linux_dsm_epyc7002/arch/powerpc/kernel/pci_64.c
Sam Bobroff 3f068aae7a powerpc/64: Adjust order in pcibios_init()
The pcibios_init() function for PowerPC 64 currently calls
pci_bus_add_devices() before pcibios_resource_survey(). This means
that at boot time, when the pcibios_bus_add_device() hooks are called
by pci_bus_add_devices(), device resources have not been allocated and
they are unable to perform EEH setup, so a separate pass is needed.

This patch adjusts that order so that it will become possible to
consolidate the EEH setup work into a single location.

The only functional change is to execute pcibios_resource_survey()
(excepting ppc_md.pcibios_fixup(), see below) before
pci_bus_add_devices() instead of after it.

Because pcibios_scan_phb() and pci_bus_add_devices() are called
together in a loop, this must be broken into one loop for each call.
Then the call to pcibios_resource_survey() is moved up in between
them. This changes the ordering but because pcibios_resource_survey()
also calls ppc_md.pcibios_fixup(), that call is extracted out into
pcibios_init() to where pcibios_resource_survey() was, so that it is
not moved.

The only other caller of pcibios_resource_survey() is the PowerPC 32
version of pcibios_init(), and therefore, that is modified to call
ppc_md.pcibios_fixup() right after pcibios_resource_survey() so that
there is no functional change there at all.

The re-arrangement will cause very few side-effects because at this
stage in the boot, pci_bus_add_devices() does very little:
- pci_create_sysfs_dev_files() does nothing (no sysfs yet)
- pci_proc_attach_device() does nothing (no proc yet)
- device_attach() does nothing (no drivers yet)
This leaves only the pci_final_fixup calls, D3 support, and marking
the device as added. Of those, only the pci_final_fixup calls have the
potential to be affected by resource allocation.

The only pci_final_fixup handlers that touch resources seem to be one
for x86 (pci_amd_enable_64bit_bar()), and a PowerPC 32 platform driver
(quirk_final_uli1575()), neither of which use this pcibios_init()
function. Even if they did, it would almost certainly be a bug, under
the current ordering, to rely on or make changes to resources before
they were allocated.

Signed-off-by: Sam Bobroff <sbobroff@linux.ibm.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/4506b0489eabd0921a3587d90bd44c7683f3472d.1565930772.git.sbobroff@linux.ibm.com
2019-08-22 23:11:48 +10:00

270 lines
7.2 KiB
C

// SPDX-License-Identifier: GPL-2.0-or-later
/*
* Port for PPC64 David Engebretsen, IBM Corp.
* Contains common pci routines for ppc64 platform, pSeries and iSeries brands.
*
* Copyright (C) 2003 Anton Blanchard <anton@au.ibm.com>, IBM
* Rework, based on alpha PCI code.
*/
#undef DEBUG
#include <linux/kernel.h>
#include <linux/pci.h>
#include <linux/string.h>
#include <linux/init.h>
#include <linux/export.h>
#include <linux/mm.h>
#include <linux/list.h>
#include <linux/syscalls.h>
#include <linux/irq.h>
#include <linux/vmalloc.h>
#include <asm/processor.h>
#include <asm/io.h>
#include <asm/prom.h>
#include <asm/pci-bridge.h>
#include <asm/byteorder.h>
#include <asm/machdep.h>
#include <asm/ppc-pci.h>
/* pci_io_base -- the base address from which io bars are offsets.
* This is the lowest I/O base address (so bar values are always positive),
* and it *must* be the start of ISA space if an ISA bus exists because
* ISA drivers use hard coded offsets. If no ISA bus exists nothing
* is mapped on the first 64K of IO space
*/
unsigned long pci_io_base;
EXPORT_SYMBOL(pci_io_base);
static int __init pcibios_init(void)
{
struct pci_controller *hose, *tmp;
printk(KERN_INFO "PCI: Probing PCI hardware\n");
/* For now, override phys_mem_access_prot. If we need it,g
* later, we may move that initialization to each ppc_md
*/
ppc_md.phys_mem_access_prot = pci_phys_mem_access_prot;
/* On ppc64, we always enable PCI domains and we keep domain 0
* backward compatible in /proc for video cards
*/
pci_add_flags(PCI_ENABLE_PROC_DOMAINS | PCI_COMPAT_DOMAIN_0);
/* Scan all of the recorded PCI controllers. */
list_for_each_entry_safe(hose, tmp, &hose_list, list_node)
pcibios_scan_phb(hose);
/* Call common code to handle resource allocation */
pcibios_resource_survey();
/* Add devices. */
list_for_each_entry_safe(hose, tmp, &hose_list, list_node)
pci_bus_add_devices(hose->bus);
/* Call machine dependent fixup */
if (ppc_md.pcibios_fixup)
ppc_md.pcibios_fixup();
printk(KERN_DEBUG "PCI: Probing PCI hardware done\n");
return 0;
}
subsys_initcall(pcibios_init);
int pcibios_unmap_io_space(struct pci_bus *bus)
{
struct pci_controller *hose;
WARN_ON(bus == NULL);
/* If this is not a PHB, we only flush the hash table over
* the area mapped by this bridge. We don't play with the PTE
* mappings since we might have to deal with sub-page alignments
* so flushing the hash table is the only sane way to make sure
* that no hash entries are covering that removed bridge area
* while still allowing other busses overlapping those pages
*
* Note: If we ever support P2P hotplug on Book3E, we'll have
* to do an appropriate TLB flush here too
*/
if (bus->self) {
#ifdef CONFIG_PPC_BOOK3S_64
struct resource *res = bus->resource[0];
#endif
pr_debug("IO unmapping for PCI-PCI bridge %s\n",
pci_name(bus->self));
#ifdef CONFIG_PPC_BOOK3S_64
__flush_hash_table_range(&init_mm, res->start + _IO_BASE,
res->end + _IO_BASE + 1);
#endif
return 0;
}
/* Get the host bridge */
hose = pci_bus_to_host(bus);
/* Check if we have IOs allocated */
if (hose->io_base_alloc == NULL)
return 0;
pr_debug("IO unmapping for PHB %pOF\n", hose->dn);
pr_debug(" alloc=0x%p\n", hose->io_base_alloc);
/* This is a PHB, we fully unmap the IO area */
vunmap(hose->io_base_alloc);
return 0;
}
EXPORT_SYMBOL_GPL(pcibios_unmap_io_space);
static int pcibios_map_phb_io_space(struct pci_controller *hose)
{
struct vm_struct *area;
unsigned long phys_page;
unsigned long size_page;
unsigned long io_virt_offset;
phys_page = _ALIGN_DOWN(hose->io_base_phys, PAGE_SIZE);
size_page = _ALIGN_UP(hose->pci_io_size, PAGE_SIZE);
/* Make sure IO area address is clear */
hose->io_base_alloc = NULL;
/* If there's no IO to map on that bus, get away too */
if (hose->pci_io_size == 0 || hose->io_base_phys == 0)
return 0;
/* Let's allocate some IO space for that guy. We don't pass
* VM_IOREMAP because we don't care about alignment tricks that
* the core does in that case. Maybe we should due to stupid card
* with incomplete address decoding but I'd rather not deal with
* those outside of the reserved 64K legacy region.
*/
area = __get_vm_area(size_page, 0, PHB_IO_BASE, PHB_IO_END);
if (area == NULL)
return -ENOMEM;
hose->io_base_alloc = area->addr;
hose->io_base_virt = (void __iomem *)(area->addr +
hose->io_base_phys - phys_page);
pr_debug("IO mapping for PHB %pOF\n", hose->dn);
pr_debug(" phys=0x%016llx, virt=0x%p (alloc=0x%p)\n",
hose->io_base_phys, hose->io_base_virt, hose->io_base_alloc);
pr_debug(" size=0x%016llx (alloc=0x%016lx)\n",
hose->pci_io_size, size_page);
/* Establish the mapping */
if (__ioremap_at(phys_page, area->addr, size_page,
pgprot_noncached(PAGE_KERNEL)) == NULL)
return -ENOMEM;
/* Fixup hose IO resource */
io_virt_offset = pcibios_io_space_offset(hose);
hose->io_resource.start += io_virt_offset;
hose->io_resource.end += io_virt_offset;
pr_debug(" hose->io_resource=%pR\n", &hose->io_resource);
return 0;
}
int pcibios_map_io_space(struct pci_bus *bus)
{
WARN_ON(bus == NULL);
/* If this not a PHB, nothing to do, page tables still exist and
* thus HPTEs will be faulted in when needed
*/
if (bus->self) {
pr_debug("IO mapping for PCI-PCI bridge %s\n",
pci_name(bus->self));
pr_debug(" virt=0x%016llx...0x%016llx\n",
bus->resource[0]->start + _IO_BASE,
bus->resource[0]->end + _IO_BASE);
return 0;
}
return pcibios_map_phb_io_space(pci_bus_to_host(bus));
}
EXPORT_SYMBOL_GPL(pcibios_map_io_space);
void pcibios_setup_phb_io_space(struct pci_controller *hose)
{
pcibios_map_phb_io_space(hose);
}
#define IOBASE_BRIDGE_NUMBER 0
#define IOBASE_MEMORY 1
#define IOBASE_IO 2
#define IOBASE_ISA_IO 3
#define IOBASE_ISA_MEM 4
SYSCALL_DEFINE3(pciconfig_iobase, long, which, unsigned long, in_bus,
unsigned long, in_devfn)
{
struct pci_controller* hose;
struct pci_bus *tmp_bus, *bus = NULL;
struct device_node *hose_node;
/* Argh ! Please forgive me for that hack, but that's the
* simplest way to get existing XFree to not lockup on some
* G5 machines... So when something asks for bus 0 io base
* (bus 0 is HT root), we return the AGP one instead.
*/
if (in_bus == 0 && of_machine_is_compatible("MacRISC4")) {
struct device_node *agp;
agp = of_find_compatible_node(NULL, NULL, "u3-agp");
if (agp)
in_bus = 0xf0;
of_node_put(agp);
}
/* That syscall isn't quite compatible with PCI domains, but it's
* used on pre-domains setup. We return the first match
*/
list_for_each_entry(tmp_bus, &pci_root_buses, node) {
if (in_bus >= tmp_bus->number &&
in_bus <= tmp_bus->busn_res.end) {
bus = tmp_bus;
break;
}
}
if (bus == NULL || bus->dev.of_node == NULL)
return -ENODEV;
hose_node = bus->dev.of_node;
hose = PCI_DN(hose_node)->phb;
switch (which) {
case IOBASE_BRIDGE_NUMBER:
return (long)hose->first_busno;
case IOBASE_MEMORY:
return (long)hose->mem_offset[0];
case IOBASE_IO:
return (long)hose->io_base_phys;
case IOBASE_ISA_IO:
return (long)isa_io_base;
case IOBASE_ISA_MEM:
return -EINVAL;
}
return -EOPNOTSUPP;
}
#ifdef CONFIG_NUMA
int pcibus_to_node(struct pci_bus *bus)
{
struct pci_controller *phb = pci_bus_to_host(bus);
return phb->node;
}
EXPORT_SYMBOL(pcibus_to_node);
#endif