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>
20 #include <asm/proto.h>
21 #include <asm/setup.h>
24 * The e820 map is the map that gets modified e.g. with command line parameters
25 * and that is also registered with modifications in the kernel resource tree
26 * with the iomem_resource as parent.
28 * The e820_saved is directly saved after the BIOS-provided memory map is
29 * copied. It doesn't get modified afterwards. It's registered for the
30 * /sys/firmware/memmap interface.
32 * That memory map is not modified and is used as base for kexec. The kexec'd
33 * kernel should get the same memory map as the firmware provides. Then the
34 * user can e.g. boot the original kernel with mem=1G while still booting the
35 * next kernel with full memory.
38 struct e820map e820_saved
;
40 /* For PCI or other memory-mapped resources */
41 unsigned long pci_mem_start
= 0xaeedbabe;
43 EXPORT_SYMBOL(pci_mem_start
);
47 * This function checks if any part of the range <start,end> is mapped
51 e820_any_mapped(u64 start
, u64 end
, unsigned type
)
55 for (i
= 0; i
< e820
.nr_map
; i
++) {
56 struct e820entry
*ei
= &e820
.map
[i
];
58 if (type
&& ei
->type
!= type
)
60 if (ei
->addr
>= end
|| ei
->addr
+ ei
->size
<= start
)
66 EXPORT_SYMBOL_GPL(e820_any_mapped
);
69 * This function checks if the entire range <start,end> is mapped with type.
71 * Note: this function only works correct if the e820 table is sorted and
72 * not-overlapping, which is the case
74 int __init
e820_all_mapped(u64 start
, u64 end
, unsigned type
)
78 for (i
= 0; i
< e820
.nr_map
; i
++) {
79 struct e820entry
*ei
= &e820
.map
[i
];
81 if (type
&& ei
->type
!= type
)
83 /* is the region (part) in overlap with the current region ?*/
84 if (ei
->addr
>= end
|| ei
->addr
+ ei
->size
<= start
)
87 /* if the region is at the beginning of <start,end> we move
88 * start to the end of the region since it's ok until there
90 if (ei
->addr
<= start
)
91 start
= ei
->addr
+ ei
->size
;
93 * if start is now at or beyond end, we're done, full
103 * Add a memory region to the kernel e820 map.
105 static void __init
__e820_add_region(struct e820map
*e820x
, u64 start
, u64 size
,
108 int x
= e820x
->nr_map
;
110 if (x
>= ARRAY_SIZE(e820x
->map
)) {
111 printk(KERN_ERR
"Ooops! Too many entries in the memory map!\n");
115 e820x
->map
[x
].addr
= start
;
116 e820x
->map
[x
].size
= size
;
117 e820x
->map
[x
].type
= type
;
121 void __init
e820_add_region(u64 start
, u64 size
, int type
)
123 __e820_add_region(&e820
, start
, size
, type
);
126 static void __init
e820_print_type(u32 type
)
130 case E820_RESERVED_KERN
:
131 printk(KERN_CONT
"(usable)");
134 printk(KERN_CONT
"(reserved)");
137 printk(KERN_CONT
"(ACPI data)");
140 printk(KERN_CONT
"(ACPI NVS)");
143 printk(KERN_CONT
"(unusable)");
146 printk(KERN_CONT
"type %u", type
);
151 void __init
e820_print_map(char *who
)
155 for (i
= 0; i
< e820
.nr_map
; i
++) {
156 printk(KERN_INFO
" %s: %016Lx - %016Lx ", who
,
157 (unsigned long long) e820
.map
[i
].addr
,
159 (e820
.map
[i
].addr
+ e820
.map
[i
].size
));
160 e820_print_type(e820
.map
[i
].type
);
161 printk(KERN_CONT
"\n");
166 * Sanitize the BIOS e820 map.
168 * Some e820 responses include overlapping entries. The following
169 * replaces the original e820 map with a new one, removing overlaps,
170 * and resolving conflicting memory types in favor of highest
173 * The input parameter biosmap points to an array of 'struct
174 * e820entry' which on entry has elements in the range [0, *pnr_map)
175 * valid, and which has space for up to max_nr_map entries.
176 * On return, the resulting sanitized e820 map entries will be in
177 * overwritten in the same location, starting at biosmap.
179 * The integer pointed to by pnr_map must be valid on entry (the
180 * current number of valid entries located at biosmap) and will
181 * be updated on return, with the new number of valid entries
182 * (something no more than max_nr_map.)
184 * The return value from sanitize_e820_map() is zero if it
185 * successfully 'sanitized' the map entries passed in, and is -1
186 * if it did nothing, which can happen if either of (1) it was
187 * only passed one map entry, or (2) any of the input map entries
188 * were invalid (start + size < start, meaning that the size was
189 * so big the described memory range wrapped around through zero.)
191 * Visually we're performing the following
192 * (1,2,3,4 = memory types)...
194 * Sample memory map (w/overlaps):
195 * ____22__________________
196 * ______________________4_
197 * ____1111________________
198 * _44_____________________
199 * 11111111________________
200 * ____________________33__
201 * ___________44___________
202 * __________33333_________
203 * ______________22________
204 * ___________________2222_
205 * _________111111111______
206 * _____________________11_
207 * _________________4______
209 * Sanitized equivalent (no overlap):
210 * 1_______________________
211 * _44_____________________
212 * ___1____________________
213 * ____22__________________
214 * ______11________________
215 * _________1______________
216 * __________3_____________
217 * ___________44___________
218 * _____________33_________
219 * _______________2________
220 * ________________1_______
221 * _________________4______
222 * ___________________2____
223 * ____________________33__
224 * ______________________4_
227 int __init
sanitize_e820_map(struct e820entry
*biosmap
, int max_nr_map
,
230 struct change_member
{
231 struct e820entry
*pbios
; /* pointer to original bios entry */
232 unsigned long long addr
; /* address for this change point */
234 static struct change_member change_point_list
[2*E820_X_MAX
] __initdata
;
235 static struct change_member
*change_point
[2*E820_X_MAX
] __initdata
;
236 static struct e820entry
*overlap_list
[E820_X_MAX
] __initdata
;
237 static struct e820entry new_bios
[E820_X_MAX
] __initdata
;
238 struct change_member
*change_tmp
;
239 unsigned long current_type
, last_type
;
240 unsigned long long last_addr
;
241 int chgidx
, still_changing
;
244 int old_nr
, new_nr
, chg_nr
;
247 /* if there's only one memory region, don't bother */
252 BUG_ON(old_nr
> max_nr_map
);
254 /* bail out if we find any unreasonable addresses in bios map */
255 for (i
= 0; i
< old_nr
; i
++)
256 if (biosmap
[i
].addr
+ biosmap
[i
].size
< biosmap
[i
].addr
)
259 /* create pointers for initial change-point information (for sorting) */
260 for (i
= 0; i
< 2 * old_nr
; i
++)
261 change_point
[i
] = &change_point_list
[i
];
263 /* record all known change-points (starting and ending addresses),
264 omitting those that are for empty memory regions */
266 for (i
= 0; i
< old_nr
; i
++) {
267 if (biosmap
[i
].size
!= 0) {
268 change_point
[chgidx
]->addr
= biosmap
[i
].addr
;
269 change_point
[chgidx
++]->pbios
= &biosmap
[i
];
270 change_point
[chgidx
]->addr
= biosmap
[i
].addr
+
272 change_point
[chgidx
++]->pbios
= &biosmap
[i
];
277 /* sort change-point list by memory addresses (low -> high) */
279 while (still_changing
) {
281 for (i
= 1; i
< chg_nr
; i
++) {
282 unsigned long long curaddr
, lastaddr
;
283 unsigned long long curpbaddr
, lastpbaddr
;
285 curaddr
= change_point
[i
]->addr
;
286 lastaddr
= change_point
[i
- 1]->addr
;
287 curpbaddr
= change_point
[i
]->pbios
->addr
;
288 lastpbaddr
= change_point
[i
- 1]->pbios
->addr
;
291 * swap entries, when:
293 * curaddr > lastaddr or
294 * curaddr == lastaddr and curaddr == curpbaddr and
295 * lastaddr != lastpbaddr
297 if (curaddr
< lastaddr
||
298 (curaddr
== lastaddr
&& curaddr
== curpbaddr
&&
299 lastaddr
!= lastpbaddr
)) {
300 change_tmp
= change_point
[i
];
301 change_point
[i
] = change_point
[i
-1];
302 change_point
[i
-1] = change_tmp
;
308 /* create a new bios memory map, removing overlaps */
309 overlap_entries
= 0; /* number of entries in the overlap table */
310 new_bios_entry
= 0; /* index for creating new bios map entries */
311 last_type
= 0; /* start with undefined memory type */
312 last_addr
= 0; /* start with 0 as last starting address */
314 /* loop through change-points, determining affect on the new bios map */
315 for (chgidx
= 0; chgidx
< chg_nr
; chgidx
++) {
316 /* keep track of all overlapping bios entries */
317 if (change_point
[chgidx
]->addr
==
318 change_point
[chgidx
]->pbios
->addr
) {
320 * add map entry to overlap list (> 1 entry
321 * implies an overlap)
323 overlap_list
[overlap_entries
++] =
324 change_point
[chgidx
]->pbios
;
327 * remove entry from list (order independent,
330 for (i
= 0; i
< overlap_entries
; i
++) {
331 if (overlap_list
[i
] ==
332 change_point
[chgidx
]->pbios
)
334 overlap_list
[overlap_entries
-1];
339 * if there are overlapping entries, decide which
340 * "type" to use (larger value takes precedence --
341 * 1=usable, 2,3,4,4+=unusable)
344 for (i
= 0; i
< overlap_entries
; i
++)
345 if (overlap_list
[i
]->type
> current_type
)
346 current_type
= overlap_list
[i
]->type
;
348 * continue building up new bios map based on this
351 if (current_type
!= last_type
) {
352 if (last_type
!= 0) {
353 new_bios
[new_bios_entry
].size
=
354 change_point
[chgidx
]->addr
- last_addr
;
356 * move forward only if the new size
359 if (new_bios
[new_bios_entry
].size
!= 0)
361 * no more space left for new
364 if (++new_bios_entry
>= max_nr_map
)
367 if (current_type
!= 0) {
368 new_bios
[new_bios_entry
].addr
=
369 change_point
[chgidx
]->addr
;
370 new_bios
[new_bios_entry
].type
= current_type
;
371 last_addr
= change_point
[chgidx
]->addr
;
373 last_type
= current_type
;
376 /* retain count for new bios entries */
377 new_nr
= new_bios_entry
;
379 /* copy new bios mapping into original location */
380 memcpy(biosmap
, new_bios
, new_nr
* sizeof(struct e820entry
));
386 static int __init
__append_e820_map(struct e820entry
*biosmap
, int nr_map
)
389 u64 start
= biosmap
->addr
;
390 u64 size
= biosmap
->size
;
391 u64 end
= start
+ size
;
392 u32 type
= biosmap
->type
;
394 /* Overflow in 64 bits? Ignore the memory map. */
398 e820_add_region(start
, size
, type
);
407 * Copy the BIOS e820 map into a safe place.
409 * Sanity-check it while we're at it..
411 * If we're lucky and live on a modern system, the setup code
412 * will have given us a memory map that we can use to properly
413 * set up memory. If we aren't, we'll fake a memory map.
415 static int __init
append_e820_map(struct e820entry
*biosmap
, int nr_map
)
417 /* Only one memory region (or negative)? Ignore it */
421 return __append_e820_map(biosmap
, nr_map
);
424 static u64 __init
__e820_update_range(struct e820map
*e820x
, u64 start
,
425 u64 size
, unsigned old_type
,
430 u64 real_updated_size
= 0;
432 BUG_ON(old_type
== new_type
);
434 if (size
> (ULLONG_MAX
- start
))
435 size
= ULLONG_MAX
- start
;
438 printk(KERN_DEBUG
"e820 update range: %016Lx - %016Lx ",
439 (unsigned long long) start
,
440 (unsigned long long) end
);
441 e820_print_type(old_type
);
442 printk(KERN_CONT
" ==> ");
443 e820_print_type(new_type
);
444 printk(KERN_CONT
"\n");
446 for (i
= 0; i
< e820x
->nr_map
; i
++) {
447 struct e820entry
*ei
= &e820x
->map
[i
];
448 u64 final_start
, final_end
;
451 if (ei
->type
!= old_type
)
454 ei_end
= ei
->addr
+ ei
->size
;
455 /* totally covered by new range? */
456 if (ei
->addr
>= start
&& ei_end
<= end
) {
458 real_updated_size
+= ei
->size
;
462 /* new range is totally covered? */
463 if (ei
->addr
< start
&& ei_end
> end
) {
464 __e820_add_region(e820x
, start
, size
, new_type
);
465 __e820_add_region(e820x
, end
, ei_end
- end
, ei
->type
);
466 ei
->size
= start
- ei
->addr
;
467 real_updated_size
+= size
;
471 /* partially covered */
472 final_start
= max(start
, ei
->addr
);
473 final_end
= min(end
, ei_end
);
474 if (final_start
>= final_end
)
477 __e820_add_region(e820x
, final_start
, final_end
- final_start
,
480 real_updated_size
+= final_end
- final_start
;
483 * left range could be head or tail, so need to update
486 ei
->size
-= final_end
- final_start
;
487 if (ei
->addr
< final_start
)
489 ei
->addr
= final_end
;
491 return real_updated_size
;
494 u64 __init
e820_update_range(u64 start
, u64 size
, unsigned old_type
,
497 return __e820_update_range(&e820
, start
, size
, old_type
, new_type
);
500 static u64 __init
e820_update_range_saved(u64 start
, u64 size
,
501 unsigned old_type
, unsigned new_type
)
503 return __e820_update_range(&e820_saved
, start
, size
, old_type
,
507 /* make e820 not cover the range */
508 u64 __init
e820_remove_range(u64 start
, u64 size
, unsigned old_type
,
513 u64 real_removed_size
= 0;
515 if (size
> (ULLONG_MAX
- start
))
516 size
= ULLONG_MAX
- start
;
519 printk(KERN_DEBUG
"e820 remove range: %016Lx - %016Lx ",
520 (unsigned long long) start
,
521 (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
;
531 if (checktype
&& ei
->type
!= old_type
)
534 ei_end
= ei
->addr
+ ei
->size
;
535 /* totally covered? */
536 if (ei
->addr
>= start
&& ei_end
<= end
) {
537 real_removed_size
+= ei
->size
;
538 memset(ei
, 0, sizeof(struct e820entry
));
542 /* new range is totally covered? */
543 if (ei
->addr
< start
&& ei_end
> end
) {
544 e820_add_region(end
, ei_end
- end
, ei
->type
);
545 ei
->size
= start
- ei
->addr
;
546 real_removed_size
+= size
;
550 /* partially covered */
551 final_start
= max(start
, ei
->addr
);
552 final_end
= min(end
, ei_end
);
553 if (final_start
>= final_end
)
555 real_removed_size
+= final_end
- final_start
;
558 * left range could be head or tail, so need to update
561 ei
->size
-= final_end
- final_start
;
562 if (ei
->addr
< final_start
)
564 ei
->addr
= final_end
;
566 return real_removed_size
;
569 void __init
update_e820(void)
573 nr_map
= e820
.nr_map
;
574 if (sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &nr_map
))
576 e820
.nr_map
= nr_map
;
577 printk(KERN_INFO
"modified physical RAM map:\n");
578 e820_print_map("modified");
580 static void __init
update_e820_saved(void)
584 nr_map
= e820_saved
.nr_map
;
585 if (sanitize_e820_map(e820_saved
.map
, ARRAY_SIZE(e820_saved
.map
), &nr_map
))
587 e820_saved
.nr_map
= nr_map
;
589 #define MAX_GAP_END 0x100000000ull
591 * Search for a gap in the e820 memory space from start_addr to end_addr.
593 __init
int e820_search_gap(unsigned long *gapstart
, unsigned long *gapsize
,
594 unsigned long start_addr
, unsigned long long end_addr
)
596 unsigned long long last
;
600 last
= (end_addr
&& end_addr
< MAX_GAP_END
) ? end_addr
: MAX_GAP_END
;
603 unsigned long long start
= e820
.map
[i
].addr
;
604 unsigned long long end
= start
+ e820
.map
[i
].size
;
606 if (end
< start_addr
)
610 * Since "last" is at most 4GB, we know we'll
611 * fit in 32 bits if this condition is true
614 unsigned long gap
= last
- end
;
616 if (gap
>= *gapsize
) {
629 * Search for the biggest gap in the low 32 bits of the e820
630 * memory space. We pass this space to PCI to assign MMIO resources
631 * for hotplug or unconfigured devices in.
632 * Hopefully the BIOS let enough space left.
634 __init
void e820_setup_gap(void)
636 unsigned long gapstart
, gapsize
;
639 gapstart
= 0x10000000;
641 found
= e820_search_gap(&gapstart
, &gapsize
, 0, MAX_GAP_END
);
645 gapstart
= (max_pfn
<< PAGE_SHIFT
) + 1024*1024;
647 "PCI: Warning: Cannot find a gap in the 32bit address range\n"
648 "PCI: Unassigned devices with 32bit resource registers may break!\n");
653 * e820_reserve_resources_late protect stolen RAM already
655 pci_mem_start
= gapstart
;
658 "Allocating PCI resources starting at %lx (gap: %lx:%lx)\n",
659 pci_mem_start
, gapstart
, gapsize
);
663 * Because of the size limitation of struct boot_params, only first
664 * 128 E820 memory entries are passed to kernel via
665 * boot_params.e820_map, others are passed via SETUP_E820_EXT node of
666 * linked list of struct setup_data, which is parsed here.
668 void __init
parse_e820_ext(struct setup_data
*sdata
, unsigned long pa_data
)
672 struct e820entry
*extmap
;
674 entries
= sdata
->len
/ sizeof(struct e820entry
);
675 map_len
= sdata
->len
+ sizeof(struct setup_data
);
676 if (map_len
> PAGE_SIZE
)
677 sdata
= early_ioremap(pa_data
, map_len
);
678 extmap
= (struct e820entry
*)(sdata
->data
);
679 __append_e820_map(extmap
, entries
);
680 sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &e820
.nr_map
);
681 if (map_len
> PAGE_SIZE
)
682 early_iounmap(sdata
, map_len
);
683 printk(KERN_INFO
"extended physical RAM map:\n");
684 e820_print_map("extended");
687 #if defined(CONFIG_X86_64) || (defined(CONFIG_X86_32) && defined(CONFIG_HIBERNATION))
689 * Find the ranges of physical addresses that do not correspond to
690 * e820 RAM areas and mark the corresponding pages as nosave for
691 * hibernation (32 bit) or software suspend and suspend to RAM (64 bit).
693 * This function requires the e820 map to be sorted and without any
694 * overlapping entries and assumes the first e820 area to be RAM.
696 void __init
e820_mark_nosave_regions(unsigned long limit_pfn
)
701 pfn
= PFN_DOWN(e820
.map
[0].addr
+ e820
.map
[0].size
);
702 for (i
= 1; i
< e820
.nr_map
; i
++) {
703 struct e820entry
*ei
= &e820
.map
[i
];
705 if (pfn
< PFN_UP(ei
->addr
))
706 register_nosave_region(pfn
, PFN_UP(ei
->addr
));
708 pfn
= PFN_DOWN(ei
->addr
+ ei
->size
);
709 if (ei
->type
!= E820_RAM
&& ei
->type
!= E820_RESERVED_KERN
)
710 register_nosave_region(PFN_UP(ei
->addr
), pfn
);
712 if (pfn
>= limit_pfn
)
718 #ifdef CONFIG_HIBERNATION
720 * Mark ACPI NVS memory region, so that we can save/restore it during
721 * hibernation and the subsequent resume.
723 static int __init
e820_mark_nvs_memory(void)
727 for (i
= 0; i
< e820
.nr_map
; i
++) {
728 struct e820entry
*ei
= &e820
.map
[i
];
730 if (ei
->type
== E820_NVS
)
731 suspend_nvs_register(ei
->addr
, ei
->size
);
736 core_initcall(e820_mark_nvs_memory
);
740 * Find a free area with specified alignment in a specific range.
742 u64 __init
find_e820_area(u64 start
, u64 end
, u64 size
, u64 align
)
746 for (i
= 0; i
< e820
.nr_map
; i
++) {
747 struct e820entry
*ei
= &e820
.map
[i
];
749 u64 ei_start
, ei_last
;
751 if (ei
->type
!= E820_RAM
)
754 ei_last
= ei
->addr
+ ei
->size
;
756 addr
= find_early_area(ei_start
, ei_last
, start
, end
,
765 u64 __init
find_fw_memmap_area(u64 start
, u64 end
, u64 size
, u64 align
)
767 return find_e820_area(start
, end
, size
, align
);
770 u64 __init
get_max_mapped(void)
772 u64 end
= max_pfn_mapped
;
779 * Find next free range after *start
781 u64 __init
find_e820_area_size(u64 start
, u64
*sizep
, u64 align
)
785 for (i
= 0; i
< e820
.nr_map
; i
++) {
786 struct e820entry
*ei
= &e820
.map
[i
];
788 u64 ei_start
, ei_last
;
790 if (ei
->type
!= E820_RAM
)
793 ei_last
= ei
->addr
+ ei
->size
;
795 addr
= find_early_area_size(ei_start
, ei_last
, start
,
806 * pre allocated 4k and reserved it in e820
808 u64 __init
early_reserve_e820(u64 startt
, u64 sizet
, u64 align
)
814 for (start
= startt
; ; start
+= size
) {
815 start
= find_e820_area_size(start
, &size
, align
);
825 if (start
+ size
> MAXMEM
)
826 size
= MAXMEM
- start
;
829 addr
= round_down(start
+ size
- sizet
, align
);
832 e820_update_range(addr
, sizet
, E820_RAM
, E820_RESERVED
);
833 e820_update_range_saved(addr
, sizet
, E820_RAM
, E820_RESERVED
);
834 printk(KERN_INFO
"update e820 for early_reserve_e820\n");
842 # ifdef CONFIG_X86_PAE
843 # define MAX_ARCH_PFN (1ULL<<(36-PAGE_SHIFT))
845 # define MAX_ARCH_PFN (1ULL<<(32-PAGE_SHIFT))
847 #else /* CONFIG_X86_32 */
848 # define MAX_ARCH_PFN MAXMEM>>PAGE_SHIFT
852 * Find the highest page frame number we have available
854 static unsigned long __init
e820_end_pfn(unsigned long limit_pfn
, unsigned type
)
857 unsigned long last_pfn
= 0;
858 unsigned long max_arch_pfn
= MAX_ARCH_PFN
;
860 for (i
= 0; i
< e820
.nr_map
; i
++) {
861 struct e820entry
*ei
= &e820
.map
[i
];
862 unsigned long start_pfn
;
863 unsigned long end_pfn
;
865 if (ei
->type
!= type
)
868 start_pfn
= ei
->addr
>> PAGE_SHIFT
;
869 end_pfn
= (ei
->addr
+ ei
->size
) >> PAGE_SHIFT
;
871 if (start_pfn
>= limit_pfn
)
873 if (end_pfn
> limit_pfn
) {
874 last_pfn
= limit_pfn
;
877 if (end_pfn
> last_pfn
)
881 if (last_pfn
> max_arch_pfn
)
882 last_pfn
= max_arch_pfn
;
884 printk(KERN_INFO
"last_pfn = %#lx max_arch_pfn = %#lx\n",
885 last_pfn
, max_arch_pfn
);
888 unsigned long __init
e820_end_of_ram_pfn(void)
890 return e820_end_pfn(MAX_ARCH_PFN
, E820_RAM
);
893 unsigned long __init
e820_end_of_low_ram_pfn(void)
895 return e820_end_pfn(1UL<<(32 - PAGE_SHIFT
), E820_RAM
);
898 * Finds an active region in the address range from start_pfn to last_pfn and
899 * returns its range in ei_startpfn and ei_endpfn for the e820 entry.
901 int __init
e820_find_active_region(const struct e820entry
*ei
,
902 unsigned long start_pfn
,
903 unsigned long last_pfn
,
904 unsigned long *ei_startpfn
,
905 unsigned long *ei_endpfn
)
907 u64 align
= PAGE_SIZE
;
909 *ei_startpfn
= round_up(ei
->addr
, align
) >> PAGE_SHIFT
;
910 *ei_endpfn
= round_down(ei
->addr
+ ei
->size
, align
) >> PAGE_SHIFT
;
912 /* Skip map entries smaller than a page */
913 if (*ei_startpfn
>= *ei_endpfn
)
916 /* Skip if map is outside the node */
917 if (ei
->type
!= E820_RAM
|| *ei_endpfn
<= start_pfn
||
918 *ei_startpfn
>= last_pfn
)
921 /* Check for overlaps */
922 if (*ei_startpfn
< start_pfn
)
923 *ei_startpfn
= start_pfn
;
924 if (*ei_endpfn
> last_pfn
)
925 *ei_endpfn
= last_pfn
;
930 /* Walk the e820 map and register active regions within a node */
931 void __init
e820_register_active_regions(int nid
, unsigned long start_pfn
,
932 unsigned long last_pfn
)
934 unsigned long ei_startpfn
;
935 unsigned long ei_endpfn
;
938 for (i
= 0; i
< e820
.nr_map
; i
++)
939 if (e820_find_active_region(&e820
.map
[i
],
941 &ei_startpfn
, &ei_endpfn
))
942 add_active_range(nid
, ei_startpfn
, ei_endpfn
);
946 * Find the hole size (in bytes) in the memory range.
947 * @start: starting address of the memory range to scan
948 * @end: ending address of the memory range to scan
950 u64 __init
e820_hole_size(u64 start
, u64 end
)
952 unsigned long start_pfn
= start
>> PAGE_SHIFT
;
953 unsigned long last_pfn
= end
>> PAGE_SHIFT
;
954 unsigned long ei_startpfn
, ei_endpfn
, ram
= 0;
957 for (i
= 0; i
< e820
.nr_map
; i
++) {
958 if (e820_find_active_region(&e820
.map
[i
],
960 &ei_startpfn
, &ei_endpfn
))
961 ram
+= ei_endpfn
- ei_startpfn
;
963 return end
- start
- ((u64
)ram
<< PAGE_SHIFT
);
966 static void early_panic(char *msg
)
972 static int userdef __initdata
;
974 /* "mem=nopentium" disables the 4MB page tables. */
975 static int __init
parse_memopt(char *p
)
983 if (!strcmp(p
, "nopentium")) {
984 setup_clear_cpu_cap(X86_FEATURE_PSE
);
990 mem_size
= memparse(p
, &p
);
991 e820_remove_range(mem_size
, ULLONG_MAX
- mem_size
, E820_RAM
, 1);
995 early_param("mem", parse_memopt
);
997 static int __init
parse_memmap_opt(char *p
)
1000 u64 start_at
, mem_size
;
1005 if (!strncmp(p
, "exactmap", 8)) {
1006 #ifdef CONFIG_CRASH_DUMP
1008 * If we are doing a crash dump, we still need to know
1009 * the real mem size before original memory map is
1012 saved_max_pfn
= e820_end_of_ram_pfn();
1020 mem_size
= memparse(p
, &p
);
1026 start_at
= memparse(p
+1, &p
);
1027 e820_add_region(start_at
, mem_size
, E820_RAM
);
1028 } else if (*p
== '#') {
1029 start_at
= memparse(p
+1, &p
);
1030 e820_add_region(start_at
, mem_size
, E820_ACPI
);
1031 } else if (*p
== '$') {
1032 start_at
= memparse(p
+1, &p
);
1033 e820_add_region(start_at
, mem_size
, E820_RESERVED
);
1035 e820_remove_range(mem_size
, ULLONG_MAX
- mem_size
, E820_RAM
, 1);
1037 return *p
== '\0' ? 0 : -EINVAL
;
1039 early_param("memmap", parse_memmap_opt
);
1041 void __init
finish_e820_parsing(void)
1044 u32 nr
= e820
.nr_map
;
1046 if (sanitize_e820_map(e820
.map
, ARRAY_SIZE(e820
.map
), &nr
) < 0)
1047 early_panic("Invalid user supplied memory map");
1050 printk(KERN_INFO
"user-defined physical RAM map:\n");
1051 e820_print_map("user");
1055 static inline const char *e820_type_to_string(int e820_type
)
1057 switch (e820_type
) {
1058 case E820_RESERVED_KERN
:
1059 case E820_RAM
: return "System RAM";
1060 case E820_ACPI
: return "ACPI Tables";
1061 case E820_NVS
: return "ACPI Non-volatile Storage";
1062 case E820_UNUSABLE
: return "Unusable memory";
1063 default: return "reserved";
1068 * Mark e820 reserved areas as busy for the resource manager.
1070 static struct resource __initdata
*e820_res
;
1071 void __init
e820_reserve_resources(void)
1074 struct resource
*res
;
1077 res
= alloc_bootmem(sizeof(struct resource
) * e820
.nr_map
);
1079 for (i
= 0; i
< e820
.nr_map
; i
++) {
1080 end
= e820
.map
[i
].addr
+ e820
.map
[i
].size
- 1;
1081 if (end
!= (resource_size_t
)end
) {
1085 res
->name
= e820_type_to_string(e820
.map
[i
].type
);
1086 res
->start
= e820
.map
[i
].addr
;
1089 res
->flags
= IORESOURCE_MEM
;
1092 * don't register the region that could be conflicted with
1093 * pci device BAR resource and insert them later in
1094 * pcibios_resource_survey()
1096 if (e820
.map
[i
].type
!= E820_RESERVED
|| res
->start
< (1ULL<<20)) {
1097 res
->flags
|= IORESOURCE_BUSY
;
1098 insert_resource(&iomem_resource
, res
);
1103 for (i
= 0; i
< e820_saved
.nr_map
; i
++) {
1104 struct e820entry
*entry
= &e820_saved
.map
[i
];
1105 firmware_map_add_early(entry
->addr
,
1106 entry
->addr
+ entry
->size
- 1,
1107 e820_type_to_string(entry
->type
));
1111 /* How much should we pad RAM ending depending on where it is? */
1112 static unsigned long ram_alignment(resource_size_t pos
)
1114 unsigned long mb
= pos
>> 20;
1116 /* To 64kB in the first megabyte */
1120 /* To 1MB in the first 16MB */
1124 /* To 64MB for anything above that */
1125 return 64*1024*1024;
1128 #define MAX_RESOURCE_SIZE ((resource_size_t)-1)
1130 void __init
e820_reserve_resources_late(void)
1133 struct resource
*res
;
1136 for (i
= 0; i
< e820
.nr_map
; i
++) {
1137 if (!res
->parent
&& res
->end
)
1138 insert_resource_expand_to_fit(&iomem_resource
, res
);
1143 * Try to bump up RAM regions to reasonable boundaries to
1146 for (i
= 0; i
< e820
.nr_map
; i
++) {
1147 struct e820entry
*entry
= &e820
.map
[i
];
1150 if (entry
->type
!= E820_RAM
)
1152 start
= entry
->addr
+ entry
->size
;
1153 end
= round_up(start
, ram_alignment(start
)) - 1;
1154 if (end
> MAX_RESOURCE_SIZE
)
1155 end
= MAX_RESOURCE_SIZE
;
1158 printk(KERN_DEBUG
"reserve RAM buffer: %016llx - %016llx ",
1160 reserve_region_with_split(&iomem_resource
, start
, end
,
1165 char *__init
default_machine_specific_memory_setup(void)
1167 char *who
= "BIOS-e820";
1170 * Try to copy the BIOS-supplied E820-map.
1172 * Otherwise fake a memory map; one section from 0k->640k,
1173 * the next section from 1mb->appropriate_mem_k
1175 new_nr
= boot_params
.e820_entries
;
1176 sanitize_e820_map(boot_params
.e820_map
,
1177 ARRAY_SIZE(boot_params
.e820_map
),
1179 boot_params
.e820_entries
= new_nr
;
1180 if (append_e820_map(boot_params
.e820_map
, boot_params
.e820_entries
)
1184 /* compare results from other methods and take the greater */
1185 if (boot_params
.alt_mem_k
1186 < boot_params
.screen_info
.ext_mem_k
) {
1187 mem_size
= boot_params
.screen_info
.ext_mem_k
;
1190 mem_size
= boot_params
.alt_mem_k
;
1195 e820_add_region(0, LOWMEMSIZE(), E820_RAM
);
1196 e820_add_region(HIGH_MEMORY
, mem_size
<< 10, E820_RAM
);
1199 /* In case someone cares... */
1203 void __init
setup_memory_map(void)
1207 who
= x86_init
.resources
.memory_setup();
1208 memcpy(&e820_saved
, &e820
, sizeof(struct e820map
));
1209 printk(KERN_INFO
"BIOS-provided physical RAM map:\n");
1210 e820_print_map(who
);