[GitHub/moto-9609/android_kernel_motorola_exynos9610.git] / mm / bootmem.c

/*
 *  linux/mm/bootmem.c
 *
 *  Copyright (C) 1999 Ingo Molnar
 *  Discontiguous memory support, Kanoj Sarcar, SGI, Nov 1999
 *
 *  simple boot-time physical memory area allocator and
 *  free memory collector. It's used to deal with reserved
 *  system memory and memory holes as well.
 */
#include <linux/init.h>
#include <linux/pfn.h>
#include <linux/bootmem.h>
#include <linux/module.h>

#include <asm/bug.h>
#include <asm/io.h>
#include <asm/processor.h>

#include "internal.h"

/*
 * Access to this subsystem has to be serialized externally. (this is
 * true for the boot process anyway)
 */
unsigned long max_low_pfn;
unsigned long min_low_pfn;
unsigned long max_pfn;

static LIST_HEAD(bdata_list);
#ifdef CONFIG_CRASH_DUMP
/*
 * If we have booted due to a crash, max_pfn will be a very low value. We need
 * to know the amount of memory that the previous kernel used.
 */
unsigned long saved_max_pfn;
#endif

/* return the number of _pages_ that will be allocated for the boot bitmap */
unsigned long __init bootmem_bootmap_pages(unsigned long pages)
{
	unsigned long mapsize;

	mapsize = (pages+7)/8;
	mapsize = (mapsize + ~PAGE_MASK) & PAGE_MASK;
	mapsize >>= PAGE_SHIFT;

	return mapsize;
}

/*
 * link bdata in order
 */
static void __init link_bootmem(bootmem_data_t *bdata)
{
	bootmem_data_t *ent;

	if (list_empty(&bdata_list)) {
		list_add(&bdata->list, &bdata_list);
		return;
	}
	/* insert in order */
	list_for_each_entry(ent, &bdata_list, list) {
		if (bdata->node_boot_start < ent->node_boot_start) {
			list_add_tail(&bdata->list, &ent->list);
			return;
		}
	}
	list_add_tail(&bdata->list, &bdata_list);
}

/*
 * Given an initialised bdata, it returns the size of the boot bitmap
 */
static unsigned long __init get_mapsize(bootmem_data_t *bdata)
{
	unsigned long mapsize;
	unsigned long start = PFN_DOWN(bdata->node_boot_start);
	unsigned long end = bdata->node_low_pfn;

	mapsize = ((end - start) + 7) / 8;
	return ALIGN(mapsize, sizeof(long));
}

/*
 * Called once to set up the allocator itself.
 */
static unsigned long __init init_bootmem_core(pg_data_t *pgdat,
	unsigned long mapstart, unsigned long start, unsigned long end)
{
	bootmem_data_t *bdata = pgdat->bdata;
	unsigned long mapsize;

	bdata->node_bootmem_map = phys_to_virt(PFN_PHYS(mapstart));
	bdata->node_boot_start = PFN_PHYS(start);
	bdata->node_low_pfn = end;
	link_bootmem(bdata);

	/*
	 * Initially all pages are reserved - setup_arch() has to
	 * register free RAM areas explicitly.
	 */
	mapsize = get_mapsize(bdata);
	memset(bdata->node_bootmem_map, 0xff, mapsize);

	return mapsize;
}

/*
 * Marks a particular physical memory range as unallocatable. Usable RAM
 * 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,
			unsigned long addr, unsigned long size, int flags)
{
	unsigned long sidx, eidx;
	unsigned long i;
	int ret;

	/*
	 * round up, partially reserved pages are considered
	 * fully reserved.
	 */
	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);
	eidx = PFN_UP(addr + size - 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,
				     unsigned long size)
{
	unsigned long sidx, eidx;
	unsigned long i;

	BUG_ON(!size);

	/* out range */
	if (addr + size < bdata->node_boot_start ||
		PFN_DOWN(addr) > bdata->node_low_pfn)
		return;
	/*
	 * round down end of usable mem, partially free pages are
	 * considered reserved.
	 */

	if (addr >= bdata->node_boot_start && addr < bdata->last_success)
		bdata->last_success = addr;

	/*
	 * Round up to index to the range.
	 */
	if (PFN_UP(addr) > PFN_DOWN(bdata->node_boot_start))
		sidx = PFN_UP(addr) - PFN_DOWN(bdata->node_boot_start);
	else
		sidx = 0;

	eidx = PFN_DOWN(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 (unlikely(!test_and_clear_bit(i, bdata->node_bootmem_map)))
			BUG();
	}
}

/*
 * We 'merge' subsequent allocations to save space. We might 'lose'
 * some fraction of a page if allocations cannot be satisfied due to
 * size constraints on boxes where there is physical RAM space
 * fragmentation - in these cases (mostly large memory boxes) this
 * is not a problem.
 *
 * On low memory boxes we get it right in 100% of the cases.
 *
 * alignment has to be a power of 2 value.
 *
 * NOTE:  This function is _not_ reentrant.
 */
void * __init
__alloc_bootmem_core(struct bootmem_data *bdata, unsigned long size,
	      unsigned long align, unsigned long goal, unsigned long limit)
{
	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");
		BUG();
	}
	BUG_ON(align & (align-1));

	/* 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(node_boot_start);

	/*
	 * We try to allocate bootmem pages above 'goal'
	 * first, then we try to allocate lower pages.
	 */
	preferred = 0;
	if (goal && PFN_DOWN(goal) < end_pfn) {
		if (goal > node_boot_start)
			preferred = goal - node_boot_start;

		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 - node_boot_start;
	}

	preferred = PFN_DOWN(ALIGN(preferred, align));
	areasize = (size + PAGE_SIZE-1) / PAGE_SIZE;
	incr = align >> PAGE_SHIFT ? : 1;

restart_scan:
	for (i = preferred; i < eidx;) {
		unsigned long j;

		i = find_next_zero_bit(node_bootmem_map, eidx, i);
		i = ALIGN(i, incr);
		if (i >= eidx)
			break;
		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, node_bootmem_map))
				goto fail_block;
		}
		start = i;
		goto found;
	fail_block:
		i = ALIGN(j, incr);
		if (i == j)
			i += incr;
	}

	if (preferred > 0) {
		preferred = 0;
		goto restart_scan;
	}
	return NULL;

found:
	bdata->last_success = PFN_PHYS(start) + node_boot_start;
	BUG_ON(start >= eidx);

	/*
	 * Is the next page of the previous allocation-end the start
	 * of this allocation's buffer? If yes then we can 'merge'
	 * the previous partial page with this allocation.
	 */
	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;
		if (size < remaining_size) {
			areasize = 0;
			/* last_pos unchanged */
			bdata->last_offset = offset + size;
			ret = phys_to_virt(bdata->last_pos * PAGE_SIZE +
					   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 + node_boot_start);
			bdata->last_pos = start + areasize - 1;
			bdata->last_offset = remaining_size;
		}
		bdata->last_offset &= ~PAGE_MASK;
	} else {
		bdata->last_pos = start + areasize - 1;
		bdata->last_offset = size & ~PAGE_MASK;
		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, node_bootmem_map)))
			BUG();
	memset(ret, 0, size);
	return ret;
}

static unsigned long __init free_all_bootmem_core(pg_data_t *pgdat)
{
	struct page *page;
	unsigned long pfn;
	bootmem_data_t *bdata = pgdat->bdata;
	unsigned long i, count, total = 0;
	unsigned long idx;
	unsigned long *map; 
	int gofast = 0;

	BUG_ON(!bdata->node_bootmem_map);

	count = 0;
	/* first extant page of the node */
	pfn = PFN_DOWN(bdata->node_boot_start);
	idx = bdata->node_low_pfn - pfn;
	map = bdata->node_bootmem_map;
	/* Check physaddr is O(LOG2(BITS_PER_LONG)) page aligned */
	if (bdata->node_boot_start == 0 ||
	    ffs(bdata->node_boot_start) - PAGE_SHIFT > ffs(BITS_PER_LONG))
		gofast = 1;
	for (i = 0; i < idx; ) {
		unsigned long v = ~map[i / BITS_PER_LONG];

		if (gofast && v == ~0UL) {
			int order;

			page = pfn_to_page(pfn);
			count += BITS_PER_LONG;
			order = ffs(BITS_PER_LONG) - 1;
			__free_pages_bootmem(page, order);
			i += BITS_PER_LONG;
			page += BITS_PER_LONG;
		} else if (v) {
			unsigned long m;

			page = pfn_to_page(pfn);
			for (m = 1; m && i < idx; m<<=1, page++, i++) {
				if (v & m) {
					count++;
					__free_pages_bootmem(page, 0);
				}
			}
		} else {
			i += BITS_PER_LONG;
		}
		pfn += BITS_PER_LONG;
	}
	total += count;

	/*
	 * Now free the allocator bitmap itself, it's not
	 * needed anymore:
	 */
	page = virt_to_page(bdata->node_bootmem_map);
	count = 0;
	idx = (get_mapsize(bdata) + PAGE_SIZE-1) >> PAGE_SHIFT;
	for (i = 0; i < idx; i++, page++) {
		__free_pages_bootmem(page, 0);
		count++;
	}
	total += count;
	bdata->node_bootmem_map = NULL;

	return total;
}

unsigned long __init init_bootmem_node(pg_data_t *pgdat, unsigned long freepfn,
				unsigned long startpfn, unsigned long endpfn)
{
	return init_bootmem_core(pgdat, freepfn, startpfn, endpfn);
}

void __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
				 unsigned long size, int flags)
{
	reserve_bootmem_core(pgdat->bdata, physaddr, size, flags);
}

void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
			      unsigned long size)
{
	free_bootmem_core(pgdat->bdata, physaddr, size);
}

unsigned long __init free_all_bootmem_node(pg_data_t *pgdat)
{
	return free_all_bootmem_core(pgdat);
}

unsigned long __init init_bootmem(unsigned long start, unsigned long pages)
{
	max_low_pfn = pages;
	min_low_pfn = start;
	return init_bootmem_core(NODE_DATA(0), start, 0, pages);
}

#ifndef CONFIG_HAVE_ARCH_BOOTMEM_NODE
int __init reserve_bootmem(unsigned long addr, unsigned long size,
			    int flags)
{
	return reserve_bootmem_core(NODE_DATA(0)->bdata, addr, size, flags);
}
#endif /* !CONFIG_HAVE_ARCH_BOOTMEM_NODE */

void __init free_bootmem(unsigned long addr, unsigned long 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)
{
	return free_all_bootmem_core(NODE_DATA(0));
}

void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align,
				      unsigned long goal)
{
	bootmem_data_t *bdata;
	void *ptr;

	list_for_each_entry(bdata, &bdata_list, list) {
		ptr = __alloc_bootmem_core(bdata, size, align, goal, 0);
		if (ptr)
			return ptr;
	}
	return NULL;
}

void * __init __alloc_bootmem(unsigned long size, unsigned long align,
			      unsigned long goal)
{
	void *mem = __alloc_bootmem_nopanic(size,align,goal);

	if (mem)
		return mem;
	/*
	 * Whoops, we cannot satisfy the allocation request.
	 */
	printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size);
	panic("Out of memory");
	return NULL;
}


void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
				   unsigned long align, unsigned long goal)
{
	void *ptr;

	ptr = __alloc_bootmem_core(pgdat->bdata, size, align, goal, 0);
	if (ptr)
		return ptr;

	return __alloc_bootmem(size, align, goal);
}

#ifndef ARCH_LOW_ADDRESS_LIMIT
#define ARCH_LOW_ADDRESS_LIMIT	0xffffffffUL
#endif

void * __init __alloc_bootmem_low(unsigned long size, unsigned long align,
				  unsigned long goal)
{
	bootmem_data_t *bdata;
	void *ptr;

	list_for_each_entry(bdata, &bdata_list, list) {
		ptr = __alloc_bootmem_core(bdata, size, align, goal,
						ARCH_LOW_ADDRESS_LIMIT);
		if (ptr)
			return ptr;
	}

	/*
	 * Whoops, we cannot satisfy the allocation request.
	 */
	printk(KERN_ALERT "low bootmem alloc of %lu bytes failed!\n", size);
	panic("Out of low memory");
	return NULL;
}

void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size,
				       unsigned long align, unsigned long goal)
{
	return __alloc_bootmem_core(pgdat->bdata, size, align, goal,
				    ARCH_LOW_ADDRESS_LIMIT);
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/bootmem.c
	3	*
	4	* Copyright (C) 1999 Ingo Molnar
	5	* Discontiguous memory support, Kanoj Sarcar, SGI, Nov 1999
	6	*
	7	* simple boot-time physical memory area allocator and
	8	* free memory collector. It's used to deal with reserved
	9	* system memory and memory holes as well.
	10	*/
1da177e4	11	#include <linux/init.h>
bbc7b92e	12	#include <linux/pfn.h>
1da177e4	13	#include <linux/bootmem.h>
1da177e4	14	#include <linux/module.h>
e786e86a FBH	15
e786e86a FBH	16	#include <asm/bug.h>
1da177e4	17	#include <asm/io.h>
dfd54cbc	18	#include <asm/processor.h>
e786e86a	19
1da177e4 LT	20	#include "internal.h"
	21
	22	/*
	23	* Access to this subsystem has to be serialized externally. (this is
	24	* true for the boot process anyway)
	25	*/
	26	unsigned long max_low_pfn;
	27	unsigned long min_low_pfn;
	28	unsigned long max_pfn;
	29
679bc9fb	30	static LIST_HEAD(bdata_list);
92aa63a5 VG	31	#ifdef CONFIG_CRASH_DUMP
	32	/*
	33	* If we have booted due to a crash, max_pfn will be a very low value. We need
	34	* to know the amount of memory that the previous kernel used.
	35	*/
	36	unsigned long saved_max_pfn;
	37	#endif
	38
1da177e4	39	/* return the number of _pages_ that will be allocated for the boot bitmap */
f71bf0ca	40	unsigned long __init bootmem_bootmap_pages(unsigned long pages)
1da177e4 LT	41	{
	42	unsigned long mapsize;
	43
	44	mapsize = (pages+7)/8;
	45	mapsize = (mapsize + ~PAGE_MASK) & PAGE_MASK;
	46	mapsize >>= PAGE_SHIFT;
	47
	48	return mapsize;
	49	}
f71bf0ca	50
679bc9fb KH	51	/*
	52	* link bdata in order
	53	*/
69d49e68	54	static void __init link_bootmem(bootmem_data_t *bdata)
679bc9fb KH	55	{
679bc9fb KH	56	bootmem_data_t *ent;
f71bf0ca	57
679bc9fb KH	58	if (list_empty(&bdata_list)) {
	59	list_add(&bdata->list, &bdata_list);
	60	return;
	61	}
	62	/* insert in order */
	63	list_for_each_entry(ent, &bdata_list, list) {
	64	if (bdata->node_boot_start < ent->node_boot_start) {
	65	list_add_tail(&bdata->list, &ent->list);
	66	return;
	67	}
	68	}
	69	list_add_tail(&bdata->list, &bdata_list);
679bc9fb KH	70	}
679bc9fb KH	71
bbc7b92e FBH	72	/*
	73	* Given an initialised bdata, it returns the size of the boot bitmap
	74	*/
	75	static unsigned long __init get_mapsize(bootmem_data_t *bdata)
	76	{
	77	unsigned long mapsize;
	78	unsigned long start = PFN_DOWN(bdata->node_boot_start);
	79	unsigned long end = bdata->node_low_pfn;
	80
	81	mapsize = ((end - start) + 7) / 8;
	82	return ALIGN(mapsize, sizeof(long));
	83	}
1da177e4 LT	84
	85	/*
	86	* Called once to set up the allocator itself.
	87	*/
f71bf0ca	88	static unsigned long __init init_bootmem_core(pg_data_t *pgdat,
1da177e4 LT	89	unsigned long mapstart, unsigned long start, unsigned long end)
	90	{
	91	bootmem_data_t *bdata = pgdat->bdata;
bbc7b92e	92	unsigned long mapsize;
1da177e4	93
bbc7b92e FBH	94	bdata->node_bootmem_map = phys_to_virt(PFN_PHYS(mapstart));
bbc7b92e FBH	95	bdata->node_boot_start = PFN_PHYS(start);
1da177e4	96	bdata->node_low_pfn = end;
679bc9fb	97	link_bootmem(bdata);
1da177e4 LT	98
	99	/*
	100	* Initially all pages are reserved - setup_arch() has to
	101	* register free RAM areas explicitly.
	102	*/
bbc7b92e	103	mapsize = get_mapsize(bdata);
1da177e4 LT	104	memset(bdata->node_bootmem_map, 0xff, mapsize);
	105
	106	return mapsize;
	107	}
	108
	109	/*
	110	* Marks a particular physical memory range as unallocatable. Usable RAM
	111	* might be used for boot-time allocations - or it might get added
	112	* to the free page pool later on.
	113	*/
72a7fe39 BW	114	static int __init reserve_bootmem_core(bootmem_data_t *bdata,
72a7fe39 BW	115	unsigned long addr, unsigned long size, int flags)
1da177e4	116	{
bbc7b92e	117	unsigned long sidx, eidx;
1da177e4	118	unsigned long i;
72a7fe39	119	int ret;
bbc7b92e	120
1da177e4 LT	121	/*
	122	* round up, partially reserved pages are considered
	123	* fully reserved.
	124	*/
1da177e4	125	BUG_ON(!size);
bbc7b92e FBH	126	BUG_ON(PFN_DOWN(addr) >= bdata->node_low_pfn);
bbc7b92e FBH	127	BUG_ON(PFN_UP(addr + size) > bdata->node_low_pfn);
5a982cbc	128	BUG_ON(addr < bdata->node_boot_start);
bbc7b92e FBH	129
	130	sidx = PFN_DOWN(addr - bdata->node_boot_start);
	131	eidx = PFN_UP(addr + size - bdata->node_boot_start);
1da177e4 LT	132
	133	for (i = sidx; i < eidx; i++)
	134	if (test_and_set_bit(i, bdata->node_bootmem_map)) {
	135	#ifdef CONFIG_DEBUG_BOOTMEM
	136	printk("hm, page %08lx reserved twice.\n", i*PAGE_SIZE);
	137	#endif
72a7fe39 BW	138	if (flags & BOOTMEM_EXCLUSIVE) {
	139	ret = -EBUSY;
	140	goto err;
	141	}
1da177e4	142	}
72a7fe39 BW	143
	144	return 0;
	145
	146	err:
	147	/* unreserve memory we accidentally reserved */
	148	for (i--; i >= sidx; i--)
	149	clear_bit(i, bdata->node_bootmem_map);
	150
	151	return ret;
1da177e4 LT	152	}
1da177e4 LT	153
bb0923a6 FBH	154	static void __init free_bootmem_core(bootmem_data_t *bdata, unsigned long addr,
bb0923a6 FBH	155	unsigned long size)
1da177e4	156	{
bbc7b92e	157	unsigned long sidx, eidx;
1da177e4	158	unsigned long i;
bbc7b92e	159
5a982cbc YL	160	BUG_ON(!size);
	161
	162	/* out range */
	163	if (addr + size < bdata->node_boot_start \|\|
	164	PFN_DOWN(addr) > bdata->node_low_pfn)
	165	return;
1da177e4 LT	166	/*
	167	* round down end of usable mem, partially free pages are
	168	* considered reserved.
	169	*/
1da177e4	170
5a982cbc	171	if (addr >= bdata->node_boot_start && addr < bdata->last_success)
1da177e4 LT	172	bdata->last_success = addr;
	173
	174	/*
5a982cbc	175	* Round up to index to the range.
1da177e4	176	*/
5a982cbc YL	177	if (PFN_UP(addr) > PFN_DOWN(bdata->node_boot_start))
	178	sidx = PFN_UP(addr) - PFN_DOWN(bdata->node_boot_start);
	179	else
	180	sidx = 0;
	181
bbc7b92e	182	eidx = PFN_DOWN(addr + size - bdata->node_boot_start);
5a982cbc YL	183	if (eidx > bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start))
5a982cbc YL	184	eidx = bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start);
1da177e4 LT	185
	186	for (i = sidx; i < eidx; i++) {
	187	if (unlikely(!test_and_clear_bit(i, bdata->node_bootmem_map)))
	188	BUG();
	189	}
	190	}
	191
	192	/*
	193	* We 'merge' subsequent allocations to save space. We might 'lose'
	194	* some fraction of a page if allocations cannot be satisfied due to
	195	* size constraints on boxes where there is physical RAM space
	196	* fragmentation - in these cases (mostly large memory boxes) this
	197	* is not a problem.
	198	*
	199	* On low memory boxes we get it right in 100% of the cases.
	200	*
	201	* alignment has to be a power of 2 value.
	202	*
	203	* NOTE: This function is _not_ reentrant.
	204	*/
267b4801	205	void * __init
1da177e4	206	__alloc_bootmem_core(struct bootmem_data *bdata, unsigned long size,
281dd25c	207	unsigned long align, unsigned long goal, unsigned long limit)
1da177e4	208	{
9a2dc04c	209	unsigned long areasize, preferred;
bbc7b92e	210	unsigned long i, start = 0, incr, eidx, end_pfn;
1da177e4	211	void *ret;
9a2dc04c YL	212	unsigned long node_boot_start;
9a2dc04c YL	213	void *node_bootmem_map;
1da177e4	214
f71bf0ca	215	if (!size) {
1da177e4 LT	216	printk("__alloc_bootmem_core(): zero-sized request\n");
	217	BUG();
	218	}
	219	BUG_ON(align & (align-1));
	220
7c309a64 CK	221	/* on nodes without memory - bootmem_map is NULL */
	222	if (!bdata->node_bootmem_map)
	223	return NULL;
	224
9a2dc04c YL	225	/* bdata->node_boot_start is supposed to be (12+6)bits alignment on x86_64 ? */
	226	node_boot_start = bdata->node_boot_start;
	227	node_bootmem_map = bdata->node_bootmem_map;
	228	if (align) {
	229	node_boot_start = ALIGN(bdata->node_boot_start, align);
	230	if (node_boot_start > bdata->node_boot_start)
	231	node_bootmem_map = (unsigned long *)bdata->node_bootmem_map +
	232	PFN_DOWN(node_boot_start - bdata->node_boot_start)/BITS_PER_LONG;
	233	}
	234
	235	if (limit && node_boot_start >= limit)
	236	return NULL;
	237
bbc7b92e FBH	238	end_pfn = bdata->node_low_pfn;
bbc7b92e FBH	239	limit = PFN_DOWN(limit);
281dd25c YG	240	if (limit && end_pfn > limit)
	241	end_pfn = limit;
	242
9a2dc04c	243	eidx = end_pfn - PFN_DOWN(node_boot_start);
1da177e4 LT	244
	245	/*
	246	* We try to allocate bootmem pages above 'goal'
	247	* first, then we try to allocate lower pages.
	248	*/
ad09315c YL	249	preferred = 0;
ad09315c YL	250	if (goal && PFN_DOWN(goal) < end_pfn) {
9a2dc04c YL	251	if (goal > node_boot_start)
9a2dc04c YL	252	preferred = goal - node_boot_start;
1da177e4	253
9a2dc04c YL	254	if (bdata->last_success > node_boot_start &&
9a2dc04c YL	255	bdata->last_success - node_boot_start >= preferred)
281dd25c	256	if (!limit \|\| (limit && limit > bdata->last_success))
9a2dc04c	257	preferred = bdata->last_success - node_boot_start;
ad09315c	258	}
1da177e4	259
9a2dc04c	260	preferred = PFN_DOWN(ALIGN(preferred, align));
bbc7b92e	261	areasize = (size + PAGE_SIZE-1) / PAGE_SIZE;
1da177e4 LT	262	incr = align >> PAGE_SHIFT ? : 1;
	263
	264	restart_scan:
ad09315c	265	for (i = preferred; i < eidx;) {
1da177e4	266	unsigned long j;
ad09315c	267
9a2dc04c	268	i = find_next_zero_bit(node_bootmem_map, eidx, i);
1da177e4	269	i = ALIGN(i, incr);
66d43e98 HM	270	if (i >= eidx)
66d43e98 HM	271	break;
9a2dc04c	272	if (test_bit(i, node_bootmem_map)) {
ad09315c	273	i += incr;
1da177e4	274	continue;
ad09315c	275	}
1da177e4 LT	276	for (j = i + 1; j < i + areasize; ++j) {
	277	if (j >= eidx)
	278	goto fail_block;
9a2dc04c	279	if (test_bit(j, node_bootmem_map))
1da177e4 LT	280	goto fail_block;
	281	}
	282	start = i;
	283	goto found;
	284	fail_block:
	285	i = ALIGN(j, incr);
ad09315c YL	286	if (i == j)
ad09315c YL	287	i += incr;
1da177e4 LT	288	}
1da177e4 LT	289
9a2dc04c YL	290	if (preferred > 0) {
9a2dc04c YL	291	preferred = 0;
1da177e4 LT	292	goto restart_scan;
	293	}
	294	return NULL;
	295
	296	found:
9a2dc04c	297	bdata->last_success = PFN_PHYS(start) + node_boot_start;
1da177e4 LT	298	BUG_ON(start >= eidx);
	299
	300	/*
	301	* Is the next page of the previous allocation-end the start
	302	* of this allocation's buffer? If yes then we can 'merge'
	303	* the previous partial page with this allocation.
	304	*/
	305	if (align < PAGE_SIZE &&
	306	bdata->last_offset && bdata->last_pos+1 == start) {
9a2dc04c	307	unsigned long offset, remaining_size;
8c0e33c1	308	offset = ALIGN(bdata->last_offset, align);
1da177e4	309	BUG_ON(offset > PAGE_SIZE);
f71bf0ca	310	remaining_size = PAGE_SIZE - offset;
1da177e4 LT	311	if (size < remaining_size) {
	312	areasize = 0;
	313	/* last_pos unchanged */
f71bf0ca FBH	314	bdata->last_offset = offset + size;
f71bf0ca FBH	315	ret = phys_to_virt(bdata->last_pos * PAGE_SIZE +
9a2dc04c	316	offset + node_boot_start);
1da177e4 LT	317	} else {
1da177e4 LT	318	remaining_size = size - remaining_size;
f71bf0ca FBH	319	areasize = (remaining_size + PAGE_SIZE-1) / PAGE_SIZE;
f71bf0ca FBH	320	ret = phys_to_virt(bdata->last_pos * PAGE_SIZE +
9a2dc04c	321	offset + node_boot_start);
f71bf0ca	322	bdata->last_pos = start + areasize - 1;
1da177e4 LT	323	bdata->last_offset = remaining_size;
	324	}
	325	bdata->last_offset &= ~PAGE_MASK;
	326	} else {
	327	bdata->last_pos = start + areasize - 1;
	328	bdata->last_offset = size & ~PAGE_MASK;
9a2dc04c	329	ret = phys_to_virt(start * PAGE_SIZE + node_boot_start);
1da177e4 LT	330	}
	331
	332	/*
	333	* Reserve the area now:
	334	*/
f71bf0ca	335	for (i = start; i < start + areasize; i++)
9a2dc04c	336	if (unlikely(test_and_set_bit(i, node_bootmem_map)))
1da177e4 LT	337	BUG();
	338	memset(ret, 0, size);
	339	return ret;
	340	}
	341
	342	static unsigned long __init free_all_bootmem_core(pg_data_t *pgdat)
	343	{
	344	struct page *page;
d41dee36	345	unsigned long pfn;
1da177e4 LT	346	bootmem_data_t *bdata = pgdat->bdata;
	347	unsigned long i, count, total = 0;
	348	unsigned long idx;
	349	unsigned long *map;
	350	int gofast = 0;
	351
	352	BUG_ON(!bdata->node_bootmem_map);
	353
	354	count = 0;
	355	/* first extant page of the node */
bbc7b92e FBH	356	pfn = PFN_DOWN(bdata->node_boot_start);
bbc7b92e FBH	357	idx = bdata->node_low_pfn - pfn;
1da177e4 LT	358	map = bdata->node_bootmem_map;
	359	/* Check physaddr is O(LOG2(BITS_PER_LONG)) page aligned */
	360	if (bdata->node_boot_start == 0 \|\|
	361	ffs(bdata->node_boot_start) - PAGE_SHIFT > ffs(BITS_PER_LONG))
	362	gofast = 1;
	363	for (i = 0; i < idx; ) {
	364	unsigned long v = ~map[i / BITS_PER_LONG];
d41dee36	365
1da177e4	366	if (gofast && v == ~0UL) {
a226f6c8	367	int order;
1da177e4	368
d41dee36	369	page = pfn_to_page(pfn);
1da177e4	370	count += BITS_PER_LONG;
1da177e4	371	order = ffs(BITS_PER_LONG) - 1;
a226f6c8	372	__free_pages_bootmem(page, order);
1da177e4 LT	373	i += BITS_PER_LONG;
	374	page += BITS_PER_LONG;
	375	} else if (v) {
	376	unsigned long m;
d41dee36 AW	377
d41dee36 AW	378	page = pfn_to_page(pfn);
1da177e4 LT	379	for (m = 1; m && i < idx; m<<=1, page++, i++) {
	380	if (v & m) {
	381	count++;
a226f6c8	382	__free_pages_bootmem(page, 0);
1da177e4 LT	383	}
	384	}
	385	} else {
f71bf0ca	386	i += BITS_PER_LONG;
1da177e4	387	}
d41dee36	388	pfn += BITS_PER_LONG;
1da177e4 LT	389	}
	390	total += count;
	391
	392	/*
	393	* Now free the allocator bitmap itself, it's not
	394	* needed anymore:
	395	*/
	396	page = virt_to_page(bdata->node_bootmem_map);
	397	count = 0;
bbc7b92e FBH	398	idx = (get_mapsize(bdata) + PAGE_SIZE-1) >> PAGE_SHIFT;
bbc7b92e FBH	399	for (i = 0; i < idx; i++, page++) {
a226f6c8	400	__free_pages_bootmem(page, 0);
bbc7b92e	401	count++;
1da177e4 LT	402	}
	403	total += count;
	404	bdata->node_bootmem_map = NULL;
	405
	406	return total;
	407	}
	408
f71bf0ca	409	unsigned long __init init_bootmem_node(pg_data_t *pgdat, unsigned long freepfn,
bb0923a6	410	unsigned long startpfn, unsigned long endpfn)
1da177e4	411	{
f71bf0ca	412	return init_bootmem_core(pgdat, freepfn, startpfn, endpfn);
1da177e4 LT	413	}
1da177e4 LT	414
f71bf0ca	415	void __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
72a7fe39	416	unsigned long size, int flags)
1da177e4	417	{
72a7fe39	418	reserve_bootmem_core(pgdat->bdata, physaddr, size, flags);
1da177e4 LT	419	}
1da177e4 LT	420
f71bf0ca FBH	421	void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
f71bf0ca FBH	422	unsigned long size)
1da177e4 LT	423	{
	424	free_bootmem_core(pgdat->bdata, physaddr, size);
	425	}
	426
f71bf0ca	427	unsigned long __init free_all_bootmem_node(pg_data_t *pgdat)
1da177e4	428	{
f71bf0ca	429	return free_all_bootmem_core(pgdat);
1da177e4 LT	430	}
1da177e4 LT	431
f71bf0ca	432	unsigned long __init init_bootmem(unsigned long start, unsigned long pages)
1da177e4 LT	433	{
	434	max_low_pfn = pages;
	435	min_low_pfn = start;
f71bf0ca	436	return init_bootmem_core(NODE_DATA(0), start, 0, pages);
1da177e4 LT	437	}
	438
	439	#ifndef CONFIG_HAVE_ARCH_BOOTMEM_NODE
72a7fe39 BW	440	int __init reserve_bootmem(unsigned long addr, unsigned long size,
72a7fe39 BW	441	int flags)
1da177e4	442	{
72a7fe39	443	return reserve_bootmem_core(NODE_DATA(0)->bdata, addr, size, flags);
1da177e4 LT	444	}
	445	#endif /* !CONFIG_HAVE_ARCH_BOOTMEM_NODE */
	446
f71bf0ca	447	void __init free_bootmem(unsigned long addr, unsigned long size)
1da177e4	448	{
5a982cbc YL	449	bootmem_data_t *bdata;
	450	list_for_each_entry(bdata, &bdata_list, list)
	451	free_bootmem_core(bdata, addr, size);
1da177e4 LT	452	}
1da177e4 LT	453
f71bf0ca	454	unsigned long __init free_all_bootmem(void)
1da177e4	455	{
f71bf0ca	456	return free_all_bootmem_core(NODE_DATA(0));
1da177e4 LT	457	}
1da177e4 LT	458
bb0923a6 FBH	459	void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align,
bb0923a6 FBH	460	unsigned long goal)
1da177e4	461	{
679bc9fb	462	bootmem_data_t *bdata;
1da177e4 LT	463	void *ptr;
1da177e4 LT	464
f71bf0ca FBH	465	list_for_each_entry(bdata, &bdata_list, list) {
	466	ptr = __alloc_bootmem_core(bdata, size, align, goal, 0);
	467	if (ptr)
	468	return ptr;
	469	}
a8062231 AK	470	return NULL;
a8062231 AK	471	}
1da177e4	472
bb0923a6 FBH	473	void * __init __alloc_bootmem(unsigned long size, unsigned long align,
bb0923a6 FBH	474	unsigned long goal)
a8062231 AK	475	{
a8062231 AK	476	void *mem = __alloc_bootmem_nopanic(size,align,goal);
f71bf0ca	477
a8062231 AK	478	if (mem)
a8062231 AK	479	return mem;
1da177e4 LT	480	/*
	481	* Whoops, we cannot satisfy the allocation request.
	482	*/
	483	printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size);
	484	panic("Out of memory");
	485	return NULL;
	486	}
	487
281dd25c	488
bb0923a6 FBH	489	void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
bb0923a6 FBH	490	unsigned long align, unsigned long goal)
1da177e4 LT	491	{
	492	void *ptr;
	493
008857c1	494	ptr = __alloc_bootmem_core(pgdat->bdata, size, align, goal, 0);
1da177e4	495	if (ptr)
f71bf0ca	496	return ptr;
1da177e4	497
008857c1	498	return __alloc_bootmem(size, align, goal);
1da177e4 LT	499	}
1da177e4 LT	500
dfd54cbc HC	501	#ifndef ARCH_LOW_ADDRESS_LIMIT
	502	#define ARCH_LOW_ADDRESS_LIMIT 0xffffffffUL
	503	#endif
008857c1	504
bb0923a6 FBH	505	void * __init __alloc_bootmem_low(unsigned long size, unsigned long align,
bb0923a6 FBH	506	unsigned long goal)
008857c1	507	{
679bc9fb	508	bootmem_data_t *bdata;
008857c1 RT	509	void *ptr;
008857c1 RT	510
f71bf0ca	511	list_for_each_entry(bdata, &bdata_list, list) {
dfd54cbc HC	512	ptr = __alloc_bootmem_core(bdata, size, align, goal,
dfd54cbc HC	513	ARCH_LOW_ADDRESS_LIMIT);
f71bf0ca FBH	514	if (ptr)
	515	return ptr;
	516	}
008857c1 RT	517
	518	/*
	519	* Whoops, we cannot satisfy the allocation request.
	520	*/
	521	printk(KERN_ALERT "low bootmem alloc of %lu bytes failed!\n", size);
	522	panic("Out of low memory");
	523	return NULL;
	524	}
	525
	526	void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size,
	527	unsigned long align, unsigned long goal)
	528	{
dfd54cbc HC	529	return __alloc_bootmem_core(pgdat->bdata, size, align, goal,
dfd54cbc HC	530	ARCH_LOW_ADDRESS_LIMIT);
008857c1	531	}