1 /*
2  * Copyright 2017 Advanced Micro Devices, Inc.
3  *
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:
10  *
11  * The above copyright notice and this permission notice shall be included in
12  * all copies or substantial portions of the Software.
13  *
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.
21  *
22  */
23 
24 #if !defined(_AMDGPU_TRACE_H) || defined(TRACE_HEADER_MULTI_READ)
25 #define _AMDGPU_TRACE_H_
26 
27 #include <linux/stringify.h>
28 #include <linux/types.h>
29 #include <linux/tracepoint.h>
30 
31 #include <drm/drmP.h>
32 
33 #undef TRACE_SYSTEM
34 #define TRACE_SYSTEM amdgpu
35 #define TRACE_INCLUDE_FILE amdgpu_trace
36 
37 #define AMDGPU_JOB_GET_TIMELINE_NAME(job) \
38 	 job->base.s_fence->finished.ops->get_timeline_name(&job->base.s_fence->finished)
39 
40 TRACE_EVENT(amdgpu_mm_rreg,
41 	    TP_PROTO(unsigned did, uint32_t reg, uint32_t value),
42 	    TP_ARGS(did, reg, value),
43 	    TP_STRUCT__entry(
44 				__field(unsigned, did)
45 				__field(uint32_t, reg)
46 				__field(uint32_t, value)
47 			    ),
48 	    TP_fast_assign(
49 			   __entry->did = did;
50 			   __entry->reg = reg;
51 			   __entry->value = value;
52 			   ),
53 	    TP_printk("0x%04lx, 0x%08lx, 0x%08lx",
54 		      (unsigned long)__entry->did,
55 		      (unsigned long)__entry->reg,
56 		      (unsigned long)__entry->value)
57 );
58 
59 TRACE_EVENT(amdgpu_mm_wreg,
60 	    TP_PROTO(unsigned did, uint32_t reg, uint32_t value),
61 	    TP_ARGS(did, reg, value),
62 	    TP_STRUCT__entry(
63 				__field(unsigned, did)
64 				__field(uint32_t, reg)
65 				__field(uint32_t, value)
66 			    ),
67 	    TP_fast_assign(
68 			   __entry->did = did;
69 			   __entry->reg = reg;
70 			   __entry->value = value;
71 			   ),
72 	    TP_printk("0x%04lx, 0x%08lx, 0x%08lx",
73 		      (unsigned long)__entry->did,
74 		      (unsigned long)__entry->reg,
75 		      (unsigned long)__entry->value)
76 );
77 
78 TRACE_EVENT(amdgpu_iv,
79 	    TP_PROTO(struct amdgpu_iv_entry *iv),
80 	    TP_ARGS(iv),
81 	    TP_STRUCT__entry(
82 			     __field(unsigned, client_id)
83 			     __field(unsigned, src_id)
84 			     __field(unsigned, ring_id)
85 			     __field(unsigned, vm_id)
86 			     __field(unsigned, vm_id_src)
87 			     __field(uint64_t, timestamp)
88 			     __field(unsigned, timestamp_src)
89 			     __field(unsigned, pas_id)
90 			     __array(unsigned, src_data, 4)
91 			    ),
92 	    TP_fast_assign(
93 			   __entry->client_id = iv->client_id;
94 			   __entry->src_id = iv->src_id;
95 			   __entry->ring_id = iv->ring_id;
96 			   __entry->vm_id = iv->vm_id;
97 			   __entry->vm_id_src = iv->vm_id_src;
98 			   __entry->timestamp = iv->timestamp;
99 			   __entry->timestamp_src = iv->timestamp_src;
100 			   __entry->pas_id = iv->pas_id;
101 			   __entry->src_data[0] = iv->src_data[0];
102 			   __entry->src_data[1] = iv->src_data[1];
103 			   __entry->src_data[2] = iv->src_data[2];
104 			   __entry->src_data[3] = iv->src_data[3];
105 			   ),
106 	    TP_printk("client_id:%u src_id:%u ring:%u vm_id:%u timestamp: %llu pas_id:%u src_data: %08x %08x %08x %08x\n",
107 		      __entry->client_id, __entry->src_id,
108 		      __entry->ring_id, __entry->vm_id,
109 		      __entry->timestamp, __entry->pas_id,
110 		      __entry->src_data[0], __entry->src_data[1],
111 		      __entry->src_data[2], __entry->src_data[3])
112 );
113 
114 
115 TRACE_EVENT(amdgpu_bo_create,
116 	    TP_PROTO(struct amdgpu_bo *bo),
117 	    TP_ARGS(bo),
118 	    TP_STRUCT__entry(
119 			     __field(struct amdgpu_bo *, bo)
120 			     __field(u32, pages)
121 			     __field(u32, type)
122 			     __field(u32, prefer)
123 			     __field(u32, allow)
124 			     __field(u32, visible)
125 			     ),
126 
127 	    TP_fast_assign(
128 			   __entry->bo = bo;
129 			   __entry->pages = bo->tbo.num_pages;
130 			   __entry->type = bo->tbo.mem.mem_type;
131 			   __entry->prefer = bo->preferred_domains;
132 			   __entry->allow = bo->allowed_domains;
133 			   __entry->visible = bo->flags;
134 			   ),
135 
136 	    TP_printk("bo=%p, pages=%u, type=%d, preferred=%d, allowed=%d, visible=%d",
137 		       __entry->bo, __entry->pages, __entry->type,
138 		       __entry->prefer, __entry->allow, __entry->visible)
139 );
140 
141 TRACE_EVENT(amdgpu_cs,
142 	    TP_PROTO(struct amdgpu_cs_parser *p, int i),
143 	    TP_ARGS(p, i),
144 	    TP_STRUCT__entry(
145 			     __field(struct amdgpu_bo_list *, bo_list)
146 			     __field(u32, ring)
147 			     __field(u32, dw)
148 			     __field(u32, fences)
149 			     ),
150 
151 	    TP_fast_assign(
152 			   __entry->bo_list = p->bo_list;
153 			   __entry->ring = p->job->ring->idx;
154 			   __entry->dw = p->job->ibs[i].length_dw;
155 			   __entry->fences = amdgpu_fence_count_emitted(
156 				p->job->ring);
157 			   ),
158 	    TP_printk("bo_list=%p, ring=%u, dw=%u, fences=%u",
159 		      __entry->bo_list, __entry->ring, __entry->dw,
160 		      __entry->fences)
161 );
162 
163 TRACE_EVENT(amdgpu_cs_ioctl,
164 	    TP_PROTO(struct amdgpu_job *job),
165 	    TP_ARGS(job),
166 	    TP_STRUCT__entry(
167 			     __field(uint64_t, sched_job_id)
168 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
169 			     __field(unsigned int, context)
170 			     __field(unsigned int, seqno)
171 			     __field(struct dma_fence *, fence)
172 			     __field(char *, ring_name)
173 			     __field(u32, num_ibs)
174 			     ),
175 
176 	    TP_fast_assign(
177 			   __entry->sched_job_id = job->base.id;
178 			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
179 			   __entry->context = job->base.s_fence->finished.context;
180 			   __entry->seqno = job->base.s_fence->finished.seqno;
181 			   __entry->ring_name = job->ring->name;
182 			   __entry->num_ibs = job->num_ibs;
183 			   ),
184 	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
185 		      __entry->sched_job_id, __get_str(timeline), __entry->context,
186 		      __entry->seqno, __entry->ring_name, __entry->num_ibs)
187 );
188 
189 TRACE_EVENT(amdgpu_sched_run_job,
190 	    TP_PROTO(struct amdgpu_job *job),
191 	    TP_ARGS(job),
192 	    TP_STRUCT__entry(
193 			     __field(uint64_t, sched_job_id)
194 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
195 			     __field(unsigned int, context)
196 			     __field(unsigned int, seqno)
197 			     __field(char *, ring_name)
198 			     __field(u32, num_ibs)
199 			     ),
200 
201 	    TP_fast_assign(
202 			   __entry->sched_job_id = job->base.id;
203 			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
204 			   __entry->context = job->base.s_fence->finished.context;
205 			   __entry->seqno = job->base.s_fence->finished.seqno;
206 			   __entry->ring_name = job->ring->name;
207 			   __entry->num_ibs = job->num_ibs;
208 			   ),
209 	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
210 		      __entry->sched_job_id, __get_str(timeline), __entry->context,
211 		      __entry->seqno, __entry->ring_name, __entry->num_ibs)
212 );
213 
214 
215 TRACE_EVENT(amdgpu_vm_grab_id,
216 	    TP_PROTO(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
217 		     struct amdgpu_job *job),
218 	    TP_ARGS(vm, ring, job),
219 	    TP_STRUCT__entry(
220 			     __field(struct amdgpu_vm *, vm)
221 			     __field(u32, ring)
222 			     __field(u32, vm_id)
223 			     __field(u32, vm_hub)
224 			     __field(u64, pd_addr)
225 			     __field(u32, needs_flush)
226 			     ),
227 
228 	    TP_fast_assign(
229 			   __entry->vm = vm;
230 			   __entry->ring = ring->idx;
231 			   __entry->vm_id = job->vm_id;
232 			   __entry->vm_hub = ring->funcs->vmhub,
233 			   __entry->pd_addr = job->vm_pd_addr;
234 			   __entry->needs_flush = job->vm_needs_flush;
235 			   ),
236 	    TP_printk("vm=%p, ring=%u, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
237 		      __entry->vm, __entry->ring, __entry->vm_id,
238 		      __entry->vm_hub, __entry->pd_addr, __entry->needs_flush)
239 );
240 
241 TRACE_EVENT(amdgpu_vm_bo_map,
242 	    TP_PROTO(struct amdgpu_bo_va *bo_va,
243 		     struct amdgpu_bo_va_mapping *mapping),
244 	    TP_ARGS(bo_va, mapping),
245 	    TP_STRUCT__entry(
246 			     __field(struct amdgpu_bo *, bo)
247 			     __field(long, start)
248 			     __field(long, last)
249 			     __field(u64, offset)
250 			     __field(u64, flags)
251 			     ),
252 
253 	    TP_fast_assign(
254 			   __entry->bo = bo_va ? bo_va->base.bo : NULL;
255 			   __entry->start = mapping->start;
256 			   __entry->last = mapping->last;
257 			   __entry->offset = mapping->offset;
258 			   __entry->flags = mapping->flags;
259 			   ),
260 	    TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%llx",
261 		      __entry->bo, __entry->start, __entry->last,
262 		      __entry->offset, __entry->flags)
263 );
264 
265 TRACE_EVENT(amdgpu_vm_bo_unmap,
266 	    TP_PROTO(struct amdgpu_bo_va *bo_va,
267 		     struct amdgpu_bo_va_mapping *mapping),
268 	    TP_ARGS(bo_va, mapping),
269 	    TP_STRUCT__entry(
270 			     __field(struct amdgpu_bo *, bo)
271 			     __field(long, start)
272 			     __field(long, last)
273 			     __field(u64, offset)
274 			     __field(u64, flags)
275 			     ),
276 
277 	    TP_fast_assign(
278 			   __entry->bo = bo_va->base.bo;
279 			   __entry->start = mapping->start;
280 			   __entry->last = mapping->last;
281 			   __entry->offset = mapping->offset;
282 			   __entry->flags = mapping->flags;
283 			   ),
284 	    TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%llx",
285 		      __entry->bo, __entry->start, __entry->last,
286 		      __entry->offset, __entry->flags)
287 );
288 
289 DECLARE_EVENT_CLASS(amdgpu_vm_mapping,
290 	    TP_PROTO(struct amdgpu_bo_va_mapping *mapping),
291 	    TP_ARGS(mapping),
292 	    TP_STRUCT__entry(
293 			     __field(u64, soffset)
294 			     __field(u64, eoffset)
295 			     __field(u64, flags)
296 			     ),
297 
298 	    TP_fast_assign(
299 			   __entry->soffset = mapping->start;
300 			   __entry->eoffset = mapping->last + 1;
301 			   __entry->flags = mapping->flags;
302 			   ),
303 	    TP_printk("soffs=%010llx, eoffs=%010llx, flags=%llx",
304 		      __entry->soffset, __entry->eoffset, __entry->flags)
305 );
306 
307 DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_update,
308 	    TP_PROTO(struct amdgpu_bo_va_mapping *mapping),
309 	    TP_ARGS(mapping)
310 );
311 
312 DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_mapping,
313 	    TP_PROTO(struct amdgpu_bo_va_mapping *mapping),
314 	    TP_ARGS(mapping)
315 );
316 
317 TRACE_EVENT(amdgpu_vm_set_ptes,
318 	    TP_PROTO(uint64_t pe, uint64_t addr, unsigned count,
319 		     uint32_t incr, uint64_t flags),
320 	    TP_ARGS(pe, addr, count, incr, flags),
321 	    TP_STRUCT__entry(
322 			     __field(u64, pe)
323 			     __field(u64, addr)
324 			     __field(u32, count)
325 			     __field(u32, incr)
326 			     __field(u64, flags)
327 			     ),
328 
329 	    TP_fast_assign(
330 			   __entry->pe = pe;
331 			   __entry->addr = addr;
332 			   __entry->count = count;
333 			   __entry->incr = incr;
334 			   __entry->flags = flags;
335 			   ),
336 	    TP_printk("pe=%010Lx, addr=%010Lx, incr=%u, flags=%llx, count=%u",
337 		      __entry->pe, __entry->addr, __entry->incr,
338 		      __entry->flags, __entry->count)
339 );
340 
341 TRACE_EVENT(amdgpu_vm_copy_ptes,
342 	    TP_PROTO(uint64_t pe, uint64_t src, unsigned count),
343 	    TP_ARGS(pe, src, count),
344 	    TP_STRUCT__entry(
345 			     __field(u64, pe)
346 			     __field(u64, src)
347 			     __field(u32, count)
348 			     ),
349 
350 	    TP_fast_assign(
351 			   __entry->pe = pe;
352 			   __entry->src = src;
353 			   __entry->count = count;
354 			   ),
355 	    TP_printk("pe=%010Lx, src=%010Lx, count=%u",
356 		      __entry->pe, __entry->src, __entry->count)
357 );
358 
359 TRACE_EVENT(amdgpu_vm_flush,
360 	    TP_PROTO(struct amdgpu_ring *ring, unsigned vm_id,
361 		     uint64_t pd_addr),
362 	    TP_ARGS(ring, vm_id, pd_addr),
363 	    TP_STRUCT__entry(
364 			     __field(u32, ring)
365 			     __field(u32, vm_id)
366 			     __field(u32, vm_hub)
367 			     __field(u64, pd_addr)
368 			     ),
369 
370 	    TP_fast_assign(
371 			   __entry->ring = ring->idx;
372 			   __entry->vm_id = vm_id;
373 			   __entry->vm_hub = ring->funcs->vmhub;
374 			   __entry->pd_addr = pd_addr;
375 			   ),
376 	    TP_printk("ring=%u, id=%u, hub=%u, pd_addr=%010Lx",
377 		      __entry->ring, __entry->vm_id,
378 		      __entry->vm_hub,__entry->pd_addr)
379 );
380 
381 TRACE_EVENT(amdgpu_bo_list_set,
382 	    TP_PROTO(struct amdgpu_bo_list *list, struct amdgpu_bo *bo),
383 	    TP_ARGS(list, bo),
384 	    TP_STRUCT__entry(
385 			     __field(struct amdgpu_bo_list *, list)
386 			     __field(struct amdgpu_bo *, bo)
387 			     __field(u64, bo_size)
388 			     ),
389 
390 	    TP_fast_assign(
391 			   __entry->list = list;
392 			   __entry->bo = bo;
393 			   __entry->bo_size = amdgpu_bo_size(bo);
394 			   ),
395 	    TP_printk("list=%p, bo=%p, bo_size=%Ld",
396 		      __entry->list,
397 		      __entry->bo,
398 		      __entry->bo_size)
399 );
400 
401 TRACE_EVENT(amdgpu_cs_bo_status,
402 	    TP_PROTO(uint64_t total_bo, uint64_t total_size),
403 	    TP_ARGS(total_bo, total_size),
404 	    TP_STRUCT__entry(
405 			__field(u64, total_bo)
406 			__field(u64, total_size)
407 			),
408 
409 	    TP_fast_assign(
410 			__entry->total_bo = total_bo;
411 			__entry->total_size = total_size;
412 			),
413 	    TP_printk("total_bo_size=%Ld, total_bo_count=%Ld",
414 			__entry->total_bo, __entry->total_size)
415 );
416 
417 TRACE_EVENT(amdgpu_ttm_bo_move,
418 	    TP_PROTO(struct amdgpu_bo* bo, uint32_t new_placement, uint32_t old_placement),
419 	    TP_ARGS(bo, new_placement, old_placement),
420 	    TP_STRUCT__entry(
421 			__field(struct amdgpu_bo *, bo)
422 			__field(u64, bo_size)
423 			__field(u32, new_placement)
424 			__field(u32, old_placement)
425 			),
426 
427 	    TP_fast_assign(
428 			__entry->bo      = bo;
429 			__entry->bo_size = amdgpu_bo_size(bo);
430 			__entry->new_placement = new_placement;
431 			__entry->old_placement = old_placement;
432 			),
433 	    TP_printk("bo=%p, from=%d, to=%d, size=%Ld",
434 			__entry->bo, __entry->old_placement,
435 			__entry->new_placement, __entry->bo_size)
436 );
437 
438 #undef AMDGPU_JOB_GET_TIMELINE_NAME
439 #endif
440 
441 /* This part must be outside protection */
442 #undef TRACE_INCLUDE_PATH
443 #define TRACE_INCLUDE_PATH ../../drivers/gpu/drm/amd/amdgpu
444 #include <trace/define_trace.h>
445