diff options
author | Heiko Carstens <heiko.carstens@de.ibm.com> | 2008-10-13 17:50:10 -0400 |
---|---|---|
committer | Rusty Russell <rusty@rustcorp.com.au> | 2008-10-21 19:00:26 -0400 |
commit | c9583e55fa2b08a230c549bd1e3c0bde6c50d9cc (patch) | |
tree | ec535aced7cdf7390290a4bf61402346b166c07b /kernel | |
parent | 0d557dc97f4bb501f086a03d0f00b99a7855d794 (diff) |
stop_machine: use workqueues instead of kernel threads
Convert stop_machine to a workqueue based approach. Instead of using kernel
threads for stop_machine we now use a an rt workqueue to synchronize all
cpus.
This has the advantage that all needed per cpu threads are already created
when stop_machine gets called. And therefore a call to stop_machine won't
fail anymore. This is needed for s390 which needs a mechanism to synchronize
all cpus without allocating any memory.
As Rusty pointed out free_module() needs a non-failing stop_machine interface
as well.
As a side effect the stop_machine code gets simplified.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Diffstat (limited to 'kernel')
-rw-r--r-- | kernel/stop_machine.c | 111 |
1 files changed, 41 insertions, 70 deletions
diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c index af3c7cea258b..0e688c6a1a63 100644 --- a/kernel/stop_machine.c +++ b/kernel/stop_machine.c | |||
@@ -37,9 +37,13 @@ struct stop_machine_data { | |||
37 | /* Like num_online_cpus(), but hotplug cpu uses us, so we need this. */ | 37 | /* Like num_online_cpus(), but hotplug cpu uses us, so we need this. */ |
38 | static unsigned int num_threads; | 38 | static unsigned int num_threads; |
39 | static atomic_t thread_ack; | 39 | static atomic_t thread_ack; |
40 | static struct completion finished; | ||
41 | static DEFINE_MUTEX(lock); | 40 | static DEFINE_MUTEX(lock); |
42 | 41 | ||
42 | static struct workqueue_struct *stop_machine_wq; | ||
43 | static struct stop_machine_data active, idle; | ||
44 | static const cpumask_t *active_cpus; | ||
45 | static void *stop_machine_work; | ||
46 | |||
43 | static void set_state(enum stopmachine_state newstate) | 47 | static void set_state(enum stopmachine_state newstate) |
44 | { | 48 | { |
45 | /* Reset ack counter. */ | 49 | /* Reset ack counter. */ |
@@ -51,21 +55,25 @@ static void set_state(enum stopmachine_state newstate) | |||
51 | /* Last one to ack a state moves to the next state. */ | 55 | /* Last one to ack a state moves to the next state. */ |
52 | static void ack_state(void) | 56 | static void ack_state(void) |
53 | { | 57 | { |
54 | if (atomic_dec_and_test(&thread_ack)) { | 58 | if (atomic_dec_and_test(&thread_ack)) |
55 | /* If we're the last one to ack the EXIT, we're finished. */ | 59 | set_state(state + 1); |
56 | if (state == STOPMACHINE_EXIT) | ||
57 | complete(&finished); | ||
58 | else | ||
59 | set_state(state + 1); | ||
60 | } | ||
61 | } | 60 | } |
62 | 61 | ||
63 | /* This is the actual thread which stops the CPU. It exits by itself rather | 62 | /* This is the actual function which stops the CPU. It runs |
64 | * than waiting for kthread_stop(), because it's easier for hotplug CPU. */ | 63 | * in the context of a dedicated stopmachine workqueue. */ |
65 | static int stop_cpu(struct stop_machine_data *smdata) | 64 | static void stop_cpu(struct work_struct *unused) |
66 | { | 65 | { |
67 | enum stopmachine_state curstate = STOPMACHINE_NONE; | 66 | enum stopmachine_state curstate = STOPMACHINE_NONE; |
68 | 67 | struct stop_machine_data *smdata = &idle; | |
68 | int cpu = smp_processor_id(); | ||
69 | |||
70 | if (!active_cpus) { | ||
71 | if (cpu == first_cpu(cpu_online_map)) | ||
72 | smdata = &active; | ||
73 | } else { | ||
74 | if (cpu_isset(cpu, *active_cpus)) | ||
75 | smdata = &active; | ||
76 | } | ||
69 | /* Simple state machine */ | 77 | /* Simple state machine */ |
70 | do { | 78 | do { |
71 | /* Chill out and ensure we re-read stopmachine_state. */ | 79 | /* Chill out and ensure we re-read stopmachine_state. */ |
@@ -90,7 +98,6 @@ static int stop_cpu(struct stop_machine_data *smdata) | |||
90 | } while (curstate != STOPMACHINE_EXIT); | 98 | } while (curstate != STOPMACHINE_EXIT); |
91 | 99 | ||
92 | local_irq_enable(); | 100 | local_irq_enable(); |
93 | do_exit(0); | ||
94 | } | 101 | } |
95 | 102 | ||
96 | /* Callback for CPUs which aren't supposed to do anything. */ | 103 | /* Callback for CPUs which aren't supposed to do anything. */ |
@@ -101,78 +108,34 @@ static int chill(void *unused) | |||
101 | 108 | ||
102 | int __stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) | 109 | int __stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) |
103 | { | 110 | { |
104 | int i, err; | 111 | struct work_struct *sm_work; |
105 | struct stop_machine_data active, idle; | 112 | int i; |
106 | struct task_struct **threads; | ||
107 | 113 | ||
114 | /* Set up initial state. */ | ||
115 | mutex_lock(&lock); | ||
116 | num_threads = num_online_cpus(); | ||
117 | active_cpus = cpus; | ||
108 | active.fn = fn; | 118 | active.fn = fn; |
109 | active.data = data; | 119 | active.data = data; |
110 | active.fnret = 0; | 120 | active.fnret = 0; |
111 | idle.fn = chill; | 121 | idle.fn = chill; |
112 | idle.data = NULL; | 122 | idle.data = NULL; |
113 | 123 | ||
114 | /* This could be too big for stack on large machines. */ | ||
115 | threads = kcalloc(NR_CPUS, sizeof(threads[0]), GFP_KERNEL); | ||
116 | if (!threads) | ||
117 | return -ENOMEM; | ||
118 | |||
119 | /* Set up initial state. */ | ||
120 | mutex_lock(&lock); | ||
121 | init_completion(&finished); | ||
122 | num_threads = num_online_cpus(); | ||
123 | set_state(STOPMACHINE_PREPARE); | 124 | set_state(STOPMACHINE_PREPARE); |
124 | 125 | ||
125 | for_each_online_cpu(i) { | 126 | /* Schedule the stop_cpu work on all cpus: hold this CPU so one |
126 | struct stop_machine_data *smdata = &idle; | ||
127 | struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; | ||
128 | |||
129 | if (!cpus) { | ||
130 | if (i == first_cpu(cpu_online_map)) | ||
131 | smdata = &active; | ||
132 | } else { | ||
133 | if (cpu_isset(i, *cpus)) | ||
134 | smdata = &active; | ||
135 | } | ||
136 | |||
137 | threads[i] = kthread_create((void *)stop_cpu, smdata, "kstop%u", | ||
138 | i); | ||
139 | if (IS_ERR(threads[i])) { | ||
140 | err = PTR_ERR(threads[i]); | ||
141 | threads[i] = NULL; | ||
142 | goto kill_threads; | ||
143 | } | ||
144 | |||
145 | /* Place it onto correct cpu. */ | ||
146 | kthread_bind(threads[i], i); | ||
147 | |||
148 | /* Make it highest prio. */ | ||
149 | if (sched_setscheduler_nocheck(threads[i], SCHED_FIFO, ¶m)) | ||
150 | BUG(); | ||
151 | } | ||
152 | |||
153 | /* We've created all the threads. Wake them all: hold this CPU so one | ||
154 | * doesn't hit this CPU until we're ready. */ | 127 | * doesn't hit this CPU until we're ready. */ |
155 | get_cpu(); | 128 | get_cpu(); |
156 | for_each_online_cpu(i) | 129 | for_each_online_cpu(i) { |
157 | wake_up_process(threads[i]); | 130 | sm_work = percpu_ptr(stop_machine_work, i); |
158 | 131 | INIT_WORK(sm_work, stop_cpu); | |
132 | queue_work_on(i, stop_machine_wq, sm_work); | ||
133 | } | ||
159 | /* This will release the thread on our CPU. */ | 134 | /* This will release the thread on our CPU. */ |
160 | put_cpu(); | 135 | put_cpu(); |
161 | wait_for_completion(&finished); | 136 | flush_workqueue(stop_machine_wq); |
162 | mutex_unlock(&lock); | 137 | mutex_unlock(&lock); |
163 | |||
164 | kfree(threads); | ||
165 | |||
166 | return active.fnret; | 138 | return active.fnret; |
167 | |||
168 | kill_threads: | ||
169 | for_each_online_cpu(i) | ||
170 | if (threads[i]) | ||
171 | kthread_stop(threads[i]); | ||
172 | mutex_unlock(&lock); | ||
173 | |||
174 | kfree(threads); | ||
175 | return err; | ||
176 | } | 139 | } |
177 | 140 | ||
178 | int stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) | 141 | int stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) |
@@ -187,3 +150,11 @@ int stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus) | |||
187 | return ret; | 150 | return ret; |
188 | } | 151 | } |
189 | EXPORT_SYMBOL_GPL(stop_machine); | 152 | EXPORT_SYMBOL_GPL(stop_machine); |
153 | |||
154 | static int __init stop_machine_init(void) | ||
155 | { | ||
156 | stop_machine_wq = create_rt_workqueue("kstop"); | ||
157 | stop_machine_work = alloc_percpu(struct work_struct); | ||
158 | return 0; | ||
159 | } | ||
160 | early_initcall(stop_machine_init); | ||