[PATCH 1/8] x86_64: check MSR to get MMCONFIG for AMD Family 10h Opteron v3

[PATCH 1/8] x86_64: check MSR to get MMCONFIG for AMD Family 10h Opteron v3

[Yinghai Lu]

From: Yinghai Lu <Yinghai.Lu@Sun.COM>

so even booting kernel with acpi=off or even MCFG is not there, we still can
use MMCONFIG.

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Andi Kleen <ak@suse.de>
Cc: Greg KH <greg@kroah.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

Index: linux-2.6/arch/x86/pci/mmconfig-shared.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/mmconfig-shared.c
+++ linux-2.6/arch/x86/pci/mmconfig-shared.c
@@ -100,33 +100,88 @@ static const char __init *pci_mmcfg_inte
 	return "Intel Corporation 945G/GZ/P/PL Express Memory Controller Hub";
 }
 
+static const char __init *pci_mmcfg_amd_fam10h(void)
+{
+	u32 low, high, address;
+	u64 base, msr;
+	int i;
+	unsigned segnbits = 0, busnbits;
+
+	address = MSR_FAM10H_MMIO_CONF_BASE;
+	if (rdmsr_safe(address, &low, &high))
+		return NULL;
+
+	msr = high;
+	msr <<= 32;
+	msr |= low;
+
+	/* mmconfig is not enable */
+	if (!(msr & FAM10H_MMIO_CONF_ENABLE))
+		return NULL;
+
+	base = msr & (FAM10H_MMIO_CONF_BASE_MASK<<FAM10H_MMIO_CONF_BASE_SHIFT);
+
+	busnbits = (msr >> FAM10H_MMIO_CONF_BUSRANGE_SHIFT) &
+			 FAM10H_MMIO_CONF_BUSRANGE_MASK;
+	if (busnbits > 8) {
+		segnbits = busnbits - 8;
+		busnbits = 8;
+	}
+
+	pci_mmcfg_config_num = (1 << segnbits);
+	pci_mmcfg_config = kzalloc(sizeof(pci_mmcfg_config[0]) *
+				   pci_mmcfg_config_num, GFP_KERNEL);
+	if (!pci_mmcfg_config)
+		return NULL;
+
+	for (i = 0; i < (1 << segnbits); i++) {
+		pci_mmcfg_config[i].address = base + (1<<28) * i;
+		pci_mmcfg_config[i].pci_segment = i;
+		pci_mmcfg_config[i].start_bus_number = 0;
+		pci_mmcfg_config[i].end_bus_number = (1 << busnbits) - 1;
+	}
+
+	return "AMD Family 10h NB";
+}
+
 struct pci_mmcfg_hostbridge_probe {
+	u32 bus;
+	u32 devfn;
 	u32 vendor;
 	u32 device;
 	const char *(*probe)(void);
 };
 
 static struct pci_mmcfg_hostbridge_probe pci_mmcfg_probes[] __initdata = {
-	{ PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_E7520_MCH, pci_mmcfg_e7520 },
-	{ PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_82945G_HB, pci_mmcfg_intel_945 },
+	{ 0, PCI_DEVFN(0, 0), PCI_VENDOR_ID_INTEL,
+	  PCI_DEVICE_ID_INTEL_E7520_MCH, pci_mmcfg_e7520 },
+	{ 0, PCI_DEVFN(0, 0), PCI_VENDOR_ID_INTEL,
+	  PCI_DEVICE_ID_INTEL_82945G_HB, pci_mmcfg_intel_945 },
+	{ 0, PCI_DEVFN(0x18, 0), PCI_VENDOR_ID_AMD,
+	  0x1200, pci_mmcfg_amd_fam10h },
+	{ 0xff, PCI_DEVFN(0, 0), PCI_VENDOR_ID_AMD,
+	  0x1200, pci_mmcfg_amd_fam10h },
 };
 
 static int __init pci_mmcfg_check_hostbridge(void)
 {
 	u32 l;
+	u32 bus, devfn;
 	u16 vendor, device;
 	int i;
 	const char *name;
 
-	pci_direct_conf1.read(0, 0, PCI_DEVFN(0,0), 0, 4, &l);
-	vendor = l & 0xffff;
-	device = (l >> 16) & 0xffff;
-
 	pci_mmcfg_config_num = 0;
 	pci_mmcfg_config = NULL;
 	name = NULL;
 
 	for (i = 0; !name && i < ARRAY_SIZE(pci_mmcfg_probes); i++) {
+		bus =  pci_mmcfg_probes[i].bus;
+		devfn = pci_mmcfg_probes[i].devfn;
+		pci_direct_conf1.read(0, bus, devfn, 0, 4, &l);
+		vendor = l & 0xffff;
+		device = (l >> 16) & 0xffff;
+
 		if (pci_mmcfg_probes[i].vendor == vendor &&
 		    pci_mmcfg_probes[i].device == device)
 			name = pci_mmcfg_probes[i].probe();

[PATCH 2/8] x86_64: check and enable MMCONFIG for AMD Family 10h Opteron v3

[Yinghai Lu]

So we can use MMCONF when MMCONF is not set by BIOS

using TOP_MEM2 msr to get memory top, and try to scan fam10h mmio routing to
make sure the range is not conflicted with some prefetch MMIO that is above 4G.
(current only LinuxBIOS assign 64 bit mmio above 4G for some co-processor)

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

Index: linux-2.6/arch/x86/kernel/setup_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/setup_64.c
+++ linux-2.6/arch/x86/kernel/setup_64.c
@@ -29,6 +29,7 @@
 #include <linux/crash_dump.h>
 #include <linux/root_dev.h>
 #include <linux/pci.h>
+#include <asm/pci-direct.h>
 #include <linux/efi.h>
 #include <linux/acpi.h>
 #include <linux/kallsyms.h>
@@ -39,6 +40,7 @@
 #include <linux/dmi.h>
 #include <linux/dma-mapping.h>
 #include <linux/ctype.h>
+#include <linux/sort.h>
 #include <linux/uaccess.h>
 #include <linux/init_ohci1394_dma.h>
 
@@ -577,6 +579,205 @@ static int __cpuinit nearby_node(int api
 }
 #endif
 
+#ifdef CONFIG_PCI_MMCONFIG
+struct pci_hostbridge_probe {
+	u32 bus;
+	u32 slot;
+	u32 vendor;
+	u32 device;
+};
+
+static u64 __cpuinitdata fam10h_pci_mmconf_base;
+static int __cpuinitdata fam10h_pci_mmconf_base_status;
+
+static struct pci_hostbridge_probe pci_probes[] __cpuinitdata = {
+	{ 0, 0x18, PCI_VENDOR_ID_AMD, 0x1200 },
+	{ 0xff, 0, PCI_VENDOR_ID_AMD, 0x1200 },
+};
+
+struct range {
+	u64 start;
+	u64 end;
+};
+
+static int __cpuinit cmp_range(const void *x1, const void *x2)
+{
+	const struct range *r1 = x1;
+	const struct range *r2 = x2;
+	int start1, start2;
+
+	start1 = r1->start >> 32;
+	start2 = r2->start >> 32;
+
+	return start1 - start2;
+}
+
+/*[47:0] */
+/* need to avoid (0xfd<<32) and (0xfe<<32), ht used space */
+#define FAM10H_PCI_MMCONF_BASE (0xfcULL<<32)
+#define BASE_VALID(b) ((b != (0xfdULL << 32)) && (b != (0xfeULL << 32)))
+static void __cpuinit get_fam10h_pci_mmconf_base(void)
+{
+	int i;
+	unsigned bus;
+	unsigned slot;
+	int found;
+
+	u64 val;
+	u32 address;
+	u64 tom2;
+	u64 base = FAM10H_PCI_MMCONF_BASE;
+
+	int hi_mmio_num;
+	struct range range[8];
+
+	/* only try to get setting from BSP */
+	/* -1 or 1 */
+	if (fam10h_pci_mmconf_base_status)
+		return;
+
+	if (!early_pci_allowed())
+		goto fail;
+
+	found = 0;
+	for (i = 0; i < ARRAY_SIZE(pci_probes); i++) {
+		u32 id;
+		u16 device;
+		u16 vendor;
+
+		bus = pci_probes[i].bus;
+		slot = pci_probes[i].slot;
+		id = read_pci_config(bus, slot, 0, PCI_VENDOR_ID);
+
+		vendor = id & 0xffff;
+		device = (id>>16) & 0xffff;
+		if (pci_probes[i].vendor == vendor &&
+		    pci_probes[i].device == device) {
+			found = 1;
+			break;
+		}
+	}
+
+	if (!found)
+		goto fail;
+
+	/* SYS_CFG */
+	address = MSR_K8_SYSCFG;
+	rdmsrl(address, val);
+
+	/* TOP_MEM2 is not enabled? */
+	if (!(val & (1<<21))) {
+		tom2 = 0;
+	} else {
+		/* TOP_MEM2 */
+		address = MSR_K8_TOP_MEM2;
+		rdmsrl(address, val);
+		tom2 = val & (0xffffULL<<32);
+	}
+
+	if (base <= tom2)
+		base = tom2 + (1ULL<<32);
+
+	/*
+	 * need to check if the range is in the high mmio range that is
+	 * above 4G
+	 */
+	hi_mmio_num = 0;
+	for (i = 0; i < 8; i++) {
+		u32 reg;
+		u64 start;
+		u64 end;
+		reg = read_pci_config(bus, slot, 1, 0x80 + (i << 3));
+		if (!(reg & 3))
+			continue;
+
+		start = (((u64)reg) << 8) & (0xffULL << 32); /* 39:16 on 31:8*/
+		reg = read_pci_config(bus, slot, 1, 0x84 + (i << 3));
+		end = (((u64)reg) << 8) & (0xffULL << 32); /* 39:16 on 31:8*/
+
+		if (!end)
+			continue;
+
+		range[hi_mmio_num].start = start;
+		range[hi_mmio_num].end = end;
+		hi_mmio_num++;
+	}
+
+	if (!hi_mmio_num)
+		goto out;
+
+	/* sort the range */
+	sort(range, hi_mmio_num, sizeof(struct range), cmp_range, NULL);
+
+	if (range[hi_mmio_num - 1].end < base)
+		goto out;
+	if (range[0].start > base)
+		goto out;
+
+	/* need to find one window */
+	base = range[0].start - (1ULL << 32);
+	if ((base > tom2) && BASE_VALID(base))
+		goto out;
+	base = range[hi_mmio_num - 1].end + (1ULL << 32);
+	if ((base > tom2) && BASE_VALID(base))
+		goto out;
+	/* need to find window between ranges */
+	if (hi_mmio_num > 1)
+	for (i = 0; i < hi_mmio_num - 1; i++) {
+		if (range[i + 1].start > (range[i].end + (1ULL << 32))) {
+			base = range[i].end + (1ULL << 32);
+			if ((base > tom2) && BASE_VALID(base))
+				goto out;
+		}
+	}
+
+fail:
+	fam10h_pci_mmconf_base_status = -1;
+	return;
+out:
+	fam10h_pci_mmconf_base = base;
+	fam10h_pci_mmconf_base_status = 1;
+}
+#endif
+
+static void __cpuinit fam10h_check_enable_mmcfg(struct cpuinfo_x86 *c)
+{
+#ifdef CONFIG_PCI_MMCONFIG
+	u64 val;
+	u32 address;
+
+	address = MSR_FAM10H_MMIO_CONF_BASE;
+	rdmsrl(address, val);
+
+	/* try to make sure that AP's setting is identical to BSP setting */
+	if (val & FAM10H_MMIO_CONF_ENABLE) {
+		u64 base;
+		base = val & (0xffffULL << 32);
+		if (fam10h_pci_mmconf_base_status <= 0) {
+			fam10h_pci_mmconf_base = base;
+			fam10h_pci_mmconf_base_status = 1;
+			return;
+		} else if (fam10h_pci_mmconf_base ==  base)
+			return;
+	}
+
+	/*
+	 * if it is not enabled, try to enable it and assume only one segment
+	 * with 256 buses
+	 */
+	get_fam10h_pci_mmconf_base();
+	if (fam10h_pci_mmconf_base_status <= 0)
+		return;
+
+	printk(KERN_INFO "Enable MMCONFIG on AMD Family 10h\n");
+	val &= ~((FAM10H_MMIO_CONF_BASE_MASK<<FAM10H_MMIO_CONF_BASE_SHIFT) |
+	     (FAM10H_MMIO_CONF_BUSRANGE_MASK<<FAM10H_MMIO_CONF_BUSRANGE_SHIFT));
+	val |= fam10h_pci_mmconf_base | (8 << FAM10H_MMIO_CONF_BUSRANGE_SHIFT) |
+	       FAM10H_MMIO_CONF_ENABLE;
+	wrmsrl(address, val);
+#endif
+}
+
 /*
  * On a AMD dual core setup the lower bits of the APIC id distingush the cores.
  * Assumes number of cores is a power of two.
@@ -760,6 +961,9 @@ static void __cpuinit init_amd(struct cp
 	/* MFENCE stops RDTSC speculation */
 	set_cpu_cap(c, X86_FEATURE_MFENCE_RDTSC);
 
+	if (c->x86 == 0x10)
+		fam10h_check_enable_mmcfg(c);
+
 	if (amd_apic_timer_broken())
 		disable_apic_timer = 1;
 }

[PATCH 4/8] x86_64: use bus conf in NB conf fun1 to get bus range on node

[Yinghai Lu]

So we use the same code with Quad core cpu as old opteron.
this patch is usful when acpi=off or _PXM is not there in DSDT

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

Index: linux-2.6/arch/x86/pci/k8-bus_64.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/k8-bus_64.c
+++ linux-2.6/arch/x86/pci/k8-bus_64.c
@@ -12,15 +12,18 @@
  * RED-PEN empty cpus get reported wrong
  */
 
-#define NODE_ID_REGISTER 0x60
-#define NODE_ID(dword) (dword & 0x07)
-#define LDT_BUS_NUMBER_REGISTER_0 0x94
-#define LDT_BUS_NUMBER_REGISTER_1 0xB4
-#define LDT_BUS_NUMBER_REGISTER_2 0xD4
-#define NR_LDT_BUS_NUMBER_REGISTERS 3
-#define SECONDARY_LDT_BUS_NUMBER(dword) ((dword >> 8) & 0xFF)
-#define SUBORDINATE_LDT_BUS_NUMBER(dword) ((dword >> 16) & 0xFF)
+#define NODE_ID(dword) ((dword>>4) & 0x07)
+#define LDT_BUS_NUMBER_REGISTER_0 0xE0
+#define LDT_BUS_NUMBER_REGISTER_1 0xE4
+#define LDT_BUS_NUMBER_REGISTER_2 0xE8
+#define LDT_BUS_NUMBER_REGISTER_3 0xEC
+#define NR_LDT_BUS_NUMBER_REGISTERS 4
+#define SECONDARY_LDT_BUS_NUMBER(dword) ((dword >> 16) & 0xFF)
+#define SUBORDINATE_LDT_BUS_NUMBER(dword) ((dword >> 24) & 0xFF)
+
 #define PCI_DEVICE_ID_K8HTCONFIG 0x1100
+#define PCI_DEVICE_ID_K8_10H_HTCONFIG 0x1200
+#define PCI_DEVICE_ID_K8_11H_HTCONFIG 0x1300
 
 #define BUS_NR 256
 
@@ -62,12 +65,19 @@ early_fill_mp_bus_to_node(void)
 {
 	int i, j;
 	unsigned slot;
-	u32 ldtbus, nid;
+	u32 ldtbus;
 	u32 id;
-	static int lbnr[3] = {
+	int node;
+	u16 deviceid;
+	u16 vendorid;
+	int min_bus;
+	int max_bus;
+
+	static int lbnr[NR_LDT_BUS_NUMBER_REGISTERS] = {
 		LDT_BUS_NUMBER_REGISTER_0,
 		LDT_BUS_NUMBER_REGISTER_1,
-		LDT_BUS_NUMBER_REGISTER_2
+		LDT_BUS_NUMBER_REGISTER_2,
+		LDT_BUS_NUMBER_REGISTER_3
 	};
 
 	for (i = 0; i < BUS_NR; i++)
@@ -76,38 +86,36 @@ early_fill_mp_bus_to_node(void)
 	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++) {
-			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
-			 * 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++) {
-					int node = NODE_ID(nid);
-					mp_bus_to_node[j] = (unsigned char)node;
-				}
-			}
-		}
+	slot = 0x18;
+	id = read_pci_config(0, slot, 0, PCI_VENDOR_ID);
+
+	vendorid = id & 0xffff;
+	if (vendorid != PCI_VENDOR_ID_AMD)
+		goto out;
+
+	deviceid = (id>>16) & 0xffff;
+	if ((deviceid != PCI_DEVICE_ID_K8HTCONFIG) &&
+	    (deviceid != PCI_DEVICE_ID_K8_10H_HTCONFIG) &&
+	    (deviceid != PCI_DEVICE_ID_K8_11H_HTCONFIG))
+		goto out;
+
+	for (i = 0; i < NR_LDT_BUS_NUMBER_REGISTERS; i++) {
+		ldtbus = read_pci_config(0, slot, 1, lbnr[i]);
+
+		/* Check if that register is enabled for bus range */
+		if ((ldtbus & 7) != 3)
+			continue;
+
+		min_bus = SECONDARY_LDT_BUS_NUMBER(ldtbus);
+		max_bus = SUBORDINATE_LDT_BUS_NUMBER(ldtbus);
+		node = NODE_ID(ldtbus);
+		for (j = min_bus; j <= max_bus; j++)
+			mp_bus_to_node[j] = (unsigned char) node;
 	}
 
+out:
 	for (i = 0; i < BUS_NR; i++) {
-		int node = mp_bus_to_node[i];
+		node = mp_bus_to_node[i];
 		if (node >= 0)
 			printk(KERN_DEBUG "bus: %02x to node: %02x\n", i, node);
 	}

[PATCH 3/8] x86_64: get mp_bus_to_node as early v4

[Yinghai Lu]

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

workaround: pcibus_to_node(bus) is used when we want to get 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 use bus->sysdata for nodeid.

the problem is when pci_add_device is called, bus->sysdata is not assigned
correct nodeid yet. the result will be numa_node always is 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 get_mp_bus_to_node
could get correct node for sysdata in root bus.

in scanning of root bus, all child bus will take parent bus sysdata. So all
pci_device->dev.numa_node will be assigned correctly 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 one update version to pci_sysdata and jeff's pci_domain patch.

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

 arch/i386/pci/Makefile         |    1
 arch/i386/pci/Makefile         |    1
 arch/i386/pci/acpi.c           |   40 +++++--------------
 arch/i386/pci/common.c         |   17 ++++++--
 arch/i386/pci/irq.c            |    4 +
 arch/i386/pci/legacy.c         |    6 ++
 arch/i386/pci/mp_bus_to_node.c |   24 +++++++++++
 arch/x86_64/pci/k8-bus.c       |   83 ++++++++++++++++++++++++++++-------------
 include/asm-i386/pci.h         |    1
 include/asm-i386/topology.h    |   10 ++++
 include/asm-x86_64/pci.h       |    1
 include/asm-x86_64/topology.h  |   13 ++++++
 11 files changed, 138 insertions(+), 62 deletions(-)

Index: linux-2.6/arch/x86/pci/Makefile_32
===================================================================
--- linux-2.6.orig/arch/x86/pci/Makefile_32
+++ linux-2.6/arch/x86/pci/Makefile_32
@@ -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
Index: linux-2.6/arch/x86/pci/acpi.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/acpi.c
+++ linux-2.6/arch/x86/pci/acpi.c
@@ -191,7 +191,10 @@ struct pci_bus * __devinit pci_acpi_scan
 {
 	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
 		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.
@@ -212,22 +226,16 @@ struct pci_bus * __devinit pci_acpi_scan
 	}
 
 	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;
 
 	bus = pci_scan_bus_parented(NULL, busnum, &pci_root_ops, sd);
 	if (!bus)
 		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, sd->node);
 		}
 	}
@@ -235,7 +243,6 @@ struct pci_bus * __devinit pci_acpi_scan
 
 	if (bus && (pci_probe & PCI_USE__CRS))
 		get_current_resources(device, busnum, bus);
-	
 	return bus;
 }
 
Index: linux-2.6/arch/x86/pci/common.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/common.c
+++ linux-2.6/arch/x86/pci/common.c
@@ -396,9 +396,14 @@ struct pci_bus * __devinit pcibios_scan_
 		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;
@@ -541,7 +546,7 @@ void pcibios_disable_device (struct pci_
 		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;
@@ -556,10 +561,15 @@ struct pci_bus *__devinit pci_scan_bus_w
 		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);
+}
Index: linux-2.6/arch/x86/pci/irq.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/irq.c
+++ linux-2.6/arch/x86/pci/irq.c
@@ -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);
 	}
Index: linux-2.6/arch/x86/pci/legacy.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/legacy.c
+++ linux-2.6/arch/x86/pci/legacy.c
@@ -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
 		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;
 			}
 		}
Index: linux-2.6/arch/x86/pci/mp_bus_to_node.c
===================================================================
--- /dev/null
+++ linux-2.6/arch/x86/pci/mp_bus_to_node.c
@@ -0,0 +1,23 @@
+#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;
+}
Index: linux-2.6/arch/x86/pci/k8-bus_64.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/k8-bus_64.c
+++ linux-2.6/arch/x86/pci/k8-bus_64.c
@@ -1,7 +1,9 @@
 #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 +22,96 @@
 #define SUBORDINATE_LDT_BUS_NUMBER(dword) ((dword >> 16) & 0xFF)
 #define PCI_DEVICE_ID_K8HTCONFIG 0x1100
 
+#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;
+}
+
 /**
- * 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;
 	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);
+	}
 	return 0;
 }
 
-fs_initcall(fill_mp_bus_to_cpumask);
+postcore_initcall(early_fill_mp_bus_to_node);
Index: linux-2.6/include/asm-x86/pci.h
===================================================================
--- linux-2.6.orig/include/asm-x86/pci.h
+++ linux-2.6/include/asm-x86/pci.h
@@ -20,6 +20,8 @@ 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)
Index: linux-2.6/include/asm-x86/topology.h
===================================================================
--- linux-2.6.orig/include/asm-x86/topology.h
+++ linux-2.6/include/asm-x86/topology.h
@@ -165,8 +165,19 @@ extern int __node_distance(int, int);
 #define node_distance(a, b) __node_distance(a, b)
 #endif
 
+int get_mp_bus_to_node(int busnum);
+void set_mp_bus_to_node(int busnum, int node);
+
 #else /* CONFIG_NUMA */
 
+static inline int get_mp_bus_to_node(int busnum)
+{
+	return 0;
+}
+static inline void set_mp_bus_to_node(int busnum, int node)
+{
+}
+
 #include <asm-generic/topology.h>
 
 #endif

[PATCH 5/8] try parent numa_node at first before using default v2

[Yinghai Lu]

in the device_add, we try to use use parent numa_node.
need to make sure pci root bus's bridge device numa_node is set.
then we could use device->numa_node direclty for all device.
and don't need to call pcibus_to_node().

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

Index: linux-2.6/drivers/pci/probe.c
===================================================================
--- linux-2.6.orig/drivers/pci/probe.c
+++ linux-2.6/drivers/pci/probe.c
@@ -964,7 +964,6 @@ void pci_device_add(struct pci_dev *dev,
 	dev->dev.release = pci_release_dev;
 	pci_dev_get(dev);
 
-	set_dev_node(&dev->dev, pcibus_to_node(bus));
 	dev->dev.dma_mask = &dev->dma_mask;
 	dev->dev.dma_parms = &dev->dma_parms;
 	dev->dev.coherent_dma_mask = 0xffffffffull;
@@ -1116,6 +1115,9 @@ struct pci_bus * pci_create_bus(struct d
 		goto dev_reg_err;
 	b->bridge = get_device(dev);
 
+	if (!parent)
+		set_dev_node(b->bridge, pcibus_to_node(b));
+
 	b->dev.class = &pcibus_class;
 	b->dev.parent = b->bridge;
 	sprintf(b->dev.bus_id, "%04x:%02x", pci_domain_nr(b), bus);
Index: linux-2.6/drivers/base/core.c
===================================================================
--- linux-2.6.orig/drivers/base/core.c
+++ linux-2.6/drivers/base/core.c
@@ -788,6 +788,10 @@ int device_add(struct device *dev)
 	parent = get_device(dev->parent);
 	setup_parent(dev, parent);
 
+	/* use parent numa_node */
+	if (parent)
+		set_dev_node(dev, dev_to_node(parent));
+
 	/* first, register with generic layer. */
 	error = kobject_add(&dev->kobj, dev->kobj.parent, "%s", dev->bus_id);
 	if (error)
@@ -1339,8 +1343,11 @@ int device_move(struct device *dev, stru
 	dev->parent = new_parent;
 	if (old_parent)
 		klist_remove(&dev->knode_parent);
-	if (new_parent)
+	if (new_parent) {
 		klist_add_tail(&dev->knode_parent, &new_parent->klist_children);
+		set_dev_node(dev, dev_to_node(new_parent));
+	}
+
 	if (!dev->class)
 		goto out_put;
 	error = device_move_class_links(dev, old_parent, new_parent);
@@ -1350,9 +1357,12 @@ int device_move(struct device *dev, stru
 		if (!kobject_move(&dev->kobj, &old_parent->kobj)) {
 			if (new_parent)
 				klist_remove(&dev->knode_parent);
-			if (old_parent)
+			dev->parent = old_parent;
+			if (old_parent) {
 				klist_add_tail(&dev->knode_parent,
 					       &old_parent->klist_children);
+				set_dev_node(dev, dev_to_node(old_parent));
+			}
 		}
 		cleanup_glue_dir(dev, new_parent_kobj);
 		put_device(new_parent);

[PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[Yinghai Lu]

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

Index: linux-2.6/net/core/skbuff.c
===================================================================
--- linux-2.6.orig/net/core/skbuff.c
+++ linux-2.6/net/core/skbuff.c
@@ -252,7 +252,7 @@ nodata:
 struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
 		unsigned int length, gfp_t gfp_mask)
 {
-	int node = dev->dev.parent ? dev_to_node(dev->dev.parent) : -1;
+	int node = dev_to_node(&dev->dev);
 	struct sk_buff *skb;
 
 	skb = __alloc_skb(length + NET_SKB_PAD, gfp_mask | __GFP_NOTRACK,

[PATCH 7/8] x86_64: get boot_cpu_id as early for k8_scan_nodes

[Yinghai Lu]

When acpi=off or there is no SRAT defined, apicid_to_node is got from K8
Northbridge PCI configuration space in k8_scan_nodes() in
arch/x86_64/mm/k8toplogy.c.
The problem is that it assumes bsp apic id is 0 at that point.
For four socket system with Quad core cpus installed, all cpus apic id
is offset by 4, and bsp apic id is 4.
For eight socket system with dual core cpus installed, all cpus apic id
is offset by 2, and bsp apic id is 2.

We need get boot_cpu_id --- bsp apic id, before k8_scan_nodes by called.

So create early_acpi_boot_init and early_get_smp_config for get boot_cpu_id.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

 arch/x86_64/mm/k8topology.c    |    8 +++
 b/arch/i386/kernel/acpi/boot.c |   70 ++++++++++++++++++++++++++++++++
 b/arch/x86_64/kernel/apic.c    |   29 +++++++++++--
 b/arch/x86_64/kernel/mpparse.c |   89 ++++++++++++++++++++++++++++++-----------
 b/arch/x86_64/kernel/setup.c   |   21 +++++++++
 b/include/asm-x86_64/apic.h    |    2
 b/include/asm-x86_64/mpspec.h  |    2
 b/include/linux/acpi.h         |    5 ++
 8 files changed, 198 insertions(+), 28 deletions(-)

Index: linux-2.6/arch/x86/mm/k8topology_64.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/k8topology_64.c
+++ linux-2.6/arch/x86/mm/k8topology_64.c
@@ -13,12 +13,15 @@
 #include <linux/nodemask.h>
 #include <asm/io.h>
 #include <linux/pci_ids.h>
+#include <linux/acpi.h>
 #include <asm/types.h>
 #include <asm/mmzone.h>
 #include <asm/proto.h>
 #include <asm/e820.h>
 #include <asm/pci-direct.h>
 #include <asm/numa.h>
+#include <asm/mpspec.h>
+#include <asm/apic.h>
 
 static __init int find_northbridge(void)
 {
@@ -44,6 +47,30 @@ static __init int find_northbridge(void)
 	return -1;
 }
 
+static __init void early_get_boot_cpu_id(void)
+{
+	/*
+	 * need to get boot_cpu_id so can use that to create apicid_to_node
+	 * in k8_scan_nodes()
+	 */
+	/*
+	 * Find possible boot-time SMP configuration:
+	 */
+	early_find_smp_config();
+#ifdef CONFIG_ACPI
+	/*
+	 * Read APIC information from ACPI tables.
+	 */
+	early_acpi_boot_init();
+#endif
+	/*
+	 * get boot-time SMP configuration:
+	 */
+	if (smp_found_config)
+		early_get_smp_config();
+	early_init_lapic_mapping();
+}
+
 int __init k8_scan_nodes(unsigned long start, unsigned long end)
 {
 	unsigned long prevbase;
@@ -56,6 +83,7 @@ int __init k8_scan_nodes(unsigned long s
 	unsigned cores;
 	unsigned bits;
 	int j;
+	unsigned apicid_base;
 
 	if (!early_pci_allowed())
 		return -1;
@@ -174,11 +202,18 @@ int __init k8_scan_nodes(unsigned long s
 	/* use the coreid bits from early_identify_cpu */
 	bits = boot_cpu_data.x86_coreid_bits;
 	cores = (1<<bits);
+	apicid_base = 0;
+	/* need to get boot_cpu_id early for system with apicid lifting */
+	early_get_boot_cpu_id();
+	if (boot_cpu_id > 0) {
+		printk(KERN_INFO "BSP APIC ID: %02x\n", boot_cpu_id);
+		apicid_base = boot_cpu_id;
+	}
 
 	for (i = 0; i < 8; i++) {
 		if (nodes[i].start != nodes[i].end) {
 			nodeid = nodeids[i];
-			for (j = 0; j < cores; j++)
+			for (j = apicid_base; j < cores + apicid_base; j++)
 				apicid_to_node[(nodeid << bits) + j] = i;
 			setup_node_bootmem(i, nodes[i].start, nodes[i].end);
 		}
Index: linux-2.6/arch/x86/kernel/acpi/boot.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/acpi/boot.c
+++ linux-2.6/arch/x86/kernel/acpi/boot.c
@@ -731,6 +731,32 @@ static int __init acpi_parse_fadt(struct
  * Parse LAPIC entries in MADT
  * returns 0 on success, < 0 on error
  */
+static int __init early_acpi_parse_madt_lapic_addr_ovr(void)
+{
+	int count;
+
+	if (!cpu_has_apic)
+		return -ENODEV;
+
+	/*
+	 * Note that the LAPIC address is obtained from the MADT (32-bit value)
+	 * and (optionally) overriden by a LAPIC_ADDR_OVR entry (64-bit value).
+	 */
+
+	count =
+	    acpi_table_parse_madt(ACPI_MADT_TYPE_LOCAL_APIC_OVERRIDE,
+				  acpi_parse_lapic_addr_ovr, 0);
+	if (count < 0) {
+		printk(KERN_ERR PREFIX
+		       "Error parsing LAPIC address override entry\n");
+		return count;
+	}
+
+	mp_register_lapic_address(acpi_lapic_addr);
+
+	return count;
+}
+
 static int __init acpi_parse_madt_lapic_entries(void)
 {
 	int count;
@@ -857,6 +883,33 @@ static inline int acpi_parse_madt_ioapic
 }
 #endif	/* !CONFIG_X86_IO_APIC */
 
+static void __init early_acpi_process_madt(void)
+{
+#ifdef CONFIG_X86_LOCAL_APIC
+	int error;
+
+	if (!acpi_table_parse(ACPI_SIG_MADT, acpi_parse_madt)) {
+
+		/*
+		 * Parse MADT LAPIC entries
+		 */
+		error = early_acpi_parse_madt_lapic_addr_ovr();
+		if (!error) {
+			acpi_lapic = 1;
+			smp_found_config = 1;
+		}
+		if (error == -EINVAL) {
+			/*
+			 * Dell Precision Workstation 410, 610 come here.
+			 */
+			printk(KERN_ERR PREFIX
+			       "Invalid BIOS MADT, disabling ACPI\n");
+			disable_acpi();
+		}
+	}
+#endif
+}
+
 static void __init acpi_process_madt(void)
 {
 #ifdef CONFIG_X86_LOCAL_APIC
@@ -1189,6 +1242,23 @@ int __init acpi_boot_table_init(void)
 	return 0;
 }
 
+int __init early_acpi_boot_init(void)
+{
+	/*
+	 * If acpi_disabled, bail out
+	 * One exception: acpi=ht continues far enough to enumerate LAPICs
+	 */
+	if (acpi_disabled && !acpi_ht)
+		return 1;
+
+	/*
+	 * Process the Multiple APIC Description Table (MADT), if present
+	 */
+	early_acpi_process_madt();
+
+	return 0;
+}
+
 int __init acpi_boot_init(void)
 {
 	/*
Index: linux-2.6/arch/x86/kernel/mpparse_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/mpparse_64.c
+++ linux-2.6/arch/x86/kernel/mpparse_64.c
@@ -217,8 +217,7 @@ static void __init MP_lintsrc_info (stru
 /*
  * Read/parse the MPC
  */
-
-static int __init smp_read_mpc(struct mp_config_table *mpc)
+static int __init smp_read_mpc(struct mp_config_table *mpc, unsigned early)
 {
 	char str[16];
 	int count=sizeof(*mpc);
@@ -259,6 +258,9 @@ static int __init smp_read_mpc(struct mp
 	if (!acpi_lapic)
 		mp_lapic_addr = mpc->mpc_lapic;
 
+	if (early)
+		return 1;
+
 	/*
 	 *	Now process the configuration blocks.
 	 */
@@ -470,27 +472,38 @@ static struct intel_mp_floating *mpf_fou
 /*
  * Scan the memory blocks for an SMP configuration block.
  */
-void __init get_smp_config (void)
+static void __init __get_smp_config(unsigned early)
 {
 	struct intel_mp_floating *mpf = mpf_found;
 
+	if (acpi_lapic && early)
+		return;
 	/*
- 	 * ACPI supports both logical (e.g. Hyper-Threading) and physical 
- 	 * processors, where MPS only supports physical.
- 	 */
- 	if (acpi_lapic && acpi_ioapic) {
- 		printk(KERN_INFO "Using ACPI (MADT) for SMP configuration information\n");
- 		return;
-	}
- 	else if (acpi_lapic)
- 		printk(KERN_INFO "Using ACPI for processor (LAPIC) configuration information\n");
+	 * ACPI supports both logical (e.g. Hyper-Threading) and physical
+	 * processors, where MPS only supports physical.
+	 */
+	if (acpi_lapic && acpi_ioapic) {
+		printk(KERN_INFO "Using ACPI (MADT) for SMP configuration "
+			"information\n");
+		return;
+	} else if (acpi_lapic)
+		printk(KERN_INFO "Using ACPI for processor (LAPIC) "
+			"configuration information\n");
 
-	printk("Intel MultiProcessor Specification v1.%d\n", mpf->mpf_specification);
+	printk(KERN_INFO "Intel MultiProcessor Specification v1.%d\n",
+		mpf->mpf_specification);
 
 	/*
 	 * Now see if we need to read further.
 	 */
 	if (mpf->mpf_feature1 != 0) {
+		if (early) {
+			/*
+			 * local APIC has default address
+			 */
+			mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
+			return;
+		}
 
 		printk(KERN_INFO "Default MP configuration #%d\n", mpf->mpf_feature1);
 		construct_default_ISA_mptable(mpf->mpf_feature1);
@@ -501,12 +514,15 @@ void __init get_smp_config (void)
 		 * Read the physical hardware table.  Anything here will
 		 * override the defaults.
 		 */
-		if (!smp_read_mpc(phys_to_virt(mpf->mpf_physptr))) {
+		if (!smp_read_mpc(phys_to_virt(mpf->mpf_physptr), early)) {
 			smp_found_config = 0;
 			printk(KERN_ERR "BIOS bug, MP table errors detected!...\n");
 			printk(KERN_ERR "... disabling SMP support. (tell your hw vendor)\n");
 			return;
 		}
+
+		if (early)
+			return;
 		/*
 		 * If there are no explicit MP IRQ entries, then we are
 		 * broken.  We set up most of the low 16 IO-APIC pins to
@@ -528,13 +544,25 @@ void __init get_smp_config (void)
 	} else
 		BUG();
 
-	printk(KERN_INFO "Processors: %d\n", num_processors);
+	if (!early)
+		printk(KERN_INFO "Processors: %d\n", num_processors);
 	/*
 	 * Only use the first configuration found.
 	 */
 }
 
-static int __init smp_scan_config (unsigned long base, unsigned long length)
+void __init early_get_smp_config(void)
+{
+	__get_smp_config(1);
+}
+
+void __init get_smp_config(void)
+{
+	__get_smp_config(0);
+}
+
+static int __init smp_scan_config(unsigned long base, unsigned long length,
+				  unsigned reserve)
 {
 	extern void __bad_mpf_size(void); 
 	unsigned int *bp = phys_to_virt(base);
@@ -553,10 +581,15 @@ static int __init smp_scan_config (unsig
 				|| (mpf->mpf_specification == 4)) ) {
 
 			smp_found_config = 1;
+			mpf_found = mpf;
+
+			if (!reserve)
+				return 1;
+
 			reserve_bootmem_generic(virt_to_phys(mpf), PAGE_SIZE);
 			if (mpf->mpf_physptr)
-				reserve_bootmem_generic(mpf->mpf_physptr, PAGE_SIZE);
-			mpf_found = mpf;
+				reserve_bootmem_generic(mpf->mpf_physptr,
+							PAGE_SIZE);
 			return 1;
 		}
 		bp += 4;
@@ -565,7 +598,7 @@ static int __init smp_scan_config (unsig
 	return 0;
 }
 
-void __init find_smp_config(void)
+static void __init __find_smp_config(unsigned reserve)
 {
 	unsigned int address;
 
@@ -577,9 +610,9 @@ void __init find_smp_config(void)
 	 * 2) Scan the top 1K of base RAM
 	 * 3) Scan the 64K of bios
 	 */
-	if (smp_scan_config(0x0,0x400) ||
-		smp_scan_config(639*0x400,0x400) ||
-			smp_scan_config(0xF0000,0x10000))
+	if (smp_scan_config(0x0, 0x400, reserve) ||
+		smp_scan_config(639*0x400, 0x400, reserve) ||
+			smp_scan_config(0xF0000, 0x10000, reserve))
 		return;
 	/*
 	 * If it is an SMP machine we should know now.
@@ -596,13 +629,23 @@ void __init find_smp_config(void)
 
 	address = *(unsigned short *)phys_to_virt(0x40E);
 	address <<= 4;
-	if (smp_scan_config(address, 0x1000))
+	if (smp_scan_config(address, 0x1000, reserve))
 		return;
 
 	/* If we have come this far, we did not find an MP table  */
 	 printk(KERN_INFO "No mptable found.\n");
 }
 
+void __init early_find_smp_config(void)
+{
+	__find_smp_config(0);
+}
+
+void __init find_smp_config(void)
+{
+	__find_smp_config(1);
+}
+
 /* --------------------------------------------------------------------------
                             ACPI-based MP Configuration
    -------------------------------------------------------------------------- */
Index: linux-2.6/include/linux/acpi.h
===================================================================
--- linux-2.6.orig/include/linux/acpi.h
+++ linux-2.6/include/linux/acpi.h
@@ -79,6 +79,7 @@ typedef int (*acpi_table_handler) (struc
 typedef int (*acpi_table_entry_handler) (struct acpi_subtable_header *header, const unsigned long end);
 
 char * __acpi_map_table (unsigned long phys_addr, unsigned long size);
+int early_acpi_boot_init(void);
 int acpi_boot_init (void);
 int acpi_boot_table_init (void);
 int acpi_numa_init (void);
@@ -235,6 +236,10 @@ int acpi_check_mem_region(resource_size_
 
 #else	/* CONFIG_ACPI */
 
+static inline int early_acpi_boot_init(void)
+{
+	return 0;
+}
 static inline int acpi_boot_init(void)
 {
 	return 0;
Index: linux-2.6/include/asm-x86/mpspec.h
===================================================================
--- linux-2.6.orig/include/asm-x86/mpspec.h
+++ linux-2.6/include/asm-x86/mpspec.h
@@ -26,6 +26,9 @@ extern int pic_mode;
 
 extern DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);
 
+extern void early_find_smp_config(void);
+extern void early_get_smp_config(void);
+
 #endif
 
 extern int mp_bus_id_to_pci_bus[MAX_MP_BUSSES];
Index: linux-2.6/include/asm-x86/apic.h
===================================================================
--- linux-2.6.orig/include/asm-x86/apic.h
+++ linux-2.6/include/asm-x86/apic.h
@@ -124,6 +124,7 @@ extern void enable_NMI_through_LVT0(void
  */
 #ifdef CONFIG_X86_64
 extern void setup_apic_routing(void);
+extern void early_init_lapic_mapping(void);
 #endif
 
 extern u8 setup_APIC_eilvt_mce(u8 vector, u8 msg_type, u8 mask);
Index: linux-2.6/arch/x86/kernel/apic_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/apic_64.c
+++ linux-2.6/arch/x86/kernel/apic_64.c
@@ -861,6 +861,30 @@ static int __init detect_init_APIC(void)
 	return 0;
 }
 
+void __init early_init_lapic_mapping(void)
+{
+	unsigned long apic_phys;
+
+	/*
+	 * If no local APIC can be found then go out
+	 * : it means there is no mpatable and MADT
+	 */
+	if (!smp_found_config)
+		return;
+
+	apic_phys = mp_lapic_addr;
+
+	set_fixmap_nocache(FIX_APIC_BASE, apic_phys);
+	apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n",
+				 APIC_BASE, apic_phys);
+
+	/*
+	 * Fetch the APIC ID of the BSP in case we have a
+	 * default configuration (or the MP table is broken).
+	 */
+	boot_cpu_id = GET_APIC_ID(apic_read(APIC_ID));
+}
+
 /**
  * init_apic_mappings - initialize APIC mappings
  */

[PATCH 8/8] x86_64: multi pci root bus with different io resource range

[Yinghai Lu]

scan AMD opteron io/mmio routing to make sure every pci root bus get correct
resource range. So later pci scan could assign correct resource to device
with unassigned resource.

this some kind make up for system without _CRS for multi pci root bus.

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

Index: linux-2.6/arch/x86/pci/k8-bus_64.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/k8-bus_64.c
+++ linux-2.6/arch/x86/pci/k8-bus_64.c
@@ -7,23 +7,31 @@
 
 /*
  * This discovers the pcibus <-> node mapping on AMD K8.
- *
- * RED-PEN need to call this again on PCI hotplug
- * RED-PEN empty cpus get reported wrong
+ * also get peer root bus resource for io,mmio
  */
 
-#define NODE_ID(dword) ((dword>>4) & 0x07)
-#define LDT_BUS_NUMBER_REGISTER_0 0xE0
-#define LDT_BUS_NUMBER_REGISTER_1 0xE4
-#define LDT_BUS_NUMBER_REGISTER_2 0xE8
-#define LDT_BUS_NUMBER_REGISTER_3 0xEC
-#define NR_LDT_BUS_NUMBER_REGISTERS 4
-#define SECONDARY_LDT_BUS_NUMBER(dword) ((dword >> 16) & 0xFF)
-#define SUBORDINATE_LDT_BUS_NUMBER(dword) ((dword >> 24) & 0xFF)
-
-#define PCI_DEVICE_ID_K8HTCONFIG 0x1100
-#define PCI_DEVICE_ID_K8_10H_HTCONFIG 0x1200
-#define PCI_DEVICE_ID_K8_11H_HTCONFIG 0x1300
+
+/*
+ * sub bus (transparent) will use entres from 3 to store extra from root,
+ * so need to make sure have enought slot there, increase PCI_BUS_NUM_RESOURCES?
+ */
+#define RES_NUM 16
+struct pci_root_info {
+	char name[12];
+	unsigned int res_num;
+	struct resource res[RES_NUM];
+	int bus_min;
+	int bus_max;
+	int node;
+	int link;
+};
+
+/* 4 at this time, it may become to 32 */
+#define PCI_ROOT_NR 4
+static int pci_root_num;
+static struct pci_root_info pci_root_info[PCI_ROOT_NR];
+
+#ifdef CONFIG_NUMA
 
 #define BUS_NR 256
 
@@ -53,6 +61,135 @@ int get_mp_bus_to_node(int busnum)
 
 	return node;
 }
+#endif
+
+void set_pci_bus_resources_arch_default(struct pci_bus *b)
+{
+	int i;
+	int j;
+	struct pci_root_info *info;
+
+	if (!pci_root_num)
+		return;
+
+	for (i = 0; i < pci_root_num; i++) {
+		if (pci_root_info[i].bus_min == b->number)
+			break;
+	}
+
+	if (i == pci_root_num)
+		return;
+
+	info = &pci_root_info[i];
+	for (j = 0; j < info->res_num; j++) {
+		struct resource *res;
+		struct resource *root;
+
+		res = &info->res[j];
+		b->resource[j] = res;
+		if (res->flags & IORESOURCE_IO)
+			root = &ioport_resource;
+		else
+			root = &iomem_resource;
+		insert_resource(root, res);
+	}
+}
+
+#define RANGE_NUM 16
+
+struct res_range {
+	size_t start;
+	size_t end;
+};
+
+static void __init update_range(struct res_range *range, size_t start,
+				size_t end)
+{
+	int i;
+	int j;
+
+	for (j = 0; j < RANGE_NUM; j++) {
+		if (!range[j].end)
+			continue;
+		if (start == range[j].start && end < range[j].end) {
+			range[j].start = end + 1;
+			break;
+		} else if (start == range[j].start && end == range[j].end) {
+			range[j].start = 0;
+			range[j].end = 0;
+			break;
+		} else if (start > range[j].start && end == range[j].end) {
+			range[j].end = start - 1;
+			break;
+		} else if (start > range[j].start && end < range[j].end) {
+			/* find the new spare */
+			for (i = 0; i < RANGE_NUM; i++) {
+				if (range[i].end == 0)
+					break;
+			}
+			if (i < RANGE_NUM) {
+				range[i].end = range[j].end;
+				range[i].start = end + 1;
+			} else {
+				printk(KERN_ERR "run of slot in ranges\n");
+			}
+			range[j].end = start - 1;
+			break;
+		}
+	}
+}
+
+static void __init update_res(struct pci_root_info *info, size_t start,
+			      size_t end, unsigned long flags, int merge)
+{
+	int i;
+	struct resource *res;
+
+	if (!merge)
+		goto addit;
+
+	/* try to merge it with old one */
+	for (i = 0; i < info->res_num; i++) {
+		res = &info->res[i];
+		if (res->flags != flags)
+			continue;
+		if (res->end + 1 == start) {
+			res->end = end;
+			return;
+		} else if (end + 1 == res->start) {
+			res->start = start;
+			return;
+		}
+	}
+
+addit:
+
+	/* need to add that */
+	if (info->res_num >= RES_NUM)
+		return;
+
+	res = &info->res[info->res_num];
+	res->name = info->name;
+	res->flags = flags;
+	res->start = start;
+	res->end = end;
+	res->child = NULL;
+	info->res_num++;
+}
+
+struct pci_hostbridge_probe {
+	u32 bus;
+	u32 slot;
+	u32 vendor;
+	u32 device;
+};
+
+static struct pci_hostbridge_probe pci_probes[] __initdata = {
+	{ 0, 0x18, PCI_VENDOR_ID_AMD, 0x1100 },
+	{ 0, 0x18, PCI_VENDOR_ID_AMD, 0x1200 },
+	{ 0xff, 0, PCI_VENDOR_ID_AMD, 0x1200 },
+	{ 0, 0x18, PCI_VENDOR_ID_AMD, 0x1300 },
+};
 
 /**
  * early_fill_mp_bus_to_node()
@@ -60,25 +197,25 @@ int get_mp_bus_to_node(int busnum)
  * fills the mp_bus_to_cpumask array based according to the LDT Bus Number
  * Registers found in the K8 northbridge
  */
-__init static int
-early_fill_mp_bus_to_node(void)
+static int __init early_fill_mp_bus_info(void)
 {
-	int i, j;
+	int i;
+	int j;
+	unsigned bus;
 	unsigned slot;
-	u32 ldtbus;
-	u32 id;
+	int found;
 	int node;
-	u16 deviceid;
-	u16 vendorid;
-	int min_bus;
-	int max_bus;
-
-	static int lbnr[NR_LDT_BUS_NUMBER_REGISTERS] = {
-		LDT_BUS_NUMBER_REGISTER_0,
-		LDT_BUS_NUMBER_REGISTER_1,
-		LDT_BUS_NUMBER_REGISTER_2,
-		LDT_BUS_NUMBER_REGISTER_3
-	};
+	int link;
+	int def_node;
+	int def_link;
+	struct pci_root_info *info;
+	u32 reg;
+	struct resource *res;
+	size_t start;
+	size_t end;
+	struct res_range range[RANGE_NUM];
+	u64 val;
+	u32 address;
 
 	for (i = 0; i < BUS_NR; i++)
 		mp_bus_to_node[i] = -1;
@@ -86,40 +223,221 @@ early_fill_mp_bus_to_node(void)
 	if (!early_pci_allowed())
 		return -1;
 
-	slot = 0x18;
-	id = read_pci_config(0, slot, 0, PCI_VENDOR_ID);
+	found = 0;
+	for (i = 0; i < ARRAY_SIZE(pci_probes); i++) {
+		u32 id;
+		u16 device;
+		u16 vendor;
+
+		bus = pci_probes[i].bus;
+		slot = pci_probes[i].slot;
+		id = read_pci_config(bus, slot, 0, PCI_VENDOR_ID);
+
+		vendor = id & 0xffff;
+		device = (id>>16) & 0xffff;
+		if (pci_probes[i].vendor == vendor &&
+		    pci_probes[i].device == device) {
+			found = 1;
+			break;
+		}
+	}
 
-	vendorid = id & 0xffff;
-	if (vendorid != PCI_VENDOR_ID_AMD)
-		goto out;
-
-	deviceid = (id>>16) & 0xffff;
-	if ((deviceid != PCI_DEVICE_ID_K8HTCONFIG) &&
-	    (deviceid != PCI_DEVICE_ID_K8_10H_HTCONFIG) &&
-	    (deviceid != PCI_DEVICE_ID_K8_11H_HTCONFIG))
-		goto out;
+	if (!found)
+		return 0;
 
-	for (i = 0; i < NR_LDT_BUS_NUMBER_REGISTERS; i++) {
-		ldtbus = read_pci_config(0, slot, 1, lbnr[i]);
+	pci_root_num = 0;
+	for (i = 0; i < 4; i++) {
+		int min_bus;
+		int max_bus;
+		reg = read_pci_config(bus, slot, 1, 0xe0 + (i << 2));
 
 		/* Check if that register is enabled for bus range */
-		if ((ldtbus & 7) != 3)
+		if ((reg & 7) != 3)
 			continue;
 
-		min_bus = SECONDARY_LDT_BUS_NUMBER(ldtbus);
-		max_bus = SUBORDINATE_LDT_BUS_NUMBER(ldtbus);
-		node = NODE_ID(ldtbus);
+		min_bus = (reg >> 16) & 0xff;
+		max_bus = (reg >> 24) & 0xff;
+		node = (reg >> 4) & 0x07;
+#ifdef CONFIG_NUMA
 		for (j = min_bus; j <= max_bus; j++)
 			mp_bus_to_node[j] = (unsigned char) node;
+#endif
+		link = (reg >> 8) & 0x03;
+
+		info = &pci_root_info[pci_root_num];
+		info->bus_min = min_bus;
+		info->bus_max = max_bus;
+		info->node = node;
+		info->link = link;
+		sprintf(info->name, "PCI Bus #%02x", min_bus);
+		pci_root_num++;
+	}
+
+	/* get the default node and link for left over res */
+	reg = read_pci_config(bus, slot, 0, 0x60);
+	def_node = (reg >> 8) & 0x07;
+	reg = read_pci_config(bus, slot, 0, 0x64);
+	def_link = (reg >> 8) & 0x03;
+
+	memset(range, 0, sizeof(range));
+	range[0].end = 0xffff;
+	/* io port resource */
+	for (i = 0; i < 4; i++) {
+		reg = read_pci_config(bus, slot, 1, 0xc0 + (i << 3));
+		if (!(reg & 3))
+			continue;
+
+		start = reg & 0xfff000;
+		reg = read_pci_config(bus, slot, 1, 0xc4 + (i << 3));
+		node = reg & 0x07;
+		link = (reg >> 4) & 0x03;
+		end = (reg & 0xfff000) | 0xfff;
+
+		/* find the position */
+		for (j = 0; j < pci_root_num; j++) {
+			info = &pci_root_info[j];
+			if (info->node == node && info->link == link)
+				break;
+		}
+		if (j == pci_root_num)
+			continue; /* not found */
+
+		info = &pci_root_info[j];
+		update_res(info, start, end, IORESOURCE_IO, 0);
+		update_range(range, start, end);
+	}
+	/* add left over io port range to def node/link, [0, 0xffff] */
+	/* find the position */
+	for (j = 0; j < pci_root_num; j++) {
+		info = &pci_root_info[j];
+		if (info->node == def_node && info->link == def_link)
+			break;
 	}
+	if (j < pci_root_num) {
+		info = &pci_root_info[j];
+		for (i = 0; i < RANGE_NUM; i++) {
+			if (!range[i].end)
+				continue;
+
+			update_res(info, range[i].start, range[i].end,
+				   IORESOURCE_IO, 1);
+		}
+	}
+
+	memset(range, 0, sizeof(range));
+	/* 0xfd00000000-0xffffffffff for HT */
+	/* 0xfc00000000-0xfcffffffff for Family 10h mmconfig*/
+	range[0].end = 0xfbffffffffULL;
+
+	/* need to take out [0, TOM) for RAM*/
+	address = MSR_K8_TOP_MEM1;
+	rdmsrl(address, val);
+	end = (val & 0xffffff8000000ULL);
+	printk(KERN_INFO "TOM: %016lx aka %ldM\n", end, end>>20);
+	if (end < (1ULL<<32))
+		update_range(range, 0, end - 1);
+
+	/* mmio resource */
+	for (i = 0; i < 8; i++) {
+		reg = read_pci_config(bus, slot, 1, 0x80 + (i << 3));
+		if (!(reg & 3))
+			continue;
 
-out:
+		start = reg & 0xffffff00; /* 39:16 on 31:8*/
+		start <<= 8;
+		reg = read_pci_config(bus, slot, 1, 0x84 + (i << 3));
+		node = reg & 0x07;
+		link = (reg >> 4) & 0x03;
+		end = (reg & 0xffffff00);
+		end <<= 8;
+		end |= 0xffff;
+
+		/* find the position */
+		for (j = 0; j < pci_root_num; j++) {
+			info = &pci_root_info[j];
+			if (info->node == node && info->link == link)
+				break;
+		}
+		if (j == pci_root_num)
+			continue; /* not found */
+
+		info = &pci_root_info[j];
+		update_res(info, start, end, IORESOURCE_MEM, 0);
+		update_range(range, start, end);
+	}
+
+	/* need to take out [4G, TOM2) for RAM*/
+	/* SYS_CFG */
+	address = MSR_K8_SYSCFG;
+	rdmsrl(address, val);
+	/* TOP_MEM2 is enabled? */
+	if (val & (1<<21)) {
+		/* TOP_MEM2 */
+		address = MSR_K8_TOP_MEM2;
+		rdmsrl(address, val);
+		end = (val & 0xffffff8000000ULL);
+		printk(KERN_INFO "TOM2: %016lx aka %ldM\n", end, end>>20);
+		update_range(range, 1ULL<<32, end - 1);
+	}
+
+	/*
+	 * add left over mmio range to def node/link ?
+	 * that is tricky, just record range in from start_min to 4G
+	 */
+	for (j = 0; j < pci_root_num; j++) {
+		info = &pci_root_info[j];
+		if (info->node == def_node && info->link == def_link)
+			break;
+	}
+	if (j < pci_root_num) {
+		info = &pci_root_info[j];
+
+		for (i = 0; i < RANGE_NUM; i++) {
+			if (!range[i].end)
+				continue;
+#if 0
+			/* don't use the one start less than 1M */
+			if (range[i].start < 0x100000)
+				continue;
+#endif
+#if 0
+			/* don't use last one near 4G */
+			if (range[i].end == 0xffffffffULL)
+				continue;
+#endif
+
+			update_res(info, range[i].start, range[i].end,
+				   IORESOURCE_MEM, 1);
+		}
+	}
+
+#ifdef CONFIG_NUMA
 	for (i = 0; i < BUS_NR; i++) {
 		node = mp_bus_to_node[i];
 		if (node >= 0)
 			printk(KERN_DEBUG "bus: %02x to node: %02x\n", i, node);
 	}
+#endif
+
+	for (i = 0; i < pci_root_num; i++) {
+		int res_num;
+		int busnum;
+
+		info = &pci_root_info[i];
+		res_num = info->res_num;
+		busnum = info->bus_min;
+		printk(KERN_DEBUG "bus: [%02x,%02x] on node %x link %x\n",
+		       info->bus_min, info->bus_max, info->node, info->link);
+		for (j = 0; j < res_num; j++) {
+			res = &info->res[j];
+			printk(KERN_DEBUG "bus: %02x index %x %s: [%llx, %llx]\n",
+			       busnum, j,
+			       (res->flags & IORESOURCE_IO)?"io port":"mmio",
+			       res->start, res->end);
+		}
+	}
+
 	return 0;
 }
 
-postcore_initcall(early_fill_mp_bus_to_node);
+postcore_initcall(early_fill_mp_bus_info);
Index: linux-2.6/include/asm-x86/topology.h
===================================================================
--- linux-2.6.orig/include/asm-x86/topology.h
+++ linux-2.6/include/asm-x86/topology.h
@@ -25,6 +25,7 @@
 #ifndef _ASM_X86_TOPOLOGY_H
 #define _ASM_X86_TOPOLOGY_H
 
+struct pci_bus;
 #ifdef CONFIG_NUMA
 #include <linux/cpumask.h>
 #include <asm/mpspec.h>
@@ -191,6 +192,8 @@ extern cpumask_t cpu_coregroup_map(int c
 #define topology_thread_siblings(cpu)		(per_cpu(cpu_sibling_map, cpu))
 #endif
 
+void set_pci_bus_resources_arch_default(struct pci_bus *b);
+
 #ifdef CONFIG_SMP
 #define mc_capable()			(boot_cpu_data.x86_max_cores > 1)
 #define smt_capable()			(smp_num_siblings > 1)
Index: linux-2.6/drivers/pci/probe.c
===================================================================
--- linux-2.6.orig/drivers/pci/probe.c
+++ linux-2.6/drivers/pci/probe.c
@@ -1076,6 +1076,10 @@ unsigned int pci_scan_child_bus(struct p
 	return max;
 }
 
+void __attribute__((weak)) set_pci_bus_resources_arch_default(struct pci_bus *b)
+{
+}
+
 struct pci_bus * pci_create_bus(struct device *parent,
 		int bus, struct pci_ops *ops, void *sysdata)
 {
@@ -1135,6 +1139,8 @@ struct pci_bus * pci_create_bus(struct d
 	b->resource[0] = &ioport_resource;
 	b->resource[1] = &iomem_resource;
 
+	set_pci_bus_resources_arch_default(b);
+
 	return b;
 
 dev_create_file_err:
Index: linux-2.6/arch/x86/pci/Makefile_64
===================================================================
--- linux-2.6.orig/arch/x86/pci/Makefile_64
+++ linux-2.6/arch/x86/pci/Makefile_64
@@ -13,5 +13,5 @@ obj-y			+= legacy.o irq.o common.o early
 # mmconfig has a 64bit special
 obj-$(CONFIG_PCI_MMCONFIG) += mmconfig_64.o direct.o mmconfig-shared.o
 
-obj-$(CONFIG_NUMA)	+= k8-bus_64.o
+obj-y		+= k8-bus_64.o
 
Index: linux-2.6/include/linux/pci.h
===================================================================
--- linux-2.6.orig/include/linux/pci.h
+++ linux-2.6/include/linux/pci.h
@@ -247,7 +247,7 @@ static inline void pci_add_saved_cap(str
 #define PCI_NUM_RESOURCES	11
 
 #ifndef PCI_BUS_NUM_RESOURCES
-#define PCI_BUS_NUM_RESOURCES	8
+#define PCI_BUS_NUM_RESOURCES	16
 #endif
 
 #define PCI_REGION_FLAG_MASK	0x0fU	/* These bits of resource flags tell us the PCI region flags */

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[Ingo Molnar]

* Yinghai Lu <Yinghai.Lu@Sun.COM> wrote:

>  struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
>  		unsigned int length, gfp_t gfp_mask)
>  {
> -	int node = dev->dev.parent ? dev_to_node(dev->dev.parent) : -1;
> +	int node = dev_to_node(&dev->dev);

i think this is a fix for the networking folks. (Dave Cc:-ed)

	Ingo

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[David Miller]

From: Ingo Molnar <mingo@elte.hu>
Date: Tue, 19 Feb 2008 12:21:46 +0100

> 
> * Yinghai Lu <Yinghai.Lu@Sun.COM> wrote:
> 
> >  struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
> >  		unsigned int length, gfp_t gfp_mask)
> >  {
> > -	int node = dev->dev.parent ? dev_to_node(dev->dev.parent) : -1;
> > +	int node = dev_to_node(&dev->dev);
> 
> i think this is a fix for the networking folks. (Dave Cc:-ed)

It keeps getting NAK's because it's wrong.

The author of the patch hasn't convinced folks why this is really
necessary, and using the net_device embedded device struct is
definitely wrong here.  It doesn't contain the NUMA node information,
the physical device does, and that is what the parent it.

Re: [PATCH 5/8] try parent numa_node at first before using default v2

[Greg KH]

On Tue, Feb 19, 2008 at 03:20:41AM -0800, Yinghai Lu wrote:
> 
> in the device_add, we try to use use parent numa_node.
> need to make sure pci root bus's bridge device numa_node is set.
> then we could use device->numa_node direclty for all device.
> and don't need to call pcibus_to_node().
> 
> Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

This should be split into different patches, one for the driver core,
and one for the PCI stuff as they are doing totally different things
here.

thanks,

greg k-h

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[Yinghai Lu]

On Tuesday 19 February 2008 03:41:10 am David Miller wrote:
> From: Ingo Molnar <mingo@elte.hu>
> Date: Tue, 19 Feb 2008 12:21:46 +0100
> 
> > 
> > * Yinghai Lu <Yinghai.Lu@Sun.COM> wrote:
> > 
> > >  struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
> > >  		unsigned int length, gfp_t gfp_mask)
> > >  {
> > > -	int node = dev->dev.parent ? dev_to_node(dev->dev.parent) : -1;
> > > +	int node = dev_to_node(&dev->dev);
> > 
> > i think this is a fix for the networking folks. (Dave Cc:-ed)
> 
> It keeps getting NAK's because it's wrong.
> 
> The author of the patch hasn't convinced folks why this is really
> necessary, and using the net_device embedded device struct is
> definitely wrong here.  It doesn't contain the NUMA node information,
> the physical device does, and that is what the parent it.

can you check the 5/8?
that will make sure every struct device get numa_node get assigned.

YH

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[Yinghai Lu]

On Feb 19, 2008 11:42 AM, Yinghai Lu <Yinghai.Lu@sun.com> wrote:
> On Tuesday 19 February 2008 03:41:10 am David Miller wrote:
> > From: Ingo Molnar <mingo@elte.hu>
> > Date: Tue, 19 Feb 2008 12:21:46 +0100
> >
> > >
> > > * Yinghai Lu <Yinghai.Lu@Sun.COM> wrote:
> > >
> > > >  struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
> > > >           unsigned int length, gfp_t gfp_mask)
> > > >  {
> > > > - int node = dev->dev.parent ? dev_to_node(dev->dev.parent) : -1;
> > > > + int node = dev_to_node(&dev->dev);
> > >
> > > i think this is a fix for the networking folks. (Dave Cc:-ed)
> >
> > It keeps getting NAK's because it's wrong.
> >
> > The author of the patch hasn't convinced folks why this is really
> > necessary, and using the net_device embedded device struct is
> > definitely wrong here.  It doesn't contain the NUMA node information,
> > the physical device does, and that is what the parent it.
>
> can you check the 5/8?
> that will make sure every struct device get numa_node get assigned.
>
> YH
>
or need to combine 5 and 6 together?

YH

Re: [PATCH 5/8] try parent numa_node at first before using default v2

[Yinghai Lu]

On Tuesday 19 February 2008 09:54:23 am Greg KH wrote:
> On Tue, Feb 19, 2008 at 03:20:41AM -0800, Yinghai Lu wrote:
> > 
> > in the device_add, we try to use use parent numa_node.
> > need to make sure pci root bus's bridge device numa_node is set.
> > then we could use device->numa_node direclty for all device.
> > and don't need to call pcibus_to_node().
> > 
> > Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
> 
> This should be split into different patches, one for the driver core,
> and one for the PCI stuff as they are doing totally different things
> here.

but one for the driver core, rely on the one for pci stuff.
without the one pci stuff, the one for the driver core doesn't work.

last time you already said it's ok.  I will dig the mail out if needed.

YH

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[David Miller]

From: Yinghai Lu <Yinghai.Lu@Sun.COM>
Date: Tue, 19 Feb 2008 11:42:48 -0800

> On Tuesday 19 February 2008 03:41:10 am David Miller wrote:
> > From: Ingo Molnar <mingo@elte.hu>
> > Date: Tue, 19 Feb 2008 12:21:46 +0100
> > 
> > > 
> > > * Yinghai Lu <Yinghai.Lu@Sun.COM> wrote:
> > > 
> > > >  struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
> > > >  		unsigned int length, gfp_t gfp_mask)
> > > >  {
> > > > -	int node = dev->dev.parent ? dev_to_node(dev->dev.parent) : -1;
> > > > +	int node = dev_to_node(&dev->dev);
> > > 
> > > i think this is a fix for the networking folks. (Dave Cc:-ed)
> > 
> > It keeps getting NAK's because it's wrong.
> > 
> > The author of the patch hasn't convinced folks why this is really
> > necessary, and using the net_device embedded device struct is
> > definitely wrong here.  It doesn't contain the NUMA node information,
> > the physical device does, and that is what the parent it.
> 
> can you check the 5/8?
> that will make sure every struct device get numa_node get assigned.

Why do we need to bother with that if the parent will have the
necessary information for us here?

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[Yinghai Lu]

On Feb 19, 2008 2:55 PM, David Miller <davem@davemloft.net> wrote:
> From: Yinghai Lu <Yinghai.Lu@Sun.COM>
> Date: Tue, 19 Feb 2008 11:42:48 -0800
>
> > On Tuesday 19 February 2008 03:41:10 am David Miller wrote:
> > > From: Ingo Molnar <mingo@elte.hu>
> > > Date: Tue, 19 Feb 2008 12:21:46 +0100
> > >
> > > >
> > > > * Yinghai Lu <Yinghai.Lu@Sun.COM> wrote:
> > > >
> > > > >  struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
> > > > >                 unsigned int length, gfp_t gfp_mask)
> > > > >  {
> > > > > -       int node = dev->dev.parent ? dev_to_node(dev->dev.parent) : -1;
> > > > > +       int node = dev_to_node(&dev->dev);
> > > >
> > > > i think this is a fix for the networking folks. (Dave Cc:-ed)
> > >
> > > It keeps getting NAK's because it's wrong.
> > >
> > > The author of the patch hasn't convinced folks why this is really
> > > necessary, and using the net_device embedded device struct is
> > > definitely wrong here.  It doesn't contain the NUMA node information,
> > > the physical device does, and that is what the parent it.
> >
> > can you check the 5/8?
> > that will make sure every struct device get numa_node get assigned.
>
> Why do we need to bother with that if the parent will have the
> necessary information for us here?

less code?

or some kind of usb or other bus interface. may have several level...

YH

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[Ingo Molnar]

* Yinghai Lu <yhlu.kernel@gmail.com> wrote:

> > > can you check the 5/8? that will make sure every struct device get 
> > > numa_node get assigned.
> >
> > Why do we need to bother with that if the parent will have the 
> > necessary information for us here?
> 
> less code?
> 
> or some kind of usb or other bus interface. may have several level...

you mean it's a small optimization: otherwise every struct device (net 
dev, usb...) need to go back to find another pci device (parent or host 
controller) to use their numa_node.

	Ingo

Re: [PATCH 6/8] net: use numa_node in net_devcice->dev instead of parent

[Yinghai Lu]

On Feb 19, 2008 10:27 PM, Ingo Molnar <mingo@elte.hu> wrote:
>
> * Yinghai Lu <yhlu.kernel@gmail.com> wrote:
>
> > > > can you check the 5/8? that will make sure every struct device get
> > > > numa_node get assigned.
> > >
> > > Why do we need to bother with that if the parent will have the
> > > necessary information for us here?
> >
> > less code?
> >
> > or some kind of usb or other bus interface. may have several level...
>
> you mean it's a small optimization: otherwise every struct device (net
> dev, usb...) need to go back to find another pci device (parent or host
> controller) to use their numa_node.
>
Yes.

YH

[PATCH] x86: skip it if Fam 10h only handle bus 0

[Yinghai Lu]

need to apply after
	x86_64: check MSR to get MMCONFIG for AMD Family 10h Opteron v3

some BIOS only let AMD fam 10h handle bus0, and nvidia mcp55/ck804
to handle other buses. at that case MCFG will cover all over them.

this patch will skip it so to use MCFG later.

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>

Index: linux-2.6/arch/x86/pci/mmconfig-shared.c
===================================================================
--- linux-2.6.orig/arch/x86/pci/mmconfig-shared.c
+++ linux-2.6/arch/x86/pci/mmconfig-shared.c
@@ -123,6 +123,14 @@ static const char __init *pci_mmcfg_amd_
 
 	busnbits = (msr >> FAM10H_MMIO_CONF_BUSRANGE_SHIFT) &
 			 FAM10H_MMIO_CONF_BUSRANGE_MASK;
+
+	/*
+	 * only handle bus 0 ?
+	 * need to skip it
+	 */
+	if (!busnbits)
+		return NULL;
+
 	if (busnbits > 8) {
 		segnbits = busnbits - 8;
 		busnbits = 8;