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(unsigned ih, struct amdgpu_iv_entry *iv),
80 	    TP_ARGS(ih, iv),
81 	    TP_STRUCT__entry(
82 			     __field(unsigned, ih)
83 			     __field(unsigned, client_id)
84 			     __field(unsigned, src_id)
85 			     __field(unsigned, ring_id)
86 			     __field(unsigned, vmid)
87 			     __field(unsigned, vmid_src)
88 			     __field(uint64_t, timestamp)
89 			     __field(unsigned, timestamp_src)
90 			     __field(unsigned, pasid)
91 			     __array(unsigned, src_data, 4)
92 			    ),
93 	    TP_fast_assign(
94 			   __entry->ih = ih;
95 			   __entry->client_id = iv->client_id;
96 			   __entry->src_id = iv->src_id;
97 			   __entry->ring_id = iv->ring_id;
98 			   __entry->vmid = iv->vmid;
99 			   __entry->vmid_src = iv->vmid_src;
100 			   __entry->timestamp = iv->timestamp;
101 			   __entry->timestamp_src = iv->timestamp_src;
102 			   __entry->pasid = iv->pasid;
103 			   __entry->src_data[0] = iv->src_data[0];
104 			   __entry->src_data[1] = iv->src_data[1];
105 			   __entry->src_data[2] = iv->src_data[2];
106 			   __entry->src_data[3] = iv->src_data[3];
107 			   ),
108 	    TP_printk("ih:%u client_id:%u src_id:%u ring:%u vmid:%u "
109 		      "timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x",
110 		      __entry->ih, __entry->client_id, __entry->src_id,
111 		      __entry->ring_id, __entry->vmid,
112 		      __entry->timestamp, __entry->pasid,
113 		      __entry->src_data[0], __entry->src_data[1],
114 		      __entry->src_data[2], __entry->src_data[3])
115 );
116 
117 
118 TRACE_EVENT(amdgpu_bo_create,
119 	    TP_PROTO(struct amdgpu_bo *bo),
120 	    TP_ARGS(bo),
121 	    TP_STRUCT__entry(
122 			     __field(struct amdgpu_bo *, bo)
123 			     __field(u32, pages)
124 			     __field(u32, type)
125 			     __field(u32, prefer)
126 			     __field(u32, allow)
127 			     __field(u32, visible)
128 			     ),
129 
130 	    TP_fast_assign(
131 			   __entry->bo = bo;
132 			   __entry->pages = bo->tbo.num_pages;
133 			   __entry->type = bo->tbo.mem.mem_type;
134 			   __entry->prefer = bo->preferred_domains;
135 			   __entry->allow = bo->allowed_domains;
136 			   __entry->visible = bo->flags;
137 			   ),
138 
139 	    TP_printk("bo=%p, pages=%u, type=%d, preferred=%d, allowed=%d, visible=%d",
140 		       __entry->bo, __entry->pages, __entry->type,
141 		       __entry->prefer, __entry->allow, __entry->visible)
142 );
143 
144 TRACE_EVENT(amdgpu_cs,
145 	    TP_PROTO(struct amdgpu_cs_parser *p, int i),
146 	    TP_ARGS(p, i),
147 	    TP_STRUCT__entry(
148 			     __field(struct amdgpu_bo_list *, bo_list)
149 			     __field(u32, ring)
150 			     __field(u32, dw)
151 			     __field(u32, fences)
152 			     ),
153 
154 	    TP_fast_assign(
155 			   __entry->bo_list = p->bo_list;
156 			   __entry->ring = to_amdgpu_ring(p->entity->rq->sched)->idx;
157 			   __entry->dw = p->job->ibs[i].length_dw;
158 			   __entry->fences = amdgpu_fence_count_emitted(
159 				to_amdgpu_ring(p->entity->rq->sched));
160 			   ),
161 	    TP_printk("bo_list=%p, ring=%u, dw=%u, fences=%u",
162 		      __entry->bo_list, __entry->ring, __entry->dw,
163 		      __entry->fences)
164 );
165 
166 TRACE_EVENT(amdgpu_cs_ioctl,
167 	    TP_PROTO(struct amdgpu_job *job),
168 	    TP_ARGS(job),
169 	    TP_STRUCT__entry(
170 			     __field(uint64_t, sched_job_id)
171 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
172 			     __field(unsigned int, context)
173 			     __field(unsigned int, seqno)
174 			     __field(struct dma_fence *, fence)
175 			     __field(char *, ring_name)
176 			     __field(u32, num_ibs)
177 			     ),
178 
179 	    TP_fast_assign(
180 			   __entry->sched_job_id = job->base.id;
181 			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
182 			   __entry->context = job->base.s_fence->finished.context;
183 			   __entry->seqno = job->base.s_fence->finished.seqno;
184 			   __entry->ring_name = to_amdgpu_ring(job->base.sched)->name;
185 			   __entry->num_ibs = job->num_ibs;
186 			   ),
187 	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
188 		      __entry->sched_job_id, __get_str(timeline), __entry->context,
189 		      __entry->seqno, __entry->ring_name, __entry->num_ibs)
190 );
191 
192 TRACE_EVENT(amdgpu_sched_run_job,
193 	    TP_PROTO(struct amdgpu_job *job),
194 	    TP_ARGS(job),
195 	    TP_STRUCT__entry(
196 			     __field(uint64_t, sched_job_id)
197 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
198 			     __field(unsigned int, context)
199 			     __field(unsigned int, seqno)
200 			     __field(char *, ring_name)
201 			     __field(u32, num_ibs)
202 			     ),
203 
204 	    TP_fast_assign(
205 			   __entry->sched_job_id = job->base.id;
206 			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
207 			   __entry->context = job->base.s_fence->finished.context;
208 			   __entry->seqno = job->base.s_fence->finished.seqno;
209 			   __entry->ring_name = to_amdgpu_ring(job->base.sched)->name;
210 			   __entry->num_ibs = job->num_ibs;
211 			   ),
212 	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
213 		      __entry->sched_job_id, __get_str(timeline), __entry->context,
214 		      __entry->seqno, __entry->ring_name, __entry->num_ibs)
215 );
216 
217 
218 TRACE_EVENT(amdgpu_vm_grab_id,
219 	    TP_PROTO(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
220 		     struct amdgpu_job *job),
221 	    TP_ARGS(vm, ring, job),
222 	    TP_STRUCT__entry(
223 			     __field(u32, pasid)
224 			     __string(ring, ring->name)
225 			     __field(u32, ring)
226 			     __field(u32, vmid)
227 			     __field(u32, vm_hub)
228 			     __field(u64, pd_addr)
229 			     __field(u32, needs_flush)
230 			     ),
231 
232 	    TP_fast_assign(
233 			   __entry->pasid = vm->pasid;
234 			   __assign_str(ring, ring->name)
235 			   __entry->vmid = job->vmid;
236 			   __entry->vm_hub = ring->funcs->vmhub,
237 			   __entry->pd_addr = job->vm_pd_addr;
238 			   __entry->needs_flush = job->vm_needs_flush;
239 			   ),
240 	    TP_printk("pasid=%d, ring=%s, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
241 		      __entry->pasid, __get_str(ring), __entry->vmid,
242 		      __entry->vm_hub, __entry->pd_addr, __entry->needs_flush)
243 );
244 
245 TRACE_EVENT(amdgpu_vm_bo_map,
246 	    TP_PROTO(struct amdgpu_bo_va *bo_va,
247 		     struct amdgpu_bo_va_mapping *mapping),
248 	    TP_ARGS(bo_va, mapping),
249 	    TP_STRUCT__entry(
250 			     __field(struct amdgpu_bo *, bo)
251 			     __field(long, start)
252 			     __field(long, last)
253 			     __field(u64, offset)
254 			     __field(u64, flags)
255 			     ),
256 
257 	    TP_fast_assign(
258 			   __entry->bo = bo_va ? bo_va->base.bo : NULL;
259 			   __entry->start = mapping->start;
260 			   __entry->last = mapping->last;
261 			   __entry->offset = mapping->offset;
262 			   __entry->flags = mapping->flags;
263 			   ),
264 	    TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%llx",
265 		      __entry->bo, __entry->start, __entry->last,
266 		      __entry->offset, __entry->flags)
267 );
268 
269 TRACE_EVENT(amdgpu_vm_bo_unmap,
270 	    TP_PROTO(struct amdgpu_bo_va *bo_va,
271 		     struct amdgpu_bo_va_mapping *mapping),
272 	    TP_ARGS(bo_va, mapping),
273 	    TP_STRUCT__entry(
274 			     __field(struct amdgpu_bo *, bo)
275 			     __field(long, start)
276 			     __field(long, last)
277 			     __field(u64, offset)
278 			     __field(u64, flags)
279 			     ),
280 
281 	    TP_fast_assign(
282 			   __entry->bo = bo_va ? bo_va->base.bo : NULL;
283 			   __entry->start = mapping->start;
284 			   __entry->last = mapping->last;
285 			   __entry->offset = mapping->offset;
286 			   __entry->flags = mapping->flags;
287 			   ),
288 	    TP_printk("bo=%p, start=%lx, last=%lx, offset=%010llx, flags=%llx",
289 		      __entry->bo, __entry->start, __entry->last,
290 		      __entry->offset, __entry->flags)
291 );
292 
293 DECLARE_EVENT_CLASS(amdgpu_vm_mapping,
294 	    TP_PROTO(struct amdgpu_bo_va_mapping *mapping),
295 	    TP_ARGS(mapping),
296 	    TP_STRUCT__entry(
297 			     __field(u64, soffset)
298 			     __field(u64, eoffset)
299 			     __field(u64, flags)
300 			     ),
301 
302 	    TP_fast_assign(
303 			   __entry->soffset = mapping->start;
304 			   __entry->eoffset = mapping->last + 1;
305 			   __entry->flags = mapping->flags;
306 			   ),
307 	    TP_printk("soffs=%010llx, eoffs=%010llx, flags=%llx",
308 		      __entry->soffset, __entry->eoffset, __entry->flags)
309 );
310 
311 DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_update,
312 	    TP_PROTO(struct amdgpu_bo_va_mapping *mapping),
313 	    TP_ARGS(mapping)
314 );
315 
316 DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_mapping,
317 	    TP_PROTO(struct amdgpu_bo_va_mapping *mapping),
318 	    TP_ARGS(mapping)
319 );
320 
321 DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_cs,
322 	    TP_PROTO(struct amdgpu_bo_va_mapping *mapping),
323 	    TP_ARGS(mapping)
324 );
325 
326 TRACE_EVENT(amdgpu_vm_set_ptes,
327 	    TP_PROTO(uint64_t pe, uint64_t addr, unsigned count,
328 		     uint32_t incr, uint64_t flags),
329 	    TP_ARGS(pe, addr, count, incr, flags),
330 	    TP_STRUCT__entry(
331 			     __field(u64, pe)
332 			     __field(u64, addr)
333 			     __field(u32, count)
334 			     __field(u32, incr)
335 			     __field(u64, flags)
336 			     ),
337 
338 	    TP_fast_assign(
339 			   __entry->pe = pe;
340 			   __entry->addr = addr;
341 			   __entry->count = count;
342 			   __entry->incr = incr;
343 			   __entry->flags = flags;
344 			   ),
345 	    TP_printk("pe=%010Lx, addr=%010Lx, incr=%u, flags=%llx, count=%u",
346 		      __entry->pe, __entry->addr, __entry->incr,
347 		      __entry->flags, __entry->count)
348 );
349 
350 TRACE_EVENT(amdgpu_vm_copy_ptes,
351 	    TP_PROTO(uint64_t pe, uint64_t src, unsigned count),
352 	    TP_ARGS(pe, src, count),
353 	    TP_STRUCT__entry(
354 			     __field(u64, pe)
355 			     __field(u64, src)
356 			     __field(u32, count)
357 			     ),
358 
359 	    TP_fast_assign(
360 			   __entry->pe = pe;
361 			   __entry->src = src;
362 			   __entry->count = count;
363 			   ),
364 	    TP_printk("pe=%010Lx, src=%010Lx, count=%u",
365 		      __entry->pe, __entry->src, __entry->count)
366 );
367 
368 TRACE_EVENT(amdgpu_vm_flush,
369 	    TP_PROTO(struct amdgpu_ring *ring, unsigned vmid,
370 		     uint64_t pd_addr),
371 	    TP_ARGS(ring, vmid, pd_addr),
372 	    TP_STRUCT__entry(
373 			     __string(ring, ring->name)
374 			     __field(u32, vmid)
375 			     __field(u32, vm_hub)
376 			     __field(u64, pd_addr)
377 			     ),
378 
379 	    TP_fast_assign(
380 			   __assign_str(ring, ring->name)
381 			   __entry->vmid = vmid;
382 			   __entry->vm_hub = ring->funcs->vmhub;
383 			   __entry->pd_addr = pd_addr;
384 			   ),
385 	    TP_printk("ring=%s, id=%u, hub=%u, pd_addr=%010Lx",
386 		      __get_str(ring), __entry->vmid,
387 		      __entry->vm_hub,__entry->pd_addr)
388 );
389 
390 DECLARE_EVENT_CLASS(amdgpu_pasid,
391 	    TP_PROTO(unsigned pasid),
392 	    TP_ARGS(pasid),
393 	    TP_STRUCT__entry(
394 			     __field(unsigned, pasid)
395 			     ),
396 	    TP_fast_assign(
397 			   __entry->pasid = pasid;
398 			   ),
399 	    TP_printk("pasid=%u", __entry->pasid)
400 );
401 
402 DEFINE_EVENT(amdgpu_pasid, amdgpu_pasid_allocated,
403 	    TP_PROTO(unsigned pasid),
404 	    TP_ARGS(pasid)
405 );
406 
407 DEFINE_EVENT(amdgpu_pasid, amdgpu_pasid_freed,
408 	    TP_PROTO(unsigned pasid),
409 	    TP_ARGS(pasid)
410 );
411 
412 TRACE_EVENT(amdgpu_bo_list_set,
413 	    TP_PROTO(struct amdgpu_bo_list *list, struct amdgpu_bo *bo),
414 	    TP_ARGS(list, bo),
415 	    TP_STRUCT__entry(
416 			     __field(struct amdgpu_bo_list *, list)
417 			     __field(struct amdgpu_bo *, bo)
418 			     __field(u64, bo_size)
419 			     ),
420 
421 	    TP_fast_assign(
422 			   __entry->list = list;
423 			   __entry->bo = bo;
424 			   __entry->bo_size = amdgpu_bo_size(bo);
425 			   ),
426 	    TP_printk("list=%p, bo=%p, bo_size=%Ld",
427 		      __entry->list,
428 		      __entry->bo,
429 		      __entry->bo_size)
430 );
431 
432 TRACE_EVENT(amdgpu_cs_bo_status,
433 	    TP_PROTO(uint64_t total_bo, uint64_t total_size),
434 	    TP_ARGS(total_bo, total_size),
435 	    TP_STRUCT__entry(
436 			__field(u64, total_bo)
437 			__field(u64, total_size)
438 			),
439 
440 	    TP_fast_assign(
441 			__entry->total_bo = total_bo;
442 			__entry->total_size = total_size;
443 			),
444 	    TP_printk("total_bo_size=%Ld, total_bo_count=%Ld",
445 			__entry->total_bo, __entry->total_size)
446 );
447 
448 TRACE_EVENT(amdgpu_bo_move,
449 	    TP_PROTO(struct amdgpu_bo* bo, uint32_t new_placement, uint32_t old_placement),
450 	    TP_ARGS(bo, new_placement, old_placement),
451 	    TP_STRUCT__entry(
452 			__field(struct amdgpu_bo *, bo)
453 			__field(u64, bo_size)
454 			__field(u32, new_placement)
455 			__field(u32, old_placement)
456 			),
457 
458 	    TP_fast_assign(
459 			__entry->bo      = bo;
460 			__entry->bo_size = amdgpu_bo_size(bo);
461 			__entry->new_placement = new_placement;
462 			__entry->old_placement = old_placement;
463 			),
464 	    TP_printk("bo=%p, from=%d, to=%d, size=%Ld",
465 			__entry->bo, __entry->old_placement,
466 			__entry->new_placement, __entry->bo_size)
467 );
468 
469 TRACE_EVENT(amdgpu_ib_pipe_sync,
470 	    TP_PROTO(struct amdgpu_job *sched_job, struct dma_fence *fence),
471 	    TP_ARGS(sched_job, fence),
472 	    TP_STRUCT__entry(
473 			     __field(const char *,name)
474 			     __field(uint64_t, id)
475 			     __field(struct dma_fence *, fence)
476 			     __field(uint64_t, ctx)
477 			     __field(unsigned, seqno)
478 			     ),
479 
480 	    TP_fast_assign(
481 			   __entry->name = sched_job->base.sched->name;
482 			   __entry->id = sched_job->base.id;
483 			   __entry->fence = fence;
484 			   __entry->ctx = fence->context;
485 			   __entry->seqno = fence->seqno;
486 			   ),
487 	    TP_printk("job ring=%s, id=%llu, need pipe sync to fence=%p, context=%llu, seq=%u",
488 		      __entry->name, __entry->id,
489 		      __entry->fence, __entry->ctx,
490 		      __entry->seqno)
491 );
492 
493 #undef AMDGPU_JOB_GET_TIMELINE_NAME
494 #endif
495 
496 /* This part must be outside protection */
497 #undef TRACE_INCLUDE_PATH
498 #define TRACE_INCLUDE_PATH ../../drivers/gpu/drm/amd/amdgpu
499 #include <trace/define_trace.h>
500