1 files changed, 425 insertions, 291 deletions

diff --git a/arch/powerpc/platforms/pseries/xics.c b/arch/powerpc/platforms/pseries/xics.c
index 2ffebe31cb2..e9886302572 100644
--- a/arch/powerpc/platforms/pseries/xics.c
+++ b/arch/powerpc/platforms/pseries/xics.c
@@ -8,6 +8,9 @@
  *  as published by the Free Software Foundation; either version
  *  2 of the License, or (at your option) any later version.
  */
+
+#undef DEBUG
+
 #include <linux/types.h>
 #include <linux/threads.h>
 #include <linux/kernel.h>
@@ -19,6 +22,7 @@
 #include <linux/gfp.h>
 #include <linux/radix-tree.h>
 #include <linux/cpu.h>
+
 #include <asm/firmware.h>
 #include <asm/prom.h>
 #include <asm/io.h>
@@ -31,26 +35,6 @@
 
 #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);
-
 #define XICS_IPI		2
 #define XICS_IRQ_SPURIOUS	0
 
@@ -81,12 +65,12 @@ struct xics_ipl {
 
 static struct xics_ipl __iomem *xics_per_cpu[NR_CPUS];
 
-static int xics_irq_8259_cascade = 0;
-static int xics_irq_8259_cascade_real = 0;
 static unsigned int default_server = 0xFF;
 static unsigned int default_distrib_server = 0;
 static unsigned int interrupt_server_size = 8;
 
+static struct irq_host *xics_host;
+
 /*
  * XICS only has a single IPI, so encode the messages per CPU
  */
@@ -98,48 +82,34 @@ 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 unsigned 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)
 {
 	return plpar_hcall_norets(H_EOI, xirr);
@@ -161,7 +131,7 @@ 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 unsigned int lpar_xirr_info_get(int n_cpu)
 {
 	unsigned long lpar_rc;
 	unsigned long return_value;
@@ -169,10 +139,10 @@ static int pSeriesLP_xirr_info_get(int n_cpu)
 	lpar_rc = plpar_xirr(&return_value);
 	if (lpar_rc != H_SUCCESS)
 		panic(" bad return code xirr - rc = %lx \n", lpar_rc);
-	return (int)return_value;
+	return (unsigned 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 +153,7 @@ 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 +162,7 @@ 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;
 
@@ -201,43 +171,16 @@ 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;
-
-	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;
+/* High level handlers and init code */
 
-	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)
+static int get_irq_server(unsigned int virq)
 {
 	unsigned int server;
 	/* For the moment only implement delivery to all cpus or one cpu */
-	cpumask_t cpumask = irq_desc[irq].affinity;
+	cpumask_t cpumask = irq_desc[virq].affinity;
 	cpumask_t tmp = CPU_MASK_NONE;
 
 	if (!distribute_irqs)
@@ -258,23 +201,28 @@ static int get_irq_server(unsigned int irq)
 
 }
 #else
-static int get_irq_server(unsigned int irq)
+static int get_irq_server(unsigned int virq)
 {
 	return default_server;
 }
 #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)
+	pr_debug("xics: unmask virq %d\n", virq);
+
+	irq = (unsigned int)irq_map[virq].hwirq;
+	pr_debug(" -> map to hwirq 0x%x\n", irq);
+	if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
 		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 +241,7 @@ 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 +266,86 @@ 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);
+	pr_debug("xics: mask virq %d\n", virq);
+
+	irq = (unsigned int)irq_map[virq].hwirq;
+	if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+		return;
+	xics_mask_real_irq(irq);
 }
 
-static void xics_end_irq(unsigned int irq)
+static unsigned int xics_startup(unsigned int virq)
+{
+	unsigned int irq;
+
+	/* force a reverse mapping of the interrupt so it gets in the cache */
+	irq = (unsigned int)irq_map[virq].hwirq;
+	irq_radix_revmap(xics_host, irq);
+
+	/* unmask it */
+	xics_unmask_irq(virq);
+	return 0;
+}
+
+static void xics_eoi_direct(unsigned int virq)
 {
 	int cpu = smp_processor_id();
+	unsigned int irq = (unsigned int)irq_map[virq].hwirq;
 
 	iosync();
-	ops->xirr_info_set(cpu, ((0xff << 24) |
-				 (virt_irq_to_real(irq_offset_down(irq)))));
-
+	direct_xirr_info_set(cpu, (0xff << 24) | irq);
 }
 
-static void xics_mask_and_ack_irq(unsigned int irq)
+
+static void xics_eoi_lpar(unsigned int virq)
 {
 	int cpu = smp_processor_id();
+	unsigned int irq = (unsigned int)irq_map[virq].hwirq;
 
-	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) | irq);
 }
 
-int xics_get_irq(struct pt_regs *regs)
+static inline unsigned int xics_remap_irq(unsigned int vec)
 {
-	unsigned int cpu = smp_processor_id();
-	unsigned int vec;
-	int irq;
+	unsigned 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 = irq_radix_revmap(xics_host, vec);
+	if (likely(irq != NO_IRQ))
+		return 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 unsigned 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 unsigned 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 +377,88 @@ 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 = (unsigned int)irq_map[virq].hwirq;
+	if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
+		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;
+	}
+}
+
 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
@@ -438,76 +467,256 @@ void xics_setup_cpu(void)
 	 *
 	 * XXX: undo of teardown on kexec needs this too, as may hotplug
 	 */
-	rtas_set_indicator(GLOBAL_INTERRUPT_QUEUE,
+	rtas_set_indicator_fast(GLOBAL_INTERRUPT_QUEUE,
 		(1UL << interrupt_server_size) - 1 - default_distrib_server, 1);
 }
 
-void xics_init_IRQ(void)
+
+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
+};
+
+
+static int xics_host_match(struct irq_host *h, struct device_node *node)
+{
+	/* IBM machines have interrupt parents of various funky types for things
+	 * like vdevices, events, etc... The trick we use here is to match
+	 * everything here except the legacy 8259 which is compatible "chrp,iic"
+	 */
+	return !device_is_compatible(node, "chrp,iic");
+}
+
+static int xics_host_map_direct(struct irq_host *h, unsigned int virq,
+				irq_hw_number_t hw)
+{
+	pr_debug("xics: map_direct virq %d, hwirq 0x%lx\n", virq, hw);
+
+	get_irq_desc(virq)->status |= IRQ_LEVEL;
+	set_irq_chip_and_handler(virq, &xics_pic_direct, handle_fasteoi_irq);
+	return 0;
+}
+
+static int xics_host_map_lpar(struct irq_host *h, unsigned int virq,
+			      irq_hw_number_t hw)
+{
+	pr_debug("xics: map_direct virq %d, hwirq 0x%lx\n", virq, hw);
+
+	get_irq_desc(virq)->status |= IRQ_LEVEL;
+	set_irq_chip_and_handler(virq, &xics_pic_lpar, handle_fasteoi_irq);
+	return 0;
+}
+
+static int xics_host_xlate(struct irq_host *h, struct device_node *ct,
+			   u32 *intspec, unsigned int intsize,
+			   irq_hw_number_t *out_hwirq, unsigned int *out_flags)
+
+{
+	/* Current xics implementation translates everything
+	 * to level. It is not technically right for MSIs but this
+	 * is irrelevant at this point. We might get smarter in the future
+	 */
+	*out_hwirq = intspec[0];
+	*out_flags = IRQ_TYPE_LEVEL_LOW;
+
+	return 0;
+}
+
+static struct irq_host_ops xics_host_direct_ops = {
+	.match = xics_host_match,
+	.map = xics_host_map_direct,
+	.xlate = xics_host_xlate,
+};
+
+static struct irq_host_ops xics_host_lpar_ops = {
+	.match = xics_host_match,
+	.map = xics_host_map_lpar,
+	.xlate = xics_host_xlate,
+};
+
+static void __init xics_init_host(void)
 {
+	struct irq_host_ops *ops;
+
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		ops = &xics_host_lpar_ops;
+	else
+		ops = &xics_host_direct_ops;
+	xics_host = irq_alloc_host(IRQ_HOST_MAP_TREE, 0, ops,
+				   XICS_IRQ_SPURIOUS);
+	BUG_ON(xics_host == NULL);
+	irq_set_default_host(xics_host);
+}
+
+static void __init xics_map_one_cpu(int hw_id, unsigned long addr,
+				     unsigned long size)
+{
+#ifdef CONFIG_SMP
 	int i;
-	unsigned long intr_size = 0;
-	struct device_node *np;
-	uint *ireg, ilen, indx = 0;
-	unsigned long intr_base = 0;
-	struct xics_interrupt_node {
-		unsigned long addr;
-		unsigned long size;
-	} intnodes[NR_CPUS];
 
-	ppc64_boot_msg(0x20, "XICS Init");
+	/* This may look gross but it's good enough for now, we don't quite
+	 * have a hard -> linux processor id matching.
+	 */
+	for_each_possible_cpu(i) {
+		if (!cpu_present(i))
+			continue;
+		if (hw_id == get_hard_smp_processor_id(i)) {
+			xics_per_cpu[i] = ioremap(addr, size);
+			return;
+		}
+	}
+#else
+	if (hw_id != 0)
+		return;
+	xics_per_cpu[0] = ioremap(addr, size);
+#endif /* CONFIG_SMP */
+}
 
-	ibm_get_xive = rtas_token("ibm,get-xive");
-	ibm_set_xive = rtas_token("ibm,set-xive");
-	ibm_int_on  = rtas_token("ibm,int-on");
-	ibm_int_off = rtas_token("ibm,int-off");
+static void __init xics_init_one_node(struct device_node *np,
+				      unsigned int *indx)
+{
+	unsigned int ilen;
+	u32 *ireg;
 
-	np = of_find_node_by_type(NULL, "PowerPC-External-Interrupt-Presentation");
-	if (!np)
-		panic("xics_init_IRQ: can't find interrupt presentation");
+	/* This code does the theorically broken assumption that the interrupt
+	 * server numbers are the same as the hard CPU numbers.
+	 * This happens to be the case so far but we are playing with fire...
+	 * should be fixed one of these days. -BenH.
+	 */
+	ireg = (u32 *)get_property(np, "ibm,interrupt-server-ranges", NULL);
 
-nextnode:
-	ireg = (uint *)get_property(np, "ibm,interrupt-server-ranges", NULL);
+	/* Do that ever happen ? we'll know soon enough... but even good'old
+	 * f80 does have that property ..
+	 */
+	WARN_ON(ireg == NULL);
 	if (ireg) {
 		/*
 		 * set node starting index for this node
 		 */
-		indx = *ireg;
+		*indx = *ireg;
 	}
-
-	ireg = (uint *)get_property(np, "reg", &ilen);
+	ireg = (u32 *)get_property(np, "reg", &ilen);
 	if (!ireg)
 		panic("xics_init_IRQ: can't find interrupt reg property");
 
-	while (ilen) {
-		intnodes[indx].addr = (unsigned long)*ireg++ << 32;
-		ilen -= sizeof(uint);
-		intnodes[indx].addr |= *ireg++;
-		ilen -= sizeof(uint);
-		intnodes[indx].size = (unsigned long)*ireg++ << 32;
-		ilen -= sizeof(uint);
-		intnodes[indx].size |= *ireg++;
-		ilen -= sizeof(uint);
-		indx++;
-		if (indx >= NR_CPUS) break;
+	while (ilen >= (4 * sizeof(u32))) {
+		unsigned long addr, size;
+
+		/* XXX Use proper OF parsing code here !!! */
+		addr = (unsigned long)*ireg++ << 32;
+		ilen -= sizeof(u32);
+		addr |= *ireg++;
+		ilen -= sizeof(u32);
+		size = (unsigned long)*ireg++ << 32;
+		ilen -= sizeof(u32);
+		size |= *ireg++;
+		ilen -= sizeof(u32);
+		xics_map_one_cpu(*indx, addr, size);
+		(*indx)++;
 	}
+}
+
+
+static void __init xics_setup_8259_cascade(void)
+{
+	struct device_node *np, *old, *found = NULL;
+	int cascade, naddr;
+	u32 *addrp;
+	unsigned long intack = 0;
+
+	for_each_node_by_type(np, "interrupt-controller")
+		if (device_is_compatible(np, "chrp,iic")) {
+			found = np;
+			break;
+		}
+	if (found == NULL) {
+		printk(KERN_DEBUG "xics: no ISA interrupt controller\n");
+		return;
+	}
+	cascade = irq_of_parse_and_map(found, 0);
+	if (cascade == NO_IRQ) {
+		printk(KERN_ERR "xics: failed to map cascade interrupt");
+		return;
+	}
+	pr_debug("xics: cascade mapped to irq %d\n", cascade);
 
-	np = of_find_node_by_type(np, "PowerPC-External-Interrupt-Presentation");
-	if ((indx < NR_CPUS) && np) goto nextnode;
+	for (old = of_node_get(found); old != NULL ; old = np) {
+		np = of_get_parent(old);
+		of_node_put(old);
+		if (np == NULL)
+			break;
+		if (strcmp(np->name, "pci") != 0)
+			continue;
+		addrp = (u32 *)get_property(np, "8259-interrupt-acknowledge", NULL);
+		if (addrp == NULL)
+			continue;
+		naddr = prom_n_addr_cells(np);
+		intack = addrp[naddr-1];
+		if (naddr > 1)
+			intack |= ((unsigned long)addrp[naddr-2]) << 32;
+	}
+	if (intack)
+		printk(KERN_DEBUG "xics: PCI 8259 intack at 0x%016lx\n", intack);
+	i8259_init(found, intack);
+	of_node_put(found);
+	set_irq_chained_handler(cascade, pseries_8259_cascade);
+}
+
+void __init xics_init_IRQ(void)
+{
+	int i;
+	struct device_node *np;
+	u32 *ireg, ilen, indx = 0;
+	int found = 0;
+
+	ppc64_boot_msg(0x20, "XICS Init");
+
+	ibm_get_xive = rtas_token("ibm,get-xive");
+	ibm_set_xive = rtas_token("ibm,set-xive");
+	ibm_int_on  = rtas_token("ibm,int-on");
+	ibm_int_off = rtas_token("ibm,int-off");
+
+	for_each_node_by_type(np, "PowerPC-External-Interrupt-Presentation") {
+		found = 1;
+		if (firmware_has_feature(FW_FEATURE_LPAR))
+			break;
+		xics_init_one_node(np, &indx);
+	}
+	if (found == 0)
+		return;
+
+	xics_init_host();
 
 	/* Find the server numbers for the boot cpu. */
 	for (np = of_find_node_by_type(NULL, "cpu");
 	     np;
 	     np = of_find_node_by_type(np, "cpu")) {
-		ireg = (uint *)get_property(np, "reg", &ilen);
+		ireg = (u32 *)get_property(np, "reg", &ilen);
 		if (ireg && ireg[0] == get_hard_smp_processor_id(boot_cpuid)) {
-			ireg = (uint *)get_property(np, "ibm,ppc-interrupt-gserver#s",
-						    &ilen);
+			ireg = (u32 *)get_property(np,
+						  "ibm,ppc-interrupt-gserver#s",
+						   &ilen);
 			i = ilen / sizeof(int);
 			if (ireg && i > 0) {
 				default_server = ireg[0];
-				default_distrib_server = ireg[i-1]; /* take last element */
+				/* take last element */
+				default_distrib_server = ireg[i-1];
 			}
-			ireg = (uint *)get_property(np,
+			ireg = (u32 *)get_property(np,
 					"ibm,interrupt-server#-size", NULL);
 			if (ireg)
 				interrupt_server_size = *ireg;
@@ -516,135 +725,56 @@ nextnode:
 	}
 	of_node_put(np);
 
-	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 {
-#ifdef CONFIG_SMP
-		for_each_possible_cpu(i) {
-			int hard_id;
-
-			/* FIXME: Do this dynamically! --RR */
-			if (!cpu_present(i))
-				continue;
-
-			hard_id = get_hard_smp_processor_id(i);
-			xics_per_cpu[i] = ioremap(intnodes[hard_id].addr,
-						  intnodes[hard_id].size);
-		}
-#else
-		xics_per_cpu[0] = ioremap(intr_base, intr_size);
-#endif /* CONFIG_SMP */
-	}
-
-	for (i = irq_offset_value(); i < NR_IRQS; ++i)
-		get_irq_desc(i)->chip = &xics_pic;
+		ppc_md.get_irq = xics_get_irq_lpar;
+	else
+		ppc_md.get_irq = xics_get_irq_direct;
 
 	xics_setup_cpu();
 
+	xics_setup_8259_cascade();
+
 	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");
-	}
-	return 0;
-}
-arch_initcall(xics_setup_i8259);
 
 #ifdef CONFIG_SMP
 void xics_request_IPIs(void)
 {
-	virt_irq_to_real_map[XICS_IPI] = XICS_IPI;
+	unsigned int ipi;
+
+	ipi = irq_create_mapping(xics_host, XICS_IPI);
+	BUG_ON(ipi == NO_IRQ);
 
 	/*
 	 * 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(ipi, handle_percpu_irq);
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		request_irq(ipi, xics_ipi_action_lpar, IRQF_DISABLED,
+			    "IPI", NULL);
+	else
+		request_irq(ipi, xics_ipi_action_direct, IRQF_DISABLED,
+			    "IPI", NULL);
 }
+#endif /* CONFIG_SMP */
 
 void xics_teardown_cpu(int secondary)
 {
 	int cpu = smp_processor_id();
+	unsigned int ipi;
+	struct irq_desc *desc;
 
-	ops->cppr_info(cpu, 0x00);
-	iosync();
+	xics_set_cpu_priority(cpu, 0);
 
-	/* Clear IPI */
-	ops->qirr_info(cpu, 0xff);
+	/*
+	 * Clear IPI
+	 */
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		lpar_qirr_info(cpu, 0xff);
+	else
+		direct_qirr_info(cpu, 0xff);
 
 	/*
 	 * we need to EOI the IPI if we got here from kexec down IPI
@@ -653,16 +783,22 @@ 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);
+
+	ipi = irq_find_mapping(xics_host, XICS_IPI);
+	if (ipi == XICS_IRQ_SPURIOUS)
+		return;
+	desc = get_irq_desc(ipi);
+	if (desc->chip && desc->chip->eoi)
+		desc->chip->eoi(ipi);
 
 	/*
 	 * Some machines need to have at least one cpu in the GIQ,
 	 * so leave the master cpu in the group.
 	 */
 	if (secondary)
-		rtas_set_indicator(GLOBAL_INTERRUPT_QUEUE,
-			(1UL << interrupt_server_size) - 1 -
-			default_distrib_server, 0);
+		rtas_set_indicator_fast(GLOBAL_INTERRUPT_QUEUE,
+				   (1UL << interrupt_server_size) - 1 -
+				   default_distrib_server, 0);
 }
 
 #ifdef CONFIG_HOTPLUG_CPU
@@ -674,33 +810,31 @@ 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,
+	status = rtas_set_indicator_fast(GLOBAL_INTERRUPT_QUEUE,
 		(1UL << interrupt_server_size) - 1 - default_distrib_server, 0);
 	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;
 
 		/* We cant set affinity on ISA interrupts */
-		if (virq < irq_offset_value())
+		if (virq < NUM_ISA_INTERRUPTS)
 			continue;
-
-		desc = get_irq_desc(virq);
-		irq = virt_irq_to_real(irq_offset_down(virq));
-
+		if (irq_map[virq].host != xics_host)
+			continue;
+		irq = (unsigned int)irq_map[virq].hwirq;
 		/* We need to get IPIs still. */
-		if (irq == XICS_IPI || irq == NO_IRQ)
+		if (irq == XICS_IPI || irq == XICS_IRQ_SPURIOUS)
 			continue;
+		desc = get_irq_desc(virq);
 
 		/* We only need to migrate enabled IRQS */
 		if (desc == NULL || desc->chip == NULL