[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>

arch/powerpc/Kconfig

@@ -300,6 +300,7 @@ config PPC_PMAC64
 	bool
 	depends on PPC_PMAC && POWER4
 	select U3_DART
+	select MPIC_BROKEN_U3
 	select GENERIC_TBSYNC
 	default y
 

arch/powerpc/kernel/pci_64.c

@@ -34,7 +34,7 @@
 
 #ifdef DEBUG
 #include <asm/udbg.h>
-#define DBG(fmt...) udbg_printf(fmt)
+#define DBG(fmt...) printk(fmt)
 #else
 #define DBG(fmt...)
 #endif
@@ -323,6 +323,7 @@ static void pci_parse_of_addrs(struct device_node *node, struct pci_dev *dev)
 	addrs = (u32 *) get_property(node, "assigned-addresses", &proplen);
 	if (!addrs)
 		return;
+	DBG("    parse addresses (%d bytes) @ %p\n", proplen, addrs);
 	for (; proplen >= 20; proplen -= 20, addrs += 5) {
 		flags = pci_parse_of_flags(addrs[0]);
 		if (!flags)
@@ -332,6 +333,9 @@ static void pci_parse_of_addrs(struct device_node *node, struct pci_dev *dev)
 		if (!size)
 			continue;
 		i = addrs[0] & 0xff;
+		DBG("  base: %llx, size: %llx, i: %x\n",
+		    (unsigned long long)base, (unsigned long long)size, i);
+
 		if (PCI_BASE_ADDRESS_0 <= i && i <= PCI_BASE_ADDRESS_5) {
 			res = &dev->resource[(i - PCI_BASE_ADDRESS_0) >> 2];
 		} else if (i == dev->rom_base_reg) {
@@ -362,6 +366,8 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
 	if (type == NULL)
 		type = "";
 
+	DBG("    create device, devfn: %x, type: %s\n", devfn, type);
+
 	memset(dev, 0, sizeof(struct pci_dev));
 	dev->bus = bus;
 	dev->sysdata = node;
@@ -381,6 +387,8 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
 		dev->bus->number, PCI_SLOT(devfn), PCI_FUNC(devfn));
 	dev->class = get_int_prop(node, "class-code", 0);
 
+	DBG("    class: 0x%x\n", dev->class);
+
 	dev->current_state = 4;		/* unknown power state */
 
 	if (!strcmp(type, "pci")) {
@@ -402,6 +410,8 @@ struct pci_dev *of_create_pci_dev(struct device_node *node,
 
 	pci_parse_of_addrs(node, dev);
 
+	DBG("    adding to system ...\n");
+
 	pci_device_add(dev, bus);
 
 	/* XXX pci_scan_msi_device(dev); */
@@ -418,15 +428,21 @@ void __devinit of_scan_bus(struct device_node *node,
 	int reglen, devfn;
 	struct pci_dev *dev;
 
+	DBG("of_scan_bus(%s) bus no %d... \n", node->full_name, bus->number);
+
 	while ((child = of_get_next_child(node, child)) != NULL) {
+		DBG("  * %s\n", child->full_name);
 		reg = (u32 *) get_property(child, "reg", &reglen);
 		if (reg == NULL || reglen < 20)
 			continue;
 		devfn = (reg[0] >> 8) & 0xff;
+
 		/* create a new pci_dev for this device */
 		dev = of_create_pci_dev(child, bus, devfn);
 		if (!dev)
 			continue;
+		DBG("dev header type: %x\n", dev->hdr_type);
+
 		if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE ||
 		    dev->hdr_type == PCI_HEADER_TYPE_CARDBUS)
 			of_scan_pci_bridge(child, dev);
@@ -446,16 +462,18 @@ void __devinit of_scan_pci_bridge(struct device_node *node,
 	unsigned int flags;
 	u64 size;
 
+	DBG("of_scan_pci_bridge(%s)\n", node->full_name);
+
 	/* parse bus-range property */
 	busrange = (u32 *) get_property(node, "bus-range", &len);
 	if (busrange == NULL || len != 8) {
-		printk(KERN_ERR "Can't get bus-range for PCI-PCI bridge %s\n",
+		printk(KERN_DEBUG "Can't get bus-range for PCI-PCI bridge %s\n",
 		       node->full_name);
 		return;
 	}
 	ranges = (u32 *) get_property(node, "ranges", &len);
 	if (ranges == NULL) {
-		printk(KERN_ERR "Can't get ranges for PCI-PCI bridge %s\n",
+		printk(KERN_DEBUG "Can't get ranges for PCI-PCI bridge %s\n",
 		       node->full_name);
 		return;
 	}
@@ -509,10 +527,13 @@ void __devinit of_scan_pci_bridge(struct device_node *node,
 	}
 	sprintf(bus->name, "PCI Bus %04x:%02x", pci_domain_nr(bus),
 		bus->number);
+	DBG("    bus name: %s\n", bus->name);
 
 	mode = PCI_PROBE_NORMAL;
 	if (ppc_md.pci_probe_mode)
 		mode = ppc_md.pci_probe_mode(bus);
+	DBG("    probe mode: %d\n", mode);
+
 	if (mode == PCI_PROBE_DEVTREE)
 		of_scan_bus(node, bus);
 	else if (mode == PCI_PROBE_NORMAL)
@@ -528,6 +549,8 @@ void __devinit scan_phb(struct pci_controller *hose)
 	int i, mode;
 	struct resource *res;
 
+	DBG("Scanning PHB %s\n", node ? node->full_name : "<NO NAME>");
+
 	bus = pci_create_bus(NULL, hose->first_busno, hose->ops, node);
 	if (bus == NULL) {
 		printk(KERN_ERR "Failed to create bus for PCI domain %04x\n",
@@ -552,8 +575,9 @@ void __devinit scan_phb(struct pci_controller *hose)
 
 	mode = PCI_PROBE_NORMAL;
 #ifdef CONFIG_PPC_MULTIPLATFORM
-	if (ppc_md.pci_probe_mode)
+	if (node && ppc_md.pci_probe_mode)
 		mode = ppc_md.pci_probe_mode(bus);
+	DBG("    probe mode: %d\n", mode);
 	if (mode == PCI_PROBE_DEVTREE) {
 		bus->subordinate = hose->last_busno;
 		of_scan_bus(node, bus);
@@ -842,8 +866,7 @@ pgprot_t pci_phys_mem_access_prot(struct file *file,
  * Returns a negative error code on failure, zero on success.
  */
 int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
-			enum pci_mmap_state mmap_state,
-			int write_combine)
+			enum pci_mmap_state mmap_state, int write_combine)
 {
 	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
 	struct resource *rp;

arch/powerpc/kernel/prom.c

@@ -298,6 +298,16 @@ static int __devinit finish_node_interrupts(struct device_node *np,
 	int i, j, n, sense;
 	unsigned int *irq, virq;
 	struct device_node *ic;
+	int trace = 0;
+
+	//#define TRACE(fmt...) do { if (trace) { printk(fmt); mdelay(1000); } } while(0)
+#define TRACE(fmt...)
+
+	if (!strcmp(np->name, "smu-doorbell"))
+		trace = 1;
+
+	TRACE("Finishing SMU doorbell ! num_interrupt_controllers = %d\n",
+	      num_interrupt_controllers);
 
 	if (num_interrupt_controllers == 0) {
 		/*
@@ -332,11 +342,12 @@ static int __devinit finish_node_interrupts(struct device_node *np,
 	}
 
 	ints = (unsigned int *) get_property(np, "interrupts", &intlen);
+	TRACE("ints=%p, intlen=%d\n", ints, intlen);
 	if (ints == NULL)
 		return 0;
 	intrcells = prom_n_intr_cells(np);
 	intlen /= intrcells * sizeof(unsigned int);
-
+	TRACE("intrcells=%d, new intlen=%d\n", intrcells, intlen);
 	np->intrs = prom_alloc(intlen * sizeof(*(np->intrs)), mem_start);
 	if (!np->intrs)
 		return -ENOMEM;
@@ -347,6 +358,7 @@ static int __devinit finish_node_interrupts(struct device_node *np,
 	intrcount = 0;
 	for (i = 0; i < intlen; ++i, ints += intrcells) {
 		n = map_interrupt(&irq, &ic, np, ints, intrcells);
+		TRACE("map, irq=%d, ic=%p, n=%d\n", irq, ic, n);
 		if (n <= 0)
 			continue;
 
@@ -357,6 +369,7 @@ static int __devinit finish_node_interrupts(struct device_node *np,
 			np->intrs[intrcount].sense = map_isa_senses[sense];
 		} else {
 			virq = virt_irq_create_mapping(irq[0]);
+			TRACE("virq=%d\n", virq);
 #ifdef CONFIG_PPC64
 			if (virq == NO_IRQ) {
 				printk(KERN_CRIT "Could not allocate interrupt"
@@ -366,6 +379,12 @@ static int __devinit finish_node_interrupts(struct device_node *np,
 #endif
 			np->intrs[intrcount].line = irq_offset_up(virq);
 			sense = (n > 1)? (irq[1] & 3): 1;
+
+			/* Apple uses bits in there in a different way, let's
+			 * only keep the real sense bit on macs
+			 */
+			if (_machine == PLATFORM_POWERMAC)
+				sense &= 0x1;
 			np->intrs[intrcount].sense = map_mpic_senses[sense];
 		}
 
@@ -375,12 +394,13 @@ static int __devinit finish_node_interrupts(struct device_node *np,
 			char *name = get_property(ic->parent, "name", NULL);
 			if (name && !strcmp(name, "u3"))
 				np->intrs[intrcount].line += 128;
-			else if (!(name && !strcmp(name, "mac-io")))
+			else if (!(name && (!strcmp(name, "mac-io") ||
+					    !strcmp(name, "u4"))))
 				/* ignore other cascaded controllers, such as
 				   the k2-sata-root */
 				break;
 		}
-#endif
+#endif /* CONFIG_PPC64 */
 		if (n > 2) {
 			printk("hmmm, got %d intr cells for %s:", n,
 			       np->full_name);

arch/powerpc/kernel/udbg.c

@@ -110,10 +110,12 @@ static int early_console_initialized;
 
 void __init disable_early_printk(void)
 {
+#if 1
 	if (!early_console_initialized)
 		return;
 	unregister_console(&udbg_console);
 	early_console_initialized = 0;
+#endif
 }
 
 /* called by setup_system */

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;
 }

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");
@@ -1547,6 +1553,7 @@ static long g5_reset_cpu(struct device_node *node, long param, long value)
  */
 void g5_phy_disable_cpu1(void)
 {
+	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

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);

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

arch/powerpc/sysdev/Makefile

@@ -4,6 +4,6 @@ obj-$(CONFIG_PPC_I8259)		+= i8259.o
 obj-$(CONFIG_PPC_MPC106)	+= grackle.o
 obj-$(CONFIG_BOOKE)		+= dcr.o
 obj-$(CONFIG_40x)		+= dcr.o
-obj-$(CONFIG_U3_DART)		+= u3_iommu.o
+obj-$(CONFIG_U3_DART)		+= dart_iommu.o
 obj-$(CONFIG_MMIO_NVRAM)	+= mmio_nvram.o
 obj-$(CONFIG_83xx)		+= ipic.o

arch/powerpc/sysdev/dart.h

@@ -20,29 +20,44 @@
 #define _POWERPC_SYSDEV_DART_H
 
 
-/* physical base of DART registers */
-#define DART_BASE        0xf8033000UL
-
 /* Offset from base to control register */
-#define DARTCNTL   0
+#define DART_CNTL	0
+
 /* Offset from base to exception register */
-#define DARTEXCP   0x10
+#define DART_EXCP_U3	0x10
 /* Offset from base to TLB tag registers */
-#define DARTTAG    0x1000
+#define DART_TAGS_U3	0x1000
 
+/* U4 registers */
+#define DART_BASE_U4	0x10
+#define DART_SIZE_U4	0x20
+#define DART_EXCP_U4	0x30
+#define DART_TAGS_U4	0x1000
 
 /* Control Register fields */
 
-/* base address of table (pfn) */
-#define DARTCNTL_BASE_MASK    0xfffff
-#define DARTCNTL_BASE_SHIFT   12
+/* U3 registers */
+#define DART_CNTL_U3_BASE_MASK	0xfffff
+#define DART_CNTL_U3_BASE_SHIFT	12
+#define DART_CNTL_U3_FLUSHTLB	0x400
+#define DART_CNTL_U3_ENABLE	0x200
+#define DART_CNTL_U3_SIZE_MASK	0x1ff
+#define DART_CNTL_U3_SIZE_SHIFT	0
+
+/* U4 registers */
+#define DART_BASE_U4_BASE_MASK	0xffffff
+#define DART_BASE_U4_BASE_SHIFT	0
+#define DART_CNTL_U4_FLUSHTLB	0x20000000
+#define DART_CNTL_U4_ENABLE	0x80000000
+#define DART_SIZE_U4_SIZE_MASK	0x1fff
+#define DART_SIZE_U4_SIZE_SHIFT	0
+
+#define DART_REG(r)	(dart + ((r) >> 2))
+#define DART_IN(r)	(in_be32(DART_REG(r)))
+#define DART_OUT(r,v)	(out_be32(DART_REG(r), (v)))
 
-#define DARTCNTL_FLUSHTLB     0x400
-#define DARTCNTL_ENABLE       0x200
 
 /* size of table in pages */
-#define DARTCNTL_SIZE_MASK    0x1ff
-#define DARTCNTL_SIZE_SHIFT   0
 
 
 /* DART table fields */

arch/powerpc/sysdev/u3_iommu.c → arch/powerpc/sysdev/dart_iommu.c

 /*
- * arch/powerpc/sysdev/u3_iommu.c
+ * arch/powerpc/sysdev/dart_iommu.c
  *
  * Copyright (C) 2004 Olof Johansson <olof@lixom.net>, IBM Corporation
+ * Copyright (C) 2005 Benjamin Herrenschmidt <benh@kernel.crashing.org>,
+ *                    IBM Corporation
  *
  * Based on pSeries_iommu.c:
  * Copyright (C) 2001 Mike Corrigan & Dave Engebretsen, IBM Corporation
  * Copyright (C) 2004 Olof Johansson <olof@lixom.net>, IBM Corporation
  *
- * Dynamic DMA mapping support, Apple U3 & IBM CPC925 "DART" iommu.
+ * Dynamic DMA mapping support, Apple U3, U4 & IBM CPC925 "DART" iommu.
  *
  *
  * This program is free software; you can redistribute it and/or modify
@@ -57,21 +59,22 @@ static unsigned long dart_tablesize;
 static u32 *dart_vbase;
 
 /* Mapped base address for the dart */
-static unsigned int *dart; 
+static unsigned int *__iomem dart;
 
 /* Dummy val that entries are set to when unused */
 static unsigned int dart_emptyval;
 
-static struct iommu_table iommu_table_u3;
-static int iommu_table_u3_inited;
+static struct iommu_table iommu_table_dart;
+static int iommu_table_dart_inited;
 static int dart_dirty;
+static int dart_is_u4;
 
 #define DBG(...)
 
 static inline void dart_tlb_invalidate_all(void)
 {
 	unsigned long l = 0;
-	unsigned int reg;
+	unsigned int reg, inv_bit;
 	unsigned long limit;
 
 	DBG("dart: flush\n");
@@ -85,25 +88,24 @@ static inline void dart_tlb_invalidate_all(void)
 
 	limit = 0;
 
+	inv_bit = dart_is_u4 ? DART_CNTL_U4_FLUSHTLB : DART_CNTL_U3_FLUSHTLB;
 retry:
-	reg = in_be32((unsigned int *)dart+DARTCNTL);
-	reg |= DARTCNTL_FLUSHTLB;
-	out_be32((unsigned int *)dart+DARTCNTL, reg);
-
 	l = 0;
-	while ((in_be32((unsigned int *)dart+DARTCNTL) & DARTCNTL_FLUSHTLB) &&
-		l < (1L<<limit)) {
+	reg = DART_IN(DART_CNTL);
+	reg |= inv_bit;
+	DART_OUT(DART_CNTL, reg);
+
+	while ((DART_IN(DART_CNTL) & inv_bit) && l < (1L << limit))
 		l++;
-	}
-	if (l == (1L<<limit)) {
+	if (l == (1L << limit)) {
 		if (limit < 4) {
 			limit++;
-		        reg = in_be32((unsigned int *)dart+DARTCNTL);
-		        reg &= ~DARTCNTL_FLUSHTLB;
-		        out_be32((unsigned int *)dart+DARTCNTL, reg);
+		        reg = DART_IN(DART_CNTL);
+		        reg &= ~inv_bit;
+			DART_OUT(DART_CNTL, reg);
 			goto retry;
 		} else
-			panic("U3-DART: TLB did not flush after waiting a long "
+			panic("DART: TLB did not flush after waiting a long "
 			      "time. Buggy U3 ?");
 	}
 }
@@ -168,20 +170,25 @@ static void dart_free(struct iommu_table *tbl, long index, long npages)
 
 static int dart_init(struct device_node *dart_node)
 {
-	unsigned int regword;
 	unsigned int i;
-	unsigned long tmp;
+	unsigned long tmp, base, size;
+	struct resource r;
 
 	if (dart_tablebase == 0 || dart_tablesize == 0) {
-		printk(KERN_INFO "U3-DART: table not allocated, using direct DMA\n");
+		printk(KERN_INFO "DART: table not allocated, using "
+		       "direct DMA\n");
 		return -ENODEV;
 	}
 
+	if (of_address_to_resource(dart_node, 0, &r))
+		panic("DART: can't get register base ! ");
+
 	/* Make sure nothing from the DART range remains in the CPU cache
 	 * from a previous mapping that existed before the kernel took
 	 * over
 	 */
-	flush_dcache_phys_range(dart_tablebase, dart_tablebase + dart_tablesize);
+	flush_dcache_phys_range(dart_tablebase,
+				dart_tablebase + dart_tablesize);
 
 	/* Allocate a spare page to map all invalid DART pages. We need to do
 	 * that to work around what looks like a problem with the HT bridge
@@ -189,21 +196,16 @@ static int dart_init(struct device_node *dart_node)
 	 */
 	tmp = lmb_alloc(DART_PAGE_SIZE, DART_PAGE_SIZE);
 	if (!tmp)
-		panic("U3-DART: Cannot allocate spare page!");
-	dart_emptyval = DARTMAP_VALID | ((tmp >> DART_PAGE_SHIFT) & DARTMAP_RPNMASK);
+		panic("DART: Cannot allocate spare page!");
+	dart_emptyval = DARTMAP_VALID | ((tmp >> DART_PAGE_SHIFT) &
+					 DARTMAP_RPNMASK);
 
-	/* Map in DART registers. FIXME: Use device node to get base address */
-	dart = ioremap(DART_BASE, 0x7000);
+	/* Map in DART registers */
+	dart = ioremap(r.start, r.end - r.start + 1);
 	if (dart == NULL)
-		panic("U3-DART: Cannot map registers!");
+		panic("DART: Cannot map registers!");
 
-	/* Set initial control register contents: table base, 
-	 * table size and enable bit
-	 */
-	regword = DARTCNTL_ENABLE | 
-		((dart_tablebase >> DART_PAGE_SHIFT) << DARTCNTL_BASE_SHIFT) |
-		(((dart_tablesize >> DART_PAGE_SHIFT) & DARTCNTL_SIZE_MASK)
-				 << DARTCNTL_SIZE_SHIFT);
+	/* Map in DART table */
 	dart_vbase = ioremap(virt_to_abs(dart_tablebase), dart_tablesize);
 
 	/* Fill initial table */
@@ -211,36 +213,50 @@ static int dart_init(struct device_node *dart_node)
 		dart_vbase[i] = dart_emptyval;
 
 	/* Initialize DART with table base and enable it. */
-	out_be32((unsigned int *)dart, regword);
+	base = dart_tablebase >> DART_PAGE_SHIFT;
+	size = dart_tablesize >> DART_PAGE_SHIFT;
+	if (dart_is_u4) {
+		BUG_ON(size & ~DART_SIZE_U4_SIZE_MASK);
+		DART_OUT(DART_BASE_U4, base);
+		DART_OUT(DART_SIZE_U4, size);
+		DART_OUT(DART_CNTL, DART_CNTL_U4_ENABLE);
+	} else {
+		BUG_ON(size & ~DART_CNTL_U3_SIZE_MASK);
+		DART_OUT(DART_CNTL,
+			 DART_CNTL_U3_ENABLE |
+			 (base << DART_CNTL_U3_BASE_SHIFT) |
+			 (size << DART_CNTL_U3_SIZE_SHIFT));
+	}
 
 	/* Invalidate DART to get rid of possible stale TLBs */
 	dart_tlb_invalidate_all();
 
-	printk(KERN_INFO "U3/CPC925 DART IOMMU initialized\n");
+	printk(KERN_INFO "DART IOMMU initialized for %s type chipset\n",
+	       dart_is_u4 ? "U4" : "U3");
 
 	return 0;
 }
 
-static void iommu_table_u3_setup(void)
+static void iommu_table_dart_setup(void)
 {
-	iommu_table_u3.it_busno = 0;
-	iommu_table_u3.it_offset = 0;
+	iommu_table_dart.it_busno = 0;
+	iommu_table_dart.it_offset = 0;
 	/* it_size is in number of entries */
-	iommu_table_u3.it_size = (dart_tablesize / sizeof(u32)) >> DART_PAGE_FACTOR;
+	iommu_table_dart.it_size = (dart_tablesize / sizeof(u32)) >> DART_PAGE_FACTOR;
 
 	/* Initialize the common IOMMU code */
-	iommu_table_u3.it_base = (unsigned long)dart_vbase;
-	iommu_table_u3.it_index = 0;
-	iommu_table_u3.it_blocksize = 1;
-	iommu_init_table(&iommu_table_u3);
+	iommu_table_dart.it_base = (unsigned long)dart_vbase;
+	iommu_table_dart.it_index = 0;
+	iommu_table_dart.it_blocksize = 1;
+	iommu_init_table(&iommu_table_dart);
 
 	/* Reserve the last page of the DART to avoid possible prefetch
 	 * past the DART mapped area
 	 */
-	set_bit(iommu_table_u3.it_size - 1, iommu_table_u3.it_map);
+	set_bit(iommu_table_dart.it_size - 1, iommu_table_dart.it_map);
 }
 
-static void iommu_dev_setup_u3(struct pci_dev *dev)
+static void iommu_dev_setup_dart(struct pci_dev *dev)
 {
 	struct device_node *dn;
 
@@ -254,35 +270,39 @@ static void iommu_dev_setup_u3(struct pci_dev *dev)
 	dn = pci_device_to_OF_node(dev);
 
 	if (dn)
-		PCI_DN(dn)->iommu_table = &iommu_table_u3;
+		PCI_DN(dn)->iommu_table = &iommu_table_dart;
 }
 
-static void iommu_bus_setup_u3(struct pci_bus *bus)
+static void iommu_bus_setup_dart(struct pci_bus *bus)
 {
 	struct device_node *dn;
 
-	if (!iommu_table_u3_inited) {
-		iommu_table_u3_inited = 1;
-		iommu_table_u3_setup();
+	if (!iommu_table_dart_inited) {
+		iommu_table_dart_inited = 1;
+		iommu_table_dart_setup();
 	}
 
 	dn = pci_bus_to_OF_node(bus);
 
 	if (dn)
-		PCI_DN(dn)->iommu_table = &iommu_table_u3;
+		PCI_DN(dn)->iommu_table = &iommu_table_dart;
 }
 
 static void iommu_dev_setup_null(struct pci_dev *dev) { }
 static void iommu_bus_setup_null(struct pci_bus *bus) { }
 
-void iommu_init_early_u3(void)
+void iommu_init_early_dart(void)
 {
 	struct device_node *dn;
 
 	/* Find the DART in the device-tree */
 	dn = of_find_compatible_node(NULL, "dart", "u3-dart");
+	if (dn == NULL) {
+		dn = of_find_compatible_node(NULL, "dart", "u4-dart");
 		if (dn == NULL)
-		return;
+			goto bail;
+		dart_is_u4 = 1;
+	}
 
 	/* Setup low level TCE operations for the core IOMMU code */
 	ppc_md.tce_build = dart_build;
@@ -290,24 +310,27 @@ void iommu_init_early_u3(void)
 	ppc_md.tce_flush = dart_flush;
 
 	/* Initialize the DART HW */
-	if (dart_init(dn)) {
+	if (dart_init(dn) == 0) {
+		ppc_md.iommu_dev_setup = iommu_dev_setup_dart;
+		ppc_md.iommu_bus_setup = iommu_bus_setup_dart;
+
+		/* Setup pci_dma ops */
+		pci_iommu_init();
+
+		return;
+	}
+
+ bail:
 	/* If init failed, use direct iommu and null setup functions */
 	ppc_md.iommu_dev_setup = iommu_dev_setup_null;
 	ppc_md.iommu_bus_setup = iommu_bus_setup_null;
 
 	/* Setup pci_dma ops */
 	pci_direct_iommu_init();
-	} else {
-		ppc_md.iommu_dev_setup = iommu_dev_setup_u3;
-		ppc_md.iommu_bus_setup = iommu_bus_setup_u3;
-
-		/* Setup pci_dma ops */
-		pci_iommu_init();
-	}
 }
 
 
-void __init alloc_u3_dart_table(void)
+void __init alloc_dart_table(void)
 {
 	/* Only reserve DART space if machine has more than 2GB of RAM
 	 * or if requested with iommu=on on cmdline.
@@ -323,5 +346,5 @@ void __init alloc_u3_dart_table(void)
 	dart_tablebase = (unsigned long)
 		abs_to_virt(lmb_alloc_base(1UL<<24, 1UL<<24, 0x80000000L));
 
-	printk(KERN_INFO "U3-DART allocated at: %lx\n", dart_tablebase);
+	printk(KERN_INFO "DART table allocated at: %lx\n", dart_tablebase);
 }

drivers/macintosh/smu.c

@@ -53,7 +53,7 @@
 #undef DEBUG_SMU
 
 #ifdef DEBUG_SMU
-#define DPRINTK(fmt, args...) do { udbg_printf(KERN_DEBUG fmt , ##args); } while (0)
+#define DPRINTK(fmt, args...) do { printk(KERN_DEBUG fmt , ##args); } while (0)
 #else
 #define DPRINTK(fmt, args...) do { } while (0)
 #endif
@@ -909,10 +909,13 @@ static struct smu_sdbp_header *smu_create_sdb_partition(int id)
 	struct property *prop;
 
 	/* First query the partition info */
+	DPRINTK("SMU: Query partition infos ... (irq=%d)\n", smu->db_irq);
 	smu_queue_simple(&cmd, SMU_CMD_PARTITION_COMMAND, 2,
 			 smu_done_complete, &comp,
 			 SMU_CMD_PARTITION_LATEST, id);
 	wait_for_completion(&comp);
+	DPRINTK("SMU: done, status: %d, reply_len: %d\n",
+		cmd.cmd.status, cmd.cmd.reply_len);
 
 	/* Partition doesn't exist (or other error) */
 	if (cmd.cmd.status != 0 || cmd.cmd.reply_len != 6)
@@ -975,6 +978,8 @@ struct smu_sdbp_header *__smu_get_sdb_partition(int id, unsigned int *size,
 
 	sprintf(pname, "sdb-partition-%02x", id);
 
+	DPRINTK("smu_get_sdb_partition(%02x)\n", id);
+
 	if (interruptible) {
 		int rc;
 		rc = down_interruptible(&smu_part_access);
@@ -986,6 +991,7 @@ struct smu_sdbp_header *__smu_get_sdb_partition(int id, unsigned int *size,
 	part = (struct smu_sdbp_header *)get_property(smu->of_node,
 						      pname, size);
 	if (part == NULL) {
+		DPRINTK("trying to extract from SMU ...\n");
 		part = smu_create_sdb_partition(id);
 		if (part != NULL && size)
 			*size = part->len << 2;

include/asm-powerpc/iommu.h

@@ -56,7 +56,7 @@ struct device_node;
 
 /* Walks all buses and creates iommu tables */
 extern void iommu_setup_pSeries(void);
-extern void iommu_setup_u3(void);
+extern void iommu_setup_dart(void);
 
 /* Frees table for an individual device node */
 extern void iommu_free_table(struct device_node *dn);
@@ -104,7 +104,7 @@ extern void iommu_unmap_single(struct iommu_table *tbl, dma_addr_t dma_handle,
 
 extern void iommu_init_early_pSeries(void);
 extern void iommu_init_early_iSeries(void);
-extern void iommu_init_early_u3(void);
+extern void iommu_init_early_dart(void);
 
 #ifdef CONFIG_PCI
 extern void pci_iommu_init(void);
@@ -113,6 +113,6 @@ extern void pci_direct_iommu_init(void);
 static inline void pci_iommu_init(void) { }
 #endif
 
-extern void alloc_u3_dart_table(void);
+extern void alloc_dart_table(void);
 
 #endif /* _ASM_IOMMU_H */

include/asm-powerpc/mpic.h

@@ -117,8 +117,9 @@ typedef int (*mpic_cascade_t)(struct pt_regs *regs, void *data);
 struct mpic_irq_fixup
 {
 	u8 __iomem	*base;
+	u8 __iomem	*applebase;
 	u32		data;
-	unsigned int	irq;
+	unsigned int	index;
 };
 #endif /* CONFIG_MPIC_BROKEN_U3 */
 

include/asm-powerpc/pmac_feature.h

@@ -121,6 +121,7 @@
 #define PMAC_TYPE_IMAC_G5		0x152	/* iMac G5 */
 #define PMAC_TYPE_XSERVE_G5		0x153	/* Xserve G5 */
 #define PMAC_TYPE_UNKNOWN_K2		0x19f	/* Any other K2 based */
+#define PMAC_TYPE_UNKNOWN_SHASTA       	0x19e	/* Any other Shasta based */
 
 /*
  * Motherboard flags
@@ -341,6 +342,7 @@ enum {
 	macio_pangea,
 	macio_intrepid,
 	macio_keylargo2,
+	macio_shasta,
 };
 
 struct macio_chip

include/linux/pci_regs.h

@@ -196,6 +196,7 @@
 #define  PCI_CAP_ID_MSI		0x05	/* Message Signalled Interrupts */
 #define  PCI_CAP_ID_CHSWP	0x06	/* CompactPCI HotSwap */
 #define  PCI_CAP_ID_PCIX	0x07	/* PCI-X */
+#define  PCI_CAP_ID_HT_IRQCONF	0x08	/* HyperTransport IRQ Configuration */
 #define  PCI_CAP_ID_SHPC 	0x0C	/* PCI Standard Hot-Plug Controller */
 #define  PCI_CAP_ID_EXP 	0x10	/* PCI Express */
 #define  PCI_CAP_ID_MSIX	0x11	/* MSI-X */