xref: /openbmc/linux/mm/memblock.c (revision e8f6f3b4)
1 /*
2  * Procedures for maintaining information about logical memory blocks.
3  *
4  * Peter Bergner, IBM Corp.	June 2001.
5  * Copyright (C) 2001 Peter Bergner.
6  *
7  *      This program is free software; you can redistribute it and/or
8  *      modify it under the terms of the GNU General Public License
9  *      as published by the Free Software Foundation; either version
10  *      2 of the License, or (at your option) any later version.
11  */
12 
13 #include <linux/kernel.h>
14 #include <linux/slab.h>
15 #include <linux/init.h>
16 #include <linux/bitops.h>
17 #include <linux/poison.h>
18 #include <linux/pfn.h>
19 #include <linux/debugfs.h>
20 #include <linux/seq_file.h>
21 #include <linux/memblock.h>
22 
23 #include <asm-generic/sections.h>
24 #include <linux/io.h>
25 
26 #include "internal.h"
27 
28 static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
29 static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
30 #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
31 static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock;
32 #endif
33 
34 struct memblock memblock __initdata_memblock = {
35 	.memory.regions		= memblock_memory_init_regions,
36 	.memory.cnt		= 1,	/* empty dummy entry */
37 	.memory.max		= INIT_MEMBLOCK_REGIONS,
38 
39 	.reserved.regions	= memblock_reserved_init_regions,
40 	.reserved.cnt		= 1,	/* empty dummy entry */
41 	.reserved.max		= INIT_MEMBLOCK_REGIONS,
42 
43 #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
44 	.physmem.regions	= memblock_physmem_init_regions,
45 	.physmem.cnt		= 1,	/* empty dummy entry */
46 	.physmem.max		= INIT_PHYSMEM_REGIONS,
47 #endif
48 
49 	.bottom_up		= false,
50 	.current_limit		= MEMBLOCK_ALLOC_ANYWHERE,
51 };
52 
53 int memblock_debug __initdata_memblock;
54 #ifdef CONFIG_MOVABLE_NODE
55 bool movable_node_enabled __initdata_memblock = false;
56 #endif
57 static int memblock_can_resize __initdata_memblock;
58 static int memblock_memory_in_slab __initdata_memblock = 0;
59 static int memblock_reserved_in_slab __initdata_memblock = 0;
60 
61 /* inline so we don't get a warning when pr_debug is compiled out */
62 static __init_memblock const char *
63 memblock_type_name(struct memblock_type *type)
64 {
65 	if (type == &memblock.memory)
66 		return "memory";
67 	else if (type == &memblock.reserved)
68 		return "reserved";
69 	else
70 		return "unknown";
71 }
72 
73 /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */
74 static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size)
75 {
76 	return *size = min(*size, (phys_addr_t)ULLONG_MAX - base);
77 }
78 
79 /*
80  * Address comparison utilities
81  */
82 static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1,
83 				       phys_addr_t base2, phys_addr_t size2)
84 {
85 	return ((base1 < (base2 + size2)) && (base2 < (base1 + size1)));
86 }
87 
88 static long __init_memblock memblock_overlaps_region(struct memblock_type *type,
89 					phys_addr_t base, phys_addr_t size)
90 {
91 	unsigned long i;
92 
93 	for (i = 0; i < type->cnt; i++) {
94 		phys_addr_t rgnbase = type->regions[i].base;
95 		phys_addr_t rgnsize = type->regions[i].size;
96 		if (memblock_addrs_overlap(base, size, rgnbase, rgnsize))
97 			break;
98 	}
99 
100 	return (i < type->cnt) ? i : -1;
101 }
102 
103 /*
104  * __memblock_find_range_bottom_up - find free area utility in bottom-up
105  * @start: start of candidate range
106  * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
107  * @size: size of free area to find
108  * @align: alignment of free area to find
109  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
110  *
111  * Utility called from memblock_find_in_range_node(), find free area bottom-up.
112  *
113  * RETURNS:
114  * Found address on success, 0 on failure.
115  */
116 static phys_addr_t __init_memblock
117 __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
118 				phys_addr_t size, phys_addr_t align, int nid)
119 {
120 	phys_addr_t this_start, this_end, cand;
121 	u64 i;
122 
123 	for_each_free_mem_range(i, nid, &this_start, &this_end, NULL) {
124 		this_start = clamp(this_start, start, end);
125 		this_end = clamp(this_end, start, end);
126 
127 		cand = round_up(this_start, align);
128 		if (cand < this_end && this_end - cand >= size)
129 			return cand;
130 	}
131 
132 	return 0;
133 }
134 
135 /**
136  * __memblock_find_range_top_down - find free area utility, in top-down
137  * @start: start of candidate range
138  * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
139  * @size: size of free area to find
140  * @align: alignment of free area to find
141  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
142  *
143  * Utility called from memblock_find_in_range_node(), find free area top-down.
144  *
145  * RETURNS:
146  * Found address on success, 0 on failure.
147  */
148 static phys_addr_t __init_memblock
149 __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
150 			       phys_addr_t size, phys_addr_t align, int nid)
151 {
152 	phys_addr_t this_start, this_end, cand;
153 	u64 i;
154 
155 	for_each_free_mem_range_reverse(i, nid, &this_start, &this_end, NULL) {
156 		this_start = clamp(this_start, start, end);
157 		this_end = clamp(this_end, start, end);
158 
159 		if (this_end < size)
160 			continue;
161 
162 		cand = round_down(this_end - size, align);
163 		if (cand >= this_start)
164 			return cand;
165 	}
166 
167 	return 0;
168 }
169 
170 /**
171  * memblock_find_in_range_node - find free area in given range and node
172  * @size: size of free area to find
173  * @align: alignment of free area to find
174  * @start: start of candidate range
175  * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
176  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
177  *
178  * Find @size free area aligned to @align in the specified range and node.
179  *
180  * When allocation direction is bottom-up, the @start should be greater
181  * than the end of the kernel image. Otherwise, it will be trimmed. The
182  * reason is that we want the bottom-up allocation just near the kernel
183  * image so it is highly likely that the allocated memory and the kernel
184  * will reside in the same node.
185  *
186  * If bottom-up allocation failed, will try to allocate memory top-down.
187  *
188  * RETURNS:
189  * Found address on success, 0 on failure.
190  */
191 phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
192 					phys_addr_t align, phys_addr_t start,
193 					phys_addr_t end, int nid)
194 {
195 	phys_addr_t kernel_end, ret;
196 
197 	/* pump up @end */
198 	if (end == MEMBLOCK_ALLOC_ACCESSIBLE)
199 		end = memblock.current_limit;
200 
201 	/* avoid allocating the first page */
202 	start = max_t(phys_addr_t, start, PAGE_SIZE);
203 	end = max(start, end);
204 	kernel_end = __pa_symbol(_end);
205 
206 	/*
207 	 * try bottom-up allocation only when bottom-up mode
208 	 * is set and @end is above the kernel image.
209 	 */
210 	if (memblock_bottom_up() && end > kernel_end) {
211 		phys_addr_t bottom_up_start;
212 
213 		/* make sure we will allocate above the kernel */
214 		bottom_up_start = max(start, kernel_end);
215 
216 		/* ok, try bottom-up allocation first */
217 		ret = __memblock_find_range_bottom_up(bottom_up_start, end,
218 						      size, align, nid);
219 		if (ret)
220 			return ret;
221 
222 		/*
223 		 * we always limit bottom-up allocation above the kernel,
224 		 * but top-down allocation doesn't have the limit, so
225 		 * retrying top-down allocation may succeed when bottom-up
226 		 * allocation failed.
227 		 *
228 		 * bottom-up allocation is expected to be fail very rarely,
229 		 * so we use WARN_ONCE() here to see the stack trace if
230 		 * fail happens.
231 		 */
232 		WARN_ONCE(1, "memblock: bottom-up allocation failed, "
233 			     "memory hotunplug may be affected\n");
234 	}
235 
236 	return __memblock_find_range_top_down(start, end, size, align, nid);
237 }
238 
239 /**
240  * memblock_find_in_range - find free area in given range
241  * @start: start of candidate range
242  * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
243  * @size: size of free area to find
244  * @align: alignment of free area to find
245  *
246  * Find @size free area aligned to @align in the specified range.
247  *
248  * RETURNS:
249  * Found address on success, 0 on failure.
250  */
251 phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
252 					phys_addr_t end, phys_addr_t size,
253 					phys_addr_t align)
254 {
255 	return memblock_find_in_range_node(size, align, start, end,
256 					    NUMA_NO_NODE);
257 }
258 
259 static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r)
260 {
261 	type->total_size -= type->regions[r].size;
262 	memmove(&type->regions[r], &type->regions[r + 1],
263 		(type->cnt - (r + 1)) * sizeof(type->regions[r]));
264 	type->cnt--;
265 
266 	/* Special case for empty arrays */
267 	if (type->cnt == 0) {
268 		WARN_ON(type->total_size != 0);
269 		type->cnt = 1;
270 		type->regions[0].base = 0;
271 		type->regions[0].size = 0;
272 		type->regions[0].flags = 0;
273 		memblock_set_region_node(&type->regions[0], MAX_NUMNODES);
274 	}
275 }
276 
277 #ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
278 
279 phys_addr_t __init_memblock get_allocated_memblock_reserved_regions_info(
280 					phys_addr_t *addr)
281 {
282 	if (memblock.reserved.regions == memblock_reserved_init_regions)
283 		return 0;
284 
285 	*addr = __pa(memblock.reserved.regions);
286 
287 	return PAGE_ALIGN(sizeof(struct memblock_region) *
288 			  memblock.reserved.max);
289 }
290 
291 phys_addr_t __init_memblock get_allocated_memblock_memory_regions_info(
292 					phys_addr_t *addr)
293 {
294 	if (memblock.memory.regions == memblock_memory_init_regions)
295 		return 0;
296 
297 	*addr = __pa(memblock.memory.regions);
298 
299 	return PAGE_ALIGN(sizeof(struct memblock_region) *
300 			  memblock.memory.max);
301 }
302 
303 #endif
304 
305 /**
306  * memblock_double_array - double the size of the memblock regions array
307  * @type: memblock type of the regions array being doubled
308  * @new_area_start: starting address of memory range to avoid overlap with
309  * @new_area_size: size of memory range to avoid overlap with
310  *
311  * Double the size of the @type regions array. If memblock is being used to
312  * allocate memory for a new reserved regions array and there is a previously
313  * allocated memory range [@new_area_start,@new_area_start+@new_area_size]
314  * waiting to be reserved, ensure the memory used by the new array does
315  * not overlap.
316  *
317  * RETURNS:
318  * 0 on success, -1 on failure.
319  */
320 static int __init_memblock memblock_double_array(struct memblock_type *type,
321 						phys_addr_t new_area_start,
322 						phys_addr_t new_area_size)
323 {
324 	struct memblock_region *new_array, *old_array;
325 	phys_addr_t old_alloc_size, new_alloc_size;
326 	phys_addr_t old_size, new_size, addr;
327 	int use_slab = slab_is_available();
328 	int *in_slab;
329 
330 	/* We don't allow resizing until we know about the reserved regions
331 	 * of memory that aren't suitable for allocation
332 	 */
333 	if (!memblock_can_resize)
334 		return -1;
335 
336 	/* Calculate new doubled size */
337 	old_size = type->max * sizeof(struct memblock_region);
338 	new_size = old_size << 1;
339 	/*
340 	 * We need to allocated new one align to PAGE_SIZE,
341 	 *   so we can free them completely later.
342 	 */
343 	old_alloc_size = PAGE_ALIGN(old_size);
344 	new_alloc_size = PAGE_ALIGN(new_size);
345 
346 	/* Retrieve the slab flag */
347 	if (type == &memblock.memory)
348 		in_slab = &memblock_memory_in_slab;
349 	else
350 		in_slab = &memblock_reserved_in_slab;
351 
352 	/* Try to find some space for it.
353 	 *
354 	 * WARNING: We assume that either slab_is_available() and we use it or
355 	 * we use MEMBLOCK for allocations. That means that this is unsafe to
356 	 * use when bootmem is currently active (unless bootmem itself is
357 	 * implemented on top of MEMBLOCK which isn't the case yet)
358 	 *
359 	 * This should however not be an issue for now, as we currently only
360 	 * call into MEMBLOCK while it's still active, or much later when slab
361 	 * is active for memory hotplug operations
362 	 */
363 	if (use_slab) {
364 		new_array = kmalloc(new_size, GFP_KERNEL);
365 		addr = new_array ? __pa(new_array) : 0;
366 	} else {
367 		/* only exclude range when trying to double reserved.regions */
368 		if (type != &memblock.reserved)
369 			new_area_start = new_area_size = 0;
370 
371 		addr = memblock_find_in_range(new_area_start + new_area_size,
372 						memblock.current_limit,
373 						new_alloc_size, PAGE_SIZE);
374 		if (!addr && new_area_size)
375 			addr = memblock_find_in_range(0,
376 				min(new_area_start, memblock.current_limit),
377 				new_alloc_size, PAGE_SIZE);
378 
379 		new_array = addr ? __va(addr) : NULL;
380 	}
381 	if (!addr) {
382 		pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n",
383 		       memblock_type_name(type), type->max, type->max * 2);
384 		return -1;
385 	}
386 
387 	memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]",
388 			memblock_type_name(type), type->max * 2, (u64)addr,
389 			(u64)addr + new_size - 1);
390 
391 	/*
392 	 * Found space, we now need to move the array over before we add the
393 	 * reserved region since it may be our reserved array itself that is
394 	 * full.
395 	 */
396 	memcpy(new_array, type->regions, old_size);
397 	memset(new_array + type->max, 0, old_size);
398 	old_array = type->regions;
399 	type->regions = new_array;
400 	type->max <<= 1;
401 
402 	/* Free old array. We needn't free it if the array is the static one */
403 	if (*in_slab)
404 		kfree(old_array);
405 	else if (old_array != memblock_memory_init_regions &&
406 		 old_array != memblock_reserved_init_regions)
407 		memblock_free(__pa(old_array), old_alloc_size);
408 
409 	/*
410 	 * Reserve the new array if that comes from the memblock.  Otherwise, we
411 	 * needn't do it
412 	 */
413 	if (!use_slab)
414 		BUG_ON(memblock_reserve(addr, new_alloc_size));
415 
416 	/* Update slab flag */
417 	*in_slab = use_slab;
418 
419 	return 0;
420 }
421 
422 /**
423  * memblock_merge_regions - merge neighboring compatible regions
424  * @type: memblock type to scan
425  *
426  * Scan @type and merge neighboring compatible regions.
427  */
428 static void __init_memblock memblock_merge_regions(struct memblock_type *type)
429 {
430 	int i = 0;
431 
432 	/* cnt never goes below 1 */
433 	while (i < type->cnt - 1) {
434 		struct memblock_region *this = &type->regions[i];
435 		struct memblock_region *next = &type->regions[i + 1];
436 
437 		if (this->base + this->size != next->base ||
438 		    memblock_get_region_node(this) !=
439 		    memblock_get_region_node(next) ||
440 		    this->flags != next->flags) {
441 			BUG_ON(this->base + this->size > next->base);
442 			i++;
443 			continue;
444 		}
445 
446 		this->size += next->size;
447 		/* move forward from next + 1, index of which is i + 2 */
448 		memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next));
449 		type->cnt--;
450 	}
451 }
452 
453 /**
454  * memblock_insert_region - insert new memblock region
455  * @type:	memblock type to insert into
456  * @idx:	index for the insertion point
457  * @base:	base address of the new region
458  * @size:	size of the new region
459  * @nid:	node id of the new region
460  * @flags:	flags of the new region
461  *
462  * Insert new memblock region [@base,@base+@size) into @type at @idx.
463  * @type must already have extra room to accomodate the new region.
464  */
465 static void __init_memblock memblock_insert_region(struct memblock_type *type,
466 						   int idx, phys_addr_t base,
467 						   phys_addr_t size,
468 						   int nid, unsigned long flags)
469 {
470 	struct memblock_region *rgn = &type->regions[idx];
471 
472 	BUG_ON(type->cnt >= type->max);
473 	memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn));
474 	rgn->base = base;
475 	rgn->size = size;
476 	rgn->flags = flags;
477 	memblock_set_region_node(rgn, nid);
478 	type->cnt++;
479 	type->total_size += size;
480 }
481 
482 /**
483  * memblock_add_range - add new memblock region
484  * @type: memblock type to add new region into
485  * @base: base address of the new region
486  * @size: size of the new region
487  * @nid: nid of the new region
488  * @flags: flags of the new region
489  *
490  * Add new memblock region [@base,@base+@size) into @type.  The new region
491  * is allowed to overlap with existing ones - overlaps don't affect already
492  * existing regions.  @type is guaranteed to be minimal (all neighbouring
493  * compatible regions are merged) after the addition.
494  *
495  * RETURNS:
496  * 0 on success, -errno on failure.
497  */
498 int __init_memblock memblock_add_range(struct memblock_type *type,
499 				phys_addr_t base, phys_addr_t size,
500 				int nid, unsigned long flags)
501 {
502 	bool insert = false;
503 	phys_addr_t obase = base;
504 	phys_addr_t end = base + memblock_cap_size(base, &size);
505 	int i, nr_new;
506 
507 	if (!size)
508 		return 0;
509 
510 	/* special case for empty array */
511 	if (type->regions[0].size == 0) {
512 		WARN_ON(type->cnt != 1 || type->total_size);
513 		type->regions[0].base = base;
514 		type->regions[0].size = size;
515 		type->regions[0].flags = flags;
516 		memblock_set_region_node(&type->regions[0], nid);
517 		type->total_size = size;
518 		return 0;
519 	}
520 repeat:
521 	/*
522 	 * The following is executed twice.  Once with %false @insert and
523 	 * then with %true.  The first counts the number of regions needed
524 	 * to accomodate the new area.  The second actually inserts them.
525 	 */
526 	base = obase;
527 	nr_new = 0;
528 
529 	for (i = 0; i < type->cnt; i++) {
530 		struct memblock_region *rgn = &type->regions[i];
531 		phys_addr_t rbase = rgn->base;
532 		phys_addr_t rend = rbase + rgn->size;
533 
534 		if (rbase >= end)
535 			break;
536 		if (rend <= base)
537 			continue;
538 		/*
539 		 * @rgn overlaps.  If it separates the lower part of new
540 		 * area, insert that portion.
541 		 */
542 		if (rbase > base) {
543 			nr_new++;
544 			if (insert)
545 				memblock_insert_region(type, i++, base,
546 						       rbase - base, nid,
547 						       flags);
548 		}
549 		/* area below @rend is dealt with, forget about it */
550 		base = min(rend, end);
551 	}
552 
553 	/* insert the remaining portion */
554 	if (base < end) {
555 		nr_new++;
556 		if (insert)
557 			memblock_insert_region(type, i, base, end - base,
558 					       nid, flags);
559 	}
560 
561 	/*
562 	 * If this was the first round, resize array and repeat for actual
563 	 * insertions; otherwise, merge and return.
564 	 */
565 	if (!insert) {
566 		while (type->cnt + nr_new > type->max)
567 			if (memblock_double_array(type, obase, size) < 0)
568 				return -ENOMEM;
569 		insert = true;
570 		goto repeat;
571 	} else {
572 		memblock_merge_regions(type);
573 		return 0;
574 	}
575 }
576 
577 int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size,
578 				       int nid)
579 {
580 	return memblock_add_range(&memblock.memory, base, size, nid, 0);
581 }
582 
583 int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
584 {
585 	return memblock_add_range(&memblock.memory, base, size,
586 				   MAX_NUMNODES, 0);
587 }
588 
589 /**
590  * memblock_isolate_range - isolate given range into disjoint memblocks
591  * @type: memblock type to isolate range for
592  * @base: base of range to isolate
593  * @size: size of range to isolate
594  * @start_rgn: out parameter for the start of isolated region
595  * @end_rgn: out parameter for the end of isolated region
596  *
597  * Walk @type and ensure that regions don't cross the boundaries defined by
598  * [@base,@base+@size).  Crossing regions are split at the boundaries,
599  * which may create at most two more regions.  The index of the first
600  * region inside the range is returned in *@start_rgn and end in *@end_rgn.
601  *
602  * RETURNS:
603  * 0 on success, -errno on failure.
604  */
605 static int __init_memblock memblock_isolate_range(struct memblock_type *type,
606 					phys_addr_t base, phys_addr_t size,
607 					int *start_rgn, int *end_rgn)
608 {
609 	phys_addr_t end = base + memblock_cap_size(base, &size);
610 	int i;
611 
612 	*start_rgn = *end_rgn = 0;
613 
614 	if (!size)
615 		return 0;
616 
617 	/* we'll create at most two more regions */
618 	while (type->cnt + 2 > type->max)
619 		if (memblock_double_array(type, base, size) < 0)
620 			return -ENOMEM;
621 
622 	for (i = 0; i < type->cnt; i++) {
623 		struct memblock_region *rgn = &type->regions[i];
624 		phys_addr_t rbase = rgn->base;
625 		phys_addr_t rend = rbase + rgn->size;
626 
627 		if (rbase >= end)
628 			break;
629 		if (rend <= base)
630 			continue;
631 
632 		if (rbase < base) {
633 			/*
634 			 * @rgn intersects from below.  Split and continue
635 			 * to process the next region - the new top half.
636 			 */
637 			rgn->base = base;
638 			rgn->size -= base - rbase;
639 			type->total_size -= base - rbase;
640 			memblock_insert_region(type, i, rbase, base - rbase,
641 					       memblock_get_region_node(rgn),
642 					       rgn->flags);
643 		} else if (rend > end) {
644 			/*
645 			 * @rgn intersects from above.  Split and redo the
646 			 * current region - the new bottom half.
647 			 */
648 			rgn->base = end;
649 			rgn->size -= end - rbase;
650 			type->total_size -= end - rbase;
651 			memblock_insert_region(type, i--, rbase, end - rbase,
652 					       memblock_get_region_node(rgn),
653 					       rgn->flags);
654 		} else {
655 			/* @rgn is fully contained, record it */
656 			if (!*end_rgn)
657 				*start_rgn = i;
658 			*end_rgn = i + 1;
659 		}
660 	}
661 
662 	return 0;
663 }
664 
665 int __init_memblock memblock_remove_range(struct memblock_type *type,
666 					  phys_addr_t base, phys_addr_t size)
667 {
668 	int start_rgn, end_rgn;
669 	int i, ret;
670 
671 	ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
672 	if (ret)
673 		return ret;
674 
675 	for (i = end_rgn - 1; i >= start_rgn; i--)
676 		memblock_remove_region(type, i);
677 	return 0;
678 }
679 
680 int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size)
681 {
682 	return memblock_remove_range(&memblock.memory, base, size);
683 }
684 
685 
686 int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size)
687 {
688 	memblock_dbg("   memblock_free: [%#016llx-%#016llx] %pF\n",
689 		     (unsigned long long)base,
690 		     (unsigned long long)base + size - 1,
691 		     (void *)_RET_IP_);
692 
693 	kmemleak_free_part(__va(base), size);
694 	return memblock_remove_range(&memblock.reserved, base, size);
695 }
696 
697 static int __init_memblock memblock_reserve_region(phys_addr_t base,
698 						   phys_addr_t size,
699 						   int nid,
700 						   unsigned long flags)
701 {
702 	struct memblock_type *_rgn = &memblock.reserved;
703 
704 	memblock_dbg("memblock_reserve: [%#016llx-%#016llx] flags %#02lx %pF\n",
705 		     (unsigned long long)base,
706 		     (unsigned long long)base + size - 1,
707 		     flags, (void *)_RET_IP_);
708 
709 	return memblock_add_range(_rgn, base, size, nid, flags);
710 }
711 
712 int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size)
713 {
714 	return memblock_reserve_region(base, size, MAX_NUMNODES, 0);
715 }
716 
717 /**
718  *
719  * This function isolates region [@base, @base + @size), and sets/clears flag
720  *
721  * Return 0 on succees, -errno on failure.
722  */
723 static int __init_memblock memblock_setclr_flag(phys_addr_t base,
724 				phys_addr_t size, int set, int flag)
725 {
726 	struct memblock_type *type = &memblock.memory;
727 	int i, ret, start_rgn, end_rgn;
728 
729 	ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
730 	if (ret)
731 		return ret;
732 
733 	for (i = start_rgn; i < end_rgn; i++)
734 		if (set)
735 			memblock_set_region_flags(&type->regions[i], flag);
736 		else
737 			memblock_clear_region_flags(&type->regions[i], flag);
738 
739 	memblock_merge_regions(type);
740 	return 0;
741 }
742 
743 /**
744  * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG.
745  * @base: the base phys addr of the region
746  * @size: the size of the region
747  *
748  * Return 0 on succees, -errno on failure.
749  */
750 int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size)
751 {
752 	return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG);
753 }
754 
755 /**
756  * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region.
757  * @base: the base phys addr of the region
758  * @size: the size of the region
759  *
760  * Return 0 on succees, -errno on failure.
761  */
762 int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size)
763 {
764 	return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG);
765 }
766 
767 /**
768  * __next__mem_range - next function for for_each_free_mem_range() etc.
769  * @idx: pointer to u64 loop variable
770  * @nid: node selector, %NUMA_NO_NODE for all nodes
771  * @type_a: pointer to memblock_type from where the range is taken
772  * @type_b: pointer to memblock_type which excludes memory from being taken
773  * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
774  * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
775  * @out_nid: ptr to int for nid of the range, can be %NULL
776  *
777  * Find the first area from *@idx which matches @nid, fill the out
778  * parameters, and update *@idx for the next iteration.  The lower 32bit of
779  * *@idx contains index into type_a and the upper 32bit indexes the
780  * areas before each region in type_b.	For example, if type_b regions
781  * look like the following,
782  *
783  *	0:[0-16), 1:[32-48), 2:[128-130)
784  *
785  * The upper 32bit indexes the following regions.
786  *
787  *	0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX)
788  *
789  * As both region arrays are sorted, the function advances the two indices
790  * in lockstep and returns each intersection.
791  */
792 void __init_memblock __next_mem_range(u64 *idx, int nid,
793 				      struct memblock_type *type_a,
794 				      struct memblock_type *type_b,
795 				      phys_addr_t *out_start,
796 				      phys_addr_t *out_end, int *out_nid)
797 {
798 	int idx_a = *idx & 0xffffffff;
799 	int idx_b = *idx >> 32;
800 
801 	if (WARN_ONCE(nid == MAX_NUMNODES,
802 	"Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
803 		nid = NUMA_NO_NODE;
804 
805 	for (; idx_a < type_a->cnt; idx_a++) {
806 		struct memblock_region *m = &type_a->regions[idx_a];
807 
808 		phys_addr_t m_start = m->base;
809 		phys_addr_t m_end = m->base + m->size;
810 		int	    m_nid = memblock_get_region_node(m);
811 
812 		/* only memory regions are associated with nodes, check it */
813 		if (nid != NUMA_NO_NODE && nid != m_nid)
814 			continue;
815 
816 		/* skip hotpluggable memory regions if needed */
817 		if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
818 			continue;
819 
820 		if (!type_b) {
821 			if (out_start)
822 				*out_start = m_start;
823 			if (out_end)
824 				*out_end = m_end;
825 			if (out_nid)
826 				*out_nid = m_nid;
827 			idx_a++;
828 			*idx = (u32)idx_a | (u64)idx_b << 32;
829 			return;
830 		}
831 
832 		/* scan areas before each reservation */
833 		for (; idx_b < type_b->cnt + 1; idx_b++) {
834 			struct memblock_region *r;
835 			phys_addr_t r_start;
836 			phys_addr_t r_end;
837 
838 			r = &type_b->regions[idx_b];
839 			r_start = idx_b ? r[-1].base + r[-1].size : 0;
840 			r_end = idx_b < type_b->cnt ?
841 				r->base : ULLONG_MAX;
842 
843 			/*
844 			 * if idx_b advanced past idx_a,
845 			 * break out to advance idx_a
846 			 */
847 			if (r_start >= m_end)
848 				break;
849 			/* if the two regions intersect, we're done */
850 			if (m_start < r_end) {
851 				if (out_start)
852 					*out_start =
853 						max(m_start, r_start);
854 				if (out_end)
855 					*out_end = min(m_end, r_end);
856 				if (out_nid)
857 					*out_nid = m_nid;
858 				/*
859 				 * The region which ends first is
860 				 * advanced for the next iteration.
861 				 */
862 				if (m_end <= r_end)
863 					idx_a++;
864 				else
865 					idx_b++;
866 				*idx = (u32)idx_a | (u64)idx_b << 32;
867 				return;
868 			}
869 		}
870 	}
871 
872 	/* signal end of iteration */
873 	*idx = ULLONG_MAX;
874 }
875 
876 /**
877  * __next_mem_range_rev - generic next function for for_each_*_range_rev()
878  *
879  * Finds the next range from type_a which is not marked as unsuitable
880  * in type_b.
881  *
882  * @idx: pointer to u64 loop variable
883  * @nid: nid: node selector, %NUMA_NO_NODE for all nodes
884  * @type_a: pointer to memblock_type from where the range is taken
885  * @type_b: pointer to memblock_type which excludes memory from being taken
886  * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
887  * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
888  * @out_nid: ptr to int for nid of the range, can be %NULL
889  *
890  * Reverse of __next_mem_range().
891  */
892 void __init_memblock __next_mem_range_rev(u64 *idx, int nid,
893 					  struct memblock_type *type_a,
894 					  struct memblock_type *type_b,
895 					  phys_addr_t *out_start,
896 					  phys_addr_t *out_end, int *out_nid)
897 {
898 	int idx_a = *idx & 0xffffffff;
899 	int idx_b = *idx >> 32;
900 
901 	if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
902 		nid = NUMA_NO_NODE;
903 
904 	if (*idx == (u64)ULLONG_MAX) {
905 		idx_a = type_a->cnt - 1;
906 		idx_b = type_b->cnt;
907 	}
908 
909 	for (; idx_a >= 0; idx_a--) {
910 		struct memblock_region *m = &type_a->regions[idx_a];
911 
912 		phys_addr_t m_start = m->base;
913 		phys_addr_t m_end = m->base + m->size;
914 		int m_nid = memblock_get_region_node(m);
915 
916 		/* only memory regions are associated with nodes, check it */
917 		if (nid != NUMA_NO_NODE && nid != m_nid)
918 			continue;
919 
920 		/* skip hotpluggable memory regions if needed */
921 		if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
922 			continue;
923 
924 		if (!type_b) {
925 			if (out_start)
926 				*out_start = m_start;
927 			if (out_end)
928 				*out_end = m_end;
929 			if (out_nid)
930 				*out_nid = m_nid;
931 			idx_a++;
932 			*idx = (u32)idx_a | (u64)idx_b << 32;
933 			return;
934 		}
935 
936 		/* scan areas before each reservation */
937 		for (; idx_b >= 0; idx_b--) {
938 			struct memblock_region *r;
939 			phys_addr_t r_start;
940 			phys_addr_t r_end;
941 
942 			r = &type_b->regions[idx_b];
943 			r_start = idx_b ? r[-1].base + r[-1].size : 0;
944 			r_end = idx_b < type_b->cnt ?
945 				r->base : ULLONG_MAX;
946 			/*
947 			 * if idx_b advanced past idx_a,
948 			 * break out to advance idx_a
949 			 */
950 
951 			if (r_end <= m_start)
952 				break;
953 			/* if the two regions intersect, we're done */
954 			if (m_end > r_start) {
955 				if (out_start)
956 					*out_start = max(m_start, r_start);
957 				if (out_end)
958 					*out_end = min(m_end, r_end);
959 				if (out_nid)
960 					*out_nid = m_nid;
961 				if (m_start >= r_start)
962 					idx_a--;
963 				else
964 					idx_b--;
965 				*idx = (u32)idx_a | (u64)idx_b << 32;
966 				return;
967 			}
968 		}
969 	}
970 	/* signal end of iteration */
971 	*idx = ULLONG_MAX;
972 }
973 
974 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
975 /*
976  * Common iterator interface used to define for_each_mem_range().
977  */
978 void __init_memblock __next_mem_pfn_range(int *idx, int nid,
979 				unsigned long *out_start_pfn,
980 				unsigned long *out_end_pfn, int *out_nid)
981 {
982 	struct memblock_type *type = &memblock.memory;
983 	struct memblock_region *r;
984 
985 	while (++*idx < type->cnt) {
986 		r = &type->regions[*idx];
987 
988 		if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size))
989 			continue;
990 		if (nid == MAX_NUMNODES || nid == r->nid)
991 			break;
992 	}
993 	if (*idx >= type->cnt) {
994 		*idx = -1;
995 		return;
996 	}
997 
998 	if (out_start_pfn)
999 		*out_start_pfn = PFN_UP(r->base);
1000 	if (out_end_pfn)
1001 		*out_end_pfn = PFN_DOWN(r->base + r->size);
1002 	if (out_nid)
1003 		*out_nid = r->nid;
1004 }
1005 
1006 /**
1007  * memblock_set_node - set node ID on memblock regions
1008  * @base: base of area to set node ID for
1009  * @size: size of area to set node ID for
1010  * @type: memblock type to set node ID for
1011  * @nid: node ID to set
1012  *
1013  * Set the nid of memblock @type regions in [@base,@base+@size) to @nid.
1014  * Regions which cross the area boundaries are split as necessary.
1015  *
1016  * RETURNS:
1017  * 0 on success, -errno on failure.
1018  */
1019 int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size,
1020 				      struct memblock_type *type, int nid)
1021 {
1022 	int start_rgn, end_rgn;
1023 	int i, ret;
1024 
1025 	ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
1026 	if (ret)
1027 		return ret;
1028 
1029 	for (i = start_rgn; i < end_rgn; i++)
1030 		memblock_set_region_node(&type->regions[i], nid);
1031 
1032 	memblock_merge_regions(type);
1033 	return 0;
1034 }
1035 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1036 
1037 static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
1038 					phys_addr_t align, phys_addr_t start,
1039 					phys_addr_t end, int nid)
1040 {
1041 	phys_addr_t found;
1042 
1043 	if (!align)
1044 		align = SMP_CACHE_BYTES;
1045 
1046 	found = memblock_find_in_range_node(size, align, start, end, nid);
1047 	if (found && !memblock_reserve(found, size)) {
1048 		/*
1049 		 * The min_count is set to 0 so that memblock allocations are
1050 		 * never reported as leaks.
1051 		 */
1052 		kmemleak_alloc(__va(found), size, 0, 0);
1053 		return found;
1054 	}
1055 	return 0;
1056 }
1057 
1058 phys_addr_t __init memblock_alloc_range(phys_addr_t size, phys_addr_t align,
1059 					phys_addr_t start, phys_addr_t end)
1060 {
1061 	return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE);
1062 }
1063 
1064 static phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size,
1065 					phys_addr_t align, phys_addr_t max_addr,
1066 					int nid)
1067 {
1068 	return memblock_alloc_range_nid(size, align, 0, max_addr, nid);
1069 }
1070 
1071 phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid)
1072 {
1073 	return memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, nid);
1074 }
1075 
1076 phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
1077 {
1078 	return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE);
1079 }
1080 
1081 phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
1082 {
1083 	phys_addr_t alloc;
1084 
1085 	alloc = __memblock_alloc_base(size, align, max_addr);
1086 
1087 	if (alloc == 0)
1088 		panic("ERROR: Failed to allocate 0x%llx bytes below 0x%llx.\n",
1089 		      (unsigned long long) size, (unsigned long long) max_addr);
1090 
1091 	return alloc;
1092 }
1093 
1094 phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align)
1095 {
1096 	return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
1097 }
1098 
1099 phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid)
1100 {
1101 	phys_addr_t res = memblock_alloc_nid(size, align, nid);
1102 
1103 	if (res)
1104 		return res;
1105 	return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
1106 }
1107 
1108 /**
1109  * memblock_virt_alloc_internal - allocate boot memory block
1110  * @size: size of memory block to be allocated in bytes
1111  * @align: alignment of the region and block's size
1112  * @min_addr: the lower bound of the memory region to allocate (phys address)
1113  * @max_addr: the upper bound of the memory region to allocate (phys address)
1114  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1115  *
1116  * The @min_addr limit is dropped if it can not be satisfied and the allocation
1117  * will fall back to memory below @min_addr. Also, allocation may fall back
1118  * to any node in the system if the specified node can not
1119  * hold the requested memory.
1120  *
1121  * The allocation is performed from memory region limited by
1122  * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE.
1123  *
1124  * The memory block is aligned on SMP_CACHE_BYTES if @align == 0.
1125  *
1126  * The phys address of allocated boot memory block is converted to virtual and
1127  * allocated memory is reset to 0.
1128  *
1129  * In addition, function sets the min_count to 0 using kmemleak_alloc for
1130  * allocated boot memory block, so that it is never reported as leaks.
1131  *
1132  * RETURNS:
1133  * Virtual address of allocated memory block on success, NULL on failure.
1134  */
1135 static void * __init memblock_virt_alloc_internal(
1136 				phys_addr_t size, phys_addr_t align,
1137 				phys_addr_t min_addr, phys_addr_t max_addr,
1138 				int nid)
1139 {
1140 	phys_addr_t alloc;
1141 	void *ptr;
1142 
1143 	if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
1144 		nid = NUMA_NO_NODE;
1145 
1146 	/*
1147 	 * Detect any accidental use of these APIs after slab is ready, as at
1148 	 * this moment memblock may be deinitialized already and its
1149 	 * internal data may be destroyed (after execution of free_all_bootmem)
1150 	 */
1151 	if (WARN_ON_ONCE(slab_is_available()))
1152 		return kzalloc_node(size, GFP_NOWAIT, nid);
1153 
1154 	if (!align)
1155 		align = SMP_CACHE_BYTES;
1156 
1157 	if (max_addr > memblock.current_limit)
1158 		max_addr = memblock.current_limit;
1159 
1160 again:
1161 	alloc = memblock_find_in_range_node(size, align, min_addr, max_addr,
1162 					    nid);
1163 	if (alloc)
1164 		goto done;
1165 
1166 	if (nid != NUMA_NO_NODE) {
1167 		alloc = memblock_find_in_range_node(size, align, min_addr,
1168 						    max_addr,  NUMA_NO_NODE);
1169 		if (alloc)
1170 			goto done;
1171 	}
1172 
1173 	if (min_addr) {
1174 		min_addr = 0;
1175 		goto again;
1176 	} else {
1177 		goto error;
1178 	}
1179 
1180 done:
1181 	memblock_reserve(alloc, size);
1182 	ptr = phys_to_virt(alloc);
1183 	memset(ptr, 0, size);
1184 
1185 	/*
1186 	 * The min_count is set to 0 so that bootmem allocated blocks
1187 	 * are never reported as leaks. This is because many of these blocks
1188 	 * are only referred via the physical address which is not
1189 	 * looked up by kmemleak.
1190 	 */
1191 	kmemleak_alloc(ptr, size, 0, 0);
1192 
1193 	return ptr;
1194 
1195 error:
1196 	return NULL;
1197 }
1198 
1199 /**
1200  * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block
1201  * @size: size of memory block to be allocated in bytes
1202  * @align: alignment of the region and block's size
1203  * @min_addr: the lower bound of the memory region from where the allocation
1204  *	  is preferred (phys address)
1205  * @max_addr: the upper bound of the memory region from where the allocation
1206  *	      is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1207  *	      allocate only from memory limited by memblock.current_limit value
1208  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1209  *
1210  * Public version of _memblock_virt_alloc_try_nid_nopanic() which provides
1211  * additional debug information (including caller info), if enabled.
1212  *
1213  * RETURNS:
1214  * Virtual address of allocated memory block on success, NULL on failure.
1215  */
1216 void * __init memblock_virt_alloc_try_nid_nopanic(
1217 				phys_addr_t size, phys_addr_t align,
1218 				phys_addr_t min_addr, phys_addr_t max_addr,
1219 				int nid)
1220 {
1221 	memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1222 		     __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1223 		     (u64)max_addr, (void *)_RET_IP_);
1224 	return memblock_virt_alloc_internal(size, align, min_addr,
1225 					     max_addr, nid);
1226 }
1227 
1228 /**
1229  * memblock_virt_alloc_try_nid - allocate boot memory block with panicking
1230  * @size: size of memory block to be allocated in bytes
1231  * @align: alignment of the region and block's size
1232  * @min_addr: the lower bound of the memory region from where the allocation
1233  *	  is preferred (phys address)
1234  * @max_addr: the upper bound of the memory region from where the allocation
1235  *	      is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1236  *	      allocate only from memory limited by memblock.current_limit value
1237  * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1238  *
1239  * Public panicking version of _memblock_virt_alloc_try_nid_nopanic()
1240  * which provides debug information (including caller info), if enabled,
1241  * and panics if the request can not be satisfied.
1242  *
1243  * RETURNS:
1244  * Virtual address of allocated memory block on success, NULL on failure.
1245  */
1246 void * __init memblock_virt_alloc_try_nid(
1247 			phys_addr_t size, phys_addr_t align,
1248 			phys_addr_t min_addr, phys_addr_t max_addr,
1249 			int nid)
1250 {
1251 	void *ptr;
1252 
1253 	memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1254 		     __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1255 		     (u64)max_addr, (void *)_RET_IP_);
1256 	ptr = memblock_virt_alloc_internal(size, align,
1257 					   min_addr, max_addr, nid);
1258 	if (ptr)
1259 		return ptr;
1260 
1261 	panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n",
1262 	      __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1263 	      (u64)max_addr);
1264 	return NULL;
1265 }
1266 
1267 /**
1268  * __memblock_free_early - free boot memory block
1269  * @base: phys starting address of the  boot memory block
1270  * @size: size of the boot memory block in bytes
1271  *
1272  * Free boot memory block previously allocated by memblock_virt_alloc_xx() API.
1273  * The freeing memory will not be released to the buddy allocator.
1274  */
1275 void __init __memblock_free_early(phys_addr_t base, phys_addr_t size)
1276 {
1277 	memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1278 		     __func__, (u64)base, (u64)base + size - 1,
1279 		     (void *)_RET_IP_);
1280 	kmemleak_free_part(__va(base), size);
1281 	memblock_remove_range(&memblock.reserved, base, size);
1282 }
1283 
1284 /*
1285  * __memblock_free_late - free bootmem block pages directly to buddy allocator
1286  * @addr: phys starting address of the  boot memory block
1287  * @size: size of the boot memory block in bytes
1288  *
1289  * This is only useful when the bootmem allocator has already been torn
1290  * down, but we are still initializing the system.  Pages are released directly
1291  * to the buddy allocator, no bootmem metadata is updated because it is gone.
1292  */
1293 void __init __memblock_free_late(phys_addr_t base, phys_addr_t size)
1294 {
1295 	u64 cursor, end;
1296 
1297 	memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1298 		     __func__, (u64)base, (u64)base + size - 1,
1299 		     (void *)_RET_IP_);
1300 	kmemleak_free_part(__va(base), size);
1301 	cursor = PFN_UP(base);
1302 	end = PFN_DOWN(base + size);
1303 
1304 	for (; cursor < end; cursor++) {
1305 		__free_pages_bootmem(pfn_to_page(cursor), 0);
1306 		totalram_pages++;
1307 	}
1308 }
1309 
1310 /*
1311  * Remaining API functions
1312  */
1313 
1314 phys_addr_t __init memblock_phys_mem_size(void)
1315 {
1316 	return memblock.memory.total_size;
1317 }
1318 
1319 phys_addr_t __init memblock_mem_size(unsigned long limit_pfn)
1320 {
1321 	unsigned long pages = 0;
1322 	struct memblock_region *r;
1323 	unsigned long start_pfn, end_pfn;
1324 
1325 	for_each_memblock(memory, r) {
1326 		start_pfn = memblock_region_memory_base_pfn(r);
1327 		end_pfn = memblock_region_memory_end_pfn(r);
1328 		start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
1329 		end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
1330 		pages += end_pfn - start_pfn;
1331 	}
1332 
1333 	return PFN_PHYS(pages);
1334 }
1335 
1336 /* lowest address */
1337 phys_addr_t __init_memblock memblock_start_of_DRAM(void)
1338 {
1339 	return memblock.memory.regions[0].base;
1340 }
1341 
1342 phys_addr_t __init_memblock memblock_end_of_DRAM(void)
1343 {
1344 	int idx = memblock.memory.cnt - 1;
1345 
1346 	return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size);
1347 }
1348 
1349 void __init memblock_enforce_memory_limit(phys_addr_t limit)
1350 {
1351 	phys_addr_t max_addr = (phys_addr_t)ULLONG_MAX;
1352 	struct memblock_region *r;
1353 
1354 	if (!limit)
1355 		return;
1356 
1357 	/* find out max address */
1358 	for_each_memblock(memory, r) {
1359 		if (limit <= r->size) {
1360 			max_addr = r->base + limit;
1361 			break;
1362 		}
1363 		limit -= r->size;
1364 	}
1365 
1366 	/* truncate both memory and reserved regions */
1367 	memblock_remove_range(&memblock.memory, max_addr,
1368 			      (phys_addr_t)ULLONG_MAX);
1369 	memblock_remove_range(&memblock.reserved, max_addr,
1370 			      (phys_addr_t)ULLONG_MAX);
1371 }
1372 
1373 static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr)
1374 {
1375 	unsigned int left = 0, right = type->cnt;
1376 
1377 	do {
1378 		unsigned int mid = (right + left) / 2;
1379 
1380 		if (addr < type->regions[mid].base)
1381 			right = mid;
1382 		else if (addr >= (type->regions[mid].base +
1383 				  type->regions[mid].size))
1384 			left = mid + 1;
1385 		else
1386 			return mid;
1387 	} while (left < right);
1388 	return -1;
1389 }
1390 
1391 int __init memblock_is_reserved(phys_addr_t addr)
1392 {
1393 	return memblock_search(&memblock.reserved, addr) != -1;
1394 }
1395 
1396 int __init_memblock memblock_is_memory(phys_addr_t addr)
1397 {
1398 	return memblock_search(&memblock.memory, addr) != -1;
1399 }
1400 
1401 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1402 int __init_memblock memblock_search_pfn_nid(unsigned long pfn,
1403 			 unsigned long *start_pfn, unsigned long *end_pfn)
1404 {
1405 	struct memblock_type *type = &memblock.memory;
1406 	int mid = memblock_search(type, PFN_PHYS(pfn));
1407 
1408 	if (mid == -1)
1409 		return -1;
1410 
1411 	*start_pfn = PFN_DOWN(type->regions[mid].base);
1412 	*end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size);
1413 
1414 	return type->regions[mid].nid;
1415 }
1416 #endif
1417 
1418 /**
1419  * memblock_is_region_memory - check if a region is a subset of memory
1420  * @base: base of region to check
1421  * @size: size of region to check
1422  *
1423  * Check if the region [@base, @base+@size) is a subset of a memory block.
1424  *
1425  * RETURNS:
1426  * 0 if false, non-zero if true
1427  */
1428 int __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size)
1429 {
1430 	int idx = memblock_search(&memblock.memory, base);
1431 	phys_addr_t end = base + memblock_cap_size(base, &size);
1432 
1433 	if (idx == -1)
1434 		return 0;
1435 	return memblock.memory.regions[idx].base <= base &&
1436 		(memblock.memory.regions[idx].base +
1437 		 memblock.memory.regions[idx].size) >= end;
1438 }
1439 
1440 /**
1441  * memblock_is_region_reserved - check if a region intersects reserved memory
1442  * @base: base of region to check
1443  * @size: size of region to check
1444  *
1445  * Check if the region [@base, @base+@size) intersects a reserved memory block.
1446  *
1447  * RETURNS:
1448  * 0 if false, non-zero if true
1449  */
1450 int __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size)
1451 {
1452 	memblock_cap_size(base, &size);
1453 	return memblock_overlaps_region(&memblock.reserved, base, size) >= 0;
1454 }
1455 
1456 void __init_memblock memblock_trim_memory(phys_addr_t align)
1457 {
1458 	phys_addr_t start, end, orig_start, orig_end;
1459 	struct memblock_region *r;
1460 
1461 	for_each_memblock(memory, r) {
1462 		orig_start = r->base;
1463 		orig_end = r->base + r->size;
1464 		start = round_up(orig_start, align);
1465 		end = round_down(orig_end, align);
1466 
1467 		if (start == orig_start && end == orig_end)
1468 			continue;
1469 
1470 		if (start < end) {
1471 			r->base = start;
1472 			r->size = end - start;
1473 		} else {
1474 			memblock_remove_region(&memblock.memory,
1475 					       r - memblock.memory.regions);
1476 			r--;
1477 		}
1478 	}
1479 }
1480 
1481 void __init_memblock memblock_set_current_limit(phys_addr_t limit)
1482 {
1483 	memblock.current_limit = limit;
1484 }
1485 
1486 phys_addr_t __init_memblock memblock_get_current_limit(void)
1487 {
1488 	return memblock.current_limit;
1489 }
1490 
1491 static void __init_memblock memblock_dump(struct memblock_type *type, char *name)
1492 {
1493 	unsigned long long base, size;
1494 	unsigned long flags;
1495 	int i;
1496 
1497 	pr_info(" %s.cnt  = 0x%lx\n", name, type->cnt);
1498 
1499 	for (i = 0; i < type->cnt; i++) {
1500 		struct memblock_region *rgn = &type->regions[i];
1501 		char nid_buf[32] = "";
1502 
1503 		base = rgn->base;
1504 		size = rgn->size;
1505 		flags = rgn->flags;
1506 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1507 		if (memblock_get_region_node(rgn) != MAX_NUMNODES)
1508 			snprintf(nid_buf, sizeof(nid_buf), " on node %d",
1509 				 memblock_get_region_node(rgn));
1510 #endif
1511 		pr_info(" %s[%#x]\t[%#016llx-%#016llx], %#llx bytes%s flags: %#lx\n",
1512 			name, i, base, base + size - 1, size, nid_buf, flags);
1513 	}
1514 }
1515 
1516 void __init_memblock __memblock_dump_all(void)
1517 {
1518 	pr_info("MEMBLOCK configuration:\n");
1519 	pr_info(" memory size = %#llx reserved size = %#llx\n",
1520 		(unsigned long long)memblock.memory.total_size,
1521 		(unsigned long long)memblock.reserved.total_size);
1522 
1523 	memblock_dump(&memblock.memory, "memory");
1524 	memblock_dump(&memblock.reserved, "reserved");
1525 }
1526 
1527 void __init memblock_allow_resize(void)
1528 {
1529 	memblock_can_resize = 1;
1530 }
1531 
1532 static int __init early_memblock(char *p)
1533 {
1534 	if (p && strstr(p, "debug"))
1535 		memblock_debug = 1;
1536 	return 0;
1537 }
1538 early_param("memblock", early_memblock);
1539 
1540 #if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK)
1541 
1542 static int memblock_debug_show(struct seq_file *m, void *private)
1543 {
1544 	struct memblock_type *type = m->private;
1545 	struct memblock_region *reg;
1546 	int i;
1547 
1548 	for (i = 0; i < type->cnt; i++) {
1549 		reg = &type->regions[i];
1550 		seq_printf(m, "%4d: ", i);
1551 		if (sizeof(phys_addr_t) == 4)
1552 			seq_printf(m, "0x%08lx..0x%08lx\n",
1553 				   (unsigned long)reg->base,
1554 				   (unsigned long)(reg->base + reg->size - 1));
1555 		else
1556 			seq_printf(m, "0x%016llx..0x%016llx\n",
1557 				   (unsigned long long)reg->base,
1558 				   (unsigned long long)(reg->base + reg->size - 1));
1559 
1560 	}
1561 	return 0;
1562 }
1563 
1564 static int memblock_debug_open(struct inode *inode, struct file *file)
1565 {
1566 	return single_open(file, memblock_debug_show, inode->i_private);
1567 }
1568 
1569 static const struct file_operations memblock_debug_fops = {
1570 	.open = memblock_debug_open,
1571 	.read = seq_read,
1572 	.llseek = seq_lseek,
1573 	.release = single_release,
1574 };
1575 
1576 static int __init memblock_init_debugfs(void)
1577 {
1578 	struct dentry *root = debugfs_create_dir("memblock", NULL);
1579 	if (!root)
1580 		return -ENXIO;
1581 	debugfs_create_file("memory", S_IRUGO, root, &memblock.memory, &memblock_debug_fops);
1582 	debugfs_create_file("reserved", S_IRUGO, root, &memblock.reserved, &memblock_debug_fops);
1583 #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
1584 	debugfs_create_file("physmem", S_IRUGO, root, &memblock.physmem, &memblock_debug_fops);
1585 #endif
1586 
1587 	return 0;
1588 }
1589 __initcall(memblock_init_debugfs);
1590 
1591 #endif /* CONFIG_DEBUG_FS */
1592