add boot partition size

hc

2023-12-11 d2ccde1c8e90d38cee87a1b0309ad2827f3fd30d

 kernel/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
..
..
@@ -28,8 +28,6 @@
28
28
 #include <linux/types.h>
29
29
 #include <linux/tracepoint.h>
30
30
 
31
-#include <drm/drmP.h>
32
-
33
31
 #undef TRACE_SYSTEM
34
32
 #define TRACE_SYSTEM amdgpu
35
33
 #define TRACE_INCLUDE_FILE amdgpu_trace
..
..
@@ -37,7 +35,7 @@
37
35
 #define AMDGPU_JOB_GET_TIMELINE_NAME(job) \
38
36
 	 job->base.s_fence->finished.ops->get_timeline_name(&job->base.s_fence->finished)
39
37
 
40
-TRACE_EVENT(amdgpu_mm_rreg,
38
+TRACE_EVENT(amdgpu_device_rreg,
41
39
 	    TP_PROTO(unsigned did, uint32_t reg, uint32_t value),
42
40
 	    TP_ARGS(did, reg, value),
43
41
 	    TP_STRUCT__entry(
..
..
@@ -56,7 +54,7 @@
56
54
 		      (unsigned long)__entry->value)
57
55
 );
58
56
 
59
-TRACE_EVENT(amdgpu_mm_wreg,
57
+TRACE_EVENT(amdgpu_device_wreg,
60
58
 	    TP_PROTO(unsigned did, uint32_t reg, uint32_t value),
61
59
 	    TP_ARGS(did, reg, value),
62
60
 	    TP_STRUCT__entry(
..
..
@@ -76,9 +74,10 @@
76
74
 );
77
75
 
78
76
 TRACE_EVENT(amdgpu_iv,
79
-	    TP_PROTO(struct amdgpu_iv_entry *iv),
80
-	    TP_ARGS(iv),
77
+	    TP_PROTO(unsigned ih, struct amdgpu_iv_entry *iv),
78
+	    TP_ARGS(ih, iv),
81
79
 	    TP_STRUCT__entry(
80
+			     __field(unsigned, ih)
82
81
 			     __field(unsigned, client_id)
83
82
 			     __field(unsigned, src_id)
84
83
 			     __field(unsigned, ring_id)
..
..
@@ -90,6 +89,7 @@
90
89
 			     __array(unsigned, src_data, 4)
91
90
 			    ),
92
91
 	    TP_fast_assign(
92
+			   __entry->ih = ih;
93
93
 			   __entry->client_id = iv->client_id;
94
94
 			   __entry->src_id = iv->src_id;
95
95
 			   __entry->ring_id = iv->ring_id;
..
..
@@ -103,8 +103,9 @@
103
103
 			   __entry->src_data[2] = iv->src_data[2];
104
104
 			   __entry->src_data[3] = iv->src_data[3];
105
105
 			   ),
106
-	    TP_printk("client_id:%u src_id:%u ring:%u vmid:%u timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x\n",
107
-		      __entry->client_id, __entry->src_id,
106
+	    TP_printk("ih:%u client_id:%u src_id:%u ring:%u vmid:%u "
107
+		      "timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x",
108
+		      __entry->ih, __entry->client_id, __entry->src_id,
108
109
 		      __entry->ring_id, __entry->vmid,
109
110
 		      __entry->timestamp, __entry->pasid,
110
111
 		      __entry->src_data[0], __entry->src_data[1],
..
..
@@ -150,10 +151,10 @@
150
151
 
151
152
 	    TP_fast_assign(
152
153
 			   __entry->bo_list = p->bo_list;
153
-			   __entry->ring = p->ring->idx;
154
+			   __entry->ring = to_amdgpu_ring(p->entity->rq->sched)->idx;
154
155
 			   __entry->dw = p->job->ibs[i].length_dw;
155
156
 			   __entry->fences = amdgpu_fence_count_emitted(
156
-				p->ring);
157
+				to_amdgpu_ring(p->entity->rq->sched));
157
158
 			   ),
158
159
 	    TP_printk("bo_list=%p, ring=%u, dw=%u, fences=%u",
159
160
 		      __entry->bo_list, __entry->ring, __entry->dw,
..
..
@@ -169,7 +170,7 @@
169
170
 			     __field(unsigned int, context)
170
171
 			     __field(unsigned int, seqno)
171
172
 			     __field(struct dma_fence *, fence)
172
-			     __field(char *, ring_name)
173
+			     __string(ring, to_amdgpu_ring(job->base.sched)->name)
173
174
 			     __field(u32, num_ibs)
174
175
 			     ),
175
176
 
..
..
@@ -178,12 +179,12 @@
178
179
 			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
179
180
 			   __entry->context = job->base.s_fence->finished.context;
180
181
 			   __entry->seqno = job->base.s_fence->finished.seqno;
181
-			   __entry->ring_name = to_amdgpu_ring(job->base.sched)->name;
182
+			   __assign_str(ring, to_amdgpu_ring(job->base.sched)->name)
182
183
 			   __entry->num_ibs = job->num_ibs;
183
184
 			   ),
184
185
 	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
185
186
 		      __entry->sched_job_id, __get_str(timeline), __entry->context,
186
-		      __entry->seqno, __entry->ring_name, __entry->num_ibs)
187
+		      __entry->seqno, __get_str(ring), __entry->num_ibs)
187
188
 );
188
189
 
189
190
 TRACE_EVENT(amdgpu_sched_run_job,
..
..
@@ -194,7 +195,7 @@
194
195
 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
195
196
 			     __field(unsigned int, context)
196
197
 			     __field(unsigned int, seqno)
197
-			     __field(char *, ring_name)
198
+			     __string(ring, to_amdgpu_ring(job->base.sched)->name)
198
199
 			     __field(u32, num_ibs)
199
200
 			     ),
200
201
 
..
..
@@ -203,12 +204,12 @@
203
204
 			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
204
205
 			   __entry->context = job->base.s_fence->finished.context;
205
206
 			   __entry->seqno = job->base.s_fence->finished.seqno;
206
-			   __entry->ring_name = to_amdgpu_ring(job->base.sched)->name;
207
+			   __assign_str(ring, to_amdgpu_ring(job->base.sched)->name)
207
208
 			   __entry->num_ibs = job->num_ibs;
208
209
 			   ),
209
210
 	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
210
211
 		      __entry->sched_job_id, __get_str(timeline), __entry->context,
211
-		      __entry->seqno, __entry->ring_name, __entry->num_ibs)
212
+		      __entry->seqno, __get_str(ring), __entry->num_ibs)
212
213
 );
213
214
 
214
215
 
..
..
@@ -218,6 +219,7 @@
218
219
 	    TP_ARGS(vm, ring, job),
219
220
 	    TP_STRUCT__entry(
220
221
 			     __field(u32, pasid)
222
+			     __string(ring, ring->name)
221
223
 			     __field(u32, ring)
222
224
 			     __field(u32, vmid)
223
225
 			     __field(u32, vm_hub)
..
..
@@ -227,14 +229,14 @@
227
229
 
228
230
 	    TP_fast_assign(
229
231
 			   __entry->pasid = vm->pasid;
230
-			   __entry->ring = ring->idx;
232
+			   __assign_str(ring, ring->name)
231
233
 			   __entry->vmid = job->vmid;
232
234
 			   __entry->vm_hub = ring->funcs->vmhub,
233
235
 			   __entry->pd_addr = job->vm_pd_addr;
234
236
 			   __entry->needs_flush = job->vm_needs_flush;
235
237
 			   ),
236
-	    TP_printk("pasid=%d, ring=%u, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
237
-		      __entry->pasid, __entry->ring, __entry->vmid,
238
+	    TP_printk("pasid=%d, ring=%s, id=%u, hub=%u, pd_addr=%010Lx needs_flush=%u",
239
+		      __entry->pasid, __get_str(ring), __entry->vmid,
238
240
 		      __entry->vm_hub, __entry->pd_addr, __entry->needs_flush)
239
241
 );
240
242
 
..
..
@@ -319,16 +321,60 @@
319
321
 	    TP_ARGS(mapping)
320
322
 );
321
323
 
324
+TRACE_EVENT(amdgpu_vm_update_ptes,
325
+	    TP_PROTO(struct amdgpu_vm_update_params *p,
326
+		     uint64_t start, uint64_t end,
327
+		     unsigned int nptes, uint64_t dst,
328
+		     uint64_t incr, uint64_t flags,
329
+		     pid_t pid, uint64_t vm_ctx),
330
+	TP_ARGS(p, start, end, nptes, dst, incr, flags, pid, vm_ctx),
331
+	TP_STRUCT__entry(
332
+			 __field(u64, start)
333
+			 __field(u64, end)
334
+			 __field(u64, flags)
335
+			 __field(unsigned int, nptes)
336
+			 __field(u64, incr)
337
+			 __field(pid_t, pid)
338
+			 __field(u64, vm_ctx)
339
+			 __dynamic_array(u64, dst, nptes)
340
+	),
341
+
342
+	TP_fast_assign(
343
+			unsigned int i;
344
+
345
+			__entry->start = start;
346
+			__entry->end = end;
347
+			__entry->flags = flags;
348
+			__entry->incr = incr;
349
+			__entry->nptes = nptes;
350
+			__entry->pid = pid;
351
+			__entry->vm_ctx = vm_ctx;
352
+			for (i = 0; i < nptes; ++i) {
353
+				u64 addr = p->pages_addr ? amdgpu_vm_map_gart(
354
+					p->pages_addr, dst) : dst;
355
+
356
+				((u64 *)__get_dynamic_array(dst))[i] = addr;
357
+				dst += incr;
358
+			}
359
+	),
360
+	TP_printk("pid:%u vm_ctx:0x%llx start:0x%010llx end:0x%010llx,"
361
+		  " flags:0x%llx, incr:%llu, dst:\n%s", __entry->pid,
362
+		  __entry->vm_ctx, __entry->start, __entry->end,
363
+		  __entry->flags, __entry->incr,  __print_array(
364
+		  __get_dynamic_array(dst), __entry->nptes, 8))
365
+);
366
+
322
367
 TRACE_EVENT(amdgpu_vm_set_ptes,
323
368
 	    TP_PROTO(uint64_t pe, uint64_t addr, unsigned count,
324
-		     uint32_t incr, uint64_t flags),
325
-	    TP_ARGS(pe, addr, count, incr, flags),
369
+		     uint32_t incr, uint64_t flags, bool direct),
370
+	    TP_ARGS(pe, addr, count, incr, flags, direct),
326
371
 	    TP_STRUCT__entry(
327
372
 			     __field(u64, pe)
328
373
 			     __field(u64, addr)
329
374
 			     __field(u32, count)
330
375
 			     __field(u32, incr)
331
376
 			     __field(u64, flags)
377
+			     __field(bool, direct)
332
378
 			     ),
333
379
 
334
380
 	    TP_fast_assign(
..
..
@@ -337,28 +383,32 @@
337
383
 			   __entry->count = count;
338
384
 			   __entry->incr = incr;
339
385
 			   __entry->flags = flags;
386
+			   __entry->direct = direct;
340
387
 			   ),
341
-	    TP_printk("pe=%010Lx, addr=%010Lx, incr=%u, flags=%llx, count=%u",
342
-		      __entry->pe, __entry->addr, __entry->incr,
343
-		      __entry->flags, __entry->count)
388
+	    TP_printk("pe=%010Lx, addr=%010Lx, incr=%u, flags=%llx, count=%u, "
389
+		      "direct=%d", __entry->pe, __entry->addr, __entry->incr,
390
+		      __entry->flags, __entry->count, __entry->direct)
344
391
 );
345
392
 
346
393
 TRACE_EVENT(amdgpu_vm_copy_ptes,
347
-	    TP_PROTO(uint64_t pe, uint64_t src, unsigned count),
348
-	    TP_ARGS(pe, src, count),
394
+	    TP_PROTO(uint64_t pe, uint64_t src, unsigned count, bool direct),
395
+	    TP_ARGS(pe, src, count, direct),
349
396
 	    TP_STRUCT__entry(
350
397
 			     __field(u64, pe)
351
398
 			     __field(u64, src)
352
399
 			     __field(u32, count)
400
+			     __field(bool, direct)
353
401
 			     ),
354
402
 
355
403
 	    TP_fast_assign(
356
404
 			   __entry->pe = pe;
357
405
 			   __entry->src = src;
358
406
 			   __entry->count = count;
407
+			   __entry->direct = direct;
359
408
 			   ),
360
-	    TP_printk("pe=%010Lx, src=%010Lx, count=%u",
361
-		      __entry->pe, __entry->src, __entry->count)
409
+	    TP_printk("pe=%010Lx, src=%010Lx, count=%u, direct=%d",
410
+		      __entry->pe, __entry->src, __entry->count,
411
+		      __entry->direct)
362
412
 );
363
413
 
364
414
 TRACE_EVENT(amdgpu_vm_flush,
..
..
@@ -366,20 +416,20 @@
366
416
 		     uint64_t pd_addr),
367
417
 	    TP_ARGS(ring, vmid, pd_addr),
368
418
 	    TP_STRUCT__entry(
369
-			     __field(u32, ring)
419
+			     __string(ring, ring->name)
370
420
 			     __field(u32, vmid)
371
421
 			     __field(u32, vm_hub)
372
422
 			     __field(u64, pd_addr)
373
423
 			     ),
374
424
 
375
425
 	    TP_fast_assign(
376
-			   __entry->ring = ring->idx;
426
+			   __assign_str(ring, ring->name)
377
427
 			   __entry->vmid = vmid;
378
428
 			   __entry->vm_hub = ring->funcs->vmhub;
379
429
 			   __entry->pd_addr = pd_addr;
380
430
 			   ),
381
-	    TP_printk("ring=%u, id=%u, hub=%u, pd_addr=%010Lx",
382
-		      __entry->ring, __entry->vmid,
431
+	    TP_printk("ring=%s, id=%u, hub=%u, pd_addr=%010Lx",
432
+		      __get_str(ring), __entry->vmid,
383
433
 		      __entry->vm_hub,__entry->pd_addr)
384
434
 );
385
435
 
..
..
@@ -462,6 +512,30 @@
462
512
 			__entry->new_placement, __entry->bo_size)
463
513
 );
464
514
 
515
+TRACE_EVENT(amdgpu_ib_pipe_sync,
516
+	    TP_PROTO(struct amdgpu_job *sched_job, struct dma_fence *fence),
517
+	    TP_ARGS(sched_job, fence),
518
+	    TP_STRUCT__entry(
519
+			     __string(ring, sched_job->base.sched->name)
520
+			     __field(uint64_t, id)
521
+			     __field(struct dma_fence *, fence)
522
+			     __field(uint64_t, ctx)
523
+			     __field(unsigned, seqno)
524
+			     ),
525
+
526
+	    TP_fast_assign(
527
+			   __assign_str(ring, sched_job->base.sched->name)
528
+			   __entry->id = sched_job->base.id;
529
+			   __entry->fence = fence;
530
+			   __entry->ctx = fence->context;
531
+			   __entry->seqno = fence->seqno;
532
+			   ),
533
+	    TP_printk("job ring=%s, id=%llu, need pipe sync to fence=%p, context=%llu, seq=%u",
534
+		      __get_str(ring), __entry->id,
535
+		      __entry->fence, __entry->ctx,
536
+		      __entry->seqno)
537
+);
538
+
465
539
 #undef AMDGPU_JOB_GET_TIMELINE_NAME
466
540
 #endif
467
541