[PATCH] remove for_each_cpu()
[linux-2.6.git] / arch / i386 / kernel / cpu / cpufreq / acpi-cpufreq.c
index 3852d0a..5fd6532 100644 (file)
@@ -48,12 +48,13 @@ MODULE_LICENSE("GPL");
 
 
 struct cpufreq_acpi_io {
-       struct acpi_processor_performance       acpi_data;
+       struct acpi_processor_performance       *acpi_data;
        struct cpufreq_frequency_table          *freq_table;
        unsigned int                            resume;
 };
 
 static struct cpufreq_acpi_io  *acpi_io_data[NR_CPUS];
+static struct acpi_processor_performance       *acpi_perf_data[NR_CPUS];
 
 static struct cpufreq_driver acpi_cpufreq_driver;
 
@@ -104,64 +105,43 @@ acpi_processor_set_performance (
 {
        u16                     port = 0;
        u8                      bit_width = 0;
+       int                     i = 0;
        int                     ret = 0;
        u32                     value = 0;
-       int                     i = 0;
-       struct cpufreq_freqs    cpufreq_freqs;
-       cpumask_t               saved_mask;
        int                     retval;
+       struct acpi_processor_performance       *perf;
 
        dprintk("acpi_processor_set_performance\n");
 
-       /*
-        * TBD: Use something other than set_cpus_allowed.
-        * As set_cpus_allowed is a bit racy, 
-        * with any other set_cpus_allowed for this process.
-        */
-       saved_mask = current->cpus_allowed;
-       set_cpus_allowed(current, cpumask_of_cpu(cpu));
-       if (smp_processor_id() != cpu) {
-               return (-EAGAIN);
-       }
-       
-       if (state == data->acpi_data.state) {
+       retval = 0;
+       perf = data->acpi_data; 
+       if (state == perf->state) {
                if (unlikely(data->resume)) {
                        dprintk("Called after resume, resetting to P%d\n", state);
                        data->resume = 0;
                } else {
                        dprintk("Already at target state (P%d)\n", state);
-                       retval = 0;
-                       goto migrate_end;
+                       return (retval);
                }
        }
 
-       dprintk("Transitioning from P%d to P%d\n",
-               data->acpi_data.state, state);
-
-       /* cpufreq frequency struct */
-       cpufreq_freqs.cpu = cpu;
-       cpufreq_freqs.old = data->freq_table[data->acpi_data.state].frequency;
-       cpufreq_freqs.new = data->freq_table[state].frequency;
-
-       /* notify cpufreq */
-       cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_PRECHANGE);
+       dprintk("Transitioning from P%d to P%d\n", perf->state, state);
 
        /*
         * First we write the target state's 'control' value to the
         * control_register.
         */
 
-       port = data->acpi_data.control_register.address;
-       bit_width = data->acpi_data.control_register.bit_width;
-       value = (u32) data->acpi_data.states[state].control;
+       port = perf->control_register.address;
+       bit_width = perf->control_register.bit_width;
+       value = (u32) perf->states[state].control;
 
        dprintk("Writing 0x%08x to port 0x%04x\n", value, port);
 
        ret = acpi_processor_write_port(port, bit_width, value);
        if (ret) {
                dprintk("Invalid port width 0x%04x\n", bit_width);
-               retval = ret;
-               goto migrate_end;
+               return (ret);
        }
 
        /*
@@ -177,49 +157,35 @@ acpi_processor_set_performance (
                 * before giving up.
                 */
 
-               port = data->acpi_data.status_register.address;
-               bit_width = data->acpi_data.status_register.bit_width;
+               port = perf->status_register.address;
+               bit_width = perf->status_register.bit_width;
 
                dprintk("Looking for 0x%08x from port 0x%04x\n",
-                       (u32) data->acpi_data.states[state].status, port);
+                       (u32) perf->states[state].status, port);
 
-               for (i=0; i<100; i++) {
+               for (i = 0; i < 100; i++) {
                        ret = acpi_processor_read_port(port, bit_width, &value);
                        if (ret) {      
                                dprintk("Invalid port width 0x%04x\n", bit_width);
-                               retval = ret;
-                               goto migrate_end;
+                               return (ret);
                        }
-                       if (value == (u32) data->acpi_data.states[state].status)
+                       if (value == (u32) perf->states[state].status)
                                break;
                        udelay(10);
                }
        } else {
-               i = 0;
-               value = (u32) data->acpi_data.states[state].status;
+               value = (u32) perf->states[state].status;
        }
 
-       /* notify cpufreq */
-       cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_POSTCHANGE);
-
-       if (unlikely(value != (u32) data->acpi_data.states[state].status)) {
-               unsigned int tmp = cpufreq_freqs.new;
-               cpufreq_freqs.new = cpufreq_freqs.old;
-               cpufreq_freqs.old = tmp;
-               cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_PRECHANGE);
-               cpufreq_notify_transition(&cpufreq_freqs, CPUFREQ_POSTCHANGE);
+       if (unlikely(value != (u32) perf->states[state].status)) {
                printk(KERN_WARNING "acpi-cpufreq: Transition failed\n");
                retval = -ENODEV;
-               goto migrate_end;
+               return (retval);
        }
 
        dprintk("Transition successful after %d microseconds\n", i * 10);
 
-       data->acpi_data.state = state;
-
-       retval = 0;
-migrate_end:
-       set_cpus_allowed(current, saved_mask);
+       perf->state = state;
        return (retval);
 }
 
@@ -231,8 +197,17 @@ acpi_cpufreq_target (
        unsigned int relation)
 {
        struct cpufreq_acpi_io *data = acpi_io_data[policy->cpu];
+       struct acpi_processor_performance *perf;
+       struct cpufreq_freqs freqs;
+       cpumask_t online_policy_cpus;
+       cpumask_t saved_mask;
+       cpumask_t set_mask;
+       cpumask_t covered_cpus;
+       unsigned int cur_state = 0;
        unsigned int next_state = 0;
        unsigned int result = 0;
+       unsigned int j;
+       unsigned int tmp;
 
        dprintk("acpi_cpufreq_setpolicy\n");
 
@@ -241,11 +216,95 @@ acpi_cpufreq_target (
                        target_freq,
                        relation,
                        &next_state);
-       if (result)
+       if (unlikely(result))
                return (result);
 
-       result = acpi_processor_set_performance (data, policy->cpu, next_state);
+       perf = data->acpi_data;
+       cur_state = perf->state;
+       freqs.old = data->freq_table[cur_state].frequency;
+       freqs.new = data->freq_table[next_state].frequency;
+
+#ifdef CONFIG_HOTPLUG_CPU
+       /* cpufreq holds the hotplug lock, so we are safe from here on */
+       cpus_and(online_policy_cpus, cpu_online_map, policy->cpus);
+#else
+       online_policy_cpus = policy->cpus;
+#endif
+
+       for_each_cpu_mask(j, online_policy_cpus) {
+               freqs.cpu = j;
+               cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
+       }
+
+       /*
+        * We need to call driver->target() on all or any CPU in
+        * policy->cpus, depending on policy->shared_type.
+        */
+       saved_mask = current->cpus_allowed;
+       cpus_clear(covered_cpus);
+       for_each_cpu_mask(j, online_policy_cpus) {
+               /*
+                * Support for SMP systems.
+                * Make sure we are running on CPU that wants to change freq
+                */
+               cpus_clear(set_mask);
+               if (policy->shared_type == CPUFREQ_SHARED_TYPE_ANY)
+                       cpus_or(set_mask, set_mask, online_policy_cpus);
+               else
+                       cpu_set(j, set_mask);
+
+               set_cpus_allowed(current, set_mask);
+               if (unlikely(!cpu_isset(smp_processor_id(), set_mask))) {
+                       dprintk("couldn't limit to CPUs in this domain\n");
+                       result = -EAGAIN;
+                       break;
+               }
+
+               result = acpi_processor_set_performance (data, j, next_state);
+               if (result) {
+                       result = -EAGAIN;
+                       break;
+               }
+
+               if (policy->shared_type == CPUFREQ_SHARED_TYPE_ANY)
+                       break;
+               cpu_set(j, covered_cpus);
+       }
+
+       for_each_cpu_mask(j, online_policy_cpus) {
+               freqs.cpu = j;
+               cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
+       }
 
+       if (unlikely(result)) {
+               /*
+                * We have failed halfway through the frequency change.
+                * We have sent callbacks to online_policy_cpus and
+                * acpi_processor_set_performance() has been called on 
+                * coverd_cpus. Best effort undo..
+                */
+
+               if (!cpus_empty(covered_cpus)) {
+                       for_each_cpu_mask(j, covered_cpus) {
+                               policy->cpu = j;
+                               acpi_processor_set_performance (data, 
+                                               j, 
+                                               cur_state);
+                       }
+               }
+
+               tmp = freqs.new;
+               freqs.new = freqs.old;
+               freqs.old = tmp;
+               for_each_cpu_mask(j, online_policy_cpus) {
+                       freqs.cpu = j;
+                       cpufreq_notify_transition(&freqs, CPUFREQ_PRECHANGE);
+                       cpufreq_notify_transition(&freqs, CPUFREQ_POSTCHANGE);
+               }
+       }
+
+       set_cpus_allowed(current, saved_mask);
        return (result);
 }
 
@@ -271,30 +330,65 @@ acpi_cpufreq_guess_freq (
        struct cpufreq_acpi_io  *data,
        unsigned int            cpu)
 {
+       struct acpi_processor_performance       *perf = data->acpi_data;
+
        if (cpu_khz) {
                /* search the closest match to cpu_khz */
                unsigned int i;
                unsigned long freq;
-               unsigned long freqn = data->acpi_data.states[0].core_frequency * 1000;
+               unsigned long freqn = perf->states[0].core_frequency * 1000;
 
-               for (i=0; i < (data->acpi_data.state_count - 1); i++) {
+               for (i = 0; i < (perf->state_count - 1); i++) {
                        freq = freqn;
-                       freqn = data->acpi_data.states[i+1].core_frequency * 1000;
+                       freqn = perf->states[i+1].core_frequency * 1000;
                        if ((2 * cpu_khz) > (freqn + freq)) {
-                               data->acpi_data.state = i;
+                               perf->state = i;
                                return (freq);
                        }
                }
-               data->acpi_data.state = data->acpi_data.state_count - 1;
+               perf->state = perf->state_count - 1;
                return (freqn);
-       } else
+       } else {
                /* assume CPU is at P0... */
-               data->acpi_data.state = 0;
-               return data->acpi_data.states[0].core_frequency * 1000;
-       
+               perf->state = 0;
+               return perf->states[0].core_frequency * 1000;
+       }
 }
 
 
+/*
+ * acpi_cpufreq_early_init - initialize ACPI P-States library
+ *
+ * Initialize the ACPI P-States library (drivers/acpi/processor_perflib.c)
+ * in order to determine correct frequency and voltage pairings. We can
+ * do _PDC and _PSD and find out the processor dependency for the
+ * actual init that will happen later...
+ */
+static int acpi_cpufreq_early_init_acpi(void)
+{
+       struct acpi_processor_performance       *data;
+       unsigned int                            i, j;
+
+       dprintk("acpi_cpufreq_early_init\n");
+
+       for_each_possible_cpu(i) {
+               data = kzalloc(sizeof(struct acpi_processor_performance), 
+                       GFP_KERNEL);
+               if (!data) {
+                       for_each_possible_cpu(j) {
+                               kfree(acpi_perf_data[j]);
+                               acpi_perf_data[j] = NULL;
+                       }
+                       return (-ENOMEM);
+               }
+               acpi_perf_data[i] = data;
+       }
+
+       /* Do initialization in ACPI core */
+       acpi_processor_preregister_performance(acpi_perf_data);
+       return 0;
+}
+
 static int
 acpi_cpufreq_cpu_init (
        struct cpufreq_policy   *policy)
@@ -304,41 +398,51 @@ acpi_cpufreq_cpu_init (
        struct cpufreq_acpi_io  *data;
        unsigned int            result = 0;
        struct cpuinfo_x86 *c = &cpu_data[policy->cpu];
+       struct acpi_processor_performance       *perf;
 
        dprintk("acpi_cpufreq_cpu_init\n");
 
+       if (!acpi_perf_data[cpu])
+               return (-ENODEV);
+
        data = kzalloc(sizeof(struct cpufreq_acpi_io), GFP_KERNEL);
        if (!data)
                return (-ENOMEM);
 
+       data->acpi_data = acpi_perf_data[cpu];
        acpi_io_data[cpu] = data;
 
-       result = acpi_processor_register_performance(&data->acpi_data, cpu);
+       result = acpi_processor_register_performance(data->acpi_data, cpu);
 
        if (result)
                goto err_free;
 
+       perf = data->acpi_data;
+       policy->cpus = perf->shared_cpu_map;
+       policy->shared_type = perf->shared_type;
+
        if (cpu_has(c, X86_FEATURE_CONSTANT_TSC)) {
                acpi_cpufreq_driver.flags |= CPUFREQ_CONST_LOOPS;
        }
 
        /* capability check */
-       if (data->acpi_data.state_count <= 1) {
+       if (perf->state_count <= 1) {
                dprintk("No P-States\n");
                result = -ENODEV;
                goto err_unreg;
        }
-       if ((data->acpi_data.control_register.space_id != ACPI_ADR_SPACE_SYSTEM_IO) ||
-           (data->acpi_data.status_register.space_id != ACPI_ADR_SPACE_SYSTEM_IO)) {
+
+       if ((perf->control_register.space_id != ACPI_ADR_SPACE_SYSTEM_IO) ||
+           (perf->status_register.space_id != ACPI_ADR_SPACE_SYSTEM_IO)) {
                dprintk("Unsupported address space [%d, %d]\n",
-                       (u32) (data->acpi_data.control_register.space_id),
-                       (u32) (data->acpi_data.status_register.space_id));
+                       (u32) (perf->control_register.space_id),
+                       (u32) (perf->status_register.space_id));
                result = -ENODEV;
                goto err_unreg;
        }
 
        /* alloc freq_table */
-       data->freq_table = kmalloc(sizeof(struct cpufreq_frequency_table) * (data->acpi_data.state_count + 1), GFP_KERNEL);
+       data->freq_table = kmalloc(sizeof(struct cpufreq_frequency_table) * (perf->state_count + 1), GFP_KERNEL);
        if (!data->freq_table) {
                result = -ENOMEM;
                goto err_unreg;
@@ -346,9 +450,9 @@ acpi_cpufreq_cpu_init (
 
        /* detect transition latency */
        policy->cpuinfo.transition_latency = 0;
-       for (i=0; i<data->acpi_data.state_count; i++) {
-               if ((data->acpi_data.states[i].transition_latency * 1000) > policy->cpuinfo.transition_latency)
-                       policy->cpuinfo.transition_latency = data->acpi_data.states[i].transition_latency * 1000;
+       for (i=0; i<perf->state_count; i++) {
+               if ((perf->states[i].transition_latency * 1000) > policy->cpuinfo.transition_latency)
+                       policy->cpuinfo.transition_latency = perf->states[i].transition_latency * 1000;
        }
        policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
 
@@ -356,11 +460,11 @@ acpi_cpufreq_cpu_init (
        policy->cur = acpi_cpufreq_guess_freq(data, policy->cpu);
 
        /* table init */
-       for (i=0; i<=data->acpi_data.state_count; i++)
+       for (i=0; i<=perf->state_count; i++)
        {
                data->freq_table[i].index = i;
-               if (i<data->acpi_data.state_count)
-                       data->freq_table[i].frequency = data->acpi_data.states[i].core_frequency * 1000;
+               if (i<perf->state_count)
+                       data->freq_table[i].frequency = perf->states[i].core_frequency * 1000;
                else
                        data->freq_table[i].frequency = CPUFREQ_TABLE_END;
        }
@@ -375,12 +479,12 @@ acpi_cpufreq_cpu_init (
 
        printk(KERN_INFO "acpi-cpufreq: CPU%u - ACPI performance management activated.\n",
               cpu);
-       for (i = 0; i < data->acpi_data.state_count; i++)
+       for (i = 0; i < perf->state_count; i++)
                dprintk("     %cP%d: %d MHz, %d mW, %d uS\n",
-                       (i == data->acpi_data.state?'*':' '), i,
-                       (u32) data->acpi_data.states[i].core_frequency,
-                       (u32) data->acpi_data.states[i].power,
-                       (u32) data->acpi_data.states[i].transition_latency);
+                       (i == perf->state?'*':' '), i,
+                       (u32) perf->states[i].core_frequency,
+                       (u32) perf->states[i].power,
+                       (u32) perf->states[i].transition_latency);
 
        cpufreq_frequency_table_get_attr(data->freq_table, policy->cpu);
        
@@ -395,7 +499,7 @@ acpi_cpufreq_cpu_init (
  err_freqfree:
        kfree(data->freq_table);
  err_unreg:
-       acpi_processor_unregister_performance(&data->acpi_data, cpu);
+       acpi_processor_unregister_performance(perf, cpu);
  err_free:
        kfree(data);
        acpi_io_data[cpu] = NULL;
@@ -416,7 +520,7 @@ acpi_cpufreq_cpu_exit (
        if (data) {
                cpufreq_frequency_table_put_attr(policy->cpu);
                acpi_io_data[policy->cpu] = NULL;
-               acpi_processor_unregister_performance(&data->acpi_data, policy->cpu);
+               acpi_processor_unregister_performance(data->acpi_data, policy->cpu);
                kfree(data);
        }
 
@@ -444,14 +548,15 @@ static struct freq_attr* acpi_cpufreq_attr[] = {
 };
 
 static struct cpufreq_driver acpi_cpufreq_driver = {
-       .verify         = acpi_cpufreq_verify,
-       .target         = acpi_cpufreq_target,
-       .init           = acpi_cpufreq_cpu_init,
-       .exit           = acpi_cpufreq_cpu_exit,
-       .resume         = acpi_cpufreq_resume,
-       .name           = "acpi-cpufreq",
-       .owner          = THIS_MODULE,
-       .attr           = acpi_cpufreq_attr,
+       .verify = acpi_cpufreq_verify,
+       .target = acpi_cpufreq_target,
+       .init   = acpi_cpufreq_cpu_init,
+       .exit   = acpi_cpufreq_cpu_exit,
+       .resume = acpi_cpufreq_resume,
+       .name   = "acpi-cpufreq",
+       .owner  = THIS_MODULE,
+       .attr   = acpi_cpufreq_attr,
+       .flags  = CPUFREQ_STICKY,
 };
 
 
@@ -462,7 +567,10 @@ acpi_cpufreq_init (void)
 
        dprintk("acpi_cpufreq_init\n");
 
-       result = cpufreq_register_driver(&acpi_cpufreq_driver);
+       result = acpi_cpufreq_early_init_acpi();
+
+       if (!result)
+               result = cpufreq_register_driver(&acpi_cpufreq_driver);
        
        return (result);
 }
@@ -471,10 +579,15 @@ acpi_cpufreq_init (void)
 static void __exit
 acpi_cpufreq_exit (void)
 {
+       unsigned int    i;
        dprintk("acpi_cpufreq_exit\n");
 
        cpufreq_unregister_driver(&acpi_cpufreq_driver);
 
+       for_each_possible_cpu(i) {
+               kfree(acpi_perf_data[i]);
+               acpi_perf_data[i] = NULL;
+       }
        return;
 }