diff options
| -rw-r--r-- | Documentation/RCU/Design/Requirements/Requirements.rst | 36 | ||||
| -rw-r--r-- | Documentation/RCU/whatisRCU.rst | 1 | ||||
| -rw-r--r-- | Documentation/admin-guide/kernel-parameters.txt | 127 | ||||
| -rw-r--r-- | Documentation/core-api/kernel-api.rst | 12 | ||||
| -rw-r--r-- | MAINTAINERS | 2 | ||||
| -rw-r--r-- | include/linux/notifier.h | 10 | ||||
| -rw-r--r-- | include/linux/rcupdate.h | 54 | ||||
| -rw-r--r-- | include/linux/srcu.h | 8 | ||||
| -rw-r--r-- | kernel/rcu/Kconfig | 18 | ||||
| -rw-r--r-- | kernel/rcu/rcu.h | 6 | ||||
| -rw-r--r-- | kernel/rcu/tasks.h | 12 | ||||
| -rw-r--r-- | kernel/rcu/tree.c | 131 | ||||
| -rw-r--r-- | kernel/rcu/tree_exp.h | 2 | ||||
| -rw-r--r-- | kernel/rcu/tree_nocb.h | 52 | ||||
| -rw-r--r-- | kernel/rcu/tree_plugin.h | 4 |
15 files changed, 274 insertions, 201 deletions
diff --git a/Documentation/RCU/Design/Requirements/Requirements.rst b/Documentation/RCU/Design/Requirements/Requirements.rst index 49387d823619..f3b605285a87 100644 --- a/Documentation/RCU/Design/Requirements/Requirements.rst +++ b/Documentation/RCU/Design/Requirements/Requirements.rst @@ -2071,41 +2071,7 @@ call. Because RCU avoids interrupting idle CPUs, it is illegal to execute an RCU read-side critical section on an idle CPU. (Kernels built with -``CONFIG_PROVE_RCU=y`` will splat if you try it.) The RCU_NONIDLE() -macro and ``_rcuidle`` event tracing is provided to work around this -restriction. In addition, rcu_is_watching() may be used to test -whether or not it is currently legal to run RCU read-side critical -sections on this CPU. I learned of the need for diagnostics on the one -hand and RCU_NONIDLE() on the other while inspecting idle-loop code. -Steven Rostedt supplied ``_rcuidle`` event tracing, which is used quite -heavily in the idle loop. However, there are some restrictions on the -code placed within RCU_NONIDLE(): - -#. Blocking is prohibited. In practice, this is not a serious - restriction given that idle tasks are prohibited from blocking to - begin with. -#. Although nesting RCU_NONIDLE() is permitted, they cannot nest - indefinitely deeply. However, given that they can be nested on the - order of a million deep, even on 32-bit systems, this should not be a - serious restriction. This nesting limit would probably be reached - long after the compiler OOMed or the stack overflowed. -#. Any code path that enters RCU_NONIDLE() must sequence out of that - same RCU_NONIDLE(). For example, the following is grossly - illegal: - - :: - - 1 RCU_NONIDLE({ - 2 do_something(); - 3 goto bad_idea; /* BUG!!! */ - 4 do_something_else();}); - 5 bad_idea: - - - It is just as illegal to transfer control into the middle of - RCU_NONIDLE()'s argument. Yes, in theory, you could transfer in - as long as you also transferred out, but in practice you could also - expect to get sharply worded review comments. +``CONFIG_PROVE_RCU=y`` will splat if you try it.) It is similarly socially unacceptable to interrupt an ``nohz_full`` CPU running in userspace. RCU must therefore track ``nohz_full`` userspace diff --git a/Documentation/RCU/whatisRCU.rst b/Documentation/RCU/whatisRCU.rst index 8eddef28d3a1..e488c8e557a9 100644 --- a/Documentation/RCU/whatisRCU.rst +++ b/Documentation/RCU/whatisRCU.rst @@ -1117,7 +1117,6 @@ All: lockdep-checked RCU utility APIs:: RCU_LOCKDEP_WARN rcu_sleep_check - RCU_NONIDLE All: Unchecked RCU-protected pointer access:: diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt index 6c8f630f4a91..5b544b3f00fc 100644 --- a/Documentation/admin-guide/kernel-parameters.txt +++ b/Documentation/admin-guide/kernel-parameters.txt @@ -4736,43 +4736,6 @@ the propagation of recent CPU-hotplug changes up the rcu_node combining tree. - rcutree.use_softirq= [KNL] - If set to zero, move all RCU_SOFTIRQ processing to - per-CPU rcuc kthreads. Defaults to a non-zero - value, meaning that RCU_SOFTIRQ is used by default. - Specify rcutree.use_softirq=0 to use rcuc kthreads. - - But note that CONFIG_PREEMPT_RT=y kernels disable - this kernel boot parameter, forcibly setting it - to zero. - - rcutree.rcu_fanout_exact= [KNL] - Disable autobalancing of the rcu_node combining - tree. This is used by rcutorture, and might - possibly be useful for architectures having high - cache-to-cache transfer latencies. - - rcutree.rcu_fanout_leaf= [KNL] - Change the number of CPUs assigned to each - leaf rcu_node structure. Useful for very - large systems, which will choose the value 64, - and for NUMA systems with large remote-access - latencies, which will choose a value aligned - with the appropriate hardware boundaries. - - rcutree.rcu_min_cached_objs= [KNL] - Minimum number of objects which are cached and - maintained per one CPU. Object size is equal - to PAGE_SIZE. The cache allows to reduce the - pressure to page allocator, also it makes the - whole algorithm to behave better in low memory - condition. - - rcutree.rcu_delay_page_cache_fill_msec= [KNL] - Set the page-cache refill delay (in milliseconds) - in response to low-memory conditions. The range - of permitted values is in the range 0:100000. - rcutree.jiffies_till_first_fqs= [KNL] Set delay from grace-period initialization to first attempt to force quiescent states. @@ -4811,21 +4774,6 @@ When RCU_NOCB_CPU is set, also adjust the priority of NOCB callback kthreads. - rcutree.rcu_divisor= [KNL] - Set the shift-right count to use to compute - the callback-invocation batch limit bl from - the number of callbacks queued on this CPU. - The result will be bounded below by the value of - the rcutree.blimit kernel parameter. Every bl - callbacks, the softirq handler will exit in - order to allow the CPU to do other work. - - Please note that this callback-invocation batch - limit applies only to non-offloaded callback - invocation. Offloaded callbacks are instead - invoked in the context of an rcuoc kthread, which - scheduler will preempt as it does any other task. - rcutree.nocb_nobypass_lim_per_jiffy= [KNL] On callback-offloaded (rcu_nocbs) CPUs, RCU reduces the lock contention that would @@ -4839,14 +4787,6 @@ the ->nocb_bypass queue. The definition of "too many" is supplied by this kernel boot parameter. - rcutree.rcu_nocb_gp_stride= [KNL] - Set the number of NOCB callback kthreads in - each group, which defaults to the square root - of the number of CPUs. Larger numbers reduce - the wakeup overhead on the global grace-period - kthread, but increases that same overhead on - each group's NOCB grace-period kthread. - rcutree.qhimark= [KNL] Set threshold of queued RCU callbacks beyond which batch limiting is disabled. @@ -4864,6 +4804,56 @@ on rcutree.qhimark at boot time and to zero to disable more aggressive help enlistment. + rcutree.rcu_delay_page_cache_fill_msec= [KNL] + Set the page-cache refill delay (in milliseconds) + in response to low-memory conditions. The range + of permitted values is in the range 0:100000. + + rcutree.rcu_divisor= [KNL] + Set the shift-right count to use to compute + the callback-invocation batch limit bl from + the number of callbacks queued on this CPU. + The result will be bounded below by the value of + the rcutree.blimit kernel parameter. Every bl + callbacks, the softirq handler will exit in + order to allow the CPU to do other work. + + Please note that this callback-invocation batch + limit applies only to non-offloaded callback + invocation. Offloaded callbacks are instead + invoked in the context of an rcuoc kthread, which + scheduler will preempt as it does any other task. + + rcutree.rcu_fanout_exact= [KNL] + Disable autobalancing of the rcu_node combining + tree. This is used by rcutorture, and might + possibly be useful for architectures having high + cache-to-cache transfer latencies. + + rcutree.rcu_fanout_leaf= [KNL] + Change the number of CPUs assigned to each + leaf rcu_node structure. Useful for very + large systems, which will choose the value 64, + and for NUMA systems with large remote-access + latencies, which will choose a value aligned + with the appropriate hardware boundaries. + + rcutree.rcu_min_cached_objs= [KNL] + Minimum number of objects which are cached and + maintained per one CPU. Object size is equal + to PAGE_SIZE. The cache allows to reduce the + pressure to page allocator, also it makes the + whole algorithm to behave better in low memory + condition. + + rcutree.rcu_nocb_gp_stride= [KNL] + Set the number of NOCB callback kthreads in + each group, which defaults to the square root + of the number of CPUs. Larger numbers reduce + the wakeup overhead on the global grace-period + kthread, but increases that same overhead on + each group's NOCB grace-period kthread. + rcutree.rcu_kick_kthreads= [KNL] Cause the grace-period kthread to get an extra wake_up() if it sleeps three times longer than @@ -4871,6 +4861,13 @@ This wake_up() will be accompanied by a WARN_ONCE() splat and an ftrace_dump(). + rcutree.rcu_resched_ns= [KNL] + Limit the time spend invoking a batch of RCU + callbacks to the specified number of nanoseconds. + By default, this limit is checked only once + every 32 callbacks in order to limit the pain + inflicted by local_clock() overhead. + rcutree.rcu_unlock_delay= [KNL] In CONFIG_RCU_STRICT_GRACE_PERIOD=y kernels, this specifies an rcu_read_unlock()-time delay @@ -4885,6 +4882,16 @@ rcu_node tree with an eye towards determining why a new grace period has not yet started. + rcutree.use_softirq= [KNL] + If set to zero, move all RCU_SOFTIRQ processing to + per-CPU rcuc kthreads. Defaults to a non-zero + value, meaning that RCU_SOFTIRQ is used by default. + Specify rcutree.use_softirq=0 to use rcuc kthreads. + + But note that CONFIG_PREEMPT_RT=y kernels disable + this kernel boot parameter, forcibly setting it + to zero. + rcuscale.gp_async= [KNL] Measure performance of asynchronous grace-period primitives such as call_rcu(). diff --git a/Documentation/core-api/kernel-api.rst b/Documentation/core-api/kernel-api.rst index 9b3f3e5f5a95..712e59ad32fa 100644 --- a/Documentation/core-api/kernel-api.rst +++ b/Documentation/core-api/kernel-api.rst @@ -412,3 +412,15 @@ Read-Copy Update (RCU) .. kernel-doc:: include/linux/rcu_sync.h .. kernel-doc:: kernel/rcu/sync.c + +.. kernel-doc:: kernel/rcu/tasks.h + +.. kernel-doc:: kernel/rcu/tree_stall.h + +.. kernel-doc:: include/linux/rcupdate_trace.h + +.. kernel-doc:: include/linux/rcupdate_wait.h + +.. kernel-doc:: include/linux/rcuref.h + +.. kernel-doc:: include/linux/rcutree.h diff --git a/MAINTAINERS b/MAINTAINERS index 7e0b87d5aa2e..8af0aabe9bf5 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -17778,7 +17778,7 @@ M: Boqun Feng <[email protected]> R: Steven Rostedt <[email protected]> R: Mathieu Desnoyers <[email protected]> R: Lai Jiangshan <[email protected]> -R: Zqiang <[email protected]> +R: Zqiang <[email protected]> S: Supported W: http://www.rdrop.com/users/paulmck/RCU/ diff --git a/include/linux/notifier.h b/include/linux/notifier.h index 2aba75145144..86544707236a 100644 --- a/include/linux/notifier.h +++ b/include/linux/notifier.h @@ -106,12 +106,22 @@ extern void srcu_init_notifier_head(struct srcu_notifier_head *nh); #define RAW_NOTIFIER_INIT(name) { \ .head = NULL } +#ifdef CONFIG_TREE_SRCU #define SRCU_NOTIFIER_INIT(name, pcpu) \ { \ .mutex = __MUTEX_INITIALIZER(name.mutex), \ .head = NULL, \ + .srcuu = __SRCU_USAGE_INIT(name.srcuu), \ .srcu = __SRCU_STRUCT_INIT(name.srcu, name.srcuu, pcpu), \ } +#else +#define SRCU_NOTIFIER_INIT(name, pcpu) \ + { \ + .mutex = __MUTEX_INITIALIZER(name.mutex), \ + .head = NULL, \ + .srcu = __SRCU_STRUCT_INIT(name.srcu, name.srcuu, pcpu), \ + } +#endif #define ATOMIC_NOTIFIER_HEAD(name) \ struct atomic_notifier_head name = \ diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h index dcd2cf1e8326..7d9c2a63b7cd 100644 --- a/include/linux/rcupdate.h +++ b/include/linux/rcupdate.h @@ -156,31 +156,6 @@ static inline int rcu_nocb_cpu_deoffload(int cpu) { return 0; } static inline void rcu_nocb_flush_deferred_wakeup(void) { } #endif /* #else #ifdef CONFIG_RCU_NOCB_CPU */ -/** - * RCU_NONIDLE - Indicate idle-loop code that needs RCU readers - * @a: Code that RCU needs to pay attention to. - * - * RCU read-side critical sections are forbidden in the inner idle loop, - * that is, between the ct_idle_enter() and the ct_idle_exit() -- RCU - * will happily ignore any such read-side critical sections. However, - * things like powertop need tracepoints in the inner idle loop. - * - * This macro provides the way out: RCU_NONIDLE(do_something_with_RCU()) - * will tell RCU that it needs to pay attention, invoke its argument - * (in this example, calling the do_something_with_RCU() function), - * and then tell RCU to go back to ignoring this CPU. It is permissible - * to nest RCU_NONIDLE() wrappers, but not indefinitely (but the limit is - * on the order of a million or so, even on 32-bit systems). It is - * not legal to block within RCU_NONIDLE(), nor is it permissible to - * transfer control either into or out of RCU_NONIDLE()'s statement. - */ -#define RCU_NONIDLE(a) \ - do { \ - ct_irq_enter_irqson(); \ - do { a; } while (0); \ - ct_irq_exit_irqson(); \ - } while (0) - /* * Note a quasi-voluntary context switch for RCU-tasks's benefit. * This is a macro rather than an inline function to avoid #include hell. @@ -957,9 +932,8 @@ static inline notrace void rcu_read_unlock_sched_notrace(void) /** * kfree_rcu() - kfree an object after a grace period. - * @ptr: pointer to kfree for both single- and double-argument invocations. - * @rhf: the name of the struct rcu_head within the type of @ptr, - * but only for double-argument invocations. + * @ptr: pointer to kfree for double-argument invocations. + * @rhf: the name of the struct rcu_head within the type of @ptr. * * Many rcu callbacks functions just call kfree() on the base structure. * These functions are trivial, but their size adds up, and furthermore @@ -984,26 +958,18 @@ static inline notrace void rcu_read_unlock_sched_notrace(void) * The BUILD_BUG_ON check must not involve any function calls, hence the * checks are done in macros here. */ -#define kfree_rcu(ptr, rhf...) kvfree_rcu(ptr, ## rhf) +#define kfree_rcu(ptr, rhf) kvfree_rcu_arg_2(ptr, rhf) +#define kvfree_rcu(ptr, rhf) kvfree_rcu_arg_2(ptr, rhf) /** - * kvfree_rcu() - kvfree an object after a grace period. - * - * This macro consists of one or two arguments and it is - * based on whether an object is head-less or not. If it - * has a head then a semantic stays the same as it used - * to be before: - * - * kvfree_rcu(ptr, rhf); - * - * where @ptr is a pointer to kvfree(), @rhf is the name - * of the rcu_head structure within the type of @ptr. + * kfree_rcu_mightsleep() - kfree an object after a grace period. + * @ptr: pointer to kfree for single-argument invocations. * * When it comes to head-less variant, only one argument * is passed and that is just a pointer which has to be * freed after a grace period. Therefore the semantic is * - * kvfree_rcu(ptr); + * kfree_rcu_mightsleep(ptr); * * where @ptr is the pointer to be freed by kvfree(). * @@ -1012,13 +978,9 @@ static inline notrace void rcu_read_unlock_sched_notrace(void) * annotation. Otherwise, please switch and embed the * rcu_head structure within the type of @ptr. */ -#define kvfree_rcu(...) KVFREE_GET_MACRO(__VA_ARGS__, \ - kvfree_rcu_arg_2, kvfree_rcu_arg_1)(__VA_ARGS__) - +#define kfree_rcu_mightsleep(ptr) kvfree_rcu_arg_1(ptr) #define kvfree_rcu_mightsleep(ptr) kvfree_rcu_arg_1(ptr) -#define kfree_rcu_mightsleep(ptr) kvfree_rcu_mightsleep(ptr) -#define KVFREE_GET_MACRO(_1, _2, NAME, ...) NAME #define kvfree_rcu_arg_2(ptr, rhf) \ do { \ typeof (ptr) ___p = (ptr); \ diff --git a/include/linux/srcu.h b/include/linux/srcu.h index 41c4b26fb1c1..eb92a50a4599 100644 --- a/include/linux/srcu.h +++ b/include/linux/srcu.h @@ -212,7 +212,7 @@ static inline int srcu_read_lock(struct srcu_struct *ssp) __acquires(ssp) srcu_check_nmi_safety(ssp, false); retval = __srcu_read_lock(ssp); - srcu_lock_acquire(&(ssp)->dep_map); + srcu_lock_acquire(&ssp->dep_map); return retval; } @@ -229,7 +229,7 @@ static inline int srcu_read_lock_nmisafe(struct srcu_struct *ssp) __acquires(ssp srcu_check_nmi_safety(ssp, true); retval = __srcu_read_lock_nmisafe(ssp); - rcu_lock_acquire(&(ssp)->dep_map); + rcu_lock_acquire(&ssp->dep_map); return retval; } @@ -284,7 +284,7 @@ static inline void srcu_read_unlock(struct srcu_struct *ssp, int idx) { WARN_ON_ONCE(idx & ~0x1); srcu_check_nmi_safety(ssp, false); - srcu_lock_release(&(ssp)->dep_map); + srcu_lock_release(&ssp->dep_map); __srcu_read_unlock(ssp, idx); } @@ -300,7 +300,7 @@ static inline void srcu_read_unlock_nmisafe(struct srcu_struct *ssp, int idx) { WARN_ON_ONCE(idx & ~0x1); srcu_check_nmi_safety(ssp, true); - rcu_lock_release(&(ssp)->dep_map); + rcu_lock_release(&ssp->dep_map); __srcu_read_unlock_nmisafe(ssp, idx); } diff --git a/kernel/rcu/Kconfig b/kernel/rcu/Kconfig index 9071182b1284..bdd7eadb33d8 100644 --- a/kernel/rcu/Kconfig +++ b/kernel/rcu/Kconfig @@ -314,4 +314,22 @@ config RCU_LAZY To save power, batch RCU callbacks and flush after delay, memory pressure, or callback list growing too big. +config RCU_DOUBLE_CHECK_CB_TIME + bool "RCU callback-batch backup time check" + depends on RCU_EXPERT + default n + help + Use this option to provide more precise enforcement of the + rcutree.rcu_resched_ns module parameter in situations where + a single RCU callback might run for hundreds of microseconds, + thus defeating the 32-callback batching used to amortize the + cost of the fine-grained but expensive local_clock() function. + + This option rounds rcutree.rcu_resched_ns up to the next + jiffy, and overrides the 32-callback batching if this limit + is exceeded. + + Say Y here if you need tighter callback-limit enforcement. + Say N here if you are unsure. + endmenu # "RCU Subsystem" diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h index 4a1b9622598b..98c1544cf572 100644 --- a/kernel/rcu/rcu.h +++ b/kernel/rcu/rcu.h @@ -642,4 +642,10 @@ void show_rcu_tasks_trace_gp_kthread(void); static inline void show_rcu_tasks_trace_gp_kthread(void) {} #endif +#ifdef CONFIG_TINY_RCU +static inline bool rcu_cpu_beenfullyonline(int cpu) { return true; } +#else +bool rcu_cpu_beenfullyonline(int cpu); +#endif + #endif /* __LINUX_RCU_H */ diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h index 5f4fc8184dd0..b770add3f843 100644 --- a/kernel/rcu/tasks.h +++ b/kernel/rcu/tasks.h @@ -241,7 +241,6 @@ static void cblist_init_generic(struct rcu_tasks *rtp) if (rcu_task_enqueue_lim < 0) { rcu_task_enqueue_lim = 1; rcu_task_cb_adjust = true; - pr_info("%s: Setting adjustable number of callback queues.\n", __func__); } else if (rcu_task_enqueue_lim == 0) { rcu_task_enqueue_lim = 1; } @@ -272,7 +271,9 @@ static void cblist_init_generic(struct rcu_tasks *rtp) raw_spin_unlock_rcu_node(rtpcp); // irqs remain disabled. } raw_spin_unlock_irqrestore(&rtp->cbs_gbl_lock, flags); - pr_info("%s: Setting shift to %d and lim to %d.\n", __func__, data_race(rtp->percpu_enqueue_shift), data_race(rtp->percpu_enqueue_lim)); + + pr_info("%s: Setting shift to %d and lim to %d rcu_task_cb_adjust=%d.\n", rtp->name, + data_race(rtp->percpu_enqueue_shift), data_race(rtp->percpu_enqueue_lim), rcu_task_cb_adjust); } // IRQ-work handler that does deferred wakeup for call_rcu_tasks_generic(). @@ -463,6 +464,7 @@ static void rcu_tasks_invoke_cbs(struct rcu_tasks *rtp, struct rcu_tasks_percpu { int cpu; int cpunext; + int cpuwq; unsigned long flags; int len; struct rcu_head *rhp; @@ -473,11 +475,13 @@ static void rcu_tasks_invoke_cbs(struct rcu_tasks *rtp, struct rcu_tasks_percpu cpunext = cpu * 2 + 1; if (cpunext < smp_load_acquire(&rtp->percpu_dequeue_lim)) { rtpcp_next = per_cpu_ptr(rtp->rtpcpu, cpunext); - queue_work_on(cpunext, system_wq, &rtpcp_next->rtp_work); + cpuwq = rcu_cpu_beenfullyonline(cpunext) ? cpunext : WORK_CPU_UNBOUND; + queue_work_on(cpuwq, system_wq, &rtpcp_next->rtp_work); cpunext++; if (cpunext < smp_load_acquire(&rtp->percpu_dequeue_lim)) { rtpcp_next = per_cpu_ptr(rtp->rtpcpu, cpunext); - queue_work_on(cpunext, system_wq, &rtpcp_next->rtp_work); + cpuwq = rcu_cpu_beenfullyonline(cpunext) ? cpunext : WORK_CPU_UNBOUND; + queue_work_on(cpuwq, system_wq, &rtpcp_next->rtp_work); } } diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c index f52ff7241041..1449cb69a0e0 100644 --- a/kernel/rcu/tree.c +++ b/kernel/rcu/tree.c @@ -2046,19 +2046,35 @@ rcu_check_quiescent_state(struct rcu_data *rdp) rcu_report_qs_rdp(rdp); } +/* Return true if callback-invocation time limit exceeded. */ +static bool rcu_do_batch_check_time(long count, long tlimit, + bool jlimit_check, unsigned long jlimit) +{ + // Invoke local_clock() only once per 32 consecutive callbacks. + return unlikely(tlimit) && + (!likely(count & 31) || + (IS_ENABLED(CONFIG_RCU_DOUBLE_CHECK_CB_TIME) && + jlimit_check && time_after(jiffies, jlimit))) && + local_clock() >= tlimit; +} + /* * Invoke any RCU callbacks that have made it to the end of their grace * period. Throttle as specified by rdp->blimit. */ static void rcu_do_batch(struct rcu_data *rdp) { + long bl; + long count = 0; int div; bool __maybe_unused empty; unsigned long flags; - struct rcu_head *rhp; + unsigned long jlimit; + bool jlimit_check = false; + long pending; struct rcu_cblist rcl = RCU_CBLIST_INITIALIZER(rcl); - long bl, count = 0; - long pending, tlimit = 0; + struct rcu_head *rhp; + long tlimit = 0; /* If no callbacks are ready, just return. */ if (!rcu_segcblist_ready_cbs(&rdp->cblist)) { @@ -2082,11 +2098,15 @@ static void rcu_do_batch(struct rcu_data *rdp) div = READ_ONCE(rcu_divisor); div = div < 0 ? 7 : div > sizeof(long) * 8 - 2 ? sizeof(long) * 8 - 2 : div; bl = max(rdp->blimit, pending >> div); - if (in_serving_softirq() && unlikely(bl > 100)) { + if ((in_serving_softirq() || rdp->rcu_cpu_kthread_status == RCU_KTHREAD_RUNNING) && + (IS_ENABLED(CONFIG_RCU_DOUBLE_CHECK_CB_TIME) || unlikely(bl > 100))) { + const long npj = NSEC_PER_SEC / HZ; long rrn = READ_ONCE(rcu_resched_ns); rrn = rrn < NSEC_PER_MSEC ? NSEC_PER_MSEC : rrn > NSEC_PER_SEC ? NSEC_PER_SEC : rrn; tlimit = local_clock() + rrn; + jlimit = jiffies + (rrn + npj + 1) / npj; + jlimit_check = true; } trace_rcu_batch_start(rcu_state.name, rcu_segcblist_n_cbs(&rdp->cblist), bl); @@ -2126,21 +2146,23 @@ static void rcu_do_batch(struct rcu_data *rdp) * Make sure we don't spend too much time here and deprive other * softirq vectors of CPU cycles. */ - if (unlikely(tlimit)) { - /* only call local_clock() every 32 callbacks */ - if (likely((count & 31) || local_clock() < tlimit)) - continue; - /* Exceeded the time limit, so leave. */ + if (rcu_do_batch_check_time(count, tlimit, jlimit_check, jlimit)) break; - } } else { - // In rcuoc context, so no worries about depriving - // other softirq vectors of CPU cycles. + // In rcuc/rcuoc context, so no worries about + // depriving other softirq vectors of CPU cycles. local_bh_enable(); lockdep_assert_irqs_enabled(); cond_resched_tasks_rcu_qs(); lockdep_assert_irqs_enabled(); local_bh_disable(); + // But rcuc kthreads can delay quiescent-state + // reporting, so check time limits for them. + if (rdp->rcu_cpu_kthread_status == RCU_KTHREAD_RUNNING && + rcu_do_batch_check_time(count, tlimit, jlimit_check, jlimit)) { + rdp->rcu_cpu_has_work = 1; + break; + } } } @@ -2459,12 +2481,12 @@ static void rcu_cpu_kthread(unsigned int cpu) *statusp = RCU_KTHREAD_RUNNING; local_irq_disable(); work = *workp; - *workp = 0; + WRITE_ONCE(*workp, 0); local_irq_enable(); if (work) rcu_core(); local_bh_enable(); - if (*workp == 0) { + if (!READ_ONCE(*workp)) { trace_rcu_utilization(TPS("End CPU kthread@rcu_wait")); *statusp = RCU_KTHREAD_WAITING; return; @@ -2756,7 +2778,7 @@ EXPORT_SYMBOL_GPL(call_rcu); */ struct kvfree_rcu_bulk_data { struct list_head list; - unsigned long gp_snap; + struct rcu_gp_oldstate gp_snap; unsigned long nr_records; void *records[]; }; @@ -2773,6 +2795,7 @@ struct kvfree_rcu_bulk_data { * struct kfree_rcu_cpu_work - single batch of kfree_rcu() requests * @rcu_work: Let queue_rcu_work() invoke workqueue handler after grace period * @head_free: List of kfree_rcu() objects waiting for a grace period + * @head_free_gp_snap: Grace-period snapshot to check for attempted premature frees. * @bulk_head_free: Bulk-List of kvfree_rcu() objects waiting for a grace period * @krcp: Pointer to @kfree_rcu_cpu structure */ @@ -2780,6 +2803,7 @@ struct kvfree_rcu_bulk_data { struct kfree_rcu_cpu_work { struct rcu_work rcu_work; struct rcu_head *head_free; + struct rcu_gp_oldstate head_free_gp_snap; struct list_head bulk_head_free[FREE_N_CHANNELS]; struct kfree_rcu_cpu *krcp; }; @@ -2900,6 +2924,9 @@ drain_page_cache(struct kfree_rcu_cpu *krcp) struct llist_node *page_list, *pos, *n; int freed = 0; + if (!rcu_min_cached_objs) + return 0; + raw_spin_lock_irqsave(&krcp->lock, flags); page_list = llist_del_all(&krcp->bkvcache); WRITE_ONCE(krcp->nr_bkv_objs, 0); @@ -2920,24 +2947,25 @@ kvfree_rcu_bulk(struct kfree_rcu_cpu *krcp, unsigned long flags; int i; - debug_rcu_bhead_unqueue(bnode); - - rcu_lock_acquire(&rcu_callback_map); - if (idx == 0) { // kmalloc() / kfree(). - trace_rcu_invoke_kfree_bulk_callback( - rcu_state.name, bnode->nr_records, - bnode->records); - - kfree_bulk(bnode->nr_records, bnode->records); - } else { // vmalloc() / vfree(). - for (i = 0; i < bnode->nr_records; i++) { - trace_rcu_invoke_kvfree_callback( - rcu_state.name, bnode->records[i], 0); - - vfree(bnode->records[i]); + if (!WARN_ON_ONCE(!poll_state_synchronize_rcu_full(&bnode->gp_snap))) { + debug_rcu_bhead_unqueue(bnode); + rcu_lock_acquire(&rcu_callback_map); + if (idx == 0) { // kmalloc() / kfree(). + trace_rcu_invoke_kfree_bulk_callback( + rcu_state.name, bnode->nr_records, + bnode->records); + + kfree_bulk(bnode->nr_records, bnode->records); + } else { // vmalloc() / vfree(). + for (i = 0; i < bnode->nr_records; i++) { + trace_rcu_invoke_kvfree_callback( + rcu_state.name, bnode->records[i], 0); + + vfree(bnode->records[i]); + } } + rcu_lock_release(&rcu_callback_map); } - rcu_lock_release(&rcu_callback_map); raw_spin_lock_irqsave(&krcp->lock, flags); if (put_cached_bnode(krcp, bnode)) @@ -2984,6 +3012,7 @@ static void kfree_rcu_work(struct work_struct *work) struct rcu_head *head; struct kfree_rcu_cpu *krcp; struct kfree_rcu_cpu_work *krwp; + struct rcu_gp_oldstate head_gp_snap; int i; krwp = container_of(to_rcu_work(work), @@ -2998,6 +3027,7 @@ static void kfree_rcu_work(struct work_struct *work) // Channel 3. head = krwp->head_free; krwp->head_free = NULL; + head_gp_snap = krwp->head_free_gp_snap; raw_spin_unlock_irqrestore(&krcp->lock, flags); // Handle the first two channels. @@ -3014,7 +3044,8 @@ static void kfree_rcu_work(struct work_struct *work) * queued on a linked list through their rcu_head structures. * This list is named "Channel 3". */ - kvfree_rcu_list(head); + if (head && !WARN_ON_ONCE(!poll_state_synchronize_rcu_full(&head_gp_snap))) + kvfree_rcu_list(head); } static bool @@ -3081,7 +3112,7 @@ kvfree_rcu_drain_ready(struct kfree_rcu_cpu *krcp) INIT_LIST_HEAD(&bulk_ready[i]); list_for_each_entry_safe_reverse(bnode, n, &krcp->bulk_head[i], list) { - if (!poll_state_synchronize_rcu(bnode->gp_snap)) + if (!poll_state_synchronize_rcu_full(&bnode->gp_snap)) break; atomic_sub(bnode->nr_records, &krcp->bulk_count[i]); @@ -3146,6 +3177,7 @@ static void kfree_rcu_monitor(struct work_struct *work) // objects queued on the linked list. if (!krwp->head_free) { krwp->head_free = krcp->head; + get_state_synchronize_rcu_full(&krwp->head_free_gp_snap); atomic_set(&krcp->head_count, 0); WRITE_ONCE(krcp->head, NULL); } @@ -3194,7 +3226,7 @@ static void fill_page_cache_func(struct work_struct *work) nr_pages = atomic_read(&krcp->backoff_page_cache_fill) ? 1 : rcu_min_cached_objs; - for (i = 0; i < nr_pages; i++) { + for (i = READ_ONCE(krcp->nr_bkv_objs); i < nr_pages; i++) { bnode = (struct kvfree_rcu_bulk_data *) __get_free_page(GFP_KERNEL | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN); @@ -3218,6 +3250,10 @@ static void fill_page_cache_func(struct work_struct *work) static void run_page_cache_worker(struct kfree_rcu_cpu *krcp) { + // If cache disabled, bail out. + if (!rcu_min_cached_objs) + return; + if (rcu_scheduler_active == RCU_SCHEDULER_RUNNING && !atomic_xchg(&krcp->work_in_progress, 1)) { if (atomic_read(&krcp->backoff_page_cache_fill)) { @@ -3272,7 +3308,7 @@ add_ptr_to_bulk_krc_lock(struct kfree_rcu_cpu **krcp, // scenarios. bnode = (struct kvfree_rcu_bulk_data *) __get_free_page(GFP_KERNEL | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN); - *krcp = krc_this_cpu_lock(flags); + raw_spin_lock_irqsave(&(*krcp)->lock, *flags); } if (!bnode) @@ -3285,7 +3321,7 @@ add_ptr_to_bulk_krc_lock(struct kfree_rcu_cpu **krcp, // Finally insert and update the GP for this page. bnode->records[bnode->nr_records++] = ptr; - bnode->gp_snap = get_state_synchronize_rcu(); + get_state_synchronize_rcu_full(&bnode->gp_snap); atomic_inc(&(*krcp)->bulk_count[idx]); return true; @@ -4283,7 +4319,6 @@ int rcutree_prepare_cpu(unsigned int cpu) */ rnp = rdp->mynode; raw_spin_lock_rcu_node(rnp); /* irqs already disabled. */ - rdp->beenonline = true; /* We have now been online. */ rdp->gp_seq = READ_ONCE(rnp->gp_seq); rdp->gp_seq_needed = rdp->gp_seq; rdp->cpu_no_qs.b.norm = true; @@ -4311,6 +4346,16 @@ static void rcutree_affinity_setting(unsigned int cpu, int outgoing) } /* + * Has the specified (known valid) CPU ever been fully online? + */ +bool rcu_cpu_beenfullyonline(int cpu) +{ + struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu); + + return smp_load_acquire(&rdp->beenonline); +} + +/* * Near the end of the CPU-online process. Pretty much all services * enabled, and the CPU is now very much alive. */ @@ -4368,15 +4413,16 @@ int rcutree_offline_cpu(unsigned int cpu) * Note that this function is special in that it is invoked directly * from the incoming CPU rather than from the cpuhp_step mechanism. * This is because this function must be invoked at a precise location. + * This incoming CPU must not have enabled interrupts yet. */ void rcu_cpu_starting(unsigned int cpu) { - unsigned long flags; unsigned long mask; struct rcu_data *rdp; struct rcu_node *rnp; bool newcpu; + lockdep_assert_irqs_disabled(); rdp = per_cpu_ptr(&rcu_data, cpu); if (rdp->cpu_started) return; @@ -4384,7 +4430,6 @@ void rcu_cpu_starting(unsigned int cpu) rnp = rdp->mynode; mask = rdp->grpmask; - local_irq_save(flags); arch_spin_lock(&rcu_state.ofl_lock); rcu_dynticks_eqs_online(); raw_spin_lock(&rcu_state.barrier_lock); @@ -4403,17 +4448,17 @@ void rcu_cpu_starting(unsigned int cpu) /* An incoming CPU should never be blocking a grace period. */ if (WARN_ON_ONCE(rnp->qsmask & mask)) { /* RCU waiting on incoming CPU? */ /* rcu_report_qs_rnp() *really* wants some flags to restore */ - unsigned long flags2; + unsigned long flags; - local_irq_save(flags2); + local_irq_save(flags); rcu_disable_urgency_upon_qs(rdp); /* Report QS -after- changing ->qsmaskinitnext! */ - rcu_report_qs_rnp(mask, rnp, rnp->gp_seq, flags2); + rcu_report_qs_rnp(mask, rnp, rnp->gp_seq, flags); } else { raw_spin_unlock_rcu_node(rnp); } arch_spin_unlock(&rcu_state.ofl_lock); - local_irq_restore(flags); + smp_store_release(&rdp->beenonline, true); smp_mb(); /* Ensure RCU read-side usage follows above initialization. */ } diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h index 3b7abb58157d..8239b39d945b 100644 --- a/kernel/rcu/tree_exp.h +++ b/kernel/rcu/tree_exp.h @@ -643,7 +643,7 @@ static void synchronize_rcu_expedited_wait(void) "O."[!!cpu_online(cpu)], "o."[!!(rdp->grpmask & rnp->expmaskinit)], "N."[!!(rdp->grpmask & rnp->expmaskinitnext)], - "D."[!!(rdp->cpu_no_qs.b.exp)]); + "D."[!!data_race(rdp->cpu_no_qs.b.exp)]); } } pr_cont(" } %lu jiffies s: %lu root: %#lx/%c\n", diff --git a/kernel/rcu/tree_nocb.h b/kernel/rcu/tree_nocb.h index f2280616f9d5..43229d2b0c44 100644 --- a/kernel/rcu/tree_nocb.h +++ b/kernel/rcu/tree_nocb.h @@ -1319,13 +1319,22 @@ lazy_rcu_shrink_count(struct shrinker *shrink, struct shrink_control *sc) int cpu; unsigned long count = 0; + if (WARN_ON_ONCE(!cpumask_available(rcu_nocb_mask))) + return 0; + + /* Protect rcu_nocb_mask against concurrent (de-)offloading. */ + if (!mutex_trylock(&rcu_state.barrier_mutex)) + return 0; + /* Snapshot count of all CPUs */ - for_each_possible_cpu(cpu) { + for_each_cpu(cpu, rcu_nocb_mask) { struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu); count += READ_ONCE(rdp->lazy_len); } + mutex_unlock(&rcu_state.barrier_mutex); + return count ? count : SHRINK_EMPTY; } @@ -1336,15 +1345,45 @@ lazy_rcu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) unsigned long flags; unsigned long count = 0; + if (WARN_ON_ONCE(!cpumask_available(rcu_nocb_mask))) + return 0; + /* + * Protect against concurrent (de-)offloading. Otherwise nocb locking + * may be ignored or imbalanced. + */ + if (!mutex_trylock(&rcu_state.barrier_mutex)) { + /* + * But really don't insist if barrier_mutex is contended since we + * can't guarantee that it will never engage in a dependency + * chain involving memory allocation. The lock is seldom contended + * anyway. + */ + return 0; + } + /* Snapshot count of all CPUs */ - for_each_possible_cpu(cpu) { + for_each_cpu(cpu, rcu_nocb_mask) { struct rcu_data *rdp = per_cpu_ptr(&rcu_data, cpu); - int _count = READ_ONCE(rdp->lazy_len); + int _count; + + if (WARN_ON_ONCE(!rcu_rdp_is_offloaded(rdp))) + continue; - if (_count == 0) + if (!READ_ONCE(rdp->lazy_len)) continue; + rcu_nocb_lock_irqsave(rdp, flags); - WRITE_ONCE(rdp->lazy_len, 0); + /* + * Recheck under the nocb lock. Since we are not holding the bypass + * lock we may still race with increments from the enqueuer but still + * we know for sure if there is at least one lazy callback. + */ + _count = READ_ONCE(rdp->lazy_len); + if (!_count) { + rcu_nocb_unlock_irqrestore(rdp, flags); + continue; + } + WARN_ON_ONCE(!rcu_nocb_flush_bypass(rdp, NULL, jiffies, false)); rcu_nocb_unlock_irqrestore(rdp, flags); wake_nocb_gp(rdp, false); sc->nr_to_scan -= _count; @@ -1352,6 +1391,9 @@ lazy_rcu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc) if (sc->nr_to_scan <= 0) break; } + + mutex_unlock(&rcu_state.barrier_mutex); + return count ? count : SHRINK_STOP; } diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h index 7b0fe741a088..41021080ad25 100644 --- a/kernel/rcu/tree_plugin.h +++ b/kernel/rcu/tree_plugin.h @@ -257,6 +257,8 @@ static void rcu_preempt_ctxt_queue(struct rcu_node *rnp, struct rcu_data *rdp) * GP should not be able to end until we report, so there should be * no need to check for a subsequent expedited GP. (Though we are * still in a quiescent state in any case.) + * + * Interrupts are disabled, so ->cpu_no_qs.b.exp cannot change. */ if (blkd_state & RCU_EXP_BLKD && rdp->cpu_no_qs.b.exp) rcu_report_exp_rdp(rdp); @@ -941,7 +943,7 @@ notrace void rcu_preempt_deferred_qs(struct task_struct *t) { struct rcu_data *rdp = this_cpu_ptr(&rcu_data); - if (rdp->cpu_no_qs.b.exp) + if (READ_ONCE(rdp->cpu_no_qs.b.exp)) rcu_report_exp_rdp(rdp); } |