1 /* SPDX-License-Identifier: MIT */
2 /*
3  * Copyright (c) 2020-2024, Intel Corporation.
4  */
5 
6 #ifndef VPU_BOOT_API_H
7 #define VPU_BOOT_API_H
8 
9 /*
10  *  The below values will be used to construct the version info this way:
11  *  fw_bin_header->api_version[VPU_BOOT_API_VER_ID] = (VPU_BOOT_API_VER_MAJOR << 16) |
12  *  VPU_BOOT_API_VER_MINOR;
13  *  VPU_BOOT_API_VER_PATCH will be ignored. KMD and compatibility is not affected if this changes
14  *  This information is collected by using vpuip_2/application/vpuFirmware/make_std_fw_image.py
15  *  If a header is missing this info we ignore the header, if a header is missing or contains
16  *  partial info a build error will be generated.
17  */
18 
19 /*
20  * Major version changes that break backward compatibility.
21  * Major version must start from 1 and can only be incremented.
22  */
23 #define VPU_BOOT_API_VER_MAJOR 3
24 
25 /*
26  * Minor version changes when API backward compatibility is preserved.
27  * Resets to 0 if Major version is incremented.
28  */
29 #define VPU_BOOT_API_VER_MINOR 26
30 
31 /*
32  * API header changed (field names, documentation, formatting) but API itself has not been changed
33  */
34 #define VPU_BOOT_API_VER_PATCH 3
35 
36 /*
37  * Index in the API version table
38  * Must be unique for each API
39  */
40 #define VPU_BOOT_API_VER_INDEX 0
41 
42 #pragma pack(push, 4)
43 
44 /*
45  * Firmware image header format
46  */
47 #define VPU_FW_HEADER_SIZE    4096
48 #define VPU_FW_HEADER_VERSION 0x1
49 #define VPU_FW_VERSION_SIZE   32
50 #define VPU_FW_API_VER_NUM    16
51 
52 struct vpu_firmware_header {
53 	u32 header_version;
54 	u32 image_format;
55 	u64 image_load_address;
56 	u32 image_size;
57 	u64 entry_point;
58 	u8 vpu_version[VPU_FW_VERSION_SIZE];
59 	u32 compression_type;
60 	u64 firmware_version_load_address;
61 	u32 firmware_version_size;
62 	u64 boot_params_load_address;
63 	u32 api_version[VPU_FW_API_VER_NUM];
64 	/* Size of memory require for firmware execution */
65 	u32 runtime_size;
66 	u32 shave_nn_fw_size;
67 	/*
68 	 * Size of primary preemption buffer, assuming a 2-job submission queue.
69 	 * NOTE: host driver is expected to adapt size accordingly to actual
70 	 * submission queue size and device capabilities.
71 	 */
72 	u32 preemption_buffer_1_size;
73 	/*
74 	 * Size of secondary preemption buffer, assuming a 2-job submission queue.
75 	 * NOTE: host driver is expected to adapt size accordingly to actual
76 	 * submission queue size and device capabilities.
77 	 */
78 	u32 preemption_buffer_2_size;
79 	/* Space reserved for future preemption-related fields. */
80 	u32 preemption_reserved[6];
81 	/* FW image read only section start address, 4KB aligned */
82 	u64 ro_section_start_address;
83 	/* FW image read only section size, 4KB aligned */
84 	u32 ro_section_size;
85 	u32 reserved;
86 };
87 
88 /*
89  * Firmware boot parameters format
90  */
91 
92 #define VPU_BOOT_PLL_COUNT     3
93 #define VPU_BOOT_PLL_OUT_COUNT 4
94 
95 /** Values for boot_type field */
96 #define VPU_BOOT_TYPE_COLDBOOT 0
97 #define VPU_BOOT_TYPE_WARMBOOT 1
98 
99 /** Value for magic filed */
100 #define VPU_BOOT_PARAMS_MAGIC 0x10000
101 
102 /** VPU scheduling mode. By default, OS scheduling is used. */
103 #define VPU_SCHEDULING_MODE_OS 0
104 #define VPU_SCHEDULING_MODE_HW 1
105 
106 enum VPU_BOOT_L2_CACHE_CFG_TYPE {
107 	VPU_BOOT_L2_CACHE_CFG_UPA = 0,
108 	VPU_BOOT_L2_CACHE_CFG_NN = 1,
109 	VPU_BOOT_L2_CACHE_CFG_NUM = 2
110 };
111 
112 /** VPU MCA ECC signalling mode. By default, no signalling is used */
113 enum VPU_BOOT_MCA_ECC_SIGNAL_TYPE {
114 	VPU_BOOT_MCA_ECC_NONE = 0,
115 	VPU_BOOT_MCA_ECC_CORR = 1,
116 	VPU_BOOT_MCA_ECC_FATAL = 2,
117 	VPU_BOOT_MCA_ECC_BOTH = 3
118 };
119 
120 /**
121  * Logging destinations.
122  *
123  * Logging output can be directed to different logging destinations. This enum
124  * defines the list of logging destinations supported by the VPU firmware (NOTE:
125  * a specific VPU FW binary may support only a subset of such output
126  * destinations, depending on the target platform and compile options).
127  */
128 enum vpu_trace_destination {
129 	VPU_TRACE_DESTINATION_PIPEPRINT = 0x1,
130 	VPU_TRACE_DESTINATION_VERBOSE_TRACING = 0x2,
131 	VPU_TRACE_DESTINATION_NORTH_PEAK = 0x4,
132 };
133 
134 /*
135  * Processor bit shifts (for loggable HW components).
136  */
137 #define VPU_TRACE_PROC_BIT_ARM	     0
138 #define VPU_TRACE_PROC_BIT_LRT	     1
139 #define VPU_TRACE_PROC_BIT_LNN	     2
140 #define VPU_TRACE_PROC_BIT_SHV_0     3
141 #define VPU_TRACE_PROC_BIT_SHV_1     4
142 #define VPU_TRACE_PROC_BIT_SHV_2     5
143 #define VPU_TRACE_PROC_BIT_SHV_3     6
144 #define VPU_TRACE_PROC_BIT_SHV_4     7
145 #define VPU_TRACE_PROC_BIT_SHV_5     8
146 #define VPU_TRACE_PROC_BIT_SHV_6     9
147 #define VPU_TRACE_PROC_BIT_SHV_7     10
148 #define VPU_TRACE_PROC_BIT_SHV_8     11
149 #define VPU_TRACE_PROC_BIT_SHV_9     12
150 #define VPU_TRACE_PROC_BIT_SHV_10    13
151 #define VPU_TRACE_PROC_BIT_SHV_11    14
152 #define VPU_TRACE_PROC_BIT_SHV_12    15
153 #define VPU_TRACE_PROC_BIT_SHV_13    16
154 #define VPU_TRACE_PROC_BIT_SHV_14    17
155 #define VPU_TRACE_PROC_BIT_SHV_15    18
156 #define VPU_TRACE_PROC_BIT_ACT_SHV_0 19
157 #define VPU_TRACE_PROC_BIT_ACT_SHV_1 20
158 #define VPU_TRACE_PROC_BIT_ACT_SHV_2 21
159 #define VPU_TRACE_PROC_BIT_ACT_SHV_3 22
160 #define VPU_TRACE_PROC_NO_OF_HW_DEVS 23
161 
162 /* VPU 30xx HW component IDs are sequential, so define first and last IDs. */
163 #define VPU_TRACE_PROC_BIT_30XX_FIRST VPU_TRACE_PROC_BIT_LRT
164 #define VPU_TRACE_PROC_BIT_30XX_LAST  VPU_TRACE_PROC_BIT_SHV_15
165 
166 struct vpu_boot_l2_cache_config {
167 	u8 use;
168 	u8 cfg;
169 };
170 
171 struct vpu_warm_boot_section {
172 	u32 src;
173 	u32 dst;
174 	u32 size;
175 	u32 core_id;
176 	u32 is_clear_op;
177 };
178 
179 /*
180  * When HW scheduling mode is enabled, a present period is defined.
181  * It will be used by VPU to swap between normal and focus priorities
182  * to prevent starving of normal priority band (when implemented).
183  * Host must provide a valid value at boot time in
184  * `vpu_focus_present_timer_ms`. If the value provided by the host is not within the
185  * defined range a default value will be used. Here we define the min. and max.
186  * allowed values and the and default value of the present period. Units are milliseconds.
187  */
188 #define VPU_PRESENT_CALL_PERIOD_MS_DEFAULT 50
189 #define VPU_PRESENT_CALL_PERIOD_MS_MIN	   16
190 #define VPU_PRESENT_CALL_PERIOD_MS_MAX	   10000
191 
192 /**
193  * Macros to enable various power profiles within the NPU.
194  * To be defined as part of 32 bit mask.
195  */
196 #define POWER_PROFILE_SURVIVABILITY 0x1
197 
198 /**
199  * Enum for dvfs_mode boot param.
200  */
201 enum vpu_governor {
202 	VPU_GOV_DEFAULT = 0, /* Default Governor for the system */
203 	VPU_GOV_MAX_PERFORMANCE = 1, /* Maximum performance governor */
204 	VPU_GOV_ON_DEMAND = 2, /* On Demand frequency control governor */
205 	VPU_GOV_POWER_SAVE = 3, /* Power save governor */
206 	VPU_GOV_ON_DEMAND_PRIORITY_AWARE = 4 /* On Demand priority based governor */
207 };
208 
209 struct vpu_boot_params {
210 	u32 magic;
211 	u32 vpu_id;
212 	u32 vpu_count;
213 	u32 pad0[5];
214 	/* Clock frequencies: 0x20 - 0xFF */
215 	u32 frequency;
216 	u32 pll[VPU_BOOT_PLL_COUNT][VPU_BOOT_PLL_OUT_COUNT];
217 	u32 perf_clk_frequency;
218 	u32 pad1[42];
219 	/* Memory regions: 0x100 - 0x1FF */
220 	u64 ipc_header_area_start;
221 	u32 ipc_header_area_size;
222 	u64 shared_region_base;
223 	u32 shared_region_size;
224 	u64 ipc_payload_area_start;
225 	u32 ipc_payload_area_size;
226 	u64 global_aliased_pio_base;
227 	u32 global_aliased_pio_size;
228 	u32 autoconfig;
229 	struct vpu_boot_l2_cache_config cache_defaults[VPU_BOOT_L2_CACHE_CFG_NUM];
230 	u64 global_memory_allocator_base;
231 	u32 global_memory_allocator_size;
232 	/**
233 	 * ShaveNN FW section VPU base address
234 	 * On VPU2.7 HW this address must be within 2GB range starting from L2C_PAGE_TABLE base
235 	 */
236 	u64 shave_nn_fw_base;
237 	u64 save_restore_ret_address; /* stores the address of FW's restore entry point */
238 	u32 pad2[43];
239 	/* IRQ re-direct numbers: 0x200 - 0x2FF */
240 	s32 watchdog_irq_mss;
241 	s32 watchdog_irq_nce;
242 	/* ARM -> VPU doorbell interrupt. ARM is notifying VPU of async command or compute job. */
243 	u32 host_to_vpu_irq;
244 	/* VPU -> ARM job done interrupt. VPU is notifying ARM of compute job completion. */
245 	u32 job_done_irq;
246 	/* VPU -> ARM IRQ line to use to request MMU update. */
247 	u32 mmu_update_request_irq;
248 	/* ARM -> VPU IRQ line to use to notify of MMU update completion. */
249 	u32 mmu_update_done_irq;
250 	/* ARM -> VPU IRQ line to use to request power level change. */
251 	u32 set_power_level_irq;
252 	/* VPU -> ARM IRQ line to use to notify of power level change completion. */
253 	u32 set_power_level_done_irq;
254 	/* VPU -> ARM IRQ line to use to notify of VPU idle state change */
255 	u32 set_vpu_idle_update_irq;
256 	/* VPU -> ARM IRQ line to use to request counter reset. */
257 	u32 metric_query_event_irq;
258 	/* ARM -> VPU IRQ line to use to notify of counter reset completion. */
259 	u32 metric_query_event_done_irq;
260 	/* VPU -> ARM IRQ line to use to notify of preemption completion. */
261 	u32 preemption_done_irq;
262 	/* Padding. */
263 	u32 pad3[52];
264 	/* Silicon information: 0x300 - 0x3FF */
265 	u32 host_version_id;
266 	u32 si_stepping;
267 	u64 device_id;
268 	u64 feature_exclusion;
269 	u64 sku;
270 	/** PLL ratio for minimum clock frequency */
271 	u32 min_freq_pll_ratio;
272 	/** PLL ratio for maximum clock frequency */
273 	u32 max_freq_pll_ratio;
274 	/**
275 	 * Initial log level threshold (messages with log level severity less than
276 	 * the threshold will not be logged); applies to every enabled logging
277 	 * destination and loggable HW component. See 'mvLog_t' enum for acceptable
278 	 * values.
279 	 * TODO: EISW-33556: Move log level definition (mvLog_t) to this file.
280 	 */
281 	u32 default_trace_level;
282 	u32 boot_type;
283 	u64 punit_telemetry_sram_base;
284 	u64 punit_telemetry_sram_size;
285 	u32 vpu_telemetry_enable;
286 	u64 crit_tracing_buff_addr;
287 	u32 crit_tracing_buff_size;
288 	u64 verbose_tracing_buff_addr;
289 	u32 verbose_tracing_buff_size;
290 	u64 verbose_tracing_sw_component_mask; /* TO BE REMOVED */
291 	/**
292 	 * Mask of destinations to which logging messages are delivered; bitwise OR
293 	 * of values defined in vpu_trace_destination enum.
294 	 */
295 	u32 trace_destination_mask;
296 	/**
297 	 * Mask of hardware components for which logging is enabled; bitwise OR of
298 	 * bits defined by the VPU_TRACE_PROC_BIT_* macros.
299 	 */
300 	u64 trace_hw_component_mask;
301 	/** Mask of trace message formats supported by the driver */
302 	u64 tracing_buff_message_format_mask;
303 	u64 trace_reserved_1[2];
304 	/**
305 	 * Period at which the VPU reads the temp sensor values into MMIO, on
306 	 * platforms where that is necessary (in ms). 0 to disable reads.
307 	 */
308 	u32 temp_sensor_period_ms;
309 	/** PLL ratio for efficient clock frequency */
310 	u32 pn_freq_pll_ratio;
311 	/**
312 	 * DVFS Mode:
313 	 * 0 - Default, DVFS mode selected by the firmware
314 	 * 1 - Max Performance
315 	 * 2 - On Demand
316 	 * 3 - Power Save
317 	 * 4 - On Demand Priority Aware
318 	 */
319 	u32 dvfs_mode;
320 	/**
321 	 * Depending on DVFS Mode:
322 	 * On-demand: Default if 0.
323 	 *    Bit 0-7   - uint8_t: Highest residency percent
324 	 *    Bit 8-15  - uint8_t: High residency percent
325 	 *    Bit 16-23 - uint8_t: Low residency percent
326 	 *    Bit 24-31 - uint8_t: Lowest residency percent
327 	 *    Bit 32-35 - unsigned 4b: PLL Ratio increase amount on highest residency
328 	 *    Bit 36-39 - unsigned 4b: PLL Ratio increase amount on high residency
329 	 *    Bit 40-43 - unsigned 4b: PLL Ratio decrease amount on low residency
330 	 *    Bit 44-47 - unsigned 4b: PLL Ratio decrease amount on lowest frequency
331 	 *    Bit 48-55 - uint8_t: Period (ms) for residency decisions
332 	 *    Bit 56-63 - uint8_t: Averaging windows (as multiples of period. Max: 30 decimal)
333 	 * Power Save/Max Performance: Unused
334 	 */
335 	u64 dvfs_param;
336 	/**
337 	 * D0i3 delayed entry
338 	 * Bit0: Disable CPU state save on D0i2 entry flow.
339 	 *       0: Every D0i2 entry saves state. Save state IPC message ignored.
340 	 *       1: IPC message required to save state on D0i3 entry flow.
341 	 */
342 	u32 d0i3_delayed_entry;
343 	/* Time spent by VPU in D0i3 state */
344 	u64 d0i3_residency_time_us;
345 	/* Value of VPU perf counter at the time of entering D0i3 state . */
346 	u64 d0i3_entry_vpu_ts;
347 	/*
348 	 * The system time of the host operating system in microseconds.
349 	 * E.g the number of microseconds since 1st of January 1970, or whatever
350 	 * date the host operating system uses to maintain system time.
351 	 * This value will be used to track system time on the VPU.
352 	 * The KMD is required to update this value on every VPU reset.
353 	 */
354 	u64 system_time_us;
355 	u32 pad4[2];
356 	/*
357 	 * The delta between device monotonic time and the current value of the
358 	 * HW timestamp register, in ticks. Written by the firmware during boot.
359 	 * Can be used by the KMD to calculate device time.
360 	 */
361 	u64 device_time_delta_ticks;
362 	u32 pad7[14];
363 	/* Warm boot information: 0x400 - 0x43F */
364 	u32 warm_boot_sections_count;
365 	u32 warm_boot_start_address_reference;
366 	u32 warm_boot_section_info_address_offset;
367 	u32 pad5[13];
368 	/* Power States transitions timestamps: 0x440 - 0x46F*/
369 	struct {
370 		/* VPU_IDLE -> VPU_ACTIVE transition initiated timestamp */
371 		u64 vpu_active_state_requested;
372 		/* VPU_IDLE -> VPU_ACTIVE transition completed timestamp */
373 		u64 vpu_active_state_achieved;
374 		/* VPU_ACTIVE -> VPU_IDLE transition initiated timestamp */
375 		u64 vpu_idle_state_requested;
376 		/* VPU_ACTIVE -> VPU_IDLE transition completed timestamp */
377 		u64 vpu_idle_state_achieved;
378 		/* VPU_IDLE -> VPU_STANDBY transition initiated timestamp */
379 		u64 vpu_standby_state_requested;
380 		/* VPU_IDLE -> VPU_STANDBY transition completed timestamp */
381 		u64 vpu_standby_state_achieved;
382 	} power_states_timestamps;
383 	/* VPU scheduling mode. Values defined by VPU_SCHEDULING_MODE_* macros. */
384 	u32 vpu_scheduling_mode;
385 	/* Present call period in milliseconds. */
386 	u32 vpu_focus_present_timer_ms;
387 	/* VPU ECC Signaling */
388 	u32 vpu_uses_ecc_mca_signal;
389 	/* Values defined by POWER_PROFILE* macros */
390 	u32 power_profile;
391 	/* Microsecond value for DCT active cycle */
392 	u32 dct_active_us;
393 	/* Microsecond value for DCT inactive cycle */
394 	u32 dct_inactive_us;
395 	/* Unused/reserved: 0x488 - 0xFFF */
396 	u32 pad6[734];
397 };
398 
399 /* Magic numbers set between host and vpu to detect corruption of tracing init */
400 #define VPU_TRACING_BUFFER_CANARY (0xCAFECAFE)
401 
402 /* Tracing buffer message format definitions */
403 #define VPU_TRACING_FORMAT_STRING 0
404 #define VPU_TRACING_FORMAT_MIPI	  2
405 /*
406  * Header of the tracing buffer.
407  * The below defined header will be stored at the beginning of
408  * each allocated tracing buffer, followed by a series of 256b
409  * of ASCII trace message entries.
410  */
411 struct vpu_tracing_buffer_header {
412 	/**
413 	 * Magic number set by host to detect corruption
414 	 * @see VPU_TRACING_BUFFER_CANARY
415 	 */
416 	u32 host_canary_start;
417 	/* offset from start of buffer for trace entries */
418 	u32 read_index;
419 	/* keeps track of wrapping on the reader side */
420 	u32 read_wrap_count;
421 	u32 pad_to_cache_line_size_0[13];
422 	/* End of first cache line */
423 
424 	/**
425 	 * Magic number set by host to detect corruption
426 	 * @see VPU_TRACING_BUFFER_CANARY
427 	 */
428 	u32 vpu_canary_start;
429 	/* offset from start of buffer from write start */
430 	u32 write_index;
431 	/* counter for buffer wrapping */
432 	u32 wrap_count;
433 	/* legacy field - do not use */
434 	u32 reserved_0;
435 	/**
436 	 * Size of the log buffer include this header (@header_size) and space
437 	 * reserved for all messages. If @alignment` is greater that 0 the @Size
438 	 * must be multiple of @Alignment.
439 	 */
440 	u32 size;
441 	/* Header version */
442 	u16 header_version;
443 	/* Header size */
444 	u16 header_size;
445 	/*
446 	 * Format of the messages in the trace buffer
447 	 * 0 - null terminated string
448 	 * 1 - size + null terminated string
449 	 * 2 - MIPI-SysT encoding
450 	 */
451 	u32 format;
452 	/*
453 	 * Message alignment
454 	 * 0 - messages are place 1 after another
455 	 * n - every message starts and multiple on offset
456 	 */
457 	u32 alignment; /* 64, 128, 256 */
458 	/* Name of the logging entity, i.e "LRT", "LNN", "SHV0", etc */
459 	char name[16];
460 	u32 pad_to_cache_line_size_1[4];
461 	/* End of second cache line */
462 };
463 
464 #pragma pack(pop)
465 
466 #endif
467