On Wed, Mar 23, 2016 at 04:01:58PM -0600, Jon Derrick wrote: > This patch adds a hint, no_acpi, to the pci_host_bridge struct which > informs the hotplug driver to not try and release the host bridge from > acpi. > > The VMD driver creates a root port which does not have an acpi entry. > This patch will allow the hotplug driver to bypass acpi release when > enumerating the VMD root port as a hotplug slot. I want to add a little context to this for the motivation for this implementation. Currently we have to set kernel parameter "pcie_ports=native" for pciehp to work on these domains since they are not known to ACPI. We don't want to force a system wide PCIe parameter for something specific to a subset of domains. If the implementation here sounds reasonable, I suggest having a helper API function in pci to check for the host bridge requested acpi disabled. > arch/x86/pci/vmd.c | 4 ++++ > drivers/pci/pcie/portdrv_acpi.c | 5 +++++ > include/linux/pci.h | 1 + > 3 files changed, 10 insertions(+) > > diff --git a/arch/x86/pci/vmd.c b/arch/x86/pci/vmd.c > index 7792aba..a196be3 100644 > --- a/arch/x86/pci/vmd.c > +++ b/arch/x86/pci/vmd.c > @@ -531,6 +531,7 @@ static int vmd_find_free_domain(void) > static int vmd_enable_domain(struct vmd_dev *vmd) > { > struct pci_sysdata *sd = &vmd->sysdata; > + struct pci_host_bridge *host_bridge; > struct resource *res; > u32 upper_bits; > unsigned long flags; > @@ -609,6 +610,9 @@ static int vmd_enable_domain(struct vmd_dev *vmd) > return -ENODEV; > } > > + host_bridge = to_pci_host_bridge(vmd->bus->bridge); > + host_bridge->no_acpi = true; > + > vmd_attach_resources(vmd); > vmd_setup_dma_ops(vmd); > dev_set_msi_domain(&vmd->bus->dev, vmd->irq_domain); > diff --git a/drivers/pci/pcie/portdrv_acpi.c b/drivers/pci/pcie/portdrv_acpi.c > index b4d2894..0e6fa69 100644 > --- a/drivers/pci/pcie/portdrv_acpi.c > +++ b/drivers/pci/pcie/portdrv_acpi.c > @@ -35,12 +35,17 @@ > int pcie_port_acpi_setup(struct pci_dev *port, int *srv_mask) > { > struct acpi_pci_root *root; > + struct pci_host_bridge *host_bridge; > acpi_handle handle; > u32 flags; > > if (acpi_pci_disabled) > return 0; > > + host_bridge = pci_find_host_bridge(port->bus); > + if (host_bridge && host_bridge->no_acpi) > + return 0; > + > handle = acpi_find_root_bridge_handle(port); > if (!handle) > return -EINVAL; > diff --git a/include/linux/pci.h b/include/linux/pci.h > index 15f466e..a6958e9b 100644 > --- a/include/linux/pci.h > +++ b/include/linux/pci.h > @@ -407,6 +407,7 @@ struct pci_host_bridge { > void (*release_fn)(struct pci_host_bridge *); > void *release_data; > unsigned int ignore_reset_delay:1; /* for entire hierarchy */ > + unsigned int no_acpi:1; /* bypasses acpi release */ > /* Resource alignment requirements */ > resource_size_t (*align_resource)(struct pci_dev *dev, > const struct resource *res, > -- > 1.8.3.1 -- 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