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/ioport.h>
16 #include <linux/string.h>
17 #include <linux/kexec.h>
18 #include <linux/module.h>
20 #include <linux/pfn.h>
21 #include <linux/suspend.h>
22 #include <linux/firmware-map.h>
24 #include <asm/pgtable.h>
27 #include <asm/proto.h>
28 #include <asm/setup.h>
29 #include <asm/trampoline.h>
32 * The e820 map is the map that gets modified e.g. with command line parameters
33 * and that is also registered with modifications in the kernel resource tree
34 * with the iomem_resource as parent.
36 * The e820_saved is directly saved after the BIOS-provided memory map is
37 * copied. It doesn't get modified afterwards. It's registered for the
38 * /sys/firmware/memmap interface.
40 * That memory map is not modified and is used as base for kexec. The kexec'd
41 * kernel should get the same memory map as the firmware provides. Then the
42 * user can e.g. boot the original kernel with mem=1G while still booting the
43 * next kernel with full memory.
46 struct e820map e820_saved
;
48 /* For PCI or other memory-mapped resources */
49 unsigned long pci_mem_start
= 0xaeedbabe;
51 EXPORT_SYMBOL(pci_mem_start
);
55 * This function checks if any part of the range <start,end> is mapped
59 e820_any_mapped(u64 start
, u64 end
, unsigned type
)
63 for (i
= 0; i
< e820
.nr_map
; i
++) {
64 struct e820entry
*ei
= &e820
.map
[i
];
66 if (type
&& ei
->type
!= type
)
68 if (ei
->addr
>= end
|| ei
->addr
+ ei
->size
<= start
)
74 EXPORT_SYMBOL_GPL(e820_any_mapped
);
77 * This function checks if the entire range <start,end> is mapped with type.
79 * Note: this function only works correct if the e820 table is sorted and
80 * not-overlapping, which is the case
82 int __init
e820_all_mapped(u64 start
, u64 end
, unsigned type
)
86 for (i
= 0; i
< e820
.nr_map
; i
++) {
87 struct e820entry
*ei
= &e820
.map
[i
];
89 if (type
&& ei
->type
!= type
)
91 /* is the region (part) in overlap with the current region ?*/
92 if (ei
->addr
>= end
|| ei
->addr
+ ei
->size
<= start
)
95 /* if the region is at the beginning of <start,end> we move
96 * start to the end of the region since it's ok until there
98 if (ei
->addr
<= start
)
99 start
= ei
->addr
+ ei
->size
;
101 * if start is now at or beyond end, we're done, full
111 * Add a memory region to the kernel e820 map.
113 static void __init
__e820_add_region(struct e820map
*e820x
, u64 start
, u64 size
,
116 int x
= e820x
->nr_map
;
118 if (x
>= ARRAY_SIZE(e820x
->map
)) {
119 printk(KERN_ERR
"Ooops! Too many entries in the memory map!\n");
123 e820x
->map
[x
].addr
= start
;
124 e820x
->map
[x
].size
= size
;
125 e820x
->map
[x
].type
= type
;
129 void __init
e820_add_region(u64 start
, u64 size
, int type
)
131 __e820_add_region(&e820
, start
, size
, type
);
134 static void __init
e820_print_type(u32 type
)
138 case E820_RESERVED_KERN
:
139 printk(KERN_CONT
"(usable)");
142 printk(KERN_CONT
"(reserved)");
145 printk(KERN_CONT
"(ACPI data)");
148 printk(KERN_CONT
"(ACPI NVS)");
151 printk(KERN_CONT
"(unusable)");
154 printk(KERN_CONT
"type %u", type
);
159 void __init
e820_print_map(char *who
)
163 for (i
= 0; i
< e820
.nr_map
; i
++) {
164 printk(KERN_INFO
" %s: %016Lx - %016Lx ", who
,
165 (unsigned long long) e820
.map
[i
].addr
,
167 (e820
.map
[i
].addr
+ e820
.map
[i
].size
));
168 e820_print_type(e820
.map
[i
].type
);
169 printk(KERN_CONT
"\n");
174 * Sanitize the BIOS e820 map.
176 * Some e820 responses include overlapping entries. The following
177 * replaces the original e820 map with a new one, removing overlaps,
178 * and resolving conflicting memory types in favor of highest
181 * The input parameter biosmap points to an array of 'struct
182 * e820entry' which on entry has elements in the range [0, *pnr_map)
183 * valid, and which has space for up to max_nr_map entries.
184 * On return, the resulting sanitized e820 map entries will be in
185 * overwritten in the same location, starting at biosmap.
187 * The integer pointed to by pnr_map must be valid on entry (the
188 * current number of valid entries located at biosmap) and will
189 * be updated on return, with the new number of valid entries
190 * (something no more than max_nr_map.)
192 * The return value from sanitize_e820_map() is zero if it
193 * successfully 'sanitized' the map entries passed in, and is -1
194 * if it did nothing, which can happen if either of (1) it was
195 * only passed one map entry, or (2) any of the input map entries
196 * were invalid (start + size < start, meaning that the size was
197 * so big the described memory range wrapped around through zero.)
199 * Visually we're performing the following
200 * (1,2,3,4 = memory types)...
202 * Sample memory map (w/overlaps):
203 * ____22__________________
204 * ______________________4_
205 * ____1111________________
206 * _44_____________________
207 * 11111111________________
208 * ____________________33__
209 * ___________44___________
210 * __________33333_________
211 * ______________22________
212 * ___________________2222_
213 * _________111111111______
214 * _____________________11_
215 * _________________4______
217 * Sanitized equivalent (no overlap):
218 * 1_______________________
219 * _44_____________________
220 * ___1____________________
221 * ____22__________________
222 * ______11________________
223 * _________1______________
224 * __________3_____________
225 * ___________44___________
226 * _____________33_________
227 * _______________2________
228 * ________________1_______
229 * _________________4______
230 * ___________________2____
231 * ____________________33__
232 * ______________________4_
235 int __init
sanitize_e820_map(struct e820entry
*biosmap
, int max_nr_map
,
238 struct change_member
{
239 struct e820entry
*pbios
; /* pointer to original bios entry */
240 unsigned long long addr
; /* address for this change point */
242 static struct change_member change_point_list
[2*E820_X_MAX
] __initdata
;
243 static struct change_member
*change_point
[2*E820_X_MAX
] __initdata
;
244 static struct e820entry
*overlap_list
[E820_X_MAX
] __initdata
;
245 static struct e820entry new_bios
[E820_X_MAX
] __initdata
;
246 struct change_member
*change_tmp
;
247 unsigned long current_type
, last_type
;
248 unsigned long long last_addr
;
249 int chgidx
, still_changing
;
252 int old_nr
, new_nr
, chg_nr
;
255 /* if there's only one memory region, don't bother */
260 BUG_ON(old_nr
> max_nr_map
);
262 /* bail out if we find any unreasonable addresses in bios map */
263 for (i
= 0; i
< old_nr
; i
++)
264 if (biosmap
[i
].addr
+ biosmap
[i
].size
< biosmap
[i
].addr
)
267 /* create pointers for initial change-point information (for sorting) */
268 for (i
= 0; i
< 2 * old_nr
; i
++)
269 change_point
[i
] = &change_point_list
[i
];
271 /* record all known change-points (starting and ending addresses),
272 omitting those that are for empty memory regions */
274 for (i
= 0; i
< old_nr
; i
++) {
275 if (biosmap
[i
].size
!= 0) {
276 change_point
[chgidx
]->addr
= biosmap
[i
].addr
;
277 change_point
[chgidx
++]->pbios
= &biosmap
[i
];
278 change_point
[chgidx
]->addr
= biosmap
[i
].addr
+
280 change_point
[chgidx
++]->pbios
= &biosmap
[i
];
285 /* sort change-point list by memory addresses (low -> high) */
287 while (still_changing
) {
289 for (i
= 1; i
< chg_nr
; i
++) {
290 unsigned long long curaddr
, lastaddr
;
291 unsigned long long curpbaddr
, lastpbaddr
;
293 curaddr
= change_point
[i
]->addr
;
294 lastaddr
= change_point
[i
- 1]->addr
;
295 curpbaddr
= change_point
[i
]->pbios
->addr
;
296 lastpbaddr
= change_point
[i
- 1]->pbios
->addr
;
299 * swap entries, when:
301 * curaddr > lastaddr or
302 * curaddr == lastaddr and curaddr == curpbaddr and
303 * lastaddr != lastpbaddr
305 if (curaddr
< lastaddr
||
306 (curaddr
== lastaddr
&& curaddr
== curpbaddr
&&
307 lastaddr
!= lastpbaddr
)) {
308 change_tmp
= change_point
[i
];
309 change_point
[i
] = change_point
[i
-1];
310 change_point
[i
-1] = change_tmp
;
316 /* create a new bios memory map, removing overlaps */
317 overlap_entries
= 0; /* number of entries in the overlap table */
318 new_bios_entry
= 0; /* index for creating new bios map entries */
319 last_type
= 0; /* start with undefined memory type */
320 last_addr
= 0; /* start with 0 as last starting address */
322 /* loop through change-points, determining affect on the new bios map */
323 for (chgidx
= 0; chgidx
< chg_nr
; chgidx
++) {
324 /* keep track of all overlapping bios entries */
325 if (change_point
[chgidx
]->addr
==
326 change_point
[chgidx
]->pbios
->addr
) {
328 * add map entry to overlap list (> 1 entry
329 * implies an overlap)
331 overlap_list
[overlap_entries
++] =
332 change_point
[chgidx
]->pbios
;
335 * remove entry from list (order independent,
338 for (i
= 0; i
< overlap_entries
; i
++) {
339 if (overlap_list
[i
] ==
340 change_point
[chgidx
]->pbios
)
342 overlap_list
[overlap_entries
-1];
347 * if there are overlapping entries, decide which
348 * "type" to use (larger value takes precedence --
349 * 1=usable, 2,3,4,4+=unusable)
352 for (i
= 0; i
< overlap_entries
; i
++)
353 if (overlap_list
[i
]->type
> current_type
)
354 current_type
= overlap_list
[i
]->type
;
356 * continue building up new bios map based on this
359 if (current_type
!= last_type
) {
360 if (last_type
!= 0) {
361 new_bios
[new_bios_entry
].size
=
362 change_point
[chgidx
]->addr
- last_addr
;
364 * move forward only if the new size
367 if (new_bios
[new_bios_entry
].size
!= 0)
369 * no more space left for new
372 if (++new_bios_entry
>= max_nr_map
)
375 if (current_type
!= 0) {
376 new_bios
[new_bios_entry
].addr
=
377 change_point
[chgidx
]->addr
;
378 new_bios
[new_bios_entry
].type
= current_type
;
379 last_addr
= change_point
[chgidx
]->addr
;
381 last_type
= current_type
;
384 /* retain count for new bios entries */
385 new_nr
= new_bios_entry
;
387 /* copy new bios mapping into original location */
388 memcpy(biosmap
, new_bios
, new_nr
* sizeof(struct e820entry
));
394 static int __init
__append_e820_map(struct e820entry
*biosmap
, int nr_map
)
397 u64 start
= biosmap
->addr
;
398 u64 size
= biosmap
->size
;
399 u64 end
= start
+ size
;
400 u32 type
= biosmap
->type
;
402 /* Overflow in 64 bits? Ignore the memory map. */
406 e820_add_region(start
, size
, type
);
415 * Copy the BIOS e820 map into a safe place.
417 * Sanity-check it while we're at it..
419 * If we're lucky and live on a modern system, the setup code
420 * will have given us a memory map that we can use to properly
421 * set up memory. If we aren't, we'll fake a memory map.
423 static int __init
append_e820_map(struct e820entry
*biosmap
, int nr_map
)
425 /* Only one memory region (or negative)? Ignore it */
429 return __append_e820_map(biosmap
, nr_map
);
432 static u64 __init
__e820_update_range(struct e820map
*e820x
, u64 start
,
433 u64 size
, unsigned old_type
,
438 u64 real_updated_size
= 0;
440 BUG_ON(old_type
== new_type
);
442 if (size
> (ULLONG_MAX
- start
))
443 size
= ULLONG_MAX
- start
;
446 printk(KERN_DEBUG
"e820 update range: %016Lx - %016Lx ",
447 (unsigned long long) start
,
448 (unsigned long long) end
);
449 e820_print_type(old_type
);
450 printk(KERN_CONT
" ==> ");
451 e820_print_type(new_type
);
452 printk(KERN_CONT
"\n");
454 for (i
= 0; i
< e820x
->nr_map
; i
++) {
455 struct e820entry
*ei
= &e820x
->map
[i
];
456 u64 final_start
, final_end
;
459 if (ei
->type
!= old_type
)
462 ei_end
= ei
->addr
+ ei
->size
;
463 /* totally covered by new range? */
464 if (ei
->addr
>= start
&& ei_end
<= end
) {
466 real_updated_size
+= ei
->size
;
470 /* new range is totally covered? */
471 if (ei
->addr
< start
&& ei_end
> end
) {
472 __e820_add_region(e820x
, start
, size
, new_type
);
473 __e820_add_region(e820x
, end
, ei_end
- end
, ei
->type
);
474 ei
->size
= start
- ei
->addr
;
475 real_updated_size
+= size
;
479 /* partially covered */
480 final_start
= max(start
, ei
->addr
);
481 final_end
= min(end
, ei_end
);
482 if (final_start
>= final_end
)
485 __e820_add_region(e820x
, final_start
, final_end
- final_start
,
488 real_updated_size
+= final_end
- final_start
;
491 * left range could be head or tail, so need to update
494 ei
->size
-= final_end
- final_start
;
495 if (ei
->addr
< final_start
)
497 ei
->addr
= final_end
;
499 return real_updated_size
;
502 u64 __init
e820_update_range(u64 start
, u64 size
, unsigned old_type
,
505 return __e820_update_range(&e820
, start
, size
, old_type
, new_type
);
508 static u64 __init
e820_update_range_saved(u64 start
, u64 size
,
509 unsigned old_type
, unsigned new_type
)
511 return __e820_update_range(&e820_saved
, start
, size
, old_type
,
515 /* make e820 not cover the range */
516 u64 __init
e820_remove_range(u64 start
, u64 size
, unsigned old_type
,
520 u64 real_removed_size
= 0;
522 if (size
> (ULLONG_MAX
- start
))
523 size
= ULLONG_MAX
- start
;
525 for (i
= 0; i
< e820
.nr_map
; i
++) {
526 struct e820entry
*ei
= &e820
.map
[i
];
527 u64 final_start
, final_end
;
529 if (checktype
&& ei
->type
!= old_type
)
531 /* totally covered? */
532 if (ei
->addr
>= start
&&
533 (ei
->addr
+ ei
->size
) <= (start
+ size
)) {
534 real_removed_size
+= ei
->size
;
535 memset(ei
, 0, sizeof(struct e820entry
));
538 /* partially covered */
539 final_start
= max(start
, ei
->addr
);
540 final_end
= min(start
+ size
, ei
->addr
+ ei
->size
);
541 if (final_start
>= final_end
)
543 real_removed_size
+= final_end
- final_start
;
545 ei
->size
-= final_end
- final_start
;
546 if (ei
->addr
< final_start
)
548 ei
->addr
= final_end
;
550 return real_removed_size
;
553 void __init
update_e820(void)
557 nr_map
= e820
.nr_map
;
558 if (sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &nr_map
))
560 e820
.nr_map
= nr_map
;
561 printk(KERN_INFO
"modified physical RAM map:\n");
562 e820_print_map("modified");
564 static void __init
update_e820_saved(void)
568 nr_map
= e820_saved
.nr_map
;
569 if (sanitize_e820_map(e820_saved
.map
, ARRAY_SIZE(e820_saved
.map
), &nr_map
))
571 e820_saved
.nr_map
= nr_map
;
573 #define MAX_GAP_END 0x100000000ull
575 * Search for a gap in the e820 memory space from start_addr to end_addr.
577 __init
int e820_search_gap(unsigned long *gapstart
, unsigned long *gapsize
,
578 unsigned long start_addr
, unsigned long long end_addr
)
580 unsigned long long last
;
584 last
= (end_addr
&& end_addr
< MAX_GAP_END
) ? end_addr
: MAX_GAP_END
;
587 unsigned long long start
= e820
.map
[i
].addr
;
588 unsigned long long end
= start
+ e820
.map
[i
].size
;
590 if (end
< start_addr
)
594 * Since "last" is at most 4GB, we know we'll
595 * fit in 32 bits if this condition is true
598 unsigned long gap
= last
- end
;
600 if (gap
>= *gapsize
) {
613 * Search for the biggest gap in the low 32 bits of the e820
614 * memory space. We pass this space to PCI to assign MMIO resources
615 * for hotplug or unconfigured devices in.
616 * Hopefully the BIOS let enough space left.
618 __init
void e820_setup_gap(void)
620 unsigned long gapstart
, gapsize
;
623 gapstart
= 0x10000000;
625 found
= e820_search_gap(&gapstart
, &gapsize
, 0, MAX_GAP_END
);
629 gapstart
= (max_pfn
<< PAGE_SHIFT
) + 1024*1024;
631 "PCI: Warning: Cannot find a gap in the 32bit address range\n"
632 "PCI: Unassigned devices with 32bit resource registers may break!\n");
637 * e820_reserve_resources_late protect stolen RAM already
639 pci_mem_start
= gapstart
;
642 "Allocating PCI resources starting at %lx (gap: %lx:%lx)\n",
643 pci_mem_start
, gapstart
, gapsize
);
647 * Because of the size limitation of struct boot_params, only first
648 * 128 E820 memory entries are passed to kernel via
649 * boot_params.e820_map, others are passed via SETUP_E820_EXT node of
650 * linked list of struct setup_data, which is parsed here.
652 void __init
parse_e820_ext(struct setup_data
*sdata
, unsigned long pa_data
)
656 struct e820entry
*extmap
;
658 entries
= sdata
->len
/ sizeof(struct e820entry
);
659 map_len
= sdata
->len
+ sizeof(struct setup_data
);
660 if (map_len
> PAGE_SIZE
)
661 sdata
= early_ioremap(pa_data
, map_len
);
662 extmap
= (struct e820entry
*)(sdata
->data
);
663 __append_e820_map(extmap
, entries
);
664 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
665 if (map_len
> PAGE_SIZE
)
666 early_iounmap(sdata
, map_len
);
667 printk(KERN_INFO
"extended physical RAM map:\n");
668 e820_print_map("extended");
671 #if defined(CONFIG_X86_64) || \
672 (defined(CONFIG_X86_32) && defined(CONFIG_HIBERNATION))
674 * Find the ranges of physical addresses that do not correspond to
675 * e820 RAM areas and mark the corresponding pages as nosave for
676 * hibernation (32 bit) or software suspend and suspend to RAM (64 bit).
678 * This function requires the e820 map to be sorted and without any
679 * overlapping entries and assumes the first e820 area to be RAM.
681 void __init
e820_mark_nosave_regions(unsigned long limit_pfn
)
686 pfn
= PFN_DOWN(e820
.map
[0].addr
+ e820
.map
[0].size
);
687 for (i
= 1; i
< e820
.nr_map
; i
++) {
688 struct e820entry
*ei
= &e820
.map
[i
];
690 if (pfn
< PFN_UP(ei
->addr
))
691 register_nosave_region(pfn
, PFN_UP(ei
->addr
));
693 pfn
= PFN_DOWN(ei
->addr
+ ei
->size
);
694 if (ei
->type
!= E820_RAM
&& ei
->type
!= E820_RESERVED_KERN
)
695 register_nosave_region(PFN_UP(ei
->addr
), pfn
);
697 if (pfn
>= limit_pfn
)
703 #ifdef CONFIG_HIBERNATION
705 * Mark ACPI NVS memory region, so that we can save/restore it during
706 * hibernation and the subsequent resume.
708 static int __init
e820_mark_nvs_memory(void)
712 for (i
= 0; i
< e820
.nr_map
; i
++) {
713 struct e820entry
*ei
= &e820
.map
[i
];
715 if (ei
->type
== E820_NVS
)
716 hibernate_nvs_register(ei
->addr
, ei
->size
);
721 core_initcall(e820_mark_nvs_memory
);
725 * Early reserved memory areas.
727 #define MAX_EARLY_RES 32
734 static struct early_res early_res
[MAX_EARLY_RES
] __initdata
= {
735 { 0, PAGE_SIZE
, "BIOS data page", 1 }, /* BIOS data page */
736 #if defined(CONFIG_X86_32) && defined(CONFIG_X86_TRAMPOLINE)
738 * But first pinch a few for the stack/trampoline stuff
739 * FIXME: Don't need the extra page at 4K, but need to fix
740 * trampoline before removing it. (see the GDT stuff)
742 { PAGE_SIZE
, PAGE_SIZE
+ PAGE_SIZE
, "EX TRAMPOLINE", 1 },
748 static int __init
find_overlapped_early(u64 start
, u64 end
)
753 for (i
= 0; i
< MAX_EARLY_RES
&& early_res
[i
].end
; i
++) {
755 if (end
> r
->start
&& start
< r
->end
)
763 * Drop the i-th range from the early reservation map,
764 * by copying any higher ranges down one over it, and
765 * clearing what had been the last slot.
767 static void __init
drop_range(int i
)
771 for (j
= i
+ 1; j
< MAX_EARLY_RES
&& early_res
[j
].end
; j
++)
774 memmove(&early_res
[i
], &early_res
[i
+ 1],
775 (j
- 1 - i
) * sizeof(struct early_res
));
777 early_res
[j
- 1].end
= 0;
781 * Split any existing ranges that:
782 * 1) are marked 'overlap_ok', and
783 * 2) overlap with the stated range [start, end)
784 * into whatever portion (if any) of the existing range is entirely
785 * below or entirely above the stated range. Drop the portion
786 * of the existing range that overlaps with the stated range,
787 * which will allow the caller of this routine to then add that
788 * stated range without conflicting with any existing range.
790 static void __init
drop_overlaps_that_are_ok(u64 start
, u64 end
)
794 u64 lower_start
, lower_end
;
795 u64 upper_start
, upper_end
;
798 for (i
= 0; i
< MAX_EARLY_RES
&& early_res
[i
].end
; i
++) {
801 /* Continue past non-overlapping ranges */
802 if (end
<= r
->start
|| start
>= r
->end
)
806 * Leave non-ok overlaps as is; let caller
807 * panic "Overlapping early reservations"
808 * when it hits this overlap.
814 * We have an ok overlap. We will drop it from the early
815 * reservation map, and add back in any non-overlapping
816 * portions (lower or upper) as separate, overlap_ok,
817 * non-overlapping ranges.
820 /* 1. Note any non-overlapping (lower or upper) ranges. */
821 strncpy(name
, r
->name
, sizeof(name
) - 1);
823 lower_start
= lower_end
= 0;
824 upper_start
= upper_end
= 0;
825 if (r
->start
< start
) {
826 lower_start
= r
->start
;
834 /* 2. Drop the original ok overlapping range */
837 i
--; /* resume for-loop on copied down entry */
839 /* 3. Add back in any non-overlapping ranges. */
841 reserve_early_overlap_ok(lower_start
, lower_end
, name
);
843 reserve_early_overlap_ok(upper_start
, upper_end
, name
);
847 static void __init
__reserve_early(u64 start
, u64 end
, char *name
,
853 i
= find_overlapped_early(start
, end
);
854 if (i
>= MAX_EARLY_RES
)
855 panic("Too many early reservations");
858 panic("Overlapping early reservations "
859 "%llx-%llx %s to %llx-%llx %s\n",
860 start
, end
- 1, name
?name
:"", r
->start
,
861 r
->end
- 1, r
->name
);
864 r
->overlap_ok
= overlap_ok
;
866 strncpy(r
->name
, name
, sizeof(r
->name
) - 1);
870 * A few early reservtations come here.
872 * The 'overlap_ok' in the name of this routine does -not- mean it
873 * is ok for these reservations to overlap an earlier reservation.
874 * Rather it means that it is ok for subsequent reservations to
877 * Use this entry point to reserve early ranges when you are doing
878 * so out of "Paranoia", reserving perhaps more memory than you need,
879 * just in case, and don't mind a subsequent overlapping reservation
880 * that is known to be needed.
882 * The drop_overlaps_that_are_ok() call here isn't really needed.
883 * It would be needed if we had two colliding 'overlap_ok'
884 * reservations, so that the second such would not panic on the
885 * overlap with the first. We don't have any such as of this
886 * writing, but might as well tolerate such if it happens in
889 void __init
reserve_early_overlap_ok(u64 start
, u64 end
, char *name
)
891 drop_overlaps_that_are_ok(start
, end
);
892 __reserve_early(start
, end
, name
, 1);
896 * Most early reservations come here.
898 * We first have drop_overlaps_that_are_ok() drop any pre-existing
899 * 'overlap_ok' ranges, so that we can then reserve this memory
900 * range without risk of panic'ing on an overlapping overlap_ok
903 void __init
reserve_early(u64 start
, u64 end
, char *name
)
908 drop_overlaps_that_are_ok(start
, end
);
909 __reserve_early(start
, end
, name
, 0);
912 void __init
free_early(u64 start
, u64 end
)
917 i
= find_overlapped_early(start
, end
);
919 if (i
>= MAX_EARLY_RES
|| r
->end
!= end
|| r
->start
!= start
)
920 panic("free_early on not reserved area: %llx-%llx!",
926 void __init
early_res_to_bootmem(u64 start
, u64 end
)
929 u64 final_start
, final_end
;
932 for (i
= 0; i
< MAX_EARLY_RES
&& early_res
[i
].end
; i
++)
935 printk(KERN_INFO
"(%d early reservations) ==> bootmem [%010llx - %010llx]\n",
937 for (i
= 0; i
< count
; i
++) {
938 struct early_res
*r
= &early_res
[i
];
939 printk(KERN_INFO
" #%d [%010llx - %010llx] %16s", i
,
940 r
->start
, r
->end
, r
->name
);
941 final_start
= max(start
, r
->start
);
942 final_end
= min(end
, r
->end
);
943 if (final_start
>= final_end
) {
944 printk(KERN_CONT
"\n");
947 printk(KERN_CONT
" ==> [%010llx - %010llx]\n",
948 final_start
, final_end
);
949 reserve_bootmem_generic(final_start
, final_end
- final_start
,
954 /* Check for already reserved areas */
955 static inline int __init
bad_addr(u64
*addrp
, u64 size
, u64 align
)
962 i
= find_overlapped_early(addr
, addr
+ size
);
964 if (i
< MAX_EARLY_RES
&& r
->end
) {
965 *addrp
= addr
= round_up(r
->end
, align
);
972 /* Check for already reserved areas */
973 static inline int __init
bad_addr_size(u64
*addrp
, u64
*sizep
, u64 align
)
976 u64 addr
= *addrp
, last
;
981 for (i
= 0; i
< MAX_EARLY_RES
&& early_res
[i
].end
; i
++) {
982 struct early_res
*r
= &early_res
[i
];
983 if (last
> r
->start
&& addr
< r
->start
) {
984 size
= r
->start
- addr
;
988 if (last
> r
->end
&& addr
< r
->end
) {
989 addr
= round_up(r
->end
, align
);
994 if (last
<= r
->end
&& addr
>= r
->start
) {
1007 * Find a free area with specified alignment in a specific range.
1009 u64 __init
find_e820_area(u64 start
, u64 end
, u64 size
, u64 align
)
1013 for (i
= 0; i
< e820
.nr_map
; i
++) {
1014 struct e820entry
*ei
= &e820
.map
[i
];
1018 if (ei
->type
!= E820_RAM
)
1020 addr
= round_up(ei
->addr
, align
);
1021 ei_last
= ei
->addr
+ ei
->size
;
1023 addr
= round_up(start
, align
);
1024 if (addr
>= ei_last
)
1026 while (bad_addr(&addr
, size
, align
) && addr
+size
<= ei_last
)
1039 * Find next free range after *start
1041 u64 __init
find_e820_area_size(u64 start
, u64
*sizep
, u64 align
)
1045 for (i
= 0; i
< e820
.nr_map
; i
++) {
1046 struct e820entry
*ei
= &e820
.map
[i
];
1050 if (ei
->type
!= E820_RAM
)
1052 addr
= round_up(ei
->addr
, align
);
1053 ei_last
= ei
->addr
+ ei
->size
;
1055 addr
= round_up(start
, align
);
1056 if (addr
>= ei_last
)
1058 *sizep
= ei_last
- addr
;
1059 while (bad_addr_size(&addr
, sizep
, align
) &&
1060 addr
+ *sizep
<= ei_last
)
1062 last
= addr
+ *sizep
;
1072 * pre allocated 4k and reserved it in e820
1074 u64 __init
early_reserve_e820(u64 startt
, u64 sizet
, u64 align
)
1080 for (start
= startt
; ; start
+= size
) {
1081 start
= find_e820_area_size(start
, &size
, align
);
1088 #ifdef CONFIG_X86_32
1089 if (start
>= MAXMEM
)
1091 if (start
+ size
> MAXMEM
)
1092 size
= MAXMEM
- start
;
1095 addr
= round_down(start
+ size
- sizet
, align
);
1098 e820_update_range(addr
, sizet
, E820_RAM
, E820_RESERVED
);
1099 e820_update_range_saved(addr
, sizet
, E820_RAM
, E820_RESERVED
);
1100 printk(KERN_INFO
"update e820 for early_reserve_e820\n");
1102 update_e820_saved();
1107 #ifdef CONFIG_X86_32
1108 # ifdef CONFIG_X86_PAE
1109 # define MAX_ARCH_PFN (1ULL<<(36-PAGE_SHIFT))
1111 # define MAX_ARCH_PFN (1ULL<<(32-PAGE_SHIFT))
1113 #else /* CONFIG_X86_32 */
1114 # define MAX_ARCH_PFN MAXMEM>>PAGE_SHIFT
1118 * Find the highest page frame number we have available
1120 static unsigned long __init
e820_end_pfn(unsigned long limit_pfn
, unsigned type
)
1123 unsigned long last_pfn
= 0;
1124 unsigned long max_arch_pfn
= MAX_ARCH_PFN
;
1126 for (i
= 0; i
< e820
.nr_map
; i
++) {
1127 struct e820entry
*ei
= &e820
.map
[i
];
1128 unsigned long start_pfn
;
1129 unsigned long end_pfn
;
1131 if (ei
->type
!= type
)
1134 start_pfn
= ei
->addr
>> PAGE_SHIFT
;
1135 end_pfn
= (ei
->addr
+ ei
->size
) >> PAGE_SHIFT
;
1137 if (start_pfn
>= limit_pfn
)
1139 if (end_pfn
> limit_pfn
) {
1140 last_pfn
= limit_pfn
;
1143 if (end_pfn
> last_pfn
)
1147 if (last_pfn
> max_arch_pfn
)
1148 last_pfn
= max_arch_pfn
;
1150 printk(KERN_INFO
"last_pfn = %#lx max_arch_pfn = %#lx\n",
1151 last_pfn
, max_arch_pfn
);
1154 unsigned long __init
e820_end_of_ram_pfn(void)
1156 return e820_end_pfn(MAX_ARCH_PFN
, E820_RAM
);
1159 unsigned long __init
e820_end_of_low_ram_pfn(void)
1161 return e820_end_pfn(1UL<<(32 - PAGE_SHIFT
), E820_RAM
);
1164 * Finds an active region in the address range from start_pfn to last_pfn and
1165 * returns its range in ei_startpfn and ei_endpfn for the e820 entry.
1167 int __init
e820_find_active_region(const struct e820entry
*ei
,
1168 unsigned long start_pfn
,
1169 unsigned long last_pfn
,
1170 unsigned long *ei_startpfn
,
1171 unsigned long *ei_endpfn
)
1173 u64 align
= PAGE_SIZE
;
1175 *ei_startpfn
= round_up(ei
->addr
, align
) >> PAGE_SHIFT
;
1176 *ei_endpfn
= round_down(ei
->addr
+ ei
->size
, align
) >> PAGE_SHIFT
;
1178 /* Skip map entries smaller than a page */
1179 if (*ei_startpfn
>= *ei_endpfn
)
1182 /* Skip if map is outside the node */
1183 if (ei
->type
!= E820_RAM
|| *ei_endpfn
<= start_pfn
||
1184 *ei_startpfn
>= last_pfn
)
1187 /* Check for overlaps */
1188 if (*ei_startpfn
< start_pfn
)
1189 *ei_startpfn
= start_pfn
;
1190 if (*ei_endpfn
> last_pfn
)
1191 *ei_endpfn
= last_pfn
;
1196 /* Walk the e820 map and register active regions within a node */
1197 void __init
e820_register_active_regions(int nid
, unsigned long start_pfn
,
1198 unsigned long last_pfn
)
1200 unsigned long ei_startpfn
;
1201 unsigned long ei_endpfn
;
1204 for (i
= 0; i
< e820
.nr_map
; i
++)
1205 if (e820_find_active_region(&e820
.map
[i
],
1206 start_pfn
, last_pfn
,
1207 &ei_startpfn
, &ei_endpfn
))
1208 add_active_range(nid
, ei_startpfn
, ei_endpfn
);
1212 * Find the hole size (in bytes) in the memory range.
1213 * @start: starting address of the memory range to scan
1214 * @end: ending address of the memory range to scan
1216 u64 __init
e820_hole_size(u64 start
, u64 end
)
1218 unsigned long start_pfn
= start
>> PAGE_SHIFT
;
1219 unsigned long last_pfn
= end
>> PAGE_SHIFT
;
1220 unsigned long ei_startpfn
, ei_endpfn
, ram
= 0;
1223 for (i
= 0; i
< e820
.nr_map
; i
++) {
1224 if (e820_find_active_region(&e820
.map
[i
],
1225 start_pfn
, last_pfn
,
1226 &ei_startpfn
, &ei_endpfn
))
1227 ram
+= ei_endpfn
- ei_startpfn
;
1229 return end
- start
- ((u64
)ram
<< PAGE_SHIFT
);
1232 static void early_panic(char *msg
)
1238 static int userdef __initdata
;
1240 /* "mem=nopentium" disables the 4MB page tables. */
1241 static int __init
parse_memopt(char *p
)
1248 #ifdef CONFIG_X86_32
1249 if (!strcmp(p
, "nopentium")) {
1250 setup_clear_cpu_cap(X86_FEATURE_PSE
);
1256 mem_size
= memparse(p
, &p
);
1257 e820_remove_range(mem_size
, ULLONG_MAX
- mem_size
, E820_RAM
, 1);
1261 early_param("mem", parse_memopt
);
1263 static int __init
parse_memmap_opt(char *p
)
1266 u64 start_at
, mem_size
;
1271 if (!strncmp(p
, "exactmap", 8)) {
1272 #ifdef CONFIG_CRASH_DUMP
1274 * If we are doing a crash dump, we still need to know
1275 * the real mem size before original memory map is
1278 saved_max_pfn
= e820_end_of_ram_pfn();
1286 mem_size
= memparse(p
, &p
);
1292 start_at
= memparse(p
+1, &p
);
1293 e820_add_region(start_at
, mem_size
, E820_RAM
);
1294 } else if (*p
== '#') {
1295 start_at
= memparse(p
+1, &p
);
1296 e820_add_region(start_at
, mem_size
, E820_ACPI
);
1297 } else if (*p
== '$') {
1298 start_at
= memparse(p
+1, &p
);
1299 e820_add_region(start_at
, mem_size
, E820_RESERVED
);
1301 e820_remove_range(mem_size
, ULLONG_MAX
- mem_size
, E820_RAM
, 1);
1303 return *p
== '\0' ? 0 : -EINVAL
;
1305 early_param("memmap", parse_memmap_opt
);
1307 void __init
finish_e820_parsing(void)
1310 u32 nr
= e820
.nr_map
;
1312 if (sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &nr
) < 0)
1313 early_panic("Invalid user supplied memory map");
1316 printk(KERN_INFO
"user-defined physical RAM map:\n");
1317 e820_print_map("user");
1321 static inline const char *e820_type_to_string(int e820_type
)
1323 switch (e820_type
) {
1324 case E820_RESERVED_KERN
:
1325 case E820_RAM
: return "System RAM";
1326 case E820_ACPI
: return "ACPI Tables";
1327 case E820_NVS
: return "ACPI Non-volatile Storage";
1328 case E820_UNUSABLE
: return "Unusable memory";
1329 default: return "reserved";
1334 * Mark e820 reserved areas as busy for the resource manager.
1336 static struct resource __initdata
*e820_res
;
1337 void __init
e820_reserve_resources(void)
1340 struct resource
*res
;
1343 res
= alloc_bootmem(sizeof(struct resource
) * e820
.nr_map
);
1345 for (i
= 0; i
< e820
.nr_map
; i
++) {
1346 end
= e820
.map
[i
].addr
+ e820
.map
[i
].size
- 1;
1347 if (end
!= (resource_size_t
)end
) {
1351 res
->name
= e820_type_to_string(e820
.map
[i
].type
);
1352 res
->start
= e820
.map
[i
].addr
;
1355 res
->flags
= IORESOURCE_MEM
;
1358 * don't register the region that could be conflicted with
1359 * pci device BAR resource and insert them later in
1360 * pcibios_resource_survey()
1362 if (e820
.map
[i
].type
!= E820_RESERVED
|| res
->start
< (1ULL<<20)) {
1363 res
->flags
|= IORESOURCE_BUSY
;
1364 insert_resource(&iomem_resource
, res
);
1369 for (i
= 0; i
< e820_saved
.nr_map
; i
++) {
1370 struct e820entry
*entry
= &e820_saved
.map
[i
];
1371 firmware_map_add_early(entry
->addr
,
1372 entry
->addr
+ entry
->size
- 1,
1373 e820_type_to_string(entry
->type
));
1377 /* How much should we pad RAM ending depending on where it is? */
1378 static unsigned long ram_alignment(resource_size_t pos
)
1380 unsigned long mb
= pos
>> 20;
1382 /* To 64kB in the first megabyte */
1386 /* To 1MB in the first 16MB */
1390 /* To 64MB for anything above that */
1391 return 64*1024*1024;
1394 #define MAX_RESOURCE_SIZE ((resource_size_t)-1)
1396 void __init
e820_reserve_resources_late(void)
1399 struct resource
*res
;
1402 for (i
= 0; i
< e820
.nr_map
; i
++) {
1403 if (!res
->parent
&& res
->end
)
1404 insert_resource_expand_to_fit(&iomem_resource
, res
);
1409 * Try to bump up RAM regions to reasonable boundaries to
1412 for (i
= 0; i
< e820
.nr_map
; i
++) {
1413 struct e820entry
*entry
= &e820
.map
[i
];
1416 if (entry
->type
!= E820_RAM
)
1418 start
= entry
->addr
+ entry
->size
;
1419 end
= round_up(start
, ram_alignment(start
)) - 1;
1420 if (end
> MAX_RESOURCE_SIZE
)
1421 end
= MAX_RESOURCE_SIZE
;
1424 reserve_region_with_split(&iomem_resource
, start
, end
,
1429 char *__init
default_machine_specific_memory_setup(void)
1431 char *who
= "BIOS-e820";
1434 * Try to copy the BIOS-supplied E820-map.
1436 * Otherwise fake a memory map; one section from 0k->640k,
1437 * the next section from 1mb->appropriate_mem_k
1439 new_nr
= boot_params
.e820_entries
;
1440 sanitize_e820_map(boot_params
.e820_map
,
1441 ARRAY_SIZE(boot_params
.e820_map
),
1443 boot_params
.e820_entries
= new_nr
;
1444 if (append_e820_map(boot_params
.e820_map
, boot_params
.e820_entries
)
1448 /* compare results from other methods and take the greater */
1449 if (boot_params
.alt_mem_k
1450 < boot_params
.screen_info
.ext_mem_k
) {
1451 mem_size
= boot_params
.screen_info
.ext_mem_k
;
1454 mem_size
= boot_params
.alt_mem_k
;
1459 e820_add_region(0, LOWMEMSIZE(), E820_RAM
);
1460 e820_add_region(HIGH_MEMORY
, mem_size
<< 10, E820_RAM
);
1463 /* In case someone cares... */
1467 void __init
setup_memory_map(void)
1471 who
= x86_init
.resources
.memory_setup();
1472 memcpy(&e820_saved
, &e820
, sizeof(struct e820map
));
1473 printk(KERN_INFO
"BIOS-provided physical RAM map:\n");
1474 e820_print_map(who
);