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

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6: (23 commits)
  sparc: sunzilog uart order
  [SPARC64]: Detect trap frames in stack backtraces.
  [SPARC64]: %l6 trap return handling no longer necessary.
  [SPARC64]: Use trap type stored in pt_regs to handle syscall restart.
  [SPARC64]: Store magic cookie and trap type in pt_regs.
  [SPARC64]: PROM debug console can be CON_ANYTIME.
  sparc64: cleanup after SunOS/Solaris binary emulation removal
  sparc: cleanup after SunOS binary emulation removal
  [SPARC64]: Add NUMA support.
  [SPARC64]: Allocate TSB node-local.
  [SPARC64]: NUMA device infrastructure.
  [SPARC64]: Kill pci_iommu_table_init() declaration.
  [SPARC64]: Once we have the boot cmdline, call parse_early_param()
  [SPARC64]: Remove unused asm-sparc64/numnodes.h
  [SPARC64]: Decrease SECTION_SIZE_BITS to 30.
  [SPARC64]: Initialize MDESC earlier and use lmb_alloc()
  [SPARC64]: Use lmb_alloc() for PROM device tree.
  [SPARC64]: Call real_setup_per_cpu_areas() earlier and use lmb_alloc().
  [SPARC64]: Fully use LMB information in bootmem_init().
  [SPARC64]: Start using LMB information in bootmem_init().
  ...

37 files changed, 1020 insertions, 450 deletions

diff --git a/arch/sparc/kernel/entry.S b/arch/sparc/kernel/entry.S
index 135644f8add..484c83d23ee 100644
--- a/arch/sparc/kernel/entry.S
+++ b/arch/sparc/kernel/entry.S
@@ -1409,7 +1409,6 @@ syscall_is_too_hard:
 
 	st	%o0, [%sp + STACKFRAME_SZ + PT_I0]
 
-	.globl	ret_sys_call
 ret_sys_call:
 	ld	[%curptr + TI_FLAGS], %l6
 	cmp	%o0, -ERESTART_RESTARTBLOCK
diff --git a/arch/sparc/kernel/signal.c b/arch/sparc/kernel/signal.c
index 1f730619a24..3e849e8e348 100644
--- a/arch/sparc/kernel/signal.c
+++ b/arch/sparc/kernel/signal.c
@@ -105,11 +105,6 @@ static int _sigpause_common(old_sigset_t set)
 	return -ERESTARTNOHAND;
 }
 
-asmlinkage int sys_sigpause(unsigned int set)
-{
-	return _sigpause_common(set);
-}
-
 asmlinkage int sys_sigsuspend(old_sigset_t set)
 {
 	return _sigpause_common(set);
diff --git a/arch/sparc64/Kconfig b/arch/sparc64/Kconfig
index df3eacb5ca1..8acc5cc3862 100644
--- a/arch/sparc64/Kconfig
+++ b/arch/sparc64/Kconfig
@@ -250,6 +250,26 @@ endchoice
 
 endmenu
 
+config NUMA
+	bool "NUMA support"
+
+config NODES_SHIFT
+	int
+	default "4"
+	depends on NEED_MULTIPLE_NODES
+
+# Some NUMA nodes have memory ranges that span
+# other nodes.  Even though a pfn is valid and
+# between a node's start and end pfns, it may not
+# reside on that node.  See memmap_init_zone()
+# for details.
+config NODES_SPAN_OTHER_NODES
+	def_bool y
+	depends on NEED_MULTIPLE_NODES
+
+config ARCH_POPULATES_NODE_MAP
+	def_bool y
+
 config ARCH_SELECT_MEMORY_MODEL
 	def_bool y
 
diff --git a/arch/sparc64/defconfig b/arch/sparc64/defconfig
index e1835868ad3..92f79680f70 100644
--- a/arch/sparc64/defconfig
+++ b/arch/sparc64/defconfig
@@ -1,7 +1,7 @@
 #
 # Automatically generated make config: don't edit
-# Linux kernel version: 2.6.25
-# Sun Apr 20 01:33:21 2008
+# Linux kernel version: 2.6.25-numa
+# Wed Apr 23 04:49:08 2008
 #
 CONFIG_SPARC=y
 CONFIG_SPARC64=y
@@ -152,6 +152,8 @@ CONFIG_GENERIC_CALIBRATE_DELAY=y
 CONFIG_HUGETLB_PAGE_SIZE_4MB=y
 # CONFIG_HUGETLB_PAGE_SIZE_512K is not set
 # CONFIG_HUGETLB_PAGE_SIZE_64K is not set
+# CONFIG_NUMA is not set
+CONFIG_ARCH_POPULATES_NODE_MAP=y
 CONFIG_ARCH_SELECT_MEMORY_MODEL=y
 CONFIG_ARCH_SPARSEMEM_ENABLE=y
 CONFIG_ARCH_SPARSEMEM_DEFAULT=y
@@ -787,7 +789,6 @@ CONFIG_I2C_ALGOBIT=y
 # CONFIG_SENSORS_PCF8574 is not set
 # CONFIG_PCF8575 is not set
 # CONFIG_SENSORS_PCF8591 is not set
-# CONFIG_TPS65010 is not set
 # CONFIG_SENSORS_MAX6875 is not set
 # CONFIG_SENSORS_TSL2550 is not set
 # CONFIG_I2C_DEBUG_CORE is not set
@@ -869,6 +870,7 @@ CONFIG_SSB_POSSIBLE=y
 # Multifunction device drivers
 #
 # CONFIG_MFD_SM501 is not set
+# CONFIG_HTC_PASIC3 is not set
 
 #
 # Multimedia devices
@@ -1219,10 +1221,6 @@ CONFIG_USB_STORAGE=m
 # CONFIG_NEW_LEDS is not set
 # CONFIG_INFINIBAND is not set
 # CONFIG_RTC_CLASS is not set
-
-#
-# Userspace I/O
-#
 # CONFIG_UIO is not set
 
 #
@@ -1399,6 +1397,7 @@ CONFIG_SCHEDSTATS=y
 CONFIG_DEBUG_BUGVERBOSE=y
 # CONFIG_DEBUG_INFO is not set
 # CONFIG_DEBUG_VM is not set
+# CONFIG_DEBUG_WRITECOUNT is not set
 # CONFIG_DEBUG_LIST is not set
 # CONFIG_DEBUG_SG is not set
 # CONFIG_BOOT_PRINTK_DELAY is not set
@@ -1425,53 +1424,82 @@ CONFIG_ASYNC_CORE=m
 CONFIG_ASYNC_MEMCPY=m
 CONFIG_ASYNC_XOR=m
 CONFIG_CRYPTO=y
+
+#
+# Crypto core or helper
+#
 CONFIG_CRYPTO_ALGAPI=y
 CONFIG_CRYPTO_AEAD=y
 CONFIG_CRYPTO_BLKCIPHER=y
-# CONFIG_CRYPTO_SEQIV is not set
 CONFIG_CRYPTO_HASH=y
 CONFIG_CRYPTO_MANAGER=y
+CONFIG_CRYPTO_GF128MUL=m
+CONFIG_CRYPTO_NULL=m
+# CONFIG_CRYPTO_CRYPTD is not set
+CONFIG_CRYPTO_AUTHENC=y
+CONFIG_CRYPTO_TEST=m
+
+#
+# Authenticated Encryption with Associated Data
+#
+# CONFIG_CRYPTO_CCM is not set
+# CONFIG_CRYPTO_GCM is not set
+# CONFIG_CRYPTO_SEQIV is not set
+
+#
+# Block modes
+#
+CONFIG_CRYPTO_CBC=y
+# CONFIG_CRYPTO_CTR is not set
+# CONFIG_CRYPTO_CTS is not set
+CONFIG_CRYPTO_ECB=m
+CONFIG_CRYPTO_LRW=m
+CONFIG_CRYPTO_PCBC=m
+CONFIG_CRYPTO_XTS=m
+
+#
+# Hash modes
+#
 CONFIG_CRYPTO_HMAC=y
 CONFIG_CRYPTO_XCBC=y
-CONFIG_CRYPTO_NULL=m
+
+#
+# Digest
+#
+CONFIG_CRYPTO_CRC32C=m
 CONFIG_CRYPTO_MD4=y
 CONFIG_CRYPTO_MD5=y
+CONFIG_CRYPTO_MICHAEL_MIC=m
 CONFIG_CRYPTO_SHA1=y
 CONFIG_CRYPTO_SHA256=m
 CONFIG_CRYPTO_SHA512=m
-CONFIG_CRYPTO_WP512=m
 CONFIG_CRYPTO_TGR192=m
-CONFIG_CRYPTO_GF128MUL=m
-CONFIG_CRYPTO_ECB=m
-CONFIG_CRYPTO_CBC=y
-CONFIG_CRYPTO_PCBC=m
-CONFIG_CRYPTO_LRW=m
-CONFIG_CRYPTO_XTS=m
-# CONFIG_CRYPTO_CTR is not set
-# CONFIG_CRYPTO_GCM is not set
-# CONFIG_CRYPTO_CCM is not set
-# CONFIG_CRYPTO_CRYPTD is not set
-CONFIG_CRYPTO_DES=y
-CONFIG_CRYPTO_FCRYPT=m
-CONFIG_CRYPTO_BLOWFISH=m
-CONFIG_CRYPTO_TWOFISH=m
-CONFIG_CRYPTO_TWOFISH_COMMON=m
-CONFIG_CRYPTO_SERPENT=m
+CONFIG_CRYPTO_WP512=m
+
+#
+# Ciphers
+#
 CONFIG_CRYPTO_AES=m
+CONFIG_CRYPTO_ANUBIS=m
+CONFIG_CRYPTO_ARC4=m
+CONFIG_CRYPTO_BLOWFISH=m
+CONFIG_CRYPTO_CAMELLIA=m
 CONFIG_CRYPTO_CAST5=m
 CONFIG_CRYPTO_CAST6=m
-CONFIG_CRYPTO_TEA=m
-CONFIG_CRYPTO_ARC4=m
+CONFIG_CRYPTO_DES=y
+CONFIG_CRYPTO_FCRYPT=m
 CONFIG_CRYPTO_KHAZAD=m
-CONFIG_CRYPTO_ANUBIS=m
-CONFIG_CRYPTO_SEED=m
 # CONFIG_CRYPTO_SALSA20 is not set
+CONFIG_CRYPTO_SEED=m
+CONFIG_CRYPTO_SERPENT=m
+CONFIG_CRYPTO_TEA=m
+CONFIG_CRYPTO_TWOFISH=m
+CONFIG_CRYPTO_TWOFISH_COMMON=m
+
+#
+# Compression
+#
 CONFIG_CRYPTO_DEFLATE=y
-CONFIG_CRYPTO_MICHAEL_MIC=m
-CONFIG_CRYPTO_CRC32C=m
-CONFIG_CRYPTO_CAMELLIA=m
-CONFIG_CRYPTO_TEST=m
-CONFIG_CRYPTO_AUTHENC=y
 # CONFIG_CRYPTO_LZO is not set
 CONFIG_CRYPTO_HW=y
 # CONFIG_CRYPTO_DEV_HIFN_795X is not set
@@ -1492,3 +1520,4 @@ CONFIG_PLIST=y
 CONFIG_HAS_IOMEM=y
 CONFIG_HAS_IOPORT=y
 CONFIG_HAS_DMA=y
+CONFIG_HAVE_LMB=y
diff --git a/arch/sparc64/kernel/ebus.c b/arch/sparc64/kernel/ebus.c
index 04ab81cb4f4..bc263227484 100644
--- a/arch/sparc64/kernel/ebus.c
+++ b/arch/sparc64/kernel/ebus.c
@@ -396,6 +396,7 @@ static void __init fill_ebus_device(struct device_node *dp, struct linux_ebus_de
 	sd->op = &dev->ofdev;
 	sd->iommu = dev->bus->ofdev.dev.parent->archdata.iommu;
 	sd->stc = dev->bus->ofdev.dev.parent->archdata.stc;
+	sd->numa_node = dev->bus->ofdev.dev.parent->archdata.numa_node;
 
 	dev->ofdev.node = dp;
 	dev->ofdev.dev.parent = &dev->bus->ofdev.dev;
diff --git a/arch/sparc64/kernel/entry.S b/arch/sparc64/kernel/entry.S
index fb43c76bdc2..fd06e937ae1 100644
--- a/arch/sparc64/kernel/entry.S
+++ b/arch/sparc64/kernel/entry.S
@@ -47,7 +47,7 @@ do_fpdis:
 	ba,pt		%xcc, etrap
 109:	 or		%g7, %lo(109b), %g7
 	add		%g0, %g0, %g0
-	ba,a,pt		%xcc, rtrap_clr_l6
+	ba,a,pt		%xcc, rtrap
 
 1:	TRAP_LOAD_THREAD_REG(%g6, %g1)
 	ldub		[%g6 + TI_FPSAVED], %g5
@@ -226,7 +226,7 @@ fp_other_bounce:
 	call		do_fpother
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	.globl		do_fpother_check_fitos
 	.align		32
@@ -489,7 +489,7 @@ utrap_trap:		/* %g3=handler,%g4=level */
         call		bad_trap
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 invoke_utrap:
 	sllx		%g3, 3, %g3
@@ -607,7 +607,7 @@ __spitfire_cee_trap_continue:
 	call		spitfire_access_error
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	/* This is the trap handler entry point for ECC correctable
 	 * errors.  They are corrected, but we listen for the trap
@@ -686,7 +686,7 @@ __spitfire_data_access_exception_tl1:
 	call		spitfire_data_access_exception_tl1
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 __spitfire_data_access_exception:
 	rdpr		%pstate, %g4
@@ -705,7 +705,7 @@ __spitfire_data_access_exception:
 	call		spitfire_data_access_exception
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	.globl		__spitfire_insn_access_exception
 	.globl		__spitfire_insn_access_exception_tl1
@@ -725,7 +725,7 @@ __spitfire_insn_access_exception_tl1:
 	call		spitfire_insn_access_exception_tl1
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 __spitfire_insn_access_exception:
 	rdpr		%pstate, %g4
@@ -743,7 +743,7 @@ __spitfire_insn_access_exception:
 	call		spitfire_insn_access_exception
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	/* These get patched into the trap table at boot time
 	 * once we know we have a cheetah processor.
@@ -937,7 +937,7 @@ do_dcpe_tl1_fatal:
 	call		cheetah_plus_parity_error
 	 add		%sp, PTREGS_OFF, %o1
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 do_icpe_tl1:
 	rdpr		%tl, %g1		! Save original trap level
@@ -979,7 +979,7 @@ do_icpe_tl1_fatal:
 	call		cheetah_plus_parity_error
 	 add		%sp, PTREGS_OFF, %o1
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 	
 dcpe_icpe_tl1_common:
 	/* Flush D-cache, re-enable D/I caches in DCU and finally
@@ -1281,7 +1281,7 @@ __do_privact:
 	call		do_privact
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	.globl		do_mna
 do_mna:
@@ -1308,7 +1308,7 @@ do_mna:
 	call		mem_address_unaligned
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	.globl		do_lddfmna
 do_lddfmna:
@@ -1326,7 +1326,7 @@ do_lddfmna:
 	call		handle_lddfmna
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	.globl		do_stdfmna
 do_stdfmna:
@@ -1344,7 +1344,7 @@ do_stdfmna:
 	call		handle_stdfmna
 	 add		%sp, PTREGS_OFF, %o0
 	ba,pt		%xcc, rtrap
-	 clr		%l6
+	 nop
 
 	.globl	breakpoint_trap
 breakpoint_trap:
@@ -1424,13 +1424,13 @@ sys32_rt_sigreturn:
 1:		ldx		[%curptr + TI_FLAGS], %l5
 		andcc		%l5, (_TIF_SYSCALL_TRACE|_TIF_SECCOMP|_TIF_SYSCALL_AUDIT), %g0
 		be,pt		%icc, rtrap
-		 clr		%l6
+		 nop
 		add		%sp, PTREGS_OFF, %o0
 		call		syscall_trace
 		 mov		1, %o1
 
 		ba,pt		%xcc, rtrap
-		 clr		%l6
+		 nop
 
 	/* This is how fork() was meant to be done, 8 instruction entry.
 	 *
@@ -1559,7 +1559,7 @@ linux_sparc_syscall32:
 
 	/* Linux native system calls enter here... */
 	.align	32
-	.globl	linux_sparc_syscall, ret_sys_call
+	.globl	linux_sparc_syscall
 linux_sparc_syscall:
 	/* Direct access to user regs, much faster. */
 	cmp		%g1, NR_SYSCALLS			! IEU1	Group
@@ -1605,7 +1605,7 @@ ret_sys_call:
 	bne,pn		%icc, linux_syscall_trace2
 	 add		%l1, 0x4, %l2			! npc = npc+4
 	stx		%l1, [%sp + PTREGS_OFF + PT_V9_TPC]
-	ba,pt		%xcc, rtrap_clr_l6
+	ba,pt		%xcc, rtrap
 	 stx		%l2, [%sp + PTREGS_OFF + PT_V9_TNPC]
 
 1:
@@ -1616,7 +1616,6 @@ ret_sys_call:
 	sub		%g0, %o0, %o0
 	or		%g3, %g2, %g3
 	stx		%o0, [%sp + PTREGS_OFF + PT_V9_I0]
-	mov		1, %l6
 	stx		%g3, [%sp + PTREGS_OFF + PT_V9_TSTATE]
 	bne,pn		%icc, linux_syscall_trace2
 	 add		%l1, 0x4, %l2			! npc = npc+4
diff --git a/arch/sparc64/kernel/entry.h b/arch/sparc64/kernel/entry.h
index 4a91e9c6d31..32fbab62085 100644
--- a/arch/sparc64/kernel/entry.h
+++ b/arch/sparc64/kernel/entry.h
@@ -20,7 +20,6 @@ extern void timer_interrupt(int irq, struct pt_regs *regs);
 
 extern void do_notify_resume(struct pt_regs *regs,
 			     unsigned long orig_i0,
-			     int restart_syscall,
 			     unsigned long thread_info_flags);
 
 extern asmlinkage void syscall_trace(struct pt_regs *regs,
diff --git a/arch/sparc64/kernel/etrap.S b/arch/sparc64/kernel/etrap.S
index 4b2bf9eb447..b49d3b60bc0 100644
--- a/arch/sparc64/kernel/etrap.S
+++ b/arch/sparc64/kernel/etrap.S
@@ -53,7 +53,11 @@ etrap_irq:
 		stx	%g3, [%g2 + STACKFRAME_SZ + PT_V9_TPC]
 		rd	%y, %g3
 		stx	%g1, [%g2 + STACKFRAME_SZ + PT_V9_TNPC]
+		rdpr	%tt, %g1
 		st	%g3, [%g2 + STACKFRAME_SZ + PT_V9_Y]
+		sethi	%hi(PT_REGS_MAGIC), %g3
+		or	%g3, %g1, %g1
+		st	%g1, [%g2 + STACKFRAME_SZ + PT_V9_MAGIC]
 
 		rdpr	%cansave, %g1
 		brnz,pt %g1, etrap_save
diff --git a/arch/sparc64/kernel/iommu.c b/arch/sparc64/kernel/iommu.c
index 756fa24eeef..2a37a6ca2a1 100644
--- a/arch/sparc64/kernel/iommu.c
+++ b/arch/sparc64/kernel/iommu.c
@@ -173,9 +173,11 @@ void iommu_range_free(struct iommu *iommu, dma_addr_t dma_addr, unsigned long np
 }
 
 int iommu_table_init(struct iommu *iommu, int tsbsize,
-		     u32 dma_offset, u32 dma_addr_mask)
+		     u32 dma_offset, u32 dma_addr_mask,
+		     int numa_node)
 {
-	unsigned long i, tsbbase, order, sz, num_tsb_entries;
+	unsigned long i, order, sz, num_tsb_entries;
+	struct page *page;
 
 	num_tsb_entries = tsbsize / sizeof(iopte_t);
 
@@ -188,11 +190,12 @@ int iommu_table_init(struct iommu *iommu, int tsbsize,
 	/* Allocate and initialize the free area map.  */
 	sz = num_tsb_entries / 8;
 	sz = (sz + 7UL) & ~7UL;
-	iommu->arena.map = kzalloc(sz, GFP_KERNEL);
+	iommu->arena.map = kmalloc_node(sz, GFP_KERNEL, numa_node);
 	if (!iommu->arena.map) {
 		printk(KERN_ERR "IOMMU: Error, kmalloc(arena.map) failed.\n");
 		return -ENOMEM;
 	}
+	memset(iommu->arena.map, 0, sz);
 	iommu->arena.limit = num_tsb_entries;
 
 	if (tlb_type != hypervisor)
@@ -201,21 +204,23 @@ int iommu_table_init(struct iommu *iommu, int tsbsize,
 	/* Allocate and initialize the dummy page which we
 	 * set inactive IO PTEs to point to.
 	 */
-	iommu->dummy_page = get_zeroed_page(GFP_KERNEL);
-	if (!iommu->dummy_page) {
+	page = alloc_pages_node(numa_node, GFP_KERNEL, 0);
+	if (!page) {
 		printk(KERN_ERR "IOMMU: Error, gfp(dummy_page) failed.\n");
 		goto out_free_map;
 	}
+	iommu->dummy_page = (unsigned long) page_address(page);
+	memset((void *)iommu->dummy_page, 0, PAGE_SIZE);
 	iommu->dummy_page_pa = (unsigned long) __pa(iommu->dummy_page);
 
 	/* Now allocate and setup the IOMMU page table itself.  */
 	order = get_order(tsbsize);
-	tsbbase = __get_free_pages(GFP_KERNEL, order);
-	if (!tsbbase) {
+	page = alloc_pages_node(numa_node, GFP_KERNEL, order);
+	if (!page) {
 		printk(KERN_ERR "IOMMU: Error, gfp(tsb) failed.\n");
 		goto out_free_dummy_page;
 	}
-	iommu->page_table = (iopte_t *)tsbbase;
+	iommu->page_table = (iopte_t *)page_address(page);
 
 	for (i = 0; i < num_tsb_entries; i++)
 		iopte_make_dummy(iommu, &iommu->page_table[i]);
@@ -276,20 +281,24 @@ static inline void iommu_free_ctx(struct iommu *iommu, int ctx)
 static void *dma_4u_alloc_coherent(struct device *dev, size_t size,
 				   dma_addr_t *dma_addrp, gfp_t gfp)
 {
+	unsigned long flags, order, first_page;
 	struct iommu *iommu;
+	struct page *page;
+	int npages, nid;
 	iopte_t *iopte;
-	unsigned long flags, order, first_page;
 	void *ret;
-	int npages;
 
 	size = IO_PAGE_ALIGN(size);
 	order = get_order(size);
 	if (order >= 10)
 		return NULL;
 
-	first_page = __get_free_pages(gfp, order);
-	if (first_page == 0UL)
+	nid = dev->archdata.numa_node;
+	page = alloc_pages_node(nid, gfp, order);
+	if (unlikely(!page))
 		return NULL;
+
+	first_page = (unsigned long) page_address(page);
 	memset((char *)first_page, 0, PAGE_SIZE << order);
 
 	iommu = dev->archdata.iommu;
diff --git a/arch/sparc64/kernel/isa.c b/arch/sparc64/kernel/isa.c
index b5f7b354084..a2af5ed784c 100644
--- a/arch/sparc64/kernel/isa.c
+++ b/arch/sparc64/kernel/isa.c
@@ -92,6 +92,7 @@ static void __init isa_fill_devices(struct sparc_isa_bridge *isa_br)
 		sd->op = &isa_dev->ofdev;
 		sd->iommu = isa_br->ofdev.dev.parent->archdata.iommu;
 		sd->stc = isa_br->ofdev.dev.parent->archdata.stc;
+		sd->numa_node = isa_br->ofdev.dev.parent->archdata.numa_node;
 
 		isa_dev->ofdev.node = dp;
 		isa_dev->ofdev.dev.parent = &isa_br->ofdev.dev;
diff --git a/arch/sparc64/kernel/mdesc.c b/arch/sparc64/kernel/mdesc.c
index 91008358956..dde52bcf5c6 100644
--- a/arch/sparc64/kernel/mdesc.c
+++ b/arch/sparc64/kernel/mdesc.c
@@ -1,10 +1,10 @@
 /* mdesc.c: Sun4V machine description handling.
  *
- * Copyright (C) 2007 David S. Miller <davem@davemloft.net>
+ * Copyright (C) 2007, 2008 David S. Miller <davem@davemloft.net>
  */
 #include <linux/kernel.h>
 #include <linux/types.h>
-#include <linux/bootmem.h>
+#include <linux/lmb.h>
 #include <linux/log2.h>
 #include <linux/list.h>
 #include <linux/slab.h>
@@ -84,24 +84,28 @@ static void mdesc_handle_init(struct mdesc_handle *hp,
 	hp->handle_size = handle_size;
 }
 
-static struct mdesc_handle * __init mdesc_bootmem_alloc(unsigned int mdesc_size)
+static struct mdesc_handle * __init mdesc_lmb_alloc(unsigned int mdesc_size)
 {
-	struct mdesc_handle *hp;
 	unsigned int handle_size, alloc_size;
+	struct mdesc_handle *hp;
+	unsigned long paddr;
 
 	handle_size = (sizeof(struct mdesc_handle) -
 		       sizeof(struct mdesc_hdr) +
 		       mdesc_size);
 	alloc_size = PAGE_ALIGN(handle_size);
 
-	hp = __alloc_bootmem(alloc_size, PAGE_SIZE, 0UL);
-	if (hp)
-		mdesc_handle_init(hp, handle_size, hp);
+	paddr = lmb_alloc(alloc_size, PAGE_SIZE);
 
+	hp = NULL;
+	if (paddr) {
+		hp = __va(paddr);
+		mdesc_handle_init(hp, handle_size, hp);
+	}
 	return hp;
 }
 
-static void mdesc_bootmem_free(struct mdesc_handle *hp)
+static void mdesc_lmb_free(struct mdesc_handle *hp)
 {
 	unsigned int alloc_size, handle_size = hp->handle_size;
 	unsigned long start, end;
@@ -124,9 +128,9 @@ static void mdesc_bootmem_free(struct mdesc_handle *hp)
 	}
 }
 
-static struct mdesc_mem_ops bootmem_mdesc_ops = {
-	.alloc = mdesc_bootmem_alloc,
-	.free  = mdesc_bootmem_free,
+static struct mdesc_mem_ops lmb_mdesc_ops = {
+	.alloc = mdesc_lmb_alloc,
+	.free  = mdesc_lmb_free,
 };
 
 static struct mdesc_handle *mdesc_kmalloc(unsigned int mdesc_size)
@@ -888,7 +892,7 @@ void __init sun4v_mdesc_init(void)
 
 	printk("MDESC: Size is %lu bytes.\n", len);
 
-	hp = mdesc_alloc(len, &bootmem_mdesc_ops);
+	hp = mdesc_alloc(len, &lmb_mdesc_ops);
 	if (hp == NULL) {
 		prom_printf("MDESC: alloc of %lu bytes failed.\n", len);
 		prom_halt();
diff --git a/arch/sparc64/kernel/of_device.c b/arch/sparc64/kernel/of_device.c
index 0fd9db95b89..9e58e8cba1c 100644
--- a/arch/sparc64/kernel/of_device.c
+++ b/arch/sparc64/kernel/of_device.c
@@ -6,6 +6,7 @@
 #include <linux/mod_devicetable.h>
 #include <linux/slab.h>
 #include <linux/errno.h>
+#include <linux/irq.h>
 #include <linux/of_device.h>
 #include <linux/of_platform.h>
 
@@ -660,6 +661,7 @@ static unsigned int __init build_one_device_irq(struct of_device *op,
 	struct device_node *dp = op->node;
 	struct device_node *pp, *ip;
 	unsigned int orig_irq = irq;
+	int nid;
 
 	if (irq == 0xffffffff)
 		return irq;
@@ -672,7 +674,7 @@ static unsigned int __init build_one_device_irq(struct of_device *op,
 			printk("%s: direct translate %x --> %x\n",
 			       dp->full_name, orig_irq, irq);
 
-		return irq;
+		goto out;
 	}
 
 	/* Something more complicated.  Walk up to the root, applying
@@ -744,6 +746,14 @@ static unsigned int __init build_one_device_irq(struct of_device *op,
 		printk("%s: Apply IRQ trans [%s] %x --> %x\n",
 		       op->node->full_name, ip->full_name, orig_irq, irq);
 
+out:
+	nid = of_node_to_nid(dp);
+	if (nid != -1) {
+		cpumask_t numa_mask = node_to_cpumask(nid);
+
+		irq_set_affinity(irq, numa_mask);
+	}
+
 	return irq;
 }
 
diff --git a/arch/sparc64/kernel/pci.c b/arch/sparc64/kernel/pci.c
index 545356b00e2..49f91276651 100644
--- a/arch/sparc64/kernel/pci.c
+++ b/arch/sparc64/kernel/pci.c
@@ -369,10 +369,12 @@ struct pci_dev *of_create_pci_dev(struct pci_pbm_info *pbm,
 	sd->host_controller = pbm;
 	sd->prom_node = node;
 	sd->op = of_find_device_by_node(node);
+	sd->numa_node = pbm->numa_node;
 
 	sd = &sd->op->dev.archdata;
 	sd->iommu = pbm->iommu;
 	sd->stc = &pbm->stc;
+	sd->numa_node = pbm->numa_node;
 
 	type = of_get_property(node, "device_type", NULL);
 	if (type == NULL)
@@ -1159,6 +1161,16 @@ int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma,
 	return 0;
 }
 
+#ifdef CONFIG_NUMA
+int pcibus_to_node(struct pci_bus *pbus)
+{
+	struct pci_pbm_info *pbm = pbus->sysdata;
+
+	return pbm->numa_node;
+}
+EXPORT_SYMBOL(pcibus_to_node);
+#endif
+
 /* Return the domain nuber for this pci bus */
 
 int pci_domain_nr(struct pci_bus *pbus)
diff --git a/arch/sparc64/kernel/pci_fire.c b/arch/sparc64/kernel/pci_fire.c
index 7571ed56314..d23bb6f53cd 100644
--- a/arch/sparc64/kernel/pci_fire.c
+++ b/arch/sparc64/kernel/pci_fire.c
@@ -71,7 +71,8 @@ static int pci_fire_pbm_iommu_init(struct pci_pbm_info *pbm)
 	 */
 	fire_write(iommu->iommu_flushinv, ~(u64)0);
 
-	err = iommu_table_init(iommu, tsbsize * 8 * 1024, vdma[0], dma_mask);
+	err = iommu_table_init(iommu, tsbsize * 8 * 1024, vdma[0], dma_mask,
+			       pbm->numa_node);
 	if (err)
 		return err;
 
@@ -449,6 +450,8 @@ static int __init pci_fire_pbm_init(struct pci_controller_info *p,
 	pbm->next = pci_pbm_root;
 	pci_pbm_root = pbm;
 
+	pbm->numa_node = -1;
+
 	pbm->scan_bus = pci_fire_scan_bus;
 	pbm->pci_ops = &sun4u_pci_ops;
 	pbm->config_space_reg_bits = 12;
diff --git a/arch/sparc64/kernel/pci_impl.h b/arch/sparc64/kernel/pci_impl.h
index 4a50da13ce4..218bac4ff79 100644
--- a/arch/sparc64/kernel/pci_impl.h
+++ b/arch/sparc64/kernel/pci_impl.h
@@ -148,6 +148,8 @@ struct pci_pbm_info {
 	struct pci_bus			*pci_bus;
 	void (*scan_bus)(struct pci_pbm_info *);
 	struct pci_ops			*pci_ops;
+
+	int				numa_node;
 };
 
 struct pci_controller_info {
@@ -161,8 +163,6 @@ extern struct pci_pbm_info *pci_pbm_root;
 extern int pci_num_pbms;
 
 /* PCI bus scanning and fixup support. */
-extern void pci_iommu_table_init(struct iommu *iommu, int tsbsize,
-				 u32 dma_offset, u32 dma_addr_mask);
 extern void pci_get_pbm_props(struct pci_pbm_info *pbm);
 extern struct pci_bus *pci_scan_one_pbm(struct pci_pbm_info *pbm);
 extern void pci_determine_mem_io_space(struct pci_pbm_info *pbm);
diff --git a/arch/sparc64/kernel/pci_msi.c b/arch/sparc64/kernel/pci_msi.c
index d6d64b44af6..db5e8fd8f67 100644
--- a/arch/sparc64/kernel/pci_msi.c
+++ b/arch/sparc64/kernel/pci_msi.c
@@ -279,11 +279,17 @@ static int bringup_one_msi_queue(struct pci_pbm_info *pbm,
 				 unsigned long devino)
 {
 	int irq = ops->msiq_build_irq(pbm, msiqid, devino);
-	int err;
+	int err, nid;
 
 	if (irq < 0)
 		return irq;
 
+	nid = pbm->numa_node;
+	if (nid != -1) {
+		cpumask_t numa_mask = node_to_cpumask(nid);
+
+		irq_set_affinity(irq, numa_mask);
+	}
 	err = request_irq(irq, sparc64_msiq_interrupt, 0,
 			  "MSIQ",
 			  &pbm->msiq_irq_cookies[msiqid - pbm->msiq_first]);
diff --git a/arch/sparc64/kernel/pci_psycho.c b/arch/sparc64/kernel/pci_psycho.c
index 0bad96e5d18..994dbe0603d 100644
--- a/arch/sparc64/kernel/pci_psycho.c
+++ b/arch/sparc64/kernel/pci_psycho.c
@@ -848,7 +848,8 @@ static int psycho_iommu_init(struct pci_pbm_info *pbm)
 	/* Leave diag mode enabled for full-flushing done
 	 * in pci_iommu.c
 	 */
-	err = iommu_table_init