[PATCH v2] PM / Domains: enable domain idle state accounting

[PATCH v2] PM / Domains: enable domain idle state accounting

[Lina Iyer]

To enable better debug of PM domains, let's keep a track of the success
and rejections in entering each domain idle state.

This statistics is exported in debugfs when reading the idle_states
node, associated with each PM domain.

Signed-off-by: Lina Iyer <ilina@codeaurora.org>
---
Changes in v2:
	- Renamed 'failed' to 'rejected'

This patch depends-on: https://lkml.org/lkml/2020/9/24/465
---
 drivers/base/power/domain.c | 7 +++++--
 include/linux/pm_domain.h   | 2 ++
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
index f001ac6326fb..dbe89454f594 100644
--- a/drivers/base/power/domain.c
+++ b/drivers/base/power/domain.c
@@ -564,6 +564,7 @@ static int genpd_power_off(struct generic_pm_domain *genpd, bool one_dev_on,
 
 	genpd->status = GENPD_STATE_OFF;
 	genpd_update_accounting(genpd);
+	genpd->states[genpd->state_idx].usage++;
 
 	list_for_each_entry(link, &genpd->child_links, child_node) {
 		genpd_sd_counter_dec(link->parent);
@@ -574,6 +575,7 @@ static int genpd_power_off(struct generic_pm_domain *genpd, bool one_dev_on,
 
 	return 0;
 busy:
+	genpd->states[genpd->state_idx].rejected++;
 	if (nr_calls)
 		__raw_notifier_call_chain(&genpd->power_notifiers,
 					  GENPD_STATE_ON, NULL,
@@ -3053,7 +3055,7 @@ static int idle_states_show(struct seq_file *s, void *data)
 	if (ret)
 		return -ERESTARTSYS;
 
-	seq_puts(s, "State          Time Spent(ms)\n");
+	seq_puts(s, "State          Time Spent(ms) Usage          Rejected\n");
 
 	for (i = 0; i < genpd->state_count; i++) {
 		ktime_t delta = 0;
@@ -3065,7 +3067,8 @@ static int idle_states_show(struct seq_file *s, void *data)
 
 		msecs = ktime_to_ms(
 			ktime_add(genpd->states[i].idle_time, delta));
-		seq_printf(s, "S%-13i %lld\n", i, msecs);
+		seq_printf(s, "S%-13i %-14lld %-14llu %llu\n", i, msecs,
+			      genpd->states[i].usage, genpd->states[i].rejected);
 	}
 
 	genpd_unlock(genpd);
diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h
index 3b2b561ce846..239647f2d27f 100644
--- a/include/linux/pm_domain.h
+++ b/include/linux/pm_domain.h
@@ -82,6 +82,8 @@ struct genpd_power_state {
 	s64 power_off_latency_ns;
 	s64 power_on_latency_ns;
 	s64 residency_ns;
+	u64 usage;
+	u64 rejected;
 	struct fwnode_handle *fwnode;
 	ktime_t idle_time;
 	void *data;
-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

Re: [PATCH v2] PM / Domains: enable domain idle state accounting

[Rafael J. Wysocki]

On Sat, Oct 3, 2020 at 5:56 PM Lina Iyer <ilina@codeaurora.org> wrote:
>
> To enable better debug of PM domains, let's keep a track of the success
> and rejections in entering each domain idle state.
>
> This statistics is exported in debugfs when reading the idle_states
> node, associated with each PM domain.
>
> Signed-off-by: Lina Iyer <ilina@codeaurora.org>
> ---
> Changes in v2:
>         - Renamed 'failed' to 'rejected'
>
> This patch depends-on: https://lkml.org/lkml/2020/9/24/465
> ---
>  drivers/base/power/domain.c | 7 +++++--
>  include/linux/pm_domain.h   | 2 ++
>  2 files changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
> index f001ac6326fb..dbe89454f594 100644
> --- a/drivers/base/power/domain.c
> +++ b/drivers/base/power/domain.c
> @@ -564,6 +564,7 @@ static int genpd_power_off(struct generic_pm_domain *genpd, bool one_dev_on,
>
>         genpd->status = GENPD_STATE_OFF;
>         genpd_update_accounting(genpd);
> +       genpd->states[genpd->state_idx].usage++;

Why not to do this in genpd_update_accounting()?

>
>         list_for_each_entry(link, &genpd->child_links, child_node) {
>                 genpd_sd_counter_dec(link->parent);
> @@ -574,6 +575,7 @@ static int genpd_power_off(struct generic_pm_domain *genpd, bool one_dev_on,
>
>         return 0;
>  busy:
> +       genpd->states[genpd->state_idx].rejected++;
>         if (nr_calls)
>                 __raw_notifier_call_chain(&genpd->power_notifiers,
>                                           GENPD_STATE_ON, NULL,

This doesn't apply to the current code, please rebase.

> @@ -3053,7 +3055,7 @@ static int idle_states_show(struct seq_file *s, void *data)
>         if (ret)
>                 return -ERESTARTSYS;
>
> -       seq_puts(s, "State          Time Spent(ms)\n");
> +       seq_puts(s, "State          Time Spent(ms) Usage          Rejected\n");
>
>         for (i = 0; i < genpd->state_count; i++) {
>                 ktime_t delta = 0;
> @@ -3065,7 +3067,8 @@ static int idle_states_show(struct seq_file *s, void *data)
>
>                 msecs = ktime_to_ms(
>                         ktime_add(genpd->states[i].idle_time, delta));
> -               seq_printf(s, "S%-13i %lld\n", i, msecs);
> +               seq_printf(s, "S%-13i %-14lld %-14llu %llu\n", i, msecs,
> +                             genpd->states[i].usage, genpd->states[i].rejected);
>         }
>
>         genpd_unlock(genpd);
> diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h
> index 3b2b561ce846..239647f2d27f 100644
> --- a/include/linux/pm_domain.h
> +++ b/include/linux/pm_domain.h
> @@ -82,6 +82,8 @@ struct genpd_power_state {
>         s64 power_off_latency_ns;
>         s64 power_on_latency_ns;
>         s64 residency_ns;
> +       u64 usage;
> +       u64 rejected;
>         struct fwnode_handle *fwnode;
>         ktime_t idle_time;
>         void *data;
> --
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
> a Linux Foundation Collaborative Project
>

Re: [PATCH v2] PM / Domains: enable domain idle state accounting

[Lina Iyer]

On Mon, Oct 05 2020 at 07:27 -0600, Rafael J. Wysocki wrote:
>On Sat, Oct 3, 2020 at 5:56 PM Lina Iyer <ilina@codeaurora.org> wrote:
>>
>> To enable better debug of PM domains, let's keep a track of the success
>> and rejections in entering each domain idle state.
>>
>> This statistics is exported in debugfs when reading the idle_states
>> node, associated with each PM domain.
>>
>> Signed-off-by: Lina Iyer <ilina@codeaurora.org>
>> ---
>> Changes in v2:
>>         - Renamed 'failed' to 'rejected'
>>
>> This patch depends-on: https://lkml.org/lkml/2020/9/24/465
>> ---
>>  drivers/base/power/domain.c | 7 +++++--
>>  include/linux/pm_domain.h   | 2 ++
>>  2 files changed, 7 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
>> index f001ac6326fb..dbe89454f594 100644
>> --- a/drivers/base/power/domain.c
>> +++ b/drivers/base/power/domain.c
>> @@ -564,6 +564,7 @@ static int genpd_power_off(struct generic_pm_domain *genpd, bool one_dev_on,
>>
>>         genpd->status = GENPD_STATE_OFF;
>>         genpd_update_accounting(genpd);
>> +       genpd->states[genpd->state_idx].usage++;
>
>Why not to do this in genpd_update_accounting()?
>
That function is clubbed with debugfs and does heavy tracking using
timers. This accounting is fairly basic and still quite useful for
debugging.

>>
>>         list_for_each_entry(link, &genpd->child_links, child_node) {
>>                 genpd_sd_counter_dec(link->parent);
>> @@ -574,6 +575,7 @@ static int genpd_power_off(struct generic_pm_domain *genpd, bool one_dev_on,
>>
>>         return 0;
>>  busy:
>> +       genpd->states[genpd->state_idx].rejected++;
>>         if (nr_calls)
>>                 __raw_notifier_call_chain(&genpd->power_notifiers,
>>                                           GENPD_STATE_ON, NULL,
>
>This doesn't apply to the current code, please rebase.
>
I believe it applies cleanly on top of
https://lkml.org/lkml/2020/9/24/465, which I believe you are applied to
your tree. Let me rebase and re-post. Sorry about that.

Thanks,
Lina

>> @@ -3053,7 +3055,7 @@ static int idle_states_show(struct seq_file *s, void *data)
>>         if (ret)
>>                 return -ERESTARTSYS;
>>
>> -       seq_puts(s, "State          Time Spent(ms)\n");
>> +       seq_puts(s, "State          Time Spent(ms) Usage          Rejected\n");
>>
>>         for (i = 0; i < genpd->state_count; i++) {
>>                 ktime_t delta = 0;
>> @@ -3065,7 +3067,8 @@ static int idle_states_show(struct seq_file *s, void *data)
>>
>>                 msecs = ktime_to_ms(
>>                         ktime_add(genpd->states[i].idle_time, delta));
>> -               seq_printf(s, "S%-13i %lld\n", i, msecs);
>> +               seq_printf(s, "S%-13i %-14lld %-14llu %llu\n", i, msecs,
>> +                             genpd->states[i].usage, genpd->states[i].rejected);
>>         }
>>
>>         genpd_unlock(genpd);
>> diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h
>> index 3b2b561ce846..239647f2d27f 100644
>> --- a/include/linux/pm_domain.h
>> +++ b/include/linux/pm_domain.h
>> @@ -82,6 +82,8 @@ struct genpd_power_state {
>>         s64 power_off_latency_ns;
>>         s64 power_on_latency_ns;
>>         s64 residency_ns;
>> +       u64 usage;
>> +       u64 rejected;
>>         struct fwnode_handle *fwnode;
>>         ktime_t idle_time;
>>         void *data;
>> --
>> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
>> a Linux Foundation Collaborative Project
>>