diff options
28 files changed, 1218 insertions, 479 deletions
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt index 40a05df05580..4cd3ca5d09a8 100644 --- a/Documentation/admin-guide/kernel-parameters.txt +++ b/Documentation/admin-guide/kernel-parameters.txt @@ -3659,6 +3659,9 @@ just as if they had also been called out in the rcu_nocbs= boot parameter. + Note that this argument takes precedence over + the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. + noiotrap [SH] Disables trapped I/O port accesses. noirqdebug [X86-32] Disables the code which attempts to detect and @@ -4557,6 +4560,9 @@ no-callback mode from boot but the mode may be toggled at runtime via cpusets. + Note that this argument takes precedence over + the CONFIG_RCU_NOCB_CPU_DEFAULT_ALL option. + rcu_nocb_poll [KNL] Rather than requiring that offloaded CPUs (specified by rcu_nocbs= above) explicitly @@ -5799,6 +5805,24 @@ expediting. Set to zero to disable automatic expediting. + srcutree.srcu_max_nodelay [KNL] + Specifies the number of no-delay instances + per jiffy for which the SRCU grace period + worker thread will be rescheduled with zero + delay. Beyond this limit, worker thread will + be rescheduled with a sleep delay of one jiffy. + + srcutree.srcu_max_nodelay_phase [KNL] + Specifies the per-grace-period phase, number of + non-sleeping polls of readers. Beyond this limit, + grace period worker thread will be rescheduled + with a sleep delay of one jiffy, between each + rescan of the readers, for a grace period phase. + + srcutree.srcu_retry_check_delay [KNL] + Specifies number of microseconds of non-sleeping + delay between each non-sleeping poll of readers. + srcutree.small_contention_lim [KNL] Specifies the number of update-side contention events per jiffy will be tolerated before diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h index 1a32036c918c..937a58b3266b 100644 --- a/include/linux/rcupdate.h +++ b/include/linux/rcupdate.h @@ -41,6 +41,7 @@ void call_rcu(struct rcu_head *head, rcu_callback_t func); void rcu_barrier_tasks(void); void rcu_barrier_tasks_rude(void); void synchronize_rcu(void); +unsigned long get_completed_synchronize_rcu(void); #ifdef CONFIG_PREEMPT_RCU @@ -169,13 +170,24 @@ void synchronize_rcu_tasks(void); # endif # ifdef CONFIG_TASKS_TRACE_RCU -# define rcu_tasks_trace_qs(t) \ - do { \ - if (!likely(READ_ONCE((t)->trc_reader_checked)) && \ - !unlikely(READ_ONCE((t)->trc_reader_nesting))) { \ - smp_store_release(&(t)->trc_reader_checked, true); \ - smp_mb(); /* Readers partitioned by store. */ \ - } \ +// Bits for ->trc_reader_special.b.need_qs field. +#define TRC_NEED_QS 0x1 // Task needs a quiescent state. +#define TRC_NEED_QS_CHECKED 0x2 // Task has been checked for needing quiescent state. + +u8 rcu_trc_cmpxchg_need_qs(struct task_struct *t, u8 old, u8 new); +void rcu_tasks_trace_qs_blkd(struct task_struct *t); + +# define rcu_tasks_trace_qs(t) \ + do { \ + int ___rttq_nesting = READ_ONCE((t)->trc_reader_nesting); \ + \ + if (likely(!READ_ONCE((t)->trc_reader_special.b.need_qs)) && \ + likely(!___rttq_nesting)) { \ + rcu_trc_cmpxchg_need_qs((t), 0, TRC_NEED_QS_CHECKED); \ + } else if (___rttq_nesting && ___rttq_nesting != INT_MIN && \ + !READ_ONCE((t)->trc_reader_special.b.blocked)) { \ + rcu_tasks_trace_qs_blkd(t); \ + } \ } while (0) # else # define rcu_tasks_trace_qs(t) do { } while (0) @@ -184,7 +196,7 @@ void synchronize_rcu_tasks(void); #define rcu_tasks_qs(t, preempt) \ do { \ rcu_tasks_classic_qs((t), (preempt)); \ - rcu_tasks_trace_qs((t)); \ + rcu_tasks_trace_qs(t); \ } while (0) # ifdef CONFIG_TASKS_RUDE_RCU diff --git a/include/linux/rcupdate_trace.h b/include/linux/rcupdate_trace.h index 6f9c35817398..9bc8cbb33340 100644 --- a/include/linux/rcupdate_trace.h +++ b/include/linux/rcupdate_trace.h @@ -75,7 +75,7 @@ static inline void rcu_read_unlock_trace(void) nesting = READ_ONCE(t->trc_reader_nesting) - 1; barrier(); // Critical section before disabling. // Disable IPI-based setting of .need_qs. - WRITE_ONCE(t->trc_reader_nesting, INT_MIN); + WRITE_ONCE(t->trc_reader_nesting, INT_MIN + nesting); if (likely(!READ_ONCE(t->trc_reader_special.s)) || nesting) { WRITE_ONCE(t->trc_reader_nesting, nesting); return; // We assume shallow reader nesting. diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h index 5fed476f977f..e6bb31a0927b 100644 --- a/include/linux/rcutiny.h +++ b/include/linux/rcutiny.h @@ -23,6 +23,16 @@ static inline void cond_synchronize_rcu(unsigned long oldstate) might_sleep(); } +static inline unsigned long start_poll_synchronize_rcu_expedited(void) +{ + return start_poll_synchronize_rcu(); +} + +static inline void cond_synchronize_rcu_expedited(unsigned long oldstate) +{ + cond_synchronize_rcu(oldstate); +} + extern void rcu_barrier(void); static inline void synchronize_rcu_expedited(void) @@ -38,7 +48,7 @@ static inline void synchronize_rcu_expedited(void) */ extern void kvfree(const void *addr); -static inline void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func) +static inline void __kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func) { if (head) { call_rcu(head, func); @@ -51,6 +61,15 @@ static inline void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func) kvfree((void *) func); } +#ifdef CONFIG_KASAN_GENERIC +void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func); +#else +static inline void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func) +{ + __kvfree_call_rcu(head, func); +} +#endif + void rcu_qs(void); static inline void rcu_softirq_qs(void) diff --git a/include/linux/rcutree.h b/include/linux/rcutree.h index 9c6cfb742504..20dbaa9a3882 100644 --- a/include/linux/rcutree.h +++ b/include/linux/rcutree.h @@ -40,6 +40,8 @@ bool rcu_eqs_special_set(int cpu); void rcu_momentary_dyntick_idle(void); void kfree_rcu_scheduler_running(void); bool rcu_gp_might_be_stalled(void); +unsigned long start_poll_synchronize_rcu_expedited(void); +void cond_synchronize_rcu_expedited(unsigned long oldstate); unsigned long get_state_synchronize_rcu(void); unsigned long start_poll_synchronize_rcu(void); bool poll_state_synchronize_rcu(unsigned long oldstate); diff --git a/include/linux/sched.h b/include/linux/sched.h index c46f3a63b758..72242bc73d85 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -843,8 +843,9 @@ struct task_struct { int trc_reader_nesting; int trc_ipi_to_cpu; union rcu_special trc_reader_special; - bool trc_reader_checked; struct list_head trc_holdout_list; + struct list_head trc_blkd_node; + int trc_blkd_cpu; #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ struct sched_info sched_info; @@ -2223,6 +2224,7 @@ static inline void set_task_cpu(struct task_struct *p, unsigned int cpu) extern bool sched_task_on_rq(struct task_struct *p); extern unsigned long get_wchan(struct task_struct *p); +extern struct task_struct *cpu_curr_snapshot(int cpu); /* * In order to reduce various lock holder preemption latencies provide an diff --git a/init/init_task.c b/init/init_task.c index 73cc8f03511a..ff6c4b9bfe6b 100644 --- a/init/init_task.c +++ b/init/init_task.c @@ -157,6 +157,7 @@ struct task_struct init_task .trc_reader_nesting = 0, .trc_reader_special.s = 0, .trc_holdout_list = LIST_HEAD_INIT(init_task.trc_holdout_list), + .trc_blkd_node = LIST_HEAD_INIT(init_task.trc_blkd_node), #endif #ifdef CONFIG_CPUSETS .mems_allowed_seq = SEQCNT_SPINLOCK_ZERO(init_task.mems_allowed_seq, diff --git a/kernel/fork.c b/kernel/fork.c index 9d44f2d46c69..1950eb870244 100644 --- a/kernel/fork.c +++ b/kernel/fork.c @@ -1814,6 +1814,7 @@ static inline void rcu_copy_process(struct task_struct *p) p->trc_reader_nesting = 0; p->trc_reader_special.s = 0; INIT_LIST_HEAD(&p->trc_holdout_list); + INIT_LIST_HEAD(&p->trc_blkd_node); #endif /* #ifdef CONFIG_TASKS_TRACE_RCU */ } diff --git a/kernel/rcu/Kconfig b/kernel/rcu/Kconfig index 1c630e573548..c05ca52cdf64 100644 --- a/kernel/rcu/Kconfig +++ b/kernel/rcu/Kconfig @@ -262,6 +262,35 @@ config RCU_NOCB_CPU Say Y here if you need reduced OS jitter, despite added overhead. Say N here if you are unsure. +config RCU_NOCB_CPU_DEFAULT_ALL + bool "Offload RCU callback processing from all CPUs by default" + depends on RCU_NOCB_CPU + default n + help + Use this option to offload callback processing from all CPUs + by default, in the absence of the rcu_nocbs or nohz_full boot + parameter. This also avoids the need to use any boot parameters + to achieve the effect of offloading all CPUs on boot. + + Say Y here if you want offload all CPUs by default on boot. + Say N here if you are unsure. + +config RCU_NOCB_CPU_CB_BOOST + bool "Offload RCU callback from real-time kthread" + depends on RCU_NOCB_CPU && RCU_BOOST + default y if PREEMPT_RT + help + Use this option to invoke offloaded callbacks as SCHED_FIFO + to avoid starvation by heavy SCHED_OTHER background load. + Of course, running as SCHED_FIFO during callback floods will + cause the rcuo[ps] kthreads to monopolize the CPU for hundreds + of milliseconds or more. Therefore, when enabling this option, + it is your responsibility to ensure that latency-sensitive + tasks either run with higher priority or run on some other CPU. + + Say Y here if you want to set RT priority for offloading kthreads. + Say N here if you are building a !PREEMPT_RT kernel and are unsure. + config TASKS_TRACE_RCU_READ_MB bool "Tasks Trace RCU readers use memory barriers in user and idle" depends on RCU_EXPERT && TASKS_TRACE_RCU diff --git a/kernel/rcu/Kconfig.debug b/kernel/rcu/Kconfig.debug index 9b64e55d4f61..4da05beb13d7 100644 --- a/kernel/rcu/Kconfig.debug +++ b/kernel/rcu/Kconfig.debug @@ -121,7 +121,7 @@ config RCU_EQS_DEBUG config RCU_STRICT_GRACE_PERIOD bool "Provide debug RCU implementation with short grace periods" - depends on DEBUG_KERNEL && RCU_EXPERT && NR_CPUS <= 4 + depends on DEBUG_KERNEL && RCU_EXPERT && NR_CPUS <= 4 && !TINY_RCU default n select PREEMPT_COUNT if PREEMPT=n help diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h index 4916077119f3..32291f4eefde 100644 --- a/kernel/rcu/rcu.h +++ b/kernel/rcu/rcu.h @@ -23,6 +23,9 @@ #define RCU_SEQ_CTR_SHIFT 2 #define RCU_SEQ_STATE_MASK ((1 << RCU_SEQ_CTR_SHIFT) - 1) +/* Low-order bit definition for polled grace-period APIs. */ +#define RCU_GET_STATE_COMPLETED 0x1 + extern int sysctl_sched_rt_runtime; /* @@ -120,6 +123,18 @@ static inline bool rcu_seq_done(unsigned long *sp, unsigned long s) } /* + * Given a snapshot from rcu_seq_snap(), determine whether or not a + * full update-side operation has occurred, but do not allow the + * (ULONG_MAX / 2) safety-factor/guard-band. + */ +static inline bool rcu_seq_done_exact(unsigned long *sp, unsigned long s) +{ + unsigned long cur_s = READ_ONCE(*sp); + + return ULONG_CMP_GE(cur_s, s) || ULONG_CMP_LT(cur_s, s - (2 * RCU_SEQ_STATE_MASK + 1)); +} + +/* * Has a grace period completed since the time the old gp_seq was collected? */ static inline bool rcu_seq_completed_gp(unsigned long old, unsigned long new) diff --git a/kernel/rcu/rcuscale.c b/kernel/rcu/rcuscale.c index 277a5bfb37d4..3ef02d4a8108 100644 --- a/kernel/rcu/rcuscale.c +++ b/kernel/rcu/rcuscale.c @@ -419,6 +419,7 @@ rcu_scale_writer(void *arg) VERBOSE_SCALEOUT_STRING("rcu_scale_writer task started"); WARN_ON(!wdpp); set_cpus_allowed_ptr(current, cpumask_of(me % nr_cpu_ids)); + current->flags |= PF_NO_SETAFFINITY; sched_set_fifo_low(current); if (holdoff) diff --git a/kernel/rcu/rcutorture.c b/kernel/rcu/rcutorture.c index 6f47d1490c4f..d8e1b270a065 100644 --- a/kernel/rcu/rcutorture.c +++ b/kernel/rcu/rcutorture.c @@ -83,9 +83,11 @@ torture_param(int, fwd_progress_div, 4, "Fraction of CPU stall to wait"); torture_param(int, fwd_progress_holdoff, 60, "Time between forward-progress tests (s)"); torture_param(bool, fwd_progress_need_resched, 1, "Hide cond_resched() behind need_resched()"); torture_param(bool, gp_cond, false, "Use conditional/async GP wait primitives"); +torture_param(bool, gp_cond_exp, false, "Use conditional/async expedited GP wait primitives"); torture_param(bool, gp_exp, false, "Use expedited GP wait primitives"); torture_param(bool, gp_normal, false, "Use normal (non-expedited) GP wait primitives"); torture_param(bool, gp_poll, false, "Use polling GP wait primitives"); +torture_param(bool, gp_poll_exp, false, "Use polling expedited GP wait primitives"); torture_param(bool, gp_sync, false, "Use synchronous GP wait primitives"); torture_param(int, irqreader, 1, "Allow RCU readers from irq handlers"); torture_param(int, leakpointer, 0, "Leak pointer dereferences from readers"); @@ -192,12 +194,16 @@ static int rcu_torture_writer_state; #define RTWS_DEF_FREE 3 #define RTWS_EXP_SYNC 4 #define RTWS_COND_GET 5 -#define RTWS_COND_SYNC 6 -#define RTWS_POLL_GET 7 -#define RTWS_POLL_WAIT 8 -#define RTWS_SYNC 9 -#define RTWS_STUTTER 10 -#define RTWS_STOPPING 11 +#define RTWS_COND_GET_EXP 6 +#define RTWS_COND_SYNC 7 +#define RTWS_COND_SYNC_EXP 8 +#define RTWS_POLL_GET 9 +#define RTWS_POLL_GET_EXP 10 +#define RTWS_POLL_WAIT 11 +#define RTWS_POLL_WAIT_EXP 12 +#define RTWS_SYNC 13 +#define RTWS_STUTTER 14 +#define RTWS_STOPPING 15 static const char * const rcu_torture_writer_state_names[] = { "RTWS_FIXED_DELAY", "RTWS_DELAY", @@ -205,9 +211,13 @@ static const char * const rcu_torture_writer_state_names[] = { "RTWS_DEF_FREE", "RTWS_EXP_SYNC", "RTWS_COND_GET", + "RTWS_COND_GET_EXP", "RTWS_COND_SYNC", + "RTWS_COND_SYNC_EXP", "RTWS_POLL_GET", + "RTWS_POLL_GET_EXP", "RTWS_POLL_WAIT", + "RTWS_POLL_WAIT_EXP", "RTWS_SYNC", "RTWS_STUTTER", "RTWS_STOPPING", @@ -320,7 +330,12 @@ struct rcu_torture_ops { void (*deferred_free)(struct rcu_torture *p); void (*sync)(void); void (*exp_sync)(void); + unsigned long (*get_gp_state_exp)(void); + unsigned long (*start_gp_poll_exp)(void); + bool (*poll_gp_state_exp)(unsigned long oldstate); + void (*cond_sync_exp)(unsigned long oldstate); unsigned long (*get_gp_state)(void); + unsigned long (*get_gp_completed)(void); unsigned long (*start_gp_poll)(void); bool (*poll_gp_state)(unsigned long oldstate); void (*cond_sync)(unsigned long oldstate); @@ -487,9 +502,14 @@ static struct rcu_torture_ops rcu_ops = { .sync = synchronize_rcu, .exp_sync = synchronize_rcu_expedited, .get_gp_state = get_state_synchronize_rcu, + .get_gp_completed = get_completed_synchronize_rcu, .start_gp_poll = start_poll_synchronize_rcu, .poll_gp_state = poll_state_synchronize_rcu, .cond_sync = cond_synchronize_rcu, + .get_gp_state_exp = get_state_synchronize_rcu, + .start_gp_poll_exp = start_poll_synchronize_rcu_expedited, + .poll_gp_state_exp = poll_state_synchronize_rcu, + .cond_sync_exp = cond_synchronize_rcu_expedited, .call = call_rcu, .cb_barrier = rcu_barrier, .fqs = rcu_force_quiescent_state, @@ -1119,9 +1139,8 @@ rcu_torture_fqs(void *arg) return 0; } -// Used by writers to randomly choose from the available grace-period -// primitives. The only purpose of the initialization is to size the array. -static int synctype[] = { RTWS_DEF_FREE, RTWS_EXP_SYNC, RTWS_COND_GET, RTWS_POLL_GET, RTWS_SYNC }; +// Used by writers to randomly choose from the available grace-period primitives. +static int synctype[ARRAY_SIZE(rcu_torture_writer_state_names)] = { }; static int nsynctypes; /* @@ -1129,18 +1148,27 @@ static int nsynctypes; */ static void rcu_torture_write_types(void) { - bool gp_cond1 = gp_cond, gp_exp1 = gp_exp, gp_normal1 = gp_normal; - bool gp_poll1 = gp_poll, gp_sync1 = gp_sync; + bool gp_cond1 = gp_cond, gp_cond_exp1 = gp_cond_exp, gp_exp1 = gp_exp; + bool gp_poll_exp1 = gp_poll_exp, gp_normal1 = gp_normal, gp_poll1 = gp_poll; + bool gp_sync1 = gp_sync; /* Initialize synctype[] array. If none set, take default. */ - if (!gp_cond1 && !gp_exp1 && !gp_normal1 && !gp_poll1 && !gp_sync1) - gp_cond1 = gp_exp1 = gp_normal1 = gp_poll1 = gp_sync1 = true; + if (!gp_cond1 && !gp_cond_exp1 && !gp_exp1 && !gp_poll_exp && + !gp_normal1 && !gp_poll1 && !gp_sync1) + gp_cond1 = gp_cond_exp1 = gp_exp1 = gp_poll_exp1 = + gp_normal1 = gp_poll1 = gp_sync1 = true; if (gp_cond1 && cur_ops->get_gp_state && cur_ops->cond_sync) { synctype[nsynctypes++] = RTWS_COND_GET; pr_info("%s: Testing conditional GPs.\n", __func__); } else if (gp_cond && (!cur_ops->get_gp_state || !cur_ops->cond_sync)) { pr_alert("%s: gp_cond without primitives.\n", __func__); } + if (gp_cond_exp1 && cur_ops->get_gp_state_exp && cur_ops->cond_sync_exp) { + synctype[nsynctypes++] = RTWS_COND_GET_EXP; + pr_info("%s: Testing conditional expedited GPs.\n", __func__); + } else if (gp_cond_exp && (!cur_ops->get_gp_state_exp || !cur_ops->cond_sync_exp)) { + pr_alert("%s: gp_cond_exp without primitives.\n", __func__); + } if (gp_exp1 && cur_ops->exp_sync) { synctype[nsynctypes++] = RTWS_EXP_SYNC; pr_info("%s: Testing expedited GPs.\n", __func__); @@ -1159,6 +1187,12 @@ static void rcu_torture_write_types(void) } else if (gp_poll && (!cur_ops->start_gp_poll || !cur_ops->poll_gp_state)) { pr_alert("%s: gp_poll without primitives.\n", __func__); } + if (gp_poll_exp1 && cur_ops->start_gp_poll_exp && cur_ops->poll_gp_state_exp) { + synctype[nsynctypes++] = RTWS_POLL_GET_EXP; + pr_info("%s: Testing polling expedited GPs.\n", __func__); + } else if (gp_poll_exp && (!cur_ops->start_gp_poll_exp || !cur_ops->poll_gp_state_exp)) { + pr_alert("%s: gp_poll_exp without primitives.\n", __func__); + } if (gp_sync1 && cur_ops->sync) { synctype[nsynctypes++] = RTWS_SYNC; pr_info("%s: Testing normal GPs.\n", __func__); @@ -1237,6 +1271,10 @@ rcu_torture_writer(void *arg) rcu_torture_writer_state_getname(), rcu_torture_writer_state, cookie, cur_ops->get_gp_state()); + if (cur_ops->get_gp_completed) { + cookie = cur_ops->get_gp_completed(); + WARN_ON_ONCE(!cur_ops->poll_gp_state(cookie)); + } cur_ops->readunlock(idx); } switch (synctype[torture_random(&rand) % nsynctypes]) { @@ -1246,7 +1284,12 @@ rcu_torture_writer(void *arg) break; case RTWS_EXP_SYNC: rcu_torture_writer_state = RTWS_EXP_SYNC; + if (cur_ops->get_gp_state && cur_ops->poll_gp_state) + cookie = cur_ops->get_gp_state(); cur_ops->exp_sync(); + cur_ops->exp_sync(); + if (cur_ops->get_gp_state && cur_ops->poll_gp_state) + WARN_ON_ONCE(!cur_ops->poll_gp_state(cookie)); rcu_torture_pipe_update(old_rp); break; case RTWS_COND_GET: @@ -1257,6 +1300,14 @@ rcu_torture_writer(void *arg) cur_ops->cond_sync(gp_snap); rcu_torture_pipe_update(old_rp); break; + case RTWS_COND_GET_EXP: + rcu_torture_writer_state = RTWS_COND_GET_EXP; + gp_snap = cur_ops->get_gp_state_exp(); + torture_hrtimeout_jiffies(torture_random(&rand) % 16, &rand); + rcu_torture_writer_state = RTWS_COND_SYNC_EXP; + cur_ops->cond_sync_exp(gp_snap); + rcu_torture_pipe_update(old_rp); + break; case RTWS_POLL_GET: rcu_torture_writer_state = RTWS_POLL_GET; gp_snap = cur_ops->start_gp_poll(); @@ -1266,9 +1317,23 @@ rcu_torture_writer(void *arg) &rand); rcu_torture_pipe_update(old_rp); break; + case RTWS_POLL_GET_EXP: + rcu_torture_writer_state = RTWS_POLL_GET_EXP; + gp_snap = cur_ops->start_gp_poll_exp(); + rcu_torture_writer_state = RTWS_POLL_WAIT_EXP; + while (!cur_ops->poll_gp_state_exp(gp_snap)) + torture_hrtimeout_jiffies(torture_random(&rand) % 16, + &rand); + rcu_torture_pipe_update(old_rp); + break; case RTWS_SYNC: rcu_torture_writer_state = RTWS_SYNC; + if (cur_ops->get_gp_state && cur_ops->poll_gp_state) + cookie = cur_ops->get_gp_state(); cur_ops->sync(); + cur_ops->sync(); + if (cur_ops->get_gp_state && cur_ops->poll_gp_state) + WARN_ON_ONCE(!cur_ops->poll_gp_state(cookie)); rcu_torture_pipe_update(old_rp); break; default: @@ -1304,8 +1369,9 @@ rcu_torture_writer(void *arg) if (list_empty(&rcu_tortures[i].rtort_free) && rcu_access_pointer(rcu_torture_current) != &rcu_tortures[i]) { - rcu_ftrace_dump(DUMP_ALL); + tracing_off(); WARN(1, "%s: rtort_pipe_count: %d\n", __func__, rcu_tortures[i].rtort_pipe_count); + rcu_ftrace_dump(DUMP_ALL); } if (stutter_waited) sched_set_normal(current, oldnice); @@ -1367,6 +1433,11 @@ rcu_torture_fakewriter(void *arg) torture_hrtimeout_jiffies(torture_random(&rand) % 16, &rand); cur_ops->cond_sync(gp_snap); break; + case RTWS_COND_GET_EXP: + gp_snap = cur_ops->get_gp_state_exp(); + torture_hrtimeout_jiffies(torture_random(&rand) % 16, &rand); + cur_ops->cond_sync_exp(gp_snap); + break; case RTWS_POLL_GET: gp_snap = cur_ops->start_gp_poll(); while (!cur_ops->poll_gp_state(gp_snap)) { @@ -1374,6 +1445,13 @@ rcu_torture_fakewriter(void *arg) &rand); } break; + case RTWS_POLL_GET_EXP: + gp_snap = cur_ops->start_gp_poll_exp(); + while (!cur_ops->poll_gp_state_exp(gp_snap)) { + torture_hrtimeout_jiffies(torture_random(&rand) % 16, + &rand); + } + break; case RTWS_SYNC: cur_ops->sync(); break; @@ -1851,7 +1929,7 @@ rcu_torture_stats_print(void) batchsummary[i] += READ_ONCE(per_cpu(rcu_torture_batch, cpu)[i]); } } - for (i = RCU_TORTURE_PIPE_LEN - 1; i >= 0; i--) { + for (i = RCU_TORTURE_PIPE_LEN; i >= 0; i--) { if (pipesummary[i] != 0) break; } @@ -1973,7 +2051,13 @@ static void rcu_torture_mem_dump_obj(void) static int z; kcp = kmem_cache_create("rcuscale", 136, 8, SLAB_STORE_USER, NULL); + if (WARN_ON_ONCE(!kcp)) + return; rhp = kmem_cache_alloc(kcp, GFP_KERNEL); + if (WARN_ON_ONCE(!rhp)) { + kmem_cache_destroy(kcp); + return; + } pr_alert("mem_dump_obj() slab test: rcu_torture_stats = %px, &rhp = %px, rhp = %px, &z = %px\n", stats_task, &rhp, rhp, &z); pr_alert("mem_dump_obj(ZERO_SIZE_PTR):"); mem_dump_obj(ZERO_SIZE_PTR); @@ -1990,6 +2074,8 @@ static void rcu_torture_mem_dump_obj(void) kmem_cache_free(kcp, rhp); kmem_cache_destroy(kcp); rhp = kmalloc(sizeof(*rhp), GFP_KERNEL); + if (WARN_ON_ONCE(!rhp)) + return; pr_alert("mem_dump_obj() kmalloc test: rcu_torture_stats = %px, &rhp = %px, rhp = %px\n", stats_task, &rhp, rhp); pr_alert("mem_dump_obj(kmalloc %px):", rhp); mem_dump_obj(rhp); @@ -1997,6 +2083,8 @@ static void rcu_torture_mem_dump_obj(void) mem_dump_obj(&rhp->func); kfree(rhp); rhp = vmalloc(4096); + if (WARN_ON_ONCE(!rhp)) + return; pr_alert("mem_dump_obj() vmalloc test: rcu_torture_stats = %px, &rhp = %px, rhp = %px\n", stats_task, &rhp, rhp); pr_alert("mem_dump_obj(vmalloc %px):", rhp); mem_dump_obj(rhp); @@ -2058,6 +2146,19 @@ static int rcutorture_booster_init(unsigned int cpu) if (boost_tasks[cpu] != NULL) return 0; /* Already created, nothing more to do. */ + // Testing RCU priority boosting requires rcutorture do + // some serious abuse. Counter this by running ksoftirqd + // at higher priority. + if (IS_BUILTIN(CONFIG_RCU_TORTURE_TEST)) { + struct sched_param sp; + struct task_struct *t; + + t = per_cpu(ksoftirqd, cpu); + WARN_ON_ONCE(!t); + sp.sched_priority = 2; + sched_setscheduler_nocheck(t, SCHED_FIFO, &sp); + } + /* Don't allow time recalculation while creating a new task. */ mutex_lock(&boost_mutex); rcu_torture_disable_rt_throttle(); @@ -2856,7 +2957,6 @@ static int rcu_torture_read_exit_child(void *trsp_in) // Parent kthread which creates and destroys read-exit child kthreads. static int rcu_torture_read_exit(void *unused) { - int count = 0; bool errexit = false; int i; struct task_struct *tsp; @@ -2868,34 +2968,28 @@ static int rcu_torture_read_exit(void *unused) // Each pass through this loop does one read-exit episode. do { - if (++count > read_exit_burst) { - VERBOSE_TOROUT_STRING("rcu_torture_read_exit: End of episode"); - rcu_barrier(); // Wait for task_struct free, avoid OOM. - for (i = 0; i < read_exit_delay; i++) { - schedule_timeout_uninterruptible(HZ); - if (READ_ONCE(read_exit_child_stop)) - break; + VERBOSE_TOROUT_STRING("rcu_torture_read_exit: Start of episode"); + for (i = 0; i < read_exit_burst; i++) { + if (READ_ONCE(read_exit_child_stop)) + break; + stutter_wait("rcu_torture_read_exit"); + // Spawn child. + tsp = kthread_run(rcu_torture_read_exit_child, + &trs, "%s", "rcu_torture_read_exit_child"); + if (IS_ERR(tsp)) { + TOROUT_ERRSTRING("out of memory"); + errexit = true; + break; } - if (!READ_ONCE(read_exit_child_stop)) - VERBOSE_TOROUT_STRING("rcu_torture_read_exit: Start of episode"); - count = 0; - } - if (READ_ONCE(read_exit_child_stop)) - break; - // Spawn child. - tsp = kthread_run(rcu_torture_read_exit_child, - &trs, "%s", - "rcu_torture_read_exit_child"); - if (IS_ERR(tsp)) { - TOROUT_ERRSTRING("out of memory"); - errexit = true; - tsp = NULL; - break; + cond_resched(); + kthread_stop(tsp); + n_read_exits++; } - cond_resched(); - kthread_stop(tsp); - n_read_exits ++; - stutter_wait("rcu_torture_read_exit"); + VERBOSE_TOROUT_STRING("rcu_torture_read_exit: End of episode"); + rcu_barrier(); // Wait for task_struct free, avoid OOM. + i = 0; + for (; !errexit && !READ_ONCE(read_exit_child_stop) && i < read_exit_delay; i++) + schedule_timeout_uninterruptible(HZ); } while (!errexit && !READ_ONCE(read_exit_child_stop)); // Clean up and exit. @@ -3105,6 +3199,7 @@ static void rcu_test_debug_objects(void) pr_alert("%s: WARN: Duplicate call_rcu() test complete.\n", KBUILD_MODNAME); destroy_rcu_head_on_stack(&rh1); destroy_rcu_head_on_stack(&rh2); + kfree(rhp); #else /* #ifdef CONFIG_DEBUG_OBJECTS_RCU_HEAD */ pr_alert("%s: !CONFIG_DEBUG_OBJECTS_RCU_HEAD, not testing duplicate call_rcu()\n", KBUILD_MODNAME); #endif /* #else #ifdef CONFIG_DEBUG_OBJECTS_RCU_HEAD */ @@ -3312,21 +3407,6 @@ rcu_torture_init(void) rcutor_hp = firsterr; if (torture_init_error(firsterr)) goto unwind; - - // Testing RCU priority boosting requires rcutorture do - // some serious abuse. Counter this by running ksoftirqd - // at higher priority. - if (IS_BUILTIN(CONFIG_RCU_TORTURE_TEST)) { - for_each_online_cpu(cpu) { - struct sched_param sp; - struct task_struct *t; - - t = per_cpu(ksoftirqd, cpu); - WARN_ON_ONCE(!t); - sp.sched_priority = 2; - sched_setscheduler_nocheck(t, SCHED_FIFO, &sp); - } - } } shutdown_jiffies = jiffies + shutdown_secs * HZ; firsterr = torture_shutdown_init(shutdown_secs, rcu_torture_cleanup); diff --git a/kernel/rcu/refscale.c b/kernel/rcu/refscale.c index 909644abee67..435c884c02b5 100644 --- a/kernel/rcu/refscale.c +++ b/kernel/rcu/refscale.c @@ -385,7 +385,7 @@ static struct ref_scale_ops rwsem_ops = { }; // Definitions for global spinlock -static DEFINE_SPINLOCK(test_lock); +static DEFINE_RAW_SPINLOCK(test_lock); static void ref_lock_section(const int nloops) { @@ -393,8 +393,8 @@ static void ref_lock_section(const int nloops) preempt_disable(); for (i = nloops; i >= 0; i--) { - spin_lock(&test_lock); - spin_unlock(&test_lock); + raw_spin_lock(&test_lock); + raw_spin_unlock(&test_lock); } preempt_enable(); } @@ -405,9 +405,9 @@ static void ref_lock_delay_section(const int nloops, const int udl, const int nd preempt_disable(); for (i = nloops; i >= 0; i--) { - spin_lock(&test_lock); + raw_spin_lock(&test_lock); un_delay(udl, ndl); - spin_unlock(&test_lock); + raw_spin_unlock(&test_lock); } preempt_enable(); } @@ -427,8 +427,8 @@ static void ref_lock_irq_section(const int nloops) preempt_disable(); for (i = nloops; i >= 0; i--) { - spin_lock_irqsave(&test_lock, flags); - spin_unlock_irqrestore(&test_lock, flags); + raw_spin_lock_irqsave(&test_lock, flags); + raw_spin_unlock_irqrestore(&test_lock, flags); } preempt_enable(); } @@ -440,9 +440,9 @@ static void ref_lock_irq_delay_section(const int nloops, const int udl, const in preempt_disable(); for (i = nloops; i >= 0; i--) { - spin_lock_irqsave(&test_lock, flags); + raw_spin_lock_irqsave(&test_lock, flags); un_delay(udl, ndl); - spin_unlock_irqrestore(&test_lock, flags); + raw_spin_unlock_irqrestore(&test_lock, flags); } preempt_enable(); } diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c index 50ba70f019de..1c304fec89c0 100644 --- a/kernel/rcu/srcutree.c +++ b/kernel/rcu/srcutree.c @@ -511,10 +511,52 @@ static bool srcu_readers_active(struct srcu_struct *ssp) return sum; } -#define SRCU_INTERVAL 1 // Base delay if no expedited GPs pending. -#define SRCU_MAX_INTERVAL 10 // Maximum incremental delay from slow readers. -#define SRCU_MAX_NODELAY_PHASE 1 // Maximum per-GP-phase consecutive no-delay instances. -#define SRCU_MAX_NODELAY 100 // Maximum consecutive no-delay instances. +/* + * We use an adaptive strategy for synchronize_srcu() and especially for + * synchronize_srcu_expedited(). We spin for a fixed time period + * (defined below, boot time configurable) to allow SRCU readers to exit + * their read-side critical sections. If there are still some readers + * after one jiffy, we repeatedly block for one jiffy time periods. + * The blocking time is increased as the grace-period age increases, + * with max blocking time capped at 10 jiffies. + */ +#define SRCU_DEFAULT_RETRY_CHECK_DELAY 5 + +static ulong srcu_retry_check_delay = SRCU_DEFAULT_RETRY_CHECK_DELAY; +module_param(srcu_retry_check_delay, ulong, 0444); + +#define SRCU_INTERVAL 1 // Base delay if no expedited GPs pending. +#define SRCU_MAX_INTERVAL 10 // Maximum incremental delay from slow readers. + +#define SRCU_DEFAULT_MAX_NODELAY_PHASE_LO 3UL // Lowmark on default per-GP-phase + // no-delay instances. +#define SRCU_DEFAULT_MAX_NODELAY_PHASE_HI 1000UL // Highmark on default per-GP-phase + // no-delay instances. + +#define SRCU_UL_CLAMP_LO(val, low) ((val) > (low) ? (val) : (low)) +#define SRCU_UL_CLAMP_HI(val, high) ((val) < (high) ? (val) : (high)) +#define SRCU_UL_CLAMP(val, low, high) SRCU_UL_CLAMP_HI(SRCU_UL_CLAMP_LO((val), (low)), (high)) +// per-GP-phase no-delay instances adjusted to allow non-sleeping poll upto +// one jiffies time duration. Mult by 2 is done to factor in the srcu_get_delay() +// called from process_srcu(). +#define SRCU_DEFAULT_MAX_NODELAY_PHASE_ADJUSTED \ + (2UL * USEC_PER_SEC / HZ / SRCU_DEFAULT_RETRY_CHECK_DELAY) + +// Maximum per-GP-phase consecutive no-delay instances. +#define SRCU_DEFAULT_MAX_NODELAY_PHASE \ + SRCU_UL_CLAMP(SRCU_DEFAULT_MAX_NODELAY_PHASE_ADJUSTED, \ + SRCU_DEFAULT_MAX_NODELAY_PHASE_LO, \ + SRCU_DEFAULT_MAX_NODELAY_PHASE_HI) + +static ulong srcu_max_nodelay_phase = SRCU_DEFAULT_MAX_NODELAY_PHASE; +module_param(srcu_max_nodelay_phase, ulong, 0444); + +// Maximum consecutive no-delay instances. +#define SRCU_DEFAULT_MAX_NODELAY (SRCU_DEFAULT_MAX_NODELAY_PHASE > 100 ? \ + SRCU_DEFAULT_MAX_NODELAY_PHASE : 100) + +static ulong srcu_max_nodelay = SRCU_DEFAULT_MAX_NODELAY; +module_param(srcu_max_nodelay, ulong, 0444); /* * Return grace-period delay, zero if there are expedited grace @@ -522,16 +564,22 @@ static bool srcu_readers_active(struct srcu_struct *ssp) */ static unsigned long srcu_get_delay(struct srcu_struct *ssp) { + unsigned long gpstart; + unsigned long j; unsigned long jbase = SRCU_INTERVAL; if (ULONG_CMP_LT(READ_ONCE(ssp->srcu_gp_seq), READ_ONCE(ssp->srcu_gp_seq_needed_exp))) jbase = 0; - if (rcu_seq_state(READ_ONCE(ssp->srcu_gp_seq))) - jbase += jiffies - READ_ONCE(ssp->srcu_gp_start); - if (!jbase) { - WRITE_ONCE(ssp->srcu_n_exp_nodelay, READ_ONCE(ssp->srcu_n_exp_nodelay) + 1); - if (READ_ONCE(ssp->srcu_n_exp_nodelay) > SRCU_MAX_NODELAY_PHASE) - jbase = 1; + if (rcu_seq_state(READ_ONCE(ssp->srcu_gp_seq))) { + j = jiffies - 1; + gpstart = READ_ONCE(ssp->srcu_gp_start); + if (time_after(j, gpstart)) + jbase += j - gpstart; + if (!jbase) { + WRITE_ONCE(ssp->srcu_n_exp_nodelay, READ_ONCE(ssp->srcu_n_exp_nodelay) + 1); + if (READ_ONCE(ssp->srcu_n_exp_nodelay) > srcu_max_nodelay_phase) + jbase = 1; + } } return jbase > SRCU_MAX_INTERVAL ? SRCU_MAX_INTERVAL : jbase; } @@ -607,15 +655,6 @@ void __srcu_read_unlock(struct srcu_struct *ssp, int idx) EXPORT_SYMBOL_GPL(__srcu_read_unlock); /* - * We use an adaptive strategy for synchronize_srcu() and especially for - * synchronize_srcu_expedited(). We spin for a fixed time period - * (defined below) to allow SRCU readers to exit their read-side critical - * sections. If there are still some readers after a few microseconds, - * we repeatedly block for 1-millisecond time periods. - */ -#define SRCU_RETRY_CHECK_DELAY 5 - -/* * Start an SRCU grace period. */ static void srcu_gp_start(struct srcu_struct *ssp) @@ -700,7 +739,7 @@ static void srcu_schedule_cbs_snp(struct srcu_struct *ssp, struct srcu_node *snp */ static void srcu_gp_end(struct srcu_struct *ssp) { - unsigned long cbdelay; + unsigned long cbdelay = 1; bool cbs; bool last_lvl; int cpu; @@ -720,7 +759,9 @@ static void srcu_gp_end(struct srcu_struct *ssp) spin_lock_irq_rcu_node(ssp); idx = rcu_seq_state(ssp->srcu_gp_seq); WARN_ON_ONCE(idx != SRCU_STATE_SCAN2); - cbdelay = !!srcu_get_delay(ssp); + if (ULONG_CMP_LT(READ_ONCE(ssp->srcu_gp_seq), READ_ONCE(ssp->srcu_gp_seq_needed_exp))) + cbdelay = 0; + WRITE_ONCE(ssp->srcu_last_gp_end, ktime_get_mono_fast_ns()); rcu_seq_end(&ssp->srcu_gp_seq); gpseq = rcu_seq_current(&ssp->srcu_gp_seq); @@ -921,12 +962,16 @@ static void srcu_funnel_gp_start(struct srcu_struct *ssp, struct srcu_data *sdp, */ static bool try_check_zero(struct srcu_struct *ssp, int idx, int trycount) { + unsigned long curdelay; + + curdelay = !srcu_get_delay(ssp); + for (;;) { if (srcu_readers_active_idx_check(ssp, idx)) return true; - if (--trycount + !srcu_get_delay(ssp) <= 0) + if ((--trycount + curdelay) <= 0) return false; - udelay(SRCU_RETRY_CHECK_DELAY); + udelay(srcu_retry_check_delay); } } @@ -1582,7 +1627,7 @@ static void process_srcu(struct work_struct *work) j = jiffies; if (READ_ONCE(ssp->reschedule_jiffies) == j) { WRITE_ONCE(ssp->reschedule_count, READ_ONCE(ssp->reschedule_count) + 1); - if (READ_ONCE(ssp->reschedule_count) > SRCU_MAX_NODELAY) + if (READ_ONCE(ssp->reschedule_count) > srcu_max_nodelay) curdelay = 1; } else { WRITE_ONCE(ssp->reschedule_count, 1); @@ -1674,6 +1719,11 @@ static int __init srcu_bootup_announce(void) pr_info("Hierarchical SRCU implementation.\n"); if (exp_holdoff != DEFAULT_SRCU_EXP_HOLDOFF) pr_info("\tNon-default auto-expedite holdoff of %lu ns.\n", exp_holdoff); + if (srcu_retry_check_delay != SRCU_DEFAULT_RETRY_CHECK_DELAY) + pr_info("\tNon-default retry check delay of %lu us.\n", srcu_retry_check_delay); + if (srcu_max_nodelay != SRCU_DEFAULT_MAX_NODELAY) + pr_info("\tNon-default max no-delay of %lu.\n", srcu_max_nodelay); + pr_info("\tMax phase no-delay instances is %lu.\n", srcu_max_nodelay_phase); return 0; } early_initcall(srcu_bootup_announce); diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h index 3925e32159b5..83c7e6620d40 100644 --- a/kernel/rcu/tasks.h +++ b/kernel/rcu/tasks.h @@ -14,7 +14,7 @@ struct rcu_tasks; typedef void (*rcu_tasks_gp_func_t)(struct rcu_tasks *rtp); -typedef void (*pregp_func_t)(void); +typedef void (*pregp_func_t)(struct list_head *hop); typedef void (*pertask_func_t)(struct task_struct *t, struct list_head *hop); typedef void (*postscan_func_t)(struct list_head *hop); typedef void (*holdouts_func_t)(struct list_head *hop, bool ndrpt, bool *frptp); @@ -29,6 +29,7 @@ typedef void (*postgp_func_t)(struct rcu_tasks *rtp); * @rtp_work: Work queue for invoking callbacks. * @rtp_irq_work: IRQ work queue for deferred wakeups. * @barrier_q_head: RCU callback for barrier operation. + * @rtp_blkd_tasks: List of tasks blocked as readers. * @cpu: CPU number corresponding to this entry. * @rtpp: Pointer to the rcu_tasks structure. */ @@ -40,6 +41,7 @@ struct rcu_tasks_percpu { struct work_struct rtp_work; struct irq_work rtp_irq_work; struct rcu_head barrier_q_head; + struct list_head rtp_blkd_tasks; int cpu; struct rcu_tasks *rtpp; }; @@ -48,6 +50,7 @@ struct rcu_tasks_percpu { * struct rcu_tasks - Definition for a Tasks-RCU-like mechanism. * @cbs_wait: RCU wait allowing a new callback to get kthread's attention. * @cbs_gbl_lock: Lock protecting callback list. + * @tasks_gp_mutex: Mutex protecting grace period, needed during mid-boot dead zone. * @kthread_ptr: This flavor's grace-period/callback-invocation kthread. * @gp_func: This flavor's grace-period-wait function. * @gp_state: Grace period's most recent state transition (debugging). @@ -79,6 +82,7 @@ struct rcu_tasks_percpu { struct rcu_tasks { struct rcuwait cbs_wait; raw_spinlock_t cbs_gbl_lock; + struct mutex tasks_gp_mutex; int gp_state; int gp_sleep; int init_fract; @@ -119,6 +123,7 @@ static struct rcu_tasks rt_name = \ { \ .cbs_wait = __RCUWAIT_INITIALIZER(rt_name.wait), \ .cbs_gbl_lock = __RAW_SPIN_LOCK_UNLOCKED(rt_name.cbs_gbl_lock), \ + .tasks_gp_mutex = __MUTEX_INITIALIZER(rt_name.tasks_gp_mutex), \ .gp_func = gp, \ .call_func = call, \ .rtpcpu = &rt_name ## __percpu, \ @@ -140,6 +145,7 @@ static int rcu_task_ipi_delay __read_mostly = RCU_TASK_IPI_DELAY; module_param(rcu_task_ipi_delay, int, 0644); /* Control stall timeouts. Disable with <= 0, otherwise jiffies till stall. */ +#define RCU_TASK_BOOT_STALL_TIMEOUT (HZ * 30) #define RCU_TASK_STALL_TIMEOUT (HZ * 60 * 10) static int rcu_task_stall_timeout __read_mostly = RCU_TASK_STALL_TIMEOUT; module_param(rcu_task_stall_timeout, int, 0644); @@ -253,6 +259,8 @@ static void cblist_init_generic(struct rcu_tasks *rtp) INIT_WORK(&rtpcp->rtp_work, rcu_tasks_invoke_cbs_wq); rtpcp->cpu = cpu; rtpcp->rtpp = rtp; + if (!rtpcp->rtp_blkd_tasks.next) + INIT_LIST_HEAD(&rtpcp->rtp_blkd_tasks); raw_spin_unlock_rcu_node(rtpcp); // irqs remain disabled. } raw_spin_unlock_irqrestore(&rtp->cbs_gbl_lock, flags); @@ -323,17 +331,6 @@ static void call_rcu_tasks_generic(struct rcu_head *rhp, rcu_callback_t func, irq_work_queue(&rtpcp->rtp_irq_work); } -// Wait for a grace period for the specified flavor of Tasks RCU. -static void synchronize_rcu_tasks_generic(struct rcu_tasks *rtp) -{ - /* Complain if the scheduler has not started. */ - RCU_LOCKDEP_WARN(rcu_scheduler_active == RCU_SCHEDULER_INACTIVE, - "synchronize_rcu_tasks called too soon"); - - /* Wait for the grace period. */ - wait_rcu_gp(rtp->call_func); -} - // RCU callback function for rcu_barrier_tasks_generic(). static void rcu_barrier_tasks_generic_cb(struct rcu_head *rhp) { @@ -439,6 +436,11 @@ static int rcu_tasks_need_gpcb(struct rcu_tasks *rtp) WRITE_ONCE(rtp->percpu_dequeue_lim, 1); pr_info("Completing switch %s to CPU-0 callback queuing.\n", rtp->name); } + for (cpu = rtp->percpu_dequeue_lim; cpu < nr_cpu_ids; cpu++) { + struct rcu_tasks_percpu *rtpcp = per_cpu_ptr(rtp->rtpcpu, cpu); + + WARN_ON_ONCE(rcu_segcblist_n_cbs(&rtpcp->cblist)); + } raw_spin_unlock_irqrestore(&rtp->cbs_gbl_lock, flags); } @@ -497,10 +499,41 @@ static void rcu_tasks_invoke_cbs_wq(struct work_struct *wp) rcu_tasks_invoke_cbs(rtp, rtpcp); } -/* RCU-tasks kthread that detects grace periods and invokes callbacks. */ -static int __noreturn rcu_tasks_kthread(void *arg) +// Wait for one grace period. +static void rcu_tasks_one_gp(struct rcu_tasks *rtp, bool midboot) { int needgpcb; + + mutex_lock(&rtp->tasks_gp_mutex); + + // If there were none, wait a bit and start over. + if (unlikely(midboot)) { + needgpcb = 0x2; + } else { + set_tasks_gp_state(rtp, RTGS_WAIT_CBS); + rcuwait_wait_event(&rtp->cbs_wait, + (needgpcb = rcu_tasks_need_gpcb(rtp)), + TASK_IDLE); + } + + if (needgpcb & 0x2) { + // Wait for one grace period. + set_tasks_gp_state(rtp, RTGS_WAIT_GP); + rtp->gp_start = jiffies; + rcu_seq_start(&rtp->tasks_gp_seq); + rtp->gp_func(rtp); + rcu_seq_end(&rtp->tasks_gp_seq); + } + + // Invoke callbacks. + set_tasks_gp_state(rtp, RTGS_INVOKE_CBS); + rcu_tasks_invoke_cbs(rtp, per_cpu_ptr(rtp->rtpcpu, 0)); + mutex_unlock(&rtp->tasks_gp_mutex); +} + +// RCU-tasks kthread that detects grace periods and invokes callbacks. +static int __noreturn rcu_tasks_kthread(void *arg) +{ struct rcu_tasks *rtp = arg; /* Run on housekeeping CPUs by default. Sysadm can move if desired. */ @@ -514,29 +547,28 @@ static int __noreturn rcu_tasks_kthread(void *arg) * This loop is terminated by the system going down. ;-) */ for (;;) { - set_tasks_gp_state(rtp, RTGS_WAIT_CBS); + // Wait for one grace period and invoke any callbacks + // that are ready. + rcu_tasks_one_gp(rtp, false); - /* If there were none, wait a bit and start over. */ - rcuwait_wait_event(&rtp->cbs_wait, - (needgpcb = rcu_tasks_need_gpcb(rtp)), - TASK_IDLE); - - if (needgpcb & 0x2) { - // Wait for one grace period. - set_tasks_gp_state(rtp, RTGS_WAIT_GP); - rtp->gp_start = jiffies; - rcu_seq_start(&rtp->tasks_gp_seq); - rtp->gp_func(rtp); - rcu_seq_end(&rtp->tasks_gp_seq); - } + // Paranoid sleep to keep this from entering a tight loop. + schedule_timeout_idle(rtp->gp_sleep); + } +} - /* Invoke callbacks. */ - set_tasks_gp_state(rtp, RTGS_INVOKE_CBS); - rcu_tasks_invoke_cbs(rtp, per_cpu_ptr(rtp->rtpcpu, 0)); +// Wait for a grace period for the specified flavor of Tasks RCU. +static void synchronize_rcu_tasks_generic(struct rcu_tasks *rtp) +{ + /* Complain if the scheduler has not started. */ + RCU_LOCKDEP_WARN(rcu_scheduler_active == RCU_SCHEDULER_INACTIVE, + "synchronize_rcu_tasks called too soon"); - /* Paranoid sleep to keep this from entering a tight loop */ - schedule_timeout_idle(rtp->gp_sleep); + // If the grace-period kthread is running, use it. + if (READ_ONCE(rtp->kthread_ptr)) { + wait_rcu_gp(rtp->call_func); + return; } + rcu_tasks_one_gp(rtp, true); } /* Spawn RCU-tasks grace-period kthread. */ @@ -630,7 +662,7 @@ static void rcu_tasks_wait_gp(struct rcu_tasks *rtp) struct task_struct *t; set_tasks_gp_state(rtp, RTGS_PRE_WAIT_GP); - rtp->pregp_func(); + rtp->pregp_func(&holdouts); /* * There were callbacks, so we need to wait for an RCU-tasks @@ -639,10 +671,12 @@ static void rcu_tasks_wait_gp(struct rcu_tasks *rtp) * and make a list of them in holdouts. */ set_tasks_gp_state(rtp, RTGS_SCAN_TASKLIST); - rcu_read_lock(); - for_each_process_thread(g, t) - rtp->pertask_func(t, &holdouts); - rcu_read_unlock(); + if (rtp->pertask_func) { + rcu_read_lock(); + for_each_process_thread(g, t) + rtp->pertask_func(t, &holdouts); + rcu_read_unlock(); + } set_tasks_gp_state(rtp, RTGS_POST_SCAN_TASKLIST); rtp->postscan_func(&holdouts); @@ -760,7 +794,7 @@ static void rcu_tasks_wait_gp(struct rcu_tasks *rtp) // disabling. /* Pre-grace-period preparation. */ -static void rcu_tasks_pregp_step(void) +static void rcu_tasks_pregp_step(struct list_head *hop) { /* * Wait for all pre-existing t->on_rq and t->nvcsw transitions @@ -1105,11 +1139,10 @@ EXPORT_SYMBOL_GPL(show_rcu_tasks_rude_gp_kthread); // 3. Avoids expensive read-side instructions, having overhead similar // to that of Preemptible RCU. // -// There are of course downsides. The grace-period code can send IPIs to -// CPUs, even when those CPUs are in the idle loop or in nohz_full userspace. -// It is necessary to scan the full tasklist, much as for Tasks RCU. There -// is a single callback queue guarded by a single lock, again, much as for -// Tasks RCU. If needed, these downsides can be at least partially remedied. +// There are of course downsides. For example, the grace-period code +// can send IPIs to CPUs, even when those CPUs are in the idle loop or +// in nohz_full userspace. If needed, these downsides can be at least +// partially remedied. // // Perhaps most important, this variant of RCU does not affect the vanilla // flavors, rcu_preempt and rcu_sched. The fact that RCU Tasks Trace @@ -1122,38 +1155,30 @@ EXPORT_SYMBOL_GPL(show_rcu_tasks_rude_gp_kthread); // invokes these functions in this order: // // rcu_tasks_trace_pregp_step(): -// Initialize the count of readers and block CPU-hotplug operations. -// rcu_tasks_trace_pertask(), invoked on every non-idle task: -// Initialize per-task state and attempt to identify an immediate -// quiescent state for that task, or, failing that, attempt to -// set that task's .need_qs flag so that task's next outermost -// rcu_read_unlock_trace() will report the quiescent state (in which -// case the count of readers is incremented). If both attempts fail, -// the task is added to a "holdout" list. Note that IPIs are used -// to invoke trc_read_check_handler() in the context of running tasks -// in order to avoid ordering overhead on common-case shared-variable -// accessses. +// Disables CPU hotplug, adds all currently executing tasks to the +// holdout list, then checks the state of all tasks that blocked +// or were preempted within their current RCU Tasks Trace read-side +// critical section, adding them to the holdout list if appropriate. +// Finally, this function re-enables CPU hotplug. +// The ->pertask_func() pointer is NULL, so there is no per-task processing. // rcu_tasks_trace_postscan(): -// Initialize state and attempt to identify an immediate quiescent -// state as above (but only for idle tasks), unblock CPU-hotplug -// operations, and wait for an RCU grace period to avoid races with -// tasks that are in the process of exiting. +// Invokes synchronize_rcu() to wait for late-stage exiting tasks +// to finish exiting. // check_all_holdout_tasks_trace(), repeatedly until holdout list is empty: // Scans the holdout list, attempting to identify a quiescent state // for each task on the list. If there is a quiescent state, the -// corresponding task is removed from the holdout list. +// corresponding task is removed from the holdout list. Once this +// list is empty, the grace period has completed. // rcu_tasks_trace_postgp(): -// Wait for the count of readers do drop to zero, reporting any stalls. -// Also execute full memory barriers to maintain ordering with code -// executing after the grace period. +// Provides the needed full memory barrier and does debug checks. // // The exit_tasks_rcu_finish_trace() synchronizes with exiting tasks. // -// Pre-grace-period update-side code is ordered before the grace -// period via the ->cbs_lock and barriers in rcu_tasks_kthread(). -// Pre-grace-period read-side code is ordered before the grace period by -// atomic_dec_and_test() of the count of readers (for IPIed readers) and by -// scheduler context-switch ordering (for locked-down non-running readers). +// Pre-grace-period update-side code is ordered before the grace period +// via the ->cbs_lock and barriers in rcu_tasks_kthread(). Pre-grace-period +// read-side code is ordered before the grace period by atomic operations +// on .b.need_qs flag of each task involved in this process, or by scheduler +// context-switch ordering (for locked-down non-running readers). // The lockdep state must be outside of #ifdef to be useful. #ifdef CONFIG_DEBUG_LOCK_ALLOC @@ -1165,9 +1190,6 @@ EXPORT_SYMBOL_GPL(rcu_trace_lock_map); #ifdef CONFIG_TASKS_TRACE_RCU -static atomic_t trc_n_readers_need_end; // Number of waited-for readers. -static DECLARE_WAIT_QUEUE_HEAD(trc_wait); // List of holdout tasks. - // Record outstanding IPIs to each CPU. No point in sending two... static DEFINE_PER_CPU(bool, trc_ipi_to_cpu); @@ -1176,44 +1198,104 @@ static DEFINE_PER_CPU(bool, trc_ipi_to_cpu); static unsigned long n_heavy_reader_attempts; static unsigned long n_heavy_reader_updates; static unsigned long n_heavy_reader_ofl_updates; +static unsigned long n_trc_holdouts; void call_rcu_tasks_trace(struct rcu_head *rhp, rcu_callback_t func); DEFINE_RCU_TASKS(rcu_tasks_trace, rcu_tasks_wait_gp, call_rcu_tasks_trace, "RCU Tasks Trace"); +/* Load from ->trc_reader_special.b.need_qs with proper ordering. */ +static u8 rcu_ld_need_qs(struct task_struct *t) +{ + smp_mb(); // Enforce full grace-period ordering. + return smp_load_acquire(&t->trc_reader_special.b.need_qs); +} + +/* Store to ->trc_reader_special.b.need_qs with proper ordering. */ +static void rcu_st_need_qs(struct task_struct *t, u8 v) +{ + smp_store_release(&t->trc_reader_special.b.need_qs, v); + smp_mb(); // Enforce full grace-period ordering. +} + /* - * This irq_work handler allows rcu_read_unlock_trace() to be invoked - * while the scheduler locks are held. + * Do a cmpxchg() on ->trc_reader_special.b.need_qs, allowing for + * the four-byte operand-size restriction of some platforms. + * Returns the old value, which is often ignored. */ -static void rcu_read_unlock_iw(struct irq_work *iwp) +u8 rcu_trc_cmpxchg_need_qs(struct task_struct *t, u8 old, u8 new) { - wake_up(&trc_wait); + union rcu_special ret; + union rcu_special trs_old = READ_ONCE(t->trc_reader_special); + union rcu_special trs_new = trs_old; + + if (trs_old.b.need_qs != old) + return trs_old.b.need_qs; + trs_new.b.need_qs = new; + ret.s = cmpxchg(&t->trc_reader_special.s, trs_old.s, trs_new.s); + return ret.b.need_qs; } -static DEFINE_IRQ_WORK(rcu_tasks_trace_iw, rcu_read_unlock_iw); +EXPORT_SYMBOL_GPL(rcu_trc_cmpxchg_need_qs); -/* If we are the last reader, wake up the grace-period kthread. */ +/* + * If we are the last reader, signal the grace-period kthread. + * Also remove from the per-CPU list of blocked tasks. + */ void rcu_read_unlock_trace_special(struct task_struct *t) { - int nq = READ_ONCE(t->trc_reader_special.b.need_qs); + unsigned long flags; + struct rcu_tasks_percpu *rtpcp; + union rcu_special trs; + + // Open-coded full-word version of rcu_ld_need_qs(). + smp_mb(); // Enforce full grace-period ordering. + trs = smp_load_acquire(&t->trc_reader_special); - if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) && - t->trc_reader_special.b.need_mb) + if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) && t->trc_reader_special.b.need_mb) smp_mb(); // Pairs with update-side barriers. // Update .need_qs before ->trc_reader_nesting for irq/NMI handlers. - if (nq) - WRITE_ONCE(t->trc_reader_special.b.need_qs, false); + if (trs.b.need_qs == (TRC_NEED_QS_CHECKED | TRC_NEED_QS)) { + u8 result = rcu_trc_cmpxchg_need_qs(t, TRC_NEED_QS_CHECKED | TRC_NEED_QS, + TRC_NEED_QS_CHECKED); + + WARN_ONCE(result != trs.b.need_qs, "%s: result = %d", __func__, result); + } + if (trs.b.blocked) { + rtpcp = per_cpu_ptr(rcu_tasks_trace.rtpcpu, t->trc_blkd_cpu); + raw_spin_lock_irqsave_rcu_node(rtpcp, flags); + list_del_init(&t->trc_blkd_node); + WRITE_ONCE(t->trc_reader_special.b.blocked, false); + raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags); + } WRITE_ONCE(t->trc_reader_nesting, 0); - if (nq && atomic_dec_and_test(&trc_n_readers_need_end)) - irq_work_queue(&rcu_tasks_trace_iw); } EXPORT_SYMBOL_GPL(rcu_read_unlock_trace_special); +/* Add a newly blocked reader task to its CPU's list. */ +void rcu_tasks_trace_qs_blkd(struct task_struct *t) +{ + unsigned long flags; + struct rcu_tasks_percpu *rtpcp; + + local_irq_save(flags); + rtpcp = this_cpu_ptr(rcu_tasks_trace.rtpcpu); + raw_spin_lock_rcu_node(rtpcp); // irqs already disabled + t->trc_blkd_cpu = smp_processor_id(); + if (!rtpcp->rtp_blkd_tasks.next) + INIT_LIST_HEAD(&rtpcp->rtp_blkd_tasks); + list_add(&t->trc_blkd_node, &rtpcp->rtp_blkd_tasks); + WRITE_ONCE(t->trc_reader_special.b.blocked, true); + raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags); +} +EXPORT_SYMBOL_GPL(rcu_tasks_trace_qs_blkd); + /* Add a task to the holdout list, if it is not already on the list. */ static void trc_add_holdout(struct task_struct *t, struct list_head *bhp) { if (list_empty(&t->trc_holdout_list)) { get_task_struct(t); list_add(&t->trc_holdout_list, bhp); + n_trc_holdouts++; } } @@ -1223,37 +1305,36 @@ static void trc_del_holdout(struct task_struct *t) if (!list_empty(&t->trc_holdout_list)) { list_del_init(&t->trc_holdout_list); put_task_struct(t); + n_trc_holdouts--; } } /* IPI handler to check task state. */ static void trc_read_check_handler(void *t_in) { + int nesting; struct task_struct *t = current; struct task_struct *texp = t_in; // If the task is no longer running on this CPU, leave. - if (unlikely(texp != t)) { + if (unlikely(texp != t)) goto reset_ipi; // Already on holdout list, so will check later. - } // If the task is not in a read-side critical section, and // if this is the last reader, awaken the grace-period kthread. - if (likely(!READ_ONCE(t->trc_reader_nesting))) { - WRITE_ONCE(t->trc_reader_checked, true); + nesting = READ_ONCE(t->trc_reader_nesting); + if (likely(!nesting)) { + rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED); goto reset_ipi; } // If we are racing with an rcu_read_unlock_trace(), try again later. - if (unlikely(READ_ONCE(t->trc_reader_nesting) < 0)) + if (unlikely(nesting < 0)) goto reset_ipi; - WRITE_ONCE(t->trc_reader_checked, true); - // Get here if the task is in a read-side critical section. Set - // its state so that it will awaken the grace-period kthread upon - // exit from that critical section. - atomic_inc(&trc_n_readers_need_end); // One more to wait on. - WARN_ON_ONCE(READ_ONCE(t->trc_reader_special.b.need_qs)); - WRITE_ONCE(t->trc_reader_special.b.need_qs, true); + // Get here if the task is in a read-side critical section. + // Set its state so that it will update state for the grace-period + // kthread upon exit from that critical section. + rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS | TRC_NEED_QS_CHECKED); reset_ipi: // Allow future IPIs to be sent on CPU and for task. @@ -1264,48 +1345,50 @@ reset_ipi: } /* Callback function for scheduler to check locked-down task. */ -static int trc_inspect_reader(struct task_struct *t, void *arg) +static int trc_inspect_reader(struct task_struct *t, void *bhp_in) { + struct list_head *bhp = bhp_in; int cpu = task_cpu(t); int nesting; bool ofl = cpu_is_offline(cpu); - if (task_curr(t)) { - WARN_ON_ONCE(ofl && !is_idle_task(t)); - + if (task_curr(t) && !ofl) { // If no chance of heavyweight readers, do it the hard way. - if (!ofl && !IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB)) + if (!IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB)) return -EINVAL; // If heavyweight readers are enabled on the remote task, // we can inspect its state despite its currently running. // However, we cannot safely change its state. n_heavy_reader_attempts++; - if (!ofl && // Check for "running" idle tasks on offline CPUs. - !rcu_dynticks_zero_in_eqs(cpu, &t->trc_reader_nesting)) + // Check for "running" idle tasks on offline CPUs. + if (!rcu_dynticks_zero_in_eqs(cpu, &t->trc_reader_nesting)) return -EINVAL; // No quiescent state, do it the hard way. n_heavy_reader_updates++; - if (ofl) - n_heavy_reader_ofl_updates++; nesting = 0; } else { // The task is not running, so C-language access is safe. nesting = t->trc_reader_nesting; + WARN_ON_ONCE(ofl && task_curr(t) && !is_idle_task(t)); + if (IS_ENABLED(CONFIG_TASKS_TRACE_RCU_READ_MB) && ofl) + n_heavy_reader_ofl_updates++; } // If not exiting a read-side critical section, mark as checked // so that the grace-period kthread will remove it from the // holdout list. - t->trc_reader_checked = nesting >= 0; - if (nesting <= 0) - return nesting ? -EINVAL : 0; // If in QS, done, otherwise try again later. + if (!nesting) { + rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED); + return 0; // In QS, so done. + } + if (nesting < 0) + return -EINVAL; // Reader transitioning, try again later. // The task is in a read-side critical section, so set up its - // state so that it will awaken the grace-period kthread upon exit - // from that critical section. - atomic_inc(&trc_n_readers_need_end); // One more to wait on. - WARN_ON_ONCE(READ_ONCE(t->trc_reader_special.b.need_qs)); - WRITE_ONCE(t->trc_reader_special.b.need_qs, true); + // state so that it will update state upon exit from that critical + // section. + if (!rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS | TRC_NEED_QS_CHECKED)) + trc_add_holdout(t, bhp); return 0; } @@ -1321,14 +1404,14 @@ static void trc_wait_for_one_reader(struct task_struct *t, // The current task had better be in a quiescent state. if (t == current) { - t->trc_reader_checked = true; + rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED); WARN_ON_ONCE(READ_ONCE(t->trc_reader_nesting)); return; } // Attempt to nail down the task for inspection. get_task_struct(t); - if (!task_call_func(t, trc_inspect_reader, NULL)) { + if (!task_call_func(t, trc_inspect_reader, bhp)) { put_task_struct(t); return; } @@ -1366,56 +1449,93 @@ static void trc_wait_for_one_reader(struct task_struct *t, } } +/* + * Initialize for first-round processing for the specified task. + * Return false if task is NULL or already taken care of, true otherwise. + */ +static bool rcu_tasks_trace_pertask_prep(struct task_struct *t, bool notself) +{ + // During early boot when there is only the one boot CPU, there + // is no idle task for the other CPUs. Also, the grace-period + // kthread is always in a quiescent state. In addition, just return + // if this task is already on the list. + if (unlikely(t == NULL) || (t == current && notself) || !list_empty(&t->trc_holdout_list)) + return false; + + rcu_st_need_qs(t, 0); + t->trc_ipi_to_cpu = -1; + return true; +} + +/* Do first-round processing for the specified task. */ +static void rcu_tasks_trace_pertask(struct task_struct *t, struct list_head *hop) +{ + if (rcu_tasks_trace_pertask_prep(t, true)) + trc_wait_for_one_reader(t, hop); +} + /* Initialize for a new RCU-tasks-trace grace period. */ -static void rcu_tasks_trace_pregp_step(void) +static void rcu_tasks_trace_pregp_step(struct list_head *hop) { + LIST_HEAD(blkd_tasks); int cpu; - - // Allow for fast-acting IPIs. - atomic_set(&trc_n_readers_need_end, 1); + unsigned long flags; + struct rcu_tasks_percpu *rtpcp; + struct task_struct *t; // There shouldn't be any old IPIs, but... for_each_possible_cpu(cpu) WARN_ON_ONCE(per_cpu(trc_ipi_to_cpu, cpu)); - // Disable CPU hotplug across the tasklist scan. - // This also waits for all readers in CPU-hotplug code paths. + // Disable CPU hotplug across the CPU scan for the benefit of + // any IPIs that might be needed. This also waits for all readers + // in CPU-hotplug code paths. cpus_read_lock(); -} -/* Do first-round processing for the specified task. */ -static void rcu_tasks_trace_pertask(struct task_struct *t, - struct list_head *hop) -{ - // During early boot when there is only the one boot CPU, there - // is no idle task for the other CPUs. Just return. - if (unlikely(t == NULL)) - return; + // These rcu_tasks_trace_pertask_prep() calls are serialized to + // allow safe access to the hop list. + for_each_online_cpu(cpu) { + rcu_read_lock(); + t = cpu_curr_snapshot(cpu); + if (rcu_tasks_trace_pertask_prep(t, true)) + trc_add_holdout(t, hop); + rcu_read_unlock(); + } - WRITE_ONCE(t->trc_reader_special.b.need_qs, false); - WRITE_ONCE(t->trc_reader_checked, false); - t->trc_ipi_to_cpu = -1; - trc_wait_for_one_reader(t, hop); + // Only after all running tasks have been accounted for is it + // safe to take care of the tasks that have blocked within their + // current RCU tasks trace read-side critical section. + for_each_possible_cpu(cpu) { + rtpcp = per_cpu_ptr(rcu_tasks_trace.rtpcpu, cpu); + raw_spin_lock_irqsave_rcu_node(rtpcp, flags); + list_splice_init(&rtpcp->rtp_blkd_tasks, &blkd_tasks); + while (!list_empty(&blkd_tasks)) { + rcu_read_lock(); + t = list_first_entry(&blkd_tasks, struct task_struct, trc_blkd_node); + list_del_init(&t->trc_blkd_node); + list_add(&t->trc_blkd_node, &rtpcp->rtp_blkd_tasks); + raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags); + rcu_tasks_trace_pertask(t, hop); + rcu_read_unlock(); + raw_spin_lock_irqsave_rcu_node(rtpcp, flags); + } + raw_spin_unlock_irqrestore_rcu_node(rtpcp, flags); + } + + // Re-enable CPU hotplug now that the holdout list is populated. + cpus_read_unlock(); } /* - * Do intermediate processing between task and holdout scans and - * pick up the idle tasks. + * Do intermediate processing between task and holdout scans. */ static void rcu_tasks_trace_postscan(struct list_head *hop) { - int cpu; - - for_each_possible_cpu(cpu) - rcu_tasks_trace_pertask(idle_task(cpu), hop); - - // Re-enable CPU hotplug now that the tasklist scan has completed. - cpus_read_unlock(); - // Wait for late-stage exiting tasks to finish exiting. // These might have passed the call to exit_tasks_rcu_finish(). synchronize_rcu(); - // Any tasks that exit after this point will set ->trc_reader_checked. + // Any tasks that exit after this point will set + // TRC_NEED_QS_CHECKED in ->trc_reader_special.b.need_qs. } /* Communicate task state back to the RCU tasks trace stall warning request. */ @@ -1429,11 +1549,11 @@ static int trc_check_slow_task(struct task_struct *t, void *arg) { struct trc_stall_chk_rdr *trc_rdrp = arg; - if (task_curr(t)) + if (task_curr(t) && cpu_online(task_cpu(t))) return false; // It is running, so decline to inspect it. trc_rdrp->nesting = READ_ONCE(t->trc_reader_nesting); trc_rdrp->ipi_to_cpu = READ_ONCE(t->trc_ipi_to_cpu); - trc_rdrp->needqs = READ_ONCE(t->trc_reader_special.b.need_qs); + trc_rdrp->needqs = rcu_ld_need_qs(t); return true; } @@ -1450,18 +1570,21 @@ static void show_stalled_task_trace(struct task_struct *t, bool *firstreport) } cpu = task_cpu(t); if (!task_call_func(t, trc_check_slow_task, &trc_rdr)) - pr_alert("P%d: %c\n", + pr_alert("P%d: %c%c\n", t->pid, + ".I"[t->trc_ipi_to_cpu >= 0], ".i"[is_idle_tsk]); else - pr_alert("P%d: %c%c%c nesting: %d%c cpu: %d\n", + pr_alert("P%d: %c%c%c%c nesting: %d%c%c cpu: %d%s\n", t->pid, ".I"[trc_rdr.ipi_to_cpu >= 0], ".i"[is_idle_tsk], ".N"[cpu >= 0 && tick_nohz_full_cpu(cpu)], + ".B"[!!data_race(t->trc_reader_special.b.blocked)], trc_rdr.nesting, - " N"[!!trc_rdr.needqs], - cpu); + " !CN"[trc_rdr.needqs & 0x3], + " ?"[trc_rdr.needqs > 0x3], + cpu, cpu_online(cpu) ? "" : "(offline)"); sched_show_task(t); } @@ -1481,18 +1604,18 @@ static void check_all_holdout_tasks_trace(struct list_head *hop, { struct task_struct *g, *t; - // Disable CPU hotplug across the holdout list scan. + // Disable CPU hotplug across the holdout list scan for IPIs. cpus_read_lock(); list_for_each_entry_safe(t, g, hop, trc_holdout_list) { // If safe and needed, try to check the current task. if (READ_ONCE(t->trc_ipi_to_cpu) == -1 && - !READ_ONCE(t->trc_reader_checked)) + !(rcu_ld_need_qs(t) & TRC_NEED_QS_CHECKED)) trc_wait_for_one_reader(t, hop); // If check succeeded, remove this task from the list. if (smp_load_acquire(&t->trc_ipi_to_cpu) == -1 && - READ_ONCE(t->trc_reader_checked)) + rcu_ld_need_qs(t) == TRC_NEED_QS_CHECKED) trc_del_holdout(t); else if (needreport) show_stalled_task_trace(t, firstreport); @@ -1516,10 +1639,6 @@ static void rcu_tasks_trace_empty_fn(void *unused) static void rcu_tasks_trace_postgp(struct rcu_tasks *rtp) { int cpu; - bool firstreport; - struct task_struct *g, *t; - LIST_HEAD(holdouts); - long ret; // Wait for any lingering IPI handlers to complete. Note that // if a CPU has gone offline or transitioned to userspace in the @@ -1530,37 +1649,6 @@ static void rcu_tasks_trace_postgp(struct rcu_tasks *rtp) if (WARN_ON_ONCE(smp_load_acquire(per_cpu_ptr(&trc_ipi_to_cpu, cpu)))) smp_call_function_single(cpu, rcu_tasks_trace_empty_fn, NULL, 1); - // Remove the safety count. - smp_mb__before_atomic(); // Order vs. earlier atomics - atomic_dec(&trc_n_readers_need_end); - smp_mb__after_atomic(); // Order vs. later atomics - - // Wait for readers. - set_tasks_gp_state(rtp, RTGS_WAIT_READERS); - for (;;) { - ret = wait_event_idle_exclusive_timeout( - trc_wait, - atomic_read(&trc_n_readers_need_end) == 0, - READ_ONCE(rcu_task_stall_timeout)); - if (ret) - break; // Count reached zero. - // Stall warning time, so make a list of the offenders. - rcu_read_lock(); - for_each_process_thread(g, t) - if (READ_ONCE(t->trc_reader_special.b.need_qs)) - trc_add_holdout(t, &holdouts); - rcu_read_unlock(); - firstreport = true; - list_for_each_entry_safe(t, g, &holdouts, trc_holdout_list) { - if (READ_ONCE(t->trc_reader_special.b.need_qs)) - show_stalled_task_trace(t, &firstreport); - trc_del_holdout(t); // Release task_struct reference. - } - if (firstreport) - pr_err("INFO: rcu_tasks_trace detected stalls? (Counter/taskslist mismatch?)\n"); - show_stalled_ipi_trace(); - pr_err("\t%d holdouts\n", atomic_read(&trc_n_readers_need_end)); - } smp_mb(); // Caller's code must be ordered after wakeup. // Pairs with pretty much every ordering primitive. } @@ -1568,11 +1656,14 @@ static void rcu_tasks_trace_postgp(struct rcu_tasks *rtp) /* Report any needed quiescent state for this exiting task. */ static void exit_tasks_rcu_finish_trace(struct task_struct *t) { - WRITE_ONCE(t->trc_reader_checked, true); + union rcu_special trs = READ_ONCE(t->trc_reader_special); + + rcu_trc_cmpxchg_need_qs(t, 0, TRC_NEED_QS_CHECKED); WARN_ON_ONCE(READ_ONCE(t->trc_reader_nesting)); - WRITE_ONCE(t->trc_reader_nesting, 0); - if (WARN_ON_ONCE(READ_ONCE(t->trc_reader_special.b.need_qs))) + if (WARN_ON_ONCE(rcu_ld_need_qs(t) & TRC_NEED_QS || trs.b.blocked)) rcu_read_unlock_trace_special(t); + else + WRITE_ONCE(t->trc_reader_nesting, 0); } /** @@ -1646,7 +1737,6 @@ static int __init rcu_spawn_tasks_trace_kthread(void) rcu_tasks_trace.init_fract = 1; } rcu_tasks_trace.pregp_func = rcu_tasks_trace_pregp_step; - rcu_tasks_trace.pertask_func = rcu_tasks_trace_pertask; rcu_tasks_trace.postscan_func = rcu_tasks_trace_postscan; rcu_tasks_trace.holdouts_func = check_all_holdout_tasks_trace; rcu_tasks_trace.postgp_func = rcu_tasks_trace_postgp; @@ -1659,7 +1749,8 @@ void show_rcu_tasks_trace_gp_kthread(void) { char buf[64]; - sprintf(buf, "N%d h:%lu/%lu/%lu", atomic_read(&trc_n_readers_need_end), + sprintf(buf, "N%lu h:%lu/%lu/%lu", + data_race(n_trc_holdouts), data_race(n_heavy_reader_ofl_updates), data_race(n_heavy_reader_updates), data_race(n_heavy_reader_attempts)); @@ -1686,23 +1777,24 @@ struct rcu_tasks_test_desc { struct rcu_head rh; const char *name; bool notrun; + unsigned long runstart; }; static struct rcu_tasks_test_desc tests[] = { { .name = "call_rcu_tasks()", /* If not defined, the test is skipped. */ - .notrun = !IS_ENABLED(CONFIG_TASKS_RCU), + .notrun = IS_ENABLED(CONFIG_TASKS_RCU), }, { .name = "call_rcu_tasks_rude()", /* If not defined, the test is skipped. */ - .notrun = !IS_ENABLED(CONFIG_TASKS_RUDE_RCU), + .notrun = IS_ENABLED(CONFIG_TASKS_RUDE_RCU), }, { .name = "call_rcu_tasks_trace()", /* If not defined, the test is skipped. */ - .notrun = !IS_ENABLED(CONFIG_TASKS_TRACE_RCU) + .notrun = IS_ENABLED(CONFIG_TASKS_TRACE_RCU) } }; @@ -1713,46 +1805,85 @@ static void test_rcu_tasks_callback(struct rcu_head *rhp) pr_info("Callback from %s invoked.\n", rttd->name); - rttd->notrun = true; + rttd->notrun = false; } static void rcu_tasks_initiate_self_tests(void) { + unsigned long j = jiffies; + pr_info("Running RCU-tasks wait API self tests\n"); #ifdef CONFIG_TASKS_RCU + tests[0].runstart = j; synchronize_rcu_tasks(); call_rcu_tasks(&tests[0].rh, test_rcu_tasks_callback); #endif #ifdef CONFIG_TASKS_RUDE_RCU + tests[1].runstart = j; synchronize_rcu_tasks_rude(); call_rcu_tasks_rude(&tests[1].rh, test_rcu_tasks_callback); #endif #ifdef CONFIG_TASKS_TRACE_RCU + tests[2].runstart = j; synchronize_rcu_tasks_trace(); call_rcu_tasks_trace(&tests[2].rh, test_rcu_tasks_callback); #endif } +/* + * Return: 0 - test passed + * 1 - test failed, but have not timed out yet + * -1 - test failed and timed out + */ static int rcu_tasks_verify_self_tests(void) { int ret = 0; int i; + unsigned long bst = rcu_task_stall_timeout; + if (bst <= 0 || bst > RCU_TASK_BOOT_STALL_TIMEOUT) + bst = RCU_TASK_BOOT_STALL_TIMEOUT; for (i = 0; i < ARRAY_SIZE(tests); i++) { - if (!tests[i].notrun) { // still hanging. - pr_err("%s has been failed.\n", tests[i].name); - ret = -1; + while (tests[i].notrun) { // still hanging. + if (time_after(jiffies, tests[i].runstart + bst)) { + pr_err("%s has failed boot-time tests.\n", tests[i].name); + ret = -1; + break; + } + ret = 1; + break; } } - - if (ret) - WARN_ON(1); + WARN_ON(ret < 0); return ret; } -late_initcall(rcu_tasks_verify_self_tests); + +/* + * Repeat the rcu_tasks_verify_self_tests() call once every second until the + * test passes or has timed out. + */ +static struct delayed_work rcu_tasks_verify_work; +static void rcu_tasks_verify_work_fn(struct work_struct *work __maybe_unused) +{ + int ret = rcu_tasks_verify_self_tests(); + + if (ret <= 0) + return; + + /* Test fails but not timed out yet, reschedule another check */ + schedule_delayed_work(&rcu_tasks_verify_work, HZ); +} + +static int rcu_tasks_verify_schedule_work(void) +{ + INIT_DELAYED_WORK(&rcu_tasks_verify_work, rcu_tasks_verify_work_fn); + rcu_tasks_verify_work_fn(NULL); + return 0; +} +late_initcall(rcu_tasks_verify_schedule_work); #else /* #ifdef CONFIG_PROVE_RCU */ static void rcu_tasks_initiate_self_tests(void) { } #endif /* #else #ifdef CONFIG_PROVE_RCU */ diff --git a/kernel/rcu/tiny.c b/kernel/rcu/tiny.c index 340b3f8b090d..f0561ee16b9c 100644 --- a/kernel/rcu/tiny.c +++ b/kernel/rcu/tiny.c @@ -58,7 +58,7 @@ void rcu_qs(void) rcu_ctrlblk.donetail = rcu_ctrlblk.curtail; raise_softirq_irqoff(RCU_SOFTIRQ); } - WRITE_ONCE(rcu_ctrlblk.gp_seq, rcu_ctrlblk.gp_seq + 1); + WRITE_ONCE(rcu_ctrlblk.gp_seq, rcu_ctrlblk.gp_seq + 2); local_irq_restore(flags); } @@ -139,8 +139,10 @@ static __latent_entropy void rcu_process_callbacks(struct softirq_action *unused /* * Wait for a grace period to elapse. But it is illegal to invoke * synchronize_rcu() from within an RCU read-side critical section. - * Therefore, any legal call to synchronize_rcu() is a quiescent - * state, and so on a UP system, synchronize_rcu() need do nothing. + * Therefore, any legal call to synchronize_rcu() is a quiescent state, + * and so on a UP system, synchronize_rcu() need do nothing, other than + * let the polled APIs know that another grace period elapsed. + * * (But Lai Jiangshan points out the benefits of doing might_sleep() * to reduce latency.) * @@ -152,6 +154,7 @@ void synchronize_rcu(void) lock_is_held(&rcu_lock_map) || lock_is_held(&rcu_sched_lock_map), "Illegal synchronize_rcu() in RCU read-side critical section"); + WRITE_ONCE(rcu_ctrlblk.gp_seq, rcu_ctrlblk.gp_seq + 2); } EXPORT_SYMBOL_GPL(synchronize_rcu); @@ -213,10 +216,24 @@ EXPORT_SYMBOL_GPL(start_poll_synchronize_rcu); */ bool poll_state_synchronize_rcu(unsigned long oldstate) { - return READ_ONCE(rcu_ctrlblk.gp_seq) != oldstate; + return oldstate == RCU_GET_STATE_COMPLETED || READ_ONCE(rcu_ctrlblk.gp_seq) != oldstate; } EXPORT_SYMBOL_GPL(poll_state_synchronize_rcu); +#ifdef CONFIG_KASAN_GENERIC +void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func) +{ + if (head) { + void *ptr = (void *) head - (unsigned long) func; + + kasan_record_aux_stack_noalloc(ptr); + } + + __kvfree_call_rcu(head, func); +} +EXPORT_SYMBOL_GPL(kvfree_call_rcu); +#endif + void __init rcu_init(void) { open_softirq(RCU_SOFTIRQ, rcu_process_callbacks); diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index c25ba442044a..62e514713f7a 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -154,7 +154,11 @@ static void sync_sched_exp_online_cleanup(int cpu); static void check_cb_ovld_locked(struct rcu_data *rdp, struct rcu_node *rnp); static bool rcu_rdp_is_offloaded(struct rcu_data *rdp); -/* rcuc/rcub/rcuop kthread realtime priority */ +/* + * rcuc/rcub/rcuop kthread realtime priority. The "rcuop" + * real-time priority(enabling/disabling) is controlled by + * the extra CONFIG_RCU_NOCB_CPU_CB_BOOST configuration. + */ static int kthread_prio = IS_ENABLED(CONFIG_RCU_BOOST) ? 1 : 0; module_param(kthread_prio, int, 0444); @@ -1775,6 +1779,79 @@ static void rcu_strict_gp_boundary(void *unused) invoke_rcu_core(); } +// Has rcu_init() been invoked? This is used (for example) to determine +// whether spinlocks may be acquired safely. +static bool rcu_init_invoked(void) +{ + return !!rcu_state.n_online_cpus; +} + +// Make the polled API aware of the beginning of a grace period. +static void rcu_poll_gp_seq_start(unsigned long *snap) +{ + struct rcu_node *rnp = rcu_get_root(); + + if (rcu_init_invoked()) + raw_lockdep_assert_held_rcu_node(rnp); + + // If RCU was idle, note beginning of GP. + if (!rcu_seq_state(rcu_state.gp_seq_polled)) + rcu_seq_start(&rcu_state.gp_seq_polled); + + // Either way, record current state. + *snap = rcu_state.gp_seq_polled; +} + +// Make the polled API aware of the end of a grace period. +static void rcu_poll_gp_seq_end(unsigned long *snap) +{ + struct rcu_node *rnp = rcu_get_root(); + + if (rcu_init_invoked()) + raw_lockdep_assert_held_rcu_node(rnp); + + // If the previously noted GP is still in effect, record the + // end of that GP. Either way, zero counter to avoid counter-wrap + // problems. + if (*snap && *snap == rcu_state.gp_seq_polled) { + rcu_seq_end(&rcu_state.gp_seq_polled); + rcu_state.gp_seq_polled_snap = 0; + rcu_state.gp_seq_polled_exp_snap = 0; + } else { + *snap = 0; + } +} + +// Make the polled API aware of the beginning of a grace period, but +// where caller does not hold the root rcu_node structure's lock. +static void rcu_poll_gp_seq_start_unlocked(unsigned long *snap) +{ + struct rcu_node *rnp = rcu_get_root(); + + if (rcu_init_invoked()) { + lockdep_assert_irqs_enabled(); + raw_spin_lock_irq_rcu_node(rnp); + } + rcu_poll_gp_seq_start(snap); + if (rcu_init_invoked()) + raw_spin_unlock_irq_rcu_node(rnp); +} + +// Make the polled API aware of the end of a grace period, but where +// caller does not hold the root rcu_node structure's lock. +static void rcu_poll_gp_seq_end_unlocked(unsigned long *snap) +{ + struct rcu_node *rnp = rcu_get_root(); + + if (rcu_init_invoked()) { + lockdep_assert_irqs_enabled(); + raw_spin_lock_irq_rcu_node(rnp); + } + rcu_poll_gp_seq_end(snap); + if (rcu_init_invoked()) + raw_spin_unlock_irq_rcu_node(rnp); +} + /* * Initialize a new grace period. Return false if no grace period required. */ @@ -1810,6 +1887,7 @@ static noinline_for_stack bool rcu_gp_init(void) rcu_seq_start(&rcu_state.gp_seq); ASSERT_EXCLUSIVE_WRITER(rcu_state.gp_seq); trace_rcu_grace_period(rcu_state.name, rcu_state.gp_seq, TPS("start")); + rcu_poll_gp_seq_start(&rcu_state.gp_seq_polled_snap); raw_spin_unlock_irq_rcu_node(rnp); /* @@ -1971,19 +2049,23 @@ static void rcu_gp_fqs(bool first_time) */ static noinline_for_stack void rcu_gp_fqs_loop(void) { - bool first_gp_fqs; + bool first_gp_fqs = true; int gf = 0; unsigned long j; int ret; struct rcu_node *rnp = rcu_get_root(); - first_gp_fqs = true; j = READ_ONCE(jiffies_till_first_fqs); if (rcu_state.cbovld) gf = RCU_GP_FLAG_OVLD; ret = 0; for (;;) { - if (!ret) { + if (rcu_state.cbovld) { + j = (j + 2) / 3; + if (j <= 0) + j = 1; + } + if (!ret || time_before(jiffies + j, rcu_state.jiffies_force_qs)) { WRITE_ONCE(rcu_state.jiffies_force_qs, jiffies + j); /* * jiffies_force_qs before RCU_GP_WAIT_FQS state @@ -2001,7 +2083,15 @@ static noinline_for_stack void rcu_gp_fqs_loop(void) rcu_gp_torture_wait(); WRITE_ONCE(rcu_state.gp_state, RCU_GP_DOING_FQS); /* Locking provides needed memory barriers. */ - /* If grace period done, leave loop. */ + /* + * Exit the loop if the root rcu_node structure indicates that the grace period + * has ended, leave the loop. The rcu_preempt_blocked_readers_cgp(rnp) check + * is required only for single-node rcu_node trees because readers blocking + * the current grace period are queued only on leaf rcu_node structures. + * For multi-node trees, checking the root node's ->qsmask suffices, because a + * given root node's ->qsmask bit is cleared only when all CPUs and tasks from + * the corresponding leaf nodes have passed through their quiescent state. + */ if (!READ_ONCE(rnp->qsmask) && !rcu_preempt_blocked_readers_cgp(rnp)) break; @@ -2069,6 +2159,7 @@ static noinline void rcu_gp_cleanup(void) * safe for us to drop the lock in order to mark the grace * period as completed in all of the rcu_node structures. */ + rcu_poll_gp_seq_end(&rcu_state.gp_seq_polled_snap); raw_spin_unlock_irq_rcu_node(rnp); /* @@ -2530,7 +2621,7 @@ static void rcu_do_batch(struct rcu_data *rdp) trace_rcu_batch_end(rcu_state.name, 0, !rcu_segcblist_empty(&rdp->cblist), need_resched(), is_idle_task(current), - rcu_is_callbacks_kthread()); + rcu_is_callbacks_kthread(rdp)); return; } @@ -2608,7 +2699,7 @@ static void rcu_do_batch(struct rcu_data *rdp) rcu_nocb_lock_irqsave(rdp, flags); rdp->n_cbs_invoked += count; trace_rcu_batch_end(rcu_state.name, count, !!rcl.head, need_resched(), - is_idle_task(current), rcu_is_callbacks_kthread()); + is_idle_task(current), rcu_is_callbacks_kthread(rdp)); /* Update counts and requeue any remaining callbacks. */ rcu_segcblist_insert_done_cbs(&rdp->cblist, &rcl); @@ -3211,7 +3302,6 @@ struct kfree_rcu_cpu_work { * @krw_arr: Array of batches of kfree_rcu() objects waiting for a grace period * @lock: Synchronize access to this structure * @monitor_work: Promote @head to @head_free after KFREE_DRAIN_JIFFIES - * @monitor_todo: Tracks whether a @monitor_work delayed work is pending * @initialized: The @rcu_work fields have been initialized * @count: Number of objects for which GP not started * @bkvcache: @@ -3236,7 +3326,6 @@ struct kfree_rcu_cpu { struct kfree_rcu_cpu_work krw_arr[KFREE_N_BATCHES]; raw_spinlock_t lock; struct delayed_work monitor_work; - bool monitor_todo; bool initialized; int count; @@ -3416,6 +3505,18 @@ static void kfree_rcu_work(struct work_struct *work) } } +static bool +need_offload_krc(struct kfree_rcu_cpu *krcp) +{ + int i; + + for (i = 0; i < FREE_N_CHANNELS; i++) + if (krcp->bkvhead[i]) + return true; + + return !!krcp->head; +} + /* * This function is invoked after the KFREE_DRAIN_JIFFIES timeout. */ @@ -3472,9 +3573,7 @@ static void kfree_rcu_monitor(struct work_struct *work) // of the channels that is still busy we should rearm the // work to repeat an attempt. Because previous batches are // still in progress. - if (!krcp->bkvhead[0] && !krcp->bkvhead[1] && !krcp->head) - krcp->monitor_todo = false; - else + if (need_offload_krc(krcp)) schedule_delayed_work(&krcp->monitor_work, KFREE_DRAIN_JIFFIES); raw_spin_unlock_irqrestore(&krcp->lock, flags); @@ -3662,11 +3761,8 @@ void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func) WRITE_ONCE(krcp->count, krcp->count + 1); // Set timer to drain after KFREE_DRAIN_JIFFIES. - if (rcu_scheduler_active == RCU_SCHEDULER_RUNNING && - !krcp->monitor_todo) { - krcp->monitor_todo = true; + if (rcu_scheduler_active == RCU_SCHEDULER_RUNNING) schedule_delayed_work(&krcp->monitor_work, KFREE_DRAIN_JIFFIES); - } unlock_return: krc_this_cpu_unlock(krcp, flags); @@ -3741,14 +3837,8 @@ void __init kfree_rcu_scheduler_running(void) struct kfree_rcu_cpu *krcp = per_cpu_ptr(&krc, cpu); raw_spin_lock_irqsave(&krcp->lock, flags); - if ((!krcp->bkvhead[0] && !krcp->bkvhead[1] && !krcp->head) || - krcp->monitor_todo) { - raw_spin_unlock_irqrestore(&krcp->lock, flags); - continue; - } - krcp->monitor_todo = true; - schedule_delayed_work_on(cpu, &krcp->monitor_work, - KFREE_DRAIN_JIFFIES); + if (need_offload_krc(krcp)) + schedule_delayed_work_on(cpu, &krcp->monitor_work, KFREE_DRAIN_JIFFIES); raw_spin_unlock_irqrestore(&krcp->lock, flags); } } @@ -3837,8 +3927,18 @@ void synchronize_rcu(void) lock_is_held(&rcu_lock_map) || lock_is_held(&rcu_sched_lock_map), "Illegal synchronize_rcu() in RCU read-side critical section"); - if (rcu_blocking_is_gp()) + if (rcu_blocking_is_gp()) { + // Note well that this code runs with !PREEMPT && !SMP. + // In addition, all code that advances grace periods runs at + // process level. Therefore, this normal GP overlaps with + // other normal GPs only by being fully nested within them, + // which allows reuse of ->gp_seq_polled_snap. + rcu_poll_gp_seq_start_unlocked(&rcu_state.gp_seq_polled_snap); + rcu_poll_gp_seq_end_unlocked(&rcu_state.gp_seq_polled_snap); + if (rcu_init_invoked()) + cond_resched_tasks_rcu_qs(); return; // Context allows vacuous grace periods. + } if (rcu_gp_is_expedited()) synchronize_rcu_expedited(); else @@ -3860,7 +3960,7 @@ unsigned long get_state_synchronize_rcu(void) * before the load from ->gp_seq. */ smp_mb(); /* ^^^ */ - return rcu_seq_snap(&rcu_state.gp_seq); + return rcu_seq_snap(&rcu_state.gp_seq_polled); } EXPORT_SYMBOL_GPL(get_state_synchronize_rcu); @@ -3889,7 +3989,13 @@ unsigned long start_poll_synchronize_rcu(void) rdp = this_cpu_ptr(&rcu_data); rnp = rdp->mynode; raw_spin_lock_rcu_node(rnp); // irqs already disabled. - needwake = rcu_start_this_gp(rnp, rdp, gp_seq); + // Note it is possible for a grace period to have elapsed between + // the above call to get_state_synchronize_rcu() and the below call + // to rcu_seq_snap. This is OK, the worst that happens is that we + // get a grace period that no one needed. These accesses are ordered + // by smp_mb(), and we are accessing them in the opposite order + // from which they are updated at grace-period start, as required. + needwake = rcu_start_this_gp(rnp, rdp, rcu_seq_snap(&rcu_state.gp_seq)); raw_spin_unlock_irqrestore_rcu_node(rnp, flags); if (needwake) rcu_gp_kthread_wake(); @@ -3911,7 +4017,7 @@ EXPORT_SYMBOL_GPL(start_poll_synchronize_rcu); * * Yes, this function does not take counter wrap into account. * But counter wrap is harmless. If the counter wraps, we have waited for - * more than 2 billion grace periods (and way more on a 64-bit system!). + * more than a billion grace periods (and way more on a 64-bit system!). * Those needing to keep oldstate values for very long time periods * (many hours even on 32-bit systems) should check them occasionally * and either refresh them or set a flag indicating that the grace period @@ -3924,7 +4030,8 @@ EXPORT_SYMBOL_GPL(start_poll_synchronize_rcu); */ bool poll_state_synchronize_rcu(unsigned long oldstate) { - if (rcu_seq_done(&rcu_state.gp_seq, oldstate)) { + if (oldstate == RCU_GET_STATE_COMPLETED || + rcu_seq_done_exact(&rcu_state.gp_seq_polled, oldstate)) { smp_mb(); /* Ensure GP ends before subsequent accesses. */ return true; } @@ -3935,20 +4042,20 @@ EXPORT_SYMBOL_GPL(poll_state_synchronize_rcu); /** * cond_synchronize_rcu - Conditionally wait for an RCU grace period * - * @oldstate: value from get_state_synchronize_rcu() or start_poll_synchronize_rcu() + * @oldstate: value from get_state_synchronize_rcu(), start_poll_synchronize_rcu(), or start_poll_synchronize_rcu_expedited() * * If a full RCU grace period has elapsed since the earlier call to * get_state_synchronize_rcu() or start_poll_synchronize_rcu(), just return. * Otherwise, invoke synchronize_rcu() to wait for a full grace period. * - * Yes, this function does not take counter wrap into account. But - * counter wrap is harmless. If the counter wraps, we have waited for + * Yes, this function does not take counter wrap into account. + * But counter wrap is harmless. If the counter wraps, we have waited for * more than 2 billion grace periods (and way more on a 64-bit system!), - * so waiting for one additional grace period should be just fine. + * so waiting for a couple of additional grace periods should be just fine. * * This function provides the same memory-ordering guarantees that * would be provided by a synchronize_rcu() that was invoked at the call - * to the function that provided @oldstate, and that returned at the end + * to the function that provided @oldstate and that returned at the end * of this function. */ void cond_synchronize_rcu(unsigned long oldstate) @@ -4441,6 +4548,7 @@ void rcu_report_dead(unsigned int cpu) rdp->rcu_ofl_gp_flags = READ_ONCE(rcu_state.gp_flags); if (rnp->qsmask & mask) { /* RCU waiting on outgoing CPU? */ /* Report quiescent state -before- changing ->qsmaskinitnext! */ + rcu_disable_urgency_upon_qs(rdp); rcu_report_qs_rnp(mask, rnp, rnp->gp_seq, flags); raw_spin_lock_irqsave_rcu_node(rnp, flags); } @@ -4486,6 +4594,7 @@ void rcutree_migrate_callbacks(int cpu) needwake = needwake || rcu_advance_cbs(my_rnp, my_rdp); rcu_segcblist_disable(&rdp->cblist); WARN_ON_ONCE(rcu_segcblist_empty(&my_rdp->cblist) != !rcu_segcblist_n_cbs(&my_rdp->cblist)); + check_cb_ovld_locked(my_rdp, my_rnp); if (rcu_rdp_is_offloaded(my_rdp)) { raw_spin_unlock_rcu_node(my_rnp); /* irqs remain disabled. */ __call_rcu_nocb_wake(my_rdp, true, flags); @@ -4701,6 +4810,9 @@ static void __init rcu_init_one(void) init_waitqueue_head(&rnp->exp_wq[3]); spin_lock_init(&rnp->exp_lock); mutex_init(&rnp->boost_kthread_mutex); + raw_spin_lock_init(&rnp->exp_poll_lock); + rnp->exp_seq_poll_rq = RCU_GET_STATE_COMPLETED; + INIT_WORK(&rnp->exp_poll_wq, sync_rcu_do_polled_gp); } } @@ -4926,6 +5038,10 @@ void __init rcu_init(void) qovld_calc = DEFAULT_RCU_QOVLD_MULT * qhimark; else qovld_calc = qovld; + + // Kick-start any polled grace periods that started early. + if (!(per_cpu_ptr(&rcu_data, cpu)->mynode->exp_seq_poll_rq & 0x1)) + (void)start_poll_synchronize_rcu_expedited(); } #include "tree_stall.h" diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h index 2ccf5845957d..3cdc18997a38 100644 --- a/kernel/rcu/tree.h +++ b/kernel/rcu/tree.h @@ -133,6 +133,10 @@ struct rcu_node { wait_queue_head_t exp_wq[4]; struct rcu_exp_work rew; bool exp_need_flush; /* Need to flush workitem? */ + raw_spinlock_t exp_poll_lock; + /* Lock and data for polled expedited grace periods. */ + unsigned long exp_seq_poll_rq; + struct work_struct exp_poll_wq; } ____cacheline_internodealigned_in_smp; /* @@ -235,6 +239,7 @@ struct rcu_data { * if rdp_gp. */ struct list_head nocb_entry_rdp; /* rcu_data node in wakeup chain. */ + struct rcu_data *nocb_toggling_rdp; /* rdp queued for (de-)offloading */ /* The following fields are used by CB kthread, hence new cacheline. */ struct rcu_data *nocb_gp_rdp ____cacheline_internodealigned_in_smp; @@ -323,6 +328,9 @@ struct rcu_state { short gp_state; /* GP kthread sleep state. */ unsigned long gp_wake_time; /* Last GP kthread wake. */ unsigned long gp_wake_seq; /* ->gp_seq at ^^^. */ + unsigned long gp_seq_polled; /* GP seq for polled API. */ + unsigned long gp_seq_polled_snap; /* ->gp_seq_polled at normal GP start. */ + unsigned long gp_seq_polled_exp_snap; /* ->gp_seq_polled at expedited GP start. */ /* End of fields guarded by root rcu_node's lock. */ @@ -425,7 +433,7 @@ static void rcu_flavor_sched_clock_irq(int user); static void dump_blkd_tasks(struct rcu_node *rnp, int ncheck); static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags); static void rcu_preempt_boost_start_gp(struct rcu_node *rnp); -static bool rcu_is_callbacks_kthread(void); +static bool rcu_is_callbacks_kthread(struct rcu_data *rdp); static void rcu_cpu_kthread_setup(unsigned int cpu); static void rcu_spawn_one_boost_kthread(struct rcu_node *rnp); static bool rcu_preempt_has_tasks(struct rcu_node *rnp); @@ -481,3 +489,6 @@ static void rcu_iw_handler(struct irq_work *iwp); static void check_cpu_stall(struct rcu_data *rdp); static void rcu_check_gp_start_stall(struct rcu_node *rnp, struct rcu_data *rdp, const unsigned long gpssdelay); + +/* Forward declarations for tree_exp.h. */ +static void sync_rcu_do_polled_gp(struct work_struct *wp); diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h index 0f70f62039a9..f092c7f18a5f 100644 --- a/kernel/rcu/tree_exp.h +++ b/kernel/rcu/tree_exp.h @@ -18,6 +18,7 @@ static int rcu_print_task_exp_stall(struct rcu_node *rnp); static void rcu_exp_gp_seq_start(void) { rcu_seq_start(&rcu_state.expedited_sequence); + rcu_poll_gp_seq_start_unlocked(&rcu_state.gp_seq_polled_exp_snap); } /* @@ -34,6 +35,7 @@ static __maybe_unused unsigned long rcu_exp_gp_seq_endval(void) */ static void rcu_exp_gp_seq_end(void) { + rcu_poll_gp_seq_end_unlocked(&rcu_state.gp_seq_polled_exp_snap); rcu_seq_end(&rcu_state.expedited_sequence); smp_mb(); /* Ensure that consecutive grace periods serialize. */ } @@ -621,7 +623,6 @@ static void synchronize_rcu_expedited_wait(void) return; if (rcu_stall_is_suppressed()) continue; - panic_on_rcu_stall(); trace_rcu_stall_warning(rcu_state.name, TPS("ExpeditedStall")); pr_err("INFO: %s detected expedited stalls on CPUs/tasks: {", rcu_state.name); @@ -636,10 +637,11 @@ static void synchronize_rcu_expedited_wait(void) continue; ndetected++; rdp = per_cpu_ptr(&rcu_data, cpu); - pr_cont(" %d-%c%c%c", cpu, + pr_cont(" %d-%c%c%c%c", cpu, "O."[!!cpu_online(cpu)], "o."[!!(rdp->grpmask & rnp->expmaskinit)], - "N."[!!(rdp->grpmask & rnp->expmaskinitnext)]); + "N."[!!(rdp->grpmask & rnp->expmaskinitnext)], + "D."[!!(rdp->cpu_no_qs.b.exp)]); } } pr_cont(" } %lu jiffies s: %lu root: %#lx/%c\n", @@ -669,6 +671,7 @@ static void synchronize_rcu_expedited_wait(void) } } jiffies_stall = 3 * rcu_exp_jiffies_till_stall_check() + 3; + panic_on_rcu_stall(); } } @@ -913,8 +916,18 @@ void synchronize_rcu_expedited(void) "Illegal synchronize_rcu_expedited() in RCU read-side critical section"); /* Is the state is such that the call is a grace period? */ - if (rcu_blocking_is_gp()) - return; + if (rcu_blocking_is_gp()) { + // Note well that this code runs with !PREEMPT && !SMP. + // In addition, all code that advances grace periods runs + // at process level. Therefore, this expedited GP overlaps + // with other expedited GPs only by being fully nested within + // them, which allows reuse of ->gp_seq_polled_exp_snap. + rcu_poll_gp_seq_start_unlocked(&rcu_state.gp_seq_polled_exp_snap); + rcu_poll_gp_seq_end_unlocked(&rcu_state.gp_seq_polled_exp_snap); + if (rcu_init_invoked()) + cond_resched(); + return; // Context allows vacuous grace periods. + } /* If expedited grace periods are prohibited, fall back to normal. */ if (rcu_gp_is_normal()) { @@ -950,3 +963,93 @@ void synchronize_rcu_expedited(void) synchronize_rcu_expedited_destroy_work(&rew); } EXPORT_SYMBOL_GPL(synchronize_rcu_expedited); + +/* + * Ensure that start_poll_synchronize_rcu_expedited() has the expedited + * RCU grace periods that it needs. + */ +static void sync_rcu_do_polled_gp(struct work_struct *wp) +{ + unsigned long flags; + int i = 0; + struct rcu_node *rnp = container_of(wp, struct rcu_node, exp_poll_wq); + unsigned long s; + + raw_spin_lock_irqsave(&rnp->exp_poll_lock, flags); + s = rnp->exp_seq_poll_rq; + rnp->exp_seq_poll_rq = RCU_GET_STATE_COMPLETED; + raw_spin_unlock_irqrestore(&rnp->exp_poll_lock, flags); + if (s == RCU_GET_STATE_COMPLETED) + return; + while (!poll_state_synchronize_rcu(s)) { + synchronize_rcu_expedited(); + if (i == 10 || i == 20) + pr_info("%s: i = %d s = %lx gp_seq_polled = %lx\n", __func__, i, s, READ_ONCE(rcu_state.gp_seq_polled)); + i++; + } + raw_spin_lock_irqsave(&rnp->exp_poll_lock, flags); + s = rnp->exp_seq_poll_rq; + if (poll_state_synchronize_rcu(s)) + rnp->exp_seq_poll_rq = RCU_GET_STATE_COMPLETED; + raw_spin_unlock_irqrestore(&rnp->exp_poll_lock, flags); +} + +/** + * start_poll_synchronize_rcu_expedited - Snapshot current RCU state and start expedited grace period + * + * Returns a cookie to pass to a call to cond_synchronize_rcu(), + * cond_synchronize_rcu_expedited(), or poll_state_synchronize_rcu(), + * allowing them to determine whether or not any sort of grace period has + * elapsed in the meantime. If the needed expedited grace period is not + * already slated to start, initiates that grace period. + */ +unsigned long start_poll_synchronize_rcu_expedited(void) +{ + unsigned long flags; + struct rcu_data *rdp; + struct rcu_node *rnp; + unsigned long s; + + s = get_state_synchronize_rcu(); + rdp = per_cpu_ptr(&rcu_data, raw_smp_processor_id()); + rnp = rdp->mynode; + if (rcu_init_invoked()) + raw_spin_lock_irqsave(&rnp->exp_poll_lock, flags); + if (!poll_state_synchronize_rcu(s)) { + rnp->exp_seq_poll_rq = s; + if (rcu_init_invoked()) + queue_work(rcu_gp_wq, &rnp->exp_poll_wq); + } + if (rcu_init_invoked()) + raw_spin_unlock_irqrestore(&rnp->exp_poll_lock, flags); + + return s; +} +EXPORT_SYMBOL_GPL(start_poll_synchronize_rcu_expedited); + +/** + * cond_synchronize_rcu_expedited - Conditionally wait for an expedited RCU grace period + * + * @oldstate: value from get_state_synchronize_rcu(), start_poll_synchronize_rcu(), or start_poll_synchronize_rcu_expedited() + * + * If any type of full RCU grace period has elapsed since the earlier + * call to get_state_synchronize_rcu(), start_poll_synchronize_rcu(), + * or start_poll_synchronize_rcu_expedited(), just return. Otherwise, + * invoke synchronize_rcu_expedited() to wait for a full grace period. + * + * Yes, this function does not take counter wrap into account. + * But counter wrap is harmless. If the counter wraps, we have waited for + * more than 2 billion grace periods (and way more on a 64-bit system!), + * so waiting for a couple of additional grace periods should be just fine. + * + * This function provides the same memory-ordering guarantees that + * would be provided by a synchronize_rcu() that was invoked at the call + * to the function that provided @oldstate and that returned at the end + * of this function. + */ +void cond_synchronize_rcu_expedited(unsigned long oldstate) +{ + if (!poll_state_synchronize_rcu(oldstate)) + synchronize_rcu_expedited(); +} +EXPORT_SYMBOL_GPL(cond_synchronize_rcu_expedited); diff --git a/kernel/rcu/tree_nocb.h b/kernel/rcu/tree_nocb.h index 46694e13398a..a8f574d8850d 100644 --- a/kernel/rcu/tree_nocb.h +++ b/kernel/rcu/tree_nocb.h @@ -546,52 +546,51 @@ static void __call_rcu_nocb_wake(struct rcu_data *rdp, bool was_alldone, } } -/* - * Check if we ignore this rdp. - * - * We check that without holding the nocb lock but - * we make sure not to miss a freshly offloaded rdp - * with the current ordering: - * - * rdp_offload_toggle() nocb_gp_enabled_cb() - * ------------------------- ---------------------------- - * WRITE flags LOCK nocb_gp_lock - * LOCK nocb_gp_lock READ/WRITE nocb_gp_sleep - * READ/WRITE nocb_gp_sleep UNLOCK nocb_gp_lock - * UNLOCK nocb_gp_lock READ flags - */ -static inline bool nocb_gp_enabled_cb(struct rcu_data *rdp) -{ - u8 flags = SEGCBLIST_OFFLOADED | SEGCBLIST_KTHREAD_GP; - - return rcu_segcblist_test_flags(&rdp->cblist, flags); -} - -static inline bool nocb_gp_update_state_deoffloading(struct rcu_data *rdp, - bool *needwake_state) +static int nocb_gp_toggle_rdp(struct rcu_data *rdp, + bool *wake_state) { struct rcu_segcblist *cblist = &rdp->cblist; + unsigned long flags; + int ret; - if (rcu_segcblist_test_flags(cblist, SEGCBLIST_OFFLOADED)) { - if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP)) { - rcu_segcblist_set_flags(cblist, SEGCBLIST_KTHREAD_GP); - if (rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB)) - *needwake_state = true; - } - return false; + rcu_nocb_lock_irqsave(rdp, flags); + if (rcu_segcblist_test_flags(cblist, SEGCBLIST_OFFLOADED) && + !rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP)) { + /* + * Offloading. Set our flag and notify the offload worker. + * We will handle this rdp until it ever gets de-offloaded. + */ + rcu_segcblist_set_flags(cblist, SEGCBLIST_KTHREAD_GP); + if (rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB)) + *wake_state = true; + ret = 1; + } else if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_OFFLOADED) && + rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP)) { + /* + * De-offloading. Clear our flag and notify the de-offload worker. + * We will ignore this rdp until it ever gets re-offloaded. + */ + rcu_segcblist_clear_flags(cblist, SEGCBLIST_KTHREAD_GP); + if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB)) + *wake_state = true; + ret = 0; + } else { + WARN_ON_ONCE(1); + ret = -1; } - /* - * De-offloading. Clear our flag and notify the de-offload worker. - * We will ignore this rdp until it ever gets re-offloaded. - */ - WARN_ON_ONCE(!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP)); - rcu_segcblist_clear_flags(cblist, SEGCBLIST_KTHREAD_GP); - if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB)) - *needwake_state = true; - return true; + rcu_nocb_unlock_irqrestore(rdp, flags); + + return ret; } +static void nocb_gp_sleep(struct rcu_data *my_rdp, int cpu) +{ + trace_rcu_nocb_wake(rcu_state.name, cpu, TPS("Sleep")); + swait_event_interruptible_exclusive(my_rdp->nocb_gp_wq, + !READ_ONCE(my_rdp->nocb_gp_sleep)); + trace_rcu_nocb_wake(rcu_state.name, cpu, TPS("EndSleep")); +} /* * No-CBs GP kthreads come here to wait for additional callbacks to show up @@ -609,7 +608,7 @@ static void nocb_gp_wait(struct rcu_data *my_rdp) bool needwait_gp = false; // This prevents actual uninitialized use. bool needwake; bool needwake_gp; - struct rcu_data *rdp; + struct rcu_data *rdp, *rdp_toggling = NULL; struct rcu_node *rnp; unsigned long wait_gp_seq = 0; // Suppress "use uninitialized" warning. bool wasempty = false; @@ -634,19 +633,10 @@ static void nocb_gp_wait(struct rcu_data *my_rdp) * is added to the list, so the skipped-over rcu_data structures * won't be ignored for long. */ - list_for_each_entry_rcu(rdp, &my_rdp->nocb_head_rdp, nocb_entry_rdp, 1) { - bool needwake_state = false; - - if (!nocb_gp_enabled_cb(rdp)) - continue; + list_for_each_entry(rdp, &my_rdp->nocb_head_rdp, nocb_entry_rdp) { trace_rcu_nocb_wake(rcu_state.name, rdp->cpu, TPS("Check")); rcu_nocb_lock_irqsave(rdp, flags); - if (nocb_gp_update_state_deoffloading(rdp, &needwake_state)) { - rcu_nocb_unlock_irqrestore(rdp, flags); - if (needwake_state) - swake_up_one(&rdp->nocb_state_wq); - continue; - } + lockdep_assert_held(&rdp->nocb_lock); bypass_ncbs = rcu_cblist_n_cbs(&rdp->nocb_bypass); if (bypass_ncbs && (time_after(j, READ_ONCE(rdp->nocb_bypass_first) + 1) || @@ -656,8 +646,6 @@ static void nocb_gp_wait(struct rcu_data *my_rdp) bypass_ncbs = rcu_cblist_n_cbs(&rdp->nocb_bypass); } else if (!bypass_ncbs && rcu_segcblist_empty(&rdp->cblist)) { rcu_nocb_unlock_irqrestore(rdp, flags); - if (needwake_state) - swake_up_one(&rdp->nocb_state_wq); continue; /* No callbacks here, try next. */ } if (bypass_ncbs) { @@ -705,8 +693,6 @@ static void nocb_gp_wait(struct rcu_data *my_rdp) } if (needwake_gp) rcu_gp_kthread_wake(); - if (needwake_state) - swake_up_one(&rdp->nocb_state_wq); } my_rdp->nocb_gp_bypass = bypass; @@ -723,13 +709,19 @@ static void nocb_gp_wait(struct rcu_data *my_rdp) /* Polling, so trace if first poll in the series. */ if (gotcbs) trace_rcu_nocb_wake(rcu_state.name, cpu, TPS("Poll")); - schedule_timeout_idle(1); + if (list_empty(&my_rdp->nocb_head_rdp)) { + raw_spin_lock_irqsave(&my_rdp->nocb_gp_lock, flags); + if (!my_rdp->nocb_toggling_rdp) + WRITE_ONCE(my_rdp->nocb_gp_sleep, true); + raw_spin_unlock_irqrestore(&my_rdp->nocb_gp_lock, flags); + /* Wait for any offloading rdp */ + nocb_gp_sleep(my_rdp, cpu); + } else { + schedule_timeout_idle(1); + } } else if (!needwait_gp) { /* Wait for callbacks to appear. */ - trace_rcu_nocb_wake(rcu_state.name, cpu, TPS("Sleep")); - swait_event_interruptible_exclusive(my_rdp->nocb_gp_wq, - !READ_ONCE(my_rdp->nocb_gp_sleep)); - trace_rcu_nocb_wake(rcu_state.name, cpu, TPS("EndSleep")); + nocb_gp_sleep(my_rdp, cpu); } else { rnp = my_rdp->mynode; trace_rcu_this_gp(rnp, my_rdp, wait_gp_seq, TPS("StartWait")); @@ -739,15 +731,49 @@ static void nocb_gp_wait(struct rcu_data *my_rdp) !READ_ONCE(my_rdp->nocb_gp_sleep)); trace_rcu_this_gp(rnp, my_rdp, wait_gp_seq, TPS("EndWait")); } + if (!rcu_nocb_poll) { raw_spin_lock_irqsave(&my_rdp->nocb_gp_lock, flags); + // (De-)queue an rdp to/from the group if its nocb state is changing + rdp_toggling = my_rdp->nocb_toggling_rdp; + if (rdp_toggling) + my_rdp->nocb_toggling_rdp = NULL; + if (my_rdp->nocb_defer_wakeup > RCU_NOCB_WAKE_NOT) { WRITE_ONCE(my_rdp->nocb_defer_wakeup, RCU_NOCB_WAKE_NOT); del_timer(&my_rdp->nocb_timer); } WRITE_ONCE(my_rdp->nocb_gp_sleep, true); raw_spin_unlock_irqrestore(&my_rdp->nocb_gp_lock, flags); + } else { + rdp_toggling = READ_ONCE(my_rdp->nocb_toggling_rdp); + if (rdp_toggling) { + /* + * Paranoid locking to make sure nocb_toggling_rdp is well + * reset *before* we (re)set SEGCBLIST_KTHREAD_GP or we could + * race with another round of nocb toggling for this rdp. + * Nocb locking should prevent from that already but we stick + * to paranoia, especially in rare path. + */ + raw_spin_lock_irqsave(&my_rdp->nocb_gp_lock, flags); + my_rdp->nocb_toggling_rdp = NULL; + raw_spin_unlock_irqrestore(&my_rdp->nocb_gp_lock, flags); + } + } + + if (rdp_toggling) { + bool wake_state = false; + int ret; + + ret = nocb_gp_toggle_rdp(rdp_toggling, &wake_state); + if (ret == 1) + list_add_tail(&rdp_toggling->nocb_entry_rdp, &my_rdp->nocb_head_rdp); + else if (ret == 0) + list_del(&rdp_toggling->nocb_entry_rdp); + if (wake_state) + swake_up_one(&rdp_toggling->nocb_state_wq); } + my_rdp->nocb_gp_seq = -1; WARN_ON(signal_pending(current)); } @@ -966,16 +992,15 @@ static int rdp_offload_toggle(struct rcu_data *rdp, swake_up_one(&rdp->nocb_cb_wq); raw_spin_lock_irqsave(&rdp_gp->nocb_gp_lock, flags); + // Queue this rdp for add/del to/from the list to iterate on rcuog + WRITE_ONCE(rdp_gp->nocb_toggling_rdp, rdp); if (rdp_gp->nocb_gp_sleep) { rdp_gp->nocb_gp_sleep = false; wake_gp = true; } raw_spin_unlock_irqrestore(&rdp_gp->nocb_gp_lock, flags); - if (wake_gp) - wake_up_process(rdp_gp->nocb_gp_kthread); - - return 0; + return wake_gp; } static long rcu_nocb_rdp_deoffload(void *arg) @@ -983,9 +1008,15 @@ static long rcu_nocb_rdp_deoffload(void *arg) struct rcu_data *rdp = arg; struct rcu_segcblist *cblist = &rdp->cblist; unsigned long flags; - int ret; + int wake_gp; + struct rcu_data *rdp_gp = rdp->nocb_gp_rdp; - WARN_ON_ONCE(rdp->cpu != raw_smp_processor_id()); + /* + * rcu_nocb_rdp_deoffload() may be called directly if + * rcuog/o[p] spawn failed, because at this time the rdp->cpu + * is not online yet. + */ + WARN_ON_ONCE((rdp->cpu != raw_smp_processor_id()) && cpu_online(rdp->cpu)); pr_info("De-offloading %d\n", rdp->cpu); @@ -1009,12 +1040,41 @@ static long rcu_nocb_rdp_deoffload(void *arg) */ rcu_segcblist_set_flags(cblist, SEGCBLIST_RCU_CORE); invoke_rcu_core(); - ret = rdp_offload_toggle(rdp, false, flags); - swait_event_exclusive(rdp->nocb_state_wq, - !rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB | - SEGCBLIST_KTHREAD_GP)); - /* Stop nocb_gp_wait() from iterating over this structure. */ - list_del_rcu(&rdp->nocb_entry_rdp); + wake_gp = rdp_offload_toggle(rdp, false, flags); + + mutex_lock(&rdp_gp->nocb_gp_kthread_mutex); + if (rdp_gp->nocb_gp_kthread) { + if (wake_gp) + wake_up_process(rdp_gp->nocb_gp_kthread); + + /* + * If rcuo[p] kthread spawn failed, directly remove SEGCBLIST_KTHREAD_CB. + * Just wait SEGCBLIST_KTHREAD_GP to be cleared by rcuog. + */ + if (!rdp->nocb_cb_kthread) { + rcu_nocb_lock_irqsave(rdp, flags); + rcu_segcblist_clear_flags(&rdp->cblist, SEGCBLIST_KTHREAD_CB); + rcu_nocb_unlock_irqrestore(rdp, flags); + } + + swait_event_exclusive(rdp->nocb_state_wq, + !rcu_segcblist_test_flags(cblist, + SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP)); + } else { + /* + * No kthread to clear the flags for us or remove the rdp from the nocb list + * to iterate. Do it here instead. Locking doesn't look stricly necessary + * but we stick to paranoia in this rare path. + */ + rcu_nocb_lock_irqsave(rdp, flags); + rcu_segcblist_clear_flags(&rdp->cblist, + SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP); + rcu_nocb_unlock_irqrestore(rdp, flags); + + list_del(&rdp->nocb_entry_rdp); + } + mutex_unlock(&rdp_gp->nocb_gp_kthread_mutex); + /* * Lock one last time to acquire latest callback updates from kthreads * so we can later handle callbacks locally without locking. @@ -1035,7 +1095,7 @@ static long rcu_nocb_rdp_deoffload(void *arg) WARN_ON_ONCE(rcu_cblist_n_cbs(&rdp->nocb_bypass)); - return ret; + return 0; } int rcu_nocb_cpu_deoffload(int cpu) @@ -1043,8 +1103,8 @@ int rcu_nocb_cpu_deoffload(int cpu) struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu); int ret = 0; - mutex_lock(&rcu_state.barrier_mutex); cpus_read_lock(); + mutex_lock(&rcu_state.barrier_mutex); if (rcu_rdp_is_offloaded(rdp)) { if (cpu_online(cpu)) { ret = work_on_cpu(cpu, rcu_nocb_rdp_deoffload, rdp); @@ -1055,8 +1115,8 @@ int rcu_nocb_cpu_deoffload(int cpu) ret = -EINVAL; } } - cpus_read_unlock(); mutex_unlock(&rcu_state.barrier_mutex); + cpus_read_unlock(); return ret; } @@ -1067,7 +1127,8 @@ static long rcu_nocb_rdp_offload(void *arg) struct rcu_data *rdp = arg; struct rcu_segcblist *cblist = &rdp->cblist; unsigned long flags; - int ret; + int wake_gp; + struct rcu_data *rdp_gp = rdp->nocb_gp_rdp; WARN_ON_ONCE(rdp->cpu != raw_smp_processor_id()); /* @@ -1077,17 +1138,10 @@ static long rcu_nocb_rdp_offload(void *arg) if (!rdp->nocb_gp_rdp) return -EINVAL; - pr_info("Offloading %d\n", rdp->cpu); + if (WARN_ON_ONCE(!rdp_gp->nocb_gp_kthread)) + return -EINVAL; - /* - * Cause future nocb_gp_wait() invocations to iterate over - * structure, resetting ->nocb_gp_sleep and waking up the related - * "rcuog". Since nocb_gp_wait() in turn locks ->nocb_gp_lock - * before setting ->nocb_gp_sleep again, we are guaranteed to - * iterate this newly added structure before "rcuog" goes to - * sleep again. - */ - list_add_tail_rcu(&rdp->nocb_entry_rdp, &rdp->nocb_gp_rdp->nocb_head_rdp); + pr_info("Offloading %d\n", rdp->cpu); /* * Can't use rcu_nocb_lock_irqsave() before SEGCBLIST_LOCKING @@ -1111,7 +1165,9 @@ static long rcu_nocb_rdp_offload(void *arg) * WRITE flags READ callbacks * rcu_nocb_unlock() rcu_nocb_unlock() */ - ret = rdp_offload_toggle(rdp, true, flags); + wake_gp = rdp_offload_toggle(rdp, true, flags); + if (wake_gp) + wake_up_process(rdp_gp->nocb_gp_kthread); swait_event_exclusive(rdp->nocb_state_wq, rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB) && rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP)); @@ -1124,7 +1180,7 @@ static long rcu_nocb_rdp_offload(void *arg) rcu_segcblist_clear_flags(cblist, SEGCBLIST_RCU_CORE); rcu_nocb_unlock_irqrestore(rdp, flags); - return ret; + return 0; } int rcu_nocb_cpu_offload(int cpu) @@ -1132,8 +1188,8 @@ int rcu_nocb_cpu_offload(int cpu) struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu); int ret = 0; - mutex_lock(&rcu_state.barrier_mutex); cpus_read_lock(); + mutex_lock(&rcu_state.barrier_mutex); if (!rcu_rdp_is_offloaded(rdp)) { if (cpu_online(cpu)) { ret = work_on_cpu(cpu, rcu_nocb_rdp_offload, rdp); @@ -1144,8 +1200,8 @@ int rcu_nocb_cpu_offload(int cpu) ret = -EINVAL; } } - cpus_read_unlock(); mutex_unlock(&rcu_state.barrier_mutex); + cpus_read_unlock(); return ret; } @@ -1155,11 +1211,21 @@ void __init rcu_init_nohz(void) { int cpu; bool need_rcu_nocb_mask = false; + bool offload_all = false; struct rcu_data *rdp; +#if defined(CONFIG_RCU_NOCB_CPU_DEFAULT_ALL) + if (!rcu_state.nocb_is_setup) { + need_rcu_nocb_mask = true; + offload_all = true; + } +#endif /* #if defined(CONFIG_RCU_NOCB_CPU_DEFAULT_ALL) */ + #if defined(CONFIG_NO_HZ_FULL) - if (tick_nohz_full_running && !cpumask_empty(tick_nohz_full_mask)) + if (tick_nohz_full_running && !cpumask_empty(tick_nohz_full_mask)) { need_rcu_nocb_mask = true; + offload_all = false; /* NO_HZ_FULL has its own mask. */ + } #endif /* #if defined(CONFIG_NO_HZ_FULL) */ if (need_rcu_nocb_mask) { @@ -1180,6 +1246,9 @@ void __init rcu_init_nohz(void) cpumask_or(rcu_nocb_mask, rcu_nocb_mask, tick_nohz_full_mask); #endif /* #if defined(CONFIG_NO_HZ_FULL) */ + if (offload_all) + cpumask_setall(rcu_nocb_mask); + if (!cpumask_subset(rcu_nocb_mask, cpu_possible_mask)) { pr_info("\tNote: kernel parameter 'rcu_nocbs=', 'nohz_full', or 'isolcpus=' contains nonexistent CPUs.\n"); cpumask_and(rcu_nocb_mask, cpu_possible_mask, @@ -1246,7 +1315,7 @@ static void rcu_spawn_cpu_nocb_kthread(int cpu) "rcuog/%d", rdp_gp->cpu); if (WARN_ONCE(IS_ERR(t), "%s: Could not start rcuo GP kthread, OOM is now expected behavior\n", __func__)) { mutex_unlock(&rdp_gp->nocb_gp_kthread_mutex); - return; + goto end; } WRITE_ONCE(rdp_gp->nocb_gp_kthread, t); if (kthread_prio) @@ -1258,12 +1327,21 @@ static void rcu_spawn_cpu_nocb_kthread(int cpu) t = kthread_run(rcu_nocb_cb_kthread, rdp, "rcuo%c/%d", rcu_state.abbr, cpu); if (WARN_ONCE(IS_ERR(t), "%s: Could not start rcuo CB kthread, OOM is now expected behavior\n", __func__)) - return; + goto end; - if (kthread_prio) + if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_CB_BOOST) && kthread_prio) sched_setscheduler_nocheck(t, SCHED_FIFO, &sp); + WRITE_ONCE(rdp->nocb_cb_kthread, t); WRITE_ONCE(rdp->nocb_gp_kthread, rdp_gp->nocb_gp_kthread); + return; +end: + mutex_lock(&rcu_state.barrier_mutex); + if (rcu_rdp_is_offloaded(rdp)) { + rcu_nocb_rdp_deoffload(rdp); + cpumask_clear_cpu(cpu, rcu_nocb_mask); + } + mutex_unlock(&rcu_state.barrier_mutex); } /* How many CB CPU IDs per GP kthread? Default of -1 for sqrt(nr_cpu_ids). */ diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index c8ba0fe17267..7ae1551479a2 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -460,7 +460,7 @@ static bool rcu_preempt_has_tasks(struct rcu_node *rnp) * be quite short, for example, in the case of the call from * rcu_read_unlock_special(). */ -static void +static notrace void rcu_preempt_deferred_qs_irqrestore(struct task_struct *t, unsigned long flags) { bool empty_exp; @@ -581,7 +581,7 @@ rcu_preempt_deferred_qs_irqrestore(struct task_struct *t, unsigned long flags) * is disabled. This function cannot be expected to understand these * nuances, so the caller must handle them. */ -static bool rcu_preempt_need_deferred_qs(struct task_struct *t) +static notrace bool rcu_preempt_need_deferred_qs(struct task_struct *t) { return (__this_cpu_read(rcu_data.cpu_no_qs.b.exp) || READ_ONCE(t->rcu_read_unlock_special.s)) && @@ -595,7 +595,7 @@ static bool rcu_preempt_need_deferred_qs(struct task_struct *t) * evaluate safety in terms of interrupt, softirq, and preemption * disabling. */ -static void rcu_preempt_deferred_qs(struct task_struct *t) +static notrace void rcu_preempt_deferred_qs(struct task_struct *t) { unsigned long flags; @@ -899,8 +899,8 @@ void rcu_note_context_switch(bool preempt) this_cpu_write(rcu_data.rcu_urgent_qs, false); if (unlikely(raw_cpu_read(rcu_data.rcu_need_heavy_qs))) rcu_momentary_dyntick_idle(); - rcu_tasks_qs(current, preempt); out: + rcu_tasks_qs(current, preempt); trace_rcu_utilization(TPS("End context switch")); } EXPORT_SYMBOL_GPL(rcu_note_context_switch); @@ -926,7 +926,7 @@ static bool rcu_preempt_has_tasks(struct rcu_node *rnp) * Because there is no preemptible RCU, there can be no deferred quiescent * states. */ -static bool rcu_preempt_need_deferred_qs(struct task_struct *t) +static notrace bool rcu_preempt_need_deferred_qs(struct task_struct *t) { return false; } @@ -935,7 +935,7 @@ static bool rcu_preempt_need_deferred_qs(struct task_struct *t) // period for a quiescent state from this CPU. Note that requests from // tasks are handled when removing the task from the blocked-tasks list // below. -static void rcu_preempt_deferred_qs(struct task_struct *t) +static notrace void rcu_preempt_deferred_qs(struct task_struct *t) { struct rcu_data *rdp = this_cpu_ptr(&rcu_data); @@ -1012,6 +1012,25 @@ static void rcu_cpu_kthread_setup(unsigned int cpu) WRITE_ONCE(rdp->rcuc_activity, jiffies); } +static bool rcu_is_callbacks_nocb_kthread(struct rcu_data *rdp) +{ +#ifdef CONFIG_RCU_NOCB_CPU + return rdp->nocb_cb_kthread == current; +#else + return false; +#endif +} + +/* + * Is the current CPU running the RCU-callbacks kthread? + * Caller must have preemption disabled. + */ +static bool rcu_is_callbacks_kthread(struct rcu_data *rdp) +{ + return rdp->rcu_cpu_kthread_task == current || + rcu_is_callbacks_nocb_kthread(rdp); +} + #ifdef CONFIG_RCU_BOOST /* @@ -1140,7 +1159,8 @@ static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags) (rnp->gp_tasks != NULL && rnp->boost_tasks == NULL && rnp->qsmask == 0 && - (!time_after(rnp->boost_time, jiffies) || rcu_state.cbovld))) { + (!time_after(rnp->boost_time, jiffies) || rcu_state.cbovld || + IS_ENABLED(CONFIG_RCU_STRICT_GRACE_PERIOD)))) { if (rnp->exp_tasks == NULL) WRITE_ONCE(rnp->boost_tasks, rnp->gp_tasks); raw_spin_unlock_irqrestore_rcu_node(rnp, flags); @@ -1151,15 +1171,6 @@ static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags) } } -/* - * Is the current CPU running the RCU-callbacks kthread? - * Caller must have preemption disabled. - */ -static bool rcu_is_callbacks_kthread(void) -{ - return __this_cpu_read(rcu_data.rcu_cpu_kthread_task) == current; -} - #define RCU_BOOST_DELAY_JIFFIES DIV_ROUND_UP(CONFIG_RCU_BOOST_DELAY * HZ, 1000) /* @@ -1242,11 +1253,6 @@ static void rcu_initiate_boost(struct rcu_node *rnp, unsigned long flags) raw_spin_unlock_irqrestore_rcu_node(rnp, flags); } -static bool rcu_is_callbacks_kthread(void) -{ - return false; -} - static void rcu_preempt_boost_start_gp(struct rcu_node *rnp) { } diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c index fc7fef575606..2e93acad1e31 100644 --- a/kernel/rcu/update.c +++ b/kernel/rcu/update.c @@ -516,6 +516,19 @@ int rcu_cpu_stall_suppress_at_boot __read_mostly; // !0 = suppress boot stalls. EXPORT_SYMBOL_GPL(rcu_cpu_stall_suppress_at_boot); module_param(rcu_cpu_stall_suppress_at_boot, int, 0444); +/** + * get_completed_synchronize_rcu - Return a pre-completed polled state cookie + * + * Returns a value that will always be treated by functions like + * poll_state_synchronize_rcu() as a cookie whose grace period has already + * completed. + */ +unsigned long get_completed_synchronize_rcu(void) +{ + return RCU_GET_STATE_COMPLETED; +} +EXPORT_SYMBOL_GPL(get_completed_synchronize_rcu); + #ifdef CONFIG_PROVE_RCU /* diff --git a/kernel/sched/core.c b/kernel/sched/core.c index da0bf6fe9ecd..9568019be124 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -4264,6 +4264,38 @@ int task_call_func(struct task_struct *p, task_call_f func, void *arg) } /** + * cpu_curr_snapshot - Return a snapshot of the currently running task + * @cpu: The CPU on which to snapshot the task. + * + * Returns the task_struct pointer of the task "currently" running on + * the specified CPU. If the same task is running on that CPU throughout, + * the return value will be a pointer to that task's task_struct structure. + * If the CPU did any context switches even vaguely concurrently with the + * execution of this function, the return value will be a pointer to the + * task_struct structure of a randomly chosen task that was running on + * that CPU somewhere around the time that this function was executing. + * + * If the specified CPU was offline, the return value is whatever it + * is, perhaps a pointer to the task_struct structure of that CPU's idle + * task, but there is no guarantee. Callers wishing a useful return + * value must take some action to ensure that the specified CPU remains + * online throughout. + * + * This function executes full memory barriers before and after fetching + * the pointer, which permits the caller to confine this function's fetch + * with respect to the caller's accesses to other shared variables. + */ +struct task_struct *cpu_curr_snapshot(int cpu) +{ + struct task_struct *t; + + smp_mb(); /* Pairing determined by caller's synchronization design. */ + t = rcu_dereference(cpu_curr(cpu)); + smp_mb(); /* Pairing determined by caller's synchronization design. */ + return t; +} + +/** * wake_up_process - Wake up a specific process * @p: The process to be woken up. * diff --git a/kernel/smp.c b/kernel/smp.c index dd215f439426..650810a6f29b 100644 --- a/kernel/smp.c +++ b/kernel/smp.c @@ -174,9 +174,9 @@ static int __init csdlock_debug(char *str) if (val) static_branch_enable(&csdlock_debug_enabled); - return 0; + return 1; } -early_param("csdlock_debug", csdlock_debug); +__setup("csdlock_debug=", csdlock_debug); static DEFINE_PER_CPU(call_single_data_t *, cur_csd); static DEFINE_PER_CPU(smp_call_func_t, cur_csd_func); diff --git a/tools/testing/selftests/rcutorture/bin/kvm-check-branches.sh b/tools/testing/selftests/rcutorture/bin/kvm-check-branches.sh index f17000a2ccf1..ed0ec7f0927e 100755 --- a/tools/testing/selftests/rcutorture/bin/kvm-check-branches.sh +++ b/tools/testing/selftests/rcutorture/bin/kvm-check-branches.sh @@ -35,7 +35,7 @@ then exit 1 fi -# Remember where we started so that we can get back and the end. +# Remember where we started so that we can get back at the end. curcommit="`git status | head -1 | awk '{ print $NF }'`" nfail=0 @@ -73,15 +73,10 @@ do # Test the specified commit. git checkout $i > $resdir/$ds/$idir/git-checkout.out 2>&1 echo git checkout return code: $? "(Commit $ntry: $i)" - kvm.sh --allcpus --duration 3 --trust-make > $resdir/$ds/$idir/kvm.sh.out 2>&1 + kvm.sh --allcpus --duration 3 --trust-make --datestamp "$ds/$idir" > $resdir/$ds/$idir/kvm.sh.out 2>&1 ret=$? echo kvm.sh return code $ret for commit $i from branch $gitbr - - # Move the build products to their resting place. - runresdir="`grep -m 1 '^Results directory:' < $resdir/$ds/$idir/kvm.sh.out | sed -e 's/^Results directory://'`" - mv $runresdir $resdir/$ds/$idir - rrd="`echo $runresdir | sed -e 's,^.*/,,'`" - echo Run results: $resdir/$ds/$idir/$rrd + echo Run results: $resdir/$ds/$idir if test "$ret" -ne 0 then # Failure, so leave all evidence intact. diff --git a/tools/testing/selftests/rcutorture/bin/kvm-remote.sh b/tools/testing/selftests/rcutorture/bin/kvm-remote.sh index 0ff59bd8b640..9f0a5d5ff2dd 100755 --- a/tools/testing/selftests/rcutorture/bin/kvm-remote.sh +++ b/tools/testing/selftests/rcutorture/bin/kvm-remote.sh @@ -262,6 +262,7 @@ echo All batches started. `date` | tee -a "$oldrun/remote-log" # Wait for all remaining scenarios to complete and collect results. for i in $systems do + echo " ---" Waiting for $i `date` | tee -a "$oldrun/remote-log" while checkremotefile "$i" "$resdir/$ds/remote.run" do sleep 30 diff --git a/tools/testing/selftests/rcutorture/bin/kvm.sh b/tools/testing/selftests/rcutorture/bin/kvm.sh index 263e16aeca0e..6c734818a875 100755 --- a/tools/testing/selftests/rcutorture/bin/kvm.sh +++ b/tools/testing/selftests/rcutorture/bin/kvm.sh @@ -164,7 +164,7 @@ do shift ;; --gdb) - TORTURE_KCONFIG_GDB_ARG="CONFIG_DEBUG_INFO=y"; export TORTURE_KCONFIG_GDB_ARG + TORTURE_KCONFIG_GDB_ARG="CONFIG_DEBUG_INFO_NONE=n CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y"; export TORTURE_KCONFIG_GDB_ARG TORTURE_BOOT_GDB_ARG="nokaslr"; export TORTURE_BOOT_GDB_ARG TORTURE_QEMU_GDB_ARG="-s -S"; export TORTURE_QEMU_GDB_ARG ;; @@ -180,7 +180,7 @@ do shift ;; --kasan) - TORTURE_KCONFIG_KASAN_ARG="CONFIG_DEBUG_INFO=y CONFIG_KASAN=y"; export TORTURE_KCONFIG_KASAN_ARG + TORTURE_KCONFIG_KASAN_ARG="CONFIG_DEBUG_INFO_NONE=n CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y CONFIG_KASAN=y"; export TORTURE_KCONFIG_KASAN_ARG if test -n "$torture_qemu_mem_default" then TORTURE_QEMU_MEM=2G @@ -192,7 +192,7 @@ do shift ;; --kcsan) - TORTURE_KCONFIG_KCSAN_ARG="CONFIG_DEBUG_INFO=y CONFIG_KCSAN=y CONFIG_KCSAN_STRICT=y CONFIG_KCSAN_REPORT_ONCE_IN_MS=100000 CONFIG_KCSAN_VERBOSE=y CONFIG_DEBUG_LOCK_ALLOC=y CONFIG_PROVE_LOCKING=y"; export TORTURE_KCONFIG_KCSAN_ARG + TORTURE_KCONFIG_KCSAN_ARG="CONFIG_DEBUG_INFO_NONE=n CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT=y CONFIG_KCSAN=y CONFIG_KCSAN_STRICT=y CONFIG_KCSAN_REPORT_ONCE_IN_MS=100000 CONFIG_KCSAN_VERBOSE=y CONFIG_DEBUG_LOCK_ALLOC=y CONFIG_PROVE_LOCKING=y"; export TORTURE_KCONFIG_KCSAN_ARG ;; --kmake-arg|--kmake-args) checkarg --kmake-arg "(kernel make arguments)" $# "$2" '.*' '^error$' |