This is what I've collected on my pci/ecam branch so far. This is
preliminary and probably doesn't even build since I made several
changes, but at least it's a strawman. I called it v10 to keep it
separate from the previous postings of the various pieces.
ACPI: Add acpi_resource_consumer() to find device that claims a resource
PCI: Search ACPI namespace to ensure ECAM space is reserved
x86/PCI: Use acpi_resource_consumer() to search ACPI namespace for MMCFG
acpi_resource_consumer() takes a resource and returns an
acpi_device that has _CRS that contains the resource. The idea is
that we call this on a resource from MCFG. We *should* find a
PNP0C02 or other device that reserves that resource. On x86 we
already do this in a more hand-coded way. I probably wouldn't
merge the x86 patch for v4.10 but I included it here as an
example.
I don't think it's worth trying to fabricate ACPI devices or _CRS
resources to compensate for firmware that doesn't describe
everything. This check will emit a warning if the MCFG region is
not reserved, and that's probably enough to motivate firmware
fixes for the next round of hardware. The ECAM code itself does
reserve the region, so that part will be safe from other users.
Unreported non-ECAM register space is still a landmine and there's
no real way to look for it.
arm64: PCI: Manage controller-specific data on per-controller basis
Tomasz's fix for pci_acpi_scan_root(), unchanged.
PCI/ACPI: Extend pci_mcfg_lookup() to return ECAM config accessors
PCI/ACPI: Check for platform-specific MCFG quirks
Tomasz's quirk infrastructure. I put this under
CONFIG_PCI_QUIRKS.
PCI/ACPI: Provide acpi_get_rc_resources() for ARM64 platform
Dongdong's interface to look up a _HID with a _UID matching the
segment. I don't have an opinion on using _UID yet. It's
encapsulated so it could be changed easily. I put this under
CONFIG_PCI_QUIRKS && CONFIG_ARM64.
PCI: Add MCFG quirks for Qualcomm QDF2432 host controller
Christopher's Qualcomm quirks. Basically unchanged except to put
the quirk ops under CONFIG_ACPI and CONFIG_PCI_QUIRKS.
PCI: Add MCFG quirks for HiSilicon Hip05/06/07 host controllers
Dongdong's HiSilicon quirks. Here's where it gets interesting. I
moved this to the existing pcie-hisi.c instead of adding
pcie-hisi-acpi.c. I changed the Makefile so we always build
pcie-hisi.c on ARM64. I added ifdefs so we get the quirk code if
CONFIG_ACPI and CONFIG_PCI_QUIRKS and we get the original platform
driver if CONFIG_PCI_HISI. It's possible to have both, and if we
process the MCFG quirk we get the quirk code.
I'm confused about why the quirk accessors are so much different
than the original accessors. hisi_pcie_acpi_rd_conf() looks much
different than hisi_pcie_cfg_read(). The original driver claims
Hipxx only supports 32-bit config accesses, but the quirk
accessors don't enforce that.
PCI: thunder-pem: Factor out resource lookup
PCI: Add MCFG quirks for Cavium ThunderX pass2.x host controller
PCI: Add MCFG quirks for Cavium ThunderX pass1.x host controller
Tomasz's ThunderX quirks. I restructured the PEM init I could do
the same ifdef tricks as in hisi. The quirks make a lot more
sense here -- they use the same accessors as the original platform
driver. Only the initialization is different.
The Makefile looks a little strange:
obj-$(CONFIG_ARM64) += pcie-hisi.o
obj-$(CONFIG_ARM64) += pci-thunder-ecam.o
obj-$(CONFIG_ARM64) += pci-thunder-pem.o
and the ifdefs inside those files are a little unwieldy. But I think
they do accomplish the goals of:
- Adding no new files,
- Including the MCFG quirks whenever CONFIG_ACPI and
CONFIG_PCI_QUIRKS are enabled, regardless of the driver-specific
config,
- Including the original platform drivers when they are specifically
enabled, and
- Sharing the same code for the platform driver and the quirks
(except for HiSi, which I don't understand)
I started looking at the Duc's X-Gene quirks, but I haven't wrapped my
head around those yet. I'm going to push this branch to get build
testing and so you can comment on it.
If by some miracle it builds and you test it, please collect the dmesg
log and /proc/iomem contents.
---
Bjorn Helgaas (5):
ACPI: Add acpi_resource_consumer() to find device that claims a resource
PCI: Search ACPI namespace to ensure ECAM space is reserved
x86/PCI: Use acpi_resource_consumer() to search ACPI namespace for MMCFG
PCI: Add MCFG quirks for HiSilicon Hip05/06/07 host controllers
PCI: thunder-pem: Factor out resource lookup
Christopher Covington (1):
PCI: Add MCFG quirks for Qualcomm QDF2432 host controller
Dongdong Liu (1):
PCI/ACPI: Provide acpi_get_rc_resources() for ARM64 platform
Tomasz Nowicki (5):
arm64: PCI: Manage controller-specific data on per-controller basis
PCI/ACPI: Extend pci_mcfg_lookup() to return ECAM config accessors
PCI/ACPI: Check for platform-specific MCFG quirks
PCI: Add MCFG quirks for Cavium ThunderX pass2.x host controller
PCI: Add MCFG quirks for Cavium ThunderX pass1.x host controller
arch/arm64/kernel/pci.c | 34 +++----
arch/x86/pci/mmconfig-shared.c | 69 ++-------------
drivers/acpi/pci_mcfg.c | 165 ++++++++++++++++++++++++++++++++++-
drivers/acpi/resource.c | 57 ++++++++++++
drivers/pci/ecam.c | 31 +++++++
drivers/pci/host/Makefile | 6 +
drivers/pci/host/pci-thunder-ecam.c | 9 ++
drivers/pci/host/pci-thunder-pem.c | 88 ++++++++++++++-----
drivers/pci/host/pcie-hisi.c | 95 ++++++++++++++++++++
drivers/pci/pci-acpi.c | 71 +++++++++++++++
drivers/pci/pci.h | 4 +
include/linux/acpi.h | 7 +
include/linux/pci-acpi.h | 4 +
include/linux/pci-ecam.h | 7 +
14 files changed, 537 insertions(+), 110 deletions(-)
--
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
