2 #define TRACE_SYSTEM rcu
4 #if !defined(_TRACE_RCU_H) || defined(TRACE_HEADER_MULTI_READ)
7 #include <linux/tracepoint.h>
10 * Tracepoint for start/end markers used for utilization calculations.
11 * By convention, the string is of the following forms:
13 * "Start <activity>" -- Mark the start of the specified activity,
14 * such as "context switch". Nesting is permitted.
15 * "End <activity>" -- Mark the end of the specified activity.
17 * An "@" character within "<activity>" is a comment character: Data
18 * reduction scripts will ignore the "@" and the remainder of the line.
20 TRACE_EVENT(rcu_utilization,
22 TP_PROTO(const char *s),
27 __field(const char *, s)
34 TP_printk("%s", __entry->s)
37 #ifdef CONFIG_RCU_TRACE
39 #if defined(CONFIG_TREE_RCU) || defined(CONFIG_TREE_PREEMPT_RCU)
42 * Tracepoint for grace-period events. Takes a string identifying the
43 * RCU flavor, the grace-period number, and a string identifying the
44 * grace-period-related event as follows:
46 * "AccReadyCB": CPU acclerates new callbacks to RCU_NEXT_READY_TAIL.
47 * "AccWaitCB": CPU accelerates new callbacks to RCU_WAIT_TAIL.
48 * "newreq": Request a new grace period.
49 * "start": Start a grace period.
50 * "cpustart": CPU first notices a grace-period start.
51 * "cpuqs": CPU passes through a quiescent state.
52 * "cpuonl": CPU comes online.
53 * "cpuofl": CPU goes offline.
54 * "reqwait": GP kthread sleeps waiting for grace-period request.
55 * "reqwaitsig": GP kthread awakened by signal from reqwait state.
56 * "fqswait": GP kthread waiting until time to force quiescent states.
57 * "fqsstart": GP kthread starts forcing quiescent states.
58 * "fqsend": GP kthread done forcing quiescent states.
59 * "fqswaitsig": GP kthread awakened by signal from fqswait state.
60 * "end": End a grace period.
61 * "cpuend": CPU first notices a grace-period end.
63 TRACE_EVENT(rcu_grace_period,
65 TP_PROTO(const char *rcuname, unsigned long gpnum, const char *gpevent),
67 TP_ARGS(rcuname, gpnum, gpevent),
70 __field(const char *, rcuname)
71 __field(unsigned long, gpnum)
72 __field(const char *, gpevent)
76 __entry->rcuname = rcuname;
77 __entry->gpnum = gpnum;
78 __entry->gpevent = gpevent;
81 TP_printk("%s %lu %s",
82 __entry->rcuname, __entry->gpnum, __entry->gpevent)
86 * Tracepoint for future grace-period events, including those for no-callbacks
87 * CPUs. The caller should pull the data from the rcu_node structure,
88 * other than rcuname, which comes from the rcu_state structure, and event,
89 * which is one of the following:
91 * "Startleaf": Request a nocb grace period based on leaf-node data.
92 * "Startedleaf": Leaf-node start proved sufficient.
93 * "Startedleafroot": Leaf-node start proved sufficient after checking root.
94 * "Startedroot": Requested a nocb grace period based on root-node data.
95 * "StartWait": Start waiting for the requested grace period.
96 * "ResumeWait": Resume waiting after signal.
97 * "EndWait": Complete wait.
98 * "Cleanup": Clean up rcu_node structure after previous GP.
99 * "CleanupMore": Clean up, and another no-CB GP is needed.
101 TRACE_EVENT(rcu_future_grace_period,
103 TP_PROTO(const char *rcuname, unsigned long gpnum, unsigned long completed,
104 unsigned long c, u8 level, int grplo, int grphi,
105 const char *gpevent),
107 TP_ARGS(rcuname, gpnum, completed, c, level, grplo, grphi, gpevent),
110 __field(const char *, rcuname)
111 __field(unsigned long, gpnum)
112 __field(unsigned long, completed)
113 __field(unsigned long, c)
117 __field(const char *, gpevent)
121 __entry->rcuname = rcuname;
122 __entry->gpnum = gpnum;
123 __entry->completed = completed;
125 __entry->level = level;
126 __entry->grplo = grplo;
127 __entry->grphi = grphi;
128 __entry->gpevent = gpevent;
131 TP_printk("%s %lu %lu %lu %u %d %d %s",
132 __entry->rcuname, __entry->gpnum, __entry->completed,
133 __entry->c, __entry->level, __entry->grplo, __entry->grphi,
138 * Tracepoint for grace-period-initialization events. These are
139 * distinguished by the type of RCU, the new grace-period number, the
140 * rcu_node structure level, the starting and ending CPU covered by the
141 * rcu_node structure, and the mask of CPUs that will be waited for.
142 * All but the type of RCU are extracted from the rcu_node structure.
144 TRACE_EVENT(rcu_grace_period_init,
146 TP_PROTO(const char *rcuname, unsigned long gpnum, u8 level,
147 int grplo, int grphi, unsigned long qsmask),
149 TP_ARGS(rcuname, gpnum, level, grplo, grphi, qsmask),
152 __field(const char *, rcuname)
153 __field(unsigned long, gpnum)
157 __field(unsigned long, qsmask)
161 __entry->rcuname = rcuname;
162 __entry->gpnum = gpnum;
163 __entry->level = level;
164 __entry->grplo = grplo;
165 __entry->grphi = grphi;
166 __entry->qsmask = qsmask;
169 TP_printk("%s %lu %u %d %d %lx",
170 __entry->rcuname, __entry->gpnum, __entry->level,
171 __entry->grplo, __entry->grphi, __entry->qsmask)
175 * Tracepoint for RCU no-CBs CPU callback handoffs. This event is intended
176 * to assist debugging of these handoffs.
178 * The first argument is the name of the RCU flavor, and the second is
179 * the number of the offloaded CPU are extracted. The third and final
180 * argument is a string as follows:
182 * "WakeEmpty": Wake rcuo kthread, first CB to empty list.
183 * "WakeOvf": Wake rcuo kthread, CB list is huge.
184 * "WakeNot": Don't wake rcuo kthread.
185 * "WakeNotPoll": Don't wake rcuo kthread because it is polling.
186 * "WokeEmpty": rcuo kthread woke to find empty list.
187 * "WokeNonEmpty": rcuo kthread woke to find non-empty list.
189 TRACE_EVENT(rcu_nocb_wake,
191 TP_PROTO(const char *rcuname, int cpu, const char *reason),
193 TP_ARGS(rcuname, cpu, reason),
196 __field(const char *, rcuname)
198 __field(const char *, reason)
202 __entry->rcuname = rcuname;
204 __entry->reason = reason;
207 TP_printk("%s %d %s", __entry->rcuname, __entry->cpu, __entry->reason)
211 * Tracepoint for tasks blocking within preemptible-RCU read-side
212 * critical sections. Track the type of RCU (which one day might
213 * include SRCU), the grace-period number that the task is blocking
214 * (the current or the next), and the task's PID.
216 TRACE_EVENT(rcu_preempt_task,
218 TP_PROTO(const char *rcuname, int pid, unsigned long gpnum),
220 TP_ARGS(rcuname, pid, gpnum),
223 __field(const char *, rcuname)
224 __field(unsigned long, gpnum)
229 __entry->rcuname = rcuname;
230 __entry->gpnum = gpnum;
234 TP_printk("%s %lu %d",
235 __entry->rcuname, __entry->gpnum, __entry->pid)
239 * Tracepoint for tasks that blocked within a given preemptible-RCU
240 * read-side critical section exiting that critical section. Track the
241 * type of RCU (which one day might include SRCU) and the task's PID.
243 TRACE_EVENT(rcu_unlock_preempted_task,
245 TP_PROTO(const char *rcuname, unsigned long gpnum, int pid),
247 TP_ARGS(rcuname, gpnum, pid),
250 __field(const char *, rcuname)
251 __field(unsigned long, gpnum)
256 __entry->rcuname = rcuname;
257 __entry->gpnum = gpnum;
261 TP_printk("%s %lu %d", __entry->rcuname, __entry->gpnum, __entry->pid)
265 * Tracepoint for quiescent-state-reporting events. These are
266 * distinguished by the type of RCU, the grace-period number, the
267 * mask of quiescent lower-level entities, the rcu_node structure level,
268 * the starting and ending CPU covered by the rcu_node structure, and
269 * whether there are any blocked tasks blocking the current grace period.
270 * All but the type of RCU are extracted from the rcu_node structure.
272 TRACE_EVENT(rcu_quiescent_state_report,
274 TP_PROTO(const char *rcuname, unsigned long gpnum,
275 unsigned long mask, unsigned long qsmask,
276 u8 level, int grplo, int grphi, int gp_tasks),
278 TP_ARGS(rcuname, gpnum, mask, qsmask, level, grplo, grphi, gp_tasks),
281 __field(const char *, rcuname)
282 __field(unsigned long, gpnum)
283 __field(unsigned long, mask)
284 __field(unsigned long, qsmask)
288 __field(u8, gp_tasks)
292 __entry->rcuname = rcuname;
293 __entry->gpnum = gpnum;
294 __entry->mask = mask;
295 __entry->qsmask = qsmask;
296 __entry->level = level;
297 __entry->grplo = grplo;
298 __entry->grphi = grphi;
299 __entry->gp_tasks = gp_tasks;
302 TP_printk("%s %lu %lx>%lx %u %d %d %u",
303 __entry->rcuname, __entry->gpnum,
304 __entry->mask, __entry->qsmask, __entry->level,
305 __entry->grplo, __entry->grphi, __entry->gp_tasks)
309 * Tracepoint for quiescent states detected by force_quiescent_state().
310 * These trace events include the type of RCU, the grace-period number
311 * that was blocked by the CPU, the CPU itself, and the type of quiescent
312 * state, which can be "dti" for dyntick-idle mode, "ofl" for CPU offline,
313 * or "kick" when kicking a CPU that has been in dyntick-idle mode for
318 TP_PROTO(const char *rcuname, unsigned long gpnum, int cpu, const char *qsevent),
320 TP_ARGS(rcuname, gpnum, cpu, qsevent),
323 __field(const char *, rcuname)
324 __field(unsigned long, gpnum)
326 __field(const char *, qsevent)
330 __entry->rcuname = rcuname;
331 __entry->gpnum = gpnum;
333 __entry->qsevent = qsevent;
336 TP_printk("%s %lu %d %s",
337 __entry->rcuname, __entry->gpnum,
338 __entry->cpu, __entry->qsevent)
341 #endif /* #if defined(CONFIG_TREE_RCU) || defined(CONFIG_TREE_PREEMPT_RCU) */
344 * Tracepoint for dyntick-idle entry/exit events. These take a string
345 * as argument: "Start" for entering dyntick-idle mode, "End" for
346 * leaving it, "--=" for events moving towards idle, and "++=" for events
347 * moving away from idle. "Error on entry: not idle task" and "Error on
348 * exit: not idle task" indicate that a non-idle task is erroneously
349 * toying with the idle loop.
351 * These events also take a pair of numbers, which indicate the nesting
352 * depth before and after the event of interest. Note that task-related
353 * events use the upper bits of each number, while interrupt-related
354 * events use the lower bits.
356 TRACE_EVENT(rcu_dyntick,
358 TP_PROTO(const char *polarity, long long oldnesting, long long newnesting),
360 TP_ARGS(polarity, oldnesting, newnesting),
363 __field(const char *, polarity)
364 __field(long long, oldnesting)
365 __field(long long, newnesting)
369 __entry->polarity = polarity;
370 __entry->oldnesting = oldnesting;
371 __entry->newnesting = newnesting;
374 TP_printk("%s %llx %llx", __entry->polarity,
375 __entry->oldnesting, __entry->newnesting)
379 * Tracepoint for RCU preparation for idle, the goal being to get RCU
380 * processing done so that the current CPU can shut off its scheduling
381 * clock and enter dyntick-idle mode. One way to accomplish this is
382 * to drain all RCU callbacks from this CPU, and the other is to have
383 * done everything RCU requires for the current grace period. In this
384 * latter case, the CPU will be awakened at the end of the current grace
385 * period in order to process the remainder of its callbacks.
387 * These tracepoints take a string as argument:
389 * "No callbacks": Nothing to do, no callbacks on this CPU.
390 * "In holdoff": Nothing to do, holding off after unsuccessful attempt.
391 * "Begin holdoff": Attempt failed, don't retry until next jiffy.
392 * "Dyntick with callbacks": Entering dyntick-idle despite callbacks.
393 * "Dyntick with lazy callbacks": Entering dyntick-idle w/lazy callbacks.
394 * "More callbacks": Still more callbacks, try again to clear them out.
395 * "Callbacks drained": All callbacks processed, off to dyntick idle!
396 * "Timer": Timer fired to cause CPU to continue processing callbacks.
397 * "Demigrate": Timer fired on wrong CPU, woke up correct CPU.
398 * "Cleanup after idle": Idle exited, timer canceled.
400 TRACE_EVENT(rcu_prep_idle,
402 TP_PROTO(const char *reason),
407 __field(const char *, reason)
411 __entry->reason = reason;
414 TP_printk("%s", __entry->reason)
418 * Tracepoint for the registration of a single RCU callback function.
419 * The first argument is the type of RCU, the second argument is
420 * a pointer to the RCU callback itself, the third element is the
421 * number of lazy callbacks queued, and the fourth element is the
422 * total number of callbacks queued.
424 TRACE_EVENT(rcu_callback,
426 TP_PROTO(const char *rcuname, struct rcu_head *rhp, long qlen_lazy,
429 TP_ARGS(rcuname, rhp, qlen_lazy, qlen),
432 __field(const char *, rcuname)
434 __field(void *, func)
435 __field(long, qlen_lazy)
440 __entry->rcuname = rcuname;
442 __entry->func = rhp->func;
443 __entry->qlen_lazy = qlen_lazy;
444 __entry->qlen = qlen;
447 TP_printk("%s rhp=%p func=%pf %ld/%ld",
448 __entry->rcuname, __entry->rhp, __entry->func,
449 __entry->qlen_lazy, __entry->qlen)
453 * Tracepoint for the registration of a single RCU callback of the special
454 * kfree() form. The first argument is the RCU type, the second argument
455 * is a pointer to the RCU callback, the third argument is the offset
456 * of the callback within the enclosing RCU-protected data structure,
457 * the fourth argument is the number of lazy callbacks queued, and the
458 * fifth argument is the total number of callbacks queued.
460 TRACE_EVENT(rcu_kfree_callback,
462 TP_PROTO(const char *rcuname, struct rcu_head *rhp, unsigned long offset,
463 long qlen_lazy, long qlen),
465 TP_ARGS(rcuname, rhp, offset, qlen_lazy, qlen),
468 __field(const char *, rcuname)
470 __field(unsigned long, offset)
471 __field(long, qlen_lazy)
476 __entry->rcuname = rcuname;
478 __entry->offset = offset;
479 __entry->qlen_lazy = qlen_lazy;
480 __entry->qlen = qlen;
483 TP_printk("%s rhp=%p func=%ld %ld/%ld",
484 __entry->rcuname, __entry->rhp, __entry->offset,
485 __entry->qlen_lazy, __entry->qlen)
489 * Tracepoint for marking the beginning rcu_do_batch, performed to start
490 * RCU callback invocation. The first argument is the RCU flavor,
491 * the second is the number of lazy callbacks queued, the third is
492 * the total number of callbacks queued, and the fourth argument is
493 * the current RCU-callback batch limit.
495 TRACE_EVENT(rcu_batch_start,
497 TP_PROTO(const char *rcuname, long qlen_lazy, long qlen, long blimit),
499 TP_ARGS(rcuname, qlen_lazy, qlen, blimit),
502 __field(const char *, rcuname)
503 __field(long, qlen_lazy)
505 __field(long, blimit)
509 __entry->rcuname = rcuname;
510 __entry->qlen_lazy = qlen_lazy;
511 __entry->qlen = qlen;
512 __entry->blimit = blimit;
515 TP_printk("%s CBs=%ld/%ld bl=%ld",
516 __entry->rcuname, __entry->qlen_lazy, __entry->qlen,
521 * Tracepoint for the invocation of a single RCU callback function.
522 * The first argument is the type of RCU, and the second argument is
523 * a pointer to the RCU callback itself.
525 TRACE_EVENT(rcu_invoke_callback,
527 TP_PROTO(const char *rcuname, struct rcu_head *rhp),
529 TP_ARGS(rcuname, rhp),
532 __field(const char *, rcuname)
534 __field(void *, func)
538 __entry->rcuname = rcuname;
540 __entry->func = rhp->func;
543 TP_printk("%s rhp=%p func=%pf",
544 __entry->rcuname, __entry->rhp, __entry->func)
548 * Tracepoint for the invocation of a single RCU callback of the special
549 * kfree() form. The first argument is the RCU flavor, the second
550 * argument is a pointer to the RCU callback, and the third argument
551 * is the offset of the callback within the enclosing RCU-protected
554 TRACE_EVENT(rcu_invoke_kfree_callback,
556 TP_PROTO(const char *rcuname, struct rcu_head *rhp, unsigned long offset),
558 TP_ARGS(rcuname, rhp, offset),
561 __field(const char *, rcuname)
563 __field(unsigned long, offset)
567 __entry->rcuname = rcuname;
569 __entry->offset = offset;
572 TP_printk("%s rhp=%p func=%ld",
573 __entry->rcuname, __entry->rhp, __entry->offset)
577 * Tracepoint for exiting rcu_do_batch after RCU callbacks have been
578 * invoked. The first argument is the name of the RCU flavor,
579 * the second argument is number of callbacks actually invoked,
580 * the third argument (cb) is whether or not any of the callbacks that
581 * were ready to invoke at the beginning of this batch are still
582 * queued, the fourth argument (nr) is the return value of need_resched(),
583 * the fifth argument (iit) is 1 if the current task is the idle task,
584 * and the sixth argument (risk) is the return value from
585 * rcu_is_callbacks_kthread().
587 TRACE_EVENT(rcu_batch_end,
589 TP_PROTO(const char *rcuname, int callbacks_invoked,
590 bool cb, bool nr, bool iit, bool risk),
592 TP_ARGS(rcuname, callbacks_invoked, cb, nr, iit, risk),
595 __field(const char *, rcuname)
596 __field(int, callbacks_invoked)
604 __entry->rcuname = rcuname;
605 __entry->callbacks_invoked = callbacks_invoked;
609 __entry->risk = risk;
612 TP_printk("%s CBs-invoked=%d idle=%c%c%c%c",
613 __entry->rcuname, __entry->callbacks_invoked,
614 __entry->cb ? 'C' : '.',
615 __entry->nr ? 'S' : '.',
616 __entry->iit ? 'I' : '.',
617 __entry->risk ? 'R' : '.')
621 * Tracepoint for rcutorture readers. The first argument is the name
622 * of the RCU flavor from rcutorture's viewpoint and the second argument
623 * is the callback address.
625 TRACE_EVENT(rcu_torture_read,
627 TP_PROTO(const char *rcutorturename, struct rcu_head *rhp,
628 unsigned long secs, unsigned long c_old, unsigned long c),
630 TP_ARGS(rcutorturename, rhp, secs, c_old, c),
633 __field(const char *, rcutorturename)
634 __field(struct rcu_head *, rhp)
635 __field(unsigned long, secs)
636 __field(unsigned long, c_old)
637 __field(unsigned long, c)
641 __entry->rcutorturename = rcutorturename;
643 __entry->secs = secs;
644 __entry->c_old = c_old;
648 TP_printk("%s torture read %p %luus c: %lu %lu",
649 __entry->rcutorturename, __entry->rhp,
650 __entry->secs, __entry->c_old, __entry->c)
654 * Tracepoint for _rcu_barrier() execution. The string "s" describes
655 * the _rcu_barrier phase:
656 * "Begin": rcu_barrier_callback() started.
657 * "Check": rcu_barrier_callback() checking for piggybacking.
658 * "EarlyExit": rcu_barrier_callback() piggybacked, thus early exit.
659 * "Inc1": rcu_barrier_callback() piggyback check counter incremented.
660 * "Offline": rcu_barrier_callback() found offline CPU
661 * "OnlineNoCB": rcu_barrier_callback() found online no-CBs CPU.
662 * "OnlineQ": rcu_barrier_callback() found online CPU with callbacks.
663 * "OnlineNQ": rcu_barrier_callback() found online CPU, no callbacks.
664 * "IRQ": An rcu_barrier_callback() callback posted on remote CPU.
665 * "CB": An rcu_barrier_callback() invoked a callback, not the last.
666 * "LastCB": An rcu_barrier_callback() invoked the last callback.
667 * "Inc2": rcu_barrier_callback() piggyback check counter incremented.
668 * The "cpu" argument is the CPU or -1 if meaningless, the "cnt" argument
669 * is the count of remaining callbacks, and "done" is the piggybacking count.
671 TRACE_EVENT(rcu_barrier,
673 TP_PROTO(const char *rcuname, const char *s, int cpu, int cnt, unsigned long done),
675 TP_ARGS(rcuname, s, cpu, cnt, done),
678 __field(const char *, rcuname)
679 __field(const char *, s)
682 __field(unsigned long, done)
686 __entry->rcuname = rcuname;
690 __entry->done = done;
693 TP_printk("%s %s cpu %d remaining %d # %lu",
694 __entry->rcuname, __entry->s, __entry->cpu, __entry->cnt,
698 #else /* #ifdef CONFIG_RCU_TRACE */
700 #define trace_rcu_grace_period(rcuname, gpnum, gpevent) do { } while (0)
701 #define trace_rcu_grace_period_init(rcuname, gpnum, level, grplo, grphi, \
702 qsmask) do { } while (0)
703 #define trace_rcu_future_grace_period(rcuname, gpnum, completed, c, \
704 level, grplo, grphi, event) \
706 #define trace_rcu_nocb_wake(rcuname, cpu, reason) do { } while (0)
707 #define trace_rcu_preempt_task(rcuname, pid, gpnum) do { } while (0)
708 #define trace_rcu_unlock_preempted_task(rcuname, gpnum, pid) do { } while (0)
709 #define trace_rcu_quiescent_state_report(rcuname, gpnum, mask, qsmask, level, \
710 grplo, grphi, gp_tasks) do { } \
712 #define trace_rcu_fqs(rcuname, gpnum, cpu, qsevent) do { } while (0)
713 #define trace_rcu_dyntick(polarity, oldnesting, newnesting) do { } while (0)
714 #define trace_rcu_prep_idle(reason) do { } while (0)
715 #define trace_rcu_callback(rcuname, rhp, qlen_lazy, qlen) do { } while (0)
716 #define trace_rcu_kfree_callback(rcuname, rhp, offset, qlen_lazy, qlen) \
718 #define trace_rcu_batch_start(rcuname, qlen_lazy, qlen, blimit) \
720 #define trace_rcu_invoke_callback(rcuname, rhp) do { } while (0)
721 #define trace_rcu_invoke_kfree_callback(rcuname, rhp, offset) do { } while (0)
722 #define trace_rcu_batch_end(rcuname, callbacks_invoked, cb, nr, iit, risk) \
724 #define trace_rcu_torture_read(rcutorturename, rhp, secs, c_old, c) \
726 #define trace_rcu_barrier(name, s, cpu, cnt, done) do { } while (0)
728 #endif /* #else #ifdef CONFIG_RCU_TRACE */
730 #endif /* _TRACE_RCU_H */
732 /* This part must be outside protection */
733 #include <trace/define_trace.h>