lkml.org 
[lkml]   [2012]   [Apr]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [RFC PATCH] PCIe: Add PCIe runtime D3cold support
Date
On Thursday, April 19, 2012, huang ying wrote:
> On Thu, Apr 19, 2012 at 4:51 AM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> > On Wednesday, April 18, 2012, huang ying wrote:
> >> On Wed, Apr 18, 2012 at 5:10 AM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> >> > On Tuesday, April 17, 2012, huang ying wrote:
> >> >> On Tue, Apr 17, 2012 at 5:30 AM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> >> >> >> >> + return 0;
> >> >> >> >> +}
> >> >> >> >> +
> >> >> >> >> +static int pcie_port_runtime_resume(struct device *dev)
> >> >> >> >> +{
> >> >> >> >> + struct pci_dev *pdev = to_pci_dev(dev);
> >> >> >> >> +
> >> >> >> >> + pci_restore_state(pdev);
> >> >> >> >> + if (pdev->runtime_d3cold)
> >> >> >> >> + msleep(100);
> >> >> >> >
> >> >> >> > What's _that_ supposed to do?
> >> >> >>
> >> >> >> When resume from d3cold, PCIe main link will be powered on again, it
> >> >> >> will take quite some time before the main link go into L0 state.
> >> >> >> Otherwise, accessing devices under the port may return wrong result.
> >> >> >
> >> >> > OK, but this is generic code and as per the standard the link should have been
> >> >> > reestablished at this point already.
> >> >> >
> >> >> > Please don't put some nonstandard-platform-specific quirks like this into
> >> >> > code that's supposed to handle _every_ PCIe system.
> >> >>
> >> >> After checking PCIe spec, I found that the 100ms here has its standard origin :)
> >> >>
> >> >> In PCI Express Base Specification Revision 2.0:
> >> >>
> >> >> Section 6.6.1 Conventional Reset
> >> >>
> >> >> "
> >> >> To allow components to perform internal initialization, system
> >> >> software must wait for at least
> >> >> 100 ms from the end of a Conventional Reset of one or more devices
> >> >> before it is permitted to
> >> >> issue Configuration Requests to those devices
> >> >> "
> >> >>
> >> >> But I think we should move the 100ms delay here to PCIe bus code or
> >> >> PCIe/ACPI code, because that is needed by all PCIe devices for D3cold
> >> >> support.
> >> >
> >> > I think it should be sufficient to wait for the PME message to arrive at
> >> > the root port (which will cause the PME interrupt to appear), at which
> >> > point the device that sent it should be able to receive configuration
> >> > requests.
> >>
> >> For remote wake up, it is sufficient. But for host wake up, we still
> >> need to wait 100ms.
> >
> > Yes, we do.
> >
> >> > At this point, I need to konw what exactly happens when the GPE is triggered
> >> > by WAKE#.
> >>
> >> - Lxx handler will be executed
> >> - in Lxx handler, Notify the ACPI handle PCIe port
> >> - Linux has registered a handler for the ACPI handle of PCIe port, in
> >> the handler, turn on _PR0 and execute _PS0, which will power on the
> >> link.
> >
> > But the handler we have is not the handler we want here.
> >
> > In fact, there are two handlers, pci_acpi_wake_bus() and pci_acpi_wake_dev()
> > and they only do useful things for ACPI_NOTIFY_DEVICE_WAKE. Is that the
> > event type we receive from that _Lxx?
>
> I check the DSDT, in _Lxx, there is
>
> Notify (\_SB.PCI0.RP03, 0x02)
>
> That is, the event type is ACPI_NOTIFY_DEVICE_WAKE.
>
> > Even if so, these routines don't seem to be suitable to handle the case at hand.
>
> Yes. Maybe add a flag named like "come_from_d3cold", and if
> come_from_d3cold == true, resume the dev itself without checking pme
> bits, because the PCIe main link is not available now.

Actually, I think we can do the full resume (however with the 100 ms wait)
in that case, because we're going to resume the device shortly anyway.

The PME would only be useful as a kind of handshake with the device, so we
know we can access its registers, but as you pointed that out, we need the
100 ms delay in the host wakeup case anyway, so perhaps it's better to make
remote wakeup and host wakeup behave identically in that respect.

Thanks,
Rafael


\
 
 \ /
  Last update: 2012-04-19 14:35    [W:0.556 / U:0.728 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site