os/pm: Add PM stop/status control and split code by seokhun-eom24 · Pull Request #6979 · Samsung/TizenRT

seokhun-eom24 · 2025-09-24T07:04:26Z

1. os/pm: Fix typo and convention in pm_initialize.c

Fix typo in pm_clock_initialize and pm_start
Fix function brace convention to enter after function name.

2. os/pm: add PMIOC_STOP ioctl to disable pm

Add PMIOC_STOP ioctl command and pm_stop function to support disable pm.

3. pm/pm_procfs: Add stopped status to /proc/power/state

Add stopped status to /proc/power/state.
It returns "stopped" when PM is stopped and PM states when PM is running.

[Example]

TASH>>cat state
* STOPPED
  NORMAL
  SLEEP

4. os/pm: Split source files for each functions

Extract pm_start(), pm_stop(), and pm_clock_initialize() functions from pm_initialize.c into separate source files.
Update Makefile to include new source files.

5. apps/examples/power: Add pm stop and status example

Add "power stop" command to stop power example.
Add "power status" command to get pm status.
Extract _pm_stop, _pm_start, _pm_status for ioctl communication.

6. os/pm: Add pm_lock function and update Makefile

Add new pm_lock function to lock the pm module.
This function automatically retries if interrupted by a signal (EINTR).

ewoodev · 2025-09-24T07:18:19Z

+		pid = task_create("stop_pm_test", 100, 1024, stop_pm_test, argv + 1);
+		if (pid < 0) {
+			printf("Fail to create stop_pm_test task(errno %d)\n", get_errno());
+			is_running = true;
+			return -1;
+		}


we don't need stop_pm_test task

it is requred only calling stop ioctl.

if(ioctl(fd, PMIOC_STOP, 0) < 0) { printf("Fail to pm start(errno %d)\n", get_errno()); close(fd); return -1; }

And how about make each testing function like below? (start / stop)

we can move log of start and stop to actually start and end entry point.

static int pm_start(void) { if(ioctl(fd, PMIOC_START, 0) < 0) ~~ } static int pm_stop(void) { if(ioctl(fd, PMIOC_STOP, 0) < 0) ~~ } .... if (strncmp(argv[1], "start", 6) == 0) { ... is_running = true; printf("######################### PM LONG TERM TEST START #########################\n"); pm_stop(); } else if (strncmp(argv[1], "stop", 5) == 0) { ... pm_stop(); printf("######################### PM LONG TERM TEST END #########################\n"); is_running = true; }

Thank you for suggest.
I updated to print start and stop log inside _pm_start and _pm_stop.

ewoodev · 2025-09-24T07:26:54Z

+void pm_stop(void)
+{
+	g_pmglobals.is_running = false;
+}


Looks good to me this stop fucntion.

pcs1265 · 2025-09-24T07:30:36Z

 #define PMIOC_METRICS            _PMIOC(0x0007)
 #define PMIOC_SUSPEND_COUNT      _PMIOC(0x0008)
 #define PMIOC_START              _PMIOC(0x0009)
+#define PMIOC_STOP               _PMIOC(0x0010)


0x10 is not a next value of 0x09.
Please change this value to 0x0A

Fixed. Thank you.

Taejun-Kwon · 2025-09-24T07:41:01Z

 {
 	enum pm_state_e pm_state;
+
+	readprint("PM %s\n\n", (g_pmglobals.is_running) ? "RUNNING" : "STOPPED");


procfs is not only for information.
sometimes application should parse data from procfs, so it shoul memcpy data to buffer.
Hence if you want reveal running state, then you should create additional directory, not here.

I updated this PR not to use procfs.
Thank you for review.

Taejun-Kwon · 2025-09-24T07:42:50Z

+
+void pm_stop(void)
+{
+	g_pmglobals.is_running = false;


pm_ioctl doesn't use semaphore, so I think this is not 'thread-safe'

I updated to protect g_pmglobals.is_running using semaphore.
Thank you.

sunghan-chang · 2025-09-24T09:16:27Z

+
+	fd = open(PM_DRVPATH, O_WRONLY);
+	if (fd < 0) {
+		printf("Fail to open pm start(errno %d)", get_errno());


It's open pm driver, not open pm start.

Fixed. Thank you.
And I modified other open error message too.

sunghan-chang · 2025-09-24T09:16:47Z

+	}
+
+	if(ioctl(fd, PMIOC_STOP, 0) < 0) {
+		printf("Fail to pm start(errno %d)\n", get_errno());


looks fail to stop pm, not start

sunghan-chang · 2025-09-24T09:17:54Z

+		return -1;
+	}
+
+	if(ioctl(fd, PMIOC_STOP, 0) < 0) {


Fixed this line and also in start_pm_test

sunghan-chang · 2025-09-24T09:21:40Z


 		is_running = false;

+		pid = task_create("stop_pm_test", 100, 1024, stop_pm_test, argv + 1);


Could you let me know why you give argv + 1?
In help message, the stop command does not have any following argument.

And the task creation for start also give it, but is it right with +1? not +2?

Yes, stop_pm_test doesn't need any argument.
So I fixed argv + 1 to NULL.

argv + 1 is the correct value to use because argv is a pointer to the first element of the command argument.

When the command line is something like:

power start -t

the argv array is initialized as:

argv[0] = "power" argv[1] = "start" argv[2] = "-t" argv[3] = NULL

To pass only the part after the program name(i.e.,{"start", "-t"}), need to give the pointer that starts at argv[1]. So argv + 1 is right.

@seokhun-eom24 You already check the argv[1] with start.

if (strncmp(argv[1], "start", 6) == 0) { if (is_running) { printf("power test is already running\n"); return 0; } is_running = true; pid = task_create("start_pm_test", 100, 1024, start_pm_test, argv + 1);

The task is created for start. Why do you need to give start argument to the task?

Right. We don't need to pass argv + 1, it's start.
I updated to pass argv + 2.
Thank you.

sunghan-chang · 2025-09-24T09:25:05Z

+		pid = task_create("stop_pm_test", 100, 1024, stop_pm_test, argv + 1);
+		if (pid < 0) {
+			printf("Fail to create stop_pm_test task(errno %d)\n", get_errno());
+			is_running = true;


If you set the variable to false after this creation, what happen?
I mean
do you need to set it to false always and then revert it with failure case?

In start_pm_test, is_running should be set first and start task,
but as you said in stop_pm_test reset is_running can handled after create task.
So I modified the flow to reset after create task.

sunghan-chang · 2025-09-24T09:29:56Z

+ *
+ ****************************************************************************/
+
+void pm_stop(void)


This file is pm_initialize. Let's add this with new file.
same for pm_start

I separated pm_start, pm_stop, and pm_clock_initialize from pm_initialize.c with new files.

sunghan-chang · 2025-09-25T05:17:42Z

 			return 0;
 		}

+		pid = task_create("stop_pm_test", 100, 1024, stop_pm_test, NULL);


Is there any reason why we should do start and stop test in separate tasks?

We don't need to separate tasks to stop.
I updated code.

ewoodev · 2025-09-25T05:22:17Z

+		return -1;
+	}
+
+	if (ioctl(fd, PMIOC_STOP, 0) < 0) {


This code can be called first than "start_pm_test" task termination

like

call resume(10) ######################### PM LONG TERM TEST END ######################### call suspend(10)

I updated to print log after _pm_stop.

gidori98 · 2025-10-28T01:04:25Z

LGTM

ewoodev · 2026-02-06T07:44:23Z

+int pm_getstatus(void)
+{
+	int status;
+
+	pm_lock();
+	status = g_pmglobals.is_running ? PM_STATUS_RUNNING : PM_STATUS_STOPPED;
+	pm_unlock();
+
+	return status;
+}


we don't need to provide status using api.
how about add it to procfs?

I added a new procfs /proc/power/running.
The reason for creating /power/running separately from /power/status is based on Taejun-Kwon's comment.
Procfs is not only for information but also allows applications to fetch data via memcpy, which is why the new procfs was added.

Modified to use "/proc/power/state" together as we discussed.

Fix typo in `pm_clock_initialize` and `Pm_start`. Fix function brace convention to enter after function name. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

Add `PMIOC_STOP` ioctl command and `pm_stop` function to support disable pm. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

Add stopped status to /proc/power/state. It returns "stopped" when PM is stopped and PM states when PM is running. [Example] ``` TASH>>cat state * STOPPED NORMAL SLEEP ``` Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

Extract `pm_start()`, `pm_stop()`, and `pm_clock_initialize()` functions from `pm_initialize.c` into separate source files. Update `Makefile` to include new source files. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

Add "power stop" command to stop power example. Add "power status" command to get pm status. Extract `_pm_stop`, `_pm_start`, `_pm_status` for ioctl communication. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

Add new `pm_lock` function to lock the pm module. This function automatically retries if interrupted by a signal (EINTR). Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

seokhun-eom24 force-pushed the 250924-add-pm-stop branch 2 times, most recently from ebae5c7 to a7ae0e3 Compare September 24, 2025 07:07

ewoodev reviewed Sep 24, 2025

View reviewed changes

pcs1265 reviewed Sep 24, 2025

View reviewed changes

Taejun-Kwon reviewed Sep 24, 2025

View reviewed changes

sunghan-chang reviewed Sep 24, 2025

View reviewed changes

seokhun-eom24 force-pushed the 250924-add-pm-stop branch from a7ae0e3 to 40a7baa Compare September 25, 2025 02:22

sunghan-chang reviewed Sep 25, 2025

View reviewed changes

ewoodev reviewed Sep 25, 2025

View reviewed changes

seokhun-eom24 force-pushed the 250924-add-pm-stop branch from 40a7baa to 4acfa9c Compare January 28, 2026 09:41

seokhun-eom24 changed the title ~~os/pm: Add PMIOC_STOP to stop pm and show state in pm_procfs~~ os/pm: Add PM stop/status control and split code Jan 28, 2026

seokhun-eom24 force-pushed the 250924-add-pm-stop branch 4 times, most recently from a947978 to 0eba68b Compare February 2, 2026 04:26

ewoodev reviewed Feb 6, 2026

View reviewed changes

seokhun-eom24 force-pushed the 250924-add-pm-stop branch 3 times, most recently from 83054cb to 39edc21 Compare February 13, 2026 05:49

seokhun-eom24 added 6 commits February 13, 2026 14:51

os/pm: Fix typo and convention in pm_initialize.c

be379fb

Fix typo in `pm_clock_initialize` and `Pm_start`. Fix function brace convention to enter after function name. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

os/pm: add PMIOC_STOP ioctl to disable pm

db42e76

Add `PMIOC_STOP` ioctl command and `pm_stop` function to support disable pm. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

pm/pm_procfs: Add stopped status to /proc/power/state

bdedea0

Add stopped status to /proc/power/state. It returns "stopped" when PM is stopped and PM states when PM is running. [Example] ``` TASH>>cat state * STOPPED NORMAL SLEEP ``` Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

os/pm: Split source files for each functions

0f3faf9

Extract `pm_start()`, `pm_stop()`, and `pm_clock_initialize()` functions from `pm_initialize.c` into separate source files. Update `Makefile` to include new source files. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

apps/examples/power: Add pm stop and status example

632b7f4

Add "power stop" command to stop power example. Add "power status" command to get pm status. Extract `_pm_stop`, `_pm_start`, `_pm_status` for ioctl communication. Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

os/pm: Add pm_lock function and update Makefile

fe98a9d

Add new `pm_lock` function to lock the pm module. This function automatically retries if interrupted by a signal (EINTR). Signed-off-by: seokhun-eom <seokhun.eom@samsung.com>

seokhun-eom24 force-pushed the 250924-add-pm-stop branch from 39edc21 to fe98a9d Compare February 13, 2026 05:52


		is_running = false;

		pid = task_create("stop_pm_test", 100, 1024, stop_pm_test, argv + 1);

Conversation

seokhun-eom24 commented Sep 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

1. os/pm: Fix typo and convention in pm_initialize.c

2. os/pm: add PMIOC_STOP ioctl to disable pm

3. pm/pm_procfs: Add stopped status to /proc/power/state

4. os/pm: Split source files for each functions

5. apps/examples/power: Add pm stop and status example

6. os/pm: Add pm_lock function and update Makefile

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

pcs1265 Sep 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

seokhun-eom24 Jan 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gidori98 commented Oct 28, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

seokhun-eom24 commented Sep 24, 2025 •

edited

Loading

pcs1265 Sep 24, 2025 •

edited

Loading

seokhun-eom24 Jan 28, 2026 •

edited

Loading