[POWERPC] Use the genirq framework

This adapts the generic powerpc interrupt handling code, and all of
the platforms except for the embedded 6xx machines, to use the new
genirq framework.

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

10 files changed, 534 insertions, 456 deletions

diff --git a/arch/powerpc/platforms/cell/interrupt.c b/arch/powerpc/platforms/cell/interrupt.c
index 22da1335445..97936f547f1 100644
--- a/arch/powerpc/platforms/cell/interrupt.c
+++ b/arch/powerpc/platforms/cell/interrupt.c
@@ -37,64 +37,51 @@
 struct iic {
 	struct cbe_iic_thread_regs __iomem *regs;
 	u8 target_id;
+	u8 eoi_stack[16];
+	int eoi_ptr;
 };
 
 static DEFINE_PER_CPU(struct iic, iic);
 
-void iic_local_enable(void)
+static void iic_mask(unsigned int irq)
 {
-	struct iic *iic = &__get_cpu_var(iic);
-	u64 tmp;
-
-	/*
-	 * There seems to be a bug that is present in DD2.x CPUs
-	 * and still only partially fixed in DD3.1.
-	 * This bug causes a value written to the priority register
-	 * not to make it there, resulting in a system hang unless we
-	 * write it again.
-	 * Masking with 0xf0 is done because the Cell BE does not
-	 * implement the lower four bits of the interrupt priority,
-	 * they always read back as zeroes, although future CPUs
-	 * might implement different bits.
-	 */
-	do {
-		out_be64(&iic->regs->prio, 0xff);
-		tmp = in_be64(&iic->regs->prio);
-	} while ((tmp & 0xf0) != 0xf0);
-}
-
-void iic_local_disable(void)
-{
-	out_be64(&__get_cpu_var(iic).regs->prio, 0x0);
-}
-
-static unsigned int iic_startup(unsigned int irq)
-{
-	return 0;
-}
-
-static void iic_enable(unsigned int irq)
-{
-	iic_local_enable();
 }
 
-static void iic_disable(unsigned int irq)
+static void iic_unmask(unsigned int irq)
 {
 }
 
-static void iic_end(unsigned int irq)
+static void iic_eoi(unsigned int irq)
 {
-	iic_local_enable();
+	struct iic *iic = &__get_cpu_var(iic);
+	out_be64(&iic->regs->prio, iic->eoi_stack[--iic->eoi_ptr]);
+	BUG_ON(iic->eoi_ptr < 0);
 }
 
-static struct hw_interrupt_type iic_pic = {
+static struct irq_chip iic_chip = {
 	.typename = " CELL-IIC ",
-	.startup = iic_startup,
-	.enable = iic_enable,
-	.disable = iic_disable,
-	.end = iic_end,
+	.mask = iic_mask,
+	.unmask = iic_unmask,
+	.eoi = iic_eoi,
 };
 
+/* XXX All of this has to be reworked completely. We need to assign a real
+ * interrupt numbers to the external interrupts and remove all the hard coded
+ * interrupt maps (rely on the device-tree whenever possible).
+ *
+ * Basically, my scheme is to define the "pendings" bits to be the HW interrupt
+ * number (ignoring the data and flags here). That means we can sort-of split
+ * external sources based on priority, and we can use request_irq() on pretty
+ * much anything.
+ *
+ * For spider or axon, they have their own interrupt space. spider will just have
+ * local "hardward" interrupts 0...xx * node stride. The node stride is not
+ * necessary (separate interrupt chips will have separate HW number space), but
+ * will allow to be compatible with existing device-trees.
+ *
+ * All of thise little world will get a standard remapping scheme to map those HW
+ * numbers into the linux flat irq number space.
+*/
 static int iic_external_get_irq(struct cbe_iic_pending_bits pending)
 {
 	int irq;
@@ -118,9 +105,10 @@ static int iic_external_get_irq(struct cbe_iic_pending_bits pending)
 		 */
 		if (pending.class != 2)
 			break;
-		irq = IIC_EXT_OFFSET
-			+ spider_get_irq(node)
-			+ node * IIC_NODE_STRIDE;
+		/* TODO: We might want to silently ignore cascade interrupts
+		 * when no cascade handler exist yet
+		 */
+		irq = IIC_EXT_CASCADE + node * IIC_NODE_STRIDE;
 		break;
 	case 0x01 ... 0x04:
 	case 0x07 ... 0x0a:
@@ -152,6 +140,8 @@ int iic_get_irq(struct pt_regs *regs)
 	iic = &__get_cpu_var(iic);
 	*(unsigned long *) &pending = 
 		in_be64((unsigned long __iomem *) &iic->regs->pending_destr);
+	iic->eoi_stack[++iic->eoi_ptr] = pending.prio;
+	BUG_ON(iic->eoi_ptr > 15);
 
 	irq = -1;
 	if (pending.flags & CBE_IIC_IRQ_VALID) {
@@ -172,7 +162,7 @@ int iic_get_irq(struct pt_regs *regs)
 
 /* hardcoded part to be compatible with older firmware */
 
-static int setup_iic_hardcoded(void)
+static int __init setup_iic_hardcoded(void)
 {
 	struct device_node *np;
 	int nodeid, cpu;
@@ -207,12 +197,13 @@ static int setup_iic_hardcoded(void)
 		printk(KERN_INFO "IIC for CPU %d at %lx\n", cpu, regs);
 		iic->regs = ioremap(regs, sizeof(struct cbe_iic_thread_regs));
 		iic->target_id = (nodeid << 4) + ((cpu & 1) ? 0xf : 0xe);
+		iic->eoi_stack[0] = 0xff;
 	}
 
 	return 0;
 }
 
-static int setup_iic(void)
+static int __init setup_iic(void)
 {
 	struct device_node *dn;
 	unsigned long *regs;
@@ -248,11 +239,14 @@ static int setup_iic(void)
  		iic = &per_cpu(iic, np[0]);
  		iic->regs = ioremap(regs[0], sizeof(struct cbe_iic_thread_regs));
 		iic->target_id = ((np[0] & 2) << 3) + ((np[0] & 1) ? 0xf : 0xe);
+		iic->eoi_stack[0] = 0xff;
  		printk("IIC for CPU %d at %lx mapped to %p\n", np[0], regs[0], iic->regs);
 
  		iic = &per_cpu(iic, np[1]);
  		iic->regs = ioremap(regs[2], sizeof(struct cbe_iic_thread_regs));
 		iic->target_id = ((np[1] & 2) << 3) + ((np[1] & 1) ? 0xf : 0xe);
+		iic->eoi_stack[0] = 0xff;
+
  		printk("IIC for CPU %d at %lx mapped to %p\n", np[1], regs[2], iic->regs);
 
 		found++;
@@ -304,10 +298,10 @@ static void iic_request_ipi(int ipi, const char *name)
 	int irq;
 
 	irq = iic_ipi_to_irq(ipi);
+
 	/* IPIs are marked IRQF_DISABLED as they must run with irqs
 	 * disabled */
-	get_irq_desc(irq)->chip = &iic_pic;
-	get_irq_desc(irq)->status |= IRQ_PER_CPU;
+ 	set_irq_chip_and_handler(irq, &iic_chip, handle_percpu_irq);
 	request_irq(irq, iic_ipi_action, IRQF_DISABLED, name, NULL);
 }
 
@@ -321,20 +315,26 @@ void iic_request_IPIs(void)
 }
 #endif /* CONFIG_SMP */
 
-static void iic_setup_spe_handlers(void)
+static void __init iic_setup_builtin_handlers(void)
 {
 	int be, isrc;
 
-	/* Assume two threads per BE are present */
+	/* XXX FIXME: Assume two threads per BE are present */
 	for (be=0; be < num_present_cpus() / 2; be++) {
+		int irq;
+
+		/* setup SPE chip and handlers */
 		for (isrc = 0; isrc < IIC_CLASS_STRIDE * 3; isrc++) {
-			int irq = IIC_NODE_STRIDE * be + IIC_SPE_OFFSET + isrc;
-			get_irq_desc(irq)->chip = &iic_pic;
+			irq = IIC_NODE_STRIDE * be + IIC_SPE_OFFSET + isrc;
+			set_irq_chip_and_handler(irq, &iic_chip, handle_fasteoi_irq);
 		}
+		/* setup cascade chip */
+		irq = IIC_EXT_CASCADE + be * IIC_NODE_STRIDE;
+		set_irq_chip_and_handler(irq, &iic_chip, handle_fasteoi_irq);
 	}
 }
 
-void iic_init_IRQ(void)
+void __init iic_init_IRQ(void)
 {
 	int cpu, irq_offset;
 	struct iic *iic;
@@ -348,5 +348,6 @@ void iic_init_IRQ(void)
 		if (iic->regs)
 			out_be64(&iic->regs->prio, 0xff);
 	}
-	iic_setup_spe_handlers();
+	iic_setup_builtin_handlers();
+
 }
diff --git a/arch/powerpc/platforms/cell/interrupt.h b/arch/powerpc/platforms/cell/interrupt.h
index 799f77d98f9..c74515aeb63 100644
--- a/arch/powerpc/platforms/cell/interrupt.h
+++ b/arch/powerpc/platforms/cell/interrupt.h
@@ -38,6 +38,7 @@
 
 enum {
 	IIC_EXT_OFFSET   = 0x00, /* Start of south bridge IRQs */
+	IIC_EXT_CASCADE  = 0x20, /* There is no interrupt 32 on spider */
 	IIC_NUM_EXT      = 0x40, /* Number of south bridge IRQs */
 	IIC_SPE_OFFSET   = 0x40, /* Start of SPE interrupts */
 	IIC_CLASS_STRIDE = 0x10, /* SPE IRQs per class    */
@@ -51,13 +52,10 @@ extern int  iic_get_irq(struct pt_regs *regs);
 extern void iic_cause_IPI(int cpu, int mesg);
 extern void iic_request_IPIs(void);
 extern void iic_setup_cpu(void);
-extern void iic_local_enable(void);
-extern void iic_local_disable(void);
 
 extern u8 iic_get_target_id(int cpu);
 
 extern void spider_init_IRQ(void);
-extern int spider_get_irq(int node);
 
 #endif
 #endif /* ASM_CELL_PIC_H */
diff --git a/arch/powerpc/platforms/cell/setup.c b/arch/powerpc/platforms/cell/setup.c
index 01ec39431fb..70a4e903f7e 100644
--- a/arch/powerpc/platforms/cell/setup.c
+++ b/arch/powerpc/platforms/cell/setup.c
@@ -80,10 +80,14 @@ static void cell_progress(char *s, unsigned short hex)
 	printk("*** %04x : %s\n", hex, s ? s : "");
 }
 
+static void __init cell_init_irq(void)
+{
+	iic_init_IRQ();
+	spider_init_IRQ();
+}
+
 static void __init cell_setup_arch(void)
 {
-	ppc_md.init_IRQ       = iic_init_IRQ;
-	ppc_md.get_irq        = iic_get_irq;
 #ifdef CONFIG_SPU_BASE
 	spu_priv1_ops         = &spu_priv1_mmio_ops;
 #endif
@@ -109,7 +113,6 @@ static void __init cell_setup_arch(void)
 	/* Find and initialize PCI host bridges */
 	init_pci_config_tokens();
 	find_and_init_phbs();
-	spider_init_IRQ();
 	cbe_pervasive_init();
 #ifdef CONFIG_DUMMY_CONSOLE
 	conswitchp = &dummy_con;
@@ -174,6 +177,9 @@ define_machine(cell) {
 	.calibrate_decr		= generic_calibrate_decr,
 	.check_legacy_ioport	= cell_check_legacy_ioport,
 	.progress		= cell_progress,
+	.init_IRQ       	= cell_init_irq,
+	.get_irq        	= iic_get_irq,
+
 #ifdef CONFIG_KEXEC
 	.machine_kexec		= default_machine_kexec,
 	.machine_kexec_prepare	= default_machine_kexec_prepare,
diff --git a/arch/powerpc/platforms/cell/spider-pic.c b/arch/powerpc/platforms/cell/spider-pic.c
index 7c3a0b6d34f..98425acb6cd 100644
--- a/arch/powerpc/platforms/cell/spider-pic.c
+++ b/arch/powerpc/platforms/cell/spider-pic.c
@@ -82,17 +82,20 @@ static void __iomem *spider_get_irq_config(int irq)
 	return pic + TIR_CFGA + 8 * spider_get_nr(irq);
 }
 
-static void spider_enable_irq(unsigned int irq)
+static void spider_unmask_irq(unsigned int irq)
 {
 	int nodeid = (irq / IIC_NODE_STRIDE) * 0x10;
 	void __iomem *cfg = spider_get_irq_config(irq);
 	irq = spider_get_nr(irq);
 
+	/* FIXME: Most of that is configuration and has nothing to do with enabling/disable,
+	 * besides, it's also partially bogus.
+	 */
 	out_be32(cfg, (in_be32(cfg) & ~0xf0)| 0x3107000eu | nodeid);
 	out_be32(cfg + 4, in_be32(cfg + 4) | 0x00020000u | irq);
 }
 
-static void spider_disable_irq(unsigned int irq)
+static void spider_mask_irq(unsigned int irq)
 {
 	void __iomem *cfg = spider_get_irq_config(irq);
 	irq = spider_get_nr(irq);
@@ -100,39 +103,21 @@ static void spider_disable_irq(unsigned int irq)
 	out_be32(cfg, in_be32(cfg) & ~0x30000000u);
 }
 
-static unsigned int spider_startup_irq(unsigned int irq)
-{
-	spider_enable_irq(irq);
-	return 0;
-}
-
-static void spider_shutdown_irq(unsigned int irq)
-{
-	spider_disable_irq(irq);
-}
-
-static void spider_end_irq(unsigned int irq)
-{
-	spider_enable_irq(irq);
-}
-
 static void spider_ack_irq(unsigned int irq)
 {
-	spider_disable_irq(irq);
-	iic_local_enable();
+	/* Should reset edge detection logic but we don't configure any edge interrupt
+	 * at the moment.
+	 */
 }
 
-static struct hw_interrupt_type spider_pic = {
+static struct irq_chip spider_pic = {
 	.typename = " SPIDER   ",
-	.startup = spider_startup_irq,
-	.shutdown = spider_shutdown_irq,
-	.enable = spider_enable_irq,
-	.disable = spider_disable_irq,
+	.unmask = spider_unmask_irq,
+	.mask = spider_mask_irq,
 	.ack = spider_ack_irq,
-	.end = spider_end_irq,
 };
 
-int spider_get_irq(int node)
+static int spider_get_irq(int node)
 {
 	unsigned long cs;
 	void __iomem *regs = spider_pics[node];
@@ -145,95 +130,89 @@ int spider_get_irq(int node)
 		return cs;
 }
 
+static void spider_irq_cascade(unsigned int irq, struct irq_desc *desc,
+			       struct pt_regs *regs)
+{
+	int node = (int)(long)desc->handler_data;
+	int cascade_irq;
+
+	cascade_irq = spider_get_irq(node);
+	generic_handle_irq(cascade_irq, regs);
+	desc->chip->eoi(irq);
+}
+
 /* hardcoded part to be compatible with older firmware */
 
-void spider_init_IRQ_hardcoded(void)
+static void __init spider_init_one(int node, unsigned long addr)
 {
-	int node;
-	long spiderpic;
-	long pics[] = { 0x24000008000, 0x34000008000 };
-	int n;
-
-	pr_debug("%s(%d): Using hardcoded defaults\n", __FUNCTION__, __LINE__);
-
-	for (node = 0; node < num_present_cpus()/2; node++) {
-		spiderpic = pics[node];
-		printk(KERN_DEBUG "SPIDER addr: %lx\n", spiderpic);
-		spider_pics[node] = ioremap(spiderpic, 0x800);
-		for (n = 0; n < IIC_NUM_EXT; n++) {
-			int irq = n + IIC_EXT_OFFSET + node * IIC_NODE_STRIDE;
-			get_irq_desc(irq)->chip = &spider_pic;
-		}
-
- 		/* do not mask any interrupts because of level */
- 		out_be32(spider_pics[node] + TIR_MSK, 0x0);
-
- 		/* disable edge detection clear */
- 		/* out_be32(spider_pics[node] + TIR_EDC, 0x0); */
-
- 		/* enable interrupt packets to be output */
- 		out_be32(spider_pics[node] + TIR_PIEN,
-			in_be32(spider_pics[node] + TIR_PIEN) | 0x1);
-
- 		/* Enable the interrupt detection enable bit. Do this last! */
- 		out_be32(spider_pics[node] + TIR_DEN,
-			in_be32(spider_pics[node] + TIR_DEN) | 0x1);
+	int n, irq;
+
+	spider_pics[node] = ioremap(addr, 0x800);
+	if (spider_pics[node] == NULL)
+		panic("spider_pic: can't map registers !");
+
+	printk(KERN_INFO "spider_pic: mapped for node %d, addr: 0x%lx mapped to %p\n",
+	       node, addr, spider_pics[node]);
+
+	for (n = 0; n < IIC_NUM_EXT; n++) {
+		if (n == IIC_EXT_CASCADE)
+			continue;
+		irq = n + IIC_EXT_OFFSET + node * IIC_NODE_STRIDE;
+		set_irq_chip_and_handler(irq, &spider_pic, handle_level_irq);
+		get_irq_desc(irq)->status |= IRQ_LEVEL;
 	}
+
+	/* do not mask any interrupts because of level */
+	out_be32(spider_pics[node] + TIR_MSK, 0x0);
+
+	/* disable edge detection clear */
+	/* out_be32(spider_pics[node] + TIR_EDC, 0x0); */
+
+	/* enable interrupt packets to be output */
+	out_be32(spider_pics[node] + TIR_PIEN,
+		 in_be32(spider_pics[node] + TIR_PIEN) | 0x1);
+
+	/* Hook up cascade */
+	irq = IIC_EXT_CASCADE + node * IIC_NODE_STRIDE;
+	set_irq_data(irq, (void *)(long)node);
+	set_irq_chained_handler(irq, spider_irq_cascade);
+
+	/* Enable the interrupt detection enable bit. Do this last! */
+	out_be32(spider_pics[node] + TIR_DEN,
+		 in_be32(spider_pics[node] + TIR_DEN) | 0x1);
 }
 
-void spider_init_IRQ(void)
+void __init spider_init_IRQ(void)
 {
-	long spider_reg;
+	unsigned long *spider_reg;
 	struct device_node *dn;
 	char *compatible;
-	int n, node = 0;
-
+	int node = 0;
+
+	/* XXX node numbers are totally bogus. We _hope_ we get the device nodes in the right
+	 * order here but that's definitely not guaranteed, we need to get the node from the
+	 * device tree instead. There is currently no proper property for it (but our whole
+	 * device-tree is bogus anyway) so all we can do is pray or maybe test the address
+	 * and deduce the node-id
+	 */
 	for (dn = NULL; (dn = of_find_node_by_name(dn, "interrupt-controller"));) {
 		compatible = (char *)get_property(dn, "compatible", NULL);
 
 		if (!compatible)
 			continue;
 
-		if (strstr(compatible, "CBEA,platform-spider-pic"))
-			spider_reg = *(long *)get_property(dn,"reg", NULL);
-		else if (strstr(compatible, "sti,platform-spider-pic")) {
-			spider_init_IRQ_hardcoded();
-			return;
+ 		if (strstr(compatible, "CBEA,platform-spider-pic"))
+			spider_reg = (unsigned long *)get_property(dn, "reg", NULL);
+		else if (strstr(compatible, "sti,platform-spider-pic") && (node < 2)) {
+			static long hard_coded_pics[] = { 0x24000008000, 0x34000008000 };
+			spider_reg = &hard_coded_pics[node];
 		} else
 			continue;
 
-		if (!spider_reg)
-			printk("interrupt controller does not have reg property !\n");
-
-		n = prom_n_addr_cells(dn);
-
-		if ( n != 2)
-			printk("reg property with invalid number of elements \n");
-
-		spider_pics[node] = ioremap(spider_reg, 0x800);
-
-		printk("SPIDER addr: %lx with %i addr_cells mapped to %p\n",
-		       spider_reg, n, spider_pics[node]);
-
-		for (n = 0; n < IIC_NUM_EXT; n++) {
-			int irq = n + IIC_EXT_OFFSET + node * IIC_NODE_STRIDE;
-			get_irq_desc(irq)->chip = &spider_pic;
-		}
-
-		/* do not mask any interrupts because of level */
-		out_be32(spider_pics[node] + TIR_MSK, 0x0);
-
-		/* disable edge detection clear */
-		/* out_be32(spider_pics[node] + TIR_EDC, 0x0); */
-
-		/* enable interrupt packets to be output */
-		out_be32(spider_pics[node] + TIR_PIEN,
-			in_be32(spider_pics[node] + TIR_PIEN) | 0x1);
-
-		/* Enable the interrupt detection enable bit. Do this last! */
-		out_be32(spider_pics[node] + TIR_DEN,
-			in_be32(spider_pics[node] + TIR_DEN) | 0x1);
+		if (spider_reg == NULL)
+			printk(KERN_ERR "spider_pic: No address for node %d\n", node);
 
+		spider_init_one(node, *spider_reg);
 		node++;
 	}
 }
diff --git a/arch/powerpc/platforms/chrp/setup.c b/arch/powerpc/platforms/chrp/setup.c
index 1f1771b212b..a5dffc8d93a 100644
--- a/arch/powerpc/platforms/chrp/setup.c
+++ b/arch/powerpc/platforms/chrp/setup.c
@@ -315,6 +315,21 @@ chrp_event_scan(unsigned long unused)
 		  jiffies + event_scan_interval);
 }
 
+void chrp_8259_cascade(unsigned int irq, struct irq_desc *desc,
+			struct pt_regs *regs)
+{
+	unsigned int max = 100;
+
+	while(max--) {
+		int irq = i8259_irq(regs);
+		if (max == 99)
+			desc->chip->eoi(irq);
+		if (irq < 0)
+			break;
+		generic_handle_irq(irq, regs);
+	};
+}
+
 /*
  * Finds the open-pic node and sets up the mpic driver.
  */
@@ -402,7 +417,7 @@ static void __init chrp_find_openpic(void)
 	}
 
 	mpic_init(chrp_mpic);
-	mpic_setup_cascade(NUM_ISA_INTERRUPTS, i8259_irq_cascade, NULL);
+	set_irq_chained_handler(NUM_ISA_INTERRUPTS, chrp_8259_cascade);
 }
 
 #if defined(CONFIG_VT) && defined(CONFIG_INPUT_ADBHID) && defined(XMON)
diff --git a/arch/powerpc/platforms/iseries/irq.c b/arch/powerpc/platforms/iseries/irq.c
index f70e820e730..7fb6a08786c 100644
--- a/arch/powerpc/platforms/iseries/irq.c
+++ b/arch/powerpc/platforms/iseries/irq.c
@@ -297,13 +297,13 @@ static void iseries_end_IRQ(unsigned int irq)
 		(REAL_IRQ_TO_IDSEL(rirq) << 4) + REAL_IRQ_TO_FUNC(rirq));
 }
 
-static hw_irq_controller iSeries_IRQ_handler = {
-	.typename = "iSeries irq controller",
-	.startup = iseries_startup_IRQ,
-	.shutdown = iseries_shutdown_IRQ,
-	.enable = iseries_enable_IRQ,
-	.disable = iseries_disable_IRQ,
-	.end = iseries_end_IRQ
+static struct irq_chip iseries_pic = {
+	.typename	= "iSeries irq controller",
+	.startup	= iseries_startup_IRQ,
+	.shutdown	= iseries_shutdown_IRQ,
+	.unmask		= iseries_enable_IRQ,
+	.mask		= iseries_disable_IRQ,
+	.eoi		= iseries_end_IRQ
 };
 
 /*
@@ -322,8 +322,7 @@ int __init iSeries_allocate_IRQ(HvBusNumber bus,
 	realirq = (((((sub_bus << 8) + (bus - 1)) << 3) + (idsel - 1)) << 3)
 		+ function;
 	virtirq = virt_irq_create_mapping(realirq);
-
-	irq_desc[virtirq].chip = &iSeries_IRQ_handler;
+	set_irq_chip_and_handler(virtirq, &iseries_pic, handle_fasteoi_irq);
 	return virtirq;
 }
 
diff --git a/arch/powerpc/platforms/powermac/pic.c b/arch/powerpc/platforms/powermac/pic.c
index c9b09a9e605..58a4c7b90b8 100644
--- a/arch/powerpc/platforms/powermac/pic.c
+++ b/arch/powerpc/platforms/powermac/pic.c
@@ -70,18 +70,19 @@ static struct interrupt_info gatwick_int_pool[GATWICK_IRQ_POOL_SIZE];
 
 #define NR_MASK_WORDS	((NR_IRQS + 31) / 32)
 static unsigned long ppc_lost_interrupts[NR_MASK_WORDS];
+static unsigned long ppc_cached_irq_mask[NR_MASK_WORDS];
+static int pmac_irq_cascade = -1;
 
-/*
- * Mark an irq as "lost".  This is only used on the pmac
- * since it can lose interrupts (see pmac_set_irq_mask).
- * -- Cort
- */
-void __set_lost(unsigned long irq_nr, int nokick)
+static void __pmac_retrigger(unsigned int irq_nr)
 {
-	if (!test_and_set_bit(irq_nr, ppc_lost_interrupts)) {
+	if (irq_nr >= max_real_irqs && pmac_irq_cascade > 0) {
+		__set_bit(irq_nr, ppc_lost_interrupts);
+		irq_nr = pmac_irq_cascade;
+		mb();
+	}
+	if (!__test_and_set_bit(irq_nr, ppc_lost_interrupts)) {
 		atomic_inc(&ppc_n_lost_interrupts);
-		if (!nokick)
-			set_dec(1);
+		set_dec(1);
 	}
 }
 
@@ -94,10 +95,10 @@ static void pmac_mask_and_ack_irq(unsigned int irq_nr)
         if ((unsigned)irq_nr >= max_irqs)
                 return;
 
-        clear_bit(irq_nr, ppc_cached_irq_mask);
-        if (test_and_clear_bit(irq_nr, ppc_lost_interrupts))
-                atomic_dec(&ppc_n_lost_interrupts);
 	spin_lock_irqsave(&pmac_pic_lock, flags);
+        __clear_bit(irq_nr, ppc_cached_irq_mask);
+        if (__test_and_clear_bit(irq_nr, ppc_lost_interrupts))
+                atomic_dec(&ppc_n_lost_interrupts);
         out_le32(&pmac_irq_hw[i]->enable, ppc_cached_irq_mask[i]);
         out_le32(&pmac_irq_hw[i]->ack, bit);
         do {
@@ -109,7 +110,7 @@ static void pmac_mask_and_ack_irq(unsigned int irq_nr)
 	spin_unlock_irqrestore(&pmac_pic_lock, flags);
 }
 
-static void pmac_set_irq_mask(unsigned int irq_nr, int nokicklost)
+static void pmac_ack_irq(unsigned int irq_nr)
 {
         unsigned long bit = 1UL << (irq_nr & 0x1f);
         int i = irq_nr >> 5;
@@ -118,7 +119,22 @@ static void pmac_set_irq_mask(unsigned int irq_nr, int nokicklost)
         if ((unsigned)irq_nr >= max_irqs)
                 return;
 
-	spin_lock_irqsave(&pmac_pic_lock, flags);
+  	spin_lock_irqsave(&pmac_pic_lock, flags);
+	if (__test_and_clear_bit(irq_nr, ppc_lost_interrupts))
+                atomic_dec(&ppc_n_lost_interrupts);
+        out_le32(&pmac_irq_hw[i]->ack, bit);
+        (void)in_le32(&pmac_irq_hw[i]->ack);
+	spin_unlock_irqrestore(&pmac_pic_lock, flags);
+}
+
+static void __pmac_set_irq_mask(unsigned int irq_nr, int nokicklost)
+{
+        unsigned long bit = 1UL << (irq_nr & 0x1f);
+        int i = irq_nr >> 5;
+
+        if ((unsigned)irq_nr >= max_irqs)
+                return;
+
         /* enable unmasked interrupts */
         out_le32(&pmac_irq_hw[i]->enable, ppc_cached_irq_mask[i]);
 
@@ -135,8 +151,7 @@ static void pmac_set_irq_mask(unsigned int irq_nr, int nokicklost)
          * the bit in the flag register or request another interrupt.
          */
         if (bit & ppc_cached_irq_mask[i] & in_le32(&pmac_irq_hw[i]->level))
-		__set_lost((ulong)irq_nr, nokicklost);
-	spin_unlock_irqrestore(&pmac_pic_lock, flags);
+		__pmac_retrigger(irq_nr);
 }
 
 /* When an irq gets requested for the first client, if it's an
@@ -144,62 +159,67 @@ static void pmac_set_irq_mask(unsigned int irq_nr, int nokicklost)
  */
 static unsigned int pmac_startup_irq(unsigned int irq_nr)
 {
+	unsigned long flags;
         unsigned long bit = 1UL << (irq_nr & 0x1f);
         int i = irq_nr >> 5;
 
+  	spin_lock_irqsave(&pmac_pic_lock, flags);
 	if ((irq_desc[irq_nr].status & IRQ_LEVEL) == 0)
 		out_le32(&pmac_irq_hw[i]->ack, bit);
-        set_bit(irq_nr, ppc_cached_irq_mask);
-        pmac_set_irq_mask(irq_nr, 0);
+        __set_bit(irq_nr, ppc_cached_irq_mask);
+        __pmac_set_irq_mask(irq_nr, 0);
+  	spin_unlock_irqrestore(&pmac_pic_lock, flags);
 
 	return 0;
 }
 
 static void pmac_mask_irq(unsigned int irq_nr)
 {
-        clear_bit(irq_nr, ppc_cached_irq_mask);
-        pmac_set_irq_mask(irq_nr, 0);
-        mb();
+	unsigned long flags;
+
+  	spin_lock_irqsave(&pmac_pic_lock, flags);
+        __clear_bit(irq_nr, ppc_cached_irq_mask);
+        __pmac_set_irq_mask(irq_nr, 0);
+  	spin_unlock_irqrestore(&pmac_pic_lock, flags);
 }
 
 static void pmac_unmask_irq(unsigned int irq_nr)
 {
-        set_bit(irq_nr, ppc_cached_irq_mask);
-        pmac_set_irq_mask(irq_nr, 0);
+	unsigned long flags;
+
+	spin_lock_irqsave(&pmac_pic_lock, flags);
+	__set_bit(irq_nr, ppc_cached_irq_mask);
+        __pmac_set_irq_mask(irq_nr, 0);
+  	spin_unlock_irqrestore(&pmac_pic_lock, flags);
 }
 
-static void pmac_end_irq(unsigned int irq_nr)
+static int pmac_retrigger(unsigned int irq_nr)
 {
-	if (!(irq_desc[irq_nr].status & (IRQ_DISABLED|IRQ_INPROGRESS))
-	    && irq_desc[irq_nr].action) {
-        	set_bit(irq_nr, ppc_cached_irq_mask);
-	        pmac_set_irq_mask(irq_nr, 1);
-	}
-}
+	unsigned long flags;
 
+  	spin_lock_irqsave(&pmac_pic_lock, flags);
+	__pmac_retrigger(irq_nr);
+  	spin_unlock_irqrestore(&pmac_pic_lock, flags);
+	return 1;
+}
 
-struct hw_interrupt_type pmac_pic = {
+static struct irq_chip pmac_pic = {
 	.typename	= " PMAC-PIC ",
 	.startup	= pmac_startup_irq,
-	.enable		= pmac_unmask_irq,
-	.disable	= pmac_mask_irq,
-	.ack		= pmac_mask_and_ack_irq,
-	.end		= pmac_end_irq,
-};
-
-struct hw_interrupt_type gatwick_pic = {
-	.typename	= " GATWICK  ",
-	.startup	= pmac_startup_irq,
-	.enable		= pmac_unmask_irq,
-	.disable	= pmac_mask_irq,
-	.ack		= pmac_mask_and_ack_irq,
-	.end		= pmac_end_irq,
+	.mask		= pmac_mask_irq,
+	.ack		= pmac_ack_irq,
+	.mask_ack	= pmac_mask_and_ack_irq,
+	.unmask		= pmac_unmask_irq,
+	.retrigger	= pmac_retrigger,
 };
 
 static irqreturn_t gatwick_action(int cpl, void *dev_id, struct pt_regs *regs)
 {
+	unsigned long flags;
 	int irq, bits;
+	int rc = IRQ_NONE;
 
+  	spin_lock_irqsave(&pmac_pic_lock, flags);
 	for (irq = max_irqs; (irq -= 32) >= max_real_irqs; ) {
 		int i = irq >> 5;
 		bits = in_le32(&pmac_irq_hw[i]->event) | ppc_lost_interrupts[i];
@@ -209,17 +229,20 @@ static irqreturn_t gatwick_action(int cpl, void *dev_id, struct pt_regs *regs)
 		if (bits == 0)
 			continue;
 		irq += __ilog2(bits);
+		spin_unlock_irqrestore(&pmac_pic_lock, flags);
 		__do_IRQ(irq, regs);
-		return IRQ_HANDLED;
+		spin_lock_irqsave(&pmac_pic_lock, flags);
+		rc = IRQ_HANDLED;
 	}
-	printk("gatwick irq not from gatwick pic\n");
-	return IRQ_NONE;
+  	spin_unlock_irqrestore(&pmac_pic_lock, flags);
+	return rc;
 }
 
 static int pmac_get_irq(struct pt_regs *regs)
 {
 	int irq;
 	unsigned long bits = 0;
+	unsigned long flags;
 
 #ifdef CONFIG_SMP
 	void psurge_smp_message_recv(struct pt_regs *);
@@ -230,6 +253,7 @@ static int pmac_get_irq(struct pt_regs *regs)
 		return -2;	/* ignore, already handled */
         }
 #endif /* CONFIG_SMP */
+  	spin_lock_irqsave(&pmac_pic_lock, flags);
 	for (irq = max_real_irqs; (irq -= 32) >= 0; ) {
 		int i = irq >> 5;
 		bits = in_le32(&pmac_irq_hw[i]->event) | ppc_lost_interrupts[i];
@@ -241,6 +265,7 @@ static int pmac_get_irq(struct pt_regs *regs)
 		irq += __ilog2(bits);
 		break;
 	}
+  	spin_unlock_irqrestore(&pmac_pic_lock, flags);
 
 	return irq;
 }
@@ -389,7 +414,6 @@ static struct irqaction gatwick_cascade_action = {
 static void __init pmac_pic_probe_oldstyle(void)
 {
         int i;
-	int irq_cascade = -1;
         struct device_node *master = NULL;
 	struct device_node *slave = NULL;
 	u8 __iomem *addr;
@@ -443,9 +467,16 @@ static void __init pmac_pic_probe_oldstyle(void)
 	}
 	BUG_ON(master == NULL);
 
-	/* Set the handler for the main PIC */
-	for ( i = 0; i < max_real_irqs ; i++ )
-		irq_desc[i].chip = &pmac_pic;
+	/* Mark level interrupts and set handlers */
+	for (i = 0; i < max_irqs; i++) {
+		int level = !!(level_mask[i >> 5] & (1UL << (i & 0x1f)));
+		if (level)
+			irq_desc[i].status |= IRQ_LEVEL;
+		else
+			irq_desc[i].status |= IRQ_DELAYED_DISABLE;
+		set_irq_chip_and_handler(i, &pmac_pic, level ?
+					 handle_level_irq : handle_edge_irq);
+	}
 
 	/* Get addresses of first controller if we have a node for it */
 	BUG_ON(of_address_to_resource(master, 0, &r));
@@ -472,29 +503,22 @@ static void __init pmac_pic_probe_oldstyle(void)
 			pmac_irq_hw[i++] =
 				(volatile struct pmac_irq_hw __iomem *)
 				(addr + 0x10);
-		irq_cascade = slave->intrs[0].line;
+		pmac_irq_cascade = slave->intrs[0].line;
 
 		printk(KERN_INFO "irq: Found slave Apple PIC %s for %d irqs"
 		       " cascade: %d\n", slave->full_name,
-		       max_irqs - max_real_irqs, irq_cascade);
+		       max_irqs - max_real_irqs, pmac_irq_cascade);
 	}
 	of_node_put(slave);
 
-	/* disable all interrupts in all controllers */
+	/* Disable all interrupts in all controllers */
 	for (i = 0; i * 32 < max_irqs; ++i)
 		out_le32(&pmac_irq_hw[i]->enable, 0);
 
-	/* mark level interrupts */
-	for (i = 0; i < max_irqs; i++)
-		if (level_mask[i >> 5] & (1UL << (i & 0x1f)))
-			irq_desc[i].status = IRQ_LEVEL;
+	/* Hookup cascade irq */
+	if (slave)
+		setup_irq(pmac_irq_cascade, &gatwick_cascade_action);
 
-	/* Setup handlers for secondary controller and hook cascade irq*/
-	if (slave) {
-		for ( i = max_real_irqs ; i < max_irqs ; i++ )
-			irq_desc[i].chip = &gatwick_pic;
-		setup_irq(irq_cascade, &gatwick_cascade_action);
-	}
 	printk(KERN_INFO "irq: System has %d possible interrupts\n", max_irqs);
 #ifdef CONFIG_XMON
 	setup_irq(20, &xmon_action);
@@ -502,9 +526,20 @@ static void __init pmac_pic_probe_oldstyle(void)
 }
 #endif /* CONFIG_PPC32 */
 
-static int pmac_u3_cascade(struct pt_regs *regs, void *data)
+static void pmac_u3_cascade(unsigned int irq, struct irq_desc *desc,
+			    struct pt_regs *regs)
 {
-	return mpic_get_one_irq((struct mpic *)data, regs);
+	struct mpic *mpic = desc->handler_data;
+	unsigned int max = 100;
+
+	while(max--) {
+		int cascade_irq = mpic_get_one_irq(mpic, regs);
+		if (max == 99)
+			desc->chip->eoi(irq);
+		if (irq < 0)
+			break;
+		generic_handle_irq(cascade_irq, regs);
+	};
 }
 
 static void __init pmac_pic_setup_mpic_nmi(struct mpic *mpic)
@@ -612,7 +647,8 @@ static int __init pmac_pic_probe_mpic(void)
 		of_node_put(slave);
 		return 0;
 	}
-	mpic_setup_cascade(slave->intrs[0].line, pmac_u3_cascade, mpic2);
+	set_irq_data(slave->intrs[0].line, mpic2);
+	set_irq_chained_handler(slave->intrs[0].line, pmac_u3_cascade);
 
 	of_node_put(slave);
 	return 0;
diff --git a/arch/powerpc/platforms/pseries/setup.c b/arch/powerpc/platforms/pseries/setup.c
index 999509d28af..476b564a208 100644
--- a/arch/powerpc/platforms/pseries/setup.c
+++ b/arch/powerpc/platforms/pseries/setup.c
@@ -118,6 +118,21 @@ static void __init fwnmi_init(void)
 		fwnmi_active = 1;
 }
 
+void pSeries_8259_cascade(unsigned int irq, struct irq_desc *desc,
+			  struct pt_regs *regs)
+{
+	unsigned int max = 100;
+
+	while(max--) {
+		int cascade_irq = i8259_irq(regs);
+		if (max == 99)
+			desc->chip->eoi(irq);
+		if (cascade_irq < 0)
+			break;
+		generic_handle_irq(cascade_irq, regs);
+	};
+}
+
 static void __init pSeries_init_mpic(void)
 {
         unsigned int *addrp;
@@ -140,7 +155,7 @@ static void __init pSeries_init_mpic(void)
 	i8259_init(intack, 0);
 
 	/* Hook cascade to mpic */
-	mpic_setup_cascade(NUM_ISA_INTERRUPTS, i8259_irq_cascade, NULL);
+	set_irq_chained_handler(NUM_ISA_INTERRUPTS, pSeries_8259_cascade);
 }
 
 static void __init pSeries_setup_mpic(void)
@@ -201,10 +216,8 @@ static void __init pSeries_setup_arch(void)
 		/* Allocate the mpic now, so that find_and_init_phbs() can
 		 * fill the ISUs */
 		pSeries_setup_mpic();
-	} else {
+	} else
 		ppc_md.init_IRQ       = xics_init_IRQ;
-		ppc_md.get_irq        = xics_get_irq;
-	}
 
 #ifdef CONFIG_SMP
 	smp_init_pSeries();
@@ -291,10 +304,7 @@ static void pSeries_mach_cpu_die(void)
 {
 	local_irq_disable();
 	idle_task_exit();
-	/* Some hardware requires clearing the CPPR, while other hardware does not
-	 * it is safe either way
-	 */
-	pSeriesLP_cppr_info(0, 0);
+	xics_teardown_cpu(0);
 	rtas_stop_self();
 	/* Should never get here... */
 	BUG();
diff --git a/arch/powerpc/platforms/pseries/xics.c b/arch/powerpc/platforms/pseries/xics.c
index 2ffebe31cb2..c7f04420066 100644
--- a/arch/powerpc/platforms/pseries/xics.c
+++ b/arch/powerpc/platforms/pseries/xics.c
@@ -31,23 +31,6 @@
 
 #include "xics.h"
 
-static unsigned int xics_startup(unsigned int irq);
-static void xics_enable_irq(unsigned int irq);
-static void xics_disable_irq(unsigned int irq);
-static void xics_mask_and_ack_irq(unsigned int irq);
-static void xics_end_irq(unsigned int irq);
-static void xics_set_affinity(unsigned int irq_nr, cpumask_t cpumask);
-
-static struct hw_interrupt_type xics_pic = {
-	.typename = " XICS     ",
-	.startup = xics_startup,
-	.enable = xics_enable_irq,
-	.disable = xics_disable_irq,
-	.ack = xics_mask_and_ack_irq,
-	.end = xics_end_irq,
-	.set_affinity = xics_set_affinity
-};
-
 /* This is used to map real irq numbers to virtual */
 static struct radix_tree_root irq_map = RADIX_TREE_INIT(GFP_ATOMIC);
 
@@ -98,47 +81,33 @@ static int ibm_set_xive;
 static int ibm_int_on;
 static int ibm_int_off;
 
-typedef struct {
-	int (*xirr_info_get)(int cpu);
-	void (*xirr_info_set)(int cpu, int val);
-	void (*cppr_info)(int cpu, u8 val);
-	void (*qirr_info)(int cpu, u8 val);
-} xics_ops;
 
+/* Direct HW low level accessors */
 
-/* SMP */
 
-static int pSeries_xirr_info_get(int n_cpu)
+static inline int direct_xirr_info_get(int n_cpu)
 {
 	return in_be32(&xics_per_cpu[n_cpu]->xirr.word);
 }
 
-static void pSeries_xirr_info_set(int n_cpu, int value)
+static inline void direct_xirr_info_set(int n_cpu, int value)
 {
 	out_be32(&xics_per_cpu[n_cpu]->xirr.word, value);
 }
 
-static void pSeries_cppr_info(int n_cpu, u8 value)
+static inline void direct_cppr_info(int n_cpu, u8 value)
 {
 	out_8(&xics_per_cpu[n_cpu]->xirr.bytes[0], value);
 }
 
-static void pSeries_qirr_info(int n_cpu, u8 value)
+static inline void direct_qirr_info(int n_cpu, u8 value)
 {
 	out_8(&xics_per_cpu[n_cpu]->qirr.bytes[0], value);
 }
 
-static xics_ops pSeries_ops = {
-	pSeries_xirr_info_get,
-	pSeries_xirr_info_set,
-	pSeries_cppr_info,
-	pSeries_qirr_info
-};
-
-static xics_ops *ops = &pSeries_ops;
 
+/* LPAR low level accessors */
 
-/* LPAR */
 
 static inline long plpar_eoi(unsigned long xirr)
 {
@@ -161,7 +130,7 @@ static inline long plpar_xirr(unsigned long *xirr_ret)
 	return plpar_hcall(H_XIRR, 0, 0, 0, 0, xirr_ret, &dummy, &dummy);
 }
 
-static int pSeriesLP_xirr_info_get(int n_cpu)
+static inline int lpar_xirr_info_get(int n_cpu)
 {
 	unsigned long lpar_rc;
 	unsigned long return_value;
@@ -172,7 +141,7 @@ static int pSeriesLP_xirr_info_get(int n_cpu)
 	return (int)return_value;
 }
 
-static void pSeriesLP_xirr_info_set(int n_cpu, int value)
+static inline void lpar_xirr_info_set(int n_cpu, int value)
 {
 	unsigned long lpar_rc;
 	unsigned long val64 = value & 0xffffffff;
@@ -183,7 +152,7 @@ static void pSeriesLP_xirr_info_set(int n_cpu, int value)
 		      val64);
 }
 
-void pSeriesLP_cppr_info(int n_cpu, u8 value)
+static inline void lpar_cppr_info(int n_cpu, u8 value)
 {
 	unsigned long lpar_rc;
 
@@ -192,7 +161,7 @@ void pSeriesLP_cppr_info(int n_cpu, u8 value)
 		panic("bad return code cppr - rc = %lx\n", lpar_rc);
 }
 
-static void pSeriesLP_qirr_info(int n_cpu , u8 value)
+static inline void lpar_qirr_info(int n_cpu , u8 value)
 {
 	unsigned long lpar_rc;
 
@@ -201,36 +170,9 @@ static void pSeriesLP_qirr_info(int n_cpu , u8 value)
 		panic("bad return code qirr - rc = %lx\n", lpar_rc);
 }
 
-xics_ops pSeriesLP_ops = {
-	pSeriesLP_xirr_info_get,
-	pSeriesLP_xirr_info_set,
-	pSeriesLP_cppr_info,
-	pSeriesLP_qirr_info
-};
 
-static unsigned int xics_startup(unsigned int virq)
-{
-	unsigned int irq;
+/* High level handlers and init code */
 
-	irq = irq_offset_down(virq);
-	if (radix_tree_insert(&irq_map, virt_irq_to_real(irq),
-			      &virt_irq_to_real_map[irq]) == -ENOMEM)
-		printk(KERN_CRIT "Out of memory creating real -> virtual"
-		       " IRQ mapping for irq %u (real 0x%x)\n",
-		       virq, virt_irq_to_real(irq));
-	xics_enable_irq(virq);
-	return 0;	/* return value is ignored */
-}
-
-static unsigned int real_irq_to_virt(unsigned int real_irq)
-{
-	unsigned int *ptr;
-
-	ptr = radix_tree_lookup(&irq_map, real_irq);
-	if (ptr == NULL)
-		return NO_IRQ;
-	return ptr - virt_irq_to_real_map;
-}
 
 #ifdef CONFIG_SMP
 static int get_irq_server(unsigned int irq)
@@ -264,17 +206,20 @@ static int get_irq_server(unsigned int irq)
 }
 #endif
 
-static void xics_enable_irq(unsigned int virq)
+
+static void xics_unmask_irq(unsigned int virq)
 {
 	unsigned int irq;
 	int call_status;
 	unsigned int server;
 
 	irq = virt_irq_to_real(irq_offset_down(virq));
-	if (irq == XICS_IPI)
+	WARN_ON(irq == NO_IRQ);
+	if (irq == XICS_IPI || irq == NO_IRQ)
 		return;
 
 	server = get_irq_server(virq);
+
 	call_status = rtas_call(ibm_set_xive, 3, 1, NULL, irq, server,
 				DEFAULT_PRIORITY);
 	if (call_status != 0) {
@@ -293,7 +238,7 @@ static void xics_enable_irq(unsigned int virq)
 	}
 }
 
-static void xics_disable_real_irq(unsigned int irq)
+static void xics_mask_real_irq(unsigned int irq)
 {
 	int call_status;
 	unsigned int server;
@@ -318,75 +263,104 @@ static void xics_disable_real_irq(unsigned int irq)
 	}
 }
 
-static void xics_disable_irq(unsigned int virq)
+static void xics_mask_irq(unsigned int virq)
 {
 	unsigned int irq;
 
 	irq = virt_irq_to_real(irq_offset_down(virq));
-	xics_disable_real_irq(irq);
+	WARN_ON(irq == NO_IRQ);
+	if (irq != NO_IRQ)
+		xics_mask_real_irq(irq);
+}
+
+static void xics_set_irq_revmap(unsigned int virq)
+{
+	unsigned int irq;
+
+	irq = irq_offset_down(virq);
+	if (radix_tree_insert(&irq_map, virt_irq_to_real(irq),
+			      &virt_irq_to_real_map[irq]) == -ENOMEM)
+		printk(KERN_CRIT "Out of memory creating real -> virtual"
+		       " IRQ mapping for irq %u (real 0x%x)\n",
+		       virq, virt_irq_to_real(irq));
 }
 
-static void xics_end_irq(unsigned int irq)
+static unsigned int xics_startup(unsigned int virq)
+{
+	xics_set_irq_revmap(virq);
+	xics_unmask_irq(virq);
+	return 0;
+}
+
+static unsigned int real_irq_to_virt(unsigned int real_irq)
+{
+	unsigned int *ptr;
+
+	ptr = radix_tree_lookup(&irq_map, real_irq);
+	if (ptr == NULL)
+		return NO_IRQ;
+	return ptr - virt_irq_to_real_map;
+}
+
+static void xics_eoi_direct(unsigned int irq)
 {
 	int cpu = smp_processor_id();
 
 	iosync();
-	ops->xirr_info_set(cpu, ((0xff << 24) |
-				 (virt_irq_to_real(irq_offset_down(irq)))));
-
+	direct_xirr_info_set(cpu, ((0xff << 24) |
+				   (virt_irq_to_real(irq_offset_down(irq)))));
 }
 
-static void xics_mask_and_ack_irq(unsigned int irq)
+
+static void xics_eoi_lpar(unsigned int irq)
 {
 	int cpu = smp_processor_id();
 
-	if (irq < irq_offset_value()) {
-		i8259_pic.ack(irq);
-		iosync();
-		ops->xirr_info_set(cpu, ((0xff<<24) |
-					 xics_irq_8259_cascade_real));
-		iosync();
-	}
+	iosync();
+	lpar_xirr_info_set(cpu, ((0xff << 24) |
+				 (virt_irq_to_real(irq_offset_down(irq)))));
+
 }
 
-int xics_get_irq(struct pt_regs *regs)
+static inline int xics_remap_irq(int vec)
 {
-	unsigned int cpu = smp_processor_id();
-	unsigned int vec;
 	int irq;
 
-	vec = ops->xirr_info_get(cpu);
-	/*  (vec >> 24) == old priority */
 	vec &= 0x00ffffff;
 
-	/* for sanity, this had better be < NR_IRQS - 16 */
-	if (vec == xics_irq_8259_cascade_real) {
-		irq = i8259_irq(regs);
-		xics_end_irq(irq_offset_up(xics_irq_8259_cascade));
-	} else if (vec == XICS_IRQ_SPURIOUS) {
-		irq = -1;
-	} else {
-		irq = real_irq_to_virt(vec);
-		if (irq == NO_IRQ)
-			irq = real_irq_to_virt_slowpath(vec);
-		if (irq == NO_IRQ) {
-			printk(KERN_ERR "Interrupt %u (real) is invalid,"
-			       " disabling it.\n", vec);
-			xics_disable_real_irq(vec);
-		} else
-			irq = irq_offset_up(irq);
-	}
-	return irq;
+	if (vec == XICS_IRQ_SPURIOUS)
+		return NO_IRQ;
+
+	irq = real_irq_to_virt(vec);
+	if (irq == NO_IRQ)
+		irq = real_irq_to_virt_slowpath(vec);
+	if (likely(irq != NO_IRQ))
+		return irq_offset_up(irq);
+
+	printk(KERN_ERR "Interrupt %u (real) is invalid,"
+	       " disabling it.\n", vec);
+	xics_mask_real_irq(vec);
+	return NO_IRQ;
 }
 
-#ifdef CONFIG_SMP
+static int xics_get_irq_direct(struct pt_regs *regs)
+{
+	unsigned int cpu = smp_processor_id();
 
-static irqreturn_t xics_ipi_action(int irq, void *dev_id, struct pt_regs *regs)
+	return xics_remap_irq(direct_xirr_info_get(cpu));
+}
+
+static int xics_get_irq_lpar(struct pt_regs *regs)
 {
-	int cpu = smp_processor_id();
+	unsigned int cpu = smp_processor_id();
 
-	ops->qirr_info(cpu, 0xff);
+	return xics_remap_irq(lpar_xirr_info_get(cpu));
+}
+
+#ifdef CONFIG_SMP
 
+static irqreturn_t xics_ipi_dispatch(int cpu, struct pt_regs *regs)
+{
 	WARN_ON(cpu_is_offline(cpu));
 
 	while (xics_ipi_message[cpu].value) {
@@ -418,18 +392,108 @@ static irqreturn_t xics_ipi_action(int irq, void *dev_id, struct pt_regs *regs)
 	return IRQ_HANDLED;
 }
 
+static irqreturn_t xics_ipi_action_direct(int irq, void *dev_id, struct pt_regs *regs)
+{
+	int cpu = smp_processor_id();
+
+	direct_qirr_info(cpu, 0xff);
+
+	return xics_ipi_dispatch(cpu, regs);
+}
+
+static irqreturn_t xics_ipi_action_lpar(int irq, void *dev_id, struct pt_regs *regs)
+{
+	int cpu = smp_processor_id();
+
+	lpar_qirr_info(cpu, 0xff);
+
+	return xics_ipi_dispatch(cpu, regs);
+}
+
 void xics_cause_IPI(int cpu)
 {
-	ops->qirr_info(cpu, IPI_PRIORITY);
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		lpar_qirr_info(cpu, IPI_PRIORITY);
+	else
+		direct_qirr_info(cpu, IPI_PRIORITY);
 }
+
 #endif /* CONFIG_SMP */
 
+static void xics_set_cpu_priority(int cpu, unsigned char cppr)
+{
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		lpar_cppr_info(cpu, cppr);
+	else
+		direct_cppr_info(cpu, cppr);
+	iosync();
+}
+
+static void xics_set_affinity(unsigned int virq, cpumask_t cpumask)
+{
+	unsigned int irq;
+	int status;
+	int xics_status[2];
+	unsigned long newmask;
+	cpumask_t tmp = CPU_MASK_NONE;
+
+	irq = virt_irq_to_real(irq_offset_down(virq));
+	if (irq == XICS_IPI || irq == NO_IRQ)
+		return;
+
+	status = rtas_call(ibm_get_xive, 1, 3, xics_status, irq);
+
+	if (status) {
+		printk(KERN_ERR "xics_set_affinity: irq=%u ibm,get-xive "
+		       "returns %d\n", irq, status);
+		return;
+	}
+
+	/* For the moment only implement delivery to all cpus or one cpu */
+	if (cpus_equal(cpumask, CPU_MASK_ALL)) {
+		newmask = default_distrib_server;
+	} else {
+		cpus_and(tmp, cpu_online_map, cpumask);
+		if (cpus_empty(tmp))
+			return;
+		newmask = get_hard_smp_processor_id(first_cpu(tmp));
+	}
+
+	status = rtas_call(ibm_set_xive, 3, 1, NULL,
+				irq, newmask, xics_status[1]);
+
+	if (status) {
+		printk(KERN_ERR "xics_set_affinity: irq=%u ibm,set-xive "
+		       "returns %d\n", irq, status);
+		return;
+	}
+}
+
+static struct irq_chip xics_pic_direct = {
+	.typename = " XICS     ",
+	.startup = xics_startup,
+	.mask = xics_mask_irq,
+	.unmask = xics_unmask_irq,
+	.eoi = xics_eoi_direct,
+	.set_affinity = xics_set_affinity
+};
+
+
+static struct irq_chip xics_pic_lpar = {
+	.typename = " XICS     ",
+	.startup = xics_startup,
+	.mask = xics_mask_irq,
+	.unmask = xics_unmask_irq,
+	.eoi = xics_eoi_lpar,
+	.set_affinity = xics_set_affinity
+};
+
+
 void xics_setup_cpu(void)
 {
 	int cpu = smp_processor_id();
 
-	ops->cppr_info(cpu, 0xff);
-	iosync();
+	xics_set_cpu_priority(cpu, 0xff);
 
 	/*
 	 * Put the calling processor into the GIQ.  This is really only
@@ -453,6 +517,7 @@ void xics_init_IRQ(void)
 		unsigned long addr;
 		unsigned long size;
 	} intnodes[NR_CPUS];
+	struct irq_chip *chip;
 
 	ppc64_boot_msg(0x20, "XICS Init");
 
@@ -519,26 +584,10 @@ nextnode:
 	intr_base = intnodes[0].addr;
 	intr_size = intnodes[0].size;
 
-	np = of_find_node_by_type(NULL, "interrupt-controller");
-	if (!np) {
-		printk(KERN_DEBUG "xics: no ISA interrupt controller\n");
-		xics_irq_8259_cascade_real = -1;
-		xics_irq_8259_cascade = -1;
-	} else {
-		ireg = (uint *) get_property(np, "interrupts", NULL);
-		if (!ireg)
-			panic("xics_init_IRQ: can't find ISA interrupts property");
-
-		xics_irq_8259_cascade_real = *ireg;
-		xics_irq_8259_cascade
-			= virt_irq_create_mapping(xics_irq_8259_cascade_real);
-		i8259_init(0, 0);
-		of_node_put(np);
-	}
-
-	if (firmware_has_feature(FW_FEATURE_LPAR))
-		ops = &pSeriesLP_ops;
-	else {
+ 	if (firmware_has_feature(FW_FEATURE_LPAR)) {
+ 		ppc_md.get_irq = xics_get_irq_lpar;
+ 		chip = &xics_pic_lpar;
+  	} else {
 #ifdef CONFIG_SMP
 		for_each_possible_cpu(i) {
 			int hard_id;
@@ -554,32 +603,54 @@ nextnode:
 #else
 		xics_per_cpu[0] = ioremap(intr_base, intr_size);
 #endif /* CONFIG_SMP */
+		ppc_md.get_irq = xics_get_irq_direct;
+		chip = &xics_pic_direct;
+
 	}
 
-	for (i = irq_offset_value(); i < NR_IRQS; ++i)
-		get_irq_desc(i)->chip = &xics_pic;
+	for (i = irq_offset_value(); i < NR_IRQS; ++i) {
+		/* All IRQs on XICS are level for now. MSI code may want to modify
+		 * that for reporting purposes
+		 */
+		get_irq_desc(i)->status |= IRQ_LEVEL;
+		set_irq_chip_and_handler(i, chip, handle_fasteoi_irq);
+	}
 
 	xics_setup_cpu();
 
 	ppc64_boot_msg(0x21, "XICS Done");
 }
 
-/*
- * We cant do this in init_IRQ because we need the memory subsystem up for
- * request_irq()
- */
-static int __init xics_setup_i8259(void)
-{
-	if (ppc64_interrupt_controller == IC_PPC_XIC &&
-	    xics_irq_8259_cascade != -1) {
-		if (request_irq(irq_offset_up(xics_irq_8259_cascade),
-				no_action, 0, "8259 cascade", NULL))
-			printk(KERN_ERR "xics_setup_i8259: couldn't get 8259 "
-					"cascade\n");
+static int xics_setup_8259_cascade(void)
+{
+	struct device_node *np;
+	uint *ireg;
+
+	np = of_find_node_by_type(NULL, "interrupt-controller");
+	if (np == NULL) {
+		printk(KERN_WARNING "xics: no ISA interrupt controller\n");
+		xics_irq_8259_cascade_real = -1;
+		xics_irq_8259_cascade = -1;
+		return 0;
 	}
+
+	ireg = (uint *) get_property(np, "interrupts", NULL);
+	if (!ireg)
+		panic("xics_init_IRQ: can't find ISA interrupts property");
+
+	xics_irq_8259_cascade_real = *ireg;
+	xics_irq_8259_cascade = irq_offset_up
+		(virt_irq_create_mapping(xics_irq_8259_cascade_real));
+	i8259_init(0, 0);
+	of_node_put(np);
+
+	xics_set_irq_revmap(xics_irq_8259_cascade);
+	set_irq_chained_handler(xics_irq_8259_cascade, pSeries_8259_cascade);
+
 	return 0;
 }
-arch_initcall(xics_setup_i8259);
+arch_initcall(xics_setup_8259_cascade);
+
 
 #ifdef CONFIG_SMP
 void xics_request_IPIs(void)
@@ -590,61 +661,22 @@ void xics_request_IPIs(void)
 	 * IPIs are marked IRQF_DISABLED as they must run with irqs
 	 * disabled
 	 */
-	request_irq(irq_offset_up(XICS_IPI), xics_ipi_action,
-		    IRQF_DISABLED, "IPI", NULL);
-	get_irq_desc(irq_offset_up(XICS_IPI))->status |= IRQ_PER_CPU;
-}
-#endif
-
-static void xics_set_affinity(unsigned int virq, cpumask_t cpumask)
-{
-	unsigned int irq;
-	int status;
-	int xics_status[2];
-	unsigned long newmask;
-	cpumask_t tmp = CPU_MASK_NONE;
-
-	irq = virt_irq_to_real(irq_offset_down(virq));
-	if (irq == XICS_IPI || irq == NO_IRQ)
-		return;
-
-	status = rtas_call(ibm_get_xive, 1, 3, xics_status, irq);
-
-	if (status) {
-		printk(KERN_ERR "xics_set_affinity: irq=%u ibm,get-xive "
-		       "returns %d\n", irq, status);
-		return;
-	}
-
-	/* For the moment only implement delivery to all cpus or one cpu */
-	if (cpus_equal(cpumask, CPU_MASK_ALL)) {
-		newmask = default_distrib_server;
-	} else {
-		cpus_and(tmp, cpu_online_map, cpumask);
-		if (cpus_empty(tmp))
-			return;
-		newmask = get_hard_smp_processor_id(first_cpu(tmp));
-	}
-
-	status = rtas_call(ibm_set_xive, 3, 1, NULL,
-				irq, newmask, xics_status[1]);
-
-	if (status) {
-		printk(KERN_ERR "xics_set_affinity: irq=%u ibm,set-xive "
-		       "returns %d\n", irq, status);
-		return;
-	}
+	set_irq_handler(irq_offset_up(XICS_IPI), handle_percpu_irq);
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		request_irq(irq_offset_up(XICS_IPI), xics_ipi_action_lpar,
+			    SA_INTERRUPT, "IPI", NULL);
+	else
+		request_irq(irq_offset_up(XICS_IPI), xics_ipi_action_direct,
+			    SA_INTERRUPT, "IPI", NULL);
 }
+#endif /* CONFIG_SMP */
 
 void xics_teardown_cpu(int secondary)
 {
+	struct irq_desc *desc = get_irq_desc(irq_offset_up(XICS_IPI));
 	int cpu = smp_processor_id();
 
-	ops->cppr_info(cpu, 0x00);
-	iosync();
-
-	/* Clear IPI */
-	ops->qirr_info(cpu, 0xff);
+ 	xics_set_cpu_priority(cpu, 0);
 
 	/*
 	 * we need to EOI the IPI if we got here from kexec down IPI
@@ -653,7 +685,8 @@ void xics_teardown_cpu(int secondary)
 	 * should we be flagging idle loop instead?
 	 * or creating some task to be scheduled?
 	 */
-	ops->xirr_info_set(cpu, XICS_IPI);
+	if (desc->chip && desc->chip->eoi)
+		desc->chip->eoi(XICS_IPI);
 
 	/*
 	 * Some machines need to have at least one cpu in the GIQ,
@@ -674,8 +707,7 @@ void xics_migrate_irqs_away(void)
 	unsigned int irq, virq, cpu = smp_processor_id();
 
 	/* Reject any interrupt that was queued to us... */
-	ops->cppr_info(cpu, 0);
-	iosync();
+	xics_set_cpu_priority(cpu, 0);
 
 	/* remove ourselves from the global interrupt queue */
 	status = rtas_set_indicator(GLOBAL_INTERRUPT_QUEUE,
@@ -683,11 +715,10 @@ void xics_migrate_irqs_away(void)
 	WARN_ON(status < 0);
 
 	/* Allow IPIs again... */
-	ops->cppr_info(cpu, DEFAULT_PRIORITY);
-	iosync();
+	xics_set_cpu_priority(cpu, DEFAULT_PRIORITY);
 
 	for_each_irq(virq) {
-		irq_desc_t *desc;
+		struct irq_desc *desc;
 		int xics_status[2];
 		unsigned long flags;
 
diff --git a/arch/powerpc/platforms/pseries/xics.h b/arch/powerpc/platforms/pseries/xics.h
index e14c70868f1..67dedf3514e 100644
--- a/arch/powerpc/platforms/pseries/xics.h
+++ b/arch/powerpc/platforms/pseries/xics.h
@@ -14,13 +14,12 @@
 
 #include <linux/cache.h>
 
-void xics_init_IRQ(void);
-int xics_get_irq(struct pt_regs *);
-void xics_setup_cpu(void);
-void xics_teardown_cpu(int secondary);
-void xics_cause_IPI(int cpu);
-void xics_request_IPIs(void);
-void xics_migrate_irqs_away(void);
+extern void xics_init_IRQ(void);
+extern void xics_setup_cpu(void);
+extern void xics_teardown_cpu(int secondary);
+extern void xics_cause_IPI(int cpu);
+extern  void xics_request_IPIs(void);
+extern void xics_migrate_irqs_away(void);
 
 /* first argument is ignored for now*/
 void pSeriesLP_cppr_info(int n_cpu, u8 value);
@@ -31,4 +30,8 @@ struct xics_ipi_struct {
 
 extern struct xics_ipi_struct xics_ipi_message[NR_CPUS] __cacheline_aligned;
 
+struct irq_desc;
+extern void pSeries_8259_cascade(unsigned int irq, struct irq_desc *desc,
+				 struct pt_regs *regs);
+
 #endif /* _POWERPC_KERNEL_XICS_H */