mbox series

[0/9] x86/pci: Work around lack of ECAM space reservation

Message ID 20231121183643.249006-1-helgaas@kernel.org (mailing list archive)
Headers show
Series x86/pci: Work around lack of ECAM space reservation | expand

Message

Bjorn Helgaas Nov. 21, 2023, 6:36 p.m. UTC 
From: Bjorn Helgaas <bhelgaas@google.com>

The ECAM space described in the ACPI MCFG table should be reserved via a
PNP0C02 _CRS method, per PCI Firmware spec (r3.3, sec 4.1.2).

Some platforms (at least Supermicro X9DRi-LN4+, HP Proliant ML30 Gen9) do
not include these PNP0C02 reservations, and at the same time, they *do*
include that space in the PNP0A03 host bridge windows, which makes it
available for assignment to PCI devices.

In this case, we treat the ECAM space as valid (after fd3a8cff4d4a
("x86/pci: Treat EfiMemoryMappedIO as reservation of ECAM space")), but
there was no actual reservation to prevent assignment to PCI devices.

This series adds that reservation to prevent assigning ECAM space to PCI
BARs, adds more debug logging, changes "MMCONFIG" to "ECAM" to match the
spec terminology, and cleans up some coding style.

I propose to merge this via the PCI tree for v6.8.

Bjorn Helgaas (9):
  x86/pci: Reserve ECAM if BIOS didn't include it in PNP0C02 _CRS
  x86/pci: Reword ECAM EfiMemoryMappedIO logging to avoid 'reserved'
  x86/pci: Add MCFG debug logging
  x86/pci: Rename 'MMCONFIG' to 'ECAM', use pr_fmt
  x86/pci: Rename acpi_mcfg_check_entry() to acpi_mcfg_valid_entry()
  x86/pci: Rename pci_mmcfg_check_reserved() to pci_mmcfg_reserved()
  x86/pci: Comment pci_mmconfig_insert() obscure MCFG dependency
  x86/pci: Return pci_mmconfig_add() failure early
  x86/pci: Reorder pci_mmcfg_arch_map() definition before calls

 arch/x86/pci/acpi.c            |   3 +
 arch/x86/pci/mmconfig-shared.c | 178 ++++++++++++++++++---------------
 arch/x86/pci/mmconfig_32.c     |   2 +-
 arch/x86/pci/mmconfig_64.c     |  42 ++++----
 4 files changed, 120 insertions(+), 105 deletions(-)