2005-04-17 06:20:36 +08:00
|
|
|
#include <linux/pci.h>
|
|
|
|
#include <linux/acpi.h>
|
|
|
|
#include <linux/init.h>
|
2005-10-01 00:34:42 +08:00
|
|
|
#include <linux/irq.h>
|
2007-10-04 06:56:14 +08:00
|
|
|
#include <linux/dmi.h>
|
2005-09-13 00:49:24 +08:00
|
|
|
#include <asm/numa.h>
|
2008-12-27 21:02:28 +08:00
|
|
|
#include <asm/pci_x86.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2007-10-04 06:56:51 +08:00
|
|
|
struct pci_root_info {
|
2009-10-07 05:33:49 +08:00
|
|
|
struct acpi_device *bridge;
|
2007-10-04 06:56:51 +08:00
|
|
|
char *name;
|
|
|
|
unsigned int res_num;
|
|
|
|
struct resource *res;
|
|
|
|
struct pci_bus *bus;
|
|
|
|
int busnum;
|
|
|
|
};
|
|
|
|
|
|
|
|
static acpi_status
|
|
|
|
resource_to_addr(struct acpi_resource *resource,
|
|
|
|
struct acpi_resource_address64 *addr)
|
|
|
|
{
|
|
|
|
acpi_status status;
|
|
|
|
|
|
|
|
status = acpi_resource_to_address64(resource, addr);
|
|
|
|
if (ACPI_SUCCESS(status) &&
|
|
|
|
(addr->resource_type == ACPI_MEMORY_RANGE ||
|
|
|
|
addr->resource_type == ACPI_IO_RANGE) &&
|
|
|
|
addr->address_length > 0 &&
|
|
|
|
addr->producer_consumer == ACPI_PRODUCER) {
|
|
|
|
return AE_OK;
|
|
|
|
}
|
|
|
|
return AE_ERROR;
|
|
|
|
}
|
|
|
|
|
|
|
|
static acpi_status
|
|
|
|
count_resource(struct acpi_resource *acpi_res, void *data)
|
|
|
|
{
|
|
|
|
struct pci_root_info *info = data;
|
|
|
|
struct acpi_resource_address64 addr;
|
|
|
|
acpi_status status;
|
|
|
|
|
|
|
|
status = resource_to_addr(acpi_res, &addr);
|
|
|
|
if (ACPI_SUCCESS(status))
|
|
|
|
info->res_num++;
|
|
|
|
return AE_OK;
|
|
|
|
}
|
|
|
|
|
2009-05-28 03:41:44 +08:00
|
|
|
static int
|
|
|
|
bus_has_transparent_bridge(struct pci_bus *bus)
|
|
|
|
{
|
|
|
|
struct pci_dev *dev;
|
|
|
|
|
|
|
|
list_for_each_entry(dev, &bus->devices, bus_list) {
|
|
|
|
u16 class = dev->class >> 8;
|
|
|
|
|
|
|
|
if (class == PCI_CLASS_BRIDGE_PCI && dev->transparent)
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2007-10-04 06:56:51 +08:00
|
|
|
static acpi_status
|
|
|
|
setup_resource(struct acpi_resource *acpi_res, void *data)
|
|
|
|
{
|
|
|
|
struct pci_root_info *info = data;
|
|
|
|
struct resource *res;
|
|
|
|
struct acpi_resource_address64 addr;
|
|
|
|
acpi_status status;
|
|
|
|
unsigned long flags;
|
|
|
|
struct resource *root;
|
2009-05-28 03:41:44 +08:00
|
|
|
int max_root_bus_resources = PCI_BUS_NUM_RESOURCES;
|
2009-06-25 10:01:19 +08:00
|
|
|
u64 start, end;
|
|
|
|
|
|
|
|
if (bus_has_transparent_bridge(info->bus))
|
|
|
|
max_root_bus_resources -= 3;
|
2007-11-17 23:27:01 +08:00
|
|
|
|
2007-10-04 06:56:51 +08:00
|
|
|
status = resource_to_addr(acpi_res, &addr);
|
|
|
|
if (!ACPI_SUCCESS(status))
|
|
|
|
return AE_OK;
|
|
|
|
|
|
|
|
if (addr.resource_type == ACPI_MEMORY_RANGE) {
|
|
|
|
root = &iomem_resource;
|
|
|
|
flags = IORESOURCE_MEM;
|
|
|
|
if (addr.info.mem.caching == ACPI_PREFETCHABLE_MEMORY)
|
|
|
|
flags |= IORESOURCE_PREFETCH;
|
|
|
|
} else if (addr.resource_type == ACPI_IO_RANGE) {
|
|
|
|
root = &ioport_resource;
|
|
|
|
flags = IORESOURCE_IO;
|
|
|
|
} else
|
|
|
|
return AE_OK;
|
|
|
|
|
2009-06-25 10:01:19 +08:00
|
|
|
start = addr.minimum + addr.translation_offset;
|
|
|
|
end = start + addr.address_length - 1;
|
2009-05-28 03:41:44 +08:00
|
|
|
if (info->res_num >= max_root_bus_resources) {
|
|
|
|
printk(KERN_WARNING "PCI: Failed to allocate 0x%lx-0x%lx "
|
|
|
|
"from %s for %s due to _CRS returning more than "
|
2009-06-25 10:01:19 +08:00
|
|
|
"%d resource descriptors\n", (unsigned long) start,
|
|
|
|
(unsigned long) end, root->name, info->name,
|
2009-05-28 03:41:44 +08:00
|
|
|
max_root_bus_resources);
|
|
|
|
return AE_OK;
|
|
|
|
}
|
|
|
|
|
2009-06-25 10:01:19 +08:00
|
|
|
res = &info->res[info->res_num];
|
|
|
|
res->name = info->name;
|
|
|
|
res->flags = flags;
|
|
|
|
res->start = start;
|
|
|
|
res->end = end;
|
|
|
|
res->child = NULL;
|
|
|
|
|
2007-10-04 06:56:51 +08:00
|
|
|
if (insert_resource(root, res)) {
|
2009-10-07 05:33:49 +08:00
|
|
|
dev_err(&info->bridge->dev, "can't allocate %pRt\n", res);
|
2007-10-04 06:56:51 +08:00
|
|
|
} else {
|
|
|
|
info->bus->resource[info->res_num] = res;
|
|
|
|
info->res_num++;
|
2009-10-07 05:33:49 +08:00
|
|
|
if (addr.translation_offset)
|
|
|
|
dev_info(&info->bridge->dev, "host bridge window: %pRt "
|
|
|
|
"(PCI address [%#llx-%#llx])\n",
|
|
|
|
res, res->start - addr.translation_offset,
|
|
|
|
res->end - addr.translation_offset);
|
|
|
|
else
|
|
|
|
dev_info(&info->bridge->dev,
|
|
|
|
"host bridge window: %pRt\n", res);
|
2007-10-04 06:56:51 +08:00
|
|
|
}
|
|
|
|
return AE_OK;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
get_current_resources(struct acpi_device *device, int busnum,
|
2008-02-09 06:00:52 +08:00
|
|
|
int domain, struct pci_bus *bus)
|
2007-10-04 06:56:51 +08:00
|
|
|
{
|
|
|
|
struct pci_root_info info;
|
|
|
|
size_t size;
|
|
|
|
|
2009-10-07 05:33:49 +08:00
|
|
|
info.bridge = device;
|
2007-10-04 06:56:51 +08:00
|
|
|
info.bus = bus;
|
|
|
|
info.res_num = 0;
|
|
|
|
acpi_walk_resources(device->handle, METHOD_NAME__CRS, count_resource,
|
|
|
|
&info);
|
|
|
|
if (!info.res_num)
|
|
|
|
return;
|
|
|
|
|
|
|
|
size = sizeof(*info.res) * info.res_num;
|
|
|
|
info.res = kmalloc(size, GFP_KERNEL);
|
|
|
|
if (!info.res)
|
|
|
|
goto res_alloc_fail;
|
|
|
|
|
2008-02-09 06:00:52 +08:00
|
|
|
info.name = kmalloc(16, GFP_KERNEL);
|
2007-10-04 06:56:51 +08:00
|
|
|
if (!info.name)
|
|
|
|
goto name_alloc_fail;
|
2008-02-09 06:00:52 +08:00
|
|
|
sprintf(info.name, "PCI Bus %04x:%02x", domain, busnum);
|
2007-10-04 06:56:51 +08:00
|
|
|
|
|
|
|
info.res_num = 0;
|
|
|
|
acpi_walk_resources(device->handle, METHOD_NAME__CRS, setup_resource,
|
|
|
|
&info);
|
|
|
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
name_alloc_fail:
|
|
|
|
kfree(info.res);
|
|
|
|
res_alloc_fail:
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
struct pci_bus * __devinit pci_acpi_scan_root(struct acpi_device *device, int domain, int busnum)
|
|
|
|
{
|
2005-09-13 00:49:24 +08:00
|
|
|
struct pci_bus *bus;
|
2007-07-22 05:23:39 +08:00
|
|
|
struct pci_sysdata *sd;
|
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: Yinghai Lu <yinghai.lu@sun.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-19 19:20:09 +08:00
|
|
|
int node;
|
|
|
|
#ifdef CONFIG_ACPI_NUMA
|
2007-07-22 05:23:39 +08:00
|
|
|
int pxm;
|
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: Yinghai Lu <yinghai.lu@sun.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-19 19:20:09 +08:00
|
|
|
#endif
|
2007-07-22 05:23:39 +08:00
|
|
|
|
2007-10-12 04:58:30 +08:00
|
|
|
if (domain && !pci_domains_supported) {
|
|
|
|
printk(KERN_WARNING "PCI: Multiple domains not supported "
|
|
|
|
"(dom %d, bus %d)\n", domain, busnum);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
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: Yinghai Lu <yinghai.lu@sun.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-19 19:20:09 +08:00
|
|
|
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
|
|
|
|
#endif
|
|
|
|
node = get_mp_bus_to_node(busnum);
|
2008-02-21 04:41:52 +08:00
|
|
|
|
|
|
|
if (node != -1 && !node_online(node))
|
|
|
|
node = -1;
|
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: Yinghai Lu <yinghai.lu@sun.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-19 19:20:09 +08:00
|
|
|
|
2007-07-22 05:23:39 +08:00
|
|
|
/* 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.
|
|
|
|
*/
|
|
|
|
sd = kzalloc(sizeof(*sd), GFP_KERNEL);
|
|
|
|
if (!sd) {
|
|
|
|
printk(KERN_ERR "PCI: OOM, not probing PCI bus %02x\n", busnum);
|
|
|
|
return NULL;
|
|
|
|
}
|
2005-09-13 00:49:24 +08:00
|
|
|
|
2007-10-12 04:58:30 +08:00
|
|
|
sd->domain = domain;
|
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: Yinghai Lu <yinghai.lu@sun.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-02-19 19:20:09 +08:00
|
|
|
sd->node = node;
|
2008-04-16 05:34:49 +08:00
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
|
|
|
bus = pci_find_bus(domain, busnum);
|
|
|
|
if (bus) {
|
|
|
|
/*
|
|
|
|
* If the desired bus exits, the content of bus->sysdata will
|
|
|
|
* be replaced by sd.
|
|
|
|
*/
|
|
|
|
memcpy(bus->sysdata, sd, sizeof(*sd));
|
|
|
|
kfree(sd);
|
2009-06-25 11:00:12 +08:00
|
|
|
} else {
|
|
|
|
bus = pci_create_bus(NULL, busnum, &pci_root_ops, sd);
|
|
|
|
if (bus) {
|
|
|
|
if (pci_probe & PCI_USE__CRS)
|
|
|
|
get_current_resources(device, busnum, domain,
|
|
|
|
bus);
|
|
|
|
bus->subordinate = pci_scan_child_bus(bus);
|
|
|
|
}
|
|
|
|
}
|
2007-07-22 05:23:39 +08:00
|
|
|
|
|
|
|
if (!bus)
|
|
|
|
kfree(sd);
|
|
|
|
|
2008-04-19 16:30:16 +08:00
|
|
|
if (bus && node != -1) {
|
2005-09-13 00:49:24 +08:00
|
|
|
#ifdef CONFIG_ACPI_NUMA
|
2008-04-19 16:30:16 +08:00
|
|
|
if (pxm >= 0)
|
2008-12-19 07:34:51 +08:00
|
|
|
dev_printk(KERN_DEBUG, &bus->dev,
|
|
|
|
"on NUMA node %d (pxm %d)\n", node, pxm);
|
2008-04-19 16:30:16 +08:00
|
|
|
#else
|
2008-12-19 07:34:51 +08:00
|
|
|
dev_printk(KERN_DEBUG, &bus->dev, "on NUMA node %d\n", node);
|
2005-09-13 00:49:24 +08:00
|
|
|
#endif
|
2008-04-19 16:30:16 +08:00
|
|
|
}
|
2007-10-04 06:56:51 +08:00
|
|
|
|
2005-09-13 00:49:24 +08:00
|
|
|
return bus;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2008-07-03 04:50:29 +08:00
|
|
|
int __init pci_acpi_init(void)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
struct pci_dev *dev = NULL;
|
|
|
|
|
|
|
|
if (pcibios_scanned)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (acpi_noirq)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
printk(KERN_INFO "PCI: Using ACPI for IRQ routing\n");
|
|
|
|
acpi_irq_penalty_init();
|
|
|
|
pcibios_scanned++;
|
|
|
|
pcibios_enable_irq = acpi_pci_irq_enable;
|
2005-07-28 11:02:00 +08:00
|
|
|
pcibios_disable_irq = acpi_pci_irq_disable;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (pci_routeirq) {
|
|
|
|
/*
|
|
|
|
* PCI IRQ routing is set up by pci_enable_device(), but we
|
|
|
|
* also do it here in case there are still broken drivers that
|
|
|
|
* don't use pci_enable_device().
|
|
|
|
*/
|
|
|
|
printk(KERN_INFO "PCI: Routing PCI interrupts for all devices because \"pci=routeirq\" specified\n");
|
2005-11-07 15:39:36 +08:00
|
|
|
for_each_pci_dev(dev)
|
2005-04-17 06:20:36 +08:00
|
|
|
acpi_pci_irq_enable(dev);
|
2008-02-19 00:44:13 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|