Skip to contentSkip to conversation

PCI PM: call platform helpers properly

PATCH In review Adam Belay 2 participants 4 messages Jun 5, 2006 – Jun 7, 2006 PCI

lore
Share & tools
lore

Adam Belay

author
#0 lore raw

PATCH 6/9 PCI PM: call platform helpers properly

It doesn't make much sense to call platform helpers (e.g. ACPI) after the PCI power state has been set when returning to D0. This patch ensures that platform helpers and PCI PM registers are called in the correct order.

---
 pm.c |   68 +++++++++++++++++++++++++++++++++++--------------------------------
 1 file changed, 36 insertions(+), 32 deletions(-)
drivers/pci/pm.c +36 -32
diff -urN a/drivers/pci/pm.c b/drivers/pci/pm.c
@@ -111,6 +111,36 @@
 
 int (*platform_pci_set_power_state)(struct pci_dev *dev, pci_power_t t);
 
+static void __pci_set_power_state(struct pci_dev *dev, pci_power_t state)
+{
+	struct pci_dev_pm *pm = dev->pm;
+	u16 pmcsr;
+
+	pci_read_config_word(dev, pm->pm_offset + PCI_PM_CTRL, &pmcsr);
+
+	/* If we're (effectively) in D3, force entire word to 0.
+	 * This doesn't affect PME_Status, disables PME_En, and
+	 * sets PowerState to 0.
+	 */
+	if (dev->current_state == PCI_D3) {
+		pmcsr = 0;
+	} else {
+		pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
+		pmcsr |= state;
+	}
+
+	/* enter specified state */
+	pci_write_config_word(dev, pm->pm_offset + PCI_PM_CTRL, pmcsr);
+
+	/* Mandatory power management transition delays */
+	/* see PCI PM 1.1 5.6.1 table 18 */
+	if (state == PCI_D3 || dev->current_state == PCI_D3)
+		msleep(10);
+	else if (state == PCI_D2 || dev->current_state == PCI_D2)
+		udelay(200);
+
+}
+
 /**
  * pci_set_power_state - Set the power state of a PCI device
  * @dev: PCI device to be suspended
@@ -127,7 +157,6 @@
  */
 int pci_set_power_state(struct pci_dev *dev, pci_power_t state)
 {
-	u16 pmcsr;
 	struct pci_dev_pm *pm = dev->pm;
 
 	if (!pm)
@@ -150,39 +179,14 @@
 	else if (state == PCI_D2 && !pm->D2_supported)
 		return -EIO;
 
-	pci_read_config_word(dev, pm->pm_offset + PCI_PM_CTRL, &pmcsr);
-
-	/* If we're (effectively) in D3, force entire word to 0.
-	 * This doesn't affect PME_Status, disables PME_En, and
-	 * sets PowerState to 0.
-	 */
-	switch (dev->current_state) {
-	case PCI_D0:
-	case PCI_D1:
-	case PCI_D2:
-		pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
-		pmcsr |= state;
-		break;
-	default:
-		pmcsr = 0;
-		break;
-	}
-
-	/* enter specified state */
-	pci_write_config_word(dev, pm->pm_offset + PCI_PM_CTRL, pmcsr);
+	/* call platform helpers (e.g. ACPI) before restoring power */
+	if (state == PCI_D0 && platform_pci_set_power_state)
+		platform_pci_set_power_state(dev, state);
 
-	/* Mandatory power management transition delays */
-	/* see PCI PM 1.1 5.6.1 table 18 */
-	if (state == PCI_D3 || dev->current_state == PCI_D3)
-		msleep(10);
-	else if (state == PCI_D2 || dev->current_state == PCI_D2)
-		udelay(200);
+	__pci_set_power_state(dev, state);
 
-	/*
-	 * Give firmware a chance to be called, such as ACPI _PRx, _PSx
-	 * Firmware method after natice method ?
-	 */
-	if (platform_pci_set_power_state)
+	/* otherwise, call platform helpers after removing power */
+	if (state != PCI_D0 && platform_pci_set_power_state)
 		platform_pci_set_power_state(dev, state);
 
 	dev->current_state = state;
#1 lore raw

drivers/pci/pm.c line 184

Quoted 5 lines
drivers/pci/pm.c +3 -1
-	pci_write_config_word(dev, pm->pm_offset + PCI_PM_CTRL, pmcsr);
+	/* call platform helpers (e.g. ACPI) before restoring power */
+	if (state == PCI_D0 && platform_pci_set_power_state)
+		platform_pci_set_power_state(dev, state);

I think the platform helper need to be able to either return from the function directly or at least you should read back the state before continuing. The platform helper may call into firmware which might handle the complete state transition. Thus, when you come back from it, you probably need to check the device-state, or provide return codes for the platform helper to affect the code flow of the rest of the function.

Ben.

author
#2 lore raw
Quoted 11 lines · Benjamin Herrenschmidt

drivers/pci/pm.c line 184

Quoted 4 lines
drivers/pci/pm.c +3 -1
-	pci_write_config_word(dev, pm->pm_offset + PCI_PM_CTRL, pmcsr);
+	/* call platform helpers (e.g. ACPI) before restoring power */
+	if (state == PCI_D0 && platform_pci_set_power_state)
+		platform_pci_set_power_state(dev, state);

I think the platform helper need to be able to either return from the function directly or at least you should read back the state before continuing. The platform helper may call into firmware which might handle the complete state transition. Thus, when you come back from it, you probably need to check the device-state, or provide return codes for the platform helper to affect the code flow of the rest of the function.

Hmm, do you know of any example firmware scenarios that make the entire state transition? We might need a separate callback. I think the changes in this patch are at least an improvement over the current behavior, especially for ACPI PM functions.

Thanks,
Adam

#3 lore raw
Quoted 4 lines · Adam Belay

Hmm, do you know of any example firmware scenarios that make the entire state transition? We might need a separate callback. I think the changes in this patch are at least an improvement over the current behavior, especially for ACPI PM functions.

Well, do we know precisely what ACPI does/needs for example here ?

Ben.