adding handles to pm_qos?
We are calling pm_qos from some of our drivers. One concern is that each call of pm_qos_update_requirement() iterates through the client list and strcmp the client names. It could be slow. A proposal is that pm_qos provides handles that can be used on pm_qos_update_requirement().
For measurement purposes, I added get/put interfaces to acquire/release the handles and a new pm_qos_update_requirement function that bypasses the iteration and strcmp. Here are some collected data:
How many clock cycles does pm_qos_update_requirement take? when there is one client on this qos_class:
using handle using name using handle/using name avg 252.4 400.4 63%
when there are 5 clients on this qos_class:
using handle using name using handle/using name avg 407.6 644.8 63%
when there are 10 clients on this qos_class:
using handle using name using handle/using name avg 582.4 938.4 62%
Given the time differences, it seems worthwhile to add handles.
~Ai
Quoted 17 lines · Ai Li
We are calling
pm_qosfrom some of our drivers. One concern is that each call ofpm_qos_update_requirement()iterates through the client list and strcmp the client names. It could be slow. A proposal is thatpm_qosprovides handles that can be used onpm_qos_update_requirement().For measurement purposes, I added get/put interfaces to acquire/release the handles and a new
pm_qos_update_requirementfunction that bypasses the iteration and strcmp. Here are some collected data:How many clock cycles does
pm_qos_update_requirementtake? when there is one client on thisqos_class:using handle using name using handle/using name avg 252.4 400.4 63%when there are 5 clients on this
qos_class:
Quoted 8 lines
using handle using name using handle/using name avg 407.6 644.8 63%when there are 10 clients on this
qos_class:using handle using name using handle/using name avg 582.4 938.4 62%Given the time differences, it seems worthwhile to add handles.
How often are you calling pm_qos_update_requirement?
I think calling pm_qos_ interfaces too often makes me wonder about my assumptions and your sanity.
Can you explain why the pm_qos_update_requirement is getting hit often enough to bother with this change?
Other than that I don't have a problem with moving to handles, if its a practical change made for reasons other than making api abuse less painful.
Further, If the implicit assumption that pmqos calls are on cold paths is wrong, then perhaps more thought is needed than just changing things to handle based searches.
--mgross
Quoted 6 lines
~Ai
_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm
Quoted 16 lines
How often are you calling
pm_qos_update_requirement?I think calling
pm_qos_interfaces too often makes me wonder about my
assumptions and your sanity.Can you explain why the
pm_qos_update_requirementis getting hit often
enough to bother with this change?Other than that I don't have a problem with moving to handles, if its a
practical change made for reasons other than making api abuse less
painful.
Quoted 6 lines
Further, If the implicit assumption that pmqos calls are on cold paths
is wrong, then perhaps more thought is needed than just changing things
to handle based searches.
Our embedded platforms support different low power modes. With the modes, the deeper the sleep, the more the power savings, and the larger the interrupt latency coming out of the low power mode.
To help the platform achieving greatest power savings, some of our device drivers set lateny qos only when there is a service request to the driver or a device transaction. When the transaction or request is done, the drivers cancel the QoS with pm_qos_update_requirement(PM_QOS_DEFAULT_VALUE), allowing the platform to reach a deeper sleep.
The approach gives us good power savings. However when there are lots of transactions, pm_qos_update_requirement() gets called a lot of times.
~Ai
Quoted 34 lines · Ai Li
Quoted 16 lines
How often are you calling
pm_qos_update_requirement?I think calling
pm_qos_interfaces too often makes me wonder about my
assumptions and your sanity.Can you explain why the
pm_qos_update_requirementis getting hit often
enough to bother with this change?Other than that I don't have a problem with moving to handles, if its a
practical change made for reasons other than making api abuse less
painful.Quoted 6 lines
Further, If the implicit assumption that pmqos calls are on cold paths
is wrong, then perhaps more thought is needed than just changing things
to handle based searches.Our embedded platforms support different low power modes. With the modes, the deeper the sleep, the more the power savings, and the larger the interrupt latency coming out of the low power mode.
To help the platform achieving greatest power savings, some of our device drivers set lateny qos only when there is a service request to the driver or a device transaction. When the transaction or request is done, the drivers cancel the QoS with
pm_qos_update_requirement(PM_QOS_DEFAULT_VALUE), allowing the platform to reach a deeper sleep.
Quoted 3 lines
The approach gives us good power savings. However when there are lots of transactions,
pm_qos_update_requirement()gets called a lot of times.
Oh.
This will not scale with the aggregation logic very well at all if pm_qos update requirement gets hit per transaction through a driver code path, then I think some thought on the scalability is needed and perhaps a change to the aggregation design for such uses.
Do you have a patch for the handle implementation I could look at?
--mgross
.
Quoted 3 lines
~Ai
Quoted 10 lines
Oh.
This will not scale with the aggregation logic very well at all if
pm_qosupdate requirement gets hit per transaction through a driver code
path, then I think some thought on the scalability is needed and perhaps
a change to the aggregation design for such uses.
Quoted 2 lines
Do you have a patch for the handle implementation I could look at?
Here is the patch that I use with the test code. There are three new functions:
void *pm_qos_get(int qos, char *name);
int pm_qos_put(void *handle);
int pm_qos_update_requirement_direct(void *handle, s32 new_value);
In the test, I wanted to keep the existing interface intact so I could compare them at the same time. For the formal code submission, a different approach may work better.
linux/pm_qos_params.h renamed from linux/pm_qos_params.h.orig +5 -1
@@ -16,7 +16,11 @@int pm_qos_add_requirement (int qos,char *name,s32 value);int pm_qos_update_requirement (int qos,char *name,s32 new_value); -void pm_qos_remove_requirement (int qos,char *name); +int pm_qos_remove_requirement (int qos,char *name); + +void *pm_qos_get (int qos,char *name); +int pm_qos_put (void *handle); +int pm_qos_update_requirement_direct (void *handle,s32 new_value);int pm_qos_requirement (int qos);
pm_qos_params.c renamed from pm_qos_params.c.orig +4 -0
@@ -55,6 +55,8 @@ struct requirement_list {s32 kbps; };char *name; +int pm_qos_class; +int handle_count; };static s32 max_compare (s32 v1,s32 v2);
@@ -210,7 +212,9 @@ EXPORT_SYMBOL_GPL(pm_qos_requirement);int pm_qos_add_requirement (int pm_qos_class,char *name,s32 value) {struct requirement_list *dep; +struct requirement_list *node;unsigned long flags; +int retval; dep =kzalloc (sizeof (struct requirement_list ),GFP_KERNEL );if (dep) {
@@ -219,10 +223,25 @@ int pm_qos_add_requirement(int pm_qos_class,
char *name, s32 value)
else
dep->value = value;
dep->name = kstrdup(name, GFP_KERNEL);
- if (!dep->name)
+ if (!dep->name) {
+ retval = -ENOMEM;
goto cleanup;
+ }
+ dep->pm_qos_class = pm_qos_class;
spin_lock_irqsave(&pm_qos_lock, flags);
+ /* make sure the new name is unique */
+ list_for_each_entry(node,
+ &pm_qos_array[pm_qos_class]->requirements.list, list) {
+ if (strcmp(node->name, name) == 0) {
+ spin_unlock_irqrestore(&pm_qos_lock,
flags);
+
+ kfree(dep->name);
+ retval = -EINVAL;
+ goto cleanup;
+ }
+ }
+
list_add(&dep->list,
&pm_qos_array[pm_qos_class]->requirements.list);
spin_unlock_irqrestore(&pm_qos_lock, flags);
@@ -231,9 +250,11 @@ int pm_qos_add_requirement(int pm_qos_class, char *name, s32 value)
return 0;
}
+ retval = -ENOMEM;
+
cleanup:
kfree(dep);
- return -ENOMEM;
+ return retval;
}
EXPORT_SYMBOL_GPL(pm_qos_add_requirement);
@@ -282,8 +303,10 @@ EXPORT_SYMBOL_GPL(pm_qos_update_requirement);
*
* Will remove named qos request from pm_qos_class list of
parameters and
* recompute the current target value for the pm_qos_class.
+ *
+ * If there are outstanding handles to the request, -EBUSY is returned.
*/
-void pm_qos_remove_requirement(int pm_qos_class, char *name) +int pm_qos_remove_requirement(int pm_qos_class, char *name)
{
unsigned long flags;
struct requirement_list *node;
@@ -293,6 +316,11 @@ void pm_qos_remove_requirement(int pm_qos_class, char *name)
list_for_each_entry(node,
&pm_qos_array[pm_qos_class]->requirements.list, list)
{
if (strcmp(node->name, name) == 0) {
+ if (node->handle_count > 0) {
+ spin_unlock_irqrestore(&pm_qos_lock,
flags);
+ return -EBUSY;
+ }
+
kfree(node->name); list_del(&node->list); kfree(node);
@@ -303,9 +331,83 @@ void pm_qos_remove_requirement(int pm_qos_class, char *name)
spin_unlock_irqrestore(&pm_qos_lock, flags);
if (pending_update)
update_target(pm_qos_class);
+
+ return 0;
}
EXPORT_SYMBOL_GPL(pm_qos_remove_requirement);
+/* pm_qos_get - returns an opaque handle to the existing qos request
+ * @pm_qos_class: identifies which list of qos request to us
+ * @name: identifies the request
+ *
+ * Will increment the reference count as well.
+ *
+ * If the named request isn't in the list then NULL is returned.
+ */
+void *pm_qos_get(int pm_qos_class, char *name)
+{
+ unsigned long flags;
+ struct requirement_list *node;
+ struct requirement_list *handle = NULL;
+
+ spin_lock_irqsave(&pm_qos_lock, flags);
+ list_for_each_entry(node,
+ &pm_qos_array[pm_qos_class]->requirements.list, list)
{
+ if (strcmp(node->name, name) == 0) {
+ node->handle_count++;
+ handle = node;
+ break;
+ }
+ }
+ spin_unlock_irqrestore(&pm_qos_lock, flags);
+ return handle;
+}
+EXPORT_SYMBOL_GPL(pm_qos_get);
+
+/* pm_qos_put - release the handle of the existing qos request
+ * @handle: previously acquired through pm_qos_get
+ *
+ * Will decrement the reference count as well. Caller should no
+ * longer use the handle after this call.
+ */
+int pm_qos_put(void *handle)
+{
+ unsigned long flags;
+ struct requirement_list *node = (struct requirement_list *)
handle;
+
+ spin_lock_irqsave(&pm_qos_lock, flags);
+ BUG_ON(node->handle_count == 0);
+ node->handle_count--;
+ spin_unlock_irqrestore(&pm_qos_lock, flags);
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(pm_qos_put);
+
+/* pm_qos_update_requirement_direct - modifies an existing qos request
+ * @handle: previously acquired through pm_qos_get
+ * @value: defines the qos request
+ *
+ * Updates an existing qos requirement for the pm_qos_class of parameters along
+ * with updating the target pm_qos_class value.
+ */
+int pm_qos_update_requirement_direct(void *handle, s32 new_value) +{
+ unsigned long flags;
+ struct requirement_list *node = (struct requirement_list *)
handle;
+
+ spin_lock_irqsave(&pm_qos_lock, flags);
+ if (new_value == PM_QOS_DEFAULT_VALUE)
+ node->value =
pm_qos_array[node->pm_qos_class]->default_value;
+ else
+ node->value = new_value;
+ spin_unlock_irqrestore(&pm_qos_lock, flags);
+ update_target(node->pm_qos_class);
+
+ return 0;
+}
+EXPORT_SYMBOL_GPL(pm_qos_update_requirement_direct);
+
/**
* pm_qos_add_notifier - sets notification entry for changes to
target value
* @pm_qos_class: identifies which qos target changes should be
notified.
~Ai
Thanks, I'll look at it over the next few days.
--mgross
Quoted 231 lines · Ai Li
Quoted 10 lines
Oh.
This will not scale with the aggregation logic very well at all if
pm_qosupdate requirement gets hit per transaction through a driver code
path, then I think some thought on the scalability is needed and perhaps
a change to the aggregation design for such uses.Quoted 2 lines
Do you have a patch for the handle implementation I could look at?
Here is the patch that I use with the test code. There are three new functions:
void *pm_qos_get(int qos, char *name);
intpm_qos_put(void *handle);
intpm_qos_update_requirement_direct(void *handle, s32 new_value);In the test, I wanted to keep the existing interface intact so I could compare them at the same time. For the formal code submission, a different approach may work better.
linux/pm_qos_params.h renamed from linux/pm_qos_params.h.orig +5
-1@@ -16,7 +16,11 @@int pm_qos_add_requirement (int qos,char *name,s32 value);int pm_qos_update_requirement (int qos,char *name,s32 new_value); -void pm_qos_remove_requirement (int qos,char *name); +int pm_qos_remove_requirement (int qos,char *name); + +void *pm_qos_get (int qos,char *name); +int pm_qos_put (void *handle); +int pm_qos_update_requirement_direct (void *handle,s32 new_value);int pm_qos_requirement (int qos);pm_qos_params.c renamed from pm_qos_params.c.orig +20
-1@@ -55,6 +55,8 @@ struct requirement_list {s32 kbps; };char *name; +int pm_qos_class; +int handle_count; };static s32 max_compare (s32 v1,s32 v2);@@ -210,7 +212,9 @@ EXPORT_SYMBOL_GPL(pm_qos_requirement);int pm_qos_add_requirement (int pm_qos_class,char *name,s32 value) {struct requirement_list *dep; +struct requirement_list *node;unsigned long flags; +int retval; dep =kzalloc (sizeof (struct requirement_list ),GFP_KERNEL );if (dep) {@@ -219,10 +223,25 @@ int pm_qos_add_requirement(int pm_qos_class, char *name, s32 value) else dep->value = value; dep->name =kstrdup (name,GFP_KERNEL ); -if (!dep->name) +if (!dep->name) { + retval = -ENOMEM ; goto cleanup; + } + dep->pm_qos_class = pm_qos_class;spin_lock_irqsave (&pm_qos_lock, flags); +/* make sure the new name is unique */ +list_for_each_entry (node, + &pm_qos_array[pm_qos_class]->requirements.list, list) { +if (strcmp (node->name, name) ==0 ) { +spin_unlock_irqrestore (&pm_qos_lock, flags); + +kfree (dep->name); + retval = -EINVAL ; + goto cleanup; + } + } +list_add(&dep->list,
&pm_qos_array[pm_qos_class]->requirements.list);
pm_qos_params.c renamed from pm_qos_params.c.orig +3
-1spin_unlock_irqrestore (&pm_qos_lock, flags);@@ -231,9 +250,11 @@ int pm_qos_add_requirement(int pm_qos_class, char *name, s32 value) return 0; } + retval = -ENOMEM ; + cleanup:kfree (dep); -return -ENOMEM ; +return retval; }EXPORT_SYMBOL_GPL (pm_qos_add_requirement);@@ -282,8 +303,10 @@ EXPORT_SYMBOL_GPL(pm_qos_update_requirement);* * Will remove named qos request from pm_qos_class list of parameters and
* recompute the current target value for thepm_qos_class.
+ *
+ * If there are outstanding handles to the request,-EBUSYis returned.
*/pm_qos_params.c renamed from pm_qos_params.c.orig +6
-1-void pm_qos_remove_requirement (int pm_qos_class,char *name) +int pm_qos_remove_requirement (int pm_qos_class,char *name) {unsigned long flags;struct requirement_list *node;@@ -293,6 +316,11 @@ void pm_qos_remove_requirement(int pm_qos_class, char *name)list_for_each_entry (node, &pm_qos_array[pm_qos_class]->requirements.list, list) {if (strcmp (node->name, name) ==0 ) { +if (node->handle_count >0 ) { +spin_unlock_irqrestore (&pm_qos_lock, flags); +return -EBUSY ; + } +pm_qos_params.c renamed from pm_qos_params.c.orig +51
-0kfree (node->name);list_del (&node->list);kfree (node);@@ -303,9 +331,83 @@ void pm_qos_remove_requirement(int pm_qos_class, char *name) spin_unlock_irqrestore(&pm_qos_lock, flags);if (pending_update)update_target (pm_qos_class); + +return 0 ; }EXPORT_SYMBOL_GPL (pm_qos_remove_requirement); +/* pm_qos_get - returns an opaque handle to the existing qos request +* @pm_qos_class: identifies which list of qos request to us +* @name: identifies the request +* +* Will increment the reference count as well. +* +* If the named request isn't in the list then NULL is returned. +*/ +void *pm_qos_get (int pm_qos_class,char *name) +{ +unsigned long flags; +struct requirement_list *node; +struct requirement_list *handle =NULL ; + +spin_lock_irqsave (&pm_qos_lock, flags); +list_for_each_entry (node, + &pm_qos_array[pm_qos_class]->requirements.list, list) { +if (strcmp (node->name, name) ==0 ) { + node->handle_count++; + handle = node; +break ; + } + } +spin_unlock_irqrestore (&pm_qos_lock, flags); +return handle; +} +EXPORT_SYMBOL_GPL (pm_qos_get); + +/* pm_qos_put - release the handle of the existing qos request +* @handle: previously acquired through pm_qos_get +* +* Will decrement the reference count as well. Caller should no +* longer use the handle after this call. +*/ +int pm_qos_put (void *handle) +{ +unsigned long flags; +struct requirement_list *node = (struct requirement_list *) handle; + +spin_lock_irqsave (&pm_qos_lock, flags); +BUG_ON (node->handle_count ==0 ); + node->handle_count--; +spin_unlock_irqrestore (&pm_qos_lock, flags); + +return 0 ; +} +EXPORT_SYMBOL_GPL (pm_qos_put); + +/* pm_qos_update_requirement_direct - modifies an existing qos request
+ * @handle: previously acquired throughpm_qos_get
+ * @value: defines the qos request
+ *
+ * Updates an existing qos requirement for thepm_qos_classof parameters along
+ * with updating the targetpm_qos_classvalue.
+ */pm_qos_params.c renamed from pm_qos_params.c.orig +4
-0+int pm_qos_update_requirement_direct (void *handle,s32 new_value) +{ +unsigned long flags; + struct requirement_list *node = (struct requirement_list *)handle;
pm_qos_params.c renamed from pm_qos_params.c.orig +4
-0+ +spin_lock_irqsave (&pm_qos_lock, flags); +if (new_value ==PM_QOS_DEFAULT_VALUE ) + node->value =pm_qos_array[node->pm_qos_class]->default_value;
pm_qos_params.c renamed from pm_qos_params.c.orig +9
-0+else + node->value = new_value; +spin_unlock_irqrestore (&pm_qos_lock, flags); +update_target (node->pm_qos_class); + +return 0 ; +} +EXPORT_SYMBOL_GPL (pm_qos_update_requirement_direct); +/** * pm_qos_add_notifier - sets notification entry for changes to target value
* @pm_qos_class: identifies which qos target changes should be
notified.~Ai
Quoted 11 lines
Thanks, I'll look at it over the next few days.
--mgross
Quoted 1 line
Here is the patch that I use with the test code. There are
three new
Quoted 5 lines
functions:
void *pm_qos_get(int qos, char *name);
intpm_qos_put(void *handle);
int pm_qos_update_requirement_direct(void *handle, s32
Quoted 7 lines
new_value);Quoted 2 lines
In the test, I wanted to keep the existing interface intact so
I
Quoted 1 line
could compare them at the same time. For the formal code
submission,
Quoted 1 line
a different approach may work better.
Here is an alternate way of adding handles to pm_qos_params that I was alluding to. This approach may be more preferable because it does not bloat the API and the handles become an integral part of pm_qos_params. In my previous patch, handles are kind of bolted onto pm_qos_params and it needs separate calls (pm_qos_get, pm_qos_put) to acquire and release the handle. In this patch, pm_qos_add_requirement and pm_qos_remove_requirement automatically take care of the handles.
linux/pm_qos_params.h renamed from linux/pm_qos_params.h.orig +3 -3
@@ -14,9 +14,11 @@#define PM_QOS_NUM_CLASSES 4#define PM_QOS_DEFAULT_VALUE -1 -int pm_qos_add_requirement (int qos,char *name,s32 value); -int pm_qos_update_requirement (int qos,char *name,s32 new_value); -void pm_qos_remove_requirement (int qos,char *name); +struct requirement_list ; + +struct requirement_list *pm_qos_add_requirement(int qos, char *name,
s32 value);
+int pm_qos_update_requirement(struct requirement_list *qos, s32 new_value);
+void pm_qos_remove_requirement(struct requirement_list *qos);
int pm_qos_requirement(int qos);
pm_qos_params.c renamed from pm_qos_params.c.orig +1 -0
@@ -49,6 +49,7 @@*/ struct requirement_list {struct list_head list; +int pm_qos_class;union {s32 value;s32 usec;
@@ -207,13 +208,16 @@ EXPORT_SYMBOL_GPL(pm_qos_requirement);* performance characteristics. It recomputes the aggregate QoS
expectations
* for the pm_qos_class of parameters.
*/
-int pm_qos_add_requirement(int pm_qos_class, char *name, s32 value) +struct requirement_list *pm_qos_add_requirement(int pm_qos_class, char *name,
+ s32 value)
{
struct requirement_list *dep;
unsigned long flags;
dep = kzalloc(sizeof(struct requirement_list), GFP_KERNEL); if (dep) {
+ dep->pm_qos_class = pm_qos_class;
+
if (value == PM_QOS_DEFAULT_VALUE)
dep->value =
pm_qos_array[pm_qos_class]->default_value;
else
@@ -228,48 +232,37 @@ int pm_qos_add_requirement(int pm_qos_class, char *name, s32 value)
spin_unlock_irqrestore(&pm_qos_lock, flags); update_target(pm_qos_class);
- return 0;
+ return dep;
}
cleanup:
kfree(dep);
- return -ENOMEM;
+ return ERR_PTR(-ENOMEM);
}
EXPORT_SYMBOL_GPL(pm_qos_add_requirement);
/**
* pm_qos_update_requirement - modifies an existing qos request
- * @pm_qos_class: identifies which list of qos request to us
- * @name: identifies the request
+ * @qos: identifies the qos request to us
* @value: defines the qos request
*
- * Updates an existing qos requirement for the pm_qos_class of parameters along
+ * Updates an existing qos requirement along
* with updating the target pm_qos_class value.
- *
- * If the named request isn't in the list then no change is made. */
-int pm_qos_update_requirement(int pm_qos_class, char *name, s32 new_value)
+int pm_qos_update_requirement(struct requirement_list *qos, s32 new_value)
{
unsigned long flags;
- struct requirement_list *node;
int pending_update = 0;
spin_lock_irqsave(&pm_qos_lock, flags);
- list_for_each_entry(node,
- &pm_qos_array[pm_qos_class]->requirements.list, list)
{
- if (strcmp(node->name, name) == 0) {
- if (new_value == PM_QOS_DEFAULT_VALUE)
- node->value =
-
pm_qos_array[pm_qos_class]->default_value;
- else
- node->value = new_value;
- pending_update = 1;
- break;
- }
- }
+ if (new_value == PM_QOS_DEFAULT_VALUE)
+ qos->value =
pm_qos_array[qos->pm_qos_class]->default_value;
+ else
+ qos->value = new_value;
+ pending_update = 1;
spin_unlock_irqrestore(&pm_qos_lock, flags); if (pending_update)
- update_target(pm_qos_class);
+ update_target(qos->pm_qos_class);
return 0;
}
@@ -277,32 +270,26 @@ EXPORT_SYMBOL_GPL(pm_qos_update_requirement);
/**
* pm_qos_remove_requirement - modifies an existing qos request
- * @pm_qos_class: identifies which list of qos request to us
+ * @qos: identifies the qos request to us
* @name: identifies the request
*
- * Will remove named qos request from pm_qos_class list of parameters and
+ * Will remove qos request from pm_qos_class list and
* recompute the current target value for the pm_qos_class. */
-void pm_qos_remove_requirement(int pm_qos_class, char *name) +void pm_qos_remove_requirement(struct requirement_list *qos)
{
unsigned long flags;
- struct requirement_list *node;
int pending_update = 0;
spin_lock_irqsave(&pm_qos_lock, flags);
- list_for_each_entry(node,
- &pm_qos_array[pm_qos_class]->requirements.list, list)
{
- if (strcmp(node->name, name) == 0) {
- kfree(node->name);
- list_del(&node->list);
- kfree(node);
- pending_update = 1;
- break;
- }
- }
+ list_del(&qos->list);
+ pending_update = 1;
spin_unlock_irqrestore(&pm_qos_lock, flags); if (pending_update)
- update_target(pm_qos_class);
+ update_target(qos->pm_qos_class);
+
+ kfree(qos->name);
+ kfree(qos);
} EXPORT_SYMBOL_GPL(pm_qos_remove_requirement);
@@ -345,37 +332,42 @@ int pm_qos_remove_notifier(int pm_qos_class, struct notifier_block *notifier)
EXPORT_SYMBOL_GPL(pm_qos_remove_notifier);
#define PID_NAME_LEN sizeof("process_1234567890")
-static char name[PID_NAME_LEN];
static int pm_qos_power_open(struct inode *inode, struct file *filp) {
int ret;
long pm_qos_class;
+ char name[PID_NAME_LEN];
+ struct requirement_list *qos;
lock_kernel();
pm_qos_class = find_pm_qos_object_by_minor(iminor(inode));
- if (pm_qos_class >= 0) {
- filp->private_data = (void *)pm_qos_class;
- sprintf(name, "process_%d", current->pid);
- ret = pm_qos_add_requirement(pm_qos_class, name,
- PM_QOS_DEFAULT_VALUE);
- if (ret >= 0) {
- unlock_kernel();
- return 0;
- }
+ if (pm_qos_class < 0) {
+ ret = -EPERM;
+ goto power_open_exit;
}
- unlock_kernel();
- return -EPERM;
+ sprintf(name, "process_%d", current->pid);
+ qos = pm_qos_add_requirement(pm_qos_class, name,PM_QOS_DEFAULT_VALUE);
+ if (IS_ERR(qos)) {
+ ret = PTR_ERR(qos);
+ goto power_open_exit;
+ }
+
+ filp->private_data = qos;
+ ret = 0;
+
+power_open_exit:
+ unlock_kernel();
+ return ret;
}
static int pm_qos_power_release(struct inode *inode, struct file
*filp)
{
- int pm_qos_class;
+ struct requirement_list *qos;
- pm_qos_class = (long)filp->private_data;
- sprintf(name, "process_%d", current->pid);
- pm_qos_remove_requirement(pm_qos_class, name);
+ qos = (struct requirement_list *)filp->private_data;
+ pm_qos_remove_requirement(qos);
return 0;
}
@@ -384,15 +376,15 @@ static ssize_t pm_qos_power_write(struct file *filp, const char __user *buf,
size_t count, loff_t *f_pos)
{
s32 value;
- int pm_qos_class;
+ struct requirement_list *qos;
- pm_qos_class = (long)filp->private_data;
if (count != sizeof(s32))
return -EINVAL;
if (copy_from_user(&value, buf, sizeof(s32)))
return -EFAULT;
- sprintf(name, "process_%d", current->pid);
- pm_qos_update_requirement(pm_qos_class, name, value);
+
+ qos = (struct requirement_list *)filp->private_data;
+ pm_qos_update_requirement(qos, value);
return sizeof(s32);
}
~Ai
Quoted 72 lines · Ai Li
Date: Tue, 17 Nov 2009 18:06:56 -0700
From: Ai Li <[email protected]>
To: [email protected]
Cc: [email protected]
Subject: Re: [linux-pm] adding handles topm_qos?Quoted 11 lines
Thanks, I'll look at it over the next few days.
--mgross
Quoted 1 line
Here is the patch that I use with the test code. There are
three new
Quoted 5 lines
functions:
void *pm_qos_get(int qos, char *name);
intpm_qos_put(void *handle);
int pm_qos_update_requirement_direct(void *handle, s32Quoted 7 lines
new_value);Quoted 2 lines
In the test, I wanted to keep the existing interface intact so
I
Quoted 1 line
could compare them at the same time. For the formal code
submission,
Quoted 1 line
a different approach may work better.
Here is an alternate way of adding handles to
pm_qos_paramsthat I was alluding to. This approach may be more preferable because it does not bloat the API and the handles become an integral part ofpm_qos_params. In my previous patch, handles are kind of bolted ontopm_qos_paramsand it needs separate calls (pm_qos_get,pm_qos_put) to acquire and release the handle. In this patch,pm_qos_add_requirementandpm_qos_remove_requirementautomatically take care of the handles.linux/pm_qos_params.h renamed from linux/pm_qos_params.h.orig +5
-3@@ -14,9 +14,11 @@#define PM_QOS_NUM_CLASSES 4#define PM_QOS_DEFAULT_VALUE -1 -int pm_qos_add_requirement (int qos,char *name,s32 value); -int pm_qos_update_requirement (int qos,char *name,s32 new_value); -void pm_qos_remove_requirement (int qos,char *name); +struct requirement_list ; + +struct requirement_list *pm_qos_add_requirement (int qos,char *name,s32 value); +int pm_qos_update_requirement (struct requirement_list *qos,s32 new_value); +void pm_qos_remove_requirement (struct requirement_list *qos);int
pm_qos_requirement(int qos);pm_qos_params.c renamed from pm_qos_params.c.orig +1
-0@@ -49,6 +49,7 @@*/ struct requirement_list {struct list_head list; +int pm_qos_class;union {s32 value;s32 usec;@@ -207,13 +208,16 @@ EXPORT_SYMBOL_GPL(pm_qos_requirement);* performance characteristics. It recomputes the aggregate QoS expectations
* for thepm_qos_classof parameters.
*/pm_qos_params.c renamed from pm_qos_params.c.orig +1
-1-int pm_qos_add_requirement (int pm_qos_class,char *name,s32 value) +struct requirement_list *pm_qos_add_requirement(int pm_qos_class,
Quoted 1 line
char *name,
looks like your email client buggered the your patch.
mgross@mgross-laptop:~/marks/Linux/linux-2.6$ patch -p0 < /home/mgross/pm_qos_handle.patch patching file include/linux/pm_qos_params.h patch: **** malformed patch at line 43: s32 value);
Please resend
--mgross
Quoted 215 lines
pm_qos_params.c renamed from pm_qos_params.c.orig +3
-0+ s32 value) {struct requirement_list *dep;unsigned long flags; dep =kzalloc (sizeof (struct requirement_list ),GFP_KERNEL );if (dep) { + dep->pm_qos_class = pm_qos_class; +if (value ==PM_QOS_DEFAULT_VALUE ) dep->value =pm_qos_array[pm_qos_class]->default_value;
pm_qos_params.c renamed from pm_qos_params.c.orig +3
-5else @@ -228,48 +232,37 @@ int pm_qos_add_requirement(int pm_qos_class, char *name, s32 value) spin_unlock_irqrestore(&pm_qos_lock, flags);update_target (pm_qos_class); -return 0 ; +return dep; } cleanup:kfree (dep); -return -ENOMEM ; +return ERR_PTR (-ENOMEM ); }EXPORT_SYMBOL_GPL (pm_qos_add_requirement);/** * pm_qos_update_requirement - modifies an existing qos request -* @pm_qos_class: identifies which list of qos request to us -* @name: identifies the request +* @qos: identifies the qos request to us * @value: defines the qos request * -* Updates an existing qos requirement for the pm_qos_class of parameters along
+ * Updates an existing qos requirement along
* with updating the targetpm_qos_classvalue.
- *
- * If the named request isn't in the list then no change is made. */
-intpm_qos_update_requirement(int pm_qos_class, char *name, s32 new_value)
+intpm_qos_update_requirement(struct requirement_list *qos, s32 new_value)
{pm_qos_params.c renamed from pm_qos_params.c.orig +0
-3unsigned long flags; -struct requirement_list *node;int pending_update =0 ;spin_lock_irqsave (&pm_qos_lock, flags); -list_for_each_entry (node, - &pm_qos_array[pm_qos_class]->requirements.list, list){
pm_qos_params.c renamed from pm_qos_params.c.orig +0
-4-if (strcmp (node->name, name) ==0 ) { -if (new_value ==PM_QOS_DEFAULT_VALUE ) - node->value = -pm_qos_array[pm_qos_class]->default_value;
pm_qos_params.c renamed from pm_qos_params.c.orig +2
-6-else - node->value = new_value; - pending_update =1 ; -break ; - } - } +if (new_value ==PM_QOS_DEFAULT_VALUE ) + qos->value =pm_qos_array[qos->pm_qos_class]->default_value;
pm_qos_params.c renamed from pm_qos_params.c.orig +5
-3+else + qos->value = new_value; + pending_update =1 ;spin_unlock_irqrestore (&pm_qos_lock, flags);if (pending_update) -update_target (pm_qos_class); +update_target (qos->pm_qos_class);return 0 ; }@@ -277,32 +270,26 @@ EXPORT_SYMBOL_GPL(pm_qos_update_requirement);/** * pm_qos_remove_requirement - modifies an existing qos request -* @pm_qos_class: identifies which list of qos request to us +* @qos: identifies the qos request to us * @name: identifies the request * -* Will remove named qos request from pm_qos_class list of parameters and
+ * Will remove qos request frompm_qos_classlist and
* recompute the current target value for thepm_qos_class. */pm_qos_params.c renamed from pm_qos_params.c.orig +1
-4-void pm_qos_remove_requirement (int pm_qos_class,char *name) +void pm_qos_remove_requirement (struct requirement_list *qos) {unsigned long flags; -struct requirement_list *node;int pending_update =0 ;spin_lock_irqsave (&pm_qos_lock, flags); -list_for_each_entry (node, - &pm_qos_array[pm_qos_class]->requirements.list, list){
pm_qos_params.c renamed from pm_qos_params.c.orig +27
-25-if (strcmp (node->name, name) ==0 ) { -kfree (node->name); -list_del (&node->list); -kfree (node); - pending_update =1 ; -break ; - } - } +list_del (&qos->list); + pending_update =1 ;spin_unlock_irqrestore (&pm_qos_lock, flags);if (pending_update) -update_target (pm_qos_class); +update_target (qos->pm_qos_class); + +kfree (qos->name); +kfree (qos); }EXPORT_SYMBOL_GPL (pm_qos_remove_requirement);@@ -345,37 +332,42 @@ int pm_qos_remove_notifier(int pm_qos_class, struct notifier_block *notifier)EXPORT_SYMBOL_GPL (pm_qos_remove_notifier );#define PID_NAME_LEN sizeof("process_1234567890") -static char name[PID_NAME_LEN ];static int pm_qos_power_open (struct inode *inode,struct file *filp) {int ret;long pm_qos_class; +char name[PID_NAME_LEN ]; +struct requirement_list *qos;lock_kernel (); pm_qos_class =find_pm_qos_object_by_minor (iminor (inode)); -if (pm_qos_class >=0 ) { - filp->private_data = (void *)pm_qos_class; -sprintf (name,"process_%d" , current->pid); - ret =pm_qos_add_requirement (pm_qos_class, name, -PM_QOS_DEFAULT_VALUE ); -if (ret >=0 ) { -unlock_kernel (); -return 0 ; - } +if (pm_qos_class <0 ) { + ret = -EPERM ; + goto power_open_exit; } -unlock_kernel (); -return -EPERM ; +sprintf (name,"process_%d" , current->pid); + qos =pm_qos_add_requirement (pm_qos_class, name,PM_QOS_DEFAULT_VALUE ); +if (IS_ERR (qos)) { + ret =PTR_ERR (qos); + goto power_open_exit; + } + + filp->private_data = qos; + ret =0 ; + +power_open_exit: +unlock_kernel (); +return ret; }static int pm_qos_power_release (struct inode *inode,struct file *filp) { -int pm_qos_class; +struct requirement_list *qos; - pm_qos_class = (long )filp->private_data; -sprintf (name,"process_%d" , current->pid); -pm_qos_remove_requirement (pm_qos_class, name); + qos = (struct requirement_list *)filp->private_data; +pm_qos_remove_requirement (qos);pm_qos_params.c renamed from pm_qos_params.c.orig +4
-4return 0 ; }@@ -384,15 +376,15 @@ static ssize_t pm_qos_power_write(struct file *filp, const char __user *buf, size_t count, loff_t *f_pos) {s32 value; -int pm_qos_class; +struct requirement_list *qos; - pm_qos_class = (long )filp->private_data;if (count !=sizeof (s32))return -EINVAL ;if (copy_from_user (&value, buf,sizeof (s32)))return -EFAULT ; -sprintf (name,"process_%d" , current->pid); -pm_qos_update_requirement (pm_qos_class, name, value); + + qos = (struct requirement_list *)filp->private_data; +pm_qos_update_requirement (qos, value);return sizeof (s32);}
~Ai
_______________________________________________
linux-pm mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/linux-pm