In 3.2, the marked lines (***) below were added:
static int pci_platform_power_transition(struct pci_dev *dev, pci_power_t state)
{
int error;
if (platform_pci_power_manageable(dev)) {
error = platform_pci_set_power_state(dev, state);
if (!error)
pci_update_current_state(dev, state);
*** /* Fall back to PCI_D0 if native PM is not supported */
*** if (!dev->pm_cap)
*** dev->current_state = PCI_D0;
} else {
error = -ENODEV;
/* Fall back to PCI_D0 if native PM is not supported */
if (!dev->pm_cap)
dev->current_state = PCI_D0;
}
return error;
}
If the target state is PCI_D3, the new code serves to *undo* the marked code below executed by the preceding call to pci_update_current_state() when PCI PM registers are not present:
void pci_update_current_state(struct pci_dev *dev, pci_power_t state)
{
if (dev->pm_cap) {
u16 pmcsr;
pci_read_config_word(dev, dev->pm_cap + PCI_PM_CTRL, &pmcsr);
dev->current_state = (pmcsr & PCI_PM_CTRL_STATE_MASK);
} else {
*** dev->current_state = state;
}
}
This is despite the fact that platform_pci_power_manageable() indicated that the platform FW could handle it and the succeeding call to platform_pci_set_power_state() succeeded. That means that devices with FW support but no PCI PM support will be left in the wrong state.
It appears the new "if" statement should be "else if":
if (!error)
pci_update_current_state(dev, state);
*** else if (!dev->pm_cap) /* Fall back to PCI_D0 if native PM is not supported */
*** dev->current_state = PCI_D0;
Rafael concurs with me.
--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
