1CONFIG_RCU_TRACE debugfs Files and Formats 2 3 4The rcutree and rcutiny implementations of RCU provide debugfs trace 5output that summarizes counters and state. This information is useful for 6debugging RCU itself, and can sometimes also help to debug abuses of RCU. 7The following sections describe the debugfs files and formats, first 8for rcutree and next for rcutiny. 9 10 11CONFIG_TREE_RCU and CONFIG_PREEMPT_RCU debugfs Files and Formats 12 13These implementations of RCU provide several debugfs directories under the 14top-level directory "rcu": 15 16rcu/rcu_bh 17rcu/rcu_preempt 18rcu/rcu_sched 19 20Each directory contains files for the corresponding flavor of RCU. 21Note that rcu/rcu_preempt is only present for CONFIG_PREEMPT_RCU. 22For CONFIG_TREE_RCU, the RCU flavor maps onto the RCU-sched flavor, 23so that activity for both appears in rcu/rcu_sched. 24 25In addition, the following file appears in the top-level directory: 26rcu/rcutorture. This file displays rcutorture test progress. The output 27of "cat rcu/rcutorture" looks as follows: 28 29rcutorture test sequence: 0 (test in progress) 30rcutorture update version number: 615 31 32The first line shows the number of rcutorture tests that have completed 33since boot. If a test is currently running, the "(test in progress)" 34string will appear as shown above. The second line shows the number of 35update cycles that the current test has started, or zero if there is 36no test in progress. 37 38 39Within each flavor directory (rcu/rcu_bh, rcu/rcu_sched, and possibly 40also rcu/rcu_preempt) the following files will be present: 41 42rcudata: 43 Displays fields in struct rcu_data. 44rcuexp: 45 Displays statistics for expedited grace periods. 46rcugp: 47 Displays grace-period counters. 48rcuhier: 49 Displays the struct rcu_node hierarchy. 50rcu_pending: 51 Displays counts of the reasons rcu_pending() decided that RCU had 52 work to do. 53rcuboost: 54 Displays RCU boosting statistics. Only present if 55 CONFIG_RCU_BOOST=y. 56 57The output of "cat rcu/rcu_preempt/rcudata" looks as follows: 58 59 0!c=30455 g=30456 pq=1/0 qp=1 dt=126535/140000000000000/0 df=2002 of=4 ql=0/0 qs=N... b=10 ci=74572 nci=0 co=1131 ca=716 60 1!c=30719 g=30720 pq=1/0 qp=0 dt=132007/140000000000000/0 df=1874 of=10 ql=0/0 qs=N... b=10 ci=123209 nci=0 co=685 ca=982 61 2!c=30150 g=30151 pq=1/1 qp=1 dt=138537/140000000000000/0 df=1707 of=8 ql=0/0 qs=N... b=10 ci=80132 nci=0 co=1328 ca=1458 62 3 c=31249 g=31250 pq=1/1 qp=0 dt=107255/140000000000000/0 df=1749 of=6 ql=0/450 qs=NRW. b=10 ci=151700 nci=0 co=509 ca=622 63 4!c=29502 g=29503 pq=1/0 qp=1 dt=83647/140000000000000/0 df=965 of=5 ql=0/0 qs=N... b=10 ci=65643 nci=0 co=1373 ca=1521 64 5 c=31201 g=31202 pq=1/0 qp=1 dt=70422/0/0 df=535 of=7 ql=0/0 qs=.... b=10 ci=58500 nci=0 co=764 ca=698 65 6!c=30253 g=30254 pq=1/0 qp=1 dt=95363/140000000000000/0 df=780 of=5 ql=0/0 qs=N... b=10 ci=100607 nci=0 co=1414 ca=1353 66 7 c=31178 g=31178 pq=1/0 qp=0 dt=91536/0/0 df=547 of=4 ql=0/0 qs=.... b=10 ci=109819 nci=0 co=1115 ca=969 67 68This file has one line per CPU, or eight for this 8-CPU system. 69The fields are as follows: 70 71o The number at the beginning of each line is the CPU number. 72 CPUs numbers followed by an exclamation mark are offline, 73 but have been online at least once since boot. There will be 74 no output for CPUs that have never been online, which can be 75 a good thing in the surprisingly common case where NR_CPUS is 76 substantially larger than the number of actual CPUs. 77 78o "c" is the count of grace periods that this CPU believes have 79 completed. Offlined CPUs and CPUs in dynticks idle mode may lag 80 quite a ways behind, for example, CPU 4 under "rcu_sched" above, 81 which has been offline through 16 RCU grace periods. It is not 82 unusual to see offline CPUs lagging by thousands of grace periods. 83 Note that although the grace-period number is an unsigned long, 84 it is printed out as a signed long to allow more human-friendly 85 representation near boot time. 86 87o "g" is the count of grace periods that this CPU believes have 88 started. Again, offlined CPUs and CPUs in dynticks idle mode 89 may lag behind. If the "c" and "g" values are equal, this CPU 90 has already reported a quiescent state for the last RCU grace 91 period that it is aware of, otherwise, the CPU believes that it 92 owes RCU a quiescent state. 93 94o "pq" indicates that this CPU has passed through a quiescent state 95 for the current grace period. It is possible for "pq" to be 96 "1" and "c" different than "g", which indicates that although 97 the CPU has passed through a quiescent state, either (1) this 98 CPU has not yet reported that fact, (2) some other CPU has not 99 yet reported for this grace period, or (3) both. 100 101o "qp" indicates that RCU still expects a quiescent state from 102 this CPU. Offlined CPUs and CPUs in dyntick idle mode might 103 well have qp=1, which is OK: RCU is still ignoring them. 104 105o "dt" is the current value of the dyntick counter that is incremented 106 when entering or leaving idle, either due to a context switch or 107 due to an interrupt. This number is even if the CPU is in idle 108 from RCU's viewpoint and odd otherwise. The number after the 109 first "/" is the interrupt nesting depth when in idle state, 110 or a large number added to the interrupt-nesting depth when 111 running a non-idle task. Some architectures do not accurately 112 count interrupt nesting when running in non-idle kernel context, 113 which can result in interesting anomalies such as negative 114 interrupt-nesting levels. The number after the second "/" 115 is the NMI nesting depth. 116 117o "df" is the number of times that some other CPU has forced a 118 quiescent state on behalf of this CPU due to this CPU being in 119 idle state. 120 121o "of" is the number of times that some other CPU has forced a 122 quiescent state on behalf of this CPU due to this CPU being 123 offline. In a perfect world, this might never happen, but it 124 turns out that offlining and onlining a CPU can take several grace 125 periods, and so there is likely to be an extended period of time 126 when RCU believes that the CPU is online when it really is not. 127 Please note that erring in the other direction (RCU believing a 128 CPU is offline when it is really alive and kicking) is a fatal 129 error, so it makes sense to err conservatively. 130 131o "ql" is the number of RCU callbacks currently residing on 132 this CPU. The first number is the number of "lazy" callbacks 133 that are known to RCU to only be freeing memory, and the number 134 after the "/" is the total number of callbacks, lazy or not. 135 These counters count callbacks regardless of what phase of 136 grace-period processing that they are in (new, waiting for 137 grace period to start, waiting for grace period to end, ready 138 to invoke). 139 140o "qs" gives an indication of the state of the callback queue 141 with four characters: 142 143 "N" Indicates that there are callbacks queued that are not 144 ready to be handled by the next grace period, and thus 145 will be handled by the grace period following the next 146 one. 147 148 "R" Indicates that there are callbacks queued that are 149 ready to be handled by the next grace period. 150 151 "W" Indicates that there are callbacks queued that are 152 waiting on the current grace period. 153 154 "D" Indicates that there are callbacks queued that have 155 already been handled by a prior grace period, and are 156 thus waiting to be invoked. Note that callbacks in 157 the process of being invoked are not counted here. 158 Callbacks in the process of being invoked are those 159 that have been removed from the rcu_data structures 160 queues by rcu_do_batch(), but which have not yet been 161 invoked. 162 163 If there are no callbacks in a given one of the above states, 164 the corresponding character is replaced by ".". 165 166o "b" is the batch limit for this CPU. If more than this number 167 of RCU callbacks is ready to invoke, then the remainder will 168 be deferred. 169 170o "ci" is the number of RCU callbacks that have been invoked for 171 this CPU. Note that ci+nci+ql is the number of callbacks that have 172 been registered in absence of CPU-hotplug activity. 173 174o "nci" is the number of RCU callbacks that have been offloaded from 175 this CPU. This will always be zero unless the kernel was built 176 with CONFIG_RCU_NOCB_CPU=y and the "rcu_nocbs=" kernel boot 177 parameter was specified. 178 179o "co" is the number of RCU callbacks that have been orphaned due to 180 this CPU going offline. These orphaned callbacks have been moved 181 to an arbitrarily chosen online CPU. 182 183o "ca" is the number of RCU callbacks that have been adopted by this 184 CPU due to other CPUs going offline. Note that ci+co-ca+ql is 185 the number of RCU callbacks registered on this CPU. 186 187 188Kernels compiled with CONFIG_RCU_BOOST=y display the following from 189/debug/rcu/rcu_preempt/rcudata: 190 191 0!c=12865 g=12866 pq=1/0 qp=1 dt=83113/140000000000000/0 df=288 of=11 ql=0/0 qs=N... kt=0/O ktl=944 b=10 ci=60709 nci=0 co=748 ca=871 192 1 c=14407 g=14408 pq=1/0 qp=0 dt=100679/140000000000000/0 df=378 of=7 ql=0/119 qs=NRW. kt=0/W ktl=9b6 b=10 ci=109740 nci=0 co=589 ca=485 193 2 c=14407 g=14408 pq=1/0 qp=0 dt=105486/0/0 df=90 of=9 ql=0/89 qs=NRW. kt=0/W ktl=c0c b=10 ci=83113 nci=0 co=533 ca=490 194 3 c=14407 g=14408 pq=1/0 qp=0 dt=107138/0/0 df=142 of=8 ql=0/188 qs=NRW. kt=0/W ktl=b96 b=10 ci=121114 nci=0 co=426 ca=290 195 4 c=14405 g=14406 pq=1/0 qp=1 dt=50238/0/0 df=706 of=7 ql=0/0 qs=.... kt=0/W ktl=812 b=10 ci=34929 nci=0 co=643 ca=114 196 5!c=14168 g=14169 pq=1/0 qp=0 dt=45465/140000000000000/0 df=161 of=11 ql=0/0 qs=N... kt=0/O ktl=b4d b=10 ci=47712 nci=0 co=677 ca=722 197 6 c=14404 g=14405 pq=1/0 qp=0 dt=59454/0/0 df=94 of=6 ql=0/0 qs=.... kt=0/W ktl=e57 b=10 ci=55597 nci=0 co=701 ca=811 198 7 c=14407 g=14408 pq=1/0 qp=1 dt=68850/0/0 df=31 of=8 ql=0/0 qs=.... kt=0/W ktl=14bd b=10 ci=77475 nci=0 co=508 ca=1042 199 200This is similar to the output discussed above, but contains the following 201additional fields: 202 203o "kt" is the per-CPU kernel-thread state. The digit preceding 204 the first slash is zero if there is no work pending and 1 205 otherwise. The character between the first pair of slashes is 206 as follows: 207 208 "S" The kernel thread is stopped, in other words, all 209 CPUs corresponding to this rcu_node structure are 210 offline. 211 212 "R" The kernel thread is running. 213 214 "W" The kernel thread is waiting because there is no work 215 for it to do. 216 217 "O" The kernel thread is waiting because it has been 218 forced off of its designated CPU or because its 219 ->cpus_allowed mask permits it to run on other than 220 its designated CPU. 221 222 "Y" The kernel thread is yielding to avoid hogging CPU. 223 224 "?" Unknown value, indicates a bug. 225 226 The number after the final slash is the CPU that the kthread 227 is actually running on. 228 229 This field is displayed only for CONFIG_RCU_BOOST kernels. 230 231o "ktl" is the low-order 16 bits (in hexadecimal) of the count of 232 the number of times that this CPU's per-CPU kthread has gone 233 through its loop servicing invoke_rcu_cpu_kthread() requests. 234 235 This field is displayed only for CONFIG_RCU_BOOST kernels. 236 237 238The output of "cat rcu/rcu_preempt/rcuexp" looks as follows: 239 240s=21872 d=21872 w=0 tf=0 wd1=0 wd2=0 n=0 sc=21872 dt=21872 dl=0 dx=21872 241 242These fields are as follows: 243 244o "s" is the starting sequence number. 245 246o "d" is the ending sequence number. When the starting and ending 247 numbers differ, there is an expedited grace period in progress. 248 249o "w" is the number of times that the sequence numbers have been 250 in danger of wrapping. 251 252o "tf" is the number of times that contention has resulted in a 253 failure to begin an expedited grace period. 254 255o "wd1" and "wd2" are the number of times that an attempt to 256 start an expedited grace period found that someone else had 257 completed an expedited grace period that satisfies the 258 attempted request. "Our work is done." 259 260o "n" is number of times that contention was so great that 261 the request was demoted from an expedited grace period to 262 a normal grace period. 263 264o "sc" is the number of times that the attempt to start a 265 new expedited grace period succeeded. 266 267o "dt" is the number of times that we attempted to update 268 the "d" counter. 269 270o "dl" is the number of times that we failed to update the "d" 271 counter. 272 273o "dx" is the number of times that we succeeded in updating 274 the "d" counter. 275 276 277The output of "cat rcu/rcu_preempt/rcugp" looks as follows: 278 279completed=31249 gpnum=31250 age=1 max=18 280 281These fields are taken from the rcu_state structure, and are as follows: 282 283o "completed" is the number of grace periods that have completed. 284 It is comparable to the "c" field from rcu/rcudata in that a 285 CPU whose "c" field matches the value of "completed" is aware 286 that the corresponding RCU grace period has completed. 287 288o "gpnum" is the number of grace periods that have started. It is 289 similarly comparable to the "g" field from rcu/rcudata in that 290 a CPU whose "g" field matches the value of "gpnum" is aware that 291 the corresponding RCU grace period has started. 292 293 If these two fields are equal, then there is no grace period 294 in progress, in other words, RCU is idle. On the other hand, 295 if the two fields differ (as they are above), then an RCU grace 296 period is in progress. 297 298o "age" is the number of jiffies that the current grace period 299 has extended for, or zero if there is no grace period currently 300 in effect. 301 302o "max" is the age in jiffies of the longest-duration grace period 303 thus far. 304 305The output of "cat rcu/rcu_preempt/rcuhier" looks as follows: 306 307c=14407 g=14408 s=0 jfq=2 j=c863 nfqs=12040/nfqsng=0(12040) fqlh=1051 oqlen=0/0 3083/3 ..>. 0:7 ^0 309e/e ..>. 0:3 ^0 d/d ..>. 4:7 ^1 310 311The fields are as follows: 312 313o "c" is exactly the same as "completed" under rcu/rcu_preempt/rcugp. 314 315o "g" is exactly the same as "gpnum" under rcu/rcu_preempt/rcugp. 316 317o "s" is the current state of the force_quiescent_state() 318 state machine. 319 320o "jfq" is the number of jiffies remaining for this grace period 321 before force_quiescent_state() is invoked to help push things 322 along. Note that CPUs in idle mode throughout the grace period 323 will not report on their own, but rather must be check by some 324 other CPU via force_quiescent_state(). 325 326o "j" is the low-order four hex digits of the jiffies counter. 327 Yes, Paul did run into a number of problems that turned out to 328 be due to the jiffies counter no longer counting. Why do you ask? 329 330o "nfqs" is the number of calls to force_quiescent_state() since 331 boot. 332 333o "nfqsng" is the number of useless calls to force_quiescent_state(), 334 where there wasn't actually a grace period active. This can 335 no longer happen due to grace-period processing being pushed 336 into a kthread. The number in parentheses is the difference 337 between "nfqs" and "nfqsng", or the number of times that 338 force_quiescent_state() actually did some real work. 339 340o "fqlh" is the number of calls to force_quiescent_state() that 341 exited immediately (without even being counted in nfqs above) 342 due to contention on ->fqslock. 343 344o Each element of the form "3/3 ..>. 0:7 ^0" represents one rcu_node 345 structure. Each line represents one level of the hierarchy, 346 from root to leaves. It is best to think of the rcu_data 347 structures as forming yet another level after the leaves. 348 Note that there might be either one, two, three, or even four 349 levels of rcu_node structures, depending on the relationship 350 between CONFIG_RCU_FANOUT, CONFIG_RCU_FANOUT_LEAF (possibly 351 adjusted using the rcu_fanout_leaf kernel boot parameter), and 352 CONFIG_NR_CPUS (possibly adjusted using the nr_cpu_ids count of 353 possible CPUs for the booting hardware). 354 355 o The numbers separated by the "/" are the qsmask followed 356 by the qsmaskinit. The qsmask will have one bit 357 set for each entity in the next lower level that has 358 not yet checked in for the current grace period ("e" 359 indicating CPUs 5, 6, and 7 in the example above). 360 The qsmaskinit will have one bit for each entity that is 361 currently expected to check in during each grace period. 362 The value of qsmaskinit is assigned to that of qsmask 363 at the beginning of each grace period. 364 365 o The characters separated by the ">" indicate the state 366 of the blocked-tasks lists. A "G" preceding the ">" 367 indicates that at least one task blocked in an RCU 368 read-side critical section blocks the current grace 369 period, while a "E" preceding the ">" indicates that 370 at least one task blocked in an RCU read-side critical 371 section blocks the current expedited grace period. 372 A "T" character following the ">" indicates that at 373 least one task is blocked within an RCU read-side 374 critical section, regardless of whether any current 375 grace period (expedited or normal) is inconvenienced. 376 A "." character appears if the corresponding condition 377 does not hold, so that "..>." indicates that no tasks 378 are blocked. In contrast, "GE>T" indicates maximal 379 inconvenience from blocked tasks. CONFIG_TREE_RCU 380 builds of the kernel will always show "..>.". 381 382 o The numbers separated by the ":" are the range of CPUs 383 served by this struct rcu_node. This can be helpful 384 in working out how the hierarchy is wired together. 385 386 For example, the example rcu_node structure shown above 387 has "0:7", indicating that it covers CPUs 0 through 7. 388 389 o The number after the "^" indicates the bit in the 390 next higher level rcu_node structure that this rcu_node 391 structure corresponds to. For example, the "d/d ..>. 4:7 392 ^1" has a "1" in this position, indicating that it 393 corresponds to the "1" bit in the "3" shown in the 394 "3/3 ..>. 0:7 ^0" entry on the next level up. 395 396 397The output of "cat rcu/rcu_sched/rcu_pending" looks as follows: 398 399 0!np=26111 qsp=29 rpq=5386 cbr=1 cng=570 gpc=3674 gps=577 nn=15903 ndw=0 400 1!np=28913 qsp=35 rpq=6097 cbr=1 cng=448 gpc=3700 gps=554 nn=18113 ndw=0 401 2!np=32740 qsp=37 rpq=6202 cbr=0 cng=476 gpc=4627 gps=546 nn=20889 ndw=0 402 3 np=23679 qsp=22 rpq=5044 cbr=1 cng=415 gpc=3403 gps=347 nn=14469 ndw=0 403 4!np=30714 qsp=4 rpq=5574 cbr=0 cng=528 gpc=3931 gps=639 nn=20042 ndw=0 404 5 np=28910 qsp=2 rpq=5246 cbr=0 cng=428 gpc=4105 gps=709 nn=18422 ndw=0 405 6!np=38648 qsp=5 rpq=7076 cbr=0 cng=840 gpc=4072 gps=961 nn=25699 ndw=0 406 7 np=37275 qsp=2 rpq=6873 cbr=0 cng=868 gpc=3416 gps=971 nn=25147 ndw=0 407 408The fields are as follows: 409 410o The leading number is the CPU number, with "!" indicating 411 an offline CPU. 412 413o "np" is the number of times that __rcu_pending() has been invoked 414 for the corresponding flavor of RCU. 415 416o "qsp" is the number of times that the RCU was waiting for a 417 quiescent state from this CPU. 418 419o "rpq" is the number of times that the CPU had passed through 420 a quiescent state, but not yet reported it to RCU. 421 422o "cbr" is the number of times that this CPU had RCU callbacks 423 that had passed through a grace period, and were thus ready 424 to be invoked. 425 426o "cng" is the number of times that this CPU needed another 427 grace period while RCU was idle. 428 429o "gpc" is the number of times that an old grace period had 430 completed, but this CPU was not yet aware of it. 431 432o "gps" is the number of times that a new grace period had started, 433 but this CPU was not yet aware of it. 434 435o "ndw" is the number of times that a wakeup of an rcuo 436 callback-offload kthread had to be deferred in order to avoid 437 deadlock. 438 439o "nn" is the number of times that this CPU needed nothing. 440 441 442The output of "cat rcu/rcuboost" looks as follows: 443 4440:3 tasks=.... kt=W ntb=0 neb=0 nnb=0 j=c864 bt=c894 445 balk: nt=0 egt=4695 bt=0 nb=0 ny=56 nos=0 4464:7 tasks=.... kt=W ntb=0 neb=0 nnb=0 j=c864 bt=c894 447 balk: nt=0 egt=6541 bt=0 nb=0 ny=126 nos=0 448 449This information is output only for rcu_preempt. Each two-line entry 450corresponds to a leaf rcu_node structure. The fields are as follows: 451 452o "n:m" is the CPU-number range for the corresponding two-line 453 entry. In the sample output above, the first entry covers 454 CPUs zero through three and the second entry covers CPUs four 455 through seven. 456 457o "tasks=TNEB" gives the state of the various segments of the 458 rnp->blocked_tasks list: 459 460 "T" This indicates that there are some tasks that blocked 461 while running on one of the corresponding CPUs while 462 in an RCU read-side critical section. 463 464 "N" This indicates that some of the blocked tasks are preventing 465 the current normal (non-expedited) grace period from 466 completing. 467 468 "E" This indicates that some of the blocked tasks are preventing 469 the current expedited grace period from completing. 470 471 "B" This indicates that some of the blocked tasks are in 472 need of RCU priority boosting. 473 474 Each character is replaced with "." if the corresponding 475 condition does not hold. 476 477o "kt" is the state of the RCU priority-boosting kernel 478 thread associated with the corresponding rcu_node structure. 479 The state can be one of the following: 480 481 "S" The kernel thread is stopped, in other words, all 482 CPUs corresponding to this rcu_node structure are 483 offline. 484 485 "R" The kernel thread is running. 486 487 "W" The kernel thread is waiting because there is no work 488 for it to do. 489 490 "Y" The kernel thread is yielding to avoid hogging CPU. 491 492 "?" Unknown value, indicates a bug. 493 494o "ntb" is the number of tasks boosted. 495 496o "neb" is the number of tasks boosted in order to complete an 497 expedited grace period. 498 499o "nnb" is the number of tasks boosted in order to complete a 500 normal (non-expedited) grace period. When boosting a task 501 that was blocking both an expedited and a normal grace period, 502 it is counted against the expedited total above. 503 504o "j" is the low-order 16 bits of the jiffies counter in 505 hexadecimal. 506 507o "bt" is the low-order 16 bits of the value that the jiffies 508 counter will have when we next start boosting, assuming that 509 the current grace period does not end beforehand. This is 510 also in hexadecimal. 511 512o "balk: nt" counts the number of times we didn't boost (in 513 other words, we balked) even though it was time to boost because 514 there were no blocked tasks to boost. This situation occurs 515 when there is one blocked task on one rcu_node structure and 516 none on some other rcu_node structure. 517 518o "egt" counts the number of times we balked because although 519 there were blocked tasks, none of them were blocking the 520 current grace period, whether expedited or otherwise. 521 522o "bt" counts the number of times we balked because boosting 523 had already been initiated for the current grace period. 524 525o "nb" counts the number of times we balked because there 526 was at least one task blocking the current non-expedited grace 527 period that never had blocked. If it is already running, it 528 just won't help to boost its priority! 529 530o "ny" counts the number of times we balked because it was 531 not yet time to start boosting. 532 533o "nos" counts the number of times we balked for other 534 reasons, e.g., the grace period ended first. 535 536 537CONFIG_TINY_RCU debugfs Files and Formats 538 539These implementations of RCU provides a single debugfs file under the 540top-level directory RCU, namely rcu/rcudata, which displays fields in 541rcu_bh_ctrlblk and rcu_sched_ctrlblk. 542 543The output of "cat rcu/rcudata" is as follows: 544 545rcu_sched: qlen: 0 546rcu_bh: qlen: 0 547 548This is split into rcu_sched and rcu_bh sections. The field is as 549follows: 550 551o "qlen" is the number of RCU callbacks currently waiting either 552 for an RCU grace period or waiting to be invoked. This is the 553 only field present for rcu_sched and rcu_bh, due to the 554 short-circuiting of grace period in those two cases. 555