From 65c011845316d3c1381f478ca0d8265c43b3b039 Mon Sep 17 00:00:00 2001 From: Mike Travis Date: Tue, 15 Jul 2008 14:14:30 -0700 Subject: cpumask: Replace cpumask_of_cpu with cpumask_of_cpu_ptr * This patch replaces the dangerous lvalue version of cpumask_of_cpu with new cpumask_of_cpu_ptr macros. These are patterned after the node_to_cpumask_ptr macros. In general terms, if there is a cpumask_of_cpu_map[] then a pointer to the cpumask_of_cpu_map[cpu] entry is used. The cpumask_of_cpu_map is provided when there is a large NR_CPUS count, reducing greatly the amount of code generated and stack space used for cpumask_of_cpu(). The pointer to the cpumask_t value is needed for calling set_cpus_allowed_ptr() to reduce the amount of stack space needed to pass the cpumask_t value. If there isn't a cpumask_of_cpu_map[], then a temporary variable is declared and filled in with value from cpumask_of_cpu(cpu) as well as a pointer variable pointing to this temporary variable. Afterwards, the pointer is used to reference the cpumask value. The compiler will optimize out the extra dereference through the pointer as well as the stack space used for the pointer, resulting in identical code. A good example of the orthogonal usages is in net/sunrpc/svc.c: case SVC_POOL_PERCPU: { unsigned int cpu = m->pool_to[pidx]; cpumask_of_cpu_ptr(cpumask, cpu); *oldmask = current->cpus_allowed; set_cpus_allowed_ptr(current, cpumask); return 1; } case SVC_POOL_PERNODE: { unsigned int node = m->pool_to[pidx]; node_to_cpumask_ptr(nodecpumask, node); *oldmask = current->cpus_allowed; set_cpus_allowed_ptr(current, nodecpumask); return 1; } Signed-off-by: Mike Travis Signed-off-by: Ingo Molnar --- kernel/stop_machine.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'kernel/stop_machine.c') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index ba9b2054ecbd..738b411ff2d3 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -33,8 +33,9 @@ static int stopmachine(void *cpu) { int irqs_disabled = 0; int prepared = 0; + cpumask_of_cpu_ptr(cpumask, (int)(long)cpu); - set_cpus_allowed_ptr(current, &cpumask_of_cpu((int)(long)cpu)); + set_cpus_allowed_ptr(current, cpumask); /* Ack: we are alive */ smp_mb(); /* Theoretically the ack = 0 might not be on this CPU yet. */ -- cgit v1.2.2 From 5c2aed622571ac7c3c6ec182d6d3c318e4b45c8b Mon Sep 17 00:00:00 2001 From: Jason Baron Date: Thu, 28 Feb 2008 11:33:03 -0500 Subject: stop_machine: add ALL_CPUS option -allow stop_mahcine_run() to call a function on all cpus. Calling stop_machine_run() with a 'ALL_CPUS' invokes this new behavior. stop_machine_run() proceeds as normal until the calling cpu has invoked 'fn'. Then, we tell all the other cpus to call 'fn'. Signed-off-by: Jason Baron Signed-off-by: Mathieu Desnoyers Signed-off-by: Rusty Russell CC: Adrian Bunk CC: Andi Kleen CC: Alexey Dobriyan CC: Christoph Hellwig CC: mingo@elte.hu CC: akpm@osdl.org --- kernel/stop_machine.c | 32 +++++++++++++++++++++++++------- 1 file changed, 25 insertions(+), 7 deletions(-) (limited to 'kernel/stop_machine.c') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 738b411ff2d3..a473bd0cb71b 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -22,9 +22,17 @@ enum stopmachine_state { STOPMACHINE_WAIT, STOPMACHINE_PREPARE, STOPMACHINE_DISABLE_IRQ, + STOPMACHINE_RUN, STOPMACHINE_EXIT, }; +struct stop_machine_data { + int (*fn)(void *); + void *data; + struct completion done; + int run_all; +} smdata; + static enum stopmachine_state stopmachine_state; static unsigned int stopmachine_num_threads; static atomic_t stopmachine_thread_ack; @@ -33,6 +41,7 @@ static int stopmachine(void *cpu) { int irqs_disabled = 0; int prepared = 0; + int ran = 0; cpumask_of_cpu_ptr(cpumask, (int)(long)cpu); set_cpus_allowed_ptr(current, cpumask); @@ -58,6 +67,11 @@ static int stopmachine(void *cpu) prepared = 1; smp_mb(); /* Must read state first. */ atomic_inc(&stopmachine_thread_ack); + } else if (stopmachine_state == STOPMACHINE_RUN && !ran) { + smdata.fn(smdata.data); + ran = 1; + smp_mb(); /* Must read state first. */ + atomic_inc(&stopmachine_thread_ack); } /* Yield in first stage: migration threads need to * help our sisters onto their CPUs. */ @@ -136,11 +150,10 @@ static void restart_machine(void) preempt_enable_no_resched(); } -struct stop_machine_data { - int (*fn)(void *); - void *data; - struct completion done; -}; +static void run_other_cpus(void) +{ + stopmachine_set_state(STOPMACHINE_RUN); +} static int do_stop(void *_smdata) { @@ -150,6 +163,8 @@ static int do_stop(void *_smdata) ret = stop_machine(); if (ret == 0) { ret = smdata->fn(smdata->data); + if (smdata->run_all) + run_other_cpus(); restart_machine(); } @@ -173,14 +188,17 @@ struct task_struct *__stop_machine_run(int (*fn)(void *), void *data, struct stop_machine_data smdata; struct task_struct *p; + mutex_lock(&stopmachine_mutex); + smdata.fn = fn; smdata.data = data; + smdata.run_all = (cpu == ALL_CPUS) ? 1 : 0; init_completion(&smdata.done); - mutex_lock(&stopmachine_mutex); + smp_wmb(); /* make sure other cpus see smdata updates */ /* If they don't care which CPU fn runs on, bind to any online one. */ - if (cpu == NR_CPUS) + if (cpu == NR_CPUS || cpu == ALL_CPUS) cpu = raw_smp_processor_id(); p = kthread_create(do_stop, &smdata, "kstopmachine"); -- cgit v1.2.2 From ffdb5976c47609c862917d4c186ecbb5706d2dda Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Mon, 28 Jul 2008 12:16:28 -0500 Subject: Simplify stop_machine stop_machine creates a kthread which creates kernel threads. We can create those threads directly and simplify things a little. Some care must be taken with CPU hotunplug, which has special needs, but that code seems more robust than it was in the past. Signed-off-by: Rusty Russell Acked-by: Christian Borntraeger --- kernel/stop_machine.c | 293 +++++++++++++++++++++----------------------------- 1 file changed, 125 insertions(+), 168 deletions(-) (limited to 'kernel/stop_machine.c') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index a473bd0cb71b..35882dccc943 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -1,4 +1,4 @@ -/* Copyright 2005 Rusty Russell rusty@rustcorp.com.au IBM Corporation. +/* Copyright 2008, 2005 Rusty Russell rusty@rustcorp.com.au IBM Corporation. * GPL v2 and any later version. */ #include @@ -13,220 +13,177 @@ #include #include -/* Since we effect priority and affinity (both of which are visible - * to, and settable by outside processes) we do indirection via a - * kthread. */ - -/* Thread to stop each CPU in user context. */ +/* This controls the threads on each CPU. */ enum stopmachine_state { - STOPMACHINE_WAIT, + /* Dummy starting state for thread. */ + STOPMACHINE_NONE, + /* Awaiting everyone to be scheduled. */ STOPMACHINE_PREPARE, + /* Disable interrupts. */ STOPMACHINE_DISABLE_IRQ, + /* Run the function */ STOPMACHINE_RUN, + /* Exit */ STOPMACHINE_EXIT, }; +static enum stopmachine_state state; struct stop_machine_data { int (*fn)(void *); void *data; - struct completion done; - int run_all; -} smdata; + int fnret; +}; -static enum stopmachine_state stopmachine_state; -static unsigned int stopmachine_num_threads; -static atomic_t stopmachine_thread_ack; +/* Like num_online_cpus(), but hotplug cpu uses us, so we need this. */ +static unsigned int num_threads; +static atomic_t thread_ack; +static struct completion finished; +static DEFINE_MUTEX(lock); -static int stopmachine(void *cpu) +static void set_state(enum stopmachine_state newstate) { - int irqs_disabled = 0; - int prepared = 0; - int ran = 0; - cpumask_of_cpu_ptr(cpumask, (int)(long)cpu); - - set_cpus_allowed_ptr(current, cpumask); - - /* Ack: we are alive */ - smp_mb(); /* Theoretically the ack = 0 might not be on this CPU yet. */ - atomic_inc(&stopmachine_thread_ack); - - /* Simple state machine */ - while (stopmachine_state != STOPMACHINE_EXIT) { - if (stopmachine_state == STOPMACHINE_DISABLE_IRQ - && !irqs_disabled) { - local_irq_disable(); - hard_irq_disable(); - irqs_disabled = 1; - /* Ack: irqs disabled. */ - smp_mb(); /* Must read state first. */ - atomic_inc(&stopmachine_thread_ack); - } else if (stopmachine_state == STOPMACHINE_PREPARE - && !prepared) { - /* Everyone is in place, hold CPU. */ - preempt_disable(); - prepared = 1; - smp_mb(); /* Must read state first. */ - atomic_inc(&stopmachine_thread_ack); - } else if (stopmachine_state == STOPMACHINE_RUN && !ran) { - smdata.fn(smdata.data); - ran = 1; - smp_mb(); /* Must read state first. */ - atomic_inc(&stopmachine_thread_ack); - } - /* Yield in first stage: migration threads need to - * help our sisters onto their CPUs. */ - if (!prepared && !irqs_disabled) - yield(); - cpu_relax(); - } - - /* Ack: we are exiting. */ - smp_mb(); /* Must read state first. */ - atomic_inc(&stopmachine_thread_ack); - - if (irqs_disabled) - local_irq_enable(); - if (prepared) - preempt_enable(); - - return 0; + /* Reset ack counter. */ + atomic_set(&thread_ack, num_threads); + smp_wmb(); + state = newstate; } -/* Change the thread state */ -static void stopmachine_set_state(enum stopmachine_state state) +/* Last one to ack a state moves to the next state. */ +static void ack_state(void) { - atomic_set(&stopmachine_thread_ack, 0); - smp_wmb(); - stopmachine_state = state; - while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads) - cpu_relax(); + if (atomic_dec_and_test(&thread_ack)) { + /* If we're the last one to ack the EXIT, we're finished. */ + if (state == STOPMACHINE_EXIT) + complete(&finished); + else + set_state(state + 1); + } } -static int stop_machine(void) +/* This is the actual thread which stops the CPU. It exits by itself rather + * than waiting for kthread_stop(), because it's easier for hotplug CPU. */ +static int stop_cpu(struct stop_machine_data *smdata) { - int i, ret = 0; - - atomic_set(&stopmachine_thread_ack, 0); - stopmachine_num_threads = 0; - stopmachine_state = STOPMACHINE_WAIT; + enum stopmachine_state curstate = STOPMACHINE_NONE; + int uninitialized_var(ret); - for_each_online_cpu(i) { - if (i == raw_smp_processor_id()) - continue; - ret = kernel_thread(stopmachine, (void *)(long)i,CLONE_KERNEL); - if (ret < 0) - break; - stopmachine_num_threads++; - } - - /* Wait for them all to come to life. */ - while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads) { - yield(); + /* Simple state machine */ + do { + /* Chill out and ensure we re-read stopmachine_state. */ cpu_relax(); - } - - /* If some failed, kill them all. */ - if (ret < 0) { - stopmachine_set_state(STOPMACHINE_EXIT); - return ret; - } - - /* Now they are all started, make them hold the CPUs, ready. */ - preempt_disable(); - stopmachine_set_state(STOPMACHINE_PREPARE); - - /* Make them disable irqs. */ - local_irq_disable(); - hard_irq_disable(); - stopmachine_set_state(STOPMACHINE_DISABLE_IRQ); - - return 0; -} + if (state != curstate) { + curstate = state; + switch (curstate) { + case STOPMACHINE_DISABLE_IRQ: + local_irq_disable(); + hard_irq_disable(); + break; + case STOPMACHINE_RUN: + /* |= allows error detection if functions on + * multiple CPUs. */ + smdata->fnret |= smdata->fn(smdata->data); + break; + default: + break; + } + ack_state(); + } + } while (curstate != STOPMACHINE_EXIT); -static void restart_machine(void) -{ - stopmachine_set_state(STOPMACHINE_EXIT); local_irq_enable(); - preempt_enable_no_resched(); + do_exit(0); } -static void run_other_cpus(void) +/* Callback for CPUs which aren't supposed to do anything. */ +static int chill(void *unused) { - stopmachine_set_state(STOPMACHINE_RUN); + return 0; } -static int do_stop(void *_smdata) +int __stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) { - struct stop_machine_data *smdata = _smdata; - int ret; + int i, err; + struct stop_machine_data active, idle; + struct task_struct **threads; + + active.fn = fn; + active.data = data; + active.fnret = 0; + idle.fn = chill; + idle.data = NULL; + + /* If they don't care which cpu fn runs on, just pick one. */ + if (cpu == NR_CPUS) + cpu = any_online_cpu(cpu_online_map); + + /* This could be too big for stack on large machines. */ + threads = kcalloc(NR_CPUS, sizeof(threads[0]), GFP_KERNEL); + if (!threads) + return -ENOMEM; + + /* Set up initial state. */ + mutex_lock(&lock); + init_completion(&finished); + num_threads = num_online_cpus(); + set_state(STOPMACHINE_PREPARE); - ret = stop_machine(); - if (ret == 0) { - ret = smdata->fn(smdata->data); - if (smdata->run_all) - run_other_cpus(); - restart_machine(); - } + for_each_online_cpu(i) { + struct stop_machine_data *smdata; + struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; - /* We're done: you can kthread_stop us now */ - complete(&smdata->done); + if (cpu == ALL_CPUS || i == cpu) + smdata = &active; + else + smdata = &idle; + + threads[i] = kthread_create((void *)stop_cpu, smdata, "kstop%u", + i); + if (IS_ERR(threads[i])) { + err = PTR_ERR(threads[i]); + threads[i] = NULL; + goto kill_threads; + } - /* Wait for kthread_stop */ - set_current_state(TASK_INTERRUPTIBLE); - while (!kthread_should_stop()) { - schedule(); - set_current_state(TASK_INTERRUPTIBLE); - } - __set_current_state(TASK_RUNNING); - return ret; -} + /* Place it onto correct cpu. */ + kthread_bind(threads[i], i); -struct task_struct *__stop_machine_run(int (*fn)(void *), void *data, - unsigned int cpu) -{ - static DEFINE_MUTEX(stopmachine_mutex); - struct stop_machine_data smdata; - struct task_struct *p; + /* Make it highest prio. */ + if (sched_setscheduler_nocheck(threads[i], SCHED_FIFO, ¶m)) + BUG(); + } - mutex_lock(&stopmachine_mutex); + /* We've created all the threads. Wake them all: hold this CPU so one + * doesn't hit this CPU until we're ready. */ + cpu = get_cpu(); + for_each_online_cpu(i) + wake_up_process(threads[i]); - smdata.fn = fn; - smdata.data = data; - smdata.run_all = (cpu == ALL_CPUS) ? 1 : 0; - init_completion(&smdata.done); + /* This will release the thread on our CPU. */ + put_cpu(); + wait_for_completion(&finished); + mutex_unlock(&lock); - smp_wmb(); /* make sure other cpus see smdata updates */ + kfree(threads); - /* If they don't care which CPU fn runs on, bind to any online one. */ - if (cpu == NR_CPUS || cpu == ALL_CPUS) - cpu = raw_smp_processor_id(); + return active.fnret; - p = kthread_create(do_stop, &smdata, "kstopmachine"); - if (!IS_ERR(p)) { - struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; +kill_threads: + for_each_online_cpu(i) + if (threads[i]) + kthread_stop(threads[i]); + mutex_unlock(&lock); - /* One high-prio thread per cpu. We'll do this one. */ - sched_setscheduler_nocheck(p, SCHED_FIFO, ¶m); - kthread_bind(p, cpu); - wake_up_process(p); - wait_for_completion(&smdata.done); - } - mutex_unlock(&stopmachine_mutex); - return p; + kfree(threads); + return err; } int stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) { - struct task_struct *p; int ret; /* No CPUs can come up or down during this. */ get_online_cpus(); - p = __stop_machine_run(fn, data, cpu); - if (!IS_ERR(p)) - ret = kthread_stop(p); - else - ret = PTR_ERR(p); + ret = __stop_machine_run(fn, data, cpu); put_online_cpus(); return ret; -- cgit v1.2.2 From eeec4fad963490821348a331cca6102ae1c4a7a3 Mon Sep 17 00:00:00 2001 From: Rusty Russell Date: Mon, 28 Jul 2008 12:16:30 -0500 Subject: stop_machine(): stop_machine_run() changed to use cpu mask Instead of a "cpu" arg with magic values NR_CPUS (any cpu) and ~0 (all cpus), pass a cpumask_t. Allow NULL for the common case (where we don't care which CPU the function is run on): temporary cpumask_t's are usually considered bad for stack space. This deprecates stop_machine_run, to be removed soon when all the callers are dead. Signed-off-by: Rusty Russell --- kernel/stop_machine.c | 27 +++++++++++++-------------- 1 file changed, 13 insertions(+), 14 deletions(-) (limited to 'kernel/stop_machine.c') diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index 35882dccc943..e446c7c7d6a9 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c @@ -100,7 +100,7 @@ static int chill(void *unused) return 0; } -int __stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) +int __stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) { int i, err; struct stop_machine_data active, idle; @@ -112,10 +112,6 @@ int __stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) idle.fn = chill; idle.data = NULL; - /* If they don't care which cpu fn runs on, just pick one. */ - if (cpu == NR_CPUS) - cpu = any_online_cpu(cpu_online_map); - /* This could be too big for stack on large machines. */ threads = kcalloc(NR_CPUS, sizeof(threads[0]), GFP_KERNEL); if (!threads) @@ -128,13 +124,16 @@ int __stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) set_state(STOPMACHINE_PREPARE); for_each_online_cpu(i) { - struct stop_machine_data *smdata; + struct stop_machine_data *smdata = &idle; struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; - if (cpu == ALL_CPUS || i == cpu) - smdata = &active; - else - smdata = &idle; + if (!cpus) { + if (i == first_cpu(cpu_online_map)) + smdata = &active; + } else { + if (cpu_isset(i, *cpus)) + smdata = &active; + } threads[i] = kthread_create((void *)stop_cpu, smdata, "kstop%u", i); @@ -154,7 +153,7 @@ int __stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) /* We've created all the threads. Wake them all: hold this CPU so one * doesn't hit this CPU until we're ready. */ - cpu = get_cpu(); + get_cpu(); for_each_online_cpu(i) wake_up_process(threads[i]); @@ -177,15 +176,15 @@ kill_threads: return err; } -int stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) +int stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) { int ret; /* No CPUs can come up or down during this. */ get_online_cpus(); - ret = __stop_machine_run(fn, data, cpu); + ret = __stop_machine(fn, data, cpus); put_online_cpus(); return ret; } -EXPORT_SYMBOL_GPL(stop_machine_run); +EXPORT_SYMBOL_GPL(stop_machine); -- cgit v1.2.2