[POWERPC] Add new interrupt mapping core and change platforms to use it

This adds the new irq remapper core and removes the old one.  Because
there are some fundamental conflicts with the old code, like the value
of NO_IRQ which I'm now setting to 0 (as per discussions with Linus),
etc..., this commit also changes the relevant platform and driver code
over to use the new remapper (so as not to cause difficulties later
in bisecting).

This patch removes the old pre-parsing of the open firmware interrupt
tree along with all the bogus assumptions it made to try to renumber
interrupts according to the platform. This is all to be handled by the
new code now.

For the pSeries XICS interrupt controller, a single remapper host is
created for the whole machine regardless of how many interrupt
presentation and source controllers are found, and it's set to match
any device node that isn't a 8259.  That works fine on pSeries and
avoids having to deal with some of the complexities of split source
controllers vs. presentation controllers in the pSeries device trees.

The powerpc i8259 PIC driver now always requests the legacy interrupt
range. It also has the feature of being able to match any device node
(including NULL) if passed no device node as an input. That will help
porting over platforms with broken device-trees like Pegasos who don't
have a proper interrupt tree.

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

5 files changed, 496 insertions, 351 deletions

diff --git a/arch/powerpc/platforms/pseries/ras.c b/arch/powerpc/platforms/pseries/ras.c
index 9639c66b453..9df783088b6 100644
--- a/arch/powerpc/platforms/pseries/ras.c
+++ b/arch/powerpc/platforms/pseries/ras.c
@@ -72,32 +72,62 @@ static irqreturn_t ras_error_interrupt(int irq, void *dev_id,
 
 /* #define DEBUG */
 
-static void request_ras_irqs(struct device_node *np, char *propname,
+
+static void request_ras_irqs(struct device_node *np,
 			irqreturn_t (*handler)(int, void *, struct pt_regs *),
 			const char *name)
 {
-	unsigned int *ireg, len, i;
-	int virq, n_intr;
-
-	ireg = (unsigned int *)get_property(np, propname, &len);
-	if (ireg == NULL)
-		return;
-	n_intr = prom_n_intr_cells(np);
-	len /= n_intr * sizeof(*ireg);
-
-	for (i = 0; i < len; i++) {
-		virq = virt_irq_create_mapping(*ireg);
-		if (virq == NO_IRQ) {
-			printk(KERN_ERR "Unable to allocate interrupt "
-			       "number for %s\n", np->full_name);
-			return;
+	int i, index, count = 0;
+	struct of_irq oirq;
+	u32 *opicprop;
+	unsigned int opicplen;
+	unsigned int virqs[16];
+
+	/* Check for obsolete "open-pic-interrupt" property. If present, then
+	 * map those interrupts using the default interrupt host and default
+	 * trigger
+	 */
+	opicprop = (u32 *)get_property(np, "open-pic-interrupt", &opicplen);
+	if (opicprop) {
+		opicplen /= sizeof(u32);
+		for (i = 0; i < opicplen; i++) {
+			if (count > 15)
+				break;
+			virqs[count] = irq_create_mapping(NULL, *(opicprop++),
+							 IRQ_TYPE_NONE);
+			if (virqs[count] == NO_IRQ)
+				printk(KERN_ERR "Unable to allocate interrupt "
+				       "number for %s\n", np->full_name);
+			else
+				count++;
+
 		}
-		if (request_irq(irq_offset_up(virq), handler, 0, name, NULL)) {
+	}
+	/* Else use normal interrupt tree parsing */
+	else {
+		/* First try to do a proper OF tree parsing */
+		for (index = 0; of_irq_map_one(np, index, &oirq) == 0;
+		     index++) {
+			if (count > 15)
+				break;
+			virqs[count] = irq_create_of_mapping(oirq.controller,
+							    oirq.specifier,
+							    oirq.size);
+			if (virqs[count] == NO_IRQ)
+				printk(KERN_ERR "Unable to allocate interrupt "
+				       "number for %s\n", np->full_name);
+			else
+				count++;
+		}
+	}
+
+	/* Now request them */
+	for (i = 0; i < count; i++) {
+		if (request_irq(virqs[i], handler, 0, name, NULL)) {
 			printk(KERN_ERR "Unable to request interrupt %d for "
-			       "%s\n", irq_offset_up(virq), np->full_name);
+			       "%s\n", virqs[i], np->full_name);
 			return;
 		}
-		ireg += n_intr;
 	}
 }
 
@@ -115,20 +145,14 @@ static int __init init_ras_IRQ(void)
 	/* Internal Errors */
 	np = of_find_node_by_path("/event-sources/internal-errors");
 	if (np != NULL) {
-		request_ras_irqs(np, "open-pic-interrupt", ras_error_interrupt,
-				 "RAS_ERROR");
-		request_ras_irqs(np, "interrupts", ras_error_interrupt,
-				 "RAS_ERROR");
+		request_ras_irqs(np, ras_error_interrupt, "RAS_ERROR");
 		of_node_put(np);
 	}
 
 	/* EPOW Events */
 	np = of_find_node_by_path("/event-sources/epow-events");
 	if (np != NULL) {
-		request_ras_irqs(np, "open-pic-interrupt", ras_epow_interrupt,
-				 "RAS_EPOW");
-		request_ras_irqs(np, "interrupts", ras_epow_interrupt,
-				 "RAS_EPOW");
+		request_ras_irqs(np, ras_epow_interrupt, "RAS_EPOW");
 		of_node_put(np);
 	}
 
@@ -162,7 +186,7 @@ ras_epow_interrupt(int irq, void *dev_id, struct pt_regs * regs)
 
 	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
 			   RAS_VECTOR_OFFSET,
-			   virt_irq_to_real(irq_offset_down(irq)),
+			   irq_map[irq].hwirq,
 			   RTAS_EPOW_WARNING | RTAS_POWERMGM_EVENTS,
 			   critical, __pa(&ras_log_buf),
 				rtas_get_error_log_max());
@@ -198,7 +222,7 @@ ras_error_interrupt(int irq, void *dev_id, struct pt_regs * regs)
 
 	status = rtas_call(ras_check_exception_token, 6, 1, NULL,
 			   RAS_VECTOR_OFFSET,
-			   virt_irq_to_real(irq_offset_down(irq)),
+			   irq_map[irq].hwirq,
 			   RTAS_INTERNAL_ERROR, 1 /*Time Critical */,
 			   __pa(&ras_log_buf),
 				rtas_get_error_log_max());
diff --git a/arch/powerpc/platforms/pseries/setup.c b/arch/powerpc/platforms/pseries/setup.c
index 476b564a208..54a52437265 100644
--- a/arch/powerpc/platforms/pseries/setup.c
+++ b/arch/powerpc/platforms/pseries/setup.c
@@ -76,6 +76,9 @@
 #define DBG(fmt...)
 #endif
 
+/* move those away to a .h */
+extern void smp_init_pseries_mpic(void);
+extern void smp_init_pseries_xics(void);
 extern void find_udbg_vterm(void);
 
 int fwnmi_active;  /* TRUE if an FWNMI handler is present */
@@ -83,7 +86,7 @@ int fwnmi_active;  /* TRUE if an FWNMI handler is present */
 static void pseries_shared_idle_sleep(void);
 static void pseries_dedicated_idle_sleep(void);
 
-struct mpic *pSeries_mpic;
+static struct device_node *pSeries_mpic_node;
 
 static void pSeries_show_cpuinfo(struct seq_file *m)
 {
@@ -118,78 +121,92 @@ static void __init fwnmi_init(void)
 		fwnmi_active = 1;
 }
 
-void pSeries_8259_cascade(unsigned int irq, struct irq_desc *desc,
+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;
+	unsigned int cascade_irq = i8259_irq(regs);
+	if (cascade_irq != NO_IRQ)
 		generic_handle_irq(cascade_irq, regs);
-	};
+	desc->chip->eoi(irq);
 }
 
-static void __init pSeries_init_mpic(void)
+static void __init pseries_mpic_init_IRQ(void)
 {
+	struct device_node *np, *old, *cascade = NULL;
         unsigned int *addrp;
-	struct device_node *np;
 	unsigned long intack = 0;
-
-	/* All ISUs are setup, complete initialization */
-	mpic_init(pSeries_mpic);
-
-	/* Check what kind of cascade ACK we have */
-        if (!(np = of_find_node_by_name(NULL, "pci"))
-            || !(addrp = (unsigned int *)
-                 get_property(np, "8259-interrupt-acknowledge", NULL)))
-                printk(KERN_ERR "Cannot find pci to get ack address\n");
-        else
-		intack = addrp[prom_n_addr_cells(np)-1];
-	of_node_put(np);
-
-	/* Setup the legacy interrupts & controller */
-	i8259_init(intack, 0);
-
-	/* Hook cascade to mpic */
-	set_irq_chained_handler(NUM_ISA_INTERRUPTS, pSeries_8259_cascade);
-}
-
-static void __init pSeries_setup_mpic(void)
-{
 	unsigned int *opprop;
 	unsigned long openpic_addr = 0;
-        unsigned char senses[NR_IRQS - NUM_ISA_INTERRUPTS];
-        struct device_node *root;
-	int irq_count;
+	unsigned int cascade_irq;
+	int naddr, n, i, opplen;
+	struct mpic *mpic;
 
-	/* Find the Open PIC if present */
-	root = of_find_node_by_path("/");
-	opprop = (unsigned int *) get_property(root, "platform-open-pic", NULL);
+	np = of_find_node_by_path("/");
+	naddr = prom_n_addr_cells(np);
+	opprop = (unsigned int *) get_property(np, "platform-open-pic", &opplen);
 	if (opprop != 0) {
-		int n = prom_n_addr_cells(root);
-
-		for (openpic_addr = 0; n > 0; --n)
-			openpic_addr = (openpic_addr << 32) + *opprop++;
+		openpic_addr = of_read_number(opprop, naddr);
 		printk(KERN_DEBUG "OpenPIC addr: %lx\n", openpic_addr);
 	}
-	of_node_put(root);
+	of_node_put(np);
 
 	BUG_ON(openpic_addr == 0);
 
-	/* Get the sense values from OF */
-	prom_get_irq_senses(senses, NUM_ISA_INTERRUPTS, NR_IRQS);
-	
 	/* Setup the openpic driver */
-	irq_count = NR_IRQS - NUM_ISA_INTERRUPTS - 4; /* leave room for IPIs */
-	pSeries_mpic = mpic_alloc(openpic_addr, MPIC_PRIMARY,
-				  16, 16, irq_count, /* isu size, irq offset, irq count */ 
-				  NR_IRQS - 4, /* ipi offset */
-				  senses, irq_count, /* sense & sense size */
-				  " MPIC     ");
+	mpic = mpic_alloc(pSeries_mpic_node, openpic_addr,
+			  MPIC_PRIMARY,
+			  16, 250, /* isu size, irq count */
+			  " MPIC     ");
+	BUG_ON(mpic == NULL);
+
+	/* Add ISUs */
+	opplen /= sizeof(u32);
+	for (n = 0, i = naddr; i < opplen; i += naddr, n++) {
+		unsigned long isuaddr = of_read_number(opprop + i, naddr);
+		mpic_assign_isu(mpic, n, isuaddr);
+	}
+
+	/* All ISUs are setup, complete initialization */
+	mpic_init(mpic);
+
+	/* Look for cascade */
+	for_each_node_by_type(np, "interrupt-controller")
+		if (device_is_compatible(np, "chrp,iic")) {
+			cascade = np;
+			break;
+		}
+	if (cascade == NULL)
+		return;
+
+	cascade_irq = irq_of_parse_and_map(cascade, 0);
+	if (cascade == NO_IRQ) {
+		printk(KERN_ERR "xics: failed to map cascade interrupt");
+		return;
+	}
+
+	/* Check ACK type */
+	for (old = of_node_get(cascade); 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 "mpic: PCI 8259 intack at 0x%016lx\n",
+		       intack);
+	i8259_init(cascade, intack);
+	of_node_put(cascade);
+	set_irq_chained_handler(cascade_irq, pseries_8259_cascade);
 }
 
 static void pseries_lpar_enable_pmcs(void)
@@ -207,21 +224,67 @@ static void pseries_lpar_enable_pmcs(void)
 		get_lppaca()->pmcregs_in_use = 1;
 }
 
-static void __init pSeries_setup_arch(void)
+#ifdef CONFIG_KEXEC
+static void pseries_kexec_cpu_down_mpic(int crash_shutdown, int secondary)
 {
-	/* Fixup ppc_md depending on the type of interrupt controller */
-	if (ppc64_interrupt_controller == IC_OPEN_PIC) {
-		ppc_md.init_IRQ       = pSeries_init_mpic;
-		ppc_md.get_irq        = mpic_get_irq;
-		/* Allocate the mpic now, so that find_and_init_phbs() can
-		 * fill the ISUs */
-		pSeries_setup_mpic();
-	} else
-		ppc_md.init_IRQ       = xics_init_IRQ;
+	mpic_teardown_this_cpu(secondary);
+}
 
+static void pseries_kexec_cpu_down_xics(int crash_shutdown, int secondary)
+{
+	/* Don't risk a hypervisor call if we're crashing */
+	if (firmware_has_feature(FW_FEATURE_SPLPAR) && !crash_shutdown) {
+		unsigned long vpa = __pa(get_lppaca());
+
+		if (unregister_vpa(hard_smp_processor_id(), vpa)) {
+			printk("VPA deregistration of cpu %u (hw_cpu_id %d) "
+					"failed\n", smp_processor_id(),
+					hard_smp_processor_id());
+		}
+	}
+	xics_teardown_cpu(secondary);
+}
+#endif /* CONFIG_KEXEC */
+
+static void __init pseries_discover_pic(void)
+{
+	struct device_node *np;
+	char *typep;
+
+	for (np = NULL; (np = of_find_node_by_name(np,
+						   "interrupt-controller"));) {
+		typep = (char *)get_property(np, "compatible", NULL);
+		if (strstr(typep, "open-pic")) {
+			pSeries_mpic_node = of_node_get(np);
+			ppc_md.init_IRQ       = pseries_mpic_init_IRQ;
+			ppc_md.get_irq        = mpic_get_irq;
+#ifdef CONFIG_KEXEC
+			ppc_md.kexec_cpu_down = pseries_kexec_cpu_down_mpic;
+#endif
 #ifdef CONFIG_SMP
-	smp_init_pSeries();
+			smp_init_pseries_mpic();
 #endif
+			return;
+		} else if (strstr(typep, "ppc-xicp")) {
+			ppc_md.init_IRQ       = xics_init_IRQ;
+#ifdef CONFIG_KEXEC
+			ppc_md.kexec_cpu_down = pseries_kexec_cpu_down_xics;
+#endif
+#ifdef CONFIG_SMP
+			smp_init_pseries_xics();
+#endif
+			return;
+		}
+	}
+	printk(KERN_ERR "pSeries_discover_pic: failed to recognize"
+	       " interrupt-controller\n");
+}
+
+static void __init pSeries_setup_arch(void)
+{
+	/* Discover PIC type and setup ppc_md accordingly */
+	pseries_discover_pic();
+
 	/* openpic global configuration register (64-bit format). */
 	/* openpic Interrupt Source Unit pointer (64-bit format). */
 	/* python0 facility area (mmio) (64-bit format) REAL address. */
@@ -273,33 +336,6 @@ static int __init pSeries_init_panel(void)
 }
 arch_initcall(pSeries_init_panel);
 
-static  void __init pSeries_discover_pic(void)
-{
-	struct device_node *np;
-	char *typep;
-
-	/*
-	 * Setup interrupt mapping options that are needed for finish_device_tree
-	 * to properly parse the OF interrupt tree & do the virtual irq mapping
-	 */
-	__irq_offset_value = NUM_ISA_INTERRUPTS;
-	ppc64_interrupt_controller = IC_INVALID;
-	for (np = NULL; (np = of_find_node_by_name(np, "interrupt-controller"));) {
-		typep = (char *)get_property(np, "compatible", NULL);
-		if (strstr(typep, "open-pic")) {
-			ppc64_interrupt_controller = IC_OPEN_PIC;
-			break;
-		} else if (strstr(typep, "ppc-xicp")) {
-			ppc64_interrupt_controller = IC_PPC_XIC;
-			break;
-		}
-	}
-	if (ppc64_interrupt_controller == IC_INVALID)
-		printk("pSeries_discover_pic: failed to recognize"
-			" interrupt-controller\n");
-
-}
-
 static void pSeries_mach_cpu_die(void)
 {
 	local_irq_disable();
@@ -342,8 +378,6 @@ static void __init pSeries_init_early(void)
 
 	iommu_init_early_pSeries();
 
-	pSeries_discover_pic();
-
 	DBG(" <- pSeries_init_early()\n");
 }
 
@@ -515,27 +549,6 @@ static int pSeries_pci_probe_mode(struct pci_bus *bus)
 	return PCI_PROBE_NORMAL;
 }
 
-#ifdef CONFIG_KEXEC
-static void pseries_kexec_cpu_down(int crash_shutdown, int secondary)
-{
-	/* Don't risk a hypervisor call if we're crashing */
-	if (firmware_has_feature(FW_FEATURE_SPLPAR) && !crash_shutdown) {
-		unsigned long vpa = __pa(get_lppaca());
-
-		if (unregister_vpa(hard_smp_processor_id(), vpa)) {
-			printk("VPA deregistration of cpu %u (hw_cpu_id %d) "
-					"failed\n", smp_processor_id(),
-					hard_smp_processor_id());
-		}
-	}
-
-	if (ppc64_interrupt_controller == IC_OPEN_PIC)
-		mpic_teardown_this_cpu(secondary);
-	else
-		xics_teardown_cpu(secondary);
-}
-#endif
-
 define_machine(pseries) {
 	.name			= "pSeries",
 	.probe			= pSeries_probe,
@@ -560,7 +573,6 @@ define_machine(pseries) {
 	.system_reset_exception = pSeries_system_reset_exception,
 	.machine_check_exception = pSeries_machine_check_exception,
 #ifdef CONFIG_KEXEC
-	.kexec_cpu_down		= pseries_kexec_cpu_down,
 	.machine_kexec		= default_machine_kexec,
 	.machine_kexec_prepare	= default_machine_kexec_prepare,
 	.machine_crash_shutdown	= default_machine_crash_shutdown,
diff --git a/arch/powerpc/platforms/pseries/smp.c b/arch/powerpc/platforms/pseries/smp.c
index 4ad144df49c..ac61098ff40 100644
--- a/arch/powerpc/platforms/pseries/smp.c
+++ b/arch/powerpc/platforms/pseries/smp.c
@@ -416,27 +416,12 @@ static struct smp_ops_t pSeries_xics_smp_ops = {
 #endif
 
 /* This is called very early */
-void __init smp_init_pSeries(void)
+static void __init smp_init_pseries(void)
 {
 	int i;
 
 	DBG(" -> smp_init_pSeries()\n");
 
-	switch (ppc64_interrupt_controller) {
-#ifdef CONFIG_MPIC
-	case IC_OPEN_PIC:
-		smp_ops = &pSeries_mpic_smp_ops;
-		break;
-#endif
-#ifdef CONFIG_XICS
-	case IC_PPC_XIC:
-		smp_ops = &pSeries_xics_smp_ops;
-		break;
-#endif
-	default:
-		panic("Invalid interrupt controller");
-	}
-
 #ifdef CONFIG_HOTPLUG_CPU
 	smp_ops->cpu_disable = pSeries_cpu_disable;
 	smp_ops->cpu_die = pSeries_cpu_die;
@@ -471,3 +456,18 @@ void __init smp_init_pSeries(void)
 	DBG(" <- smp_init_pSeries()\n");
 }
 
+#ifdef CONFIG_MPIC
+void __init smp_init_pseries_mpic(void)
+{
+	smp_ops = &pSeries_mpic_smp_ops;
+
+	smp_init_pseries();
+}
+#endif
+
+void __init smp_init_pseries_xics(void)
+{
+	smp_ops = &pSeries_xics_smp_ops;
+
+	smp_init_pseries();
+}
diff --git a/arch/powerpc/platforms/pseries/xics.c b/arch/powerpc/platforms/pseries/xics.c
index c7f04420066..716972aa977 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>
@@ -31,9 +35,6 @@
 
 #include "xics.h"
 
-/* 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
 
@@ -64,12 +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
  */
@@ -85,7 +86,7 @@ static int ibm_int_off;
 /* Direct HW low level accessors */
 
 
-static inline int direct_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);
 }
@@ -130,7 +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 inline int lpar_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;
@@ -138,7 +139,7 @@ static inline int lpar_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 inline void lpar_xirr_info_set(int n_cpu, int value)
@@ -175,11 +176,11 @@ static inline void lpar_qirr_info(int n_cpu , u8 value)
 
 
 #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)
@@ -200,7 +201,7 @@ 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;
 }
@@ -213,9 +214,11 @@ static void xics_unmask_irq(unsigned int virq)
 	int call_status;
 	unsigned int server;
 
-	irq = virt_irq_to_real(irq_offset_down(virq));
-	WARN_ON(irq == NO_IRQ);
-	if (irq == XICS_IPI || irq == NO_IRQ)
+	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);
@@ -267,75 +270,57 @@ static void xics_mask_irq(unsigned int virq)
 {
 	unsigned int irq;
 
-	irq = virt_irq_to_real(irq_offset_down(virq));
-	WARN_ON(irq == NO_IRQ);
-	if (irq != NO_IRQ)
-		xics_mask_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_set_irq_revmap(unsigned int virq)
+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));
-}
+	/* 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);
 
-static unsigned int xics_startup(unsigned int virq)
-{
-	xics_set_irq_revmap(virq);
+	/* unmask it */
 	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)
+static void xics_eoi_direct(unsigned int virq)
 {
 	int cpu = smp_processor_id();
+	unsigned int irq = (unsigned int)irq_map[virq].hwirq;
 
 	iosync();
-	direct_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_eoi_lpar(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;
 
 	iosync();
-	lpar_xirr_info_set(cpu, ((0xff << 24) |
-				 (virt_irq_to_real(irq_offset_down(irq)))));
-
+	lpar_xirr_info_set(cpu, (0xff << 24) | irq);
 }
 
-static inline int xics_remap_irq(int vec)
+static inline unsigned int xics_remap_irq(unsigned int vec)
 {
-	int irq;
+	unsigned int irq;
 
 	vec &= 0x00ffffff;
 
 	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);
+	irq = irq_radix_revmap(xics_host, vec);
 	if (likely(irq != NO_IRQ))
-		return irq_offset_up(irq);
+		return irq;
 
 	printk(KERN_ERR "Interrupt %u (real) is invalid,"
 	       " disabling it.\n", vec);
@@ -343,14 +328,14 @@ static inline int xics_remap_irq(int vec)
 	return NO_IRQ;
 }
 
-static int xics_get_irq_direct(struct pt_regs *regs)
+static unsigned int xics_get_irq_direct(struct pt_regs *regs)
 {
 	unsigned int cpu = smp_processor_id();
 
 	return xics_remap_irq(direct_xirr_info_get(cpu));
 }
 
-static int xics_get_irq_lpar(struct pt_regs *regs)
+static unsigned int xics_get_irq_lpar(struct pt_regs *regs)
 {
 	unsigned int cpu = smp_processor_id();
 
@@ -437,8 +422,8 @@ static void xics_set_affinity(unsigned int virq, cpumask_t cpumask)
 	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)
+	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);
@@ -469,6 +454,24 @@ static void xics_set_affinity(unsigned int virq, cpumask_t cpumask)
 	}
 }
 
+void xics_setup_cpu(void)
+{
+	int cpu = smp_processor_id();
+
+	xics_set_cpu_priority(cpu, 0xff);
+
+	/*
+	 * Put the calling processor into the GIQ.  This is really only
+	 * necessary from a secondary thread as the OF start-cpu interface
+	 * performs this function for us on primary threads.
+	 *
+	 * XXX: undo of teardown on kexec needs this too, as may hotplug
+	 */
+	rtas_set_indicator(GLOBAL_INTERRUPT_QUEUE,
+		(1UL << interrupt_server_size) - 1 - default_distrib_server, 1);
+}
+
+
 static struct irq_chip xics_pic_direct = {
 	.typename = " XICS     ",
 	.startup = xics_startup,
@@ -489,90 +492,245 @@ static struct irq_chip xics_pic_lpar = {
 };
 
 
-void xics_setup_cpu(void)
+static int xics_host_match(struct irq_host *h, struct device_node *node)
 {
-	int cpu = smp_processor_id();
+	/* 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");
+}
 
-	xics_set_cpu_priority(cpu, 0xff);
+static int xics_host_map_direct(struct irq_host *h, unsigned int virq,
+				irq_hw_number_t hw, unsigned int flags)
+{
+	unsigned int sense = flags & IRQ_TYPE_SENSE_MASK;
 
-	/*
-	 * Put the calling processor into the GIQ.  This is really only
-	 * necessary from a secondary thread as the OF start-cpu interface
-	 * performs this function for us on primary threads.
-	 *
-	 * XXX: undo of teardown on kexec needs this too, as may hotplug
+	pr_debug("xics: map_direct virq %d, hwirq 0x%lx, flags: 0x%x\n",
+		 virq, hw, flags);
+
+	if (sense && sense != IRQ_TYPE_LEVEL_LOW)
+		printk(KERN_WARNING "xics: using unsupported sense 0x%x"
+		       " for irq %d (h: 0x%lx)\n", flags, 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, unsigned int flags)
+{
+	unsigned int sense = flags & IRQ_TYPE_SENSE_MASK;
+
+	pr_debug("xics: map_lpar virq %d, hwirq 0x%lx, flags: 0x%x\n",
+		 virq, hw, flags);
+
+	if (sense && sense != IRQ_TYPE_LEVEL_LOW)
+		printk(KERN_WARNING "xics: using unsupported sense 0x%x"
+		       " for irq %d (h: 0x%lx)\n", flags, 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
 	 */
-	rtas_set_indicator(GLOBAL_INTERRUPT_QUEUE,
-		(1UL << interrupt_server_size) - 1 - default_distrib_server, 1);
+	*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);
 }
 
-void xics_init_IRQ(void)
+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];
-	struct irq_chip *chip;
 
-	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);
+
+	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);
+}
 
-	np = of_find_node_by_type(np, "PowerPC-External-Interrupt-Presentation");
-	if ((indx < NR_CPUS) && np) goto nextnode;
+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;
@@ -581,102 +739,48 @@ nextnode:
 	}
 	of_node_put(np);
 
-	intr_base = intnodes[0].addr;
-	intr_size = intnodes[0].size;
-
- 	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;
-
-			/* 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 */
+	if (firmware_has_feature(FW_FEATURE_LPAR))
+		ppc_md.get_irq = xics_get_irq_lpar;
+	else
 		ppc_md.get_irq = xics_get_irq_direct;
-		chip = &xics_pic_direct;
-
-	}
-
-	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");
-}
+	xics_setup_8259_cascade();
 
-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;
+	ppc64_boot_msg(0x21, "XICS Done");
 }
-arch_initcall(xics_setup_8259_cascade);
 
 
 #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, 0);
+	BUG_ON(ipi == NO_IRQ);
 
 	/*
 	 * IPIs are marked IRQF_DISABLED as they must run with irqs
 	 * disabled
 	 */
-	set_irq_handler(irq_offset_up(XICS_IPI), handle_percpu_irq);
+	set_irq_handler(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);
+		request_irq(ipi, xics_ipi_action_lpar, IRQF_DISABLED,
+			    "IPI", NULL);
 	else
-		request_irq(irq_offset_up(XICS_IPI), xics_ipi_action_direct,
-			    SA_INTERRUPT, "IPI", NULL);
+		request_irq(ipi, xics_ipi_action_direct, IRQF_DISABLED,
+			    "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();
+	unsigned int ipi;
+	struct irq_desc *desc;
 
- 	xics_set_cpu_priority(cpu, 0);
+	xics_set_cpu_priority(cpu, 0);
 
 	/*
 	 * we need to EOI the IPI if we got here from kexec down IPI
@@ -685,6 +789,11 @@ void xics_teardown_cpu(int secondary)
 	 * should we be flagging idle loop instead?
 	 * or creating some task to be scheduled?
 	 */
+
+	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(XICS_IPI);
 
@@ -694,8 +803,8 @@ void xics_teardown_cpu(int secondary)
 	 */
 	if (secondary)
 		rtas_set_indicator(GLOBAL_INTERRUPT_QUEUE,
-			(1UL << interrupt_server_size) - 1 -
-			default_distrib_server, 0);
+				   (1UL << interrupt_server_size) - 1 -
+				   default_distrib_server, 0);
 }
 
 #ifdef CONFIG_HOTPLUG_CPU
@@ -723,15 +832,15 @@ void xics_migrate_irqs_away(void)
 		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
diff --git a/arch/powerpc/platforms/pseries/xics.h b/arch/powerpc/platforms/pseries/xics.h
index 67dedf3514e..6ee1055b0ff 100644
--- a/arch/powerpc/platforms/pseries/xics.h
+++ b/arch/powerpc/platforms/pseries/xics.h
@@ -31,7 +31,7 @@ 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,
+extern void pseries_8259_cascade(unsigned int irq, struct irq_desc *desc,
 				 struct pt_regs *regs);
 
 #endif /* _POWERPC_KERNEL_XICS_H */