Commit 871d5f8d authored by Yinghai Lu's avatar Yinghai Lu Committed by Ingo Molnar

x86: get mp_bus_to_node early

Currently, on an amd k8 system with multi ht chains, the numa_node of
pci devices under /sys/devices/pci0000:80/* is always 0, even if that
chain is on node 1 or 2 or 3.

Workaround: pcibus_to_node(bus) is used when we want to get the node that
pci_device is on.

In struct device, we already have numa_node member, and we could use
dev_to_node()/set_dev_node() to get and set numa_node in the device.
set_dev_node is called in pci_device_add() with pcibus_to_node(bus),
and pcibus_to_node uses bus->sysdata for nodeid.

The problem is when pci_add_device is called, bus->sysdata is not assigned
correct nodeid yet. The result is that numa_node will always be 0.

pcibios_scan_root and pci_scan_root could take sysdata. So we need to get
mp_bus_to_node mapping before these two are called, and thus
get_mp_bus_to_node could get correct node for sysdata in root bus.

In scanning of the root bus, all child busses will take parent bus sysdata.
So all pci_device->dev.numa_node will be assigned correctly and automatically.

Later we could use dev_to_node(&pci_dev->dev) to get numa_node, and we
could also could make other bus specific device get the correct numa_node
too.

This is an updated version of pci_sysdata and Jeff's pci_domain patch.

[ mingo@elte.hu: build fix ]
Signed-off-by: default avatarYinghai Lu <yinghai.lu@sun.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
Signed-off-by: default avatarThomas Gleixner <tglx@linutronix.de>
parent bb63b421
......@@ -10,5 +10,6 @@ pci-y += legacy.o irq.o
pci-$(CONFIG_X86_VISWS) := visws.o fixup.o
pci-$(CONFIG_X86_NUMAQ) := numa.o irq.o
pci-$(CONFIG_NUMA) += mp_bus_to_node.o
obj-y += $(pci-y) common.o early.o
......@@ -191,7 +191,10 @@ struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int do
{
struct pci_bus *bus;
struct pci_sysdata *sd;
int node;
#ifdef CONFIG_ACPI_NUMA
int pxm;
#endif
dmi_check_system(acpi_pciprobe_dmi_table);
......@@ -201,6 +204,17 @@ struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int do
return NULL;
}
node = -1;
#ifdef CONFIG_ACPI_NUMA
pxm = acpi_get_pxm(device->handle);
if (pxm >= 0)
node = pxm_to_node(pxm);
if (node != -1)
set_mp_bus_to_node(busnum, node);
else
node = get_mp_bus_to_node(busnum);
#endif
/* Allocate per-root-bus (not per bus) arch-specific data.
* TODO: leak; this memory is never freed.
* It's arguable whether it's worth the trouble to care.
......@@ -212,13 +226,7 @@ struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int do
}
sd->domain = domain;
sd->node = -1;
pxm = acpi_get_pxm(device->handle);
#ifdef CONFIG_ACPI_NUMA
if (pxm >= 0)
sd->node = pxm_to_node(pxm);
#endif
sd->node = node;
/*
* Maybe the desired pci bus has been already scanned. In such case
* it is unnecessary to scan the pci bus with the given domain,busnum.
......@@ -238,9 +246,9 @@ struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int do
kfree(sd);
#ifdef CONFIG_ACPI_NUMA
if (bus != NULL) {
if (bus) {
if (pxm >= 0) {
printk("bus %d -> pxm %d -> node %d\n",
printk(KERN_DEBUG "bus %02x -> pxm %d -> node %d\n",
busnum, pxm, pxm_to_node(pxm));
}
}
......@@ -248,7 +256,6 @@ struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int do
if (bus && (pci_probe & PCI_USE__CRS))
get_current_resources(device, busnum, domain, bus);
return bus;
}
......
......@@ -342,9 +342,14 @@ struct pci_bus * __devinit pcibios_scan_root(int busnum)
return NULL;
}
sd->node = get_mp_bus_to_node(busnum);
printk(KERN_DEBUG "PCI: Probing PCI hardware (bus %02x)\n", busnum);
bus = pci_scan_bus_parented(NULL, busnum, &pci_root_ops, sd);
if (!bus)
kfree(sd);
return pci_scan_bus_parented(NULL, busnum, &pci_root_ops, sd);
return bus;
}
extern u8 pci_cache_line_size;
......@@ -480,7 +485,7 @@ void pcibios_disable_device (struct pci_dev *dev)
pcibios_disable_irq(dev);
}
struct pci_bus *__devinit pci_scan_bus_with_sysdata(int busno)
struct pci_bus *pci_scan_bus_on_node(int busno, struct pci_ops *ops, int node)
{
struct pci_bus *bus = NULL;
struct pci_sysdata *sd;
......@@ -495,10 +500,15 @@ struct pci_bus *__devinit pci_scan_bus_with_sysdata(int busno)
printk(KERN_ERR "PCI: OOM, skipping PCI bus %02x\n", busno);
return NULL;
}
sd->node = -1;
bus = pci_scan_bus(busno, &pci_root_ops, sd);
sd->node = node;
bus = pci_scan_bus(busno, ops, sd);
if (!bus)
kfree(sd);
return bus;
}
struct pci_bus *pci_scan_bus_with_sysdata(int busno)
{
return pci_scan_bus_on_node(busno, &pci_root_ops, -1);
}
......@@ -136,9 +136,11 @@ static void __init pirq_peer_trick(void)
busmap[e->bus] = 1;
}
for(i = 1; i < 256; i++) {
int node;
if (!busmap[i] || pci_find_bus(0, i))
continue;
if (pci_scan_bus_with_sysdata(i))
node = get_mp_bus_to_node(i);
if (pci_scan_bus_on_node(i, &pci_root_ops, node))
printk(KERN_INFO "PCI: Discovered primary peer "
"bus %02x [IRQ]\n", i);
}
......
#include <linux/init.h>
#include <linux/pci.h>
#include <asm/pci-direct.h>
#include <asm/mpspec.h>
#include <linux/cpumask.h>
#include <linux/topology.h>
/*
* This discovers the pcibus <-> node mapping on AMD K8.
......@@ -20,64 +22,102 @@
#define SUBORDINATE_LDT_BUS_NUMBER(dword) ((dword >> 16) & 0xFF)
#define PCI_DEVICE_ID_K8HTCONFIG 0x1100
#ifdef CONFIG_NUMA
#define BUS_NR 256
static int mp_bus_to_node[BUS_NR];
void set_mp_bus_to_node(int busnum, int node)
{
if (busnum >= 0 && busnum < BUS_NR)
mp_bus_to_node[busnum] = node;
}
int get_mp_bus_to_node(int busnum)
{
int node = -1;
if (busnum < 0 || busnum > (BUS_NR - 1))
return node;
node = mp_bus_to_node[busnum];
/*
* let numa_node_id to decide it later in dma_alloc_pages
* if there is no ram on that node
*/
if (node != -1 && !node_online(node))
node = -1;
return node;
}
#endif
/**
* fill_mp_bus_to_cpumask()
* early_fill_mp_bus_to_node()
* called before pcibios_scan_root and pci_scan_bus
* fills the mp_bus_to_cpumask array based according to the LDT Bus Number
* Registers found in the K8 northbridge
*/
__init static int
fill_mp_bus_to_cpumask(void)
early_fill_mp_bus_to_node(void)
{
struct pci_dev *nb_dev = NULL;
#ifdef CONFIG_NUMA
int i, j;
unsigned slot;
u32 ldtbus, nid;
u32 id;
static int lbnr[3] = {
LDT_BUS_NUMBER_REGISTER_0,
LDT_BUS_NUMBER_REGISTER_1,
LDT_BUS_NUMBER_REGISTER_2
};
while ((nb_dev = pci_get_device(PCI_VENDOR_ID_AMD,
PCI_DEVICE_ID_K8HTCONFIG, nb_dev))) {
pci_read_config_dword(nb_dev, NODE_ID_REGISTER, &nid);
for (i = 0; i < BUS_NR; i++)
mp_bus_to_node[i] = -1;
if (!early_pci_allowed())
return -1;
for (slot = 0x18; slot < 0x20; slot++) {
id = read_pci_config(0, slot, 0, PCI_VENDOR_ID);
if (id != (PCI_VENDOR_ID_AMD | (PCI_DEVICE_ID_K8HTCONFIG<<16)))
break;
nid = read_pci_config(0, slot, 0, NODE_ID_REGISTER);
for (i = 0; i < NR_LDT_BUS_NUMBER_REGISTERS; i++) {
pci_read_config_dword(nb_dev, lbnr[i], &ldtbus);
ldtbus = read_pci_config(0, slot, 0, lbnr[i]);
/*
* if there are no busses hanging off of the current
* ldt link then both the secondary and subordinate
* bus number fields are set to 0.
*
*
* RED-PEN
* This is slightly broken because it assumes
* HT node IDs == Linux node ids, which is not always
* HT node IDs == Linux node ids, which is not always
* true. However it is probably mostly true.
*/
if (!(SECONDARY_LDT_BUS_NUMBER(ldtbus) == 0
&& SUBORDINATE_LDT_BUS_NUMBER(ldtbus) == 0)) {
for (j = SECONDARY_LDT_BUS_NUMBER(ldtbus);
j <= SUBORDINATE_LDT_BUS_NUMBER(ldtbus);
j++) {
struct pci_bus *bus;
struct pci_sysdata *sd;
long node = NODE_ID(nid);
/* Algorithm a bit dumb, but
it shouldn't matter here */
bus = pci_find_bus(0, j);
if (!bus)
continue;
if (!node_online(node))
node = 0;
sd = bus->sysdata;
sd->node = node;
}
j++) {
int node = NODE_ID(nid);
mp_bus_to_node[j] = (unsigned char)node;
}
}
}
}
for (i = 0; i < BUS_NR; i++) {
int node = mp_bus_to_node[i];
if (node >= 0)
printk(KERN_DEBUG "bus: %02x to node: %02x\n", i, node);
}
#endif
return 0;
}
fs_initcall(fill_mp_bus_to_cpumask);
postcore_initcall(early_fill_mp_bus_to_node);
......@@ -12,6 +12,7 @@
static void __devinit pcibios_fixup_peer_bridges(void)
{
int n, devfn;
long node;
if (pcibios_last_bus <= 0 || pcibios_last_bus >= 0xff)
return;
......@@ -21,12 +22,13 @@ static void __devinit pcibios_fixup_peer_bridges(void)
u32 l;
if (pci_find_bus(0, n))
continue;
node = get_mp_bus_to_node(n);
for (devfn = 0; devfn < 256; devfn += 8) {
if (!raw_pci_read(0, n, devfn, PCI_VENDOR_ID, 2, &l) &&
l != 0x0000 && l != 0xffff) {
DBG("Found device at %02x:%02x [%04x]\n", n, devfn, l);
printk(KERN_INFO "PCI: Discovered peer bus %02x\n", n);
pci_scan_bus_with_sysdata(n);
pci_scan_bus_on_node(n, &pci_root_ops, node);
break;
}
}
......
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/topology.h>
#define BUS_NR 256
static unsigned char mp_bus_to_node[BUS_NR];
void set_mp_bus_to_node(int busnum, int node)
{
if (busnum >= 0 && busnum < BUS_NR)
mp_bus_to_node[busnum] = (unsigned char) node;
}
int get_mp_bus_to_node(int busnum)
{
int node;
if (busnum < 0 || busnum > (BUS_NR - 1))
return 0;
node = mp_bus_to_node[busnum];
return node;
}
......@@ -19,6 +19,8 @@ struct pci_sysdata {
};
/* scan a bus after allocating a pci_sysdata for it */
extern struct pci_bus *pci_scan_bus_on_node(int busno, struct pci_ops *ops,
int node);
extern struct pci_bus *pci_scan_bus_with_sysdata(int busno);
static inline int pci_domain_nr(struct pci_bus *bus)
......
......@@ -198,4 +198,17 @@ extern cpumask_t cpu_coregroup_map(int cpu);
#define smt_capable() (smp_num_siblings > 1)
#endif
#ifdef CONFIG_NUMA
extern int get_mp_bus_to_node(int busnum);
extern void set_mp_bus_to_node(int busnum, int node);
#else
static inline int get_mp_bus_to_node(int busnum)
{
return 0;
}
static inline void set_mp_bus_to_node(int busnum, int node)
{
}
#endif
#endif
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment