[PATCH] x86: trim mtrr don't close gap for resource allocation.

[PATCH] x86: trim mtrr don't close gap for resource allocation.

[Yinghai Lu]

[PATCH] x86: trim mtrr don't close gap for resource allocation.

for
	http://bugzilla.kernel.org/show_bug.cgi?id=10232

use update_memory_range instead of add_memory_range directly
to avoid closing that gap.

Signed-off-by: Yinghai Lu <yhlu.kenrel@gmail.com>

Index: linux-2.6/arch/x86/kernel/cpu/mtrr/main.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/cpu/mtrr/main.c
+++ linux-2.6/arch/x86/kernel/cpu/mtrr/main.c
@@ -711,7 +711,8 @@ int __init mtrr_trim_uncached_memory(uns
 		trim_size = end_pfn;
 		trim_size <<= PAGE_SHIFT;
 		trim_size -= trim_start;
-		add_memory_region(trim_start, trim_size, E820_RESERVED);
+		update_memory_range(trim_start, trim_size, E820_RAM,
+					E820_RESERVED);
 		update_e820();
 		return 1;
 	}
Index: linux-2.6/arch/x86/kernel/e820_32.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/e820_32.c
+++ linux-2.6/arch/x86/kernel/e820_32.c
@@ -749,6 +749,32 @@ static int __init parse_memmap(char *arg
 	return 0;
 }
 early_param("memmap", parse_memmap);
+void __init update_memory_range(u64 start, u64 size, unsigned old_type,
+				unsigned new_type)
+{
+	int i;
+
+	BUG_ON(old_type == new_type);
+
+	for (i = 0; i < e820.nr_map; i++) {
+		struct e820entry *ei = &e820.map[i];
+		u64 final_start, final_end;
+		if (ei->type != old_type)
+			continue;
+		/* totally covered? */
+		if (ei->addr >= start && ei->size <= size) {
+			ei->type = new_type;
+			continue;
+		}
+		/* partially covered */
+		final_start = max(start, ei->addr);
+		final_end = min(start + size, ei->addr + ei->size);
+		if (final_start >= final_end)
+			continue;
+		add_memory_region(final_start, final_end - final_start,
+					 new_type);
+	}
+}
 void __init update_e820(void)
 {
 	u8 nr_map;
Index: linux-2.6/arch/x86/kernel/e820_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/e820_64.c
+++ linux-2.6/arch/x86/kernel/e820_64.c
@@ -735,6 +735,33 @@ void __init finish_e820_parsing(void)
 	}
 }
 
+void __init update_memory_range(u64 start, u64 size, unsigned old_type,
+				unsigned new_type)
+{
+	int i;
+
+	BUG_ON(old_type == new_type);
+
+	for (i = 0; i < e820.nr_map; i++) {
+		struct e820entry *ei = &e820.map[i];
+		u64 final_start, final_end;
+		if (ei->type != old_type)
+			continue;
+		/* totally covered? */
+		if (ei->addr >= start && ei->size <= size) {
+			ei->type = new_type;
+			continue;
+		}
+		/* partially covered */
+		final_start = max(start, ei->addr);
+		final_end = min(start + size, ei->addr + ei->size);
+		if (final_start >= final_end)
+			continue;
+		add_memory_region(final_start, final_end - final_start,
+					 new_type);
+	}
+}
+
 void __init update_e820(void)
 {
 	u8 nr_map;
Index: linux-2.6/include/asm-x86/e820_32.h
===================================================================
--- linux-2.6.orig/include/asm-x86/e820_32.h
+++ linux-2.6/include/asm-x86/e820_32.h
@@ -28,6 +28,8 @@ extern void find_max_pfn(void);
 extern void register_bootmem_low_pages(unsigned long max_low_pfn);
 extern void add_memory_region(unsigned long long start,
 			      unsigned long long size, int type);
+extern void update_memory_range(u64 start, u64 size, unsigned old_type,
+				unsigned new_type);
 extern void e820_register_memory(void);
 extern void limit_regions(unsigned long long size);
 extern void print_memory_map(char *who);
Index: linux-2.6/include/asm-x86/e820_64.h
===================================================================
--- linux-2.6.orig/include/asm-x86/e820_64.h
+++ linux-2.6/include/asm-x86/e820_64.h
@@ -18,6 +18,8 @@ extern unsigned long find_e820_area(unsi
 				    unsigned long size, unsigned long align);
 extern void add_memory_region(unsigned long start, unsigned long size, 
 			      int type);
+extern void update_memory_range(u64 start, u64 size, unsigned old_type,
+				unsigned new_type);
 extern void setup_memory_region(void);
 extern void contig_e820_setup(void); 
 extern unsigned long e820_end_of_ram(void);

Re: [PATCH] x86: trim mtrr don't close gap for resource allocation.

[Ingo Molnar]

* Yinghai Lu <yhlu.kernel.send@gmail.com> wrote:

> [PATCH] x86: trim mtrr don't close gap for resource allocation.
> 
> for
> 	http://bugzilla.kernel.org/show_bug.cgi?id=10232
> 
> use update_memory_range instead of add_memory_range directly to avoid 
> closing that gap.

thanks Yinghai, applied, and queued up for 2.6.25.

	Ingo

[PATCH 02/12] mm: fix boundary checking in free_bootmem_core fix

[Yinghai Lu]

[PATCH] mm: fix boundary checking in free_bootmem_core fix

	[PATCH] mm: fix boundary checking in free_bootmem_core

make the free_bootmem_core could handle out of range case. we could use
bdata_list to make sure the range can be freed for sure.

so next time, we don't need to loop online nodes and could use
free_bootmem directly.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/mm/bootmem.c
===================================================================
--- linux-2.6.orig/mm/bootmem.c
+++ linux-2.6/mm/bootmem.c
@@ -432,7 +432,9 @@ int __init reserve_bootmem(unsigned long
 
 void __init free_bootmem(unsigned long addr, unsigned long size)
 {
-	free_bootmem_core(NODE_DATA(0)->bdata, addr, size);
+	bootmem_data_t *bdata;
+	list_for_each_entry(bdata, &bdata_list, list)
+		free_bootmem_core(bdata, addr, size);
 }
 
 unsigned long __init free_all_bootmem(void)

[PATCH 03/12] x86_64: free_bootmem should take phys

[Yinghai Lu]

[PATCH] x86_64: free_bootmem should take phys

so use nodedata_phys directly.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/arch/x86/mm/numa_64.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/numa_64.c
+++ linux-2.6/arch/x86/mm/numa_64.c
@@ -221,8 +221,7 @@ void __init setup_node_bootmem(int nodei
 				 bootmap_pages<<PAGE_SHIFT, PAGE_SIZE);
 	if (bootmap == NULL)  {
 		if (nodedata_phys < start || nodedata_phys >= end)
-			free_bootmem((unsigned long)node_data[nodeid],
-				     pgdat_size);
+			free_bootmem(nodedata_phys, pgdat_size);
 		node_data[nodeid] = NULL;
 		return;
 	}

[PATCH 04/12] x86_64: reserve dma32 early for gart

[Yinghai Lu]

[PATCH] x86_64: reserve dma32 early for gart

one system with 256g when numa is disabled said:

Your BIOS doesn't leave a aperture memory hole
Please enable the IOMMU option in the BIOS setup
This costs you 64 MB of RAM
Cannot allocate aperture memory hole (ffff8101c0000000,65536K)
Kernel panic - not syncing: Not enough memory for aperture
Pid: 0, comm: swapper Not tainted 2.6.25-rc4-x86-latest.git #33

Call Trace:
 [<ffffffff84037c62>] panic+0xb2/0x190
 [<ffffffff840381fc>] ? release_console_sem+0x7c/0x250
 [<ffffffff847b1628>] ? __alloc_bootmem_nopanic+0x48/0x90
 [<ffffffff847b0ac9>] ? free_bootmem+0x29/0x50
 [<ffffffff847ac1f7>] gart_iommu_hole_init+0x5e7/0x680
 [<ffffffff847b255b>] ? alloc_large_system_hash+0x16b/0x310
 [<ffffffff84506a2f>] ? _etext+0x0/0x1
 [<ffffffff847a2e8c>] pci_iommu_alloc+0x1c/0x40
 [<ffffffff847ac795>] mem_init+0x45/0x1a0
 [<ffffffff8479ff35>] start_kernel+0x295/0x380
 [<ffffffff8479f1c2>] _sinittext+0x1c2/0x230

the root cause is : memmap PMD is too big,
[ffffe200e0600000-ffffe200e07fffff] PMD ->ffff81383c000000 on node 0
almost near 4G..., and vmemmap_alloc_block will use up the ram under 4G.

solution will be:
1. make memmap allocation get memory above 4G...
2. reserve some dma32 range early before we try to set up memmap for all.
and release that before pci_iommu_alloc, so gart or swiotlb could get some
range under 4g limit for sure.

the patch is using method 2.
because method1 may need more code to handle SPARSEMEM and SPASEMEM_VMEMMAP

will get
Your BIOS doesn't leave a aperture memory hole
Please enable the IOMMU option in the BIOS setup
This costs you 64 MB of RAM
Mapping aperture over 65536 KB of RAM @ 4000000
Memory: 264245736k/268959744k available (8484k kernel code, 4187464k reserved, 4004k data, 724k init)

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/arch/x86/kernel/pci-dma_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/pci-dma_64.c
+++ linux-2.6/arch/x86/kernel/pci-dma_64.c
@@ -8,6 +8,8 @@
 #include <linux/pci.h>
 #include <linux/module.h>
 #include <linux/dmar.h>
+#include <linux/bootmem.h>
+#include <asm/proto.h>
 #include <asm/io.h>
 #include <asm/gart.h>
 #include <asm/calgary.h>
@@ -286,8 +288,53 @@ static __init int iommu_setup(char *p)
 }
 early_param("iommu", iommu_setup);
 
+static __initdata void *dma32_bootmem_ptr;
+static unsigned long dma32_bootmem_size __initdata = (128ULL<<20);
+
+static int __init parse_dma32_size_opt(char *p)
+{
+	if (!p)
+		return -EINVAL;
+	dma32_bootmem_size = memparse(p, &p);
+	return 0;
+}
+early_param("dma32_size", parse_dma32_size_opt);
+
+void __init dma32_reserve_bootmem(void)
+{
+	unsigned long size, align;
+	if (end_pfn <= MAX_DMA32_PFN)
+		return;
+
+	align = 64ULL<<20;
+	size = round_up(dma32_bootmem_size, align);
+	dma32_bootmem_ptr = __alloc_bootmem_nopanic(size, align,
+				 __pa(MAX_DMA_ADDRESS));
+	if (dma32_bootmem_ptr)
+		dma32_bootmem_size = size;
+	else
+		dma32_bootmem_size = 0;
+}
+static void __init dma32_free_bootmem(void)
+{
+	int node;
+
+	if (end_pfn <= MAX_DMA32_PFN)
+		return;
+
+	if (!dma32_bootmem_ptr)
+		return;
+
+	free_bootmem(__pa(dma32_bootmem_ptr), dma32_bootmem_size);
+
+	dma32_bootmem_ptr = NULL;
+	dma32_bootmem_size = 0;
+}
+
 void __init pci_iommu_alloc(void)
 {
+	/* free the range so iommu could get some range less than 4G */
+	dma32_free_bootmem();
 	/*
 	 * The order of these functions is important for
 	 * fall-back/fail-over reasons
Index: linux-2.6/arch/x86/kernel/setup_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/setup_64.c
+++ linux-2.6/arch/x86/kernel/setup_64.c
@@ -397,6 +397,8 @@ void __init setup_arch(char **cmdline_p)
 
 	early_res_to_bootmem();
 
+	dma32_reserve_bootmem();
+
 #ifdef CONFIG_ACPI_SLEEP
 	/*
 	 * Reserve low memory region for sleep support.
Index: linux-2.6/include/asm-x86/pci_64.h
===================================================================
--- linux-2.6.orig/include/asm-x86/pci_64.h
+++ linux-2.6/include/asm-x86/pci_64.h
@@ -25,6 +25,7 @@ extern int (*pci_config_write)(int seg, 
 
 
 
+extern void dma32_reserve_bootmem(void);
 extern void pci_iommu_alloc(void);
 
 /* The PCI address space does equal the physical memory

[PATCH 05/12] mm: make mem_map allocation continuous.

[Yinghai Lu]

[PATCH] mm: make mem_map allocation continuous.

vmemmap allocation current got
 [ffffe20000000000-ffffe200001fffff] PMD ->ffff810001400000 on node 0
 [ffffe20000200000-ffffe200003fffff] PMD ->ffff810001800000 on node 0
 [ffffe20000400000-ffffe200005fffff] PMD ->ffff810001c00000 on node 0
 [ffffe20000600000-ffffe200007fffff] PMD ->ffff810002000000 on node 0
 [ffffe20000800000-ffffe200009fffff] PMD ->ffff810002400000 on node 0
...

there is 2M hole between them.

the rootcause is that usemap (24 bytes) will be allocated after every 2M
mem_map. and it will push next vmemmap (2M) to next align (2M).

solution:
try to allocate mem_map continously.

after patch, will get
 [ffffe20000000000-ffffe200001fffff] PMD ->ffff810001400000 on node 0
 [ffffe20000200000-ffffe200003fffff] PMD ->ffff810001600000 on node 0
 [ffffe20000400000-ffffe200005fffff] PMD ->ffff810001800000 on node 0
 [ffffe20000600000-ffffe200007fffff] PMD ->ffff810001a00000 on node 0
 [ffffe20000800000-ffffe200009fffff] PMD ->ffff810001c00000 on node 0
...
and usemap will share in page because of they are allocated continuously too.
sparse_early_usemap_alloc: usemap = ffff810024e00000 size = 24
sparse_early_usemap_alloc: usemap = ffff810024e00080 size = 24
sparse_early_usemap_alloc: usemap = ffff810024e00100 size = 24
sparse_early_usemap_alloc: usemap = ffff810024e00180 size = 24
...

so we make the bootmem allocation more compact and use less memory for usemap.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/mm/sparse.c
===================================================================
--- linux-2.6.orig/mm/sparse.c
+++ linux-2.6/mm/sparse.c
@@ -285,6 +286,8 @@ struct page __init *sparse_early_mem_map
 	return NULL;
 }
 
+/* section_map pointer array is 64k */
+static __initdata struct page *section_map[NR_MEM_SECTIONS];
 /*
  * Allocate the accumulated non-linear sections, allocate a mem_map
  * for each and record the physical to section mapping.
@@ -295,14 +298,29 @@ void __init sparse_init(void)
 	struct page *map;
 	unsigned long *usemap;
 
+	/*
+	 * map is using big page (aka 2M in x86 64 bit)
+	 * usemap is less one page (aka 24 bytes)
+	 * so alloc 2M (with 2M align) and 24 bytes in turn will
+	 * make next 2M slip to one more 2M later.
+	 * then in big system, the memmory will have a lot hole...
+	 * here try to allocate 2M pages continously.
+	 */
 	for (pnum = 0; pnum < NR_MEM_SECTIONS; pnum++) {
 		if (!present_section_nr(pnum))
 			continue;
+		section_map[pnum] = sparse_early_mem_map_alloc(pnum);
+	}
 
-		map = sparse_early_mem_map_alloc(pnum);
-		if (!map)
+
+	for (pnum = 0; pnum < NR_MEM_SECTIONS; pnum++) {
+		if (!present_section_nr(pnum))
 			continue;
 
+		map = section_map[pnum];
+		if (!map)
+			 continue;
+
 		usemap = sparse_early_usemap_alloc(pnum);
 		if (!usemap)
 			continue;

[PATCH 06/12] mm: fix alloc_bootmem_core to use fast searching for all nodes

[Yinghai Lu]

[PATCH] mm: fix alloc_bootmem_core to use fast searching for all nodes

make the nodes other than node 0 could use bdata->last_success for fast search too.

we need to use __alloc_bootmem_core for vmemmap allocation for other nodes when
numa and sparsemem/vmemmap are enabled.

also make fail_block path increase i with incr only needed after ALIGN to avoid
extra increase when size is large than align.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/mm/bootmem.c
===================================================================
--- linux-2.6.orig/mm/bootmem.c
+++ linux-2.6/mm/bootmem.c
@@ -238,28 +238,32 @@ __alloc_bootmem_core(struct bootmem_data
 	 * We try to allocate bootmem pages above 'goal'
 	 * first, then we try to allocate lower pages.
 	 */
-	if (goal && goal >= bdata->node_boot_start && PFN_DOWN(goal) < end_pfn) {
-		preferred = goal - bdata->node_boot_start;
+	preferred = 0;
+	if (goal && PFN_DOWN(goal) < end_pfn) {
+		if (goal > bdata->node_boot_start)
+			preferred = goal - bdata->node_boot_start;
 
 		if (bdata->last_success >= preferred)
 			if (!limit || (limit && limit > bdata->last_success))
 				preferred = bdata->last_success;
-	} else
-		preferred = 0;
+	}
 
 	preferred = PFN_DOWN(ALIGN(preferred, align)) + offset;
 	areasize = (size + PAGE_SIZE-1) / PAGE_SIZE;
 	incr = align >> PAGE_SHIFT ? : 1;
 
 restart_scan:
-	for (i = preferred; i < eidx; i += incr) {
+	for (i = preferred; i < eidx;) {
 		unsigned long j;
+
 		i = find_next_zero_bit(bdata->node_bootmem_map, eidx, i);
 		i = ALIGN(i, incr);
 		if (i >= eidx)
 			break;
-		if (test_bit(i, bdata->node_bootmem_map))
+		if (test_bit(i, bdata->node_bootmem_map)) {
+			i += incr;
 			continue;
+		}
 		for (j = i + 1; j < i + areasize; ++j) {
 			if (j >= eidx)
 				goto fail_block;
@@ -270,6 +274,8 @@ restart_scan:
 		goto found;
 	fail_block:
 		i = ALIGN(j, incr);
+		if (i == j)
+			i += incr;
 	}
 
 	if (preferred > offset) {

[PATCH 07/12] mm: offset align in alloc_bootmem v3

[Yinghai Lu]

[PATCH] mm: offset align in alloc_bootmem v3

need offset alignment when node_boot_start's alignment is less than
align required

use local node_boot_start to match align. so don't add extra opteration in
search loop.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/mm/bootmem.c
===================================================================
--- linux-2.6.orig/mm/bootmem.c
+++ linux-2.6/mm/bootmem.c
@@ -206,9 +206,11 @@ void * __init
 __alloc_bootmem_core(struct bootmem_data *bdata, unsigned long size,
 	      unsigned long align, unsigned long goal, unsigned long limit)
 {
-	unsigned long offset, remaining_size, areasize, preferred;
+	unsigned long areasize, preferred;
 	unsigned long i, start = 0, incr, eidx, end_pfn;
 	void *ret;
+	unsigned long node_boot_start;
+	void *node_bootmem_map;
 
 	if (!size) {
 		printk("__alloc_bootmem_core(): zero-sized request\n");
@@ -216,23 +218,29 @@ __alloc_bootmem_core(struct bootmem_data
 	}
 	BUG_ON(align & (align-1));
 
-	if (limit && bdata->node_boot_start >= limit)
-		return NULL;
-
 	/* on nodes without memory - bootmem_map is NULL */
 	if (!bdata->node_bootmem_map)
 		return NULL;
 
+	/* bdata->node_boot_start is supposed to be (12+6)bits alignment on x86_64 ? */
+	node_boot_start = bdata->node_boot_start;
+	node_bootmem_map = bdata->node_bootmem_map;
+	if (align) {
+		node_boot_start = ALIGN(bdata->node_boot_start, align);
+		if (node_boot_start > bdata->node_boot_start)
+			node_bootmem_map = (unsigned long *)bdata->node_bootmem_map +
+			    PFN_DOWN(node_boot_start - bdata->node_boot_start)/BITS_PER_LONG;
+	}
+
+	if (limit && node_boot_start >= limit)
+		return NULL;
+
 	end_pfn = bdata->node_low_pfn;
 	limit = PFN_DOWN(limit);
 	if (limit && end_pfn > limit)
 		end_pfn = limit;
 
-	eidx = end_pfn - PFN_DOWN(bdata->node_boot_start);
-	offset = 0;
-	if (align && (bdata->node_boot_start & (align - 1UL)) != 0)
-		offset = align - (bdata->node_boot_start & (align - 1UL));
-	offset = PFN_DOWN(offset);
+	eidx = end_pfn - PFN_DOWN(node_boot_start);
 
 	/*
 	 * We try to allocate bootmem pages above 'goal'
@@ -240,15 +248,16 @@ __alloc_bootmem_core(struct bootmem_data
 	 */
 	preferred = 0;
 	if (goal && PFN_DOWN(goal) < end_pfn) {
-		if (goal > bdata->node_boot_start)
-			preferred = goal - bdata->node_boot_start;
+		if (goal > node_boot_start)
+			preferred = goal - node_boot_start;
 
-		if (bdata->last_success >= preferred)
+		if (bdata->last_success > node_boot_start &&
+			bdata->last_success - node_boot_start >= preferred)
 			if (!limit || (limit && limit > bdata->last_success))
-				preferred = bdata->last_success;
+				preferred = bdata->last_success - node_boot_start;
 	}
 
-	preferred = PFN_DOWN(ALIGN(preferred, align)) + offset;
+	preferred = PFN_DOWN(ALIGN(preferred, align));
 	areasize = (size + PAGE_SIZE-1) / PAGE_SIZE;
 	incr = align >> PAGE_SHIFT ? : 1;
 
@@ -256,18 +265,18 @@ restart_scan:
 	for (i = preferred; i < eidx;) {
 		unsigned long j;
 
-		i = find_next_zero_bit(bdata->node_bootmem_map, eidx, i);
+		i = find_next_zero_bit(node_bootmem_map, eidx, i);
 		i = ALIGN(i, incr);
 		if (i >= eidx)
 			break;
-		if (test_bit(i, bdata->node_bootmem_map)) {
+		if (test_bit(i, node_bootmem_map)) {
 			i += incr;
 			continue;
 		}
 		for (j = i + 1; j < i + areasize; ++j) {
 			if (j >= eidx)
 				goto fail_block;
-			if (test_bit(j, bdata->node_bootmem_map))
+			if (test_bit(j, node_bootmem_map))
 				goto fail_block;
 		}
 		start = i;
@@ -278,14 +287,14 @@ restart_scan:
 			i += incr;
 	}
 
-	if (preferred > offset) {
-		preferred = offset;
+	if (preferred > 0) {
+		preferred = 0;
 		goto restart_scan;
 	}
 	return NULL;
 
 found:
-	bdata->last_success = PFN_PHYS(start);
+	bdata->last_success = PFN_PHYS(start) + node_boot_start;
 	BUG_ON(start >= eidx);
 
 	/*
@@ -295,6 +304,7 @@ found:
 	 */
 	if (align < PAGE_SIZE &&
 	    bdata->last_offset && bdata->last_pos+1 == start) {
+		unsigned long offset, remaining_size;
 		offset = ALIGN(bdata->last_offset, align);
 		BUG_ON(offset > PAGE_SIZE);
 		remaining_size = PAGE_SIZE - offset;
@@ -303,14 +313,12 @@ found:
 			/* last_pos unchanged */
 			bdata->last_offset = offset + size;
 			ret = phys_to_virt(bdata->last_pos * PAGE_SIZE +
-					   offset +
-					   bdata->node_boot_start);
+					   offset + node_boot_start);
 		} else {
 			remaining_size = size - remaining_size;
 			areasize = (remaining_size + PAGE_SIZE-1) / PAGE_SIZE;
 			ret = phys_to_virt(bdata->last_pos * PAGE_SIZE +
-					   offset +
-					   bdata->node_boot_start);
+					   offset + node_boot_start);
 			bdata->last_pos = start + areasize - 1;
 			bdata->last_offset = remaining_size;
 		}
@@ -318,14 +326,14 @@ found:
 	} else {
 		bdata->last_pos = start + areasize - 1;
 		bdata->last_offset = size & ~PAGE_MASK;
-		ret = phys_to_virt(start * PAGE_SIZE + bdata->node_boot_start);
+		ret = phys_to_virt(start * PAGE_SIZE + node_boot_start);
 	}
 
 	/*
 	 * Reserve the area now:
 	 */
 	for (i = start; i < start + areasize; i++)
-		if (unlikely(test_and_set_bit(i, bdata->node_bootmem_map)))
+		if (unlikely(test_and_set_bit(i, node_bootmem_map)))
 			BUG();
 	memset(ret, 0, size);
 	return ret;

[PATCH 08/12] mm: allocate section_map for sparse_init

[Yinghai Lu]

[PATCH] mm: allocate section_map for sparse_init

allocate section_map in bootmem instead of using __initdata.

need to apply after
	[PATCH] mm: fix boundary checking in free_bootmem_core
	[PATCH] mm: make mem_map allocation continuous.


Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/mm/sparse.c
===================================================================
--- linux-2.6.orig/mm/sparse.c
+++ linux-2.6/mm/sparse.c
@@ -285,8 +285,6 @@ struct page __init *sparse_early_mem_map
 	return NULL;
 }
 
-/* section_map pointer array is 64k */
-static __initdata struct page *section_map[NR_MEM_SECTIONS];
 /*
  * Allocate the accumulated non-linear sections, allocate a mem_map
  * for each and record the physical to section mapping.
@@ -296,6 +294,9 @@ void __init sparse_init(void)
 	unsigned long pnum;
 	struct page *map;
 	unsigned long *usemap;
+	struct page **section_map;
+	int size;
+	int node;
 
 	/*
 	 * map is using big page (aka 2M in x86 64 bit)
@@ -305,13 +306,17 @@ void __init sparse_init(void)
 	 * then in big system, the memmory will have a lot hole...
 	 * here try to allocate 2M pages continously.
 	 */
+	size = sizeof(struct page *) * NR_MEM_SECTIONS;
+	section_map = alloc_bootmem(size);
+	if (!section_map)
+		panic("can not allocate section_map\n");
+
 	for (pnum = 0; pnum < NR_MEM_SECTIONS; pnum++) {
 		if (!present_section_nr(pnum))
 			continue;
 		section_map[pnum] = sparse_early_mem_map_alloc(pnum);
 	}
 
-
 	for (pnum = 0; pnum < NR_MEM_SECTIONS; pnum++) {
 		if (!present_section_nr(pnum))
 			continue;
@@ -327,6 +332,8 @@ void __init sparse_init(void)
 		sparse_init_one_section(__nr_to_section(pnum), pnum, map,
 								usemap);
 	}
+
+	free_bootmem(__pa(section_map), size);
 }
 
 #ifdef CONFIG_MEMORY_HOTPLUG

[PATCH 09/12] mm: make reserve_bootmem can crossed the nodes v2

[Yinghai Lu]

[PATCH] mm: make reserve_bootmem can crossed the nodes v2

split reserve_bootmem_core to two function, one check conflicts, and one set bits.

and make reserve_bootmem to loop bdata_list to cross the nodes.

user could be crashkernel and ramdisk..., in case the range cross the nodes

v2, fix out of range check

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/mm/bootmem.c
===================================================================
--- linux-2.6.orig/mm/bootmem.c
+++ linux-2.6/mm/bootmem.c
@@ -111,44 +111,71 @@ static unsigned long __init init_bootmem
  * might be used for boot-time allocations - or it might get added
  * to the free page pool later on.
  */
-static int __init reserve_bootmem_core(bootmem_data_t *bdata,
+static int __init can_reserve_bootmem_core(bootmem_data_t *bdata,
 			unsigned long addr, unsigned long size, int flags)
 {
 	unsigned long sidx, eidx;
 	unsigned long i;
-	int ret;
+
+	BUG_ON(!size);
+
+	/* out of range, don't hold other */
+	if (addr + size < bdata->node_boot_start ||
+		PFN_DOWN(addr) > bdata->node_low_pfn)
+		return 0;
 
 	/*
-	 * round up, partially reserved pages are considered
-	 * fully reserved.
+	 * Round up to index to the range.
 	 */
+	if (addr > bdata->node_boot_start)
+		sidx= PFN_DOWN(addr - bdata->node_boot_start);
+	else
+		sidx = 0;
+
+	eidx = PFN_UP(addr + size - bdata->node_boot_start);
+	if (eidx > bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start))
+		eidx = bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start);
+
+	for (i = sidx; i < eidx; i++)
+		if (test_bit(i, bdata->node_bootmem_map)) {
+			if (flags & BOOTMEM_EXCLUSIVE)
+				return -EBUSY;
+		}
+
+	return 0;
+
+}
+static void __init reserve_bootmem_core(bootmem_data_t *bdata,
+			unsigned long addr, unsigned long size, int flags)
+{
+	unsigned long sidx, eidx;
+	unsigned long i;
+
 	BUG_ON(!size);
-	BUG_ON(PFN_DOWN(addr) >= bdata->node_low_pfn);
-	BUG_ON(PFN_UP(addr + size) > bdata->node_low_pfn);
-	BUG_ON(addr < bdata->node_boot_start);
 
-	sidx = PFN_DOWN(addr - bdata->node_boot_start);
+	/* out of range */
+	if (addr + size < bdata->node_boot_start ||
+		PFN_DOWN(addr) > bdata->node_low_pfn)
+		return;
+
+	/*
+	 * Round up to index to the range.
+	 */
+	if (addr > bdata->node_boot_start)
+		sidx= PFN_DOWN(addr - bdata->node_boot_start);
+	else
+		sidx = 0;
+
 	eidx = PFN_UP(addr + size - bdata->node_boot_start);
+	if (eidx > bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start))
+		eidx = bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start);
 
 	for (i = sidx; i < eidx; i++)
 		if (test_and_set_bit(i, bdata->node_bootmem_map)) {
 #ifdef CONFIG_DEBUG_BOOTMEM
 			printk("hm, page %08lx reserved twice.\n", i*PAGE_SIZE);
 #endif
-			if (flags & BOOTMEM_EXCLUSIVE) {
-				ret = -EBUSY;
-				goto err;
-			}
 		}
-
-	return 0;
-
-err:
-	/* unreserve memory we accidentally reserved */
-	for (i--; i >= sidx; i--)
-		clear_bit(i, bdata->node_bootmem_map);
-
-	return ret;
 }
 
 static void __init free_bootmem_core(bootmem_data_t *bdata, unsigned long addr,
@@ -415,6 +442,11 @@ unsigned long __init init_bootmem_node(p
 void __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
 				 unsigned long size, int flags)
 {
+	int ret;
+
+	ret = can_reserve_bootmem_core(pgdat->bdata, physaddr, size, flags);
+	if (ret < 0)
+		return;
 	reserve_bootmem_core(pgdat->bdata, physaddr, size, flags);
 }
 
@@ -440,7 +472,16 @@ unsigned long __init init_bootmem(unsign
 int __init reserve_bootmem(unsigned long addr, unsigned long size,
 			    int flags)
 {
-	return reserve_bootmem_core(NODE_DATA(0)->bdata, addr, size, flags);
+	int ret;
+	bootmem_data_t *bdata;
+	list_for_each_entry(bdata, &bdata_list, list) {
+		ret = can_reserve_bootmem_core(bdata, addr, size, flags);
+		if (ret < 0)
+			return ret;
+	}
+	list_for_each_entry(bdata, &bdata_list, list)
+		reserve_bootmem_core(bdata, addr, size, flags);
+	return 0;
 }
 #endif /* !CONFIG_HAVE_ARCH_BOOTMEM_NODE */
 

[PATCH 10/12] x86_64: make reserve_bootmem_generic to use new reserve_bootmem

[Yinghai Lu]

[PATCH] x86_64: make reserve_bootmem_generic to use new reserve_bootmem

	[PATCH] mm: make reserve_bootmem can crossed the nodes

provides new resever_bootmem, let reserve_bootmem_generic use that.

acctualy reserve_bootme_generic is used to reserve initramdisk.
so we can make sure even bootloader or kexec load that cross the nodes, reserve_bootmem
still works.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>


Index: linux-2.6/arch/x86/mm/init_64.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/init_64.c
+++ linux-2.6/arch/x86/mm/init_64.c
@@ -648,7 +648,7 @@ void free_initrd_mem(unsigned long start
 void __init reserve_bootmem_generic(unsigned long phys, unsigned len)
 {
 #ifdef CONFIG_NUMA
-	int nid = phys_to_nid(phys);
+	int nid, next_nid;
 #endif
 	unsigned long pfn = phys >> PAGE_SHIFT;
 
@@ -667,10 +667,14 @@ void __init reserve_bootmem_generic(unsi
 
 	/* Should check here against the e820 map to avoid double free */
 #ifdef CONFIG_NUMA
+	nid = phys_to_nid(phys);
+	next_nid = phys_to_nid(phys + len - 1);
+	if (nid == next_nid)
 	reserve_bootmem_node(NODE_DATA(nid), phys, len, BOOTMEM_DEFAULT);
-#else
-	reserve_bootmem(phys, len, BOOTMEM_DEFAULT);
+	else
 #endif
+	reserve_bootmem(phys, len, BOOTMEM_DEFAULT);
+
 	if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) {
 		dma_reserve += len / PAGE_SIZE;
 		set_dma_reserve(dma_reserve);

Re: [PATCH 10/12] x86_64: make reserve_bootmem_generic to use new reserve_bootmem

[Ingo Molnar]

* Yinghai Lu <yhlu.kernel.send@gmail.com> wrote:

> [PATCH] x86_64: make reserve_bootmem_generic to use new reserve_bootmem
> 
> 	[PATCH] mm: make reserve_bootmem can crossed the nodes
> 
> provides new resever_bootmem, let reserve_bootmem_generic use that.
> 
> acctualy reserve_bootme_generic is used to reserve initramdisk. so we 
> can make sure even bootloader or kexec load that cross the nodes, 
> reserve_bootmem still works.

.25 fix, correct? But due to dependency on the mm/bootmem.c change this 
patch should go via Andrew.

Acked-by: Ingo Molnar <mingo@elte.hu>

	Ingo

[PATCH 11/12] x86_64: do not reserve ramdisk two times

[Yinghai Lu]

[PATCH] x86_64: do not reserve ramdisk two times

ramdisk is reserved via reserve_early in x86_64_start_kernel,
later early_res_to_bootmem() will convert to reservation in bootmem.

so don't need to reserve that again.

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/arch/x86/kernel/head64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/head64.c
+++ linux-2.6/arch/x86/kernel/head64.c
@@ -110,6 +110,7 @@ void __init x86_64_start_kernel(char * r
 
 	reserve_early(__pa_symbol(&_text), __pa_symbol(&_end), "TEXT DATA BSS");
 
+#ifdef CONFIG_BLK_DEV_INITRD
 	/* Reserve INITRD */
 	if (boot_params.hdr.type_of_loader && boot_params.hdr.ramdisk_image) {
 		unsigned long ramdisk_image = boot_params.hdr.ramdisk_image;
@@ -117,6 +118,7 @@ void __init x86_64_start_kernel(char * r
 		unsigned long ramdisk_end   = ramdisk_image + ramdisk_size;
 		reserve_early(ramdisk_image, ramdisk_end, "RAMDISK");
 	}
+#endif
 
 	reserve_ebda();
 
Index: linux-2.6/arch/x86/kernel/setup_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/setup_64.c
+++ linux-2.6/arch/x86/kernel/setup_64.c
@@ -421,11 +421,14 @@ void __init setup_arch(char **cmdline_p)
 		unsigned long end_of_mem    = end_pfn << PAGE_SHIFT;
 
 		if (ramdisk_end <= end_of_mem) {
-			reserve_bootmem_generic(ramdisk_image, ramdisk_size);
+			/*
+			 * don't need to reserve again, already reserved early
+			 * in x86_64_start_kernel, and early_res_to_bootmem
+			 * convert that to reserved in bootmem
+			 */
 			initrd_start = ramdisk_image + PAGE_OFFSET;
 			initrd_end = initrd_start+ramdisk_size;
 		} else {
-			/* Assumes everything on node 0 */
 			free_bootmem(ramdisk_image, ramdisk_size);
 			printk(KERN_ERR "initrd extends beyond end of memory "
 			       "(0x%08lx > 0x%08lx)\ndisabling initrd\n",

[PATCH 12/12] x86_64: fix setup_node_bootmem to support big mem excluding with memmap

[Yinghai Lu]

[PATCH] x86_64: fix setup_node_bootmem to support big mem excluding with memmap

typical case: four sockets system, every node has 4g ram, and we are using
	memmap=10g$4g to mask out memory on node1 and node2

when numa is enabled, early_node_mem is used to get node_data and node_bootmap

if it can not get from same node with find_e820_area, it will use alloc_bootmem
to get buff from previous nodes.

so check it and issue info about it.

need to move early_res_to_bootmem into every setup_node_bootmem.
and it takes range that node has. otherwise alloc_bootmem could return addr
that reserved early.

need to apply it after
	[PATCH] mm: make reserve_bootmem can crossed the nodes

Signed-off-by: Yinghai Lu <yhlu.kernel@gmail.com>

Index: linux-2.6/arch/x86/mm/numa_64.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/numa_64.c
+++ linux-2.6/arch/x86/mm/numa_64.c
@@ -188,6 +188,7 @@ void __init setup_node_bootmem(int nodei
 	unsigned long bootmap_start, nodedata_phys;
 	void *bootmap;
 	const int pgdat_size = round_up(sizeof(pg_data_t), PAGE_SIZE);
+	int nid;
 
 	start = round_up(start, ZONE_ALIGN);
 
@@ -210,9 +211,20 @@ void __init setup_node_bootmem(int nodei
 	NODE_DATA(nodeid)->node_start_pfn = start_pfn;
 	NODE_DATA(nodeid)->node_spanned_pages = end_pfn - start_pfn;
 
-	/* Find a place for the bootmem map */
+	/*
+	 * Find a place for the bootmem map
+	 * nodedata_phys could be on other nodes by alloc_bootmem,
+	 * so need to sure bootmap_start not to be small, otherwise
+	 * early_node_mem will get that with find_e820_area instead
+	 * of alloc_bootmem, that could clash with reserved range
+	 */
 	bootmap_pages = bootmem_bootmap_pages(end_pfn - start_pfn);
-	bootmap_start = round_up(nodedata_phys + pgdat_size, PAGE_SIZE);
+	nid = phys_to_nid(nodedata_phys);
+	if (nid == nodeid)
+		bootmap_start = round_up(nodedata_phys + pgdat_size,
+					 PAGE_SIZE);
+	else
+		bootmap_start = round_up(start, PAGE_SIZE);
 	/*
 	 * SMP_CAHCE_BYTES could be enough, but init_bootmem_node like
 	 * to use that to align to PAGE_SIZE
@@ -237,10 +249,29 @@ void __init setup_node_bootmem(int nodei
 
 	free_bootmem_with_active_regions(nodeid, end);
 
-	reserve_bootmem_node(NODE_DATA(nodeid), nodedata_phys, pgdat_size,
-			BOOTMEM_DEFAULT);
-	reserve_bootmem_node(NODE_DATA(nodeid), bootmap_start,
-			bootmap_pages<<PAGE_SHIFT, BOOTMEM_DEFAULT);
+	/*
+	 * convert early reserve to bootmem reserve earlier
+	 * otherwise early_node_mem could use early reserved mem
+	 * on previous node
+	 */
+	early_res_to_bootmem(start, end);
+
+	/*
+	 * in some case early_node_mem could use alloc_bootmem
+	 * to get range on other node, don't reserve that again
+	 */
+	if (nid != nodeid)
+		printk(KERN_INFO "    NODE_DATA(%d) on node %d\n", nodeid, nid);
+	else
+		reserve_bootmem_node(NODE_DATA(nodeid), nodedata_phys,
+					pgdat_size, BOOTMEM_DEFAULT);
+	nid = phys_to_nid(bootmap_start);
+	if (nid != nodeid)
+		printk(KERN_INFO "    bootmap(%d) on node %d\n", nodeid, nid);
+	else
+		reserve_bootmem_node(NODE_DATA(nodeid), bootmap_start,
+				 bootmap_pages<<PAGE_SHIFT, BOOTMEM_DEFAULT);
+
 #ifdef CONFIG_ACPI_NUMA
 	srat_reserve_add_area(nodeid);
 #endif
Index: linux-2.6/arch/x86/kernel/e820_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/e820_64.c
+++ linux-2.6/arch/x86/kernel/e820_64.c
@@ -83,14 +83,19 @@ void __init reserve_early(unsigned long 
 		strncpy(r->name, name, sizeof(r->name) - 1);
 }
 
-void __init early_res_to_bootmem(void)
+void __init early_res_to_bootmem(unsigned long start, unsigned long end)
 {
 	int i;
+	unsigned long final_start, final_end;
 	for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) {
 		struct early_res *r = &early_res[i];
-		printk(KERN_INFO "early res: %d [%lx-%lx] %s\n", i,
-			r->start, r->end - 1, r->name);
-		reserve_bootmem_generic(r->start, r->end - r->start);
+		final_start = max(start, r->start);
+		final_end = min(end, r->end);
+		if (final_start >= final_end)
+			continue;
+		printk(KERN_INFO "  early res: %d [%lx-%lx] %s\n", i,
+			final_start, final_end - 1, r->name);
+		reserve_bootmem_generic(final_start, final_end - final_start);
 	}
 }
 
Index: linux-2.6/include/asm-x86/e820_64.h
===================================================================
--- linux-2.6.orig/include/asm-x86/e820_64.h
+++ linux-2.6/include/asm-x86/e820_64.h
@@ -41,7 +41,7 @@ extern struct e820map e820;
 extern void update_e820(void);
 
 extern void reserve_early(unsigned long start, unsigned long end, char *name);
-extern void early_res_to_bootmem(void);
+extern void early_res_to_bootmem(unsigned long start, unsigned long end);
 
 #endif/*!__ASSEMBLY__*/
 
Index: linux-2.6/arch/x86/kernel/setup_64.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/setup_64.c
+++ linux-2.6/arch/x86/kernel/setup_64.c
@@ -190,6 +190,7 @@ contig_initmem_init(unsigned long start_
 	bootmap_size = init_bootmem(bootmap >> PAGE_SHIFT, end_pfn);
 	e820_register_active_regions(0, start_pfn, end_pfn);
 	free_bootmem_with_active_regions(0, end_pfn);
+	early_res_to_bootmem(0, end_pfn<<PAGE_SHIFT);
 	reserve_bootmem(bootmap, bootmap_size, BOOTMEM_DEFAULT);
 }
 #endif
@@ -395,8 +396,6 @@ void __init setup_arch(char **cmdline_p)
 	contig_initmem_init(0, end_pfn);
 #endif
 
-	early_res_to_bootmem();
-
 	dma32_reserve_bootmem();
 
 #ifdef CONFIG_ACPI_SLEEP

Re: [PATCH 12/12] x86_64: fix setup_node_bootmem to support big mem excluding with memmap

[Ingo Molnar]

* Yinghai Lu <yhlu.kernel.send@gmail.com> wrote:

> need to apply it after
> 	[PATCH] mm: make reserve_bootmem can crossed the nodes

this too should go via Andrew due to the mm/bootmem.c dependency. (can 
pick it up into x86.git as well if the mm/bootmem.c goes into -git)

	Ingo