[PATCH] powerpc: Experimental support for new G5 Macs (#2)

This adds some very basic support for the new machines, including the
Quad G5 (tested), and other new dual core based machines and iMac G5
iSight (untested). This is still experimental !  There is no thermal
control yet, there is no proper handing of MSIs, etc.. but it
boots, I have all 4 cores up on my machine. Compared to the previous
version of this patch, this one adds DART IOMMU support for the U4
chipset and thus should work fine on setups with more than 2Gb of RAM.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>

6 files changed, 449 insertions, 234 deletions

diff --git a/arch/powerpc/platforms/maple/setup.c b/arch/powerpc/platforms/maple/setup.c
index 65fe4c166a6..dd73e38bfb7 100644
--- a/arch/powerpc/platforms/maple/setup.c
+++ b/arch/powerpc/platforms/maple/setup.c
@@ -195,7 +195,7 @@ static void __init maple_init_early(void)
 	/* Setup interrupt mapping options */
 	ppc64_interrupt_controller = IC_OPEN_PIC;
 
-	iommu_init_early_u3();
+	iommu_init_early_dart();
 
 	DBG(" <- maple_init_early\n");
 }
@@ -257,7 +257,7 @@ static int __init maple_probe(int platform)
 	 * occupies having to be broken up so the DART itself is not
 	 * part of the cacheable linar mapping
 	 */
-	alloc_u3_dart_table();
+	alloc_dart_table();
 
 	return 1;
 }
diff --git a/arch/powerpc/platforms/powermac/feature.c b/arch/powerpc/platforms/powermac/feature.c
index b1f896952b1..d2915d64d45 100644
--- a/arch/powerpc/platforms/powermac/feature.c
+++ b/arch/powerpc/platforms/powermac/feature.c
@@ -101,7 +101,8 @@ static const char *macio_names[] =
 	"Keylargo",
 	"Pangea",
 	"Intrepid",
-	"K2"
+	"K2",
+	"Shasta",
 };
 
 
@@ -119,7 +120,7 @@ static const char *macio_names[] =
 static struct device_node *uninorth_node;
 static u32 __iomem *uninorth_base;
 static u32 uninorth_rev;
-static int uninorth_u3;
+static int uninorth_maj;
 static void __iomem *u3_ht;
 
 /*
@@ -1399,8 +1400,15 @@ static long g5_fw_enable(struct device_node *node, long param, long value)
 static long g5_mpic_enable(struct device_node *node, long param, long value)
 {
 	unsigned long flags;
+	struct device_node *parent = of_get_parent(node);
+	int is_u3;
 
-	if (node->parent == NULL || strcmp(node->parent->name, "u3"))
+	if (parent == NULL)
+		return 0;
+	is_u3 = strcmp(parent->name, "u3") == 0 ||
+		strcmp(parent->name, "u4") == 0;
+	of_node_put(parent);
+	if (!is_u3)
 		return 0;
 
 	LOCK(flags);
@@ -1464,7 +1472,7 @@ static long g5_i2s_enable(struct device_node *node, long param, long value)
 		},
 	};
 
-	if (macio->type != macio_keylargo2 /* && macio->type != macio_shasta*/)
+	if (macio->type != macio_keylargo2 && macio->type != macio_shasta)
 		return -ENODEV;
 	if (strncmp(node->name, "i2s-", 4))
 		return -ENODEV;
@@ -1473,11 +1481,9 @@ static long g5_i2s_enable(struct device_node *node, long param, long value)
 	case 0:
 	case 1:
 		break;
-#if 0
 	case 2:
 		if (macio->type == macio_shasta)
 			break;
-#endif
 	default:
 		return -ENODEV;
 	}
@@ -1508,7 +1514,7 @@ static long g5_reset_cpu(struct device_node *node, long param, long value)
 	struct device_node *np;
 
 	macio = &macio_chips[0];
-	if (macio->type != macio_keylargo2)
+	if (macio->type != macio_keylargo2 && macio->type != macio_shasta)
 		return -ENODEV;
 
 	np = find_path_device("/cpus");
@@ -1547,7 +1553,8 @@ static long g5_reset_cpu(struct device_node *node, long param, long value)
  */
 void g5_phy_disable_cpu1(void)
 {
-	UN_OUT(U3_API_PHY_CONFIG_1, 0);
+	if (uninorth_maj == 3)
+		UN_OUT(U3_API_PHY_CONFIG_1, 0);
 }
 #endif /* CONFIG_POWER4 */
 
@@ -2462,6 +2469,14 @@ static struct pmac_mb_def pmac_mb_defs[] = {
 		PMAC_TYPE_POWERMAC_G5_U3L,	g5_features,
 		0,
 	},
+	{	"PowerMac11,2",			"PowerMac G5 Dual Core",
+		PMAC_TYPE_POWERMAC_G5_U3L,	g5_features,
+		0,
+	},
+	{	"PowerMac12,1",			"iMac G5 (iSight)",
+		PMAC_TYPE_POWERMAC_G5_U3L,	g5_features,
+		0,
+	},
 	{       "RackMac3,1",                   "XServe G5",
 		PMAC_TYPE_XSERVE_G5,		g5_features,
 		0,
@@ -2574,6 +2589,11 @@ static int __init probe_motherboard(void)
 		pmac_mb.model_name = "Unknown K2-based";
 		pmac_mb.features = g5_features;
 		break;
+	case macio_shasta:
+		pmac_mb.model_id = PMAC_TYPE_UNKNOWN_SHASTA;
+		pmac_mb.model_name = "Unknown Shasta-based";
+		pmac_mb.features = g5_features;
+		break;
 #endif /* CONFIG_POWER4 */
 	default:
 		return -ENODEV;
@@ -2651,7 +2671,12 @@ static void __init probe_uninorth(void)
 	/* Locate G5 u3 */
 	if (uninorth_node == NULL) {
 		uninorth_node = of_find_node_by_name(NULL, "u3");
-		uninorth_u3 = 1;
+		uninorth_maj = 3;
+	}
+	/* Locate G5 u4 */
+	if (uninorth_node == NULL) {
+		uninorth_node = of_find_node_by_name(NULL, "u4");
+		uninorth_maj = 4;
 	}
 	if (uninorth_node == NULL)
 		return;
@@ -2664,12 +2689,13 @@ static void __init probe_uninorth(void)
 		return;
 	uninorth_base = ioremap(address, 0x40000);
 	uninorth_rev = in_be32(UN_REG(UNI_N_VERSION));
-	if (uninorth_u3)
+	if (uninorth_maj == 3 || uninorth_maj == 4)
 		u3_ht = ioremap(address + U3_HT_CONFIG_BASE, 0x1000);
 
-	printk(KERN_INFO "Found %s memory controller & host bridge,"
-	       " revision: %d\n", uninorth_u3 ? "U3" : "UniNorth",
-	       uninorth_rev);
+	printk(KERN_INFO "Found %s memory controller & host bridge"
+	       " @ 0x%08x revision: 0x%02x\n", uninorth_maj == 3 ? "U3" :
+	       uninorth_maj == 4 ? "U4" : "UniNorth",
+	       (unsigned int)address, uninorth_rev);
 	printk(KERN_INFO "Mapped at 0x%08lx\n", (unsigned long)uninorth_base);
 
 	/* Set the arbitrer QAck delay according to what Apple does
@@ -2677,7 +2703,8 @@ static void __init probe_uninorth(void)
 	if (uninorth_rev < 0x11) {
 		actrl = UN_IN(UNI_N_ARB_CTRL) & ~UNI_N_ARB_CTRL_QACK_DELAY_MASK;
 		actrl |= ((uninorth_rev < 3) ? UNI_N_ARB_CTRL_QACK_DELAY105 :
-			UNI_N_ARB_CTRL_QACK_DELAY) << UNI_N_ARB_CTRL_QACK_DELAY_SHIFT;
+			UNI_N_ARB_CTRL_QACK_DELAY) <<
+			UNI_N_ARB_CTRL_QACK_DELAY_SHIFT;
 		UN_OUT(UNI_N_ARB_CTRL, actrl);
 	}
 
@@ -2685,7 +2712,8 @@ static void __init probe_uninorth(void)
 	 * revs 1.5 to 2.O and Pangea. Seem to toggle the UniN Maxbus/PCI
 	 * memory timeout
 	 */
-	if ((uninorth_rev >= 0x11 && uninorth_rev <= 0x24) || uninorth_rev == 0xc0)
+	if ((uninorth_rev >= 0x11 && uninorth_rev <= 0x24) ||
+	    uninorth_rev == 0xc0)
 		UN_OUT(0x2160, UN_IN(0x2160) & 0x00ffffff);
 }
 
@@ -2736,12 +2764,14 @@ static void __init probe_one_macio(const char *name, const char *compat, int typ
 		       node->full_name);
 		return;
 	}
-	if (type == macio_keylargo) {
+	if (type == macio_keylargo || type == macio_keylargo2) {
 		u32 *did = (u32 *)get_property(node, "device-id", NULL);
 		if (*did == 0x00000025)
 			type = macio_pangea;
 		if (*did == 0x0000003e)
 			type = macio_intrepid;
+		if (*did == 0x0000004f)
+			type = macio_shasta;
 	}
 	macio_chips[i].of_node	= node;
 	macio_chips[i].type	= type;
@@ -2840,7 +2870,8 @@ set_initial_features(void)
 	}
 
 #ifdef CONFIG_POWER4
-	if (macio_chips[0].type == macio_keylargo2) {
+	if (macio_chips[0].type == macio_keylargo2 ||
+	    macio_chips[0].type == macio_shasta) {
 #ifndef CONFIG_SMP
 		/* On SMP machines running UP, we have the second CPU eating
 		 * bus cycles. We need to take it off the bus. This is done
diff --git a/arch/powerpc/platforms/powermac/pci.c b/arch/powerpc/platforms/powermac/pci.c
index 5aab261075d..f671ed25390 100644
--- a/arch/powerpc/platforms/powermac/pci.c
+++ b/arch/powerpc/platforms/powermac/pci.c
@@ -1,7 +1,7 @@
 /*
  * Support for PCI bridges found on Power Macintoshes.
  *
- * Copyright (C) 2003 Benjamin Herrenschmuidt (benh@kernel.crashing.org)
+ * Copyright (C) 2003-2005 Benjamin Herrenschmuidt (benh@kernel.crashing.org)
  * Copyright (C) 1997 Paul Mackerras (paulus@samba.org)
  *
  * This program is free software; you can redistribute it and/or
@@ -25,7 +25,7 @@
 #include <asm/pmac_feature.h>
 #include <asm/grackle.h>
 #ifdef CONFIG_PPC64
-#include <asm/iommu.h>
+//#include <asm/iommu.h>
 #include <asm/ppc-pci.h>
 #endif
 
@@ -44,6 +44,7 @@ static int add_bridge(struct device_node *dev);
 static int has_uninorth;
 #ifdef CONFIG_PPC64
 static struct pci_controller *u3_agp;
+static struct pci_controller *u4_pcie;
 static struct pci_controller *u3_ht;
 #endif /* CONFIG_PPC64 */
 
@@ -97,11 +98,8 @@ static void __init fixup_bus_range(struct device_node *bridge)
 
 	/* Lookup the "bus-range" property for the hose */
 	bus_range = (int *) get_property(bridge, "bus-range", &len);
-	if (bus_range == NULL || len < 2 * sizeof(int)) {
-		printk(KERN_WARNING "Can't get bus-range for %s\n",
-			       bridge->full_name);
+	if (bus_range == NULL || len < 2 * sizeof(int))
 		return;
-	}
 	bus_range[1] = fixup_one_level_bus_range(bridge->child, bus_range[1]);
 }
 
@@ -128,14 +126,14 @@ static void __init fixup_bus_range(struct device_node *bridge)
  */
 
 #define MACRISC_CFA0(devfn, off)	\
-	((1 << (unsigned long)PCI_SLOT(dev_fn)) \
-	| (((unsigned long)PCI_FUNC(dev_fn)) << 8) \
-	| (((unsigned long)(off)) & 0xFCUL))
+	((1 << (unsigned int)PCI_SLOT(dev_fn)) \
+	| (((unsigned int)PCI_FUNC(dev_fn)) << 8) \
+	| (((unsigned int)(off)) & 0xFCUL))
 
 #define MACRISC_CFA1(bus, devfn, off)	\
-	((((unsigned long)(bus)) << 16) \
-	|(((unsigned long)(devfn)) << 8) \
-	|(((unsigned long)(off)) & 0xFCUL) \
+	((((unsigned int)(bus)) << 16) \
+	|(((unsigned int)(devfn)) << 8) \
+	|(((unsigned int)(off)) & 0xFCUL) \
 	|1UL)
 
 static unsigned long macrisc_cfg_access(struct pci_controller* hose,
@@ -168,7 +166,8 @@ static int macrisc_read_config(struct pci_bus *bus, unsigned int devfn,
 	hose = pci_bus_to_host(bus);
 	if (hose == NULL)
 		return PCIBIOS_DEVICE_NOT_FOUND;
-
+	if (offset >= 0x100)
+		return  PCIBIOS_BAD_REGISTER_NUMBER;
 	addr = macrisc_cfg_access(hose, bus->number, devfn, offset);
 	if (!addr)
 		return PCIBIOS_DEVICE_NOT_FOUND;
@@ -199,7 +198,8 @@ static int macrisc_write_config(struct pci_bus *bus, unsigned int devfn,
 	hose = pci_bus_to_host(bus);
 	if (hose == NULL)
 		return PCIBIOS_DEVICE_NOT_FOUND;
-
+	if (offset >= 0x100)
+		return  PCIBIOS_BAD_REGISTER_NUMBER;
 	addr = macrisc_cfg_access(hose, bus->number, devfn, offset);
 	if (!addr)
 		return PCIBIOS_DEVICE_NOT_FOUND;
@@ -234,12 +234,13 @@ static struct pci_ops macrisc_pci_ops =
 /*
  * Verify that a specific (bus, dev_fn) exists on chaos
  */
-static int
-chaos_validate_dev(struct pci_bus *bus, int devfn, int offset)
+static int chaos_validate_dev(struct pci_bus *bus, int devfn, int offset)
 {
 	struct device_node *np;
 	u32 *vendor, *device;
 
+	if (offset >= 0x100)
+		return  PCIBIOS_BAD_REGISTER_NUMBER;
 	np = pci_busdev_to_OF_node(bus, devfn);
 	if (np == NULL)
 		return PCIBIOS_DEVICE_NOT_FOUND;
@@ -341,10 +342,10 @@ static int u3_ht_skip_device(struct pci_controller *hose,
 }
 
 #define U3_HT_CFA0(devfn, off)		\
-		((((unsigned long)devfn) << 8) | offset)
+		((((unsigned int)devfn) << 8) | offset)
 #define U3_HT_CFA1(bus, devfn, off)	\
 		(U3_HT_CFA0(devfn, off) \
-		+ (((unsigned long)bus) << 16) \
+		+ (((unsigned int)bus) << 16) \
 		+ 0x01000000UL)
 
 static unsigned long u3_ht_cfg_access(struct pci_controller* hose,
@@ -370,7 +371,8 @@ static int u3_ht_read_config(struct pci_bus *bus, unsigned int devfn,
 	hose = pci_bus_to_host(bus);
 	if (hose == NULL)
 		return PCIBIOS_DEVICE_NOT_FOUND;
-
+	if (offset >= 0x100)
+		return  PCIBIOS_BAD_REGISTER_NUMBER;
 	addr = u3_ht_cfg_access(hose, bus->number, devfn, offset);
 	if (!addr)
 		return PCIBIOS_DEVICE_NOT_FOUND;
@@ -419,7 +421,8 @@ static int u3_ht_write_config(struct pci_bus *bus, unsigned int devfn,
 	hose = pci_bus_to_host(bus);
 	if (hose == NULL)
 		return PCIBIOS_DEVICE_NOT_FOUND;
-
+	if (offset >= 0x100)
+		return  PCIBIOS_BAD_REGISTER_NUMBER;
 	addr = u3_ht_cfg_access(hose, bus->number, devfn, offset);
 	if (!addr)
 		return PCIBIOS_DEVICE_NOT_FOUND;
@@ -459,6 +462,112 @@ static struct pci_ops u3_ht_pci_ops =
 	u3_ht_read_config,
 	u3_ht_write_config
 };
+
+#define U4_PCIE_CFA0(devfn, off)	\
+	((1 << ((unsigned int)PCI_SLOT(dev_fn)))	\
+	 | (((unsigned int)PCI_FUNC(dev_fn)) << 8)	\
+	 | ((((unsigned int)(off)) >> 8) << 28) \
+	 | (((unsigned int)(off)) & 0xfcU))
+
+#define U4_PCIE_CFA1(bus, devfn, off)	\
+	((((unsigned int)(bus)) << 16) \
+	 |(((unsigned int)(devfn)) << 8)	\
+	 | ((((unsigned int)(off)) >> 8) << 28) \
+	 |(((unsigned int)(off)) & 0xfcU)	\
+	 |1UL)
+
+static unsigned long u4_pcie_cfg_access(struct pci_controller* hose,
+					u8 bus, u8 dev_fn, int offset)
+{
+	unsigned int caddr;
+
+	if (bus == hose->first_busno) {
+		caddr = U4_PCIE_CFA0(dev_fn, offset);
+	} else
+		caddr = U4_PCIE_CFA1(bus, dev_fn, offset);
+
+	/* Uninorth will return garbage if we don't read back the value ! */
+	do {
+		out_le32(hose->cfg_addr, caddr);
+	} while (in_le32(hose->cfg_addr) != caddr);
+
+	offset &= 0x03;
+	return ((unsigned long)hose->cfg_data) + offset;
+}
+
+static int u4_pcie_read_config(struct pci_bus *bus, unsigned int devfn,
+			       int offset, int len, u32 *val)
+{
+	struct pci_controller *hose;
+	unsigned long addr;
+
+	hose = pci_bus_to_host(bus);
+	if (hose == NULL)
+		return PCIBIOS_DEVICE_NOT_FOUND;
+	if (offset >= 0x1000)
+		return  PCIBIOS_BAD_REGISTER_NUMBER;
+	addr = u4_pcie_cfg_access(hose, bus->number, devfn, offset);
+	if (!addr)
+		return PCIBIOS_DEVICE_NOT_FOUND;
+	/*
+	 * Note: the caller has already checked that offset is
+	 * suitably aligned and that len is 1, 2 or 4.
+	 */
+	switch (len) {
+	case 1:
+		*val = in_8((u8 *)addr);
+		break;
+	case 2:
+		*val = in_le16((u16 *)addr);
+		break;
+	default:
+		*val = in_le32((u32 *)addr);
+		break;
+	}
+	return PCIBIOS_SUCCESSFUL;
+}
+
+static int u4_pcie_write_config(struct pci_bus *bus, unsigned int devfn,
+				int offset, int len, u32 val)
+{
+	struct pci_controller *hose;
+	unsigned long addr;
+
+	hose = pci_bus_to_host(bus);
+	if (hose == NULL)
+		return PCIBIOS_DEVICE_NOT_FOUND;
+	if (offset >= 0x1000)
+		return  PCIBIOS_BAD_REGISTER_NUMBER;
+	addr = u4_pcie_cfg_access(hose, bus->number, devfn, offset);
+	if (!addr)
+		return PCIBIOS_DEVICE_NOT_FOUND;
+	/*
+	 * Note: the caller has already checked that offset is
+	 * suitably aligned and that len is 1, 2 or 4.
+	 */
+	switch (len) {
+	case 1:
+		out_8((u8 *)addr, val);
+		(void) in_8((u8 *)addr);
+		break;
+	case 2:
+		out_le16((u16 *)addr, val);
+		(void) in_le16((u16 *)addr);
+		break;
+	default:
+		out_le32((u32 *)addr, val);
+		(void) in_le32((u32 *)addr);
+		break;
+	}
+	return PCIBIOS_SUCCESSFUL;
+}
+
+static struct pci_ops u4_pcie_pci_ops =
+{
+	u4_pcie_read_config,
+	u4_pcie_write_config
+};
+
 #endif /* CONFIG_PPC64 */
 
 #ifdef CONFIG_PPC32
@@ -628,15 +737,36 @@ static void __init setup_u3_agp(struct pci_controller* hose)
 	hose->ops = &macrisc_pci_ops;
 	hose->cfg_addr = ioremap(0xf0000000 + 0x800000, 0x1000);
 	hose->cfg_data = ioremap(0xf0000000 + 0xc00000, 0x1000);
-
 	u3_agp = hose;
 }
 
+static void __init setup_u4_pcie(struct pci_controller* hose)
+{
+	/* We currently only implement the "non-atomic" config space, to
+	 * be optimised later.
+	 */
+	hose->ops = &u4_pcie_pci_ops;
+	hose->cfg_addr = ioremap(0xf0000000 + 0x800000, 0x1000);
+	hose->cfg_data = ioremap(0xf0000000 + 0xc00000, 0x1000);
+
+	/* The bus contains a bridge from root -> device, we need to
+	 * make it visible on bus 0 so that we pick the right type
+	 * of config cycles. If we didn't, we would have to force all
+	 * config cycles to be type 1. So we override the "bus-range"
+	 * property here
+	 */
+	hose->first_busno = 0x00;
+	hose->last_busno = 0xff;
+	u4_pcie = hose;
+}
+
 static void __init setup_u3_ht(struct pci_controller* hose)
 {
 	struct device_node *np = (struct device_node *)hose->arch_data;
+	struct pci_controller *other = NULL;
 	int i, cur;
 
+
 	hose->ops = &u3_ht_pci_ops;
 
 	/* We hard code the address because of the different size of
@@ -670,11 +800,20 @@ static void __init setup_u3_ht(struct pci_controller* hose)
 
 	u3_ht = hose;
 
-	if (u3_agp == NULL) {
-		DBG("U3 has no AGP, using full resource range\n");
+	if (u3_agp != NULL)
+		other = u3_agp;
+	else if (u4_pcie != NULL)
+		other = u4_pcie;
+
+	if (other == NULL) {
+		DBG("U3/4 has no AGP/PCIE, using full resource range\n");
 		return;
 	}
 
+	/* Fixup bus range vs. PCIE */
+	if (u4_pcie)
+		hose->last_busno = u4_pcie->first_busno - 1;
+
 	/* We "remove" the AGP resources from the resources allocated to HT,
 	 * that is we create "holes". However, that code does assumptions
 	 * that so far happen to be true (cross fingers...), typically that
@@ -682,7 +821,7 @@ static void __init setup_u3_ht(struct pci_controller* hose)
 	 */
 	cur = 0;
 	for (i=0; i<3; i++) {
-		struct resource *res = &u3_agp->mem_resources[i];
+		struct resource *res = &other->mem_resources[i];
 		if (res->flags != IORESOURCE_MEM)
 			continue;
 		/* We don't care about "fine" resources */
@@ -777,9 +916,13 @@ static int __init add_bridge(struct device_node *dev)
 		setup_u3_ht(hose);
 		disp_name = "U3-HT";
 		primary = 1;
+	} else if (device_is_compatible(dev, "u4-pcie")) {
+		setup_u4_pcie(hose);
+		disp_name = "U4-PCIE";
+		primary = 0;
 	}
-	printk(KERN_INFO "Found %s PCI host bridge.  Firmware bus number: %d->%d\n",
-		disp_name, hose->first_busno, hose->last_busno);
+	printk(KERN_INFO "Found %s PCI host bridge.  Firmware bus number:"
+	       " %d->%d\n", disp_name, hose->first_busno, hose->last_busno);
 #endif /* CONFIG_PPC64 */
 
 	/* 32 bits only bridges */
@@ -900,6 +1043,8 @@ void __init pmac_pci_init(void)
 		pci_setup_phb_io(u3_ht, 1);
 	if (u3_agp)
 		pci_setup_phb_io(u3_agp, 0);
+	if (u4_pcie)
+		pci_setup_phb_io(u4_pcie, 0);
 
 	/*
 	 * On ppc64, fixup the IO resources on our host bridges as
@@ -912,7 +1057,8 @@ void __init pmac_pci_init(void)
 
 	/* Fixup the PCI<->OF mapping for U3 AGP due to bus renumbering. We
 	 * assume there is no P2P bridge on the AGP bus, which should be a
-	 * safe assumptions hopefully.
+	 * safe assumptions for now. We should do something better in the
+	 * future though
 	 */
 	if (u3_agp) {
 		struct device_node *np = u3_agp->arch_data;
@@ -920,7 +1066,6 @@ void __init pmac_pci_init(void)
 		for (np = np->child; np; np = np->sibling)
 			PCI_DN(np)->busno = 0xf0;
 	}
-
 	/* pmac_check_ht_link(); */
 
 	/* Tell pci.c to not use the common resource allocation mechanism */
@@ -1127,7 +1272,8 @@ void pmac_pci_fixup_pciata(struct pci_dev* dev)
  good:
 	pci_read_config_byte(dev, PCI_CLASS_PROG, &progif);
 	if ((progif & 5) != 5) {
-		printk(KERN_INFO "Forcing PCI IDE into native mode: %s\n", pci_name(dev));
+		printk(KERN_INFO "Forcing PCI IDE into native mode: %s\n",
+		       pci_name(dev));
 		(void) pci_write_config_byte(dev, PCI_CLASS_PROG, progif|5);
 		if (pci_read_config_byte(dev, PCI_CLASS_PROG, &progif) ||
 		    (progif & 5) != 5)
@@ -1153,7 +1299,8 @@ static void fixup_k2_sata(struct pci_dev* dev)
 		for (i = 0; i < 6; i++) {
 			dev->resource[i].start = dev->resource[i].end = 0;
 			dev->resource[i].flags = 0;
-			pci_write_config_dword(dev, PCI_BASE_ADDRESS_0 + 4 * i, 0);
+			pci_write_config_dword(dev, PCI_BASE_ADDRESS_0 + 4 * i,
+					       0);
 		}
 	} else {
 		pci_read_config_word(dev, PCI_COMMAND, &cmd);
@@ -1162,7 +1309,8 @@ static void fixup_k2_sata(struct pci_dev* dev)
 		for (i = 0; i < 5; i++) {
 			dev->resource[i].start = dev->resource[i].end = 0;
 			dev->resource[i].flags = 0;
-			pci_write_config_dword(dev, PCI_BASE_ADDRESS_0 + 4 * i, 0);
+			pci_write_config_dword(dev, PCI_BASE_ADDRESS_0 + 4 * i,
+					       0);
 		}
 	}
 }
diff --git a/arch/powerpc/platforms/powermac/pic.c b/arch/powerpc/platforms/powermac/pic.c
index dbb524a851a..18bf3011d1e 100644
--- a/arch/powerpc/platforms/powermac/pic.c
+++ b/arch/powerpc/platforms/powermac/pic.c
@@ -524,18 +524,56 @@ static void __init pmac_pic_setup_mpic_nmi(struct mpic *mpic)
 #endif	/* defined(CONFIG_XMON) && defined(CONFIG_PPC32) */
 }
 
+static struct mpic * __init pmac_setup_one_mpic(struct device_node *np,
+						int master)
+{
+	unsigned char senses[128];
+	int offset = master ? 0 : 128;
+	int count = master ? 128 : 124;
+	const char *name = master ? " MPIC 1   " : " MPIC 2   ";
+	struct resource r;
+	struct mpic *mpic;
+	unsigned int flags = master ? MPIC_PRIMARY : 0;
+	int rc;
+
+	rc = of_address_to_resource(np, 0, &r);
+	if (rc)
+		return NULL;
+
+	pmac_call_feature(PMAC_FTR_ENABLE_MPIC, np, 0, 0);
+
+	prom_get_irq_senses(senses, offset, offset + count);
+
+	flags |= MPIC_WANTS_RESET;
+	if (get_property(np, "big-endian", NULL))
+		flags |= MPIC_BIG_ENDIAN;
+
+	/* Primary Big Endian means HT interrupts. This is quite dodgy
+	 * but works until I find a better way
+	 */
+	if (master && (flags & MPIC_BIG_ENDIAN))
+		flags |= MPIC_BROKEN_U3;
+
+	mpic = mpic_alloc(r.start, flags, 0, offset, count, master ? 252 : 0,
+			  senses, count, name);
+	if (mpic == NULL)
+		return NULL;
+
+	mpic_init(mpic);
+
+	return mpic;
+ }
+
 static int __init pmac_pic_probe_mpic(void)
 {
 	struct mpic *mpic1, *mpic2;
 	struct device_node *np, *master = NULL, *slave = NULL;
-	unsigned char senses[128];
-	struct resource r;
 
 	/* We can have up to 2 MPICs cascaded */
 	for (np = NULL; (np = of_find_node_by_type(np, "open-pic"))
 		     != NULL;) {
 		if (master == NULL &&
-		    get_property(np, "interrupt-parent", NULL) != NULL)
+		    get_property(np, "interrupts", NULL) == NULL)
 			master = of_node_get(np);
 		else if (slave == NULL)
 			slave = of_node_get(np);
@@ -557,13 +595,8 @@ static int __init pmac_pic_probe_mpic(void)
 	ppc_md.get_irq = mpic_get_irq;
 
 	/* Setup master */
-	BUG_ON(of_address_to_resource(master, 0, &r));
-	pmac_call_feature(PMAC_FTR_ENABLE_MPIC, master, 0, 0);
-	prom_get_irq_senses(senses, 0, 128);
-	mpic1 = mpic_alloc(r.start, MPIC_PRIMARY | MPIC_WANTS_RESET,
-			   0, 0, 128, 252, senses, 128, " OpenPIC  ");
+	mpic1 = pmac_setup_one_mpic(master, 1);
 	BUG_ON(mpic1 == NULL);
-	mpic_init(mpic1);
 
 	/* Install NMI if any */
 	pmac_pic_setup_mpic_nmi(mpic1);
@@ -574,27 +607,12 @@ static int __init pmac_pic_probe_mpic(void)
 	if (slave == NULL || slave->n_intrs < 1)
 		return 0;
 
-	/* Setup slave, failures are non-fatal */
-	if (of_address_to_resource(slave, 0, &r)) {
-		printk(KERN_ERR "Can't get address of MPIC %s\n",
-		       slave->full_name);
-		return 0;
-	}
-	pmac_call_feature(PMAC_FTR_ENABLE_MPIC, slave, 0, 0);
-	prom_get_irq_senses(senses, 128, 128 + 124);
-
-	/* We don't need to set MPIC_BROKEN_U3 here since we don't have
-	 * hypertransport interrupts routed to it, at least not on currently
-	 * supported machines, that may change.
-	 */
-	mpic2 = mpic_alloc(r.start, MPIC_BIG_ENDIAN | MPIC_WANTS_RESET,
-			   0, 128, 124, 0, senses, 124, " U3-MPIC  ");
+	mpic2 = pmac_setup_one_mpic(slave, 0);
 	if (mpic2 == NULL) {
-		printk(KERN_ERR "Can't create slave MPIC %s\n",
-		       slave->full_name);
+		printk(KERN_ERR "Failed to setup slave MPIC\n");
+		of_node_put(slave);
 		return 0;
 	}
-	mpic_init(mpic2);
 	mpic_setup_cascade(slave->intrs[0].line, pmac_u3_cascade, mpic2);
 
 	of_node_put(slave);
diff --git a/arch/powerpc/platforms/powermac/setup.c b/arch/powerpc/platforms/powermac/setup.c
index 18c5620f87f..1daa5a06e9e 100644
--- a/arch/powerpc/platforms/powermac/setup.c
+++ b/arch/powerpc/platforms/powermac/setup.c
@@ -345,7 +345,7 @@ void __init pmac_setup_arch(void)
 
 #ifdef CONFIG_SMP
 	/* Check for Core99 */
-	if (find_devices("uni-n") || find_devices("u3"))
+	if (find_devices("uni-n") || find_devices("u3") || find_devices("u4"))
 		smp_ops = &core99_smp_ops;
 #ifdef CONFIG_PPC32
 	else
@@ -635,7 +635,7 @@ static void __init pmac_init_early(void)
 	/* Setup interrupt mapping options */
 	ppc64_interrupt_controller = IC_OPEN_PIC;
 
-	iommu_init_early_u3();
+	iommu_init_early_dart();
 #endif
 }
 
@@ -711,7 +711,7 @@ static int __init pmac_probe(int platform)
 	 * occupies having to be broken up so the DART itself is not
 	 * part of the cacheable linar mapping
 	 */
-	alloc_u3_dart_table();
+	alloc_dart_table();
 #endif
 
 #ifdef CONFIG_PMAC_SMU
@@ -733,10 +733,11 @@ static int pmac_pci_probe_mode(struct pci_bus *bus)
 	struct device_node *node = bus->sysdata;
 
 	/* We need to use normal PCI probing for the AGP bus,
-	   since the device for the AGP bridge isn't in the tree. */
-	if (bus->self == NULL && device_is_compatible(node, "u3-agp"))
+	 * since the device for the AGP bridge isn't in the tree.
+	 */
+	if (bus->self == NULL && (device_is_compatible(node, "u3-agp") ||
+				  device_is_compatible(node, "u4-pcie")))
 		return PCI_PROBE_NORMAL;
-
 	return PCI_PROBE_DEVTREE;
 }
 #endif
diff --git a/arch/powerpc/platforms/powermac/smp.c b/arch/powerpc/platforms/powermac/smp.c
index 862f1e985c1..df01bb8feb1 100644
--- a/arch/powerpc/platforms/powermac/smp.c
+++ b/arch/powerpc/platforms/powermac/smp.c
@@ -361,7 +361,6 @@ static void __init psurge_dual_sync_tb(int cpu_nr)
 	set_dec(tb_ticks_per_jiffy);
 	/* XXX fixme */
 	set_tb(0, 0);
-	last_jiffy_stamp(cpu_nr) = 0;
 
 	if (cpu_nr > 0) {
 		mb();
@@ -429,15 +428,62 @@ struct smp_ops_t psurge_smp_ops = {
 };
 #endif /* CONFIG_PPC32 - actually powersurge support */
 
+/*
+ * Core 99 and later support
+ */
+
+static void (*pmac_tb_freeze)(int freeze);
+static unsigned long timebase;
+static int tb_req;
+
+static void smp_core99_give_timebase(void)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+
+	while(!tb_req)
+		barrier();
+	tb_req = 0;
+	(*pmac_tb_freeze)(1);
+	mb();
+	timebase = get_tb();
+	mb();
+	while (timebase)
+		barrier();
+	mb();
+	(*pmac_tb_freeze)(0);
+	mb();
+
+	local_irq_restore(flags);
+}
+
+
+static void __devinit smp_core99_take_timebase(void)
+{
+	unsigned long flags;
+
+	local_irq_save(flags);
+
+	tb_req = 1;
+	mb();
+	while (!timebase)
+		barrier();
+	mb();
+	set_tb(timebase >> 32, timebase & 0xffffffff);
+	timebase = 0;
+	mb();
+	set_dec(tb_ticks_per_jiffy/2);
+
+	local_irq_restore(flags);
+}
+
 #ifdef CONFIG_PPC64
 /*
  * G5s enable/disable the timebase via an i2c-connected clock chip.
  */
 static struct device_node *pmac_tb_clock_chip_host;
 static u8 pmac_tb_pulsar_addr;
-static void (*pmac_tb_freeze)(int freeze);
-static DEFINE_SPINLOCK(timebase_lock);
-static unsigned long timebase;
 
 static void smp_core99_cypress_tb_freeze(int freeze)
 {
@@ -447,7 +493,8 @@ static void smp_core99_cypress_tb_freeze(int freeze)
 	/* Strangely, the device-tree says address is 0xd2, but darwin
 	 * accesses 0xd0 ...
 	 */
-	pmac_low_i2c_setmode(pmac_tb_clock_chip_host, pmac_low_i2c_mode_combined);
+	pmac_low_i2c_setmode(pmac_tb_clock_chip_host,
+			     pmac_low_i2c_mode_combined);
 	rc = pmac_low_i2c_xfer(pmac_tb_clock_chip_host,
 			       0xd0 | pmac_low_i2c_read,
 			       0x81, &data, 1);
@@ -475,7 +522,8 @@ static void smp_core99_pulsar_tb_freeze(int freeze)
 	u8 data;
 	int rc;
 
-	pmac_low_i2c_setmode(pmac_tb_clock_chip_host, pmac_low_i2c_mode_combined);
+	pmac_low_i2c_setmode(pmac_tb_clock_chip_host,
+			     pmac_low_i2c_mode_combined);
 	rc = pmac_low_i2c_xfer(pmac_tb_clock_chip_host,
 			       pmac_tb_pulsar_addr | pmac_low_i2c_read,
 			       0x2e, &data, 1);
@@ -496,54 +544,14 @@ static void smp_core99_pulsar_tb_freeze(int freeze)
 	}
 }
 
-
-static void smp_core99_give_timebase(void)
-{
-	/* Open i2c bus for synchronous access */
-	if (pmac_low_i2c_open(pmac_tb_clock_chip_host, 0))
-		panic("Can't open i2c for TB sync !\n");
-
-	spin_lock(&timebase_lock);
-	(*pmac_tb_freeze)(1);
-	mb();
-	timebase = get_tb();
-	spin_unlock(&timebase_lock);
-
-	while (timebase)
-		barrier();
-
-	spin_lock(&timebase_lock);
-	(*pmac_tb_freeze)(0);
-	spin_unlock(&timebase_lock);
-
-	/* Close i2c bus */
-	pmac_low_i2c_close(pmac_tb_clock_chip_host);
-}
-
-
-static void __devinit smp_core99_take_timebase(void)
-{
-	while (!timebase)
-		barrier();
-	spin_lock(&timebase_lock);
-	set_tb(timebase >> 32, timebase & 0xffffffff);
-	timebase = 0;
-	spin_unlock(&timebase_lock);
-}
-
-static void __init smp_core99_setup(int ncpus)
+static void __init smp_core99_setup_i2c_hwsync(int ncpus)
 {
 	struct device_node *cc = NULL;	
 	struct device_node *p;
+	const char *name = NULL;
 	u32 *reg;
 	int ok;
 
-	/* HW sync only on these platforms */
-	if (!machine_is_compatible("PowerMac7,2") &&
-	    !machine_is_compatible("PowerMac7,3") &&
-	    !machine_is_compatible("RackMac3,1"))
-		return;
-
 	/* Look for the clock chip */
 	while ((cc = of_find_node_by_name(cc, "i2c-hwclock")) != NULL) {
 		p = of_get_parent(cc);
@@ -561,114 +569,64 @@ static void __init smp_core99_setup(int ncpus)
 			if (device_is_compatible(cc, "pulsar-legacy-slewing")) {
 				pmac_tb_freeze = smp_core99_pulsar_tb_freeze;
 				pmac_tb_pulsar_addr = 0xd2;
-				printk(KERN_INFO "Timebase clock is Pulsar chip\n");
+				name = "Pulsar";
 			} else if (device_is_compatible(cc, "cy28508")) {
 				pmac_tb_freeze = smp_core99_cypress_tb_freeze;
-				printk(KERN_INFO "Timebase clock is Cypress chip\n");
+				name = "Cypress";
 			}
 			break;
 		case 0xd4:
 			pmac_tb_freeze = smp_core99_pulsar_tb_freeze;
 			pmac_tb_pulsar_addr = 0xd4;
-			printk(KERN_INFO "Timebase clock is Pulsar chip\n");
+			name = "Pulsar";
 			break;
 		}
-		if (pmac_tb_freeze != NULL) {
-			pmac_tb_clock_chip_host = of_get_parent(cc);
-			of_node_put(cc);
+		if (pmac_tb_freeze != NULL)
 			break;
-		}
 	}
-	if (pmac_tb_freeze == NULL) {
-		smp_ops->give_timebase = smp_generic_give_timebase;
-		smp_ops->take_timebase = smp_generic_take_timebase;
+	if (pmac_tb_freeze != NULL) {
+		struct device_node *p = of_get_parent(cc);
+		of_node_put(cc);
+		while(p && strcmp(p->type, "i2c")) {
+			cc = of_get_parent(p);
+			of_node_put(p);
+			p = cc;
+		}
+		if (p == NULL)
+			goto no_i2c_sync;
+		/* Open i2c bus for synchronous access */
+		if (pmac_low_i2c_open(p, 0)) {
+			printk(KERN_ERR "Failed top open i2c bus %s for clock"
+			       " sync, fallback to software sync !\n",
+			       p->full_name);
+			of_node_put(p);
+			goto no_i2c_sync;
+		}
+		pmac_tb_clock_chip_host = p;
+		printk(KERN_INFO "Processor timebase sync using %s i2c clock\n",
+		       name);
+		return;
 	}
+ no_i2c_sync:
+	pmac_tb_freeze = NULL;
 }
 
-/* nothing to do here, caches are already set up by service processor */
-static inline void __devinit core99_init_caches(int cpu)
-{
-}
+#endif /* CONFIG_PPC64 */
 
-#else /* CONFIG_PPC64 */
 
 /*
- * SMP G4 powermacs use a GPIO to enable/disable the timebase.
+ * SMP G4 and newer G5 use a GPIO to enable/disable the timebase.
  */
 
 static unsigned int core99_tb_gpio;	/* Timebase freeze GPIO */
 
-static unsigned int pri_tb_hi, pri_tb_lo;
-static unsigned int pri_tb_stamp;
-
-/* not __init, called in sleep/wakeup code */
-void smp_core99_give_timebase(void)
+static void smp_core99_gpio_tb_freeze(int freeze)
 {
-	unsigned long flags;
-	unsigned int t;
-
-	/* wait for the secondary to be in take_timebase */
-	for (t = 100000; t > 0 && !sec_tb_reset; --t)
-		udelay(10);
-	if (!sec_tb_reset) {
-		printk(KERN_WARNING "Timeout waiting sync on second CPU\n");
-		return;
-	}
-
-	/* freeze the timebase and read it */
-	/* disable interrupts so the timebase is disabled for the
-	   shortest possible time */
-	local_irq_save(flags);
-	pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, core99_tb_gpio, 4);
+	if (freeze)
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, core99_tb_gpio, 4);
+	else
+		pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, core99_tb_gpio, 0);
 	pmac_call_feature(PMAC_FTR_READ_GPIO, NULL, core99_tb_gpio, 0);
-	mb();
-	pri_tb_hi = get_tbu();
-	pri_tb_lo = get_tbl();
-	pri_tb_stamp = last_jiffy_stamp(smp_processor_id());
-	mb();
-
-	/* tell the secondary we're ready */
-	sec_tb_reset = 2;
-	mb();
-
-	/* wait for the secondary to have taken it */
-	/* note: can't use udelay here, since it needs the timebase running */
-	for (t = 10000000; t > 0 && sec_tb_reset; --t)
-		barrier();
-	if (sec_tb_reset)
-		/* XXX BUG_ON here? */
-		printk(KERN_WARNING "Timeout waiting sync(2) on second CPU\n");
-
-	/* Now, restart the timebase by leaving the GPIO to an open collector */
-       	pmac_call_feature(PMAC_FTR_WRITE_GPIO, NULL, core99_tb_gpio, 0);
-        pmac_call_feature(PMAC_FTR_READ_GPIO, NULL, core99_tb_gpio, 0);
-	local_irq_restore(flags);
-}
-
-/* not __init, called in sleep/wakeup code */
-void smp_core99_take_timebase(void)
-{
-	unsigned long flags;
-
-	/* tell the primary we're here */
-	sec_tb_reset = 1;
-	mb();
-
-	/* wait for the primary to set pri_tb_hi/lo */
-	while (sec_tb_reset < 2)
-		mb();
-
-	/* set our stuff the same as the primary */
-	local_irq_save(flags);
-	set_dec(1);
-	set_tb(pri_tb_hi, pri_tb_lo);
-	last_jiffy_stamp(smp_processor_id()) = pri_tb_stamp;
-	mb();
-
-	/* tell the primary we're done */
-       	sec_tb_reset = 0;
-	mb();
-	local_irq_restore(flags);
 }
 
 /* L2 and L3 cache settings to pass from CPU0 to CPU1 on G4 cpus */
@@ -677,6 +635,7 @@ volatile static long int core99_l3_cache;
 
 static void __devinit core99_init_caches(int cpu)
 {
+#ifndef CONFIG_PPC64
 	if (!cpu_has_feature(CPU_FTR_L2CR))
 		return;
 
@@ -702,30 +661,80 @@ static void __devinit core99_init_caches(int cpu)
 		_set_L3CR(core99_l3_cache);
 		printk("CPU%d: L3CR set to %lx\n", cpu, core99_l3_cache);
 	}
+#endif /* !CONFIG_PPC64 */
 }
 
 static void __init smp_core99_setup(int ncpus)
 {
-	struct device_node *cpu;
-	u32 *tbprop = NULL;
-	int i;
+#ifdef CONFIG_PPC64
 
-	core99_tb_gpio = KL_GPIO_TB_ENABLE;	/* default value */
-	cpu = of_find_node_by_type(NULL, "cpu");
-	if (cpu != NULL) {
-		tbprop = (u32 *)get_property(cpu, "timebase-enable", NULL);
-		if (tbprop)
-			core99_tb_gpio = *tbprop;
-		of_node_put(cpu);
+	/* i2c based HW sync on some G5s */
+	if (machine_is_compatible("PowerMac7,2") ||
+	    machine_is_compatible("PowerMac7,3") ||
+	    machine_is_compatible("RackMac3,1"))
+		smp_core99_setup_i2c_hwsync(ncpus);
+
+	/* GPIO based HW sync on recent G5s */
+	if (pmac_tb_freeze == NULL) {
+		struct device_node *np =
+			of_find_node_by_name(NULL, "timebase-enable");
+		u32 *reg = (u32 *)get_property(np, "reg", NULL);
+
+		if (np && reg && !strcmp(np->type, "gpio")) {
+			core99_tb_gpio = *reg;
+			if (core99_tb_gpio < 0x50)
+				core99_tb_gpio += 0x50;
+			pmac_tb_freeze = smp_core99_gpio_tb_freeze;
+			printk(KERN_INFO "Processor timebase sync using"
+			       " GPIO 0x%02x\n", core99_tb_gpio);
+		}
 	}
 
-	/* XXX should get this from reg properties */
-	for (i = 1; i < ncpus; ++i)
-		smp_hw_index[i] = i;
-	powersave_nap = 0;
-}
+#else /* CONFIG_PPC64 */
+
+	/* GPIO based HW sync on ppc32 Core99 */
+	if (pmac_tb_freeze == NULL && !machine_is_compatible("MacRISC4")) {
+		struct device_node *cpu;
+		u32 *tbprop = NULL;
+
+		core99_tb_gpio = KL_GPIO_TB_ENABLE;	/* default value */
+		cpu = of_find_node_by_type(NULL, "cpu");
+		if (cpu != NULL) {
+			tbprop = (u32 *)get_property(cpu, "timebase-enable",
+						     NULL);
+			if (tbprop)
+				core99_tb_gpio = *tbprop;
+			of_node_put(cpu);
+		}
+		pmac_tb_freeze = smp_core99_gpio_tb_freeze;
+		printk(KERN_INFO "Processor timebase sync using"
+		       " GPIO 0x%02x\n", core99_tb_gpio);
+	}
+
+#endif /* CONFIG_PPC64 */
+
+	/* No timebase sync, fallback to software */
+	if (pmac_tb_freeze == NULL) {
+		smp_ops->give_timebase = smp_generic_give_timebase;
+		smp_ops->take_timebase = smp_generic_take_timebase;
+		printk(KERN_INFO "Processor timebase sync using software\n");
+	}
+
+#ifndef CONFIG_PPC64
+	{
+		int i;
+
+		/* XXX should get this from reg properties */
+		for (i = 1; i < ncpus; ++i)
+			smp_hw_index[i] = i;
+	}
 #endif
 
+	/* 32 bits SMP can't NAP */
+	if (!machine_is_compatible("MacRISC4"))
+		powersave_nap = 0;
+}
+
 static int __init smp_core99_probe(void)
 {
 	struct device_node *cpus;
@@ -803,17 +812,25 @@ static void __devinit smp_core99_setup_cpu(int cpu_nr)
 	mpic_setup_this_cpu();
 
 	if (cpu_nr == 0) {
-#ifdef CONFIG_POWER4
+#ifdef CONFIG_PPC64
 		extern void g5_phy_disable_cpu1(void);
 
+		/* Close i2c bus if it was used for tb sync */
+		if (pmac_tb_clock_chip_host) {
+			pmac_low_i2c_close(pmac_tb_clock_chip_host);
+			pmac_tb_clock_chip_host	= NULL;
+		}
+
 		/* If we didn't start the second CPU, we must take
 		 * it off the bus
 		 */
 		if (machine_is_compatible("MacRISC4") &&
 		    num_online_cpus() < 2)		
 			g5_phy_disable_cpu1();
-#endif /* CONFIG_POWER4 */
-		if (ppc_md.progress) ppc_md.progress("core99_setup_cpu 0 done", 0x349);
+#endif /* CONFIG_PPC64 */
+
+		if (ppc_md.progress)
+			ppc_md.progress("core99_setup_cpu 0 done", 0x349);
 	}
 }