Hi all,
Defining PCI_IOBASE for MIPS ralink platform results in resource handling working
but the addresses generated for IO access being wrong, because the iomap tries to
ioremap it to a fixed virtual address (PCI_IOBASE), which can't work for KSEG1
addresses. To get it working this way, we would need to put PCI_IOBASE somewhere
into KSEG2, which will create TLB entries for IO addresses, which most of the
time isn't needed on MIPS because of access via KSEG1. Instead of doing that and
taking into account that we need to get a valid IO address from 'pci_address_to_pio'
and ralink platforms have IO addresses higher than 0xffff, the following approach
will be preferred to get expected working behaviour from PCI core APIs and pci
drivers working together:
1) Avoid to define PCI_IOBASE.
2) Set IO_SPACE_LIMIT to 0x1fffffff which is a valid range for this SoCs.
3) Return zero from 'pci_remap_iospace' if PCI_IOBASE is not defined.
3) Set ioport_resource end limit to this new IO_SPACE_LIMIT.
Doing in this way we end up with a properly working PCI IO in ralink SoCs.
These changes metioned above are in the three patches included in this series.
Thanks in advance for your time and comments.
Changes in v2 (only PATCH v2 2/3 afected):
- Instead of avoid to call 'devm_pci_remap_iospace' fix 'pci_remap_iospace'
to return zero for PCI_IOBASE not defined architectures.
Bjorn, I don't know if I should add any kind of 'Fixes' tag for this or
what is the way of handle this kind of changes inside the PCI tree. Thanks
in advance for clarification.
Sergio Paracuellos (3):
MIPS: ralink: don't define PC_IOBASE but increase IO_SPACE_LIMIT
PCI: fix 'pci_remap_iospace' for architectures with PCI_IOBASE not
defined
staging: mt7621-pci: set end limit for 'ioport_resource'
arch/mips/include/asm/mach-ralink/spaces.h | 4 +---
drivers/pci/pci.c | 12 ++++++------
drivers/staging/mt7621-pci/pci-mt7621.c | 2 ++
3 files changed, 9 insertions(+), 9 deletions(-)
--
2.25.1
