Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
  [SPARC64]: Update defconfig.
  [SPARC64]: Don't double-export synchronize_irq.
  [SPARC64]: Move over to GENERIC_HARDIRQS.
  [SPARC64]: Virtualize IRQ numbers.
  [SPARC64]: Kill ino_bucket->pil
  [SPARC]: Kill __irq_itoa().
  [SPARC64]: bp->pil can never be zero
  [SPARC64]: Send all device interrupts via one PIL.
  [SPARC]: Fix iommu_flush_iotlb end address
  [SPARC]: Mark smp init functions as cpuinit
  [SPARC]: Add missing rw can_lock macros
  [SPARC]: Setup cpu_possible_map
  [SPARC]: Add topology_init()

25 files changed, 515 insertions, 1204 deletions

diff --git a/arch/sparc/kernel/pcic.c b/arch/sparc/kernel/pcic.c
index 42002b742de..bcdf5ad0f03 100644
--- a/arch/sparc/kernel/pcic.c
+++ b/arch/sparc/kernel/pcic.c
@@ -896,13 +896,6 @@ static inline unsigned long get_irqmask(int irq_nr)
 	return 1 << irq_nr;
 }
 
-static inline char *pcic_irq_itoa(unsigned int irq)
-{
-	static char buff[16];
-	sprintf(buff, "%d", irq);
-	return buff;
-}
-
 static void pcic_disable_irq(unsigned int irq_nr)
 {
 	unsigned long mask, flags;
@@ -955,7 +948,6 @@ void __init sun4m_pci_init_IRQ(void)
 	BTFIXUPSET_CALL(clear_clock_irq, pcic_clear_clock_irq, BTFIXUPCALL_NORM);
 	BTFIXUPSET_CALL(clear_profile_irq, pcic_clear_profile_irq, BTFIXUPCALL_NORM);
 	BTFIXUPSET_CALL(load_profile_irq, pcic_load_profile_irq, BTFIXUPCALL_NORM);
-	BTFIXUPSET_CALL(__irq_itoa, pcic_irq_itoa, BTFIXUPCALL_NORM);
 }
 
 int pcibios_assign_resource(struct pci_dev *pdev, int resource)
diff --git a/arch/sparc/kernel/setup.c b/arch/sparc/kernel/setup.c
index 3509e430553..2cbf282f0d0 100644
--- a/arch/sparc/kernel/setup.c
+++ b/arch/sparc/kernel/setup.c
@@ -31,6 +31,7 @@
 #include <linux/console.h>
 #include <linux/spinlock.h>
 #include <linux/root_dev.h>
+#include <linux/cpu.h>
 
 #include <asm/system.h>
 #include <asm/io.h>
@@ -348,6 +349,8 @@ void __init setup_arch(char **cmdline_p)
 	init_mm.context = (unsigned long) NO_CONTEXT;
 	init_task.thread.kregs = &fake_swapper_regs;
 
+	smp_setup_cpu_possible_map();
+
 	paging_init();
 }
 
@@ -389,6 +392,8 @@ console_initcall(set_preferred_console);
 extern char *sparc_cpu_type;
 extern char *sparc_fpu_type;
 
+static int ncpus_probed;
+
 static int show_cpuinfo(struct seq_file *m, void *__unused)
 {
 	seq_printf(m,
@@ -411,7 +416,7 @@ static int show_cpuinfo(struct seq_file *m, void *__unused)
 		   romvec->pv_printrev >> 16,
 		   romvec->pv_printrev & 0xffff,
 		   &cputypval,
-		   num_possible_cpus(),
+		   ncpus_probed,
 		   num_online_cpus()
 #ifndef CONFIG_SMP
 		   , cpu_data(0).udelay_val/(500000/HZ),
@@ -471,3 +476,30 @@ void sun_do_break(void)
 
 int serial_console = -1;
 int stop_a_enabled = 1;
+
+static int __init topology_init(void)
+{
+	int i, ncpus, err;
+
+	/* Count the number of physically present processors in
+	 * the machine, even on uniprocessor, so that /proc/cpuinfo
+	 * output is consistent with 2.4.x
+	 */
+	ncpus = 0;
+	while (!cpu_find_by_instance(ncpus, NULL, NULL))
+		ncpus++;
+	ncpus_probed = ncpus;
+
+	err = 0;
+	for_each_online_cpu(i) {
+		struct cpu *p = kzalloc(sizeof(*p), GFP_KERNEL);
+		if (!p)
+			err = -ENOMEM;
+		else
+			register_cpu(p, i, NULL);
+	}
+
+	return err;
+}
+
+subsys_initcall(topology_init);
diff --git a/arch/sparc/kernel/smp.c b/arch/sparc/kernel/smp.c
index 40b42c88e6a..6135d4faeee 100644
--- a/arch/sparc/kernel/smp.c
+++ b/arch/sparc/kernel/smp.c
@@ -58,7 +58,7 @@ cpumask_t smp_commenced_mask = CPU_MASK_NONE;
 /* Used to make bitops atomic */
 unsigned char bitops_spinlock = 0;
 
-void __init smp_store_cpu_info(int id)
+void __cpuinit smp_store_cpu_info(int id)
 {
 	int cpu_node;
 
@@ -267,22 +267,18 @@ int setup_profiling_timer(unsigned int multiplier)
 void __init smp_prepare_cpus(unsigned int max_cpus)
 {
 	extern void smp4m_boot_cpus(void);
-	int i, cpuid, ncpus, extra;
+	int i, cpuid, extra;
 
 	BUG_ON(sparc_cpu_model != sun4m);
 	printk("Entering SMP Mode...\n");
 
-	ncpus = 1;
 	extra = 0;
 	for (i = 0; !cpu_find_by_instance(i, NULL, &cpuid); i++) {
-		if (cpuid == boot_cpu_id)
-			continue;
-		if (cpuid < NR_CPUS && ncpus++ < max_cpus)
-			cpu_set(cpuid, phys_cpu_present_map);
-		else
+		if (cpuid >= NR_CPUS)
 			extra++;
 	}
-	if (max_cpus >= NR_CPUS && extra)
+	/* i = number of cpus */
+	if (extra && max_cpus > i - extra)
 		printk("Warning: NR_CPUS is too low to start all cpus\n");
 
 	smp_store_cpu_info(boot_cpu_id);
@@ -290,7 +286,25 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 	smp4m_boot_cpus();
 }
 
-void __devinit smp_prepare_boot_cpu(void)
+/* Set this up early so that things like the scheduler can init
+ * properly.  We use the same cpu mask for both the present and
+ * possible cpu map.
+ */
+void __init smp_setup_cpu_possible_map(void)
+{
+	int instance, mid;
+
+	instance = 0;
+	while (!cpu_find_by_instance(instance, NULL, &mid)) {
+		if (mid < NR_CPUS) {
+			cpu_set(mid, phys_cpu_present_map);
+			cpu_set(mid, cpu_present_map);
+		}
+		instance++;
+	}
+}
+
+void __init smp_prepare_boot_cpu(void)
 {
 	int cpuid = hard_smp_processor_id();
 
@@ -306,7 +320,7 @@ void __devinit smp_prepare_boot_cpu(void)
 	cpu_set(cpuid, phys_cpu_present_map);
 }
 
-int __devinit __cpu_up(unsigned int cpu)
+int __cpuinit __cpu_up(unsigned int cpu)
 {
 	extern int smp4m_boot_one_cpu(int);
 	int ret;
diff --git a/arch/sparc/kernel/sparc_ksyms.c b/arch/sparc/kernel/sparc_ksyms.c
index 4b376fae752..fd7deabf998 100644
--- a/arch/sparc/kernel/sparc_ksyms.c
+++ b/arch/sparc/kernel/sparc_ksyms.c
@@ -163,7 +163,6 @@ EXPORT_SYMBOL(BTFIXUP_CALL(__hard_smp_processor_id));
 #endif
 EXPORT_SYMBOL(BTFIXUP_CALL(enable_irq));
 EXPORT_SYMBOL(BTFIXUP_CALL(disable_irq));
-EXPORT_SYMBOL(BTFIXUP_CALL(__irq_itoa));
 EXPORT_SYMBOL(BTFIXUP_CALL(mmu_unlockarea));
 EXPORT_SYMBOL(BTFIXUP_CALL(mmu_lockarea));
 EXPORT_SYMBOL(BTFIXUP_CALL(mmu_get_scsi_sgl));
diff --git a/arch/sparc/kernel/sun4c_irq.c b/arch/sparc/kernel/sun4c_irq.c
index 3d6a99073c4..0f2d8d9cbdb 100644
--- a/arch/sparc/kernel/sun4c_irq.c
+++ b/arch/sparc/kernel/sun4c_irq.c
@@ -198,8 +198,6 @@ static void __init sun4c_init_timers(irqreturn_t (*counter_fn)(int, void *, stru
 static void sun4c_nop(void) {}
 #endif
 
-extern char *sun4m_irq_itoa(unsigned int irq);
-
 void __init sun4c_init_IRQ(void)
 {
 	struct linux_prom_registers int_regs[2];
@@ -238,7 +236,6 @@ void __init sun4c_init_IRQ(void)
 	BTFIXUPSET_CALL(clear_clock_irq, sun4c_clear_clock_irq, BTFIXUPCALL_NORM);
 	BTFIXUPSET_CALL(clear_profile_irq, sun4c_clear_profile_irq, BTFIXUPCALL_NOP);
 	BTFIXUPSET_CALL(load_profile_irq, sun4c_load_profile_irq, BTFIXUPCALL_NOP);
-	BTFIXUPSET_CALL(__irq_itoa, sun4m_irq_itoa, BTFIXUPCALL_NORM);
 	sparc_init_timers = sun4c_init_timers;
 #ifdef CONFIG_SMP
 	BTFIXUPSET_CALL(set_cpu_int, sun4c_nop, BTFIXUPCALL_NOP);
diff --git a/arch/sparc/kernel/sun4d_irq.c b/arch/sparc/kernel/sun4d_irq.c
index ca656d9bd6f..9c30e35c88f 100644
--- a/arch/sparc/kernel/sun4d_irq.c
+++ b/arch/sparc/kernel/sun4d_irq.c
@@ -560,17 +560,6 @@ void __init sun4d_init_sbi_irq(void)
 	}
 }
 
-static char *sun4d_irq_itoa(unsigned int irq)
-{
-	static char buff[16];
-	
-	if (irq < (1 << 5))
-		sprintf(buff, "%d", irq);
-	else
-		sprintf(buff, "%d,%x", sbus_to_pil[(irq >> 2) & 7], irq);
-	return buff;
-}
-
 void __init sun4d_init_IRQ(void)
 {
 	local_irq_disable();
@@ -581,7 +570,6 @@ void __init sun4d_init_IRQ(void)
 	BTFIXUPSET_CALL(clear_clock_irq, sun4d_clear_clock_irq, BTFIXUPCALL_NORM);
 	BTFIXUPSET_CALL(clear_profile_irq, sun4d_clear_profile_irq, BTFIXUPCALL_NORM);
 	BTFIXUPSET_CALL(load_profile_irq, sun4d_load_profile_irq, BTFIXUPCALL_NORM);
-	BTFIXUPSET_CALL(__irq_itoa, sun4d_irq_itoa, BTFIXUPCALL_NORM);
 	sparc_init_timers = sun4d_init_timers;
 #ifdef CONFIG_SMP
 	BTFIXUPSET_CALL(set_cpu_int, sun4d_set_cpu_int, BTFIXUPCALL_NORM);
diff --git a/arch/sparc/kernel/sun4m_irq.c b/arch/sparc/kernel/sun4m_irq.c
index 39d712c3c80..a296c13ac18 100644
--- a/arch/sparc/kernel/sun4m_irq.c
+++ b/arch/sparc/kernel/sun4m_irq.c
@@ -229,13 +229,6 @@ static void sun4m_load_profile_irq(int cpu, unsigned int limit)
 	sun4m_timers->cpu_timers[cpu].l14_timer_limit = limit;
 }
 
-char *sun4m_irq_itoa(unsigned int irq)
-{
-	static char buff[16];
-	sprintf(buff, "%d", irq);
-	return buff;
-}
-
 static void __init sun4m_init_timers(irqreturn_t (*counter_fn)(int, void *, struct pt_regs *))
 {
 	int reg_count, irq, cpu;
@@ -388,7 +381,6 @@ void __init sun4m_init_IRQ(void)
 	BTFIXUPSET_CALL(clear_clock_irq, sun4m_clear_clock_irq, BTFIXUPCALL_NORM);
 	BTFIXUPSET_CALL(clear_profile_irq, sun4m_clear_profile_irq, BTFIXUPCALL_NORM);
 	BTFIXUPSET_CALL(load_profile_irq, sun4m_load_profile_irq, BTFIXUPCALL_NORM);
-	BTFIXUPSET_CALL(__irq_itoa, sun4m_irq_itoa, BTFIXUPCALL_NORM);
 	sparc_init_timers = sun4m_init_timers;
 #ifdef CONFIG_SMP
 	BTFIXUPSET_CALL(set_cpu_int, sun4m_send_ipi, BTFIXUPCALL_NORM);
diff --git a/arch/sparc/kernel/sun4m_smp.c b/arch/sparc/kernel/sun4m_smp.c
index 70b375a4c2c..3b32096134a 100644
--- a/arch/sparc/kernel/sun4m_smp.c
+++ b/arch/sparc/kernel/sun4m_smp.c
@@ -66,7 +66,7 @@ static inline unsigned long swap(volatile unsigned long *ptr, unsigned long val)
 static void smp_setup_percpu_timer(void);
 extern void cpu_probe(void);
 
-void __init smp4m_callin(void)
+void __cpuinit smp4m_callin(void)
 {
 	int cpuid = hard_smp_processor_id();
 
@@ -112,13 +112,8 @@ void __init smp4m_callin(void)
 	local_irq_enable();
 
 	cpu_set(cpuid, cpu_online_map);
-	/* last one in gets all the interrupts (for testing) */
-	set_irq_udt(boot_cpu_id);
 }
 
-extern void init_IRQ(void);
-extern void cpu_panic(void);
-
 /*
  *	Cycle through the processors asking the PROM to start each one.
  */
@@ -134,7 +129,7 @@ void __init smp4m_boot_cpus(void)
 	local_flush_cache_all();
 }
 
-int smp4m_boot_one_cpu(int i)
+int __cpuinit smp4m_boot_one_cpu(int i)
 {
 	extern unsigned long sun4m_cpu_startup;
 	unsigned long *entry = &sun4m_cpu_startup;
diff --git a/arch/sparc/mm/iommu.c b/arch/sparc/mm/iommu.c
index 77840c80478..7215849db39 100644
--- a/arch/sparc/mm/iommu.c
+++ b/arch/sparc/mm/iommu.c
@@ -144,8 +144,9 @@ static void iommu_flush_iotlb(iopte_t *iopte, unsigned int niopte)
 	unsigned long start;
 	unsigned long end;
 
-	start = (unsigned long)iopte & PAGE_MASK;
+	start = (unsigned long)iopte;
 	end = PAGE_ALIGN(start + niopte*sizeof(iopte_t));
+	start &= PAGE_MASK;
 	if (viking_mxcc_present) {
 		while(start < end) {
 			viking_mxcc_flush_page(start);
diff --git a/arch/sparc64/Kconfig b/arch/sparc64/Kconfig
index 43a66f5407f..a7a111db25b 100644
--- a/arch/sparc64/Kconfig
+++ b/arch/sparc64/Kconfig
@@ -87,6 +87,10 @@ config SYSVIPC_COMPAT
 	depends on COMPAT && SYSVIPC
 	default y
 
+config GENERIC_HARDIRQS
+	bool
+	default y
+
 menu "General machine setup"
 
 config SMP
diff --git a/arch/sparc64/defconfig b/arch/sparc64/defconfig
index f09a70b8aab..9da75f89fe2 100644
--- a/arch/sparc64/defconfig
+++ b/arch/sparc64/defconfig
@@ -1,7 +1,7 @@
 #
 # Automatically generated make config: don't edit
-# Linux kernel version: 2.6.17-rc3
-# Fri May 12 12:43:49 2006
+# Linux kernel version: 2.6.17
+# Tue Jun 20 01:26:43 2006
 #
 CONFIG_SPARC=y
 CONFIG_SPARC64=y
@@ -87,6 +87,7 @@ CONFIG_DEFAULT_AS=y
 # CONFIG_DEFAULT_NOOP is not set
 CONFIG_DEFAULT_IOSCHED="anticipatory"
 CONFIG_SYSVIPC_COMPAT=y
+CONFIG_GENERIC_HARDIRQS=y
 
 #
 # General machine setup
@@ -183,6 +184,8 @@ CONFIG_INET_ESP=y
 CONFIG_INET_IPCOMP=y
 CONFIG_INET_XFRM_TUNNEL=y
 CONFIG_INET_TUNNEL=y
+CONFIG_INET_XFRM_MODE_TRANSPORT=y
+CONFIG_INET_XFRM_MODE_TUNNEL=y
 CONFIG_INET_DIAG=y
 CONFIG_INET_TCP_DIAG=y
 CONFIG_TCP_CONG_ADVANCED=y
@@ -198,6 +201,9 @@ CONFIG_TCP_CONG_HSTCP=m
 CONFIG_TCP_CONG_HYBLA=m
 CONFIG_TCP_CONG_VEGAS=m
 CONFIG_TCP_CONG_SCALABLE=m
+CONFIG_TCP_CONG_LP=m
+CONFIG_TCP_CONG_VENO=m
+CONFIG_TCP_CONG_COMPOUND=m
 CONFIG_IPV6=m
 CONFIG_IPV6_PRIVACY=y
 CONFIG_IPV6_ROUTER_PREF=y
@@ -207,7 +213,10 @@ CONFIG_INET6_ESP=m
 CONFIG_INET6_IPCOMP=m
 CONFIG_INET6_XFRM_TUNNEL=m
 CONFIG_INET6_TUNNEL=m
+CONFIG_INET6_XFRM_MODE_TRANSPORT=m
+CONFIG_INET6_XFRM_MODE_TUNNEL=m
 CONFIG_IPV6_TUNNEL=m
+# CONFIG_NETWORK_SECMARK is not set
 # CONFIG_NETFILTER is not set
 
 #
@@ -260,6 +269,7 @@ CONFIG_VLAN_8021Q=m
 # Network testing
 #
 CONFIG_NET_PKTGEN=m
+CONFIG_NET_TCPPROBE=m
 # CONFIG_HAMRADIO is not set
 # CONFIG_IRDA is not set
 # CONFIG_BT is not set
@@ -562,6 +572,7 @@ CONFIG_BNX2=m
 # CONFIG_CHELSIO_T1 is not set
 # CONFIG_IXGB is not set
 # CONFIG_S2IO is not set
+# CONFIG_MYRI10GE is not set
 
 #
 # Token Ring devices
@@ -811,6 +822,7 @@ CONFIG_HWMON=y
 # Multimedia devices
 #
 # CONFIG_VIDEO_DEV is not set
+CONFIG_VIDEO_V4L2=y
 
 #
 # Digital Video Broadcasting Devices
@@ -1135,6 +1147,19 @@ CONFIG_USB_HIDDEV=y
 # CONFIG_RTC_CLASS is not set
 
 #
+# DMA Engine support
+#
+# CONFIG_DMA_ENGINE is not set
+
+#
+# DMA Clients
+#
+
+#
+# DMA Devices
+#
+
+#
 # Misc Linux/SPARC drivers
 #
 CONFIG_SUN_OPENPROMIO=m
diff --git a/arch/sparc64/kernel/devices.c b/arch/sparc64/kernel/devices.c
index 007e8922cd1..0dd95ae50e1 100644
--- a/arch/sparc64/kernel/devices.c
+++ b/arch/sparc64/kernel/devices.c
@@ -157,7 +157,7 @@ unsigned int sun4v_vdev_device_interrupt(unsigned int dev_node)
 		return 0;
 	}
 
-	return sun4v_build_irq(sun4v_vdev_devhandle, irq, 5, 0);
+	return sun4v_build_irq(sun4v_vdev_devhandle, irq);
 }
 
 static const char *cpu_mid_prop(void)
diff --git a/arch/sparc64/kernel/entry.S b/arch/sparc64/kernel/entry.S
index 6d0b3ed77a0..be85ce2a4ad 100644
--- a/arch/sparc64/kernel/entry.S
+++ b/arch/sparc64/kernel/entry.S
@@ -22,6 +22,7 @@
 #include <asm/estate.h>
 #include <asm/auxio.h>
 #include <asm/sfafsr.h>
+#include <asm/pil.h>
 
 #define curptr      g6
 
@@ -431,20 +432,16 @@ do_ivec:
 	membar		#Sync
 
 	sethi		%hi(ivector_table), %g2
-	sllx		%g3, 5, %g3
+	sllx		%g3, 3, %g3
 	or		%g2, %lo(ivector_table), %g2
 	add		%g2, %g3, %g3
-	ldub		[%g3 + 0x04], %g4	/* pil */
-	mov		1, %g2
-	sllx		%g2, %g4, %g2
-	sllx		%g4, 2, %g4
 
 	TRAP_LOAD_IRQ_WORK(%g6, %g1)
 
-	lduw		[%g6 + %g4], %g5	/* g5 = irq_work(cpu, pil) */
+	lduw		[%g6], %g5		/* g5 = irq_work(cpu) */
 	stw		%g5, [%g3 + 0x00]	/* bucket->irq_chain = g5 */
-	stw		%g3, [%g6 + %g4]	/* irq_work(cpu, pil) = bucket */
-	wr		%g2, 0x0, %set_softint
+	stw		%g3, [%g6]		/* irq_work(cpu) = bucket */
+	wr		%g0, 1 << PIL_DEVICE_IRQ, %set_softint
 	retry
 do_ivec_xcall:
 	mov		0x50, %g1
diff --git a/arch/sparc64/kernel/irq.c b/arch/sparc64/kernel/irq.c
index 11e645c9ec5..a8c9dc8d195 100644
--- a/arch/sparc64/kernel/irq.c
+++ b/arch/sparc64/kernel/irq.c
@@ -22,6 +22,7 @@
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
 #include <linux/bootmem.h>
+#include <linux/irq.h>
 
 #include <asm/ptrace.h>
 #include <asm/processor.h>
@@ -42,10 +43,6 @@
 #include <asm/auxio.h>
 #include <asm/head.h>
 
-#ifdef CONFIG_SMP
-static void distribute_irqs(void);
-#endif
-
 /* UPA nodes send interrupt packet to UltraSparc with first data reg
  * value low 5 (7 on Starfire) bits holding the IRQ identifier being
  * delivered.  We must translate this into a non-vector IRQ so we can
@@ -57,10 +54,29 @@ static void distribute_irqs(void);
  * The IVEC handler does not need to act atomically, the PIL dispatch
  * code uses CAS to get an atomic snapshot of the list and clear it
  * at the same time.
+ *
+ * If you make changes to ino_bucket, please update hand coded assembler
+ * of the vectored interrupt trap handler(s) in entry.S and sun4v_ivec.S
  */
+struct ino_bucket {
+	/* Next handler in per-CPU IRQ worklist.  We know that
+	 * bucket pointers have the high 32-bits clear, so to
+	 * save space we only store the bits we need.
+	 */
+/*0x00*/unsigned int irq_chain;
 
+	/* Virtual interrupt number assigned to this INO.  */
+/*0x04*/unsigned int virt_irq;
+};
+
+#define NUM_IVECS	(IMAP_INR + 1)
 struct ino_bucket ivector_table[NUM_IVECS] __attribute__ ((aligned (SMP_CACHE_BYTES)));
 
+#define __irq_ino(irq) \
+        (((struct ino_bucket *)(unsigned long)(irq)) - &ivector_table[0])
+#define __bucket(irq) ((struct ino_bucket *)(unsigned long)(irq))
+#define __irq(bucket) ((unsigned int)(unsigned long)(bucket))
+
 /* This has to be in the main kernel image, it cannot be
  * turned into per-cpu data.  The reason is that the main
  * kernel image is locked into the TLB and this structure
@@ -68,71 +84,82 @@ struct ino_bucket ivector_table[NUM_IVECS] __attribute__ ((aligned (SMP_CACHE_BY
  * access to this structure takes a TLB miss it could cause
  * the 5-level sparc v9 trap stack to overflow.
  */
-struct irq_work_struct {
-	unsigned int	irq_worklists[16];
-};
-struct irq_work_struct __irq_work[NR_CPUS];
-#define irq_work(__cpu, __pil)	&(__irq_work[(__cpu)].irq_worklists[(__pil)])
+#define irq_work(__cpu)	&(trap_block[(__cpu)].irq_worklist)
 
-static struct irqaction *irq_action[NR_IRQS+1];
+static unsigned int virt_to_real_irq_table[NR_IRQS];
+static unsigned char virt_irq_cur = 1;
 
-/* This only synchronizes entities which modify IRQ handler
- * state and some selected user-level spots that want to
- * read things in the table.  IRQ handler processing orders
- * its' accesses such that no locking is needed.
- */
-static DEFINE_SPINLOCK(irq_action_lock);
+static unsigned char virt_irq_alloc(unsigned int real_irq)
+{
+	unsigned char ent;
+
+	BUILD_BUG_ON(NR_IRQS >= 256);
+
+	ent = virt_irq_cur;
+	if (ent >= NR_IRQS) {
+		printk(KERN_ERR "IRQ: Out of virtual IRQs.\n");
+		return 0;
+	}
+
+	virt_irq_cur = ent + 1;
+	virt_to_real_irq_table[ent] = real_irq;
+
+	return ent;
+}
+
+#if 0 /* Currently unused. */
+static unsigned char real_to_virt_irq(unsigned int real_irq)
+{
+	struct ino_bucket *bucket = __bucket(real_irq);
 
-static void register_irq_proc (unsigned int irq);
+	return bucket->virt_irq;
+}
+#endif
+
+static unsigned int virt_to_real_irq(unsigned char virt_irq)
+{
+	return virt_to_real_irq_table[virt_irq];
+}
 
 /*
- * Upper 2b of irqaction->flags holds the ino.
- * irqaction->mask holds the smp affinity information.
+ * /proc/interrupts printing:
  */
-#define put_ino_in_irqaction(action, irq) \
-	action->flags &= 0xffffffffffffUL; \
-	if (__bucket(irq) == &pil0_dummy_bucket) \
-		action->flags |= 0xdeadUL << 48;  \
-	else \
-		action->flags |= __irq_ino(irq) << 48;
-#define get_ino_in_irqaction(action)	(action->flags >> 48)
-
-#define put_smpaff_in_irqaction(action, smpaff)	(action)->mask = (smpaff)
-#define get_smpaff_in_irqaction(action) 	((action)->mask)
 
 int show_interrupts(struct seq_file *p, void *v)
 {
+	int i = *(loff_t *) v, j;
+	struct irqaction * action;
 	unsigned long flags;
-	int i = *(loff_t *) v;
-	struct irqaction *action;
-#ifdef CONFIG_SMP
-	int j;
-#endif
 
-	spin_lock_irqsave(&irq_action_lock, flags);
-	if (i <= NR_IRQS) {
-		if (!(action = *(i + irq_action)))
-			goto out_unlock;
-		seq_printf(p, "%3d: ", i);
+	if (i == 0) {
+		seq_printf(p, "           ");
+		for_each_online_cpu(j)
+			seq_printf(p, "CPU%d       ",j);
+		seq_putc(p, '\n');
+	}
+
+	if (i < NR_IRQS) {
+		spin_lock_irqsave(&irq_desc[i].lock, flags);
+		action = irq_desc[i].action;
+		if (!action)
+			goto skip;
+		seq_printf(p, "%3d: ",i);
 #ifndef CONFIG_SMP
 		seq_printf(p, "%10u ", kstat_irqs(i));
 #else
-		for_each_online_cpu(j) {
-			seq_printf(p, "%10u ",
-				   kstat_cpu(j).irqs[i]);
-		}
+		for_each_online_cpu(j)
+			seq_printf(p, "%10u ", kstat_cpu(j).irqs[i]);
 #endif
-		seq_printf(p, " %s:%lx", action->name,
-			   get_ino_in_irqaction(action));
-		for (action = action->next; action; action = action->next) {
-			seq_printf(p, ", %s:%lx", action->name,
-				   get_ino_in_irqaction(action));
-		}
+		seq_printf(p, " %9s", irq_desc[i].handler->typename);
+		seq_printf(p, "  %s", action->name);
+
+		for (action=action->next; action; action = action->next)
+			seq_printf(p, ", %s", action->name);
+
 		seq_putc(p, '\n');
+skip:
+		spin_unlock_irqrestore(&irq_desc[i].lock, flags);
 	}
-out_unlock:
-	spin_unlock_irqrestore(&irq_action_lock, flags);
-
 	return 0;
 }
 
@@ -173,556 +200,365 @@ static unsigned int sun4u_compute_tid(unsigned long imap, unsigned long cpuid)
 	return tid;
 }
 
-/* Now these are always passed a true fully specified sun4u INO. */
-void enable_irq(unsigned int irq)
-{
-	struct ino_bucket *bucket = __bucket(irq);
-	unsigned long imap, cpuid;
-
-	imap = bucket->imap;
-	if (imap == 0UL)
-		return;
-
-	preempt_disable();
-
-	/* This gets the physical processor ID, even on uniprocessor,
-	 * so we can always program the interrupt target correctly.
-	 */
-	cpuid = real_hard_smp_processor_id();
-
-	if (tlb_type == hypervisor) {
-		unsigned int ino = __irq_ino(irq);
-		int err;
+struct irq_handler_data {
+	unsigned long	iclr;
+	unsigned long	imap;
 
-		err = sun4v_intr_settarget(ino, cpuid);
-		if (err != HV_EOK)
-			printk("sun4v_intr_settarget(%x,%lu): err(%d)\n",
-			       ino, cpuid, err);
-		err = sun4v_intr_setenabled(ino, HV_INTR_ENABLED);
-		if (err != HV_EOK)
-			printk("sun4v_intr_setenabled(%x): err(%d)\n",
-			       ino, err);
-	} else {
-		unsigned int tid = sun4u_compute_tid(imap, cpuid);
-
-		/* NOTE NOTE NOTE, IGN and INO are read-only, IGN is a product
-		 * of this SYSIO's preconfigured IGN in the SYSIO Control
-		 * Register, the hardware just mirrors that value here.
-		 * However for Graphics and UPA Slave devices the full
-		 * IMAP_INR field can be set by the programmer here.
-		 *
-		 * Things like FFB can now be handled via the new IRQ
-		 * mechanism.
-		 */
-		upa_writel(tid | IMAP_VALID, imap);
-	}
-
-	preempt_enable();
-}
+	void		(*pre_handler)(unsigned int, void *, void *);
+	void		*pre_handler_arg1;
+	void		*pre_handler_arg2;
+};
 
-/* This now gets passed true ino's as well. */
-void disable_irq(unsigned int irq)
+static inline struct ino_bucket *virt_irq_to_bucket(unsigned int virt_irq)
 {
-	struct ino_bucket *bucket = __bucket(irq);
-	unsigned long imap;
-
-	imap = bucket->imap;
-	if (imap != 0UL) {
-		if (tlb_type == hypervisor) {
-			unsigned int ino = __irq_ino(irq);
-			int err;
-
-			err = sun4v_intr_setenabled(ino, HV_INTR_DISABLED);
-			if (err != HV_EOK)
-				printk("sun4v_intr_setenabled(%x): "
-				       "err(%d)\n", ino, err);
-		} else {
-			u32 tmp;
-
-			/* NOTE: We do not want to futz with the IRQ clear registers
-			 *       and move the state to IDLE, the SCSI code does call
-			 *       disable_irq() to assure atomicity in the queue cmd
-			 *       SCSI adapter driver code.  Thus we'd lose interrupts.
-			 */
-			tmp = upa_readl(imap);
-			tmp &= ~IMAP_VALID;
-			upa_writel(tmp, imap);
-		}
-	}
-}
+	unsigned int real_irq = virt_to_real_irq(virt_irq);
+	struct ino_bucket *bucket = NULL;
 
-/* The timer is the one "weird" interrupt which is generated by
- * the CPU %tick register and not by some normal vectored interrupt
- * source.  To handle this special case, we use this dummy INO bucket.
- */
-static struct irq_desc pil0_dummy_desc;
-static struct ino_bucket pil0_dummy_bucket = {
-	.irq_info	=	&pil0_dummy_desc,
-};
+	if (likely(real_irq))
+		bucket = __bucket(real_irq);
 
-static void build_irq_error(const char *msg, unsigned int ino, int pil, int inofixup,
-			    unsigned long iclr, unsigned long imap,
-			    struct ino_bucket *bucket)
-{
-	prom_printf("IRQ: INO %04x (%d:%016lx:%016lx) --> "
-		    "(%d:%d:%016lx:%016lx), halting...\n",
-		    ino, bucket->pil, bucket->iclr, bucket->imap,
-		    pil, inofixup, iclr, imap);
-	prom_halt();
+	return bucket;
 }
 
-unsigned int build_irq(int pil, int inofixup, unsigned long iclr, unsigned long imap)
+#ifdef CONFIG_SMP
+static int irq_choose_cpu(unsigned int virt_irq)
 {
-	struct ino_bucket *bucket;
-	int ino;
+	cpumask_t mask = irq_affinity[virt_irq];
+	int cpuid;
 
-	if (pil == 0) {
-		if (iclr != 0UL || imap != 0UL) {
-			prom_printf("Invalid dummy bucket for PIL0 (%lx:%lx)\n",
-				    iclr, imap);
-			prom_halt();
-		}
-		return __irq(&pil0_dummy_bucket);
-	}
+	if (cpus_equal(mask, CPU_MASK_ALL)) {
+		static int irq_rover;
+		static DEFINE_SPINLOCK(irq_rover_lock);
+		unsigned long flags;
 
-	BUG_ON(tlb_type == hypervisor);
+		/* Round-robin distribution... */
+	do_round_robin:
+		spin_lock_irqsave(&irq_rover_lock, flags);
 
-	/* RULE: Both must be specified in all other cases. */
-	if (iclr == 0UL || imap == 0UL) {
-		prom_printf("Invalid build_irq %d %d %016lx %016lx\n",
-			    pil, inofixup, iclr, imap);
-		prom_halt();
-	}
-	
-	ino = (upa_readl(imap) & (IMAP_IGN | IMAP_INO)) + inofixup;
-	if (ino > NUM_IVECS) {
-		prom_printf("Invalid INO %04x (%d:%d:%016lx:%016lx)\n",
-			    ino, pil, inofixup, iclr, imap);
-		prom_halt();
-	}
+		while (!cpu_online(irq_rover)) {
+			if (++irq_rover >= NR_CPUS)
+				irq_rover = 0;
+		}
+		cpuid = irq_rover;
+		do {
+			if (++irq_rover >= NR_CPUS)
+				irq_rover = 0;
+		} while (!cpu_online(irq_rover));
 
-	bucket = &ivector_table[ino];
-	if (bucket->flags & IBF_ACTIVE)
-		build_irq_error("IRQ: Trying to build active INO bucket.\n",
-				ino, pil, inofixup, iclr, imap, bucket);
+		spin_unlock_irqrestore(&irq_rover_lock, flags);
+	} else {
+		cpumask_t tmp;
 
-	if (bucket->irq_info) {
-		if (bucket->imap != imap || bucket->iclr != iclr)
-			build_irq_error("IRQ: Trying to reinit INO bucket.\n",
-					ino, pil, inofixup, iclr, imap, bucket);
+		cpus_and(tmp, cpu_online_map, mask);
 
-		goto out;
-	}
+		if (cpus_empty(tmp))
+			goto do_round_robin;
 
-	bucket->irq_info = kzalloc(sizeof(struct irq_desc), GFP_ATOMIC);
-	if (!bucket->irq_info) {
-		prom_printf("IRQ: Error, kmalloc(irq_desc) failed.\n");
-		prom_halt();
+		cpuid = first_cpu(tmp);
 	}
 
-	/* Ok, looks good, set it up.  Don't touch the irq_chain or
-	 * the pending flag.
-	 */
-	bucket->imap  = imap;
-	bucket->iclr  = iclr;
-	bucket->pil   = pil;
-	bucket->flags = 0;
-
-out:
-	return __irq(bucket);
+	return cpuid;
 }
-
-unsigned int sun4v_build_irq(u32 devhandle, unsigned int devino, int pil, unsigned char flags)
+#else
+static int irq_choose_cpu(unsigned int virt_irq)
 {
-	struct ino_bucket *bucket;
-	unsigned long sysino;
+	return real_hard_smp_processor_id();
+}
+#endif
 
-	sysino = sun4v_devino_to_sysino(devhandle, devino);
+static void sun4u_irq_enable(unsigned int virt_irq)
+{
+	irq_desc_t *desc = irq_desc + virt_irq;
+	struct irq_handler_data *data = desc->handler_data;
 
-	bucket = &ivector_table[sysino];
+	if (likely(data)) {
+		unsigned long cpuid, imap;
+		unsigned int tid;
 
-	/* Catch accidental accesses to these things.  I