From 00b42959106a9ca1c2899e591ae4e9a83ad6af05 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 7 Nov 2012 21:03:07 +0100 Subject: irq_work: Don't stop the tick with pending works Don't stop the tick if we have pending irq works on the queue, otherwise if the arch can't raise self-IPIs, we may not find an opportunity to execute the pending works for a while. Signed-off-by: Frederic Weisbecker Acked-by: Steven Rostedt Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Ingo Molnar Cc: Andrew Morton Cc: Paul Gortmaker --- kernel/irq_work.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'kernel/irq_work.c') diff --git a/kernel/irq_work.c b/kernel/irq_work.c index 64eddd59ed83..b3c113a14727 100644 --- a/kernel/irq_work.c +++ b/kernel/irq_work.c @@ -99,6 +99,17 @@ bool irq_work_queue(struct irq_work *work) } EXPORT_SYMBOL_GPL(irq_work_queue); +bool irq_work_needs_cpu(void) +{ + struct llist_head *this_list; + + this_list = &__get_cpu_var(irq_work_list); + if (llist_empty(this_list)) + return false; + + return true; +} + /* * Run the irq_work entries on this cpu. Requires to be ran from hardirq * context with local IRQs disabled. -- cgit From c0e980a4bd7fc5c9b748f2f0209d2a48c0fdf0ab Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 15 Nov 2012 11:34:21 -0500 Subject: irq_work: Flush work on CPU_DYING In order not to offline a CPU with pending irq works, flush the queue from CPU_DYING. The notifier is called by stop_machine on the CPU that is going down. The code will not be called from irq context (so things like get_irq_regs() wont work) but I'm not sure what the requirements are for irq_work in that regard (Peter?). But irqs are disabled and the CPU is about to go offline. Might as well flush the work. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Ingo Molnar Cc: Andrew Morton Cc: Paul Gortmaker Signed-off-by: Frederic Weisbecker --- kernel/irq_work.c | 51 +++++++++++++++++++++++++++++++++++++++++++++------ 1 file changed, 45 insertions(+), 6 deletions(-) (limited to 'kernel/irq_work.c') diff --git a/kernel/irq_work.c b/kernel/irq_work.c index b3c113a14727..4ed17490f629 100644 --- a/kernel/irq_work.c +++ b/kernel/irq_work.c @@ -12,6 +12,8 @@ #include #include #include +#include +#include #include /* @@ -110,11 +112,7 @@ bool irq_work_needs_cpu(void) return true; } -/* - * Run the irq_work entries on this cpu. Requires to be ran from hardirq - * context with local IRQs disabled. - */ -void irq_work_run(void) +static void __irq_work_run(void) { struct irq_work *work; struct llist_head *this_list; @@ -124,7 +122,6 @@ void irq_work_run(void) if (llist_empty(this_list)) return; - BUG_ON(!in_irq()); BUG_ON(!irqs_disabled()); llnode = llist_del_all(this_list); @@ -149,6 +146,16 @@ void irq_work_run(void) (void)cmpxchg(&work->flags, IRQ_WORK_BUSY, 0); } } + +/* + * Run the irq_work entries on this cpu. Requires to be ran from hardirq + * context with local IRQs disabled. + */ +void irq_work_run(void) +{ + BUG_ON(!in_irq()); + __irq_work_run(); +} EXPORT_SYMBOL_GPL(irq_work_run); /* @@ -163,3 +170,35 @@ void irq_work_sync(struct irq_work *work) cpu_relax(); } EXPORT_SYMBOL_GPL(irq_work_sync); + +#ifdef CONFIG_HOTPLUG_CPU +static int irq_work_cpu_notify(struct notifier_block *self, + unsigned long action, void *hcpu) +{ + long cpu = (long)hcpu; + + switch (action) { + case CPU_DYING: + /* Called from stop_machine */ + if (WARN_ON_ONCE(cpu != smp_processor_id())) + break; + __irq_work_run(); + break; + default: + break; + } + return NOTIFY_OK; +} + +static struct notifier_block cpu_notify; + +static __init int irq_work_init_cpu_notifier(void) +{ + cpu_notify.notifier_call = irq_work_cpu_notify; + cpu_notify.priority = 0; + register_cpu_notifier(&cpu_notify); + return 0; +} +device_initcall(irq_work_init_cpu_notifier); + +#endif /* CONFIG_HOTPLUG_CPU */ -- cgit From 8aa2accee41f7045dc904fa41d4475b2f6ffae3e Mon Sep 17 00:00:00 2001 From: Steven Rostedt Date: Thu, 15 Nov 2012 12:52:44 -0500 Subject: irq_work: Warn if there's still work on cpu_down If we are in nohz and there's still irq_work to be done when the idle task is about to go offline, give a nasty warning. Everything should have been flushed from the CPU_DYING notifier already. Further attempts to enqueue an irq_work are buggy because irqs are disabled by __cpu_disable(). The best we can do is to report the issue to the user. Signed-off-by: Steven Rostedt Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Ingo Molnar Cc: Andrew Morton Cc: Paul Gortmaker Signed-off-by: Frederic Weisbecker --- kernel/irq_work.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'kernel/irq_work.c') diff --git a/kernel/irq_work.c b/kernel/irq_work.c index 4ed17490f629..480f74715ba9 100644 --- a/kernel/irq_work.c +++ b/kernel/irq_work.c @@ -109,6 +109,9 @@ bool irq_work_needs_cpu(void) if (llist_empty(this_list)) return false; + /* All work should have been flushed before going offline */ + WARN_ON_ONCE(cpu_is_offline(smp_processor_id())); + return true; } -- cgit From bc6679aef673f9dcb8f718528fc3df49ff661af9 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 19 Oct 2012 16:43:41 -0400 Subject: irq_work: Make self-IPIs optable On irq work initialization, let the user choose to define it as "lazy" or not. "Lazy" means that we don't want to send an IPI (provided the arch can anyway) when we enqueue this work but we rather prefer to wait for the next timer tick to execute our work if possible. This is going to be a benefit for non-urgent enqueuers (like printk in the future) that may prefer not to raise an IPI storm in case of frequent enqueuing on short periods of time. Signed-off-by: Frederic Weisbecker Acked-by: Steven Rostedt Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Ingo Molnar Cc: Andrew Morton Cc: Paul Gortmaker --- kernel/irq_work.c | 47 +++++++++++++++++++++++++++-------------------- 1 file changed, 27 insertions(+), 20 deletions(-) (limited to 'kernel/irq_work.c') diff --git a/kernel/irq_work.c b/kernel/irq_work.c index 480f74715ba9..7f3a59bc8e3d 100644 --- a/kernel/irq_work.c +++ b/kernel/irq_work.c @@ -12,24 +12,15 @@ #include #include #include +#include +#include #include #include #include -/* - * An entry can be in one of four states: - * - * free NULL, 0 -> {claimed} : free to be used - * claimed NULL, 3 -> {pending} : claimed to be enqueued - * pending next, 3 -> {busy} : queued, pending callback - * busy NULL, 2 -> {free, claimed} : callback in progress, can be claimed - */ - -#define IRQ_WORK_PENDING 1UL -#define IRQ_WORK_BUSY 2UL -#define IRQ_WORK_FLAGS 3UL static DEFINE_PER_CPU(struct llist_head, irq_work_list); +static DEFINE_PER_CPU(int, irq_work_raised); /* * Claim the entry so that no one else will poke at it. @@ -69,14 +60,19 @@ void __weak arch_irq_work_raise(void) */ static void __irq_work_queue(struct irq_work *work) { - bool empty; - preempt_disable(); - empty = llist_add(&work->llnode, &__get_cpu_var(irq_work_list)); - /* The list was empty, raise self-interrupt to start processing. */ - if (empty) - arch_irq_work_raise(); + llist_add(&work->llnode, &__get_cpu_var(irq_work_list)); + + /* + * If the work is not "lazy" or the tick is stopped, raise the irq + * work interrupt (if supported by the arch), otherwise, just wait + * for the next tick. + */ + if (!(work->flags & IRQ_WORK_LAZY) || tick_nohz_tick_stopped()) { + if (!this_cpu_cmpxchg(irq_work_raised, 0, 1)) + arch_irq_work_raise(); + } preempt_enable(); } @@ -117,10 +113,19 @@ bool irq_work_needs_cpu(void) static void __irq_work_run(void) { + unsigned long flags; struct irq_work *work; struct llist_head *this_list; struct llist_node *llnode; + + /* + * Reset the "raised" state right before we check the list because + * an NMI may enqueue after we find the list empty from the runner. + */ + __this_cpu_write(irq_work_raised, 0); + barrier(); + this_list = &__get_cpu_var(irq_work_list); if (llist_empty(this_list)) return; @@ -140,13 +145,15 @@ static void __irq_work_run(void) * to claim that work don't rely on us to handle their data * while we are in the middle of the func. */ - xchg(&work->flags, IRQ_WORK_BUSY); + flags = work->flags & ~IRQ_WORK_PENDING; + xchg(&work->flags, flags); + work->func(work); /* * Clear the BUSY bit and return to the free state if * no-one else claimed it meanwhile. */ - (void)cmpxchg(&work->flags, IRQ_WORK_BUSY, 0); + (void)cmpxchg(&work->flags, flags, flags & ~IRQ_WORK_BUSY); } } -- cgit