Skip to contentSkip to conversation

cpufreq: Don't find governor for setpolicy drivers in cpufreq_init_policy()

PATCH In review Yue Hu 4 participants 8 messages Apr 29, 2019 – May 14, 2019 CPU frequency scaling framework

lore
Share & tools
lore

Yue Hu

author
#0 lore raw

PATCH v3 1/1 cpufreq: Don't find governor for setpolicy drivers in cpufreq_init_policy()

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

---
v2: fix ->setplicy typo.
v3:
- let cpufreq_parse_governor() only handle !set_policy.
- fix using {} in the if block.
- change helper function name.
- update comment, commit message.

 drivers/cpufreq/cpufreq.c | 116 ++++++++++++++++++++++++++--------------------
 1 file changed, 65 insertions(+), 51 deletions(-)
drivers/cpufreq/cpufreq.c +65 -51
index 0322cce..ce8a01d 100644
@@ -578,50 +578,52 @@ static struct cpufreq_governor *find_governor(const char *str_governor)
 	return NULL;
 }
 
+static int cpufreq_parse_policy(char *str_governor,
+				struct cpufreq_policy *policy)
+{
+	if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
+		policy->policy = CPUFREQ_POLICY_PERFORMANCE;
+		return 0;
+	}
+	if (!strncasecmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) {
+		policy->policy = CPUFREQ_POLICY_POWERSAVE;
+		return 0;
+	}
+	return -EINVAL;
+}
+
 /**
- * cpufreq_parse_governor - parse a governor string
+ * cpufreq_parse_governor - parse a governor string only for !setpolicy
  */
 static int cpufreq_parse_governor(char *str_governor,
 				  struct cpufreq_policy *policy)
 {
-	if (cpufreq_driver->setpolicy) {
-		if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
-			policy->policy = CPUFREQ_POLICY_PERFORMANCE;
-			return 0;
-		}
-
-		if (!strncasecmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) {
-			policy->policy = CPUFREQ_POLICY_POWERSAVE;
-			return 0;
-		}
-	} else {
-		struct cpufreq_governor *t;
+	struct cpufreq_governor *t;
 
-		mutex_lock(&cpufreq_governor_mutex);
+	mutex_lock(&cpufreq_governor_mutex);
 
-		t = find_governor(str_governor);
-		if (!t) {
-			int ret;
+	t = find_governor(str_governor);
+	if (!t) {
+		int ret;
 
-			mutex_unlock(&cpufreq_governor_mutex);
+		mutex_unlock(&cpufreq_governor_mutex);
 
-			ret = request_module("cpufreq_%s", str_governor);
-			if (ret)
-				return -EINVAL;
+		ret = request_module("cpufreq_%s", str_governor);
+		if (ret)
+			return -EINVAL;
 
-			mutex_lock(&cpufreq_governor_mutex);
+		mutex_lock(&cpufreq_governor_mutex);
 
-			t = find_governor(str_governor);
-		}
-		if (t && !try_module_get(t->owner))
-			t = NULL;
+		t = find_governor(str_governor);
+	}
+	if (t && !try_module_get(t->owner))
+		t = NULL;
 
-		mutex_unlock(&cpufreq_governor_mutex);
+	mutex_unlock(&cpufreq_governor_mutex);
 
-		if (t) {
-			policy->governor = t;
-			return 0;
-		}
+	if (t) {
+		policy->governor = t;
+		return 0;
 	}
 
 	return -EINVAL;
@@ -746,8 +748,13 @@ static ssize_t store_scaling_governor(struct cpufreq_policy *policy,
 	if (ret != 1)
 		return -EINVAL;
 
-	if (cpufreq_parse_governor(str_governor, &new_policy))
-		return -EINVAL;
+	if (cpufreq_driver->setpolicy) {
+		if (cpufreq_parse_policy(str_governor, &new_policy))
+			return -EINVAL;
+	} else {
+		if (cpufreq_parse_governor(str_governor, &new_policy))
+			return -EINVAL;
+	}
 
 	ret = cpufreq_set_policy(policy, &new_policy);
 
@@ -1020,32 +1027,39 @@ __weak struct cpufreq_governor *cpufreq_default_governor(void)
 
 static int cpufreq_init_policy(struct cpufreq_policy *policy)
 {
-	struct cpufreq_governor *gov = NULL;
+	struct cpufreq_governor *gov = NULL, *def_gov = NULL;
 	struct cpufreq_policy new_policy;
 
 	memcpy(&new_policy, policy, sizeof(*policy));
 
-	/* Update governor of new_policy to the governor used before hotplug */
-	gov = find_governor(policy->last_governor);
-	if (gov) {
-		pr_debug("Restoring governor %s for cpu %d\n",
+	def_gov = cpufreq_default_governor();
+
+	if (has_target()) {
+		/*
+		 * Update governor of new_policy to the governor used before
+		 * hotplug
+		 */
+		gov = find_governor(policy->last_governor);
+		if (gov) {
+			pr_debug("Restoring governor %s for cpu %d\n",
 				policy->governor->name, policy->cpu);
+		} else {
+			if (!def_gov)
+				return -ENODATA;
+			gov = def_gov;
+		}
+		new_policy.governor = gov;
 	} else {
-		gov = cpufreq_default_governor();
-		if (!gov)
-			return -ENODATA;
-	}
-
-	new_policy.governor = gov;
-
-	/* Use the default policy if there is no last_policy. */
-	if (cpufreq_driver->setpolicy) {
-		if (policy->last_policy)
+		/* Use the default policy if there is no last_policy. */
+		if (policy->last_policy) {
 			new_policy.policy = policy->last_policy;
-		else
-			cpufreq_parse_governor(gov->name, &new_policy);
+		} else {
+			if (!def_gov)
+				return -ENODATA;
+			cpufreq_parse_policy(def_gov->name, &new_policy);
+		}
 	}
-	/* set default policy */
+
 	return cpufreq_set_policy(policy, &new_policy);
 }
 
Signature · 1 line
1.9.1
maintainer Acked-by on v3 1/1
#1 lore raw
Quoted 20 lines · Yue Hu

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

---
v2: fix ->setplicy typo.
v3:
- let cpufreq_parse_governor() only handle !set_policy.
- fix using {} in the if block.
- change helper function name.
- update comment, commit message.

Quoted 2 lines
 drivers/cpufreq/cpufreq.c | 116 ++++++++++++++++++++++++++--------------------
 1 file changed, 65 insertions(+), 51 deletions(-)
Signature · 1 line
viresh
#2 lore raw
In reply to Yue Hu
Quoted 9 lines · Yue Hu

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Quoted 4 lines

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

Have you tested this with the intel_pstate driver (in the active mode)?

Quoted 176 lines

---
v2: fix ->setplicy typo.
v3:
- let cpufreq_parse_governor() only handle !set_policy.
- fix using {} in the if block.
- change helper function name.
- update comment, commit message.

 drivers/cpufreq/cpufreq.c | 116 ++++++++++++++++++++++++++--------------------
 1 file changed, 65 insertions(+), 51 deletions(-)
drivers/cpufreq/cpufreq.c +65 -52
index 0322cce..ce8a01d 100644
@@ -578,50 +578,52 @@ static struct cpufreq_governor *find_governor(const char *str_governor)
        return NULL;
 }
 
+static int cpufreq_parse_policy(char *str_governor,
+                               struct cpufreq_policy *policy)
+{
+       if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
+               policy->policy = CPUFREQ_POLICY_PERFORMANCE;
+               return 0;
+       }
+       if (!strncasecmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) {
+               policy->policy = CPUFREQ_POLICY_POWERSAVE;
+               return 0;
+       }
+       return -EINVAL;
+}
+
 /**
- * cpufreq_parse_governor - parse a governor string
+ * cpufreq_parse_governor - parse a governor string only for !setpolicy
  */
 static int cpufreq_parse_governor(char *str_governor,
                                  struct cpufreq_policy *policy)
 {
-       if (cpufreq_driver->setpolicy) {
-               if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
-                       policy->policy = CPUFREQ_POLICY_PERFORMANCE;
-                       return 0;
-               }
-
-               if (!strncasecmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) {
-                       policy->policy = CPUFREQ_POLICY_POWERSAVE;
-                       return 0;
-               }
-       } else {
-               struct cpufreq_governor *t;
+       struct cpufreq_governor *t;
 
-               mutex_lock(&cpufreq_governor_mutex);
+       mutex_lock(&cpufreq_governor_mutex);
 
-               t = find_governor(str_governor);
-               if (!t) {
-                       int ret;
+       t = find_governor(str_governor);
+       if (!t) {
+               int ret;
 
-                       mutex_unlock(&cpufreq_governor_mutex);
+               mutex_unlock(&cpufreq_governor_mutex);
 
-                       ret = request_module("cpufreq_%s", str_governor);
-                       if (ret)
-                               return -EINVAL;
+               ret = request_module("cpufreq_%s", str_governor);
+               if (ret)
+                       return -EINVAL;
 
-                       mutex_lock(&cpufreq_governor_mutex);
+               mutex_lock(&cpufreq_governor_mutex);
 
-                       t = find_governor(str_governor);
-               }
-               if (t && !try_module_get(t->owner))
-                       t = NULL;
+               t = find_governor(str_governor);
+       }
+       if (t && !try_module_get(t->owner))
+               t = NULL;
 
-               mutex_unlock(&cpufreq_governor_mutex);
+       mutex_unlock(&cpufreq_governor_mutex);
 
-               if (t) {
-                       policy->governor = t;
-                       return 0;
-               }
+       if (t) {
+               policy->governor = t;
+               return 0;
        }
 
        return -EINVAL;
@@ -746,8 +748,13 @@ static ssize_t store_scaling_governor(struct cpufreq_policy *policy,
        if (ret != 1)
                return -EINVAL;
 
-       if (cpufreq_parse_governor(str_governor, &new_policy))
-               return -EINVAL;
+       if (cpufreq_driver->setpolicy) {
+               if (cpufreq_parse_policy(str_governor, &new_policy))
+                       return -EINVAL;
+       } else {
+               if (cpufreq_parse_governor(str_governor, &new_policy))
+                       return -EINVAL;
+       }
 
        ret = cpufreq_set_policy(policy, &new_policy);
 
@@ -1020,32 +1027,39 @@ __weak struct cpufreq_governor *cpufreq_default_governor(void)
 
 static int cpufreq_init_policy(struct cpufreq_policy *policy)
 {
-       struct cpufreq_governor *gov = NULL;
+       struct cpufreq_governor *gov = NULL, *def_gov = NULL;
        struct cpufreq_policy new_policy;
 
        memcpy(&new_policy, policy, sizeof(*policy));
 
-       /* Update governor of new_policy to the governor used before hotplug */
-       gov = find_governor(policy->last_governor);
-       if (gov) {
-               pr_debug("Restoring governor %s for cpu %d\n",
+       def_gov = cpufreq_default_governor();
+
+       if (has_target()) {
+               /*
+                * Update governor of new_policy to the governor used before
+                * hotplug
+                */
+               gov = find_governor(policy->last_governor);
+               if (gov) {
+                       pr_debug("Restoring governor %s for cpu %d\n",
                                policy->governor->name, policy->cpu);
+               } else {
+                       if (!def_gov)
+                               return -ENODATA;
+                       gov = def_gov;
+               }
+               new_policy.governor = gov;
        } else {
-               gov = cpufreq_default_governor();
-               if (!gov)
-                       return -ENODATA;
-       }
-
-       new_policy.governor = gov;
-
-       /* Use the default policy if there is no last_policy. */
-       if (cpufreq_driver->setpolicy) {
-               if (policy->last_policy)
+               /* Use the default policy if there is no last_policy. */
+               if (policy->last_policy) {
                        new_policy.policy = policy->last_policy;
-               else
-                       cpufreq_parse_governor(gov->name, &new_policy);
+               } else {
+                       if (!def_gov)
+                               return -ENODATA;
+                       cpufreq_parse_policy(def_gov->name, &new_policy);
+               }
        }
-       /* set default policy */
+
        return cpufreq_set_policy(policy, &new_policy);
 }
 
--

1.9.1

author
#3 lore raw
Quoted 10 lines · On Mon, 29 Apr 2019 09:37:27 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 9 lines · Yue Hu

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Quoted 6 lines
Quoted 4 lines

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

Have you tested this with the intel_pstate driver (in the active mode)?

No, just tested for ARM. It should be common logic from code perspective. Has any issue in this change?

Quoted 177 lines
Quoted 177 lines

---
v2: fix ->setplicy typo.
v3:
- let cpufreq_parse_governor() only handle !set_policy.
- fix using {} in the if block.
- change helper function name.
- update comment, commit message.

 drivers/cpufreq/cpufreq.c | 116 ++++++++++++++++++++++++++--------------------
 1 file changed, 65 insertions(+), 51 deletions(-)
drivers/cpufreq/cpufreq.c +65 -52
index 0322cce..ce8a01d 100644
@@ -578,50 +578,52 @@ static struct cpufreq_governor *find_governor(const char *str_governor)
        return NULL;
 }
 
+static int cpufreq_parse_policy(char *str_governor,
+                               struct cpufreq_policy *policy)
+{
+       if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
+               policy->policy = CPUFREQ_POLICY_PERFORMANCE;
+               return 0;
+       }
+       if (!strncasecmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) {
+               policy->policy = CPUFREQ_POLICY_POWERSAVE;
+               return 0;
+       }
+       return -EINVAL;
+}
+
 /**
- * cpufreq_parse_governor - parse a governor string
+ * cpufreq_parse_governor - parse a governor string only for !setpolicy
  */
 static int cpufreq_parse_governor(char *str_governor,
                                  struct cpufreq_policy *policy)
 {
-       if (cpufreq_driver->setpolicy) {
-               if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
-                       policy->policy = CPUFREQ_POLICY_PERFORMANCE;
-                       return 0;
-               }
-
-               if (!strncasecmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) {
-                       policy->policy = CPUFREQ_POLICY_POWERSAVE;
-                       return 0;
-               }
-       } else {
-               struct cpufreq_governor *t;
+       struct cpufreq_governor *t;
 
-               mutex_lock(&cpufreq_governor_mutex);
+       mutex_lock(&cpufreq_governor_mutex);
 
-               t = find_governor(str_governor);
-               if (!t) {
-                       int ret;
+       t = find_governor(str_governor);
+       if (!t) {
+               int ret;
 
-                       mutex_unlock(&cpufreq_governor_mutex);
+               mutex_unlock(&cpufreq_governor_mutex);
 
-                       ret = request_module("cpufreq_%s", str_governor);
-                       if (ret)
-                               return -EINVAL;
+               ret = request_module("cpufreq_%s", str_governor);
+               if (ret)
+                       return -EINVAL;
 
-                       mutex_lock(&cpufreq_governor_mutex);
+               mutex_lock(&cpufreq_governor_mutex);
 
-                       t = find_governor(str_governor);
-               }
-               if (t && !try_module_get(t->owner))
-                       t = NULL;
+               t = find_governor(str_governor);
+       }
+       if (t && !try_module_get(t->owner))
+               t = NULL;
 
-               mutex_unlock(&cpufreq_governor_mutex);
+       mutex_unlock(&cpufreq_governor_mutex);
 
-               if (t) {
-                       policy->governor = t;
-                       return 0;
-               }
+       if (t) {
+               policy->governor = t;
+               return 0;
        }
 
        return -EINVAL;
@@ -746,8 +748,13 @@ static ssize_t store_scaling_governor(struct cpufreq_policy *policy,
        if (ret != 1)
                return -EINVAL;
 
-       if (cpufreq_parse_governor(str_governor, &new_policy))
-               return -EINVAL;
+       if (cpufreq_driver->setpolicy) {
+               if (cpufreq_parse_policy(str_governor, &new_policy))
+                       return -EINVAL;
+       } else {
+               if (cpufreq_parse_governor(str_governor, &new_policy))
+                       return -EINVAL;
+       }
 
        ret = cpufreq_set_policy(policy, &new_policy);
 
@@ -1020,32 +1027,39 @@ __weak struct cpufreq_governor *cpufreq_default_governor(void)
 
 static int cpufreq_init_policy(struct cpufreq_policy *policy)
 {
-       struct cpufreq_governor *gov = NULL;
+       struct cpufreq_governor *gov = NULL, *def_gov = NULL;
        struct cpufreq_policy new_policy;
 
        memcpy(&new_policy, policy, sizeof(*policy));
 
-       /* Update governor of new_policy to the governor used before hotplug */
-       gov = find_governor(policy->last_governor);
-       if (gov) {
-               pr_debug("Restoring governor %s for cpu %d\n",
+       def_gov = cpufreq_default_governor();
+
+       if (has_target()) {
+               /*
+                * Update governor of new_policy to the governor used before
+                * hotplug
+                */
+               gov = find_governor(policy->last_governor);
+               if (gov) {
+                       pr_debug("Restoring governor %s for cpu %d\n",
                                policy->governor->name, policy->cpu);
+               } else {
+                       if (!def_gov)
+                               return -ENODATA;
+                       gov = def_gov;
+               }
+               new_policy.governor = gov;
        } else {
-               gov = cpufreq_default_governor();
-               if (!gov)
-                       return -ENODATA;
-       }
-
-       new_policy.governor = gov;
-
-       /* Use the default policy if there is no last_policy. */
-       if (cpufreq_driver->setpolicy) {
-               if (policy->last_policy)
+               /* Use the default policy if there is no last_policy. */
+               if (policy->last_policy) {
                        new_policy.policy = policy->last_policy;
-               else
-                       cpufreq_parse_governor(gov->name, &new_policy);
+               } else {
+                       if (!def_gov)
+                               return -ENODATA;
+                       cpufreq_parse_policy(def_gov->name, &new_policy);
+               }
        }
-       /* set default policy */
+
        return cpufreq_set_policy(policy, &new_policy);
 }
 
--

1.9.1

#4 lore raw
Quoted 13 lines · CEST Yue Hu
Quoted 10 lines · On Mon, 29 Apr 2019 09:37:27 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 9 lines · Yue Hu

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

Have you tested this with the intel_pstate driver (in the active mode)?

No, just tested for ARM. It should be common logic from code perspective.

But it is prudent to test changes on various configurations that may be affected.

Testing intel_pstate shouldn't be too difficult.

Quoted 1 line

Has any issue in this change?

Not in principle, but I need to check the details.

In general I'm a bit hesitant to take changes that haven't been tested properly.

Thanks,
Rafael

author
#5 lore raw
Quoted 25 lines · On Mon, 29 Apr 2019 10:37:45 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 13 lines · CEST Yue Hu
Quoted 10 lines · On Mon, 29 Apr 2019 09:37:27 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 9 lines · Yue Hu

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

Have you tested this with the intel_pstate driver (in the active mode)?

No, just tested for ARM. It should be common logic from code perspective.

But it is prudent to test changes on various configurations that may be affected.

Quoted 7 lines

Testing intel_pstate shouldn't be too difficult.

Quoted 1 line

Has any issue in this change?

Not in principle, but I need to check the details.

In general I'm a bit hesitant to take changes that haven't been tested properly.

Ok, i will try to test it with intel_pstate driver later.

Quoted 6 lines

Thanks,
Rafael

author
#6 lore raw
In reply to Rafael J. Wysocki
Quoted 19 lines · On Mon, 29 Apr 2019 10:37:45 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 18 lines · CEST Yue Hu
Quoted 15 lines · On Mon, 29 Apr 2019 09:37:27 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 13 lines · Yue Hu

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

Have you tested this with the intel_pstate driver (in the active mode)?

No, just tested for ARM. It should be common logic from code perspective.

But it is prudent to test changes on various configurations that may be affected.

Testing intel_pstate shouldn't be too difficult.

Hi Rafael,

Sorry for late test since i take time to emulate/find the board with Sandy brigdge processor.

I'm sure intel_pstate setpolicy driver is working fine by tests containing policy init, hotplug and changing scaling_governor.

Thx.

Quoted 12 lines
Quoted 2 lines

Has any issue in this change?

Not in principle, but I need to check the details.

In general I'm a bit hesitant to take changes that haven't been tested properly.

Thanks,
Rafael

#7 lore raw
Quoted 30 lines · CEST Yue Hu
Quoted 19 lines · On Mon, 29 Apr 2019 10:37:45 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 18 lines · CEST Yue Hu
Quoted 15 lines · On Mon, 29 Apr 2019 09:37:27 +0200 "Rafael J. Wysocki" <[email protected]>…
Quoted 13 lines · Yue Hu

From: Yue Hu <[email protected]>

In cpufreq_init_policy() we will check if there's last_governor for target and setpolicy type. However last_governor is set only if has_target() is true in cpufreq_offline(). That means find last_governor for setpolicy type is pointless. Also new_policy.governor will not be used if ->setpolicy callback is set in cpufreq_set_policy().

Moreover, there's duplicate ->setpolicy check in using default policy path. Let's add a new helper function to avoid it. Also update a little comment.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

Have you tested this with the intel_pstate driver (in the active mode)?

No, just tested for ARM. It should be common logic from code perspective.

But it is prudent to test changes on various configurations that may be affected.

Testing intel_pstate shouldn't be too difficult.

Quoted 7 lines

Hi Rafael,

Sorry for late test since i take time to emulate/find the board with Sandy brigdge processor.

I'm sure intel_pstate setpolicy driver is working fine by tests containing policy init, hotplug and changing scaling_governor.

Now applied, thanks!