1
linux/arch/riscv/include/asm/pci.h

34 lines
728 B
C
Raw Permalink Normal View History

/* SPDX-License-Identifier: GPL-2.0-only */
/*
* Copyright (C) 2016 SiFive
*/
#ifndef _ASM_RISCV_PCI_H
#define _ASM_RISCV_PCI_H
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/dma-mapping.h>
#include <asm/io.h>
RISC-V: PCI: Avoid handing out address 0 to devices For RISC-V platforms we permit assigning addresses from 0 to PCI devices, both in the memory and the I/O bus space, and we happily do so if there is no conflict, e.g.: pci 0000:07:00.0: BAR 0: assigned [io 0x0000-0x0007] pci 0000:07:00.1: BAR 0: assigned [io 0x0008-0x000f] pci 0000:06:01.0: PCI bridge to [bus 07] pci 0000:06:01.0: bridge window [io 0x0000-0x0fff] (with the SiFive HiFive Unmatched RISC-V board and a dual serial port option card based on the OxSemi OXPCIe952 device wired for the legacy UART mode). Address 0 is treated specially however in many places, for example in `pci_iomap_range' and `pci_iomap_wc_range' we require that the start address is non-zero, and even if we let such an address through, then individual device drivers could reject a request to handle a device at such an address, such as in `uart_configure_port'. Consequently given devices configured as shown above only one is actually usable: Serial: 8250/16550 driver, 4 ports, IRQ sharing disabled serial 0000:07:00.0: enabling device (0000 -> 0001) serial: probe of 0000:07:00.0 failed with error -12 serial 0000:07:00.1: enabling device (0000 -> 0001) serial 0000:07:00.1: detected caps 00000700 should be 00000500 0000:07:00.1: ttyS0 at I/O 0x8 (irq = 39, base_baud = 15625000) is a 16C950/954 Therefore avoid handing out address 0, by bumping the lowest address available to PCI via PCIBIOS_MIN_IO and PCIBIOS_MIN_MEM up by 4 and 16 respectively, which is the minimum allocation size for I/O and memory BARs. With this in place the system in question we have: pci 0000:07:00.0: BAR 0: assigned [io 0x1000-0x1007] pci 0000:07:00.1: BAR 0: assigned [io 0x1008-0x100f] pci 0000:06:01.0: PCI bridge to [bus 07] pci 0000:06:01.0: bridge window [io 0x1000-0x1fff] and then devices work correctly: Serial: 8250/16550 driver, 4 ports, IRQ sharing disabled serial 0000:07:00.0: enabling device (0000 -> 0001) serial 0000:07:00.0: detected caps 00000700 should be 00000500 0000:07:00.0: ttyS0 at I/O 0x1000 (irq = 38, base_baud = 15625000) is a 16C950/954 serial 0000:07:00.1: enabling device (0000 -> 0001) serial 0000:07:00.1: detected caps 00000700 should be 00000500 0000:07:00.1: ttyS1 at I/O 0x1008 (irq = 39, base_baud = 15625000) is a 16C950/954 Especially I/O space ranges are particularly valuable, because bridges only decode bits from 12 up and consequently where 16-bit addressing is in effect, as few as 16 separate ranges can be assigned to individual buses only, however a generic change to avoid handing out address 0 only has turned out controversial as per the discussion referred via the link below. Conversely sorting this out in platform code has been standard practice since forever to avoid a clash with legacy devices subtractively decoded by the southbridge where present. This can be revised should a generic solution be adopted sometime. Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk> Link: https://lore.kernel.org/r/alpine.DEB.2.21.2202260044180.25061@angie.orcam.me.uk Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2022-04-27 15:15:47 -07:00
#define PCIBIOS_MIN_IO 4
#define PCIBIOS_MIN_MEM 16
#if defined(CONFIG_PCI) && defined(CONFIG_NUMA)
static inline int pcibus_to_node(struct pci_bus *bus)
{
return dev_to_node(&bus->dev);
}
#ifndef cpumask_of_pcibus
#define cpumask_of_pcibus(bus) (pcibus_to_node(bus) == -1 ? \
cpu_all_mask : \
cpumask_of_node(pcibus_to_node(bus)))
#endif
#endif /* defined(CONFIG_PCI) && defined(CONFIG_NUMA) */
/* Generic PCI */
#include <asm-generic/pci.h>
#endif /* _ASM_RISCV_PCI_H */