Re: [PATCH] kexec: fix pci device initialization fail after kexec (2.6.23-rc2). (Related to e1000 doesn't resume properly from standby)

[Joonwoo Park]

2007/8/6, Rafael J. Wysocki <rjw@xxxxxxx>:
> On Monday, 6 August 2007 15:42, Joonwoo Park wrote:
> > Hi.
> > I think that the pci_set_power_state() has bug.
> > The specification says that some delays is required.
>
> And they are in place, AFAICS (from drivers/pci/pci.c):
>
>        /* Mandatory power management transition delays */
>        /* see PCI PM 1.1 5.6.1 table 18 */
>        if (state == PCI_D3hot || dev->current_state == PCI_D3hot)
>                msleep(pci_pm_d3_delay);
>        else if (state == PCI_D2 || dev->current_state == PCI_D2)
>                udelay(200);
>

The problem is occurred when state is 'PCI_D0', so those codes can't cover it.
But pci pm specification 5.4.1 says that when programmed to D0 the
equivalent of a warm reset, delay for the duration of the D3hot to D0
Uninitialized state
transition (10ms) to pci signal drivers remain disabled is required.

> > Simon, please can I know whether it works to you?
> > It is seems to the things are related.
> >
> > Joonwoo Park.
> >
> > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> > index 37c00f6..9f78064 100644
> > --- a/drivers/pci/pci.c
> > +++ b/drivers/pci/pci.c
> > @@ -493,8 +493,14 @@ pci_set_power_state(struct pci_dev *dev, pci_power_t state)
> >    * restore at least the BARs so that the device will be
> >    * accessible to its driver.
> >    */
> > - if (need_restore)
> > + if (need_restore) {
> > +   /* The specification also says that "Must ensure that all of
> > +    * its PCI signal drivers remain disabled for the duration of
> > +    * the D3hot to D0 Uninitialized state transition".
> > +    */
> > +   msleep(10);
>
> It's too late for the delay.
>

How about this?

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 37c00f6..974dd04 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -471,6 +471,8 @@ pci_set_power_state(struct pci_dev *dev, pci_power_t state)
                msleep(pci_pm_d3_delay);
        else if (state == PCI_D2 || dev->current_state == PCI_D2)
                udelay(200);
+      else if (state == PCI_D0)
+              msleep(10);

        /*
         * Give firmware a chance to be called, such as ACPI _PRx, _PSx


> >     pci_restore_bars(dev);
> > + }
> >
> >   return 0;
> >  }
>
> Greetings,
> Rafael
>
>
> --
> "Premature optimization is the root of all evil." - Donald Knuth
>

Thanks,
Joonwoo Park.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/