1 /* SPDX-License-Identifier: GPL-2.0 */
3 #define TRACE_SYSTEM rcu
5 #if !defined(_TRACE_RCU_H) || defined(TRACE_HEADER_MULTI_READ)
8 #include <linux/tracepoint.h>
10 #ifdef CONFIG_RCU_TRACE
11 #define TRACE_EVENT_RCU TRACE_EVENT
13 #define TRACE_EVENT_RCU TRACE_EVENT_NOP
17 * Tracepoint for start/end markers used for utilization calculations.
18 * By convention, the string is of the following forms:
20 * "Start <activity>" -- Mark the start of the specified activity,
21 * such as "context switch". Nesting is permitted.
22 * "End <activity>" -- Mark the end of the specified activity.
24 * An "@" character within "<activity>" is a comment character: Data
25 * reduction scripts will ignore the "@" and the remainder of the line.
27 TRACE_EVENT(rcu_utilization,
29 TP_PROTO(const char *s),
34 __field(const char *, s)
41 TP_printk("%s", __entry->s)
44 #if defined(CONFIG_TREE_RCU)
47 * Tracepoint for grace-period events. Takes a string identifying the
48 * RCU flavor, the grace-period number, and a string identifying the
49 * grace-period-related event as follows:
51 * "AccReadyCB": CPU acclerates new callbacks to RCU_NEXT_READY_TAIL.
52 * "AccWaitCB": CPU accelerates new callbacks to RCU_WAIT_TAIL.
53 * "newreq": Request a new grace period.
54 * "start": Start a grace period.
55 * "cpustart": CPU first notices a grace-period start.
56 * "cpuqs": CPU passes through a quiescent state.
57 * "cpuonl": CPU comes online.
58 * "cpuofl": CPU goes offline.
59 * "cpuofl-bgp": CPU goes offline while blocking a grace period.
60 * "reqwait": GP kthread sleeps waiting for grace-period request.
61 * "reqwaitsig": GP kthread awakened by signal from reqwait state.
62 * "fqswait": GP kthread waiting until time to force quiescent states.
63 * "fqsstart": GP kthread starts forcing quiescent states.
64 * "fqsend": GP kthread done forcing quiescent states.
65 * "fqswaitsig": GP kthread awakened by signal from fqswait state.
66 * "end": End a grace period.
67 * "cpuend": CPU first notices a grace-period end.
69 TRACE_EVENT_RCU(rcu_grace_period,
71 TP_PROTO(const char *rcuname, unsigned long gp_seq, const char *gpevent),
73 TP_ARGS(rcuname, gp_seq, gpevent),
76 __field(const char *, rcuname)
77 __field(unsigned long, gp_seq)
78 __field(const char *, gpevent)
82 __entry->rcuname = rcuname;
83 __entry->gp_seq = gp_seq;
84 __entry->gpevent = gpevent;
87 TP_printk("%s %lu %s",
88 __entry->rcuname, __entry->gp_seq, __entry->gpevent)
92 * Tracepoint for future grace-period events. The caller should pull
93 * the data from the rcu_node structure, other than rcuname, which comes
94 * from the rcu_state structure, and event, which is one of the following:
96 * "Cleanup": Clean up rcu_node structure after previous GP.
97 * "CleanupMore": Clean up, and another GP is needed.
98 * "EndWait": Complete wait.
99 * "NoGPkthread": The RCU grace-period kthread has not yet started.
100 * "Prestarted": Someone beat us to the request
101 * "Startedleaf": Leaf node marked for future GP.
102 * "Startedleafroot": All nodes from leaf to root marked for future GP.
103 * "Startedroot": Requested a nocb grace period based on root-node data.
104 * "Startleaf": Request a grace period based on leaf-node data.
105 * "StartWait": Start waiting for the requested grace period.
107 TRACE_EVENT_RCU(rcu_future_grace_period,
109 TP_PROTO(const char *rcuname, unsigned long gp_seq,
110 unsigned long gp_seq_req, u8 level, int grplo, int grphi,
111 const char *gpevent),
113 TP_ARGS(rcuname, gp_seq, gp_seq_req, level, grplo, grphi, gpevent),
116 __field(const char *, rcuname)
117 __field(unsigned long, gp_seq)
118 __field(unsigned long, gp_seq_req)
122 __field(const char *, gpevent)
126 __entry->rcuname = rcuname;
127 __entry->gp_seq = gp_seq;
128 __entry->gp_seq_req = gp_seq_req;
129 __entry->level = level;
130 __entry->grplo = grplo;
131 __entry->grphi = grphi;
132 __entry->gpevent = gpevent;
135 TP_printk("%s %lu %lu %u %d %d %s",
136 __entry->rcuname, __entry->gp_seq, __entry->gp_seq_req, __entry->level,
137 __entry->grplo, __entry->grphi, __entry->gpevent)
141 * Tracepoint for grace-period-initialization events. These are
142 * distinguished by the type of RCU, the new grace-period number, the
143 * rcu_node structure level, the starting and ending CPU covered by the
144 * rcu_node structure, and the mask of CPUs that will be waited for.
145 * All but the type of RCU are extracted from the rcu_node structure.
147 TRACE_EVENT_RCU(rcu_grace_period_init,
149 TP_PROTO(const char *rcuname, unsigned long gp_seq, u8 level,
150 int grplo, int grphi, unsigned long qsmask),
152 TP_ARGS(rcuname, gp_seq, level, grplo, grphi, qsmask),
155 __field(const char *, rcuname)
156 __field(unsigned long, gp_seq)
160 __field(unsigned long, qsmask)
164 __entry->rcuname = rcuname;
165 __entry->gp_seq = gp_seq;
166 __entry->level = level;
167 __entry->grplo = grplo;
168 __entry->grphi = grphi;
169 __entry->qsmask = qsmask;
172 TP_printk("%s %lu %u %d %d %lx",
173 __entry->rcuname, __entry->gp_seq, __entry->level,
174 __entry->grplo, __entry->grphi, __entry->qsmask)
178 * Tracepoint for expedited grace-period events. Takes a string identifying
179 * the RCU flavor, the expedited grace-period sequence number, and a string
180 * identifying the grace-period-related event as follows:
182 * "snap": Captured snapshot of expedited grace period sequence number.
183 * "start": Started a real expedited grace period.
184 * "reset": Started resetting the tree
185 * "select": Started selecting the CPUs to wait on.
186 * "selectofl": Selected CPU partially offline.
187 * "startwait": Started waiting on selected CPUs.
188 * "end": Ended a real expedited grace period.
189 * "endwake": Woke piggybackers up.
190 * "done": Someone else did the expedited grace period for us.
192 TRACE_EVENT_RCU(rcu_exp_grace_period,
194 TP_PROTO(const char *rcuname, unsigned long gpseq, const char *gpevent),
196 TP_ARGS(rcuname, gpseq, gpevent),
199 __field(const char *, rcuname)
200 __field(unsigned long, gpseq)
201 __field(const char *, gpevent)
205 __entry->rcuname = rcuname;
206 __entry->gpseq = gpseq;
207 __entry->gpevent = gpevent;
210 TP_printk("%s %lu %s",
211 __entry->rcuname, __entry->gpseq, __entry->gpevent)
215 * Tracepoint for expedited grace-period funnel-locking events. Takes a
216 * string identifying the RCU flavor, an integer identifying the rcu_node
217 * combining-tree level, another pair of integers identifying the lowest-
218 * and highest-numbered CPU associated with the current rcu_node structure,
219 * and a string. identifying the grace-period-related event as follows:
221 * "nxtlvl": Advance to next level of rcu_node funnel
222 * "wait": Wait for someone else to do expedited GP
224 TRACE_EVENT_RCU(rcu_exp_funnel_lock,
226 TP_PROTO(const char *rcuname, u8 level, int grplo, int grphi,
227 const char *gpevent),
229 TP_ARGS(rcuname, level, grplo, grphi, gpevent),
232 __field(const char *, rcuname)
236 __field(const char *, gpevent)
240 __entry->rcuname = rcuname;
241 __entry->level = level;
242 __entry->grplo = grplo;
243 __entry->grphi = grphi;
244 __entry->gpevent = gpevent;
247 TP_printk("%s %d %d %d %s",
248 __entry->rcuname, __entry->level, __entry->grplo,
249 __entry->grphi, __entry->gpevent)
252 #ifdef CONFIG_RCU_NOCB_CPU
254 * Tracepoint for RCU no-CBs CPU callback handoffs. This event is intended
255 * to assist debugging of these handoffs.
257 * The first argument is the name of the RCU flavor, and the second is
258 * the number of the offloaded CPU are extracted. The third and final
259 * argument is a string as follows:
261 * "AlreadyAwake": The to-be-awakened rcuo kthread is already awake.
262 * "Bypass": rcuo GP kthread sees non-empty ->nocb_bypass.
263 * "CBSleep": rcuo CB kthread sleeping waiting for CBs.
264 * "Check": rcuo GP kthread checking specified CPU for work.
265 * "DeferredWake": Timer expired or polled check, time to wake.
266 * "DoWake": The to-be-awakened rcuo kthread needs to be awakened.
267 * "EndSleep": Done waiting for GP for !rcu_nocb_poll.
268 * "FirstBQ": New CB to empty ->nocb_bypass (->cblist maybe non-empty).
269 * "FirstBQnoWake": FirstBQ plus rcuo kthread need not be awakened.
270 * "FirstBQwake": FirstBQ plus rcuo kthread must be awakened.
271 * "FirstQ": New CB to empty ->cblist (->nocb_bypass maybe non-empty).
272 * "NeedWaitGP": rcuo GP kthread must wait on a grace period.
273 * "Poll": Start of new polling cycle for rcu_nocb_poll.
274 * "Sleep": Sleep waiting for GP for !rcu_nocb_poll.
275 * "Timer": Deferred-wake timer expired.
276 * "WakeEmptyIsDeferred": Wake rcuo kthread later, first CB to empty list.
277 * "WakeEmpty": Wake rcuo kthread, first CB to empty list.
278 * "WakeNot": Don't wake rcuo kthread.
279 * "WakeNotPoll": Don't wake rcuo kthread because it is polling.
280 * "WakeOvfIsDeferred": Wake rcuo kthread later, CB list is huge.
281 * "WokeEmpty": rcuo CB kthread woke to find empty list.
283 TRACE_EVENT_RCU(rcu_nocb_wake,
285 TP_PROTO(const char *rcuname, int cpu, const char *reason),
287 TP_ARGS(rcuname, cpu, reason),
290 __field(const char *, rcuname)
292 __field(const char *, reason)
296 __entry->rcuname = rcuname;
298 __entry->reason = reason;
301 TP_printk("%s %d %s", __entry->rcuname, __entry->cpu, __entry->reason)
306 * Tracepoint for tasks blocking within preemptible-RCU read-side
307 * critical sections. Track the type of RCU (which one day might
308 * include SRCU), the grace-period number that the task is blocking
309 * (the current or the next), and the task's PID.
311 TRACE_EVENT_RCU(rcu_preempt_task,
313 TP_PROTO(const char *rcuname, int pid, unsigned long gp_seq),
315 TP_ARGS(rcuname, pid, gp_seq),
318 __field(const char *, rcuname)
319 __field(unsigned long, gp_seq)
324 __entry->rcuname = rcuname;
325 __entry->gp_seq = gp_seq;
329 TP_printk("%s %lu %d",
330 __entry->rcuname, __entry->gp_seq, __entry->pid)
334 * Tracepoint for tasks that blocked within a given preemptible-RCU
335 * read-side critical section exiting that critical section. Track the
336 * type of RCU (which one day might include SRCU) and the task's PID.
338 TRACE_EVENT_RCU(rcu_unlock_preempted_task,
340 TP_PROTO(const char *rcuname, unsigned long gp_seq, int pid),
342 TP_ARGS(rcuname, gp_seq, pid),
345 __field(const char *, rcuname)
346 __field(unsigned long, gp_seq)
351 __entry->rcuname = rcuname;
352 __entry->gp_seq = gp_seq;
356 TP_printk("%s %lu %d", __entry->rcuname, __entry->gp_seq, __entry->pid)
360 * Tracepoint for quiescent-state-reporting events. These are
361 * distinguished by the type of RCU, the grace-period number, the
362 * mask of quiescent lower-level entities, the rcu_node structure level,
363 * the starting and ending CPU covered by the rcu_node structure, and
364 * whether there are any blocked tasks blocking the current grace period.
365 * All but the type of RCU are extracted from the rcu_node structure.
367 TRACE_EVENT_RCU(rcu_quiescent_state_report,
369 TP_PROTO(const char *rcuname, unsigned long gp_seq,
370 unsigned long mask, unsigned long qsmask,
371 u8 level, int grplo, int grphi, int gp_tasks),
373 TP_ARGS(rcuname, gp_seq, mask, qsmask, level, grplo, grphi, gp_tasks),
376 __field(const char *, rcuname)
377 __field(unsigned long, gp_seq)
378 __field(unsigned long, mask)
379 __field(unsigned long, qsmask)
383 __field(u8, gp_tasks)
387 __entry->rcuname = rcuname;
388 __entry->gp_seq = gp_seq;
389 __entry->mask = mask;
390 __entry->qsmask = qsmask;
391 __entry->level = level;
392 __entry->grplo = grplo;
393 __entry->grphi = grphi;
394 __entry->gp_tasks = gp_tasks;
397 TP_printk("%s %lu %lx>%lx %u %d %d %u",
398 __entry->rcuname, __entry->gp_seq,
399 __entry->mask, __entry->qsmask, __entry->level,
400 __entry->grplo, __entry->grphi, __entry->gp_tasks)
404 * Tracepoint for quiescent states detected by force_quiescent_state().
405 * These trace events include the type of RCU, the grace-period number
406 * that was blocked by the CPU, the CPU itself, and the type of quiescent
407 * state, which can be "dti" for dyntick-idle mode or "kick" when kicking
408 * a CPU that has been in dyntick-idle mode for too long.
410 TRACE_EVENT_RCU(rcu_fqs,
412 TP_PROTO(const char *rcuname, unsigned long gp_seq, int cpu, const char *qsevent),
414 TP_ARGS(rcuname, gp_seq, cpu, qsevent),
417 __field(const char *, rcuname)
418 __field(unsigned long, gp_seq)
420 __field(const char *, qsevent)
424 __entry->rcuname = rcuname;
425 __entry->gp_seq = gp_seq;
427 __entry->qsevent = qsevent;
430 TP_printk("%s %lu %d %s",
431 __entry->rcuname, __entry->gp_seq,
432 __entry->cpu, __entry->qsevent)
435 #endif /* #if defined(CONFIG_TREE_RCU) */
438 * Tracepoint for dyntick-idle entry/exit events. These take a string
439 * as argument: "Start" for entering dyntick-idle mode, "Startirq" for
440 * entering it from irq/NMI, "End" for leaving it, "Endirq" for leaving it
441 * to irq/NMI, "--=" for events moving towards idle, and "++=" for events
442 * moving away from idle.
444 * These events also take a pair of numbers, which indicate the nesting
445 * depth before and after the event of interest, and a third number that is
446 * the ->dynticks counter. Note that task-related and interrupt-related
447 * events use two separate counters, and that the "++=" and "--=" events
448 * for irq/NMI will change the counter by two, otherwise by one.
450 TRACE_EVENT_RCU(rcu_dyntick,
452 TP_PROTO(const char *polarity, long oldnesting, long newnesting, int dynticks),
454 TP_ARGS(polarity, oldnesting, newnesting, dynticks),
457 __field(const char *, polarity)
458 __field(long, oldnesting)
459 __field(long, newnesting)
460 __field(int, dynticks)
464 __entry->polarity = polarity;
465 __entry->oldnesting = oldnesting;
466 __entry->newnesting = newnesting;
467 __entry->dynticks = dynticks;
470 TP_printk("%s %lx %lx %#3x", __entry->polarity,
471 __entry->oldnesting, __entry->newnesting,
472 __entry->dynticks & 0xfff)
476 * Tracepoint for the registration of a single RCU callback function.
477 * The first argument is the type of RCU, the second argument is
478 * a pointer to the RCU callback itself, the third element is the
479 * number of lazy callbacks queued, and the fourth element is the
480 * total number of callbacks queued.
482 TRACE_EVENT_RCU(rcu_callback,
484 TP_PROTO(const char *rcuname, struct rcu_head *rhp, long qlen),
486 TP_ARGS(rcuname, rhp, qlen),
489 __field(const char *, rcuname)
491 __field(void *, func)
496 __entry->rcuname = rcuname;
498 __entry->func = rhp->func;
499 __entry->qlen = qlen;
502 TP_printk("%s rhp=%p func=%ps %ld",
503 __entry->rcuname, __entry->rhp, __entry->func,
508 * Tracepoint for the registration of a single RCU callback of the special
509 * kfree() form. The first argument is the RCU type, the second argument
510 * is a pointer to the RCU callback, the third argument is the offset
511 * of the callback within the enclosing RCU-protected data structure,
512 * the fourth argument is the number of lazy callbacks queued, and the
513 * fifth argument is the total number of callbacks queued.
515 TRACE_EVENT_RCU(rcu_kfree_callback,
517 TP_PROTO(const char *rcuname, struct rcu_head *rhp, unsigned long offset,
520 TP_ARGS(rcuname, rhp, offset, qlen),
523 __field(const char *, rcuname)
525 __field(unsigned long, offset)
530 __entry->rcuname = rcuname;
532 __entry->offset = offset;
533 __entry->qlen = qlen;
536 TP_printk("%s rhp=%p func=%ld %ld",
537 __entry->rcuname, __entry->rhp, __entry->offset,
542 * Tracepoint for marking the beginning rcu_do_batch, performed to start
543 * RCU callback invocation. The first argument is the RCU flavor,
544 * the second is the number of lazy callbacks queued, the third is
545 * the total number of callbacks queued, and the fourth argument is
546 * the current RCU-callback batch limit.
548 TRACE_EVENT_RCU(rcu_batch_start,
550 TP_PROTO(const char *rcuname, long qlen, long blimit),
552 TP_ARGS(rcuname, qlen, blimit),
555 __field(const char *, rcuname)
557 __field(long, blimit)
561 __entry->rcuname = rcuname;
562 __entry->qlen = qlen;
563 __entry->blimit = blimit;
566 TP_printk("%s CBs=%ld bl=%ld",
567 __entry->rcuname, __entry->qlen, __entry->blimit)
571 * Tracepoint for the invocation of a single RCU callback function.
572 * The first argument is the type of RCU, and the second argument is
573 * a pointer to the RCU callback itself.
575 TRACE_EVENT_RCU(rcu_invoke_callback,
577 TP_PROTO(const char *rcuname, struct rcu_head *rhp),
579 TP_ARGS(rcuname, rhp),
582 __field(const char *, rcuname)
584 __field(void *, func)
588 __entry->rcuname = rcuname;
590 __entry->func = rhp->func;
593 TP_printk("%s rhp=%p func=%ps",
594 __entry->rcuname, __entry->rhp, __entry->func)
598 * Tracepoint for the invocation of a single RCU callback of the special
599 * kfree() form. The first argument is the RCU flavor, the second
600 * argument is a pointer to the RCU callback, and the third argument
601 * is the offset of the callback within the enclosing RCU-protected
604 TRACE_EVENT_RCU(rcu_invoke_kfree_callback,
606 TP_PROTO(const char *rcuname, struct rcu_head *rhp, unsigned long offset),
608 TP_ARGS(rcuname, rhp, offset),
611 __field(const char *, rcuname)
613 __field(unsigned long, offset)
617 __entry->rcuname = rcuname;
619 __entry->offset = offset;
622 TP_printk("%s rhp=%p func=%ld",
623 __entry->rcuname, __entry->rhp, __entry->offset)
627 * Tracepoint for the invocation of a single RCU callback of the special
628 * kfree_bulk() form. The first argument is the RCU flavor, the second
629 * argument is a number of elements in array to free, the third is an
630 * address of the array holding nr_records entries.
632 TRACE_EVENT_RCU(rcu_invoke_kfree_bulk_callback,
634 TP_PROTO(const char *rcuname, unsigned long nr_records, void **p),
636 TP_ARGS(rcuname, nr_records, p),
639 __field(const char *, rcuname)
640 __field(unsigned long, nr_records)
645 __entry->rcuname = rcuname;
646 __entry->nr_records = nr_records;
650 TP_printk("%s bulk=0x%p nr_records=%lu",
651 __entry->rcuname, __entry->p, __entry->nr_records)
655 * Tracepoint for exiting rcu_do_batch after RCU callbacks have been
656 * invoked. The first argument is the name of the RCU flavor,
657 * the second argument is number of callbacks actually invoked,
658 * the third argument (cb) is whether or not any of the callbacks that
659 * were ready to invoke at the beginning of this batch are still
660 * queued, the fourth argument (nr) is the return value of need_resched(),
661 * the fifth argument (iit) is 1 if the current task is the idle task,
662 * and the sixth argument (risk) is the return value from
663 * rcu_is_callbacks_kthread().
665 TRACE_EVENT_RCU(rcu_batch_end,
667 TP_PROTO(const char *rcuname, int callbacks_invoked,
668 char cb, char nr, char iit, char risk),
670 TP_ARGS(rcuname, callbacks_invoked, cb, nr, iit, risk),
673 __field(const char *, rcuname)
674 __field(int, callbacks_invoked)
682 __entry->rcuname = rcuname;
683 __entry->callbacks_invoked = callbacks_invoked;
687 __entry->risk = risk;
690 TP_printk("%s CBs-invoked=%d idle=%c%c%c%c",
691 __entry->rcuname, __entry->callbacks_invoked,
692 __entry->cb ? 'C' : '.',
693 __entry->nr ? 'S' : '.',
694 __entry->iit ? 'I' : '.',
695 __entry->risk ? 'R' : '.')
699 * Tracepoint for rcutorture readers. The first argument is the name
700 * of the RCU flavor from rcutorture's viewpoint and the second argument
701 * is the callback address. The third argument is the start time in
702 * seconds, and the last two arguments are the grace period numbers
703 * at the beginning and end of the read, respectively. Note that the
704 * callback address can be NULL.
706 #define RCUTORTURENAME_LEN 8
707 TRACE_EVENT_RCU(rcu_torture_read,
709 TP_PROTO(const char *rcutorturename, struct rcu_head *rhp,
710 unsigned long secs, unsigned long c_old, unsigned long c),
712 TP_ARGS(rcutorturename, rhp, secs, c_old, c),
715 __field(char, rcutorturename[RCUTORTURENAME_LEN])
716 __field(struct rcu_head *, rhp)
717 __field(unsigned long, secs)
718 __field(unsigned long, c_old)
719 __field(unsigned long, c)
723 strncpy(__entry->rcutorturename, rcutorturename,
725 __entry->rcutorturename[RCUTORTURENAME_LEN - 1] = 0;
727 __entry->secs = secs;
728 __entry->c_old = c_old;
732 TP_printk("%s torture read %p %luus c: %lu %lu",
733 __entry->rcutorturename, __entry->rhp,
734 __entry->secs, __entry->c_old, __entry->c)
738 * Tracepoint for rcu_barrier() execution. The string "s" describes
739 * the rcu_barrier phase:
740 * "Begin": rcu_barrier() started.
741 * "EarlyExit": rcu_barrier() piggybacked, thus early exit.
742 * "Inc1": rcu_barrier() piggyback check counter incremented.
743 * "OfflineNoCBQ": rcu_barrier() found offline no-CBs CPU with callbacks.
744 * "OnlineQ": rcu_barrier() found online CPU with callbacks.
745 * "OnlineNQ": rcu_barrier() found online CPU, no callbacks.
746 * "IRQ": An rcu_barrier_callback() callback posted on remote CPU.
747 * "IRQNQ": An rcu_barrier_callback() callback found no callbacks.
748 * "CB": An rcu_barrier_callback() invoked a callback, not the last.
749 * "LastCB": An rcu_barrier_callback() invoked the last callback.
750 * "Inc2": rcu_barrier() piggyback check counter incremented.
751 * The "cpu" argument is the CPU or -1 if meaningless, the "cnt" argument
752 * is the count of remaining callbacks, and "done" is the piggybacking count.
754 TRACE_EVENT_RCU(rcu_barrier,
756 TP_PROTO(const char *rcuname, const char *s, int cpu, int cnt, unsigned long done),
758 TP_ARGS(rcuname, s, cpu, cnt, done),
761 __field(const char *, rcuname)
762 __field(const char *, s)
765 __field(unsigned long, done)
769 __entry->rcuname = rcuname;
773 __entry->done = done;
776 TP_printk("%s %s cpu %d remaining %d # %lu",
777 __entry->rcuname, __entry->s, __entry->cpu, __entry->cnt,
781 #endif /* _TRACE_RCU_H */
783 /* This part must be outside protection */
784 #include <trace/define_trace.h>