Skip to contentSkip to conversation

platform/x86/intel/pmc: Refactor platform resume functions to use cnl_resume()

PATCH Superseded David E. Box 3 participants 7 messages Oct 11, 2024 – Oct 13, 2024

Subsystems (2)
x86 platform drivers Intel PMC core driver

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

David E. Box

author
#0 lore raw

PATCH v2 1/2 platform/x86/intel/pmc: Refactor platform resume functions to use cnl_resume()

Several platform resume functions currently call pmc_core_send_ltr_ignore() and pmc_core_resume_common(), both of which are already called by cnl_resume(). Simplify the code by having these functions call cnl_resume() directly.

Signed-off-by David E. Box <[email protected]>

---

V2 - New patch. Split from V1

 drivers/platform/x86/intel/pmc/arl.c | 3 +--
 drivers/platform/x86/intel/pmc/lnl.c | 3 +--
 drivers/platform/x86/intel/pmc/mtl.c | 3 +--
 3 files changed, 3 insertions(+), 6 deletions(-)
drivers/platform/x86/intel/pmc/arl.c +1 -2
index e10527c4e3e0..05dec4f5019f 100644
@@ -687,9 +687,8 @@ static void arl_d3_fixup(void)
 static int arl_resume(struct pmc_dev *pmcdev)
 {
 	arl_d3_fixup();
-	pmc_core_send_ltr_ignore(pmcdev, 3, 0);
 
-	return pmc_core_resume_common(pmcdev);
+	return cnl_resume(pmcdev);
 }
 
 int arl_core_init(struct pmc_dev *pmcdev)
drivers/platform/x86/intel/pmc/lnl.c +1 -2
index e7a8077d1a3e..be029f12cdf4 100644
@@ -546,9 +546,8 @@ static void lnl_d3_fixup(void)
 static int lnl_resume(struct pmc_dev *pmcdev)
 {
 	lnl_d3_fixup();
-	pmc_core_send_ltr_ignore(pmcdev, 3, 0);
 
-	return pmc_core_resume_common(pmcdev);
+	return cnl_resume(pmcdev);
 }
 
 int lnl_core_init(struct pmc_dev *pmcdev)
drivers/platform/x86/intel/pmc/mtl.c +1 -2
index 91f2fa728f5c..fc6a89b8979f 100644
@@ -988,9 +988,8 @@ static void mtl_d3_fixup(void)
 static int mtl_resume(struct pmc_dev *pmcdev)
 {
 	mtl_d3_fixup();
-	pmc_core_send_ltr_ignore(pmcdev, 3, 0);
 
-	return pmc_core_resume_common(pmcdev);
+	return cnl_resume(pmcdev);
 }
 
 int mtl_core_init(struct pmc_dev *pmcdev)

base-commit: 9852d85ec9d492ebef56dc5f229416c925758edc

Signature · 1 line
2.43.0
author
#1 lore raw

PATCH v2 2/2 platform/x86/intel/pmc: Disable C1 auto-demotion during suspend

On some platforms, aggressive C1 auto-demotion may lead to failure to enter the deepest C-state during suspend-to-idle, causing high power consumption. To prevent this, disable C1 auto-demotion during suspend and re-enable on resume.

Signed-off-by David E. Box <[email protected]>

---

V2 - Remove #define DEBUG
- Move refactor of cnl_resume() to separate patch
- Use smp_call_function() to disable and restore C1_AUTO_DEMOTE
- Add comment that the MSR is per core, not per package.
- Add comment that the online cpu mask remains unchanged during suspend due to frozen userspace.

 drivers/platform/x86/intel/pmc/cnp.c | 53 ++++++++++++++++++++++++++++
 1 file changed, 53 insertions(+)
drivers/platform/x86/intel/pmc/cnp.c +53 -0
index 513c02670c5a..f12d4f0f9e93 100644
@@ -8,6 +8,8 @@
  *
  */
 
+#include <linux/smp.h>
+#include <linux/suspend.h>
 #include "core.h"
 
 /* Cannon Lake: PGD PFET Enable Ack Status Register(s) bitmap */

@@ -206,8 +208,52 @@ const struct pmc_reg_map cnp_reg_map = {
 	.etr3_offset = ETR3_OFFSET,
 };
 
+
+/*
+ * Disable C1 auto-demotion
+ *
+ * Aggressive C1 auto-demotion may lead to failure to enter the deepest C-state
+ * during suspend-to-idle, causing high power consumption. To prevent this, we
+ * disable C1 auto-demotion during suspend and re-enable on resume.
+ *
+ * Note that, although MSR_PKG_CST_CONFIG_CONTROL has 'package' in its name, it
+ * is actually a per-core MSR on client platforms, affecting only a single CPU.
+ * Therefore, it must be configured on all online CPUs. The online cpu mask is
+ * unchanged during the phase of suspend/resume as user space is frozen.
+ */
+
+static DEFINE_PER_CPU(u64, pkg_cst_config);
+
+static void disable_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
+	u64 val;
+
+	rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	per_cpu(pkg_cst_config, cpunum) = val;
+	val &= ~NHM_C1_AUTO_DEMOTE;
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum, val);
+}
+
+static void restore_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
+
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum,
+		 per_cpu(pkg_cst_config, cpunum));
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, per_cpu(pkg_cst_config, cpunum));
+}
+
 void cnl_suspend(struct pmc_dev *pmcdev)
 {
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();
+		disable_c1_auto_demote(NULL);
+		smp_call_function(disable_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}
+
 	/*
 	 * Due to a hardware limitation, the GBE LTR blocks PC10
 	 * when a cable is attached. To unblock PC10 during suspend,

@@ -218,6 +264,13 @@ void cnl_suspend(struct pmc_dev *pmcdev)
 
 int cnl_resume(struct pmc_dev *pmcdev)
 {
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();
+		restore_c1_auto_demote(NULL);
+		smp_call_function(restore_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}
+
 	pmc_core_send_ltr_ignore(pmcdev, 3, 0);
 
 	return pmc_core_resume_common(pmcdev);
Signature · 1 line
2.43.0
#2 lore raw

Re: [PATCH V2 2/2] platform/x86/intel/pmc: Disable C1 auto-demotion during suspend

Quoted 62 lines · David E. Box

On some platforms, aggressive C1 auto-demotion may lead to failure to enter
the deepest C-state during suspend-to-idle, causing high power consumption.
To prevent this, disable C1 auto-demotion during suspend and re- enable on
resume.

Signed-off-by David E. Box <[email protected]>

---

V2 - Remove #define DEBUG
   - Move refactor of cnl_resume() to separate patch
   - Use smp_call_function() to disable and restore C1_AUTO_DEMOTE
   - Add comment that the MSR is per core, not per package.
   - Add comment that the online cpu mask remains unchanged during suspend due to frozen userspace.

 drivers/platform/x86/intel/pmc/cnp.c | 53
++++++++++++++++++++++++++++

 1 file changed, 53 insertions(+)
drivers/platform/x86/intel/pmc/cnp.c +0 -0
index 513c02670c5a..f12d4f0f9e93 100644
@@ -8,6 +8,8 @@
   *
   */

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+#include <linux/smp.h>
+#include <linux/suspend.h>

 #include "core.h"

 /* Cannon Lake: PGD PFET Enable Ack Status Register(s) bitmap */

drivers/platform/x86/intel/pmc/cnp.c +0 -0
@@ -206,8 +208,52 @@ const struct pmc_reg_map cnp_reg_map = {
  	.etr3_offset = ETR3_OFFSET,
  };

drivers/platform/x86/intel/pmc/cnp.c +5 -0
+
+/*
+ * Disable C1 auto-demotion
+ *
+ * Aggressive C1 auto-demotion may lead to failure to enter the

deepest C-state
+ * during suspend-to-idle, causing high power consumption. To prevent this, we
+ * disable C1 auto-demotion during suspend and re-enable on resume.
+ *
+ * Note that, although MSR_PKG_CST_CONFIG_CONTROL has 'package' in its name, it
+ * is actually a per-core MSR on client platforms, affecting only a single CPU.
+ * Therefore, it must be configured on all online CPUs. The online cpu mask is
+ * unchanged during the phase of suspend/resume as user space is frozen.
+ */

drivers/platform/x86/intel/pmc/cnp.c +3 -0
+
+static DEFINE_PER_CPU(u64, pkg_cst_config);
+

drivers/platform/x86/intel/pmc/cnp.c const struct pmc_reg_map cnp_reg_map = { · line 236

Quoted 10 lines
drivers/platform/x86/intel/pmc/cnp.c +10 -0
+static void disable_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
+	u64 val;
+
+	rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	per_cpu(pkg_cst_config, cpunum) = val;
+	val &= ~NHM_C1_AUTO_DEMOTE;
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum, val);

Do you want to leave pr_debug?

Quoted 14 lines
drivers/platform/x86/intel/pmc/cnp.c +9 -0
+}
+
+static void restore_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
+
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum,
+		 per_cpu(pkg_cst_config, cpunum));
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, per_cpu(pkg_cst_config,

cpunum));

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+}
+

 void cnl_suspend(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c

Quoted 2 lines
drivers/platform/x86/intel/pmc/cnp.c +2 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();

Why do you need this?

Thanks,
Srinivas

Quoted 22 lines
drivers/platform/x86/intel/pmc/cnp.c +5 -0
+		disable_c1_auto_demote(NULL);
+		smp_call_function(disable_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}
+

  /*
  * Due to a hardware limitation, the GBE LTR blocks PC10
  * when a cable is attached. To unblock PC10 during suspend,

drivers/platform/x86/intel/pmc/cnp.c +0 -0
@@ -218,6 +264,13 @@ void cnl_suspend(struct pmc_dev *pmcdev)


 int cnl_resume(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c +7 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();
+		restore_c1_auto_demote(NULL);
+		smp_call_function(restore_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}
+

  pmc_core_send_ltr_ignore(pmcdev, 3, 0);

  return pmc_core_resume_common(pmcdev);

author
#3 lore raw

Re: [PATCH V2 2/2] platform/x86/intel/pmc: Disable C1 auto-demotion during suspend

Quoted 66 lines · srinivas pandruvada
Quoted 65 lines · David E. Box

On some platforms, aggressive C1 auto-demotion may lead to failure to enter
the deepest C-state during suspend-to-idle, causing high power consumption.
To prevent this, disable C1 auto-demotion during suspend and re- enable on
resume.

Signed-off-by David E. Box <[email protected]>

---

V2 - Remove #define DEBUG
   - Move refactor of cnl_resume() to separate patch
   - Use smp_call_function() to disable and restore C1_AUTO_DEMOTE
   - Add comment that the MSR is per core, not per package.
   - Add comment that the online cpu mask remains unchanged during suspend due to frozen userspace.

 drivers/platform/x86/intel/pmc/cnp.c | 53
++++++++++++++++++++++++++++

 1 file changed, 53 insertions(+)
drivers/platform/x86/intel/pmc/cnp.c +0 -0
index 513c02670c5a..f12d4f0f9e93 100644
@@ -8,6 +8,8 @@
   *
   */

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+#include <linux/smp.h>
+#include <linux/suspend.h>

 #include "core.h"

 /* Cannon Lake: PGD PFET Enable Ack Status Register(s) bitmap */

drivers/platform/x86/intel/pmc/cnp.c +0 -0
@@ -206,8 +208,52 @@ const struct pmc_reg_map cnp_reg_map = {
  	.etr3_offset = ETR3_OFFSET,
  };

drivers/platform/x86/intel/pmc/cnp.c +5 -0
+
+/*
+ * Disable C1 auto-demotion
+ *
+ * Aggressive C1 auto-demotion may lead to failure to enter the

deepest C-state
+ * during suspend-to-idle, causing high power consumption. To prevent this, we
+ * disable C1 auto-demotion during suspend and re-enable on resume.
+ *
+ * Note that, although MSR_PKG_CST_CONFIG_CONTROL has 'package' in its name, it
+ * is actually a per-core MSR on client platforms, affecting only a single CPU.
+ * Therefore, it must be configured on all online CPUs. The online cpu mask is
+ * unchanged during the phase of suspend/resume as user space is frozen.
+ */

drivers/platform/x86/intel/pmc/cnp.c +6 -0
+
+static DEFINE_PER_CPU(u64, pkg_cst_config);
+
+static void disable_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
Quoted 8 lines

drivers/platform/x86/intel/pmc/cnp.c const struct pmc_reg_map cnp_reg_map = { · line 236

Quoted 7 lines
drivers/platform/x86/intel/pmc/cnp.c +7 -0
+	u64 val;
+
+	rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	per_cpu(pkg_cst_config, cpunum) = val;
+	val &= ~NHM_C1_AUTO_DEMOTE;
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum, val);

Do you want to leave pr_debug?

Thought it could be useful but it can be removed.

Quoted 10 lines
Quoted 10 lines
drivers/platform/x86/intel/pmc/cnp.c +9 -0
+}
+
+static void restore_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
+
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum,
+		 per_cpu(pkg_cst_config, cpunum));
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, per_cpu(pkg_cst_config,
Quoted 8 lines

drivers/platform/x86/intel/pmc/cnp.c

Quoted 7 lines

cpunum));

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+}
+

 void cnl_suspend(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();

Why do you need this?

To ensure that the cpu doesn't change between the next two calls.

David

Quoted 28 lines

Thanks,
Srinivas

Quoted 22 lines
drivers/platform/x86/intel/pmc/cnp.c +5 -0
+		disable_c1_auto_demote(NULL);
+		smp_call_function(disable_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}
+

  /*
  * Due to a hardware limitation, the GBE LTR blocks PC10
  * when a cable is attached. To unblock PC10 during suspend,

drivers/platform/x86/intel/pmc/cnp.c +0 -0
@@ -218,6 +264,13 @@ void cnl_suspend(struct pmc_dev *pmcdev)


 int cnl_resume(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c +7 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();
+		restore_c1_auto_demote(NULL);
+		smp_call_function(restore_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}
+

  pmc_core_send_ltr_ignore(pmcdev, 3, 0);

  return pmc_core_resume_common(pmcdev);

#4 lore raw

Re: [PATCH V2 2/2] platform/x86/intel/pmc: Disable C1 auto-demotion during suspend

Quoted 106 lines · David E. Box
Quoted 74 lines · srinivas pandruvada
Quoted 73 lines · David E. Box

On some platforms, aggressive C1 auto-demotion may lead to failure to
enter
the deepest C-state during suspend-to-idle, causing high power consumption.
To prevent this, disable C1 auto-demotion during suspend and re- enable on
resume.

Signed-off-by David E. Box <[email protected]>

---

V2 - Remove #define DEBUG
   - Move refactor of cnl_resume() to separate patch
   - Use smp_call_function() to disable and restore C1_AUTO_DEMOTE
   - Add comment that the MSR is per core, not per package.
   - Add comment that the online cpu mask remains unchanged during
     suspend due to frozen userspace.

 drivers/platform/x86/intel/pmc/cnp.c | 53
++++++++++++++++++++++++++++

 1 file changed, 53 insertions(+)
drivers/platform/x86/intel/pmc/cnp.c +0 -0
index 513c02670c5a..f12d4f0f9e93 100644
@@ -8,6 +8,8 @@
   *
   */

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+#include <linux/smp.h>
+#include <linux/suspend.h>

 #include "core.h"

 /* Cannon Lake: PGD PFET Enable Ack Status Register(s) bitmap */

drivers/platform/x86/intel/pmc/cnp.c +0 -0
@@ -206,8 +208,52 @@ const struct pmc_reg_map cnp_reg_map = {
  	.etr3_offset = ETR3_OFFSET,
  };

drivers/platform/x86/intel/pmc/cnp.c +5 -0
+
+/*
+ * Disable C1 auto-demotion
+ *
+ * Aggressive C1 auto-demotion may lead to failure to enter the

deepest C-state
+ * during suspend-to-idle, causing high power consumption. To prevent this, we
+ * disable C1 auto-demotion during suspend and re-enable on resume.
+ *
+ * Note that, although MSR_PKG_CST_CONFIG_CONTROL has 'package' in
its name, it
+ * is actually a per-core MSR on client platforms, affecting only a
single CPU.
+ * Therefore, it must be configured on all online CPUs. The online
cpu mask is
+ * unchanged during the phase of suspend/resume as user space is frozen.
+ */

drivers/platform/x86/intel/pmc/cnp.c +7 -0
+
+static DEFINE_PER_CPU(u64, pkg_cst_config);
+
+static void disable_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
+	u64 val;
Quoted 8 lines
Quoted 7 lines
drivers/platform/x86/intel/pmc/cnp.c +6 -0
+
+	rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	per_cpu(pkg_cst_config, cpunum) = val;
+	val &= ~NHM_C1_AUTO_DEMOTE;
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum,

val);

Do you want to leave pr_debug?

Thought it could be useful but it can be removed.

Quoted 11 lines
Quoted 11 lines
drivers/platform/x86/intel/pmc/cnp.c +9 -0
+}
+
+static void restore_c1_auto_demote(void *unused)
+{
+	int cpunum = smp_processor_id();
+
+	pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum,
+		 per_cpu(pkg_cst_config, cpunum));
+	wrmsrl(MSR_PKG_CST_CONFIG_CONTROL,

per_cpu(pkg_cst_config,

Quoted 8 lines

drivers/platform/x86/intel/pmc/cnp.c

Quoted 7 lines

cpunum));

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+}
+

 void cnl_suspend(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c +2 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();

Why do you need this?

Quoted 1 line

To ensure that the cpu doesn't change between the next two calls.

Correct. You need for smp_processor_id()

Generally for this avoiding issue with smp_processor_id(), you can use get_cpu(), which gives current cpu as return value and also disable preemption.

Something like this

                this_cpu = get_cpu();
                disable_c1_auto_demote(&this_cpu);
smp_call_function_many(cpu_online_mask, disable_c1_auto_demote, NULL,
0);
                put_cpu();

But fine, this makes your code more complex as you have to pass now a param and use for local_cpu and use smp_procesor_id() for remote call via smp_call..

Thanks,
Srinivas

Quoted 36 lines

David

Quoted 31 lines

Thanks,
Srinivas

Quoted 25 lines
drivers/platform/x86/intel/pmc/cnp.c +2 -0
+		disable_c1_auto_demote(NULL);
+		smp_call_function(disable_c1_auto_demote, NULL,

0);

drivers/platform/x86/intel/pmc/cnp.c +3 -0
+		preempt_enable();
+	}
+

  /*
  * Due to a hardware limitation, the GBE LTR blocks PC10
  * when a cable is attached. To unblock PC10 during suspend,

drivers/platform/x86/intel/pmc/cnp.c +0 -0
@@ -218,6 +264,13 @@ void cnl_suspend(struct pmc_dev *pmcdev)


 int cnl_resume(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c +4 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();
+		restore_c1_auto_demote(NULL);
+		smp_call_function(restore_c1_auto_demote, NULL,

0);

drivers/platform/x86/intel/pmc/cnp.c +3 -0
+		preempt_enable();
+	}
+

  pmc_core_send_ltr_ignore(pmcdev, 3, 0);

  return pmc_core_resume_common(pmcdev);

#5 lore raw

Re: [PATCH V2 2/2] platform/x86/intel/pmc: Disable C1 auto-demotion during suspend

In reply to David E. Box
Quoted 53 lines · David E. Box

On some platforms, aggressive C1 auto-demotion may lead to failure to enter the deepest C-state during suspend-to-idle, causing high power consumption. To prevent this, disable C1 auto-demotion during suspend and re-enable on resume.

Signed-off-by David E. Box <[email protected]>

---

V2 - Remove #define DEBUG
- Move refactor of cnl_resume() to separate patch
- Use smp_call_function() to disable and restore C1_AUTO_DEMOTE
- Add comment that the MSR is per core, not per package.
- Add comment that the online cpu mask remains unchanged during suspend due to frozen userspace.

 drivers/platform/x86/intel/pmc/cnp.c | 53 ++++++++++++++++++++++++++++
 1 file changed, 53 insertions(+)
drivers/platform/x86/intel/pmc/cnp.c +19 -0
index 513c02670c5a..f12d4f0f9e93 100644
@@ -8,6 +8,8 @@
  *
  */
 
+#include <linux/smp.h>
+#include <linux/suspend.h>
 #include "core.h"
 
 /* Cannon Lake: PGD PFET Enable Ack Status Register(s) bitmap */

@@ -206,8 +208,52 @@ const struct pmc_reg_map cnp_reg_map = {
        .etr3_offset = ETR3_OFFSET,
 };
 
+
+/*
+ * Disable C1 auto-demotion
+ *
+ * Aggressive C1 auto-demotion may lead to failure to enter the deepest C-state
+ * during suspend-to-idle, causing high power consumption. To prevent this, we
+ * disable C1 auto-demotion during suspend and re-enable on resume.
+ *
+ * Note that, although MSR_PKG_CST_CONFIG_CONTROL has 'package' in its name, it
+ * is actually a per-core MSR on client platforms, affecting only a single CPU.
+ * Therefore, it must be configured on all online CPUs. The online cpu mask is
+ * unchanged during the phase of suspend/resume as user space is frozen.
+ */
+
+static DEFINE_PER_CPU(u64, pkg_cst_config);
+
+static void disable_c1_auto_demote(void *unused)

drivers/platform/x86/intel/pmc/cnp.c const struct pmc_reg_map cnp_reg_map = { · line 237

Quoted 10 lines
drivers/platform/x86/intel/pmc/cnp.c +10 -0
+{
+       int cpunum = smp_processor_id();
+       u64 val;
+
+       rdmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+       per_cpu(pkg_cst_config, cpunum) = val;
+       val &= ~NHM_C1_AUTO_DEMOTE;
+       wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, val);
+       pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum, val);
+}

If you decide to retain the debug statements, I generally prefer to separate them from the rest of the code with empty lines, for clarity.

drivers/platform/x86/intel/pmc/cnp.c const struct pmc_reg_map cnp_reg_map = { · line 247

Quoted 10 lines
drivers/platform/x86/intel/pmc/cnp.c +10 -0
+
+static void restore_c1_auto_demote(void *unused)
+{
+       int cpunum = smp_processor_id();
+
+       pr_debug("%s: cpu:%d cst %llx\n", __func__, cpunum,
+                per_cpu(pkg_cst_config, cpunum));
+       wrmsrl(MSR_PKG_CST_CONFIG_CONTROL, per_cpu(pkg_cst_config, cpunum));
+}
+

I would add something like

static void s2idle_cpu_quirk(smp_call_func_t func)

{
       if (pm_suspend_via_firmware())
              return;

       preempt_disable();
       func();
       smp_call_function(func, NULL, 0);
       preempt_enable();
}

drivers/platform/x86/intel/pmc/cnp.c const struct pmc_reg_map cnp_reg_map = { · line 256

Quoted 9 lines
drivers/platform/x86/intel/pmc/cnp.c +7 -0
 void cnl_suspend(struct pmc_dev *pmcdev)
 {
+       if (!pm_suspend_via_firmware()) {
+               preempt_disable();
+               disable_c1_auto_demote(NULL);
+               smp_call_function(disable_c1_auto_demote, NULL, 0);
+               preempt_enable();
+       }
+

And here I'd just do

      s2idle_cpu_quirk(disable_c1_auto_demote);

and analogously below.

Quoted 18 lines
drivers/platform/x86/intel/pmc/cnp.c +7 -1
        /*
         * Due to a hardware limitation, the GBE LTR blocks PC10
         * when a cable is attached. To unblock PC10 during suspend,

@@ -218,6 +264,13 @@ void cnl_suspend(struct pmc_dev *pmcdev)
 
 int cnl_resume(struct pmc_dev *pmcdev)
 {
+       if (!pm_suspend_via_firmware()) {
+               preempt_disable();
+               restore_c1_auto_demote(NULL);
+               smp_call_function(restore_c1_auto_demote, NULL, 0);
+               preempt_enable();
+       }
+
        pmc_core_send_ltr_ignore(pmcdev, 3, 0);
 
        return pmc_core_resume_common(pmcdev);
--
#6 lore raw

Re: [PATCH V2 2/2] platform/x86/intel/pmc: Disable C1 auto-demotion during suspend

In reply to David E. Box

+Ricardo

Quoted 11 lines · David E. Box

On some platforms, aggressive C1 auto-demotion may lead to failure to enter
the deepest C-state during suspend-to-idle, causing high power consumption.
To prevent this, disable C1 auto-demotion during suspend and re- enable on
resume.

Signed-off-by David E. Box <[email protected]>

---

[...]

drivers/platform/x86/intel/pmc/cnp.c const struct pmc_reg_map cnp_reg_map = { · line 255

Quoted 9 lines

 void cnl_suspend(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c +6 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();
+		disable_c1_auto_demote(NULL);
+		smp_call_function(disable_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}

As suggested by Ricardo using the following will work avoiding separate local CPU call. Preemption will be disabled, no need separate call. Also cpu_online_mask can't be changed during these callbacks.

if (!pm_suspend_via_firmware())

	on_each_cpu(disable_c1_auto_demote, NULL, true);

I think you need wait=true. I have seen on some occasions you will miss msr write as we called mwait before before async calls finished.

Thanks,
Srinivas

Quoted 18 lines

+
  /*
  * Due to a hardware limitation, the GBE LTR blocks PC10
  * when a cable is attached. To unblock PC10 during suspend,

(quoted diff) +0 -0
@@ -218,6 +264,13 @@ void cnl_suspend(struct pmc_dev *pmcdev)


 int cnl_resume(struct pmc_dev *pmcdev)
 {

drivers/platform/x86/intel/pmc/cnp.c +7 -0
+	if (!pm_suspend_via_firmware()) {
+		preempt_disable();
+		restore_c1_auto_demote(NULL);
+		smp_call_function(restore_c1_auto_demote, NULL, 0);
+		preempt_enable();
+	}
+

  pmc_core_send_ltr_ignore(pmcdev, 3, 0);

  return pmc_core_resume_common(pmcdev);