Re: [PATCH 00/17] x86, ia64 NUMA cleanup

[Bjorn Helgaas]

[+cc Tony]

On Wed, Jan 29, 2014 at 11:17 AM, Bjorn Helgaas <bhelgaas@xxxxxxxxxx> wrote:
> This is all cleanup: it adds no new functionality and should make no
> functional difference at all except a printk change or two.
>
> The main ideas are:
>
>   - Clean up the x86 PCI bus scanning interfaces.  We had:
>       struct pci_bus *pci_scan_bus_on_node(int bus, struct pci_ops *, int node)
>       struct pci_bus *pcibios_scan_root(int bus)
>       struct pci_bus *pci_scan_bus_with_sysdata(int bus)
>       void pcibios_scan_specific_bus(int bus)
>     These were all basically the same, so I kept only pcibios_scan_root().
>
>   - Get rid of mp_bus_to_node[].  This was basically a way of passing node
>     info from the AMD "native host bridg driver" (amd_bus.c) into the PCI
>     core.  It was clunky and inconsistent between x86-64 and i386.  I
>     replaced it with x86_pci_root_bus_node(), analogous to the existing
>     x86_pci_root_bus_resources().
>
>   - Remove _PXM knowledge from the PCI host bridge, IOMMU, and APIC code.
>     They really only care about the NUMA node number, not the raw _PXM
>     value.
>
> ---
>
> Bjorn Helgaas (17):
>       x86/PCI: Drop pcibios_scan_root() check for bus already scanned
>       x86/PCI: Use pcibios_scan_root() instead of pci_scan_bus_with_sysdata()
>       x86/PCI: Use pcibios_scan_root() instead of pci_scan_bus_on_node()
>       x86/PCI: Merge pci_scan_bus_on_node() into pcibios_scan_root()
>       x86/PCI: Drop return value of pcibios_scan_root()
>       x86/PCI: Add x86_pci_root_bus_node() to look up NUMA node from PCI bus
>       x86/PCI: Use x86_pci_root_bus_node() instead of get_mp_bus_to_node()
>       x86/PCI: Remove mp_bus_to_node[], set_mp_bus_to_node(), get_mp_bus_to_node()
>       x86/PCI: Remove unnecessary list_empty(&pci_root_infos) check
>       x86/PCI: Use NUMA_NO_NODE, not -1, for unknown node
>       x86/PCI: Remove acpi_get_pxm() usage
>       ia64 / sba_iommu: Use NUMA_NO_NODE, not MAX_NUMNODES, for unknown node
>       ia64: Remove acpi_get_pxm() usage
>       ACPI / numa: Fix acpi_get_node() prototype
>       ACPI / numa: Simplify acpi_get_node() style
>       ACPI / numa: Make __acpi_map_pxm_to_node(), acpi_get_pxm() static
>       ACPI / numa: Use __weak, not the gcc-specific version
>
>
>  arch/ia64/hp/common/sba_iommu.c |   32 +++-------
>  arch/ia64/include/asm/pci.h     |    2 -
>  arch/ia64/kernel/acpi.c         |   28 ++-------
>  arch/ia64/pci/pci.c             |   10 +--
>  arch/x86/include/asm/pci.h      |    7 --
>  arch/x86/include/asm/topology.h |   14 ----
>  arch/x86/pci/acpi.c             |   29 ++-------
>  arch/x86/pci/amd_bus.c          |   10 ---
>  arch/x86/pci/bus_numa.c         |   13 +++-
>  arch/x86/pci/common.c           |  128 +++++----------------------------------
>  arch/x86/pci/fixup.c            |    6 +-
>  arch/x86/pci/irq.c              |    6 --
>  arch/x86/pci/legacy.c           |    4 -
>  arch/x86/pci/numaq_32.c         |    6 +-
>  arch/x86/pci/visws.c            |    4 +
>  drivers/acpi/numa.c             |   16 ++---
>  include/acpi/acpi_numa.h        |    1
>  include/linux/acpi.h            |    9 +--
>  18 files changed, 75 insertions(+), 250 deletions(-)

I added this series to my "next" branch for v3.14.  Tony, let me know
if you see any ia64 issues.

Bjorn
--
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/