Re: [PATCH] pci: Added flag to pci_host_bridge to hint not to use acpi

From: Jon Derrick <jonathan.derrick@intel.com>
To: Bjorn Helgaas <helgaas@kernel.org>
Cc: Keith Busch <keith.busch@intel.com>,
	linux-pci@vger.kernel.org,
	"Rafael J. Wysocki" <rjw@rjwysocki.net>,
	linux-acpi@vger.kernel.org
Subject: Re: [PATCH] pci: Added flag to pci_host_bridge to hint not to use acpi
Date: Thu, 7 Apr 2016 14:12:16 -0600	[thread overview]
Message-ID: <20160407201216.GA9986@localhost.localdomain> (raw)
In-Reply-To: <20160407174224.GE8780@localhost>

Hi Bjorn,

First, thanks for the review

> The spec (PCI Firmware spec 3.0, sec 4.5) says _OSC is a child of an
> ACPI device and applies to that device.  In your case, I think there
> is no ACPI PNP0A0x device for the VMD host bridge, and hence there is
> no _OSC method for it.  There's no ACPI handle for the VMD bridge, so
> I think pcie_port_acpi_setup() returns -EINVAL, which causes
> get_port_device_capability() to give up and decide that the port
> can't support *any* PCIe services.
> 
> I think that's the real problem: the fact that there's no ACPI device
> corresponding to the bridge should not be an error.  It should just
> mean the platform doesn't have an opportunity to limit what services
> the kernel can manage.  I think pcie_port_acpi_setup() should return 0
> in all cases.  Can you try that and see if it fixes the problem?
> 

I believe what you are suggesting is:

        handle = acpi_find_root_bridge_handle(port);
        if (!handle)
-               return -EINVAL;
+               return 0;

        root = acpi_pci_find_root(handle);
        if (!root)
-               return -ENODEV;
+               return 0;

This works for me. I will follow up with a proper patch.