From 7347aefbccb47ce2d076d8ac7dff2cdd1435974c Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 23 Apr 2008 23:30:59 -0700
Subject: [LMB]: Fix lmb allocation regression.

Changeset d9024df02ffe74d723d97d552f86de3b34beb8cc ("[LMB] Restructure
allocation loops to avoid unsigned underflow") removed the alignment
of the 'size' argument to call lmb_add_region() done by __lmb_alloc_base().

In doing so it reintroduced the bug fixed by changeset
eea89e13a9c61d3928223d2f9bf2295e22e0efb6 ("[LMB]: Fix bug in
__lmb_alloc_base().").

This puts it back.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 lib/lmb.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lib/lmb.c b/lib/lmb.c
index 896e2832099..207147ab25e 100644
--- a/lib/lmb.c
+++ b/lib/lmb.c
@@ -346,7 +346,7 @@ u64 __init __lmb_alloc_base(u64 size, u64 align, u64 max_addr)
 			if (j < 0) {
 				/* this area isn't reserved, take it */
 				if (lmb_add_region(&lmb.reserved, base,
-						   size) < 0)
+						   lmb_align_up(size, align)) < 0)
 					return 0;
 				return base;
 			}
-- 
cgit v1.2.3-18-g5258


From 4e82c9a606da2b1c1c7ea7dfd8052626a4c6d5d6 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 18:00:03 -0800
Subject: [SPARC64]: Move ramdisk discovery code out to seperate function.

And add some comments explaining all of the quirks involved in
the way the bootloader provides this information.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/mm/init.c | 57 +++++++++++++++++++++++++++++---------------------
 1 file changed, 33 insertions(+), 24 deletions(-)

diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index f37078d9640..21e9267608c 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -883,6 +883,37 @@ static void __init trim_pavail(unsigned long *cur_size_p,
 	}
 }
 
+static void __init find_ramdisk(unsigned long phys_base)
+{
+#ifdef CONFIG_BLK_DEV_INITRD
+	if (sparc_ramdisk_image || sparc_ramdisk_image64) {
+		unsigned long ramdisk_image;
+
+		/* Older versions of the bootloader only supported a
+		 * 32-bit physical address for the ramdisk image
+		 * location, stored at sparc_ramdisk_image.  Newer
+		 * SILO versions set sparc_ramdisk_image to zero and
+		 * provide a full 64-bit physical address at
+		 * sparc_ramdisk_image64.
+		 */
+		ramdisk_image = sparc_ramdisk_image;
+		if (!ramdisk_image)
+			ramdisk_image = sparc_ramdisk_image64;
+
+		/* Another bootloader quirk.  The bootloader normalizes
+		 * the physical address to KERNBASE, so we have to
+		 * factor that back out and add in the lowest valid
+		 * physical page address to get the true physical address.
+		 */
+		ramdisk_image -= KERNBASE;
+		ramdisk_image += phys_base;
+
+		initrd_start = ramdisk_image;
+		initrd_end = ramdisk_image + sparc_ramdisk_size;
+	}
+#endif
+}
+
 /* About pages_avail, this is the value we will use to calculate
  * the zholes_size[] argument given to free_area_init_node().  The
  * page allocator uses this to calculate nr_kernel_pages,
@@ -912,30 +943,6 @@ static unsigned long __init bootmem_init(unsigned long *pages_avail,
 		bytes_avail += pavail[i].reg_size;
 	}
 
-	/* Determine the location of the initial ramdisk before trying
-	 * to honor the "mem=xxx" command line argument.  We must know
-	 * where the kernel image and the ramdisk image are so that we
-	 * do not trim those two areas from the physical memory map.
-	 */
-
-#ifdef CONFIG_BLK_DEV_INITRD
-	/* Now have to check initial ramdisk, so that bootmap does not overwrite it */
-	if (sparc_ramdisk_image || sparc_ramdisk_image64) {
-		unsigned long ramdisk_image = sparc_ramdisk_image ?
-			sparc_ramdisk_image : sparc_ramdisk_image64;
-		ramdisk_image -= KERNBASE;
-		initrd_start = ramdisk_image + phys_base;
-		initrd_end = initrd_start + sparc_ramdisk_size;
-		if (initrd_end > end_of_phys_memory) {
-			printk(KERN_CRIT "initrd extends beyond end of memory "
-		                 	 "(0x%016lx > 0x%016lx)\ndisabling initrd\n",
-			       initrd_end, end_of_phys_memory);
-			initrd_start = 0;
-			initrd_end = 0;
-		}
-	}
-#endif	
-
 	if (cmdline_memory_size &&
 	    bytes_avail > cmdline_memory_size)
 		trim_pavail(&bytes_avail,
@@ -1337,6 +1344,8 @@ void __init paging_init(void)
 	for (i = 0; i < pavail_ents; i++)
 		phys_base = min(phys_base, pavail[i].phys_addr);
 
+	find_ramdisk(phys_base);
+
 	set_bit(0, mmu_context_bmap);
 
 	shift = kern_base + PAGE_OFFSET - ((unsigned long)KERNBASE);
-- 
cgit v1.2.3-18-g5258


From 3b2a7e23a9808e349bc5fb32327bacc5e81be79c Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 18:13:20 -0800
Subject: [SPARC64]: Initialize LMB tables.

Call lmb_add() on available regions, and call lmb_reserve()
on the main kernel image and the ramdisk (if any).

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/mm/init.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 21e9267608c..6eb76243fa9 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -24,6 +24,7 @@
 #include <linux/cache.h>
 #include <linux/sort.h>
 #include <linux/percpu.h>
+#include <linux/lmb.h>
 
 #include <asm/head.h>
 #include <asm/system.h>
@@ -910,6 +911,8 @@ static void __init find_ramdisk(unsigned long phys_base)
 
 		initrd_start = ramdisk_image;
 		initrd_end = ramdisk_image + sparc_ramdisk_size;
+
+		lmb_reserve(initrd_start, initrd_end);
 	}
 #endif
 }
@@ -1337,15 +1340,24 @@ void __init paging_init(void)
 		sun4v_ktsb_init();
 	}
 
+	lmb_init();
+
 	/* Find available physical memory... */
 	read_obp_memory("available", &pavail[0], &pavail_ents);
 
 	phys_base = 0xffffffffffffffffUL;
-	for (i = 0; i < pavail_ents; i++)
+	for (i = 0; i < pavail_ents; i++) {
 		phys_base = min(phys_base, pavail[i].phys_addr);
+		lmb_add(pavail[i].phys_addr, pavail[i].reg_size);
+	}
+
+	lmb_reserve(kern_base, kern_size);
 
 	find_ramdisk(phys_base);
 
+	lmb_analyze();
+	lmb_dump_all();
+
 	set_bit(0, mmu_context_bmap);
 
 	shift = kern_base + PAGE_OFFSET - ((unsigned long)KERNBASE);
-- 
cgit v1.2.3-18-g5258


From 25b0c659dfb94f1ddaeda7a8c88ef7043f57f419 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 18:20:14 -0800
Subject: [SPARC64]: Start using LMB information in bootmem_init().

This allows us to kill the incredibly complicated and stupid function
trim_pavail().

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/mm/init.c | 132 +++----------------------------------------------
 1 file changed, 6 insertions(+), 126 deletions(-)

diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 6eb76243fa9..3e6c4ecb8fe 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -775,115 +775,6 @@ static unsigned long __init choose_bootmap_pfn(unsigned long start_pfn,
 	prom_halt();
 }
 
-static void __init trim_pavail(unsigned long *cur_size_p,
-			       unsigned long *end_of_phys_p)
-{
-	unsigned long to_trim = *cur_size_p - cmdline_memory_size;
-	unsigned long avoid_start, avoid_end;
-	int i;
-
-	to_trim = PAGE_ALIGN(to_trim);
-
-	avoid_start = avoid_end = 0;
-#ifdef CONFIG_BLK_DEV_INITRD
-	avoid_start = initrd_start;
-	avoid_end = PAGE_ALIGN(initrd_end);
-#endif
-
-	/* Trim some pavail[] entries in order to satisfy the
-	 * requested "mem=xxx" kernel command line specification.
-	 *
-	 * We must not trim off the kernel image area nor the
-	 * initial ramdisk range (if any).  Also, we must not trim
-	 * any pavail[] entry down to zero in order to preserve
-	 * the invariant that all pavail[] entries have a non-zero
-	 * size which is assumed by all of the code in here.
-	 */
-	for (i = 0; i < pavail_ents; i++) {
-		unsigned long start, end, kern_end;
-		unsigned long trim_low, trim_high, n;
-
-		kern_end = PAGE_ALIGN(kern_base + kern_size);
-
-		trim_low = start = pavail[i].phys_addr;
-		trim_high = end = start + pavail[i].reg_size;
-
-		if (kern_base >= start &&
-		    kern_base < end) {
-			trim_low = kern_base;
-			if (kern_end >= end)
-				continue;
-		}
-		if (kern_end >= start &&
-		    kern_end < end) {
-			trim_high = kern_end;
-		}
-		if (avoid_start &&
-		    avoid_start >= start &&
-		    avoid_start < end) {
-			if (trim_low > avoid_start)
-				trim_low = avoid_start;
-			if (avoid_end >= end)
-				continue;
-		}
-		if (avoid_end &&
-		    avoid_end >= start &&
-		    avoid_end < end) {
-			if (trim_high < avoid_end)
-				trim_high = avoid_end;
-		}
-
-		if (trim_high <= trim_low)
-			continue;
-
-		if (trim_low == start && trim_high == end) {
-			/* Whole chunk is available for trimming.
-			 * Trim all except one page, in order to keep
-			 * entry non-empty.
-			 */
-			n = (end - start) - PAGE_SIZE;
-			if (n > to_trim)
-				n = to_trim;
-
-			if (n) {
-				pavail[i].phys_addr += n;
-				pavail[i].reg_size -= n;
-				to_trim -= n;
-			}
-		} else {
-			n = (trim_low - start);
-			if (n > to_trim)
-				n = to_trim;
-
-			if (n) {
-				pavail[i].phys_addr += n;
-				pavail[i].reg_size -= n;
-				to_trim -= n;
-			}
-			if (to_trim) {
-				n = end - trim_high;
-				if (n > to_trim)
-					n = to_trim;
-				if (n) {
-					pavail[i].reg_size -= n;
-					to_trim -= n;
-				}
-			}
-		}
-
-		if (!to_trim)
-			break;
-	}
-
-	/* Recalculate.  */
-	*cur_size_p = 0UL;
-	for (i = 0; i < pavail_ents; i++) {
-		*end_of_phys_p = pavail[i].phys_addr +
-			pavail[i].reg_size;
-		*cur_size_p += pavail[i].reg_size;
-	}
-}
-
 static void __init find_ramdisk(unsigned long phys_base)
 {
 #ifdef CONFIG_BLK_DEV_INITRD
@@ -935,25 +826,11 @@ static unsigned long __init bootmem_init(unsigned long *pages_avail,
 					 unsigned long phys_base)
 {
 	unsigned long bootmap_size, end_pfn;
-	unsigned long end_of_phys_memory = 0UL;
-	unsigned long bootmap_pfn, bytes_avail, size;
+	unsigned long bootmap_pfn, size;
 	int i;
 
-	bytes_avail = 0UL;
-	for (i = 0; i < pavail_ents; i++) {
-		end_of_phys_memory = pavail[i].phys_addr +
-			pavail[i].reg_size;
-		bytes_avail += pavail[i].reg_size;
-	}
-
-	if (cmdline_memory_size &&
-	    bytes_avail > cmdline_memory_size)
-		trim_pavail(&bytes_avail,
-			    &end_of_phys_memory);
-
-	*pages_avail = bytes_avail >> PAGE_SHIFT;
-
-	end_pfn = end_of_phys_memory >> PAGE_SHIFT;
+	*pages_avail = lmb_phys_mem_size() >> PAGE_SHIFT;
+	end_pfn = lmb_end_of_DRAM() >> PAGE_SHIFT;
 
 	/* Initialize the boot-time allocator. */
 	max_pfn = max_low_pfn = end_pfn;
@@ -1355,6 +1232,9 @@ void __init paging_init(void)
 
 	find_ramdisk(phys_base);
 
+	if (cmdline_memory_size)
+		lmb_enforce_memory_limit(phys_base + cmdline_memory_size);
+
 	lmb_analyze();
 	lmb_dump_all();
 
-- 
cgit v1.2.3-18-g5258


From 9422273ba7d139537720c8c47514925d9a621e0d Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 18:31:41 -0800
Subject: [SPARC64]: Fully use LMB information in bootmem_init().

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/mm/init.c | 100 +++++++++----------------------------------------
 1 file changed, 18 insertions(+), 82 deletions(-)

diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 3e6c4ecb8fe..90e644a0e93 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -722,57 +722,12 @@ out:
 static unsigned long __init choose_bootmap_pfn(unsigned long start_pfn,
 					       unsigned long end_pfn)
 {
-	unsigned long avoid_start, avoid_end, bootmap_size;
-	int i;
+	unsigned long bootmap_size;
 
 	bootmap_size = bootmem_bootmap_pages(end_pfn - start_pfn);
 	bootmap_size <<= PAGE_SHIFT;
 
-	avoid_start = avoid_end = 0;
-#ifdef CONFIG_BLK_DEV_INITRD
-	avoid_start = initrd_start;
-	avoid_end = PAGE_ALIGN(initrd_end);
-#endif
-
-	for (i = 0; i < pavail_ents; i++) {
-		unsigned long start, end;
-
-		start = pavail[i].phys_addr;
-		end = start + pavail[i].reg_size;
-
-		while (start < end) {
-			if (start >= kern_base &&
-			    start < PAGE_ALIGN(kern_base + kern_size)) {
-				start = PAGE_ALIGN(kern_base + kern_size);
-				continue;
-			}
-			if (start >= avoid_start && start < avoid_end) {
-				start = avoid_end;
-				continue;
-			}
-
-			if ((end - start) < bootmap_size)
-				break;
-
-			if (start < kern_base &&
-			    (start + bootmap_size) > kern_base) {
-				start = PAGE_ALIGN(kern_base + kern_size);
-				continue;
-			}
-
-			if (start < avoid_start &&
-			    (start + bootmap_size) > avoid_start) {
-				start = avoid_end;
-				continue;
-			}
-
-			/* OK, it doesn't overlap anything, use it.  */
-			return start >> PAGE_SHIFT;
-		}
-	}
-
-	prom_printf("Cannot find free area for bootmap, aborting.\n");
-	prom_halt();
+	return lmb_alloc(bootmap_size, PAGE_SIZE) >> PAGE_SHIFT;
 }
 
 static void __init find_ramdisk(unsigned long phys_base)
@@ -825,8 +780,7 @@ static void __init find_ramdisk(unsigned long phys_base)
 static unsigned long __init bootmem_init(unsigned long *pages_avail,
 					 unsigned long phys_base)
 {
-	unsigned long bootmap_size, end_pfn;
-	unsigned long bootmap_pfn, size;
+	unsigned long end_pfn;
 	int i;
 
 	*pages_avail = lmb_phys_mem_size() >> PAGE_SHIFT;
@@ -836,49 +790,31 @@ static unsigned long __init bootmem_init(unsigned long *pages_avail,
 	max_pfn = max_low_pfn = end_pfn;
 	min_low_pfn = (phys_base >> PAGE_SHIFT);
 
-	bootmap_pfn = choose_bootmap_pfn(min_low_pfn, end_pfn);
-
-	bootmap_size = init_bootmem_node(NODE_DATA(0), bootmap_pfn,
-					 min_low_pfn, end_pfn);
+	init_bootmem_node(NODE_DATA(0),
+			  choose_bootmap_pfn(min_low_pfn, end_pfn),
+			  min_low_pfn, end_pfn);
 
 	/* Now register the available physical memory with the
 	 * allocator.
 	 */
-	for (i = 0; i < pavail_ents; i++)
-		free_bootmem(pavail[i].phys_addr, pavail[i].reg_size);
+	for (i = 0; i < lmb.memory.cnt; i++)
+		free_bootmem(lmb.memory.region[i].base,
+			     lmb_size_bytes(&lmb.memory, i));
 
-#ifdef CONFIG_BLK_DEV_INITRD
-	if (initrd_start) {
-		size = initrd_end - initrd_start;
-
-		/* Reserve the initrd image area. */
-		reserve_bootmem(initrd_start, size, BOOTMEM_DEFAULT);
+	for (i = 0; i < lmb.reserved.cnt; i++)
+		reserve_bootmem(lmb.reserved.region[i].base,
+				lmb_size_bytes(&lmb.reserved, i),
+				BOOTMEM_DEFAULT);
 
-		initrd_start += PAGE_OFFSET;
-		initrd_end += PAGE_OFFSET;
-	}
-#endif
-	/* Reserve the kernel text/data/bss. */
-	reserve_bootmem(kern_base, kern_size, BOOTMEM_DEFAULT);
 	*pages_avail -= PAGE_ALIGN(kern_size) >> PAGE_SHIFT;
 
-	/* Add back in the initmem pages. */
-	size = ((unsigned long)(__init_end) & PAGE_MASK) -
-		PAGE_ALIGN((unsigned long)__init_begin);
-	*pages_avail += size >> PAGE_SHIFT;
+	for (i = 0; i < lmb.memory.cnt; ++i) {
+		unsigned long start_pfn, end_pfn, pages;
 
-	/* Reserve the bootmem map.   We do not account for it
-	 * in pages_avail because we will release that memory
-	 * in free_all_bootmem.
-	 */
-	size = bootmap_size;
-	reserve_bootmem((bootmap_pfn << PAGE_SHIFT), size, BOOTMEM_DEFAULT);
-
-	for (i = 0; i < pavail_ents; i++) {
-		unsigned long start_pfn, end_pfn;
+		pages = lmb_size_pages(&lmb.memory, i);
+		start_pfn = lmb.memory.region[i].base >> PAGE_SHIFT;
+		end_pfn = start_pfn + pages;
 
-		start_pfn = pavail[i].phys_addr >> PAGE_SHIFT;
-		end_pfn = (start_pfn + (pavail[i].reg_size >> PAGE_SHIFT));
 		memory_present(0, start_pfn, end_pfn);
 	}
 
-- 
cgit v1.2.3-18-g5258


From b97094560b991af5c62391014e72bfa4c3a3701f Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 19:20:45 -0800
Subject: [SPARC64]: Call real_setup_per_cpu_areas() earlier and use
 lmb_alloc().

We have to do it like this before we can move the PROM and MDESC device
tree code over to using lmb_alloc().

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/kernel/smp.c | 11 ++++++++---
 arch/sparc64/mm/init.c    |  8 ++++++--
 2 files changed, 14 insertions(+), 5 deletions(-)

diff --git a/arch/sparc64/kernel/smp.c b/arch/sparc64/kernel/smp.c
index 59f020d69d4..524b8892094 100644
--- a/arch/sparc64/kernel/smp.c
+++ b/arch/sparc64/kernel/smp.c
@@ -20,7 +20,7 @@
 #include <linux/cache.h>
 #include <linux/jiffies.h>
 #include <linux/profile.h>
-#include <linux/bootmem.h>
+#include <linux/lmb.h>
 
 #include <asm/head.h>
 #include <asm/ptrace.h>
@@ -1431,7 +1431,7 @@ EXPORT_SYMBOL(__per_cpu_shift);
 
 void __init real_setup_per_cpu_areas(void)
 {
-	unsigned long goal, size, i;
+	unsigned long paddr, goal, size, i;
 	char *ptr;
 
 	/* Copy section for each CPU (we discard the original) */
@@ -1441,8 +1441,13 @@ void __init real_setup_per_cpu_areas(void)
 	for (size = PAGE_SIZE; size < goal; size <<= 1UL)
 		__per_cpu_shift++;
 
-	ptr = alloc_bootmem_pages(size * NR_CPUS);
+	paddr = lmb_alloc(size * NR_CPUS, PAGE_SIZE);
+	if (!paddr) {
+		prom_printf("Cannot allocate per-cpu memory.\n");
+		prom_halt();
+	}
 
+	ptr = __va(paddr);
 	__per_cpu_base = ptr - __per_cpu_start;
 
 	for (i = 0; i < NR_CPUS; i++, ptr += size)
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 90e644a0e93..658ec462ed4 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -1208,6 +1208,12 @@ void __init paging_init(void)
 	if (tlb_type == hypervisor)
 		sun4v_ktsb_register();
 
+	/* We must setup the per-cpu areas before we pull in the
+	 * PROM and the MDESC.  The code there fills in cpu and
+	 * other information into per-cpu data structures.
+	 */
+	real_setup_per_cpu_areas();
+
 	/* Setup bootmem... */
 	pages_avail = 0;
 	last_valid_pfn = end_pfn = bootmem_init(&pages_avail, phys_base);
@@ -1216,8 +1222,6 @@ void __init paging_init(void)
 
 	kernel_physical_mapping_init();
 
-	real_setup_per_cpu_areas();
-
 	prom_build_devicetree();
 
 	if (tlb_type == hypervisor)
-- 
cgit v1.2.3-18-g5258


From ad072004ca35a9918964ca7aee2bf00d79c8657f Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 19:21:51 -0800
Subject: [SPARC64]: Use lmb_alloc() for PROM device tree.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/kernel/prom.c | 14 +++++++++-----
 arch/sparc64/mm/init.c     |  4 ++--
 2 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/arch/sparc64/kernel/prom.c b/arch/sparc64/kernel/prom.c
index 68964ddcde1..ed03a18d3b3 100644
--- a/arch/sparc64/kernel/prom.c
+++ b/arch/sparc64/kernel/prom.c
@@ -19,8 +19,8 @@
 #include <linux/types.h>
 #include <linux/string.h>
 #include <linux/mm.h>
-#include <linux/bootmem.h>
 #include <linux/module.h>
+#include <linux/lmb.h>
 
 #include <asm/prom.h>
 #include <asm/of_device.h>
@@ -122,16 +122,20 @@ int of_find_in_proplist(const char *list, const char *match, int len)
 }
 EXPORT_SYMBOL(of_find_in_proplist);
 
-static unsigned int prom_early_allocated;
+static unsigned int prom_early_allocated __initdata;
 
 static void * __init prom_early_alloc(unsigned long size)
 {
+	unsigned long paddr = lmb_alloc(size, SMP_CACHE_BYTES);
 	void *ret;
 
-	ret = __alloc_bootmem(size, SMP_CACHE_BYTES, 0UL);
-	if (ret != NULL)
-		memset(ret, 0, size);
+	if (!paddr) {
+		prom_printf("prom_early_alloc(%lu) failed\n");
+		prom_halt();
+	}
 
+	ret = __va(paddr);
+	memset(ret, 0, size);
 	prom_early_allocated += size;
 
 	return ret;
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 658ec462ed4..0abefc8ca40 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -1214,6 +1214,8 @@ void __init paging_init(void)
 	 */
 	real_setup_per_cpu_areas();
 
+	prom_build_devicetree();
+
 	/* Setup bootmem... */
 	pages_avail = 0;
 	last_valid_pfn = end_pfn = bootmem_init(&pages_avail, phys_base);
@@ -1222,8 +1224,6 @@ void __init paging_init(void)
 
 	kernel_physical_mapping_init();
 
-	prom_build_devicetree();
-
 	if (tlb_type == hypervisor)
 		sun4v_mdesc_init();
 
-- 
cgit v1.2.3-18-g5258


From 4a28333984be123d9c063df23175c48749c4b4a0 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 19:22:23 -0800
Subject: [SPARC64]: Initialize MDESC earlier and use lmb_alloc()

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/kernel/mdesc.c | 28 ++++++++++++++++------------
 arch/sparc64/mm/init.c      |  6 +++---
 2 files changed, 19 insertions(+), 15 deletions(-)

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/mm/init.c b/arch/sparc64/mm/init.c
index 0abefc8ca40..8e0e8678712 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -1216,6 +1216,9 @@ void __init paging_init(void)
 
 	prom_build_devicetree();
 
+	if (tlb_type == hypervisor)
+		sun4v_mdesc_init();
+
 	/* Setup bootmem... */
 	pages_avail = 0;
 	last_valid_pfn = end_pfn = bootmem_init(&pages_avail, phys_base);
@@ -1224,9 +1227,6 @@ void __init paging_init(void)
 
 	kernel_physical_mapping_init();
 
-	if (tlb_type == hypervisor)
-		sun4v_mdesc_init();
-
 	{
 		unsigned long zones_size[MAX_NR_ZONES];
 		unsigned long zholes_size[MAX_NR_ZONES];
-- 
cgit v1.2.3-18-g5258


From 86cfc751cb4d5da20904ae26c67fb6fd2a68a9c4 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 22:01:57 -0800
Subject: [SPARC64]: Decrease SECTION_SIZE_BITS to 30.

We'll need this to handle NUMA properly on some systems.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 include/asm-sparc64/sparsemem.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/asm-sparc64/sparsemem.h b/include/asm-sparc64/sparsemem.h
index 77bcd2bfa53..b99d4e4b6d2 100644
--- a/include/asm-sparc64/sparsemem.h
+++ b/include/asm-sparc64/sparsemem.h
@@ -3,7 +3,7 @@
 
 #ifdef __KERNEL__
 
-#define SECTION_SIZE_BITS       31
+#define SECTION_SIZE_BITS       30
 #define MAX_PHYSADDR_BITS       42
 #define MAX_PHYSMEM_BITS        42
 
-- 
cgit v1.2.3-18-g5258


From 30957a86c2442fb6d73aaa9a3153f023fe0a682b Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 22:07:34 -0800
Subject: [SPARC64]: Remove unused asm-sparc64/numnodes.h

This is handled with config options now.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 include/asm-sparc64/numnodes.h | 6 ------
 1 file changed, 6 deletions(-)
 delete mode 100644 include/asm-sparc64/numnodes.h

diff --git a/include/asm-sparc64/numnodes.h b/include/asm-sparc64/numnodes.h
deleted file mode 100644
index 017e7e74f5e..00000000000
--- a/include/asm-sparc64/numnodes.h
+++ /dev/null
@@ -1,6 +0,0 @@
-#ifndef _SPARC64_NUMNODES_H
-#define _SPARC64_NUMNODES_H
-
-#define NODES_SHIFT	0
-
-#endif /* !(_SPARC64_NUMNODES_H) */
-- 
cgit v1.2.3-18-g5258


From ce3b1d47a8605905b666649ea309b6471a75b4ed Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 19 Mar 2008 03:54:09 -0700
Subject: [SPARC64]: Once we have the boot cmdline, call parse_early_param()

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/kernel/setup.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/sparc64/kernel/setup.c b/arch/sparc64/kernel/setup.c
index 6acb4c51cfe..46bcf41097a 100644
--- a/arch/sparc64/kernel/setup.c
+++ b/arch/sparc64/kernel/setup.c
@@ -281,6 +281,7 @@ void __init setup_arch(char **cmdline_p)
 	/* Initialize PROM console and command line. */
 	*cmdline_p = prom_getbootargs();
 	strcpy(boot_command_line, *cmdline_p);
+	parse_early_param();
 
 	boot_flags_init(*cmdline_p);
 	register_console(&prom_early_console);
-- 
cgit v1.2.3-18-g5258


From 0c49a573ea93f777fd27f26b7853e7bf5a20e1a3 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 13 Feb 2008 22:15:46 -0800
Subject: [SPARC64]: Kill pci_iommu_table_init() declaration.

No longer exists.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/kernel/pci_impl.h | 2 --
 1 file changed, 2 deletions(-)

diff --git a/arch/sparc64/kernel/pci_impl.h b/arch/sparc64/kernel/pci_impl.h
index 4a50da13ce4..f79c923fdd8 100644
--- a/arch/sparc64/kernel/pci_impl.h
+++ b/arch/sparc64/kernel/pci_impl.h
@@ -161,8 +161,6 @@ 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);
-- 
cgit v1.2.3-18-g5258


From c1b1a5f1f1b2612b69b67381b223bce9f8ec4da5 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 19 Mar 2008 04:52:48 -0700
Subject: [SPARC64]: NUMA device infrastructure.

Record and propagate NUMA information for devices.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/kernel/ebus.c       |  1 +
 arch/sparc64/kernel/iommu.c      | 33 +++++++++++++++++++++------------
 arch/sparc64/kernel/isa.c        |  1 +
 arch/sparc64/kernel/of_device.c  | 12 +++++++++++-
 arch/sparc64/kernel/pci.c        | 12 ++++++++++++
 arch/sparc64/kernel/pci_fire.c   |  5 ++++-
 arch/sparc64/kernel/pci_impl.h   |  2 ++
 arch/sparc64/kernel/pci_msi.c    |  8 +++++++-
 arch/sparc64/kernel/pci_psycho.c |  5 ++++-
 arch/sparc64/kernel/pci_sabre.c  |  4 +++-
 arch/sparc64/kernel/pci_schizo.c |  5 ++++-
 arch/sparc64/kernel/pci_sun4v.c  | 13 ++++++++++---
 arch/sparc64/kernel/sbus.c       |  3 ++-
 include/asm-sparc/device.h       |  2 ++
 include/asm-sparc/prom.h         |  5 +++++
 include/asm-sparc64/iommu.h      |  3 ++-
 16 files changed, 91 insertions(+), 23 deletions(-)

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/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/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 f79c923fdd8..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 {
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(iommu, IO_TSB_SIZE, 0xc0000000, 0xffffffff);
+	err = iommu_table_init(iommu, IO_TSB_SIZE, 0xc0000000, 0xffffffff,
+			       pbm->numa_node);
 	if (err)
 		return err;
 
@@ -979,6 +980,8 @@ static void __init psycho_pbm_init(struct pci_controller_info *p,
 	pbm->next = pci_pbm_root;
 	pci_pbm_root = pbm;
 
+	pbm->numa_node = -1;
+
 	pbm->scan_bus = psycho_scan_bus;
 	pbm->pci_ops = &sun4u_pci_ops;
 	pbm->config_space_reg_bits = 8;
diff --git a/arch/sparc64/kernel/pci_sabre.c b/arch/sparc64/kernel/pci_sabre.c
index 1c5f5fa2339..4c34195baf3 100644
--- a/arch/sparc64/kernel/pci_sabre.c
+++ b/arch/sparc64/kernel/pci_sabre.c
@@ -704,7 +704,7 @@ static int sabre_iommu_init(struct pci_pbm_info *pbm,
 	 * in pci_iommu.c
 	 */
 	err = iommu_table_init(iommu, tsbsize * 1024 * 8,
-			       dvma_offset, dma_mask);
+			       dvma_offset, dma_mask, pbm->numa_node);
 	if (err)
 		return err;
 
@@ -737,6 +737,8 @@ static void __init sabre_pbm_init(struct pci_controller_info *p,
 	pbm->name = dp->full_name;
 	printk("%s: SABRE PCI Bus Module\n", pbm->name);
 
+	pbm->numa_node = -1;
+
 	pbm->scan_bus = sabre_scan_bus;
 	pbm->pci_ops = &sun4u_pci_ops;
 	pbm->config_space_reg_bits = 8;
diff --git a/arch/sparc64/kernel/pci_schizo.c b/arch/sparc64/kernel/pci_schizo.c
index e3060936232..615edd9c8e2 100644
--- a/arch/sparc64/kernel/pci_schizo.c
+++ b/arch/sparc64/kernel/pci_schizo.c
@@ -1220,7 +1220,8 @@ static int schizo_pbm_iommu_init(struct pci_pbm_info *pbm)
 	/* Leave diag mode enabled for full-flushing done
 	 * in pci_iommu.c
 	 */
-	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;
 
@@ -1379,6 +1380,8 @@ static int __init schizo_pbm_init(struct pci_controller_info *p,
 	pbm->next = pci_pbm_root;
 	pci_pbm_root = pbm;
 
+	pbm->numa_node = -1;
+
 	pbm->scan_bus = schizo_scan_bus;
 	pbm->pci_ops = &sun4u_pci_ops;
 	pbm->config_space_reg_bits = 8;
diff --git a/arch/sparc64/kernel/pci_sun4v.c b/arch/sparc64/kernel/pci_sun4v.c
index 01839706bd5..e2bb9790039 100644
--- a/arch/sparc64/kernel/pci_sun4v.c
+++ b/arch/sparc64/kernel/pci_sun4v.c
@@ -127,10 +127,12 @@ static inline long iommu_batch_end(void)
 static void *dma_4v_alloc_coherent(struct device *dev, size_t size,
 				   dma_addr_t *dma_addrp, gfp_t gfp)
 {
-	struct iommu *iommu;
 	unsigned long flags, order, first_page, npages, n;
+	struct iommu *iommu;
+	struct page *page;
 	void *ret;
 	long entry;
+	int nid;
 
 	size = IO_PAGE_ALIGN(size);
 	order = get_order(size);
@@ -139,10 +141,12 @@ static void *dma_4v_alloc_coherent(struct device *dev, size_t size,
 
 	npages = size >> IO_PAGE_SHIFT;
 
-	first_page = __get_free_pages(gfp, order);
-	if (unlikely(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;
@@ -899,6 +903,8 @@ static void __init pci_sun4v_pbm_init(struct pci_controller_info *p,
 	pbm->next = pci_pbm_root;
 	pci_pbm_root = pbm;
 
+	pbm->numa_node = of_node_to_nid(dp);
+
 	pbm->scan_bus = pci_sun4v_scan_bus;
 	pbm->pci_ops = &sun4v_pci_ops;
 	pbm->config_space_reg_bits = 12;
@@ -913,6 +919,7 @@ static void __init pci_sun4v_pbm_init(struct pci_controller_info *p,
 	pbm->name = dp->full_name;
 
 	printk("%s: SUN4V PCI Bus Module\n", pbm->name);
+	printk("%s: On NUMA node %d\n", pbm->name, pbm->numa_node);
 
 	pci_determine_mem_io_space(pbm);
 
diff --git a/arch/sparc64/kernel/sbus.c b/arch/sparc64/kernel/sbus.c
index d1fb13ba02b..fa2827c4a3a 100644
--- a/arch/sparc64/kernel/sbus.c
+++ b/arch/sparc64/kernel/sbus.c
@@ -544,6 +544,7 @@ static void __init sbus_iommu_init(int __node, struct sbus_bus *sbus)
 
 	sbus->ofdev.dev.archdata.iommu = iommu;
 	sbus->ofdev.dev.archdata.stc = strbuf;
+	sbus->ofdev.dev.archdata.numa_node = -1;
 
 	reg_base = regs + SYSIO_IOMMUREG_BASE;
 	iommu->iommu_control = reg_base + IOMMU_CONTROL;
@@ -575,7 +576,7 @@ static void __init sbus_iommu_init(int __node, struct sbus_bus *sbus)
 	       sbus->portid, regs);
 
 	/* Setup for TSB_SIZE=7, TBW_SIZE=0, MMU_DE=1, MMU_EN=1 */
-	if (iommu_table_init(iommu, IO_TSB_SIZE, MAP_BASE, 0xffffffff))
+	if (iommu_table_init(iommu, IO_TSB_SIZE, MAP_BASE, 0xffffffff, -1))
 		goto fatal_memory_error;
 
 	control = upa_readq(iommu->iommu_control);
diff --git a/include/asm-sparc/device.h b/include/asm-sparc/device.h
index 680e51d8737..19790eb99cc 100644
--- a/include/asm-sparc/device.h
+++ b/include/asm-sparc/device.h
@@ -16,6 +16,8 @@ struct dev_archdata {
 
 	struct device_node	*prom_node;
 	struct of_device	*op;
+
+	int			numa_node;
 };
 
 #endif /* _ASM_SPARC_DEVICE_H */
diff --git a/include/asm-sparc/prom.h b/include/asm-sparc/prom.h
index df5dc442248..fd55522481c 100644
--- a/include/asm-sparc/prom.h
+++ b/include/asm-sparc/prom.h
@@ -77,6 +77,11 @@ extern int of_getintprop_default(struct device_node *np,
 				 const char *name,
 				 int def);
 extern int of_find_in_proplist(const char *list, const char *match, int len);
+#ifdef CONFIG_NUMA
+extern int of_node_to_nid(struct device_node *dp);
+#else
+#define of_node_to_nid(dp)	(-1)
+#endif
 
 extern void prom_build_devicetree(void);
 
diff --git a/include/asm-sparc64/iommu.h b/include/asm-sparc64/iommu.h
index 46325ddee23..d7b9afcba08 100644
--- a/include/asm-sparc64/iommu.h
+++ b/include/asm-sparc64/iommu.h
@@ -56,6 +56,7 @@ struct strbuf {
 };
 
 extern 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);
 
 #endif /* !(_SPARC64_IOMMU_H) */
-- 
cgit v1.2.3-18-g5258


From 1f261ef53ba06658dfeb5a9c3007d0ad1b85cadf Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 19 Mar 2008 04:53:58 -0700
Subject: [SPARC64]: Allocate TSB node-local.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 arch/sparc64/mm/tsb.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/sparc64/mm/tsb.c b/arch/sparc64/mm/tsb.c
index a3e6e4b635b..fe70c8a557b 100644
--- a/arch/sparc64/mm/tsb.c
+++ b/arch/sparc64/mm/tsb.c
@@ -321,7 +321,8 @@ retry_tsb_alloc:
 	if (new_size > (PAGE_SIZE * 2))
 		gfp_flags = __GFP_NOWARN | __GFP_NORETRY;
 
-	new_tsb = kmem_cache_alloc(tsb_caches[new_cache_index], gfp_flags);
+	new_tsb = kmem_cache_alloc_node(tsb_caches[new_cache_index],
+					gfp_flags, numa_node_id());
 	if (unlikely(!new_tsb)) {
 		/* Not being able to fork due to a high-order TSB
 		 * allocation failure is very bad behavior.  Just back
-- 
cgit v1.2.3-18-g5258


From 919ee677b656c52c5f86d3d916786891220d5452 Mon Sep 17 00:00:00 2001
From: "David S. Miller" <davem@davemloft.net>
Date: Wed, 23 Apr 2008 05:40:25 -0700
Subject: [SPARC64]: Add NUMA support.

Currently there is only code to parse NUMA attributes on
sun4v/niagara systems, but later on we will add such parsing
for older systems.

Signed-off-by: David S. Miller <davem@davemloft.net>
---
 Makefile                       |   2 +-
 arch/sparc64/Kconfig           |  20 ++
 arch/sparc64/defconfig         |  99 +++--
 arch/sparc64/kernel/sysfs.c    |  12 +
 arch/sparc64/mm/init.c         | 796 +++++++++++++++++++++++++++++++++++------
 include/asm-sparc64/mmzone.h   |  17 +
 include/asm-sparc64/topology.h |  73 +++-
 7 files changed, 881 insertions(+), 138 deletions(-)
 create mode 100644 include/asm-sparc64/mmzone.h

diff --git a/Makefile b/Makefile
index 3dbc826bb8e..d35c5246fce 100644
--- a/Makefile
+++ b/Makefile
@@ -1,7 +1,7 @@
 VERSION = 2
 PATCHLEVEL = 6
 SUBLEVEL = 25
-EXTRAVERSION =
+EXTRAVERSION = -numa
 NAME = Funky Weasel is Jiggy wit it
 
 # *DOCUMENTATION*
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/sysfs.c b/arch/sparc64/kernel/sysfs.c
index 52816c7be0b..e885034a6b7 100644
--- a/arch/sparc64/kernel/sysfs.c
+++ b/arch/sparc64/kernel/sysfs.c
@@ -273,10 +273,22 @@ static void __init check_mmu_stats(void)
 		mmu_stats_supported = 1;
 }
 
+static void register_nodes(void)
+{
+#ifdef CONFIG_NUMA
+	int i;
+
+	for (i = 0; i < MAX_NUMNODES; i++)
+		register_one_node(i);
+#endif
+}
+
 static int __init topology_init(void)
 {
 	int cpu;
 
+	register_nodes();
+
 	check_mmu_stats();
 
 	register_cpu_notifier(&sysfs_cpu_nb);
diff --git a/arch/sparc64/mm/init.c b/arch/sparc64/mm/init.c
index 8e0e8678712..177d8aaeec4 100644
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -25,6 +25,7 @@
 #include <linux/sort.h>
 #include <linux/percpu.h>
 #include <linux/lmb.h>
+#include <linux/mmzone.h>
 
 #include <asm/head.h>
 #include <asm/system.h>
@@ -73,9 +74,7 @@ extern struct tsb swapper_4m_tsb[KERNEL_TSB4M_NENTRIES];
 #define MAX_BANKS	32
 
 static struct linux_prom64_registers pavail[MAX_BANKS] __initdata;
-static struct linux_prom64_registers pavail_rescan[MAX_BANKS] __initdata;
 static int pavail_ents __initdata;
-static int pavail_rescan_ents __initdata;
 
 static int cmp_p64(const void *a, const void *b)
 {
@@ -716,19 +715,28 @@ out:
 		smp_new_mmu_context_version();
 }
 
-/* Find a free area for the bootmem map, avoiding the kernel image
- * and the initial ramdisk.
- */
-static unsigned long __init choose_bootmap_pfn(unsigned long start_pfn,
-					       unsigned long end_pfn)
+static int numa_enabled = 1;
+static int numa_debug;
+
+static int __init early_numa(char *p)
 {
-	unsigned long bootmap_size;
+	if (!p)
+		return 0;
+
+	if (strstr(p, "off"))
+		numa_enabled = 0;
 
-	bootmap_size = bootmem_bootmap_pages(end_pfn - start_pfn);
-	bootmap_size <<= PAGE_SHIFT;
+	if (strstr(p, "debug"))
+		numa_debug = 1;
 
-	return lmb_alloc(bootmap_size, PAGE_SIZE) >> PAGE_SHIFT;
+	return 0;
 }
+early_param("numa", early_numa);
+
+#define numadbg(f, a...) \
+do {	if (numa_debug) \
+		printk(KERN_INFO f, ## a); \
+} while (0)
 
 static void __init find_ramdisk(unsigned long phys_base)
 {
@@ -755,6 +763,9 @@ static void __init find_ramdisk(unsigned long phys_base)
 		ramdisk_image -= KERNBASE;
 		ramdisk_image += phys_base;
 
+		numadbg("Found ramdisk at physical address 0x%lx, size %u\n",
+			ramdisk_image, sparc_ramdisk_size);
+
 		initrd_start = ramdisk_image;
 		initrd_end = ramdisk_image + sparc_ramdisk_size;
 
@@ -763,60 +774,625 @@ static void __init find_ramdisk(unsigned long phys_base)
 #endif
 }
 
-/* About pages_avail, this is the value we will use to calculate
- * the zholes_size[] argument given to free_area_init_node().  The
- * page allocator uses this to calculate nr_kernel_pages,
- * nr_all_pages and zone->present_pages.  On NUMA it is used
- * to calculate zone->min_unmapped_pages and zone->min_slab_pages.
- *
- * So this number should really be set to what the page allocator
- * actually ends up with.  This means:
- * 1) It should include bootmem map pages, we'll release those.
- * 2) It should not include the kernel image, except for the
- *    __init sections which we will also release.
- * 3) It should include the initrd image, since we'll release
- *    that too.
+struct node_mem_mask {
+	unsigned long mask;
+	unsigned long val;
+	unsigned long bootmem_paddr;
+};
+static struct node_mem_mask node_masks[MAX_NUMNODES];
+static int num_node_masks;
+
+int numa_cpu_lookup_table[NR_CPUS];
+cpumask_t numa_cpumask_lookup_table[MAX_NUMNODES];
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+static bootmem_data_t plat_node_bdata[MAX_NUMNODES];
+
+struct mdesc_mblock {
+	u64	base;
+	u64	size;
+	u64	offset; /* RA-to-PA */
+};
+static struct mdesc_mblock *mblocks;
+static int num_mblocks;
+
+static unsigned long ra_to_pa(unsigned long addr)
+{
+	int i;
+
+	for (i = 0; i < num_mblocks; i++) {
+		struct mdesc_mblock *m = &mblocks[i];
+
+		if (addr >= m->base &&
+		    addr < (m->base + m->size)) {
+			addr += m->offset;
+			break;
+		}
+	}
+	return addr;
+}
+
+static int find_node(unsigned long addr)
+{
+	int i;
+
+	addr = ra_to_pa(addr);
+	for (i = 0; i < num_node_masks; i++) {
+		struct node_mem_mask *p = &node_masks[i];
+
+		if ((addr & p->mask) == p->val)
+			return i;
+	}
+	return -1;
+}
+
+static unsigned long nid_range(unsigned long start, unsigned long end,
+			       int *nid)
+{
+	*nid = find_node(start);
+	start += PAGE_SIZE;
+	while (start < end) {
+		int n = find_node(start);
+
+		if (n != *nid)
+			break;
+		start += PAGE_SIZE;
+	}
+
+	return start;
+}
+#else
+static unsigned long nid_range(unsigned long start, unsigned long end,
+			       int *nid)
+{
+	*nid = 0;
+	return end;
+}
+#endif
+
+/* This must be invoked after performing all of the necessary
+ * add_active_range() calls for 'nid'.  We need to be able to get
+ * correct data from get_pfn_range_for_nid().
  */
-static unsigned long __init bootmem_init(unsigned long *pages_avail,
-					 unsigned long phys_base)
+static void __init allocate_node_data(int nid)
+{
+	unsigned long paddr, num_pages, start_pfn, end_pfn;
+	struct pglist_data *p;
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+	paddr = lmb_alloc_nid(sizeof(struct pglist_data),
+			      SMP_CACHE_BYTES, nid, nid_range);
+	if (!paddr) {
+		prom_printf("Cannot allocate pglist_data for nid[%d]\n", nid);
+		prom_halt();
+	}
+	NODE_DATA(nid) = __va(paddr);
+	memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
+
+	NODE_DATA(nid)->bdata = &plat_node_bdata[nid];
+#endif
+
+	p = NODE_DATA(nid);
+
+	get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
+	p->node_start_pfn = start_pfn;
+	p->node_spanned_pages = end_pfn - start_pfn;
+
+	if (p->node_spanned_pages) {
+		num_pages = bootmem_bootmap_pages(p->node_spanned_pages);
+
+		paddr = lmb_alloc_nid(num_pages << PAGE_SHIFT, PAGE_SIZE, nid,
+				      nid_range);
+		if (!paddr) {
+			prom_printf("Cannot allocate bootmap for nid[%d]\n",
+				  nid);
+			prom_halt();
+		}
+		node_masks[nid].bootmem_paddr = paddr;
+	}
+}
+
+static void init_node_masks_nonnuma(void)
 {
-	unsigned long end_pfn;
 	int i;
 
-	*pages_avail = lmb_phys_mem_size() >> PAGE_SHIFT;
-	end_pfn = lmb_end_of_DRAM() >> PAGE_SHIFT;
+	numadbg("Initializing tables for non-numa.\n");
 
-	/* Initialize the boot-time allocator. */
-	max_pfn = max_low_pfn = end_pfn;
-	min_low_pfn = (phys_base >> PAGE_SHIFT);
+	node_masks[0].mask = node_masks[0].val = 0;
+	num_node_masks = 1;
 
-	init_bootmem_node(NODE_DATA(0),
-			  choose_bootmap_pfn(min_low_pfn, end_pfn),
-			  min_low_pfn, end_pfn);
+	for (i = 0; i < NR_CPUS; i++)
+		numa_cpu_lookup_table[i] = 0;
 
-	/* Now register the available physical memory with the
-	 * allocator.
-	 */
-	for (i = 0; i < lmb.memory.cnt; i++)
-		free_bootmem(lmb.memory.region[i].base,
-			     lmb_size_bytes(&lmb.memory, i));
+	numa_cpumask_lookup_table[0] = CPU_MASK_ALL;
+}
+
+#ifdef CONFIG_NEED_MULTIPLE_NODES
+struct pglist_data *node_data[MAX_NUMNODES];
+
+EXPORT_SYMBOL(numa_cpu_lookup_table);
+EXPORT_SYMBOL(numa_cpumask_lookup_table);
+EXPORT_SYMBOL(node_data);
+
+struct mdesc_mlgroup {
+	u64	node;
+	u64	latency;
+	u64	match;
+	u64	mask;
+};
+static struct mdesc_mlgroup *mlgroups;
+static int num_mlgroups;
+
+static int scan_pio_for_cfg_handle(struct mdesc_handle *md, u64 pio,
+				   u32 cfg_handle)
+{
+	u64 arc;
+
+	mdesc_for_each_arc(arc, md, pio, MDESC_ARC_TYPE_FWD) {
+		u64 target = mdesc_arc_target(md, arc);
+		const u64 *val;
+
+		val = mdesc_get_property(md, target,
+					 "cfg-handle", NULL);
+		if (val && *val == cfg_handle)
+			return 0;
+	}
+	return -ENODEV;
+}
+
+static int scan_arcs_for_cfg_handle(struct mdesc_handle *md, u64 grp,
+				    u32 cfg_handle)
+{
+	u64 arc, candidate, best_latency = ~(u64)0;
+
+	candidate = MDESC_NODE_NULL;
+	mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_FWD) {
+		u64 target = mdesc_arc_target(md, arc);
+		const char *name = mdesc_node_name(md, target);
+		const u64 *val;
+
+		if (strcmp(name, "pio-latency-group"))
+			continue;
+
+		val = mdesc_get_property(md, target, "latency", NULL);
+		if (!val)
+			continue;
+
+		if (*val < best_latency) {
+			candidate = target;
+			best_latency = *val;
+		}
+	}
+
+	if (candidate == MDESC_NODE_NULL)
+		return -ENODEV;
+
+	return scan_pio_for_cfg_handle(md, candidate, cfg_handle);
+}
+
+int of_node_to_nid(struct device_node *dp)
+{
+	const struct linux_prom64_registers *regs;
+	struct mdesc_handle *md;
+	u32 cfg_handle;
+	int count, nid;
+	u64 grp;
+
+	if (!mlgroups)
+		return -1;
+
+	regs = of_get_property(dp, "reg", NULL);
+	if (!regs)
+		return -1;
+
+	cfg_handle = (regs->phys_addr >> 32UL) & 0x0fffffff;
+
+	md = mdesc_grab();
+
+	count = 0;
+	nid = -1;
+	mdesc_for_each_node_by_name(md, grp, "group") {
+		if (!scan_arcs_for_cfg_handle(md, grp, cfg_handle)) {
+			nid = count;
+			break;
+		}
+		count++;
+	}
+
+	mdesc_release(md);
+
+	return nid;
+}
+
+static void add_node_ranges(void)
+{
+	int i;
+
+	for (i = 0; i < lmb.memory.cnt; i++) {
+		unsigned long size = lmb_size_bytes(&lmb.memory, i);
+		unsigned long start, end;
+
+		start = lmb.memory.region[i].base;
+		end = start + size;
+		while (start < end) {
+			unsigned long this_end;
+			int nid;
+
+			this_end = nid_range(start, end, &nid);
+
+			numadbg("Adding active range nid[%d] "
+				"start[%lx] end[%lx]\n",
+				nid, start, this_end);
+
+			add_active_range(nid,
+					 start >> PAGE_SHIFT,
+					 this_end >> PAGE_SHIFT);
+
+			start = this_end;
+		}
+	}
+}
 
-	for (i = 0; i < lmb.reserved.cnt; i++)
-		reserve_bootmem(lmb.reserved.region[i].base,
-				lmb_size_bytes(&lmb.reserved, i),
-				BOOTMEM_DEFAULT);
+static int __init grab_mlgroups(struct mdesc_handle *md)
+{
+	unsigned long paddr;
+	int count = 0;
+	u64 node;
+
+	mdesc_for_each_node_by_name(md, node, "memory-latency-group")
+		count++;
+	if (!count)
+		return -ENOENT;
+
+	paddr = lmb_alloc(count * sizeof(struct mdesc_mlgroup),
+			  SMP_CACHE_BYTES);
+	if (!paddr)
+		return -ENOMEM;
+
+	mlgroups = __va(paddr);
+	num_mlgroups = count;
+
+	count = 0;
+	mdesc_for_each_node_by_name(md, node, "memory-latency-group") {
+		struct mdesc_mlgroup *m = &mlgroups[count++];
+		const u64 *val;
+
+		m->node = node;
+
+		val = mdesc_get_property(md, node, "latency", NULL);
+		m->latency = *val;
+		val = mdesc_get_property(md, node, "address-match", NULL);
+		m->match = *val;
+		val = mdesc_get_property(md, node, "address-mask", NULL);
+		m->mask = *val;
+
+		numadbg("MLGROUP[%d]: node[%lx] latency[%lx] "
+			"match[%lx] mask[%lx]\n",
+			count - 1, m->node, m->latency, m->match, m->mask);
+	}
 
-	*pages_avail -= PAGE_ALIGN(kern_size) >> PAGE_SHIFT;
+	return 0;
+}
 
-	for (i = 0; i < lmb.memory.cnt; ++i) {
-		unsigned long start_pfn, end_pfn, pages;
+static int __init grab_mblocks(struct mdesc_handle *md)
+{
+	unsigned long paddr;
+	int count = 0;
+	u64 node;
+
+	mdesc_for_each_node_by_name(md, node, "mblock")
+		count++;
+	if (!count)
+		return -ENOENT;
+
+	paddr = lmb_alloc(count * sizeof(struct mdesc_mblock),
+			  SMP_CACHE_BYTES);
+	if (!paddr)
+		return -ENOMEM;
+
+	mblocks = __va(paddr);
+	num_mblocks = count;
+
+	count = 0;
+	mdesc_for_each_node_by_name(md, node, "mblock") {
+		struct mdesc_mblock *m = &mblocks[count++];
+		const u64 *val;
+
+		val = mdesc_get_property(md, node, "base", NULL);
+		m->base = *val;
+		val = mdesc_get_property(md, node, "size", NULL);
+		m->size = *val;
+		val = mdesc_get_property(md, node,
+					 "address-congruence-offset", NULL);
+		m->offset = *val;
+
+		numadbg("MBLOCK[%d]: base[%lx] size[%lx] offset[%lx]\n",
+			count - 1, m->base, m->size, m->offset);
+	}
+
+	return 0;
+}
+
+static void __init numa_parse_mdesc_group_cpus(struct mdesc_handle *md,
+					       u64 grp, cpumask_t *mask)
+{
+	u64 arc;
+
+	cpus_clear(*mask);
+
+	mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_BACK) {
+		u64 target = mdesc_arc_target(md, arc);
+		const char *name = mdesc_node_name(md, target);
+		const u64 *id;
+
+		if (strcmp(name, "cpu"))
+			continue;
+		id = mdesc_get_property(md, target, "id", NULL);
+		if (*id < NR_CPUS)
+			cpu_set(*id, *mask);
+	}
+}
+
+static struct mdesc_mlgroup * __init find_mlgroup(u64 node)
+{
+	int i;
+
+	for (i = 0; i < num_mlgroups; i++) {
+		struct mdesc_mlgroup *m = &mlgroups[i];
+		if (m->node == node)
+			return m;
+	}
+	return NULL;
+}
+
+static int __init numa_attach_mlgroup(struct mdesc_handle *md, u64 grp,
+				      int index)
+{
+	struct mdesc_mlgroup *candidate = NULL;
+	u64 arc, best_latency = ~(u64)0;
+	struct node_mem_mask *n;
+
+	mdesc_for_each_arc(arc, md, grp, MDESC_ARC_TYPE_FWD) {
+		u64 target = mdesc_arc_target(md, arc);
+		struct mdesc_mlgroup *m = find_mlgroup(target);
+		if (!m)
+			continue;
+		if (m->latency < best_latency) {
+			candidate = m;
+			best_latency = m->latency;
+		}
+	}
+	if (!candidate)
+		return -ENOENT;
+
+	if (num_node_masks != index) {
+		printk(KERN_ERR "Inconsistent NUMA state, "
+		       "index[%d] != num_node_masks[%d]\n",
+		       index, num_node_masks);
+		return -EINVAL;
+	}
+
+	n = &node_masks[num_node_masks++];
+
+	n->mask = candidate->mask;
+	n->val = candidate->match;
+
+	numadbg("NUMA NODE[%d]: mask[%lx] val[%lx] (latency[%lx])\n",
+		index, n->mask, n->val, candidate->latency);
+
+	return 0;
+}
+
+static int __init numa_parse_mdesc_group(struct mdesc_handle *md, u64 grp,
+					 int index)
+{
+	cpumask_t mask;
+	int cpu;
+
+	numa_parse_mdesc_group_cpus(md, grp, &mask);
+
+	for_each_cpu_mask(cpu, mask)
+		numa_cpu_lookup_table[cpu] = index;
+	numa_cpumask_lookup_table[index] = mask;
+
+	if (numa_debug) {
+		printk(KERN_INFO "NUMA GROUP[%d]: cpus [ ", index);
+		for_each_cpu_mask(cpu, mask)
+			printk("%d ", cpu);
+		printk("]\n");
+	}
+
+	return numa_attach_mlgroup(md, grp, index);
+}
+
+static int __init numa_parse_mdesc(void)
+{
+	struct mdesc_handle *md = mdesc_grab();
+	int i, err, count;
+	u64 node;
+
+	node = mdesc_node_by_name(md, MDESC_NODE_NULL, "latency-groups");
+	if (node == MDESC_NODE_NULL) {
+		mdesc_release(md);
+		return -ENOENT;
+	}
+
+	err = grab_mblocks(md);
+	if (err < 0)
+		goto out;
+
+	err = grab_mlgroups(md);
+	if (err < 0)
+		goto out;
+
+	count = 0;
+	mdesc_for_each_node_by_name(md, node, "group") {
+		err = numa_parse_mdesc_group(md, node, count);
+		if (err < 0)
+			break;
+		count++;
+	}
+
+	add_node_ranges();
+
+	for (i = 0; i < num_node_masks; i++) {
+		allocate_node_data(i);
+		node_set_online(i);
+	}
+
+	err = 0;
+out:
+	mdesc_release(md);
+	return err;
+}
+
+static int __init numa_parse_sun4u(void)
+{
+	return -1;
+}
+
+static int __init bootmem_init_numa(void)
+{
+	int err = -1;
+
+	numadbg("bootmem_init_numa()\n");
+
+	if (numa_enabled) {
+		if (tlb_type == hypervisor)
+			err = numa_parse_mdesc();
+		else
+			err = numa_parse_sun4u();
+	}
+	return err;
+}
+
+#else
+
+static int bootmem_init_numa(void)
+{
+	return -1;
+}
+
+#endif
+
+static void __init bootmem_init_nonnuma(void)
+{
+	unsigned long top_of_ram = lmb_end_of_DRAM();
+	unsigned long total_ram = lmb_phys_mem_size();
+	unsigned int i;
+
+	numadbg("bootmem_init_nonnuma()\n");
+
+	printk(KERN_INFO "Top of RAM: 0x%lx, Total RAM: 0x%lx\n",
+	       top_of_ram, total_ram);
+	printk(KERN_INFO "Memory hole size: %ldMB\n",
+	       (top_of_ram - total_ram) >> 20);
+
+	init_node_masks_nonnuma();
+
+	for (i = 0; i < lmb.memory.cnt; i++) {
+		unsigned long size = lmb_size_bytes(&lmb.memory, i);
+		unsigned long start_pfn, end_pfn;
+
+		if (!size)
+			continue;
 
-		pages = lmb_size_pages(&lmb.memory, i);
 		start_pfn = lmb.memory.region[i].base >> PAGE_SHIFT;
-		end_pfn = start_pfn + pages;
+		end_pfn = start_pfn + lmb_size_pages(&lmb.memory, i);
+		add_active_range(0, start_pfn, end_pfn);
+	}
 
-		memory_present(0, start_pfn, end_pfn);
+	allocate_node_data(0);
+
+	node_set_online(0);
+}
+
+static void __init reserve_range_in_node(int nid, unsigned long start,
+					 unsigned long end)
+{
+	numadbg("    reserve_range_in_node(nid[%d],start[%lx],end[%lx]\n",
+		nid, start, end);
+	while (start < end) {
+		unsigned long this_end;
+		int n;
+
+		this_end = nid_range(start, end, &n);
+		if (n == nid) {
+			numadbg("      MATCH reserving range [%lx:%lx]\n",
+				start, this_end);
+			reserve_bootmem_node(NODE_DATA(nid), start,
+					     (this_end - start), BOOTMEM_DEFAULT);
+		} else
+			numadbg("      NO MATCH, advancing start to %lx\n",
+				this_end