[PATCH 0/2] ata: ahci: Enable DEVSLP by default on SLP_S0 support

[PATCH 0/2] ata: ahci: Enable DEVSLP by default on SLP_S0 support

[Srinivas Pandruvada]

One of the requirement for modern x86 system to enter lowest power mode
(SLP_S0) is SATA IP block to be off. This is true even during when
platform is suspended to idle and not only in opportunistic (runtime)
suspend.
This series is to enable DEVSLP by default.

-current (non-rfc) series
Implemented suggestions from Hans
- no override of policy set via module param
- Overide lpm policy per host not global policy
- changed the new policy name to "min_power_with_partial"

rfc-v2
- As suggested by Hans, it is possible to have ASP with DEVSLP, so
add a new state.
- Removed usage of mem_sleep_current, instead just rely on low power
idle flag. Don't feel good to EXPORT from core suspend code.
- Depending host policy to decide if we enable DEVSLP by default.

Srinivas Pandruvada (2):
  ata: ahci: Support state with min power but Partial low power state
  ata: ahci: Enable DEVSLP by default on x86 with SLP_S0

 drivers/ata/ahci.c        | 33 +++++++++++++++++++++++++++++----
 drivers/ata/libahci.c     |  6 +++++-
 drivers/ata/libata-core.c |  1 +
 drivers/ata/libata-scsi.c |  1 +
 include/linux/libata.h    |  3 ++-
 5 files changed, 38 insertions(+), 6 deletions(-)

-- 
2.17.1

[PATCH 1/2] ata: ahci: Support state with min power but Partial low power state

[Srinivas Pandruvada]

Currently when min_power policy is selected, the partial low power state
is not entered and link will try aggressively enter to only slumber state.
Add a new policy which still enable DEVSLP but also try to enter partial
low power state. This policy is presented as "min_power_with_partial".

For information the difference between partial and slumber
Partial – PHY logic is powered up, and in a reduced power state. The link
PM exit latency to active state maximum is 10 ns.
Slumber – PHY logic is powered up, and in a reduced power state. The link
PM exit latency to active state maximum is 10 ms.
Devslp – PHY logic is powered down. The link PM exit latency from this
state to active state maximum is 20 ms, unless otherwise specified by
DETO.

Suggested-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
---
 drivers/ata/libahci.c     | 5 ++++-
 drivers/ata/libata-core.c | 1 +
 drivers/ata/libata-scsi.c | 1 +
 include/linux/libata.h    | 3 ++-
 4 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/ata/libahci.c b/drivers/ata/libahci.c
index 4c49c7c2a42c..94dcec2a8a5b 100644
--- a/drivers/ata/libahci.c
+++ b/drivers/ata/libahci.c
@@ -800,6 +800,8 @@ static int ahci_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,
 			cmd |= PORT_CMD_ALPE;
 			if (policy == ATA_LPM_MIN_POWER)
 				cmd |= PORT_CMD_ASP;
+			else if (policy == ATA_LPM_MIN_POWER_WITH_PARTIAL)
+				cmd &= ~PORT_CMD_ASP;
 
 			/* write out new cmd value */
 			writel(cmd, port_mmio + PORT_CMD);
@@ -810,7 +812,8 @@ static int ahci_set_lpm(struct ata_link *link, enum ata_lpm_policy policy,
 	if ((hpriv->cap2 & HOST_CAP2_SDS) &&
 	    (hpriv->cap2 & HOST_CAP2_SADM) &&
 	    (link->device->flags & ATA_DFLAG_DEVSLP)) {
-		if (policy == ATA_LPM_MIN_POWER)
+		if (policy == ATA_LPM_MIN_POWER ||
+		    policy == ATA_LPM_MIN_POWER_WITH_PARTIAL)
 			ahci_set_aggressive_devslp(ap, true);
 		else
 			ahci_set_aggressive_devslp(ap, false);
diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
index 27d15ed7fa3d..773b5f8eed1e 100644
--- a/drivers/ata/libata-core.c
+++ b/drivers/ata/libata-core.c
@@ -3967,6 +3967,7 @@ int sata_link_scr_lpm(struct ata_link *link, enum ata_lpm_policy policy,
 		scontrol |= (0x6 << 8);
 		break;
 	case ATA_LPM_MED_POWER_WITH_DIPM:
+	case ATA_LPM_MIN_POWER_WITH_PARTIAL:
 	case ATA_LPM_MIN_POWER:
 		if (ata_link_nr_enabled(link) > 0)
 			/* no restrictions on LPM transitions */
diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
index 6a91d04351d9..4648bd5cee15 100644
--- a/drivers/ata/libata-scsi.c
+++ b/drivers/ata/libata-scsi.c
@@ -110,6 +110,7 @@ static const char *ata_lpm_policy_names[] = {
 	[ATA_LPM_MAX_POWER]		= "max_performance",
 	[ATA_LPM_MED_POWER]		= "medium_power",
 	[ATA_LPM_MED_POWER_WITH_DIPM]	= "med_power_with_dipm",
+	[ATA_LPM_MIN_POWER_WITH_PARTIAL] = "min_power_with_partial",
 	[ATA_LPM_MIN_POWER]		= "min_power",
 };
 
diff --git a/include/linux/libata.h b/include/linux/libata.h
index 8b8946dd63b9..975543f8c1ff 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -522,7 +522,8 @@ enum ata_lpm_policy {
 	ATA_LPM_MAX_POWER,
 	ATA_LPM_MED_POWER,
 	ATA_LPM_MED_POWER_WITH_DIPM, /* Med power + DIPM as win IRST does */
-	ATA_LPM_MIN_POWER,
+	ATA_LPM_MIN_POWER_WITH_PARTIAL, /* Min Power + partial and slumber */
+	ATA_LPM_MIN_POWER, /* Min power + no partial (slumber only) */
 };
 
 enum ata_lpm_hints {
-- 
2.17.1

[PATCH 2/2] ata: ahci: Enable DEVSLP by default on x86 with SLP_S0

[Srinivas Pandruvada]

One of the requirement for modern x86 system to enter lowest power mode
(SLP_S0) is SATA IP block to be off. This is true even during when
platform is suspended to idle and not only in opportunistic (runtime)
suspend.

Several of these system don't have traditional ACPI S3, so it is
important that they enter SLP_S0 state, to avoid draining battery even
during suspend. So it is important that out of the box Linux installation
reach this state.

SATA IP block doesn't get turned off till SATA is in DEVSLP mode. Here
user has to either use scsi-host sysfs or tools like powertop to set
the sata-host link_power_management_policy to min_power.

This change sets by default link power management policy to min_power
with partial (preferred) or slumber support on idle for some platforms.

To avoid regressions, the following conditions are used:
- User didn't override the policy from module parameter
- The kernel config is already set to use med_power_with_dipm or deeper
- System is a SLP_S0 capable using ACPI low power idle flag
This combination will make sure that systems are fairly recent and
since getting shipped with SLP_S0 support, the DEVSLP function
is already validated.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
---
 drivers/ata/ahci.c | 33 +++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)

diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
index 738fb22978dd..72dc9edbc221 100644
--- a/drivers/ata/ahci.c
+++ b/drivers/ata/ahci.c
@@ -1550,6 +1550,34 @@ static int ahci_init_msi(struct pci_dev *pdev, unsigned int n_ports,
 	return pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_MSIX);
 }
 
+void ahci_update_initial_lpm_policy(struct ata_port *ap,
+				    struct ahci_host_priv *hpriv)
+{
+	int policy = mobile_lpm_policy;
+
+	/* Ignore processing for non mobile platforms */
+	if (!(hpriv->flags & AHCI_HFLAG_IS_MOBILE))
+		return;
+
+	/* user modified policy via module param */
+	if (policy != CONFIG_SATA_MOBILE_LPM_POLICY)
+		goto update_policy;
+
+#ifdef CONFIG_ACPI
+	if (policy > ATA_LPM_MED_POWER &&
+	    (acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)) {
+		if (hpriv->cap & HOST_CAP_PART)
+			policy = ATA_LPM_MIN_POWER_WITH_PARTIAL;
+		else if (hpriv->cap & HOST_CAP_SSC)
+			policy = ATA_LPM_MIN_POWER;
+	}
+#endif
+
+update_policy:
+	if (policy >= ATA_LPM_UNKNOWN && policy <= ATA_LPM_MIN_POWER)
+		ap->target_lpm_policy = policy;
+}
+
 static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 {
 	unsigned int board_id = ent->driver_data;
@@ -1747,10 +1775,7 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 		if (ap->flags & ATA_FLAG_EM)
 			ap->em_message_type = hpriv->em_msg_type;
 
-		if ((hpriv->flags & AHCI_HFLAG_IS_MOBILE) &&
-		    mobile_lpm_policy >= ATA_LPM_UNKNOWN &&
-		    mobile_lpm_policy <= ATA_LPM_MIN_POWER)
-			ap->target_lpm_policy = mobile_lpm_policy;
+		ahci_update_initial_lpm_policy(ap, hpriv);
 
 		/* disabled/not-implemented port */
 		if (!(hpriv->port_map & (1 << i)))
-- 
2.17.1

Re: [PATCH 2/2] ata: ahci: Enable DEVSLP by default on x86 with SLP_S0

[Hans de Goede]

Hi,

On 07/20/2018 02:01 AM, Srinivas Pandruvada wrote:
> One of the requirement for modern x86 system to enter lowest power mode
> (SLP_S0) is SATA IP block to be off. This is true even during when
> platform is suspended to idle and not only in opportunistic (runtime)
> suspend.
> 
> Several of these system don't have traditional ACPI S3, so it is
> important that they enter SLP_S0 state, to avoid draining battery even
> during suspend. So it is important that out of the box Linux installation
> reach this state.
> 
> SATA IP block doesn't get turned off till SATA is in DEVSLP mode. Here
> user has to either use scsi-host sysfs or tools like powertop to set
> the sata-host link_power_management_policy to min_power.
> 
> This change sets by default link power management policy to min_power
> with partial (preferred) or slumber support on idle for some platforms.
> 
> To avoid regressions, the following conditions are used:
> - User didn't override the policy from module parameter
> - The kernel config is already set to use med_power_with_dipm or deeper
> - System is a SLP_S0 capable using ACPI low power idle flag
> This combination will make sure that systems are fairly recent and
> since getting shipped with SLP_S0 support, the DEVSLP function
> is already validated.
> 
> Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
> ---
>   drivers/ata/ahci.c | 33 +++++++++++++++++++++++++++++----
>   1 file changed, 29 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
> index 738fb22978dd..72dc9edbc221 100644
> --- a/drivers/ata/ahci.c
> +++ b/drivers/ata/ahci.c
> @@ -1550,6 +1550,34 @@ static int ahci_init_msi(struct pci_dev *pdev, unsigned int n_ports,
>   	return pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_MSIX);
>   }
>   
> +void ahci_update_initial_lpm_policy(struct ata_port *ap,
> +				    struct ahci_host_priv *hpriv)
> +{
> +	int policy = mobile_lpm_policy;
> +
> +	/* Ignore processing for non mobile platforms */
> +	if (!(hpriv->flags & AHCI_HFLAG_IS_MOBILE))
> +		return;
> +
> +	/* user modified policy via module param */
> +	if (policy != CONFIG_SATA_MOBILE_LPM_POLICY)
> +		goto update_policy;


This means that if the user explicitly requests the
default CONFIG_SATA_MOBILE_LPM_POLICY, it will still
be overridden below.

I think it would be better to initialize the global mobile_lpm_policy
to -1 and then do:

	int policy = CONFIG_SATA_MOBILE_LPM_POLICY;

	...

	/* user modified policy via module param */
	if (mobile_lpm_policy != -1) {
		policy = mobile_lpm_policy;
		goto update_policy;
	}


Otherwise this series looks good to me.

Regards,

Hans


> +
> +#ifdef CONFIG_ACPI
> +	if (policy > ATA_LPM_MED_POWER &&
> +	    (acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)) {
> +		if (hpriv->cap & HOST_CAP_PART)
> +			policy = ATA_LPM_MIN_POWER_WITH_PARTIAL;
> +		else if (hpriv->cap & HOST_CAP_SSC)
> +			policy = ATA_LPM_MIN_POWER;
> +	}
> +#endif
> +
> +update_policy:
> +	if (policy >= ATA_LPM_UNKNOWN && policy <= ATA_LPM_MIN_POWER)
> +		ap->target_lpm_policy = policy;
> +}
> +
>   static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>   {
>   	unsigned int board_id = ent->driver_data;
> @@ -1747,10 +1775,7 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>   		if (ap->flags & ATA_FLAG_EM)
>   			ap->em_message_type = hpriv->em_msg_type;
>   
> -		if ((hpriv->flags & AHCI_HFLAG_IS_MOBILE) &&
> -		    mobile_lpm_policy >= ATA_LPM_UNKNOWN &&
> -		    mobile_lpm_policy <= ATA_LPM_MIN_POWER)
> -			ap->target_lpm_policy = mobile_lpm_policy;
> +		ahci_update_initial_lpm_policy(ap, hpriv);
>   
>   		/* disabled/not-implemented port */
>   		if (!(hpriv->port_map & (1 << i)))
>