Skip to contentSkip to conversation

nvme-pci: Add sleep quirk for Kingston drives

PATCH Applied Werner Sembach 4 participants 6 messages Sep 20, 2023 – Dec 7, 2023 NVMe

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 nvme-pci: Add sleep quirk for Kingston drives

From: Georg Gottleuber <[email protected]>

Some Kingston NV1 and A2000 are wasting a lot of power on specific TUXEDO platforms in s2idle sleep if 'Simple Suspend' is used.

This patch applies a new quirk 'Force No Simple Suspend' to achieve a low power sleep without 'Simple Suspend'.

---
 drivers/nvme/host/nvme.h |  5 +++++
 drivers/nvme/host/pci.c  | 16 +++++++++++++++-
 2 files changed, 20 insertions(+), 1 deletion(-)
drivers/nvme/host/nvme.h +5 -0
index f35647c470afa..74f74b459f5fa 100644
@@ -156,6 +156,11 @@ enum nvme_quirks {
 	 * No temperature thresholds for channels other than 0 (Composite).
 	 */
 	NVME_QUIRK_NO_SECONDARY_TEMP_THRESH	= (1 << 19),
+
+	/*
+	 * Disables simple suspend/resume path.
+	 */
+	NVME_QUIRK_FORCE_NO_SIMPLE_SUSPEND	= (1 << 20),
 };
 
 /*
drivers/nvme/host/pci.c +15 -1
index 347cb5daebc3c..b4e6f0d21a44c 100644
@@ -2903,6 +2903,18 @@ static unsigned long check_vendor_combination_bug(struct pci_dev *pdev)
 		if ((dmi_match(DMI_BOARD_VENDOR, "LENOVO")) &&
 		     dmi_match(DMI_BOARD_NAME, "LNVNB161216"))
 			return NVME_QUIRK_SIMPLE_SUSPEND;
+	} else if (pdev->vendor == 0x2646 && (pdev->device == 0x2263 ||
+		   pdev->device == 0x500f)) {
+		/*
+		 * Exclude some Kingston NV1 and A2000 devices from
+		 * NVME_QUIRK_SIMPLE_SUSPEND. Do a full suspend to save a
+		 * lot fo energy with s2idle sleep on some TUXEDO platforms.
+		 */
+		if (dmi_match(DMI_BOARD_NAME, "NS5X_NS7XAU") ||
+		    dmi_match(DMI_BOARD_NAME, "NS5x_7xAU") ||
+		    dmi_match(DMI_BOARD_NAME, "NS5x_7xPU") ||
+		    dmi_match(DMI_BOARD_NAME, "PH4PRX1_PH6PRX1"))
+			return NVME_QUIRK_FORCE_NO_SIMPLE_SUSPEND;
 	}
 
 	return 0;
@@ -2933,7 +2945,9 @@ static struct nvme_dev *nvme_pci_alloc_dev(struct pci_dev *pdev,
 	dev->dev = get_device(&pdev->dev);
 
 	quirks |= check_vendor_combination_bug(pdev);
-	if (!noacpi && acpi_storage_d3(&pdev->dev)) {
+	if (!noacpi &&
+	    !(quirks & NVME_QUIRK_FORCE_NO_SIMPLE_SUSPEND) &&
+	    acpi_storage_d3(&pdev->dev)) {
 		/*
 		 * Some systems use a bios work around to ask for D3 on
 		 * platforms that support kernel managed suspend.
Signature · 1 line
2.34.1
#1 lore raw

Hi,

Am 20.09.23 um 10:52 schrieb Werner Sembach:

Quoted 11 lines

From: Georg Gottleuber <[email protected]>

Some Kingston NV1 and A2000 are wasting a lot of power on specific TUXEDO platforms in s2idle sleep if 'Simple Suspend' is used.

This patch applies a new quirk 'Force No Simple Suspend' to achieve a low power sleep without 'Simple Suspend'.

Want to leave a gentle bump here. Any issues with this patch?

Kind regards,

Werner Sembach

Quoted 55 lines
---
  drivers/nvme/host/nvme.h |  5 +++++
  drivers/nvme/host/pci.c  | 16 +++++++++++++++-
  2 files changed, 20 insertions(+), 1 deletion(-)
drivers/nvme/host/nvme.h +5 -0
index f35647c470afa..74f74b459f5fa 100644
@@ -156,6 +156,11 @@ enum nvme_quirks {
  	 * No temperature thresholds for channels other than 0 (Composite).
  	 */
  	NVME_QUIRK_NO_SECONDARY_TEMP_THRESH	= (1 << 19),
+
+	/*
+	 * Disables simple suspend/resume path.
+	 */
+	NVME_QUIRK_FORCE_NO_SIMPLE_SUSPEND	= (1 << 20),
  };
 
  /*
drivers/nvme/host/pci.c +15 -1
index 347cb5daebc3c..b4e6f0d21a44c 100644
@@ -2903,6 +2903,18 @@ static unsigned long check_vendor_combination_bug(struct pci_dev *pdev)
  		if ((dmi_match(DMI_BOARD_VENDOR, "LENOVO")) &&
  		     dmi_match(DMI_BOARD_NAME, "LNVNB161216"))
  			return NVME_QUIRK_SIMPLE_SUSPEND;
+	} else if (pdev->vendor == 0x2646 && (pdev->device == 0x2263 ||
+		   pdev->device == 0x500f)) {
+		/*
+		 * Exclude some Kingston NV1 and A2000 devices from
+		 * NVME_QUIRK_SIMPLE_SUSPEND. Do a full suspend to save a
+		 * lot fo energy with s2idle sleep on some TUXEDO platforms.
+		 */
+		if (dmi_match(DMI_BOARD_NAME, "NS5X_NS7XAU") ||
+		    dmi_match(DMI_BOARD_NAME, "NS5x_7xAU") ||
+		    dmi_match(DMI_BOARD_NAME, "NS5x_7xPU") ||
+		    dmi_match(DMI_BOARD_NAME, "PH4PRX1_PH6PRX1"))
+			return NVME_QUIRK_FORCE_NO_SIMPLE_SUSPEND;
  	}
 
  	return 0;
@@ -2933,7 +2945,9 @@ static struct nvme_dev *nvme_pci_alloc_dev(struct pci_dev *pdev,
  	dev->dev = get_device(&pdev->dev);
 
  	quirks |= check_vendor_combination_bug(pdev);
-	if (!noacpi && acpi_storage_d3(&pdev->dev)) {
+	if (!noacpi &&
+	    !(quirks & NVME_QUIRK_FORCE_NO_SIMPLE_SUSPEND) &&
+	    acpi_storage_d3(&pdev->dev)) {
  		/*
  		 * Some systems use a bios work around to ask for D3 on
  		 * platforms that support kernel managed suspend.
#2 lore raw

What does save a lot of power mean, and why does it only apply to some platforms? I suspect the more likely case is that APST is simply completely broken on this device?

#3 lore raw

Am 06.12.23 um 10:24 schrieb Christoph Hellwig:

Quoted 3 lines

What does save a lot of power mean, and why does it only apply to some platforms? I suspect the more likely case is that APST is simply completely broken on this device?

Hello,

power consumption in s0ix varies depending on mainboard and NVMe SSD. Some examples:

Board,           nvme(id), mW w/o patch, mW with patch, kernel
NS5X_SN7XAU,     0x500f,   2691mW,       487mW,         6.2.0
NS5X_SN7XAU,     0x2263,   2262mW,       741mW,         6.2.0
PH4PRX1_PH6PRX1, 0x2263,   3042mW,       760mW,         6.2.0
NS5X_SN7XPU,     0x500f,   3135mW,       779mW,         6.5.0

All measurements were taken with the battery disconnected and a modified adapter plug.

I don't have a deeper insight into APST to judge whether something is not working. But all other NVMe SSDs we have tested sleep with a power consumption below 2W (Samsung 970 Evo, Samsung 980, Corsair MP600 PRO NH, other Kingston NV1 with ID != 0x500f).

We originally noticed that newer Kingston NV1 (with id 500f) led to higher standby consumption. With some trial and error I came to the conclusion that the problem lies with 'Simple Suspend'.

Kind regards,

Georg Gottleuber