John W. Linville wrote:
> Certain (SGI?) ia64 boxes object to having their PCI BARs
> restored unless absolutely necessary. This patch restricts calling
> pci_restore_bars from pci_set_power_state unless the current state
> is PCI_UNKNOWN, the actual (i.e. physical) state of the device is
> PCI_D3hot, and the device indicates that it will lose its configuration
> when transitioning to PCI_D0.
>
> Signed-off-by: John W. Linville <linville@xxxxxxxxxxxxx>
> ---
> Many thanks to Keith Owens <kaos@xxxxxxx> for a) narrowing-down the
> problem; and, b) quickly testing the fix and reporting the results.
>
> drivers/pci/pci.c | 16 ++++++++++++----
> 1 files changed, 12 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -309,17 +309,25 @@ pci_set_power_state(struct pci_dev *dev,
>
> pci_read_config_word(dev, pm + PCI_PM_CTRL, &pmcsr);
>
> - /* If we're in D3, force entire word to 0.
> + /* 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_D3hot) {
> - if (!(pmcsr & PCI_PM_CTRL_NO_SOFT_RESET))
> + switch (dev->current_state) {
> + case PCI_UNKNOWN: /* Boot-up */
> + if ((pmcsr & PCI_PM_CTRL_STATE_MASK) == PCI_D3hot
> + && !(pmcsr & PCI_PM_CTRL_NO_SOFT_RESET))
> need_restore = 1;
> + /* Fall-through: force to D0 */
> + case PCI_D3hot:
> + case PCI_D3cold:
> + case PCI_POWER_ERROR:
> pmcsr = 0;
> - } else {
> + break;
> + default:
> pmcsr &= ~PCI_PM_CTRL_STATE_MASK;
> pmcsr |= state;
> + break;
This seems like it will break a lot of stuff that -does- need the BARs
restored when resuming from D3.
Jeff
