x86: Move back find_e820_area to e820.c
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / arch / x86 / kernel / e820.c
blobb4e512b03aa710febe6542424ba8d99accf2002a
1 /*
2 * Handle the memory map.
3 * The functions here do the job until bootmem takes over.
5 * Getting sanitize_e820_map() in sync with i386 version by applying change:
6 * - Provisions for empty E820 memory regions (reported by certain BIOSes).
7 * Alex Achenbach <xela@slit.de>, December 2002.
8 * Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
11 #include <linux/kernel.h>
12 #include <linux/types.h>
13 #include <linux/init.h>
14 #include <linux/bootmem.h>
15 #include <linux/pfn.h>
16 #include <linux/suspend.h>
17 #include <linux/firmware-map.h>
19 #include <asm/e820.h>
20 #include <asm/early_res.h>
21 #include <asm/proto.h>
22 #include <asm/setup.h>
25 * The e820 map is the map that gets modified e.g. with command line parameters
26 * and that is also registered with modifications in the kernel resource tree
27 * with the iomem_resource as parent.
29 * The e820_saved is directly saved after the BIOS-provided memory map is
30 * copied. It doesn't get modified afterwards. It's registered for the
31 * /sys/firmware/memmap interface.
33 * That memory map is not modified and is used as base for kexec. The kexec'd
34 * kernel should get the same memory map as the firmware provides. Then the
35 * user can e.g. boot the original kernel with mem=1G while still booting the
36 * next kernel with full memory.
38 struct e820map e820;
39 struct e820map e820_saved;
41 /* For PCI or other memory-mapped resources */
42 unsigned long pci_mem_start = 0xaeedbabe;
43 #ifdef CONFIG_PCI
44 EXPORT_SYMBOL(pci_mem_start);
45 #endif
48 * This function checks if any part of the range <start,end> is mapped
49 * with type.
51 int
52 e820_any_mapped(u64 start, u64 end, unsigned type)
54 int i;
56 for (i = 0; i < e820.nr_map; i++) {
57 struct e820entry *ei = &e820.map[i];
59 if (type && ei->type != type)
60 continue;
61 if (ei->addr >= end || ei->addr + ei->size <= start)
62 continue;
63 return 1;
65 return 0;
67 EXPORT_SYMBOL_GPL(e820_any_mapped);
70 * This function checks if the entire range <start,end> is mapped with type.
72 * Note: this function only works correct if the e820 table is sorted and
73 * not-overlapping, which is the case
75 int __init e820_all_mapped(u64 start, u64 end, unsigned type)
77 int i;
79 for (i = 0; i < e820.nr_map; i++) {
80 struct e820entry *ei = &e820.map[i];
82 if (type && ei->type != type)
83 continue;
84 /* is the region (part) in overlap with the current region ?*/
85 if (ei->addr >= end || ei->addr + ei->size <= start)
86 continue;
88 /* if the region is at the beginning of <start,end> we move
89 * start to the end of the region since it's ok until there
91 if (ei->addr <= start)
92 start = ei->addr + ei->size;
94 * if start is now at or beyond end, we're done, full
95 * coverage
97 if (start >= end)
98 return 1;
100 return 0;
104 * Add a memory region to the kernel e820 map.
106 static void __init __e820_add_region(struct e820map *e820x, u64 start, u64 size,
107 int type)
109 int x = e820x->nr_map;
111 if (x >= ARRAY_SIZE(e820x->map)) {
112 printk(KERN_ERR "Ooops! Too many entries in the memory map!\n");
113 return;
116 e820x->map[x].addr = start;
117 e820x->map[x].size = size;
118 e820x->map[x].type = type;
119 e820x->nr_map++;
122 void __init e820_add_region(u64 start, u64 size, int type)
124 __e820_add_region(&e820, start, size, type);
127 static void __init e820_print_type(u32 type)
129 switch (type) {
130 case E820_RAM:
131 case E820_RESERVED_KERN:
132 printk(KERN_CONT "(usable)");
133 break;
134 case E820_RESERVED:
135 printk(KERN_CONT "(reserved)");
136 break;
137 case E820_ACPI:
138 printk(KERN_CONT "(ACPI data)");
139 break;
140 case E820_NVS:
141 printk(KERN_CONT "(ACPI NVS)");
142 break;
143 case E820_UNUSABLE:
144 printk(KERN_CONT "(unusable)");
145 break;
146 default:
147 printk(KERN_CONT "type %u", type);
148 break;
152 void __init e820_print_map(char *who)
154 int i;
156 for (i = 0; i < e820.nr_map; i++) {
157 printk(KERN_INFO " %s: %016Lx - %016Lx ", who,
158 (unsigned long long) e820.map[i].addr,
159 (unsigned long long)
160 (e820.map[i].addr + e820.map[i].size));
161 e820_print_type(e820.map[i].type);
162 printk(KERN_CONT "\n");
167 * Sanitize the BIOS e820 map.
169 * Some e820 responses include overlapping entries. The following
170 * replaces the original e820 map with a new one, removing overlaps,
171 * and resolving conflicting memory types in favor of highest
172 * numbered type.
174 * The input parameter biosmap points to an array of 'struct
175 * e820entry' which on entry has elements in the range [0, *pnr_map)
176 * valid, and which has space for up to max_nr_map entries.
177 * On return, the resulting sanitized e820 map entries will be in
178 * overwritten in the same location, starting at biosmap.
180 * The integer pointed to by pnr_map must be valid on entry (the
181 * current number of valid entries located at biosmap) and will
182 * be updated on return, with the new number of valid entries
183 * (something no more than max_nr_map.)
185 * The return value from sanitize_e820_map() is zero if it
186 * successfully 'sanitized' the map entries passed in, and is -1
187 * if it did nothing, which can happen if either of (1) it was
188 * only passed one map entry, or (2) any of the input map entries
189 * were invalid (start + size < start, meaning that the size was
190 * so big the described memory range wrapped around through zero.)
192 * Visually we're performing the following
193 * (1,2,3,4 = memory types)...
195 * Sample memory map (w/overlaps):
196 * ____22__________________
197 * ______________________4_
198 * ____1111________________
199 * _44_____________________
200 * 11111111________________
201 * ____________________33__
202 * ___________44___________
203 * __________33333_________
204 * ______________22________
205 * ___________________2222_
206 * _________111111111______
207 * _____________________11_
208 * _________________4______
210 * Sanitized equivalent (no overlap):
211 * 1_______________________
212 * _44_____________________
213 * ___1____________________
214 * ____22__________________
215 * ______11________________
216 * _________1______________
217 * __________3_____________
218 * ___________44___________
219 * _____________33_________
220 * _______________2________
221 * ________________1_______
222 * _________________4______
223 * ___________________2____
224 * ____________________33__
225 * ______________________4_
228 int __init sanitize_e820_map(struct e820entry *biosmap, int max_nr_map,
229 u32 *pnr_map)
231 struct change_member {
232 struct e820entry *pbios; /* pointer to original bios entry */
233 unsigned long long addr; /* address for this change point */
235 static struct change_member change_point_list[2*E820_X_MAX] __initdata;
236 static struct change_member *change_point[2*E820_X_MAX] __initdata;
237 static struct e820entry *overlap_list[E820_X_MAX] __initdata;
238 static struct e820entry new_bios[E820_X_MAX] __initdata;
239 struct change_member *change_tmp;
240 unsigned long current_type, last_type;
241 unsigned long long last_addr;
242 int chgidx, still_changing;
243 int overlap_entries;
244 int new_bios_entry;
245 int old_nr, new_nr, chg_nr;
246 int i;
248 /* if there's only one memory region, don't bother */
249 if (*pnr_map < 2)
250 return -1;
252 old_nr = *pnr_map;
253 BUG_ON(old_nr > max_nr_map);
255 /* bail out if we find any unreasonable addresses in bios map */
256 for (i = 0; i < old_nr; i++)
257 if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr)
258 return -1;
260 /* create pointers for initial change-point information (for sorting) */
261 for (i = 0; i < 2 * old_nr; i++)
262 change_point[i] = &change_point_list[i];
264 /* record all known change-points (starting and ending addresses),
265 omitting those that are for empty memory regions */
266 chgidx = 0;
267 for (i = 0; i < old_nr; i++) {
268 if (biosmap[i].size != 0) {
269 change_point[chgidx]->addr = biosmap[i].addr;
270 change_point[chgidx++]->pbios = &biosmap[i];
271 change_point[chgidx]->addr = biosmap[i].addr +
272 biosmap[i].size;
273 change_point[chgidx++]->pbios = &biosmap[i];
276 chg_nr = chgidx;
278 /* sort change-point list by memory addresses (low -> high) */
279 still_changing = 1;
280 while (still_changing) {
281 still_changing = 0;
282 for (i = 1; i < chg_nr; i++) {
283 unsigned long long curaddr, lastaddr;
284 unsigned long long curpbaddr, lastpbaddr;
286 curaddr = change_point[i]->addr;
287 lastaddr = change_point[i - 1]->addr;
288 curpbaddr = change_point[i]->pbios->addr;
289 lastpbaddr = change_point[i - 1]->pbios->addr;
292 * swap entries, when:
294 * curaddr > lastaddr or
295 * curaddr == lastaddr and curaddr == curpbaddr and
296 * lastaddr != lastpbaddr
298 if (curaddr < lastaddr ||
299 (curaddr == lastaddr && curaddr == curpbaddr &&
300 lastaddr != lastpbaddr)) {
301 change_tmp = change_point[i];
302 change_point[i] = change_point[i-1];
303 change_point[i-1] = change_tmp;
304 still_changing = 1;
309 /* create a new bios memory map, removing overlaps */
310 overlap_entries = 0; /* number of entries in the overlap table */
311 new_bios_entry = 0; /* index for creating new bios map entries */
312 last_type = 0; /* start with undefined memory type */
313 last_addr = 0; /* start with 0 as last starting address */
315 /* loop through change-points, determining affect on the new bios map */
316 for (chgidx = 0; chgidx < chg_nr; chgidx++) {
317 /* keep track of all overlapping bios entries */
318 if (change_point[chgidx]->addr ==
319 change_point[chgidx]->pbios->addr) {
321 * add map entry to overlap list (> 1 entry
322 * implies an overlap)
324 overlap_list[overlap_entries++] =
325 change_point[chgidx]->pbios;
326 } else {
328 * remove entry from list (order independent,
329 * so swap with last)
331 for (i = 0; i < overlap_entries; i++) {
332 if (overlap_list[i] ==
333 change_point[chgidx]->pbios)
334 overlap_list[i] =
335 overlap_list[overlap_entries-1];
337 overlap_entries--;
340 * if there are overlapping entries, decide which
341 * "type" to use (larger value takes precedence --
342 * 1=usable, 2,3,4,4+=unusable)
344 current_type = 0;
345 for (i = 0; i < overlap_entries; i++)
346 if (overlap_list[i]->type > current_type)
347 current_type = overlap_list[i]->type;
349 * continue building up new bios map based on this
350 * information
352 if (current_type != last_type) {
353 if (last_type != 0) {
354 new_bios[new_bios_entry].size =
355 change_point[chgidx]->addr - last_addr;
357 * move forward only if the new size
358 * was non-zero
360 if (new_bios[new_bios_entry].size != 0)
362 * no more space left for new
363 * bios entries ?
365 if (++new_bios_entry >= max_nr_map)
366 break;
368 if (current_type != 0) {
369 new_bios[new_bios_entry].addr =
370 change_point[chgidx]->addr;
371 new_bios[new_bios_entry].type = current_type;
372 last_addr = change_point[chgidx]->addr;
374 last_type = current_type;
377 /* retain count for new bios entries */
378 new_nr = new_bios_entry;
380 /* copy new bios mapping into original location */
381 memcpy(biosmap, new_bios, new_nr * sizeof(struct e820entry));
382 *pnr_map = new_nr;
384 return 0;
387 static int __init __append_e820_map(struct e820entry *biosmap, int nr_map)
389 while (nr_map) {
390 u64 start = biosmap->addr;
391 u64 size = biosmap->size;
392 u64 end = start + size;
393 u32 type = biosmap->type;
395 /* Overflow in 64 bits? Ignore the memory map. */
396 if (start > end)
397 return -1;
399 e820_add_region(start, size, type);
401 biosmap++;
402 nr_map--;
404 return 0;
408 * Copy the BIOS e820 map into a safe place.
410 * Sanity-check it while we're at it..
412 * If we're lucky and live on a modern system, the setup code
413 * will have given us a memory map that we can use to properly
414 * set up memory. If we aren't, we'll fake a memory map.
416 static int __init append_e820_map(struct e820entry *biosmap, int nr_map)
418 /* Only one memory region (or negative)? Ignore it */
419 if (nr_map < 2)
420 return -1;
422 return __append_e820_map(biosmap, nr_map);
425 static u64 __init __e820_update_range(struct e820map *e820x, u64 start,
426 u64 size, unsigned old_type,
427 unsigned new_type)
429 u64 end;
430 unsigned int i;
431 u64 real_updated_size = 0;
433 BUG_ON(old_type == new_type);
435 if (size > (ULLONG_MAX - start))
436 size = ULLONG_MAX - start;
438 end = start + size;
439 printk(KERN_DEBUG "e820 update range: %016Lx - %016Lx ",
440 (unsigned long long) start,
441 (unsigned long long) end);
442 e820_print_type(old_type);
443 printk(KERN_CONT " ==> ");
444 e820_print_type(new_type);
445 printk(KERN_CONT "\n");
447 for (i = 0; i < e820x->nr_map; i++) {
448 struct e820entry *ei = &e820x->map[i];
449 u64 final_start, final_end;
450 u64 ei_end;
452 if (ei->type != old_type)
453 continue;
455 ei_end = ei->addr + ei->size;
456 /* totally covered by new range? */
457 if (ei->addr >= start && ei_end <= end) {
458 ei->type = new_type;
459 real_updated_size += ei->size;
460 continue;
463 /* new range is totally covered? */
464 if (ei->addr < start && ei_end > end) {
465 __e820_add_region(e820x, start, size, new_type);
466 __e820_add_region(e820x, end, ei_end - end, ei->type);
467 ei->size = start - ei->addr;
468 real_updated_size += size;
469 continue;
472 /* partially covered */
473 final_start = max(start, ei->addr);
474 final_end = min(end, ei_end);
475 if (final_start >= final_end)
476 continue;
478 __e820_add_region(e820x, final_start, final_end - final_start,
479 new_type);
481 real_updated_size += final_end - final_start;
484 * left range could be head or tail, so need to update
485 * size at first.
487 ei->size -= final_end - final_start;
488 if (ei->addr < final_start)
489 continue;
490 ei->addr = final_end;
492 return real_updated_size;
495 u64 __init e820_update_range(u64 start, u64 size, unsigned old_type,
496 unsigned new_type)
498 return __e820_update_range(&e820, start, size, old_type, new_type);
501 static u64 __init e820_update_range_saved(u64 start, u64 size,
502 unsigned old_type, unsigned new_type)
504 return __e820_update_range(&e820_saved, start, size, old_type,
505 new_type);
508 /* make e820 not cover the range */
509 u64 __init e820_remove_range(u64 start, u64 size, unsigned old_type,
510 int checktype)
512 int i;
513 u64 end;
514 u64 real_removed_size = 0;
516 if (size > (ULLONG_MAX - start))
517 size = ULLONG_MAX - start;
519 end = start + size;
520 printk(KERN_DEBUG "e820 remove range: %016Lx - %016Lx ",
521 (unsigned long long) start,
522 (unsigned long long) end);
523 e820_print_type(old_type);
524 printk(KERN_CONT "\n");
526 for (i = 0; i < e820.nr_map; i++) {
527 struct e820entry *ei = &e820.map[i];
528 u64 final_start, final_end;
530 if (checktype && ei->type != old_type)
531 continue;
532 /* totally covered? */
533 if (ei->addr >= start &&
534 (ei->addr + ei->size) <= (start + size)) {
535 real_removed_size += ei->size;
536 memset(ei, 0, sizeof(struct e820entry));
537 continue;
539 /* partially covered */
540 final_start = max(start, ei->addr);
541 final_end = min(start + size, ei->addr + ei->size);
542 if (final_start >= final_end)
543 continue;
544 real_removed_size += final_end - final_start;
546 ei->size -= final_end - final_start;
547 if (ei->addr < final_start)
548 continue;
549 ei->addr = final_end;
551 return real_removed_size;
554 void __init update_e820(void)
556 u32 nr_map;
558 nr_map = e820.nr_map;
559 if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr_map))
560 return;
561 e820.nr_map = nr_map;
562 printk(KERN_INFO "modified physical RAM map:\n");
563 e820_print_map("modified");
565 static void __init update_e820_saved(void)
567 u32 nr_map;
569 nr_map = e820_saved.nr_map;
570 if (sanitize_e820_map(e820_saved.map, ARRAY_SIZE(e820_saved.map), &nr_map))
571 return;
572 e820_saved.nr_map = nr_map;
574 #define MAX_GAP_END 0x100000000ull
576 * Search for a gap in the e820 memory space from start_addr to end_addr.
578 __init int e820_search_gap(unsigned long *gapstart, unsigned long *gapsize,
579 unsigned long start_addr, unsigned long long end_addr)
581 unsigned long long last;
582 int i = e820.nr_map;
583 int found = 0;
585 last = (end_addr && end_addr < MAX_GAP_END) ? end_addr : MAX_GAP_END;
587 while (--i >= 0) {
588 unsigned long long start = e820.map[i].addr;
589 unsigned long long end = start + e820.map[i].size;
591 if (end < start_addr)
592 continue;
595 * Since "last" is at most 4GB, we know we'll
596 * fit in 32 bits if this condition is true
598 if (last > end) {
599 unsigned long gap = last - end;
601 if (gap >= *gapsize) {
602 *gapsize = gap;
603 *gapstart = end;
604 found = 1;
607 if (start < last)
608 last = start;
610 return found;
614 * Search for the biggest gap in the low 32 bits of the e820
615 * memory space. We pass this space to PCI to assign MMIO resources
616 * for hotplug or unconfigured devices in.
617 * Hopefully the BIOS let enough space left.
619 __init void e820_setup_gap(void)
621 unsigned long gapstart, gapsize;
622 int found;
624 gapstart = 0x10000000;
625 gapsize = 0x400000;
626 found = e820_search_gap(&gapstart, &gapsize, 0, MAX_GAP_END);
628 #ifdef CONFIG_X86_64
629 if (!found) {
630 gapstart = (max_pfn << PAGE_SHIFT) + 1024*1024;
631 printk(KERN_ERR
632 "PCI: Warning: Cannot find a gap in the 32bit address range\n"
633 "PCI: Unassigned devices with 32bit resource registers may break!\n");
635 #endif
638 * e820_reserve_resources_late protect stolen RAM already
640 pci_mem_start = gapstart;
642 printk(KERN_INFO
643 "Allocating PCI resources starting at %lx (gap: %lx:%lx)\n",
644 pci_mem_start, gapstart, gapsize);
648 * Because of the size limitation of struct boot_params, only first
649 * 128 E820 memory entries are passed to kernel via
650 * boot_params.e820_map, others are passed via SETUP_E820_EXT node of
651 * linked list of struct setup_data, which is parsed here.
653 void __init parse_e820_ext(struct setup_data *sdata, unsigned long pa_data)
655 u32 map_len;
656 int entries;
657 struct e820entry *extmap;
659 entries = sdata->len / sizeof(struct e820entry);
660 map_len = sdata->len + sizeof(struct setup_data);
661 if (map_len > PAGE_SIZE)
662 sdata = early_ioremap(pa_data, map_len);
663 extmap = (struct e820entry *)(sdata->data);
664 __append_e820_map(extmap, entries);
665 sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map);
666 if (map_len > PAGE_SIZE)
667 early_iounmap(sdata, map_len);
668 printk(KERN_INFO "extended physical RAM map:\n");
669 e820_print_map("extended");
672 #if defined(CONFIG_X86_64) || \
673 (defined(CONFIG_X86_32) && defined(CONFIG_HIBERNATION))
675 * Find the ranges of physical addresses that do not correspond to
676 * e820 RAM areas and mark the corresponding pages as nosave for
677 * hibernation (32 bit) or software suspend and suspend to RAM (64 bit).
679 * This function requires the e820 map to be sorted and without any
680 * overlapping entries and assumes the first e820 area to be RAM.
682 void __init e820_mark_nosave_regions(unsigned long limit_pfn)
684 int i;
685 unsigned long pfn;
687 pfn = PFN_DOWN(e820.map[0].addr + e820.map[0].size);
688 for (i = 1; i < e820.nr_map; i++) {
689 struct e820entry *ei = &e820.map[i];
691 if (pfn < PFN_UP(ei->addr))
692 register_nosave_region(pfn, PFN_UP(ei->addr));
694 pfn = PFN_DOWN(ei->addr + ei->size);
695 if (ei->type != E820_RAM && ei->type != E820_RESERVED_KERN)
696 register_nosave_region(PFN_UP(ei->addr), pfn);
698 if (pfn >= limit_pfn)
699 break;
702 #endif
704 #ifdef CONFIG_HIBERNATION
706 * Mark ACPI NVS memory region, so that we can save/restore it during
707 * hibernation and the subsequent resume.
709 static int __init e820_mark_nvs_memory(void)
711 int i;
713 for (i = 0; i < e820.nr_map; i++) {
714 struct e820entry *ei = &e820.map[i];
716 if (ei->type == E820_NVS)
717 hibernate_nvs_register(ei->addr, ei->size);
720 return 0;
722 core_initcall(e820_mark_nvs_memory);
723 #endif
726 * Find a free area with specified alignment in a specific range.
728 u64 __init find_e820_area(u64 start, u64 end, u64 size, u64 align)
730 int i;
732 for (i = 0; i < e820.nr_map; i++) {
733 struct e820entry *ei = &e820.map[i];
734 u64 addr;
735 u64 ei_start, ei_last;
737 if (ei->type != E820_RAM)
738 continue;
740 ei_last = ei->addr + ei->size;
741 ei_start = ei->addr;
742 addr = find_early_area(ei_start, ei_last, start, end,
743 size, align);
745 if (addr != -1ULL)
746 return addr;
748 return -1ULL;
752 * Find next free range after *start
754 u64 __init find_e820_area_size(u64 start, u64 *sizep, u64 align)
756 int i;
758 for (i = 0; i < e820.nr_map; i++) {
759 struct e820entry *ei = &e820.map[i];
760 u64 addr;
761 u64 ei_start, ei_last;
763 if (ei->type != E820_RAM)
764 continue;
766 ei_last = ei->addr + ei->size;
767 ei_start = ei->addr;
768 addr = find_early_area_size(ei_start, ei_last, start,
769 sizep, align);
771 if (addr != -1ULL)
772 return addr;
775 return -1ULL;
779 * pre allocated 4k and reserved it in e820
781 u64 __init early_reserve_e820(u64 startt, u64 sizet, u64 align)
783 u64 size = 0;
784 u64 addr;
785 u64 start;
787 for (start = startt; ; start += size) {
788 start = find_e820_area_size(start, &size, align);
789 if (!(start + 1))
790 return 0;
791 if (size >= sizet)
792 break;
795 #ifdef CONFIG_X86_32
796 if (start >= MAXMEM)
797 return 0;
798 if (start + size > MAXMEM)
799 size = MAXMEM - start;
800 #endif
802 addr = round_down(start + size - sizet, align);
803 if (addr < start)
804 return 0;
805 e820_update_range(addr, sizet, E820_RAM, E820_RESERVED);
806 e820_update_range_saved(addr, sizet, E820_RAM, E820_RESERVED);
807 printk(KERN_INFO "update e820 for early_reserve_e820\n");
808 update_e820();
809 update_e820_saved();
811 return addr;
814 #ifdef CONFIG_X86_32
815 # ifdef CONFIG_X86_PAE
816 # define MAX_ARCH_PFN (1ULL<<(36-PAGE_SHIFT))
817 # else
818 # define MAX_ARCH_PFN (1ULL<<(32-PAGE_SHIFT))
819 # endif
820 #else /* CONFIG_X86_32 */
821 # define MAX_ARCH_PFN MAXMEM>>PAGE_SHIFT
822 #endif
825 * Find the highest page frame number we have available
827 static unsigned long __init e820_end_pfn(unsigned long limit_pfn, unsigned type)
829 int i;
830 unsigned long last_pfn = 0;
831 unsigned long max_arch_pfn = MAX_ARCH_PFN;
833 for (i = 0; i < e820.nr_map; i++) {
834 struct e820entry *ei = &e820.map[i];
835 unsigned long start_pfn;
836 unsigned long end_pfn;
838 if (ei->type != type)
839 continue;
841 start_pfn = ei->addr >> PAGE_SHIFT;
842 end_pfn = (ei->addr + ei->size) >> PAGE_SHIFT;
844 if (start_pfn >= limit_pfn)
845 continue;
846 if (end_pfn > limit_pfn) {
847 last_pfn = limit_pfn;
848 break;
850 if (end_pfn > last_pfn)
851 last_pfn = end_pfn;
854 if (last_pfn > max_arch_pfn)
855 last_pfn = max_arch_pfn;
857 printk(KERN_INFO "last_pfn = %#lx max_arch_pfn = %#lx\n",
858 last_pfn, max_arch_pfn);
859 return last_pfn;
861 unsigned long __init e820_end_of_ram_pfn(void)
863 return e820_end_pfn(MAX_ARCH_PFN, E820_RAM);
866 unsigned long __init e820_end_of_low_ram_pfn(void)
868 return e820_end_pfn(1UL<<(32 - PAGE_SHIFT), E820_RAM);
871 * Finds an active region in the address range from start_pfn to last_pfn and
872 * returns its range in ei_startpfn and ei_endpfn for the e820 entry.
874 int __init e820_find_active_region(const struct e820entry *ei,
875 unsigned long start_pfn,
876 unsigned long last_pfn,
877 unsigned long *ei_startpfn,
878 unsigned long *ei_endpfn)
880 u64 align = PAGE_SIZE;
882 *ei_startpfn = round_up(ei->addr, align) >> PAGE_SHIFT;
883 *ei_endpfn = round_down(ei->addr + ei->size, align) >> PAGE_SHIFT;
885 /* Skip map entries smaller than a page */
886 if (*ei_startpfn >= *ei_endpfn)
887 return 0;
889 /* Skip if map is outside the node */
890 if (ei->type != E820_RAM || *ei_endpfn <= start_pfn ||
891 *ei_startpfn >= last_pfn)
892 return 0;
894 /* Check for overlaps */
895 if (*ei_startpfn < start_pfn)
896 *ei_startpfn = start_pfn;
897 if (*ei_endpfn > last_pfn)
898 *ei_endpfn = last_pfn;
900 return 1;
903 /* Walk the e820 map and register active regions within a node */
904 void __init e820_register_active_regions(int nid, unsigned long start_pfn,
905 unsigned long last_pfn)
907 unsigned long ei_startpfn;
908 unsigned long ei_endpfn;
909 int i;
911 for (i = 0; i < e820.nr_map; i++)
912 if (e820_find_active_region(&e820.map[i],
913 start_pfn, last_pfn,
914 &ei_startpfn, &ei_endpfn))
915 add_active_range(nid, ei_startpfn, ei_endpfn);
919 * Find the hole size (in bytes) in the memory range.
920 * @start: starting address of the memory range to scan
921 * @end: ending address of the memory range to scan
923 u64 __init e820_hole_size(u64 start, u64 end)
925 unsigned long start_pfn = start >> PAGE_SHIFT;
926 unsigned long last_pfn = end >> PAGE_SHIFT;
927 unsigned long ei_startpfn, ei_endpfn, ram = 0;
928 int i;
930 for (i = 0; i < e820.nr_map; i++) {
931 if (e820_find_active_region(&e820.map[i],
932 start_pfn, last_pfn,
933 &ei_startpfn, &ei_endpfn))
934 ram += ei_endpfn - ei_startpfn;
936 return end - start - ((u64)ram << PAGE_SHIFT);
939 static void early_panic(char *msg)
941 early_printk(msg);
942 panic(msg);
945 static int userdef __initdata;
947 /* "mem=nopentium" disables the 4MB page tables. */
948 static int __init parse_memopt(char *p)
950 u64 mem_size;
952 if (!p)
953 return -EINVAL;
955 #ifdef CONFIG_X86_32
956 if (!strcmp(p, "nopentium")) {
957 setup_clear_cpu_cap(X86_FEATURE_PSE);
958 return 0;
960 #endif
962 userdef = 1;
963 mem_size = memparse(p, &p);
964 e820_remove_range(mem_size, ULLONG_MAX - mem_size, E820_RAM, 1);
966 return 0;
968 early_param("mem", parse_memopt);
970 static int __init parse_memmap_opt(char *p)
972 char *oldp;
973 u64 start_at, mem_size;
975 if (!p)
976 return -EINVAL;
978 if (!strncmp(p, "exactmap", 8)) {
979 #ifdef CONFIG_CRASH_DUMP
981 * If we are doing a crash dump, we still need to know
982 * the real mem size before original memory map is
983 * reset.
985 saved_max_pfn = e820_end_of_ram_pfn();
986 #endif
987 e820.nr_map = 0;
988 userdef = 1;
989 return 0;
992 oldp = p;
993 mem_size = memparse(p, &p);
994 if (p == oldp)
995 return -EINVAL;
997 userdef = 1;
998 if (*p == '@') {
999 start_at = memparse(p+1, &p);
1000 e820_add_region(start_at, mem_size, E820_RAM);
1001 } else if (*p == '#') {
1002 start_at = memparse(p+1, &p);
1003 e820_add_region(start_at, mem_size, E820_ACPI);
1004 } else if (*p == '$') {
1005 start_at = memparse(p+1, &p);
1006 e820_add_region(start_at, mem_size, E820_RESERVED);
1007 } else
1008 e820_remove_range(mem_size, ULLONG_MAX - mem_size, E820_RAM, 1);
1010 return *p == '\0' ? 0 : -EINVAL;
1012 early_param("memmap", parse_memmap_opt);
1014 void __init finish_e820_parsing(void)
1016 if (userdef) {
1017 u32 nr = e820.nr_map;
1019 if (sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &nr) < 0)
1020 early_panic("Invalid user supplied memory map");
1021 e820.nr_map = nr;
1023 printk(KERN_INFO "user-defined physical RAM map:\n");
1024 e820_print_map("user");
1028 static inline const char *e820_type_to_string(int e820_type)
1030 switch (e820_type) {
1031 case E820_RESERVED_KERN:
1032 case E820_RAM: return "System RAM";
1033 case E820_ACPI: return "ACPI Tables";
1034 case E820_NVS: return "ACPI Non-volatile Storage";
1035 case E820_UNUSABLE: return "Unusable memory";
1036 default: return "reserved";
1041 * Mark e820 reserved areas as busy for the resource manager.
1043 static struct resource __initdata *e820_res;
1044 void __init e820_reserve_resources(void)
1046 int i;
1047 struct resource *res;
1048 u64 end;
1050 res = alloc_bootmem(sizeof(struct resource) * e820.nr_map);
1051 e820_res = res;
1052 for (i = 0; i < e820.nr_map; i++) {
1053 end = e820.map[i].addr + e820.map[i].size - 1;
1054 if (end != (resource_size_t)end) {
1055 res++;
1056 continue;
1058 res->name = e820_type_to_string(e820.map[i].type);
1059 res->start = e820.map[i].addr;
1060 res->end = end;
1062 res->flags = IORESOURCE_MEM;
1065 * don't register the region that could be conflicted with
1066 * pci device BAR resource and insert them later in
1067 * pcibios_resource_survey()
1069 if (e820.map[i].type != E820_RESERVED || res->start < (1ULL<<20)) {
1070 res->flags |= IORESOURCE_BUSY;
1071 insert_resource(&iomem_resource, res);
1073 res++;
1076 for (i = 0; i < e820_saved.nr_map; i++) {
1077 struct e820entry *entry = &e820_saved.map[i];
1078 firmware_map_add_early(entry->addr,
1079 entry->addr + entry->size - 1,
1080 e820_type_to_string(entry->type));
1084 /* How much should we pad RAM ending depending on where it is? */
1085 static unsigned long ram_alignment(resource_size_t pos)
1087 unsigned long mb = pos >> 20;
1089 /* To 64kB in the first megabyte */
1090 if (!mb)
1091 return 64*1024;
1093 /* To 1MB in the first 16MB */
1094 if (mb < 16)
1095 return 1024*1024;
1097 /* To 64MB for anything above that */
1098 return 64*1024*1024;
1101 #define MAX_RESOURCE_SIZE ((resource_size_t)-1)
1103 void __init e820_reserve_resources_late(void)
1105 int i;
1106 struct resource *res;
1108 res = e820_res;
1109 for (i = 0; i < e820.nr_map; i++) {
1110 if (!res->parent && res->end)
1111 insert_resource_expand_to_fit(&iomem_resource, res);
1112 res++;
1116 * Try to bump up RAM regions to reasonable boundaries to
1117 * avoid stolen RAM:
1119 for (i = 0; i < e820.nr_map; i++) {
1120 struct e820entry *entry = &e820.map[i];
1121 u64 start, end;
1123 if (entry->type != E820_RAM)
1124 continue;
1125 start = entry->addr + entry->size;
1126 end = round_up(start, ram_alignment(start)) - 1;
1127 if (end > MAX_RESOURCE_SIZE)
1128 end = MAX_RESOURCE_SIZE;
1129 if (start >= end)
1130 continue;
1131 printk(KERN_DEBUG "reserve RAM buffer: %016llx - %016llx ",
1132 start, end);
1133 reserve_region_with_split(&iomem_resource, start, end,
1134 "RAM buffer");
1138 char *__init default_machine_specific_memory_setup(void)
1140 char *who = "BIOS-e820";
1141 u32 new_nr;
1143 * Try to copy the BIOS-supplied E820-map.
1145 * Otherwise fake a memory map; one section from 0k->640k,
1146 * the next section from 1mb->appropriate_mem_k
1148 new_nr = boot_params.e820_entries;
1149 sanitize_e820_map(boot_params.e820_map,
1150 ARRAY_SIZE(boot_params.e820_map),
1151 &new_nr);
1152 boot_params.e820_entries = new_nr;
1153 if (append_e820_map(boot_params.e820_map, boot_params.e820_entries)
1154 < 0) {
1155 u64 mem_size;
1157 /* compare results from other methods and take the greater */
1158 if (boot_params.alt_mem_k
1159 < boot_params.screen_info.ext_mem_k) {
1160 mem_size = boot_params.screen_info.ext_mem_k;
1161 who = "BIOS-88";
1162 } else {
1163 mem_size = boot_params.alt_mem_k;
1164 who = "BIOS-e801";
1167 e820.nr_map = 0;
1168 e820_add_region(0, LOWMEMSIZE(), E820_RAM);
1169 e820_add_region(HIGH_MEMORY, mem_size << 10, E820_RAM);
1172 /* In case someone cares... */
1173 return who;
1176 void __init setup_memory_map(void)
1178 char *who;
1180 who = x86_init.resources.memory_setup();
1181 memcpy(&e820_saved, &e820, sizeof(struct e820map));
1182 printk(KERN_INFO "BIOS-provided physical RAM map:\n");
1183 e820_print_map(who);