[CPUFREQ] cpumask: avoid playing with cpus_allowed in powernow-k8.c
authorRusty Russell <rusty@rustcorp.com.au>
Fri, 12 Jun 2009 11:25:37 +0000 (20:55 +0930)
committerDave Jones <davej@redhat.com>
Mon, 15 Jun 2009 15:49:43 +0000 (11:49 -0400)
cpumask: avoid playing with cpus_allowed in powernow-k8.c

It's generally a very bad idea to mug some process's cpumask: it could
legitimately and reasonably be changed by root, which could break us
(if done before our code) or them (if we restore the wrong value).

I did not replace powernowk8_target; it needs fixing, but it grabs a
mutex (so no smp_call_function_single here) but Mark points out it can
be called multiple times per second, so work_on_cpu is too heavy.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
To: cpufreq@vger.kernel.org
Acked-by: Mark Langsdorf <mark.langsdorf@amd.com>
Tested-by: Mark Langsdorf <mark.langsdorf@amd.com>
Signed-off-by: Dave Jones <davej@redhat.com>
arch/x86/kernel/cpu/cpufreq/powernow-k8.c

index 20c7b99..1f55547 100644 (file)
@@ -508,41 +508,34 @@ static int core_voltage_post_transition(struct powernow_k8_data *data,
        return 0;
 }
 
-static int check_supported_cpu(unsigned int cpu)
+static void check_supported_cpu(void *_rc)
 {
-       cpumask_t oldmask;
        u32 eax, ebx, ecx, edx;
-       unsigned int rc = 0;
-
-       oldmask = current->cpus_allowed;
-       set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
+       int *rc = _rc;
 
-       if (smp_processor_id() != cpu) {
-               printk(KERN_ERR PFX "limiting to cpu %u failed\n", cpu);
-               goto out;
-       }
+       *rc = -ENODEV;
 
        if (current_cpu_data.x86_vendor != X86_VENDOR_AMD)
-               goto out;
+               return;
 
        eax = cpuid_eax(CPUID_PROCESSOR_SIGNATURE);
        if (((eax & CPUID_XFAM) != CPUID_XFAM_K8) &&
            ((eax & CPUID_XFAM) < CPUID_XFAM_10H))
-               goto out;
+               return;
 
        if ((eax & CPUID_XFAM) == CPUID_XFAM_K8) {
                if (((eax & CPUID_USE_XFAM_XMOD) != CPUID_USE_XFAM_XMOD) ||
                    ((eax & CPUID_XMOD) > CPUID_XMOD_REV_MASK)) {
                        printk(KERN_INFO PFX
                                "Processor cpuid %x not supported\n", eax);
-                       goto out;
+                       return;
                }
 
                eax = cpuid_eax(CPUID_GET_MAX_CAPABILITIES);
                if (eax < CPUID_FREQ_VOLT_CAPABILITIES) {
                        printk(KERN_INFO PFX
                               "No frequency change capabilities detected\n");
-                       goto out;
+                       return;
                }
 
                cpuid(CPUID_FREQ_VOLT_CAPABILITIES, &eax, &ebx, &ecx, &edx);
@@ -550,21 +543,17 @@ static int check_supported_cpu(unsigned int cpu)
                        != P_STATE_TRANSITION_CAPABLE) {
                        printk(KERN_INFO PFX
                                "Power state transitions not supported\n");
-                       goto out;
+                       return;
                }
        } else { /* must be a HW Pstate capable processor */
                cpuid(CPUID_FREQ_VOLT_CAPABILITIES, &eax, &ebx, &ecx, &edx);
                if ((edx & USE_HW_PSTATE) == USE_HW_PSTATE)
                        cpu_family = CPU_HW_PSTATE;
                else
-                       goto out;
+                       return;
        }
 
-       rc = 1;
-
-out:
-       set_cpus_allowed_ptr(current, &oldmask);
-       return rc;
+       *rc = 0;
 }
 
 static int check_pst_table(struct powernow_k8_data *data, struct pst_s *pst,
@@ -1247,6 +1236,32 @@ static int powernowk8_verify(struct cpufreq_policy *pol)
        return cpufreq_frequency_table_verify(pol, data->powernow_table);
 }
 
+struct init_on_cpu {
+       struct powernow_k8_data *data;
+       int rc;
+};
+
+static void __cpuinit powernowk8_cpu_init_on_cpu(void *_init_on_cpu)
+{
+       struct init_on_cpu *init_on_cpu = _init_on_cpu;
+
+       if (pending_bit_stuck()) {
+               printk(KERN_ERR PFX "failing init, change pending bit set\n");
+               init_on_cpu->rc = -ENODEV;
+               return;
+       }
+
+       if (query_current_values_with_pending_wait(init_on_cpu->data)) {
+               init_on_cpu->rc = -ENODEV;
+               return;
+       }
+
+       if (cpu_family == CPU_OPTERON)
+               fidvid_msr_init();
+
+       init_on_cpu->rc = 0;
+}
+
 /* per CPU init entry point to the driver */
 static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
 {
@@ -1254,13 +1269,14 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
                KERN_ERR FW_BUG PFX "No compatible ACPI _PSS objects found.\n"
                KERN_ERR FW_BUG PFX "Try again with latest BIOS.\n";
        struct powernow_k8_data *data;
-       cpumask_t oldmask;
+       struct init_on_cpu init_on_cpu;
        int rc;
 
        if (!cpu_online(pol->cpu))
                return -ENODEV;
 
-       if (!check_supported_cpu(pol->cpu))
+       smp_call_function_single(pol->cpu, check_supported_cpu, &rc, 1);
+       if (rc)
                return -ENODEV;
 
        data = kzalloc(sizeof(struct powernow_k8_data), GFP_KERNEL);
@@ -1300,27 +1316,12 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
                pol->cpuinfo.transition_latency = get_transition_latency(data);
 
        /* only run on specific CPU from here on */
-       oldmask = current->cpus_allowed;
-       set_cpus_allowed_ptr(current, &cpumask_of_cpu(pol->cpu));
-
-       if (smp_processor_id() != pol->cpu) {
-               printk(KERN_ERR PFX "limiting to cpu %u failed\n", pol->cpu);
-               goto err_out_unmask;
-       }
-
-       if (pending_bit_stuck()) {
-               printk(KERN_ERR PFX "failing init, change pending bit set\n");
-               goto err_out_unmask;
-       }
-
-       if (query_current_values_with_pending_wait(data))
-               goto err_out_unmask;
-
-       if (cpu_family == CPU_OPTERON)
-               fidvid_msr_init();
-
-       /* run on any CPU again */
-       set_cpus_allowed_ptr(current, &oldmask);
+       init_on_cpu.data = data;
+       smp_call_function_single(data->cpu, powernowk8_cpu_init_on_cpu,
+                                &init_on_cpu, 1);
+       rc = init_on_cpu.rc;
+       if (rc != 0)
+               goto err_out_exit_acpi;
 
        if (cpu_family == CPU_HW_PSTATE)
                cpumask_copy(pol->cpus, cpumask_of(pol->cpu));
@@ -1357,8 +1358,7 @@ static int __cpuinit powernowk8_cpu_init(struct cpufreq_policy *pol)
 
        return 0;
 
-err_out_unmask:
-       set_cpus_allowed_ptr(current, &oldmask);
+err_out_exit_acpi:
        powernow_k8_cpu_exit_acpi(data);
 
 err_out:
@@ -1383,24 +1383,25 @@ static int __devexit powernowk8_cpu_exit(struct cpufreq_policy *pol)
        return 0;
 }
 
+static void query_values_on_cpu(void *_err)
+{
+       int *err = _err;
+       struct powernow_k8_data *data = __get_cpu_var(powernow_data);
+
+       *err = query_current_values_with_pending_wait(data);
+}
+
 static unsigned int powernowk8_get(unsigned int cpu)
 {
        struct powernow_k8_data *data = per_cpu(powernow_data, cpu);
-       cpumask_t oldmask = current->cpus_allowed;
        unsigned int khz = 0;
+       int err;
 
        if (!data)
                return -EINVAL;
 
-       set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
-       if (smp_processor_id() != cpu) {
-               printk(KERN_ERR PFX
-                       "limiting to CPU %d failed in powernowk8_get\n", cpu);
-               set_cpus_allowed_ptr(current, &oldmask);
-               return 0;
-       }
-
-       if (query_current_values_with_pending_wait(data))
+       smp_call_function_single(cpu, query_values_on_cpu, &err, true);
+       if (err)
                goto out;
 
        if (cpu_family == CPU_HW_PSTATE)
@@ -1411,7 +1412,6 @@ static unsigned int powernowk8_get(unsigned int cpu)
 
 
 out:
-       set_cpus_allowed_ptr(current, &oldmask);
        return khz;
 }
 
@@ -1437,7 +1437,9 @@ static int __cpuinit powernowk8_init(void)
        unsigned int i, supported_cpus = 0;
 
        for_each_online_cpu(i) {
-               if (check_supported_cpu(i))
+               int rc;
+               smp_call_function_single(i, check_supported_cpu, &rc, 1);
+               if (rc == 0)
                        supported_cpus++;
        }