An important part of validating that s0ix worked properly is to check how
much of a cycle was spent in a hardware sleep state.
The reporting of hardware sleep is a mix of kernel messages and sysfs
files that vary from vendor to vendor. Collecting this information
requires extra information on the kernel command line or fetching from
debugfs.
To make this information more readily accessible introduce a new file in
suspend_stats that drivers can report into during their resume routine.
Userspace can fetch this information and compare it against the duration
of the cycle to allow determining residency percentages and flagging
problems.
Mario Limonciello (4):
PM: Add sysfs files to represent time spent in hardware sleep state
platform/x86/amd: pmc: Report duration of time in hw sleep state
platform/x86/intel/pmc: core: Always capture counters on suspend
platform/x86/intel/pmc: core: Report duration of time in HW sleep
state
Documentation/ABI/testing/sysfs-power | 29 +++++++++++++
drivers/platform/x86/amd/pmc.c | 6 +--
drivers/platform/x86/intel/pmc/core.c | 17 ++++----
drivers/platform/x86/intel/pmc/core.h | 4 +-
include/linux/suspend.h | 8 ++++
kernel/power/main.c | 59 +++++++++++++++++++++------
6 files changed, 98 insertions(+), 25 deletions(-)
base-commit: 09a9639e56c01c7a00d6c0ca63f4c7c41abe075d
--
2.34.1
Currently counters are only captured during suspend when the
warn_on_s0ix_failures module parameter is set.
In order to relay this counter information to the kernel reporting
infrastructure adjust it so that the counters are always captured.
warn_on_s0ix_failures will be utilized solely for messaging by
the driver instead.
Reviewed-by: David E. Box <[email protected]>
Signed-off-by: Mario Limonciello <[email protected]>
---
v5->v6:
* Pick up tag
v4->v5:
* Squash patches together
* Add extra pm_suspend_via_firmware() check for resume routine too
---
drivers/platform/x86/intel/pmc/core.c | 13 +++++--------
drivers/platform/x86/intel/pmc/core.h | 2 --
2 files changed, 5 insertions(+), 10 deletions(-)
diff --git a/drivers/platform/x86/intel/pmc/core.c b/drivers/platform/x86/intel/pmc/core.c
index b9591969e0fa..925c5d676a43 100644
--- a/drivers/platform/x86/intel/pmc/core.c
+++ b/drivers/platform/x86/intel/pmc/core.c
@@ -1179,12 +1179,6 @@ static __maybe_unused int pmc_core_suspend(struct device *dev)
{
struct pmc_dev *pmcdev = dev_get_drvdata(dev);
- pmcdev->check_counters = false;
-
- /* No warnings on S0ix failures */
- if (!warn_on_s0ix_failures)
- return 0;
-
/* Check if the syspend will actually use S0ix */
if (pm_suspend_via_firmware())
return 0;
@@ -1197,7 +1191,6 @@ static __maybe_unused int pmc_core_suspend(struct device *dev)
if (pmc_core_dev_state_get(pmcdev, &pmcdev->s0ix_counter))
return -EIO;
- pmcdev->check_counters = true;
return 0;
}
@@ -1233,12 +1226,16 @@ static __maybe_unused int pmc_core_resume(struct device *dev)
const struct pmc_bit_map **maps = pmcdev->map->lpm_sts;
int offset = pmcdev->map->lpm_status_offset;
- if (!pmcdev->check_counters)
+ /* Check if the syspend used S0ix */
+ if (pm_suspend_via_firmware())
return 0;
if (!pmc_core_is_s0ix_failed(pmcdev))
return 0;
+ if (!warn_on_s0ix_failures)
+ return 0;
+
if (pmc_core_is_pc10_failed(pmcdev)) {
/* S0ix failed because of PC10 entry failure */
dev_info(dev, "CPU did not enter PC10!!! (PC10 cnt=0x%llx)\n",
diff --git a/drivers/platform/x86/intel/pmc/core.h b/drivers/platform/x86/intel/pmc/core.h
index 810204d758ab..51d73efceaf3 100644
--- a/drivers/platform/x86/intel/pmc/core.h
+++ b/drivers/platform/x86/intel/pmc/core.h
@@ -319,7 +319,6 @@ struct pmc_reg_map {
* @pmc_xram_read_bit: flag to indicate whether PMC XRAM shadow registers
* used to read MPHY PG and PLL status are available
* @mutex_lock: mutex to complete one transcation
- * @check_counters: On resume, check if counters are getting incremented
* @pc10_counter: PC10 residency counter
* @s0ix_counter: S0ix residency (step adjusted)
* @num_lpm_modes: Count of enabled modes
@@ -338,7 +337,6 @@ struct pmc_dev {
int pmc_xram_read_bit;
struct mutex lock; /* generic mutex lock for PMC Core */
- bool check_counters; /* Check for counter increments on resume */
u64 pc10_counter;
u64 s0ix_counter;
int num_lpm_modes;
--
2.34.1
Userspace can't easily discover how much of a sleep cycle was spent in a
hardware sleep state without using kernel tracing and vendor specific sysfs
or debugfs files.
To make this information more discoverable, introduce 3 new sysfs files:
1) The time spent in a hw sleep state for last cycle.
2) The time spent in a hw sleep state since the kernel booted
3) The maximum time that the hardware can report for a sleep cycle.
All of these files will be present only if the system supports s2idle.
Signed-off-by: Mario Limonciello <[email protected]>
---
v7-v8:
* Add max_hw_sleep sysfs file
* Add function for drivers to set max_hw_sleep
* Modify macro to display attributes
---
Documentation/ABI/testing/sysfs-power | 29 +++++++++++++
include/linux/suspend.h | 8 ++++
kernel/power/main.c | 59 +++++++++++++++++++++------
3 files changed, 84 insertions(+), 12 deletions(-)
diff --git a/Documentation/ABI/testing/sysfs-power b/Documentation/ABI/testing/sysfs-power
index f99d433ff311..a3942b1036e2 100644
--- a/Documentation/ABI/testing/sysfs-power
+++ b/Documentation/ABI/testing/sysfs-power
@@ -413,6 +413,35 @@ Description:
The /sys/power/suspend_stats/last_failed_step file contains
the last failed step in the suspend/resume path.
+What: /sys/power/suspend_stats/last_hw_sleep
+Date: June 2023
+Contact: Mario Limonciello <[email protected]>
+Description:
+ The /sys/power/suspend_stats/last_hw_sleep file
+ contains the duration of time spent in a hardware sleep
+ state in the most recent system suspend-resume cycle.
+ This number is measured in microseconds.
+
+What: /sys/power/suspend_stats/total_hw_sleep
+Date: June 2023
+Contact: Mario Limonciello <[email protected]>
+Description:
+ The /sys/power/suspend_stats/total_hw_sleep file
+ contains the aggregate of time spent in a hardware sleep
+ state since the kernel was booted. This number
+ is measured in microseconds.
+
+What: /sys/power/suspend_stats/max_hw_sleep
+Date: June 2023
+Contact: Mario Limonciello <[email protected]>
+Description:
+ The /sys/power/suspend_stats/max_hw_sleep file
+ contains the maximum amount of time that the hardware can
+ report for time spent in a hardware sleep state. When sleep
+ cycles are longer than this time, the values for
+ 'total_hw_sleep' and 'last_hw_sleep' may not be accurate.
+ This number is measured in microseconds.
+
What: /sys/power/sync_on_suspend
Date: October 2019
Contact: Jonas Meurer <[email protected]>
diff --git a/include/linux/suspend.h b/include/linux/suspend.h
index cfe19a028918..d0d4598a7b3f 100644
--- a/include/linux/suspend.h
+++ b/include/linux/suspend.h
@@ -68,6 +68,9 @@ struct suspend_stats {
int last_failed_errno;
int errno[REC_FAILED_NUM];
int last_failed_step;
+ u64 last_hw_sleep;
+ u64 total_hw_sleep;
+ u64 max_hw_sleep;
enum suspend_stat_step failed_steps[REC_FAILED_NUM];
};
@@ -489,6 +492,8 @@ void restore_processor_state(void);
extern int register_pm_notifier(struct notifier_block *nb);
extern int unregister_pm_notifier(struct notifier_block *nb);
extern void ksys_sync_helper(void);
+extern void pm_report_hw_sleep_time(u64 t);
+extern void pm_report_max_hw_sleep(u64 t);
#define pm_notifier(fn, pri) { \
static struct notifier_block fn##_nb = \
@@ -526,6 +531,9 @@ static inline int unregister_pm_notifier(struct notifier_block *nb)
return 0;
}
+static inline void pm_report_hw_sleep_time(u64 t) {};
+static inline void pm_report_max_hw_sleep(u64 t) {};
+
static inline void ksys_sync_helper(void) {}
#define pm_notifier(fn, pri) do { (void)(fn); } while (0)
diff --git a/kernel/power/main.c b/kernel/power/main.c
index 31ec4a9b9d70..3113ec2f1db4 100644
--- a/kernel/power/main.c
+++ b/kernel/power/main.c
@@ -6,6 +6,7 @@
* Copyright (c) 2003 Open Source Development Lab
*/
+#include <linux/acpi.h>
#include <linux/export.h>
#include <linux/kobject.h>
#include <linux/string.h>
@@ -83,6 +84,19 @@ int unregister_pm_notifier(struct notifier_block *nb)
}
EXPORT_SYMBOL_GPL(unregister_pm_notifier);
+void pm_report_hw_sleep_time(u64 t)
+{
+ suspend_stats.last_hw_sleep = t;
+ suspend_stats.total_hw_sleep += t;
+}
+EXPORT_SYMBOL_GPL(pm_report_hw_sleep_time);
+
+void pm_report_max_hw_sleep(u64 t)
+{
+ suspend_stats.max_hw_sleep = t;
+}
+EXPORT_SYMBOL_GPL(pm_report_max_hw_sleep);
+
int pm_notifier_call_chain_robust(unsigned long val_up, unsigned long val_down)
{
int ret;
@@ -314,24 +328,27 @@ static char *suspend_step_name(enum suspend_stat_step step)
}
}
-#define suspend_attr(_name) \
+#define suspend_attr(_name, format_str) \
static ssize_t _name##_show(struct kobject *kobj, \
struct kobj_attribute *attr, char *buf) \
{ \
- return sprintf(buf, "%d\n", suspend_stats._name); \
+ return sprintf(buf, format_str, suspend_stats._name); \
} \
static struct kobj_attribute _name = __ATTR_RO(_name)
-suspend_attr(success);
-suspend_attr(fail);
-suspend_attr(failed_freeze);
-suspend_attr(failed_prepare);
-suspend_attr(failed_suspend);
-suspend_attr(failed_suspend_late);
-suspend_attr(failed_suspend_noirq);
-suspend_attr(failed_resume);
-suspend_attr(failed_resume_early);
-suspend_attr(failed_resume_noirq);
+suspend_attr(success, "%d\n");
+suspend_attr(fail, "%d\n");
+suspend_attr(failed_freeze, "%d\n");
+suspend_attr(failed_prepare, "%d\n");
+suspend_attr(failed_suspend, "%d\n");
+suspend_attr(failed_suspend_late, "%d\n");
+suspend_attr(failed_suspend_noirq, "%d\n");
+suspend_attr(failed_resume, "%d\n");
+suspend_attr(failed_resume_early, "%d\n");
+suspend_attr(failed_resume_noirq, "%d\n");
+suspend_attr(last_hw_sleep, "%llu\n");
+suspend_attr(total_hw_sleep, "%llu\n");
+suspend_attr(max_hw_sleep, "%llu\n");
static ssize_t last_failed_dev_show(struct kobject *kobj,
struct kobj_attribute *attr, char *buf)
@@ -391,12 +408,30 @@ static struct attribute *suspend_attrs[] = {
&last_failed_dev.attr,
&last_failed_errno.attr,
&last_failed_step.attr,
+ &last_hw_sleep.attr,
+ &total_hw_sleep.attr,
+ &max_hw_sleep.attr,
NULL,
};
+static umode_t suspend_attr_is_visible(struct kobject *kobj, struct attribute *attr, int idx)
+{
+ if (attr != &last_hw_sleep.attr &&
+ attr != &total_hw_sleep.attr &&
+ attr != &max_hw_sleep.attr)
+ return 0444;
+
+#ifdef CONFIG_ACPI
+ if (acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)
+ return 0444;
+#endif
+ return 0;
+}
+
static const struct attribute_group suspend_attr_group = {
.name = "suspend_stats",
.attrs = suspend_attrs,
+ .is_visible = suspend_attr_is_visible,
};
#ifdef CONFIG_DEBUG_FS
--
2.34.1
intel_pmc_core displays a warning when the module parameter
`warn_on_s0ix_failures` is set and a suspend didn't get to a HW sleep
state.
Report this to the standard kernel reporting infrastructure so that
userspace software can query after the suspend cycle is done.
Suggested-by: Ilpo Järvinen <[email protected]>
Reviewed-by: David E. Box <[email protected]>
Signed-off-by: Mario Limonciello <[email protected]>
---
v8->v9:
* Use FIELD_MAX and GENMASK
* Add tags
v7->v8:
* Report max sleep as well
---
drivers/platform/x86/intel/pmc/core.c | 4 ++++
drivers/platform/x86/intel/pmc/core.h | 2 ++
2 files changed, 6 insertions(+)
diff --git a/drivers/platform/x86/intel/pmc/core.c b/drivers/platform/x86/intel/pmc/core.c
index 925c5d676a43..298f27ba1e10 100644
--- a/drivers/platform/x86/intel/pmc/core.c
+++ b/drivers/platform/x86/intel/pmc/core.c
@@ -1153,6 +1153,8 @@ static int pmc_core_probe(struct platform_device *pdev)
pmc_core_do_dmi_quirks(pmcdev);
pmc_core_dbgfs_register(pmcdev);
+ pm_report_max_hw_sleep(FIELD_MAX(SLP_S0_RES_COUNTER_MASK) *
+ pmc_core_adjust_slp_s0_step(pmcdev, 1));
device_initialized = true;
dev_info(&pdev->dev, " initialized\n");
@@ -1214,6 +1216,8 @@ static inline bool pmc_core_is_s0ix_failed(struct pmc_dev *pmcdev)
if (pmc_core_dev_state_get(pmcdev, &s0ix_counter))
return false;
+ pm_report_hw_sleep_time((u32)(s0ix_counter - pmcdev->s0ix_counter));
+
if (s0ix_counter == pmcdev->s0ix_counter)
return true;
diff --git a/drivers/platform/x86/intel/pmc/core.h b/drivers/platform/x86/intel/pmc/core.h
index 51d73efceaf3..9ca9b9746719 100644
--- a/drivers/platform/x86/intel/pmc/core.h
+++ b/drivers/platform/x86/intel/pmc/core.h
@@ -16,6 +16,8 @@
#include <linux/bits.h>
#include <linux/platform_device.h>
+#define SLP_S0_RES_COUNTER_MASK GENMASK(31, 0)
+
#define PMC_BASE_ADDR_DEFAULT 0xFE000000
/* Sunrise Point Power Management Controller PCI Device ID */
--
2.34.1
Hi Mario, et al.,
On 4/14/23 03:23, Mario Limonciello wrote:
> An important part of validating that s0ix worked properly is to check how
> much of a cycle was spent in a hardware sleep state.
>
> The reporting of hardware sleep is a mix of kernel messages and sysfs
> files that vary from vendor to vendor. Collecting this information
> requires extra information on the kernel command line or fetching from
> debugfs.
>
> To make this information more readily accessible introduce a new file in
> suspend_stats that drivers can report into during their resume routine.
>
> Userspace can fetch this information and compare it against the duration
> of the cycle to allow determining residency percentages and flagging
> problems.
>
> Mario Limonciello (4):
> PM: Add sysfs files to represent time spent in hardware sleep state
> platform/x86/amd: pmc: Report duration of time in hw sleep state
> platform/x86/intel/pmc: core: Always capture counters on suspend
> platform/x86/intel/pmc: core: Report duration of time in HW sleep
> state
>
> Documentation/ABI/testing/sysfs-power | 29 +++++++++++++
> drivers/platform/x86/amd/pmc.c | 6 +--
> drivers/platform/x86/intel/pmc/core.c | 17 ++++----
> drivers/platform/x86/intel/pmc/core.h | 4 +-
> include/linux/suspend.h | 8 ++++
> kernel/power/main.c | 59 +++++++++++++++++++++------
> 6 files changed, 98 insertions(+), 25 deletions(-)
>
>
> base-commit: 09a9639e56c01c7a00d6c0ca63f4c7c41abe075d
Thank you for working on this, this looks good to me:
Reviewed-by: Hans de Goede <[email protected]>
For the series. Since this also touches kernel/power/main.c
I think it would be best if the entire series is merged
through the linux-pm tree and I'm fine with the pdx86 bits
also getting merged through linux-pm.
Rafael ?
Regards,
Hans
On Mon, Apr 17, 2023 at 10:25 AM Hans de Goede <[email protected]> wrote:
>
> Hi Mario, et al.,
>
> On 4/14/23 03:23, Mario Limonciello wrote:
> > An important part of validating that s0ix worked properly is to check how
> > much of a cycle was spent in a hardware sleep state.
> >
> > The reporting of hardware sleep is a mix of kernel messages and sysfs
> > files that vary from vendor to vendor. Collecting this information
> > requires extra information on the kernel command line or fetching from
> > debugfs.
> >
> > To make this information more readily accessible introduce a new file in
> > suspend_stats that drivers can report into during their resume routine.
> >
> > Userspace can fetch this information and compare it against the duration
> > of the cycle to allow determining residency percentages and flagging
> > problems.
> >
> > Mario Limonciello (4):
> > PM: Add sysfs files to represent time spent in hardware sleep state
> > platform/x86/amd: pmc: Report duration of time in hw sleep state
> > platform/x86/intel/pmc: core: Always capture counters on suspend
> > platform/x86/intel/pmc: core: Report duration of time in HW sleep
> > state
> >
> > Documentation/ABI/testing/sysfs-power | 29 +++++++++++++
> > drivers/platform/x86/amd/pmc.c | 6 +--
> > drivers/platform/x86/intel/pmc/core.c | 17 ++++----
> > drivers/platform/x86/intel/pmc/core.h | 4 +-
> > include/linux/suspend.h | 8 ++++
> > kernel/power/main.c | 59 +++++++++++++++++++++------
> > 6 files changed, 98 insertions(+), 25 deletions(-)
> >
> >
> > base-commit: 09a9639e56c01c7a00d6c0ca63f4c7c41abe075d
>
> Thank you for working on this, this looks good to me:
>
> Reviewed-by: Hans de Goede <[email protected]>
>
> For the series. Since this also touches kernel/power/main.c
> I think it would be best if the entire series is merged
> through the linux-pm tree and I'm fine with the pdx86 bits
> also getting merged through linux-pm.
>
> Rafael ?
That would be fine with me, but I've only got the [1/4].
Mario, can you please resend the series with CCs to linux-pm and with
R-by from Hans?