Commit | Line | Data |
---|---|---|
269c2d81 | 1 | #include <linux/kernel.h> |
2 | #include <linux/types.h> | |
3 | #include <linux/init.h> | |
4 | #include <linux/bootmem.h> | |
5 | #include <linux/ioport.h> | |
6 | #include <linux/string.h> | |
7 | #include <linux/kexec.h> | |
8 | #include <linux/module.h> | |
9 | #include <linux/mm.h> | |
10 | #include <linux/efi.h> | |
b2dff6a8 | 11 | #include <linux/pfn.h> |
bd472c79 | 12 | #include <linux/uaccess.h> |
1c10070a | 13 | #include <linux/suspend.h> |
269c2d81 | 14 | |
15 | #include <asm/pgtable.h> | |
16 | #include <asm/page.h> | |
17 | #include <asm/e820.h> | |
90611fe9 | 18 | #include <asm/setup.h> |
269c2d81 | 19 | |
20 | #ifdef CONFIG_EFI | |
21 | int efi_enabled = 0; | |
22 | EXPORT_SYMBOL(efi_enabled); | |
23 | #endif | |
24 | ||
25 | struct e820map e820; | |
8e3342f7 | 26 | struct change_member { |
27 | struct e820entry *pbios; /* pointer to original bios entry */ | |
28 | unsigned long long addr; /* address for this change point */ | |
29 | }; | |
30 | static struct change_member change_point_list[2*E820MAX] __initdata; | |
31 | static struct change_member *change_point[2*E820MAX] __initdata; | |
32 | static struct e820entry *overlap_list[E820MAX] __initdata; | |
33 | static struct e820entry new_bios[E820MAX] __initdata; | |
b5b24057 | 34 | /* For PCI or other memory-mapped resources */ |
35 | unsigned long pci_mem_start = 0x10000000; | |
36 | #ifdef CONFIG_PCI | |
37 | EXPORT_SYMBOL(pci_mem_start); | |
38 | #endif | |
cef518e8 | 39 | extern int user_defined_memmap; |
269c2d81 | 40 | struct resource data_resource = { |
41 | .name = "Kernel data", | |
42 | .start = 0, | |
43 | .end = 0, | |
44 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
45 | }; | |
46 | ||
47 | struct resource code_resource = { | |
48 | .name = "Kernel code", | |
49 | .start = 0, | |
50 | .end = 0, | |
51 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
52 | }; | |
53 | ||
54 | static struct resource system_rom_resource = { | |
55 | .name = "System ROM", | |
56 | .start = 0xf0000, | |
57 | .end = 0xfffff, | |
58 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
59 | }; | |
60 | ||
61 | static struct resource extension_rom_resource = { | |
62 | .name = "Extension ROM", | |
63 | .start = 0xe0000, | |
64 | .end = 0xeffff, | |
65 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
66 | }; | |
67 | ||
68 | static struct resource adapter_rom_resources[] = { { | |
69 | .name = "Adapter ROM", | |
70 | .start = 0xc8000, | |
71 | .end = 0, | |
72 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
73 | }, { | |
74 | .name = "Adapter ROM", | |
75 | .start = 0, | |
76 | .end = 0, | |
77 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
78 | }, { | |
79 | .name = "Adapter ROM", | |
80 | .start = 0, | |
81 | .end = 0, | |
82 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
83 | }, { | |
84 | .name = "Adapter ROM", | |
85 | .start = 0, | |
86 | .end = 0, | |
87 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
88 | }, { | |
89 | .name = "Adapter ROM", | |
90 | .start = 0, | |
91 | .end = 0, | |
92 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
93 | }, { | |
94 | .name = "Adapter ROM", | |
95 | .start = 0, | |
96 | .end = 0, | |
97 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
98 | } }; | |
99 | ||
100 | static struct resource video_rom_resource = { | |
101 | .name = "Video ROM", | |
102 | .start = 0xc0000, | |
103 | .end = 0xc7fff, | |
104 | .flags = IORESOURCE_BUSY | IORESOURCE_READONLY | IORESOURCE_MEM | |
105 | }; | |
106 | ||
107 | static struct resource video_ram_resource = { | |
108 | .name = "Video RAM area", | |
109 | .start = 0xa0000, | |
110 | .end = 0xbffff, | |
111 | .flags = IORESOURCE_BUSY | IORESOURCE_MEM | |
112 | }; | |
113 | ||
114 | static struct resource standard_io_resources[] = { { | |
115 | .name = "dma1", | |
116 | .start = 0x0000, | |
117 | .end = 0x001f, | |
118 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
119 | }, { | |
120 | .name = "pic1", | |
121 | .start = 0x0020, | |
122 | .end = 0x0021, | |
123 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
124 | }, { | |
125 | .name = "timer0", | |
126 | .start = 0x0040, | |
127 | .end = 0x0043, | |
128 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
129 | }, { | |
130 | .name = "timer1", | |
131 | .start = 0x0050, | |
132 | .end = 0x0053, | |
133 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
134 | }, { | |
135 | .name = "keyboard", | |
136 | .start = 0x0060, | |
137 | .end = 0x006f, | |
138 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
139 | }, { | |
140 | .name = "dma page reg", | |
141 | .start = 0x0080, | |
142 | .end = 0x008f, | |
143 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
144 | }, { | |
145 | .name = "pic2", | |
146 | .start = 0x00a0, | |
147 | .end = 0x00a1, | |
148 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
149 | }, { | |
150 | .name = "dma2", | |
151 | .start = 0x00c0, | |
152 | .end = 0x00df, | |
153 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
154 | }, { | |
155 | .name = "fpu", | |
156 | .start = 0x00f0, | |
157 | .end = 0x00ff, | |
158 | .flags = IORESOURCE_BUSY | IORESOURCE_IO | |
159 | } }; | |
160 | ||
3b3d5e1d RH |
161 | #define ROMSIGNATURE 0xaa55 |
162 | ||
163 | static int __init romsignature(const unsigned char *rom) | |
bd472c79 | 164 | { |
0adad171 | 165 | const unsigned short * const ptr = (const unsigned short *)rom; |
bd472c79 | 166 | unsigned short sig; |
3b3d5e1d | 167 | |
0adad171 | 168 | return probe_kernel_address(ptr, sig) == 0 && sig == ROMSIGNATURE; |
bd472c79 | 169 | } |
269c2d81 | 170 | |
0adad171 | 171 | static int __init romchecksum(const unsigned char *rom, unsigned long length) |
269c2d81 | 172 | { |
0adad171 | 173 | unsigned char sum, c; |
269c2d81 | 174 | |
0adad171 RH |
175 | for (sum = 0; length && probe_kernel_address(rom++, c) == 0; length--) |
176 | sum += c; | |
177 | return !length && !sum; | |
269c2d81 | 178 | } |
179 | ||
180 | static void __init probe_roms(void) | |
181 | { | |
0adad171 | 182 | const unsigned char *rom; |
269c2d81 | 183 | unsigned long start, length, upper; |
0adad171 RH |
184 | unsigned char c; |
185 | int i; | |
269c2d81 | 186 | |
187 | /* video rom */ | |
188 | upper = adapter_rom_resources[0].start; | |
189 | for (start = video_rom_resource.start; start < upper; start += 2048) { | |
190 | rom = isa_bus_to_virt(start); | |
191 | if (!romsignature(rom)) | |
192 | continue; | |
193 | ||
194 | video_rom_resource.start = start; | |
195 | ||
0adad171 RH |
196 | if (probe_kernel_address(rom + 2, c) != 0) |
197 | continue; | |
198 | ||
269c2d81 | 199 | /* 0 < length <= 0x7f * 512, historically */ |
0adad171 | 200 | length = c * 512; |
269c2d81 | 201 | |
202 | /* if checksum okay, trust length byte */ | |
203 | if (length && romchecksum(rom, length)) | |
204 | video_rom_resource.end = start + length - 1; | |
205 | ||
206 | request_resource(&iomem_resource, &video_rom_resource); | |
207 | break; | |
208 | } | |
209 | ||
210 | start = (video_rom_resource.end + 1 + 2047) & ~2047UL; | |
211 | if (start < upper) | |
212 | start = upper; | |
213 | ||
214 | /* system rom */ | |
215 | request_resource(&iomem_resource, &system_rom_resource); | |
216 | upper = system_rom_resource.start; | |
217 | ||
218 | /* check for extension rom (ignore length byte!) */ | |
219 | rom = isa_bus_to_virt(extension_rom_resource.start); | |
220 | if (romsignature(rom)) { | |
221 | length = extension_rom_resource.end - extension_rom_resource.start + 1; | |
222 | if (romchecksum(rom, length)) { | |
223 | request_resource(&iomem_resource, &extension_rom_resource); | |
224 | upper = extension_rom_resource.start; | |
225 | } | |
226 | } | |
227 | ||
228 | /* check for adapter roms on 2k boundaries */ | |
229 | for (i = 0; i < ARRAY_SIZE(adapter_rom_resources) && start < upper; start += 2048) { | |
230 | rom = isa_bus_to_virt(start); | |
231 | if (!romsignature(rom)) | |
232 | continue; | |
233 | ||
0adad171 RH |
234 | if (probe_kernel_address(rom + 2, c) != 0) |
235 | continue; | |
236 | ||
269c2d81 | 237 | /* 0 < length <= 0x7f * 512, historically */ |
0adad171 | 238 | length = c * 512; |
269c2d81 | 239 | |
240 | /* but accept any length that fits if checksum okay */ | |
241 | if (!length || start + length > upper || !romchecksum(rom, length)) | |
242 | continue; | |
243 | ||
244 | adapter_rom_resources[i].start = start; | |
245 | adapter_rom_resources[i].end = start + length - 1; | |
246 | request_resource(&iomem_resource, &adapter_rom_resources[i]); | |
247 | ||
248 | start = adapter_rom_resources[i++].end & ~2047UL; | |
249 | } | |
250 | } | |
251 | ||
252 | /* | |
253 | * Request address space for all standard RAM and ROM resources | |
254 | * and also for regions reported as reserved by the e820. | |
255 | */ | |
256 | static void __init | |
257 | legacy_init_iomem_resources(struct resource *code_resource, struct resource *data_resource) | |
258 | { | |
259 | int i; | |
260 | ||
261 | probe_roms(); | |
262 | for (i = 0; i < e820.nr_map; i++) { | |
263 | struct resource *res; | |
264 | #ifndef CONFIG_RESOURCES_64BIT | |
265 | if (e820.map[i].addr + e820.map[i].size > 0x100000000ULL) | |
266 | continue; | |
267 | #endif | |
268 | res = kzalloc(sizeof(struct resource), GFP_ATOMIC); | |
269 | switch (e820.map[i].type) { | |
270 | case E820_RAM: res->name = "System RAM"; break; | |
271 | case E820_ACPI: res->name = "ACPI Tables"; break; | |
272 | case E820_NVS: res->name = "ACPI Non-volatile Storage"; break; | |
273 | default: res->name = "reserved"; | |
274 | } | |
275 | res->start = e820.map[i].addr; | |
276 | res->end = res->start + e820.map[i].size - 1; | |
277 | res->flags = IORESOURCE_MEM | IORESOURCE_BUSY; | |
278 | if (request_resource(&iomem_resource, res)) { | |
279 | kfree(res); | |
280 | continue; | |
281 | } | |
282 | if (e820.map[i].type == E820_RAM) { | |
283 | /* | |
284 | * We don't know which RAM region contains kernel data, | |
285 | * so we try it repeatedly and let the resource manager | |
286 | * test it. | |
287 | */ | |
288 | request_resource(res, code_resource); | |
289 | request_resource(res, data_resource); | |
290 | #ifdef CONFIG_KEXEC | |
291 | request_resource(res, &crashk_res); | |
292 | #endif | |
293 | } | |
294 | } | |
295 | } | |
296 | ||
297 | /* | |
298 | * Request address space for all standard resources | |
299 | * | |
300 | * This is called just before pcibios_init(), which is also a | |
301 | * subsys_initcall, but is linked in later (in arch/i386/pci/common.c). | |
302 | */ | |
303 | static int __init request_standard_resources(void) | |
304 | { | |
305 | int i; | |
306 | ||
307 | printk("Setting up standard PCI resources\n"); | |
308 | if (efi_enabled) | |
309 | efi_initialize_iomem_resources(&code_resource, &data_resource); | |
310 | else | |
311 | legacy_init_iomem_resources(&code_resource, &data_resource); | |
312 | ||
313 | /* EFI systems may still have VGA */ | |
314 | request_resource(&iomem_resource, &video_ram_resource); | |
315 | ||
316 | /* request I/O space for devices used on all i[345]86 PCs */ | |
317 | for (i = 0; i < ARRAY_SIZE(standard_io_resources); i++) | |
318 | request_resource(&ioport_resource, &standard_io_resources[i]); | |
319 | return 0; | |
320 | } | |
321 | ||
322 | subsys_initcall(request_standard_resources); | |
8e3342f7 | 323 | |
b0cb1a19 | 324 | #if defined(CONFIG_PM) && defined(CONFIG_HIBERNATION) |
1c10070a RW |
325 | /** |
326 | * e820_mark_nosave_regions - Find the ranges of physical addresses that do not | |
327 | * correspond to e820 RAM areas and mark the corresponding pages as nosave for | |
328 | * hibernation. | |
329 | * | |
330 | * This function requires the e820 map to be sorted and without any | |
331 | * overlapping entries and assumes the first e820 area to be RAM. | |
332 | */ | |
333 | void __init e820_mark_nosave_regions(void) | |
334 | { | |
335 | int i; | |
336 | unsigned long pfn; | |
337 | ||
338 | pfn = PFN_DOWN(e820.map[0].addr + e820.map[0].size); | |
339 | for (i = 1; i < e820.nr_map; i++) { | |
340 | struct e820entry *ei = &e820.map[i]; | |
341 | ||
342 | if (pfn < PFN_UP(ei->addr)) | |
343 | register_nosave_region(pfn, PFN_UP(ei->addr)); | |
344 | ||
345 | pfn = PFN_DOWN(ei->addr + ei->size); | |
346 | if (ei->type != E820_RAM) | |
347 | register_nosave_region(PFN_UP(ei->addr), pfn); | |
348 | ||
349 | if (pfn >= max_low_pfn) | |
350 | break; | |
351 | } | |
352 | } | |
353 | #endif | |
354 | ||
8e3342f7 | 355 | void __init add_memory_region(unsigned long long start, |
356 | unsigned long long size, int type) | |
357 | { | |
358 | int x; | |
359 | ||
360 | if (!efi_enabled) { | |
361 | x = e820.nr_map; | |
362 | ||
363 | if (x == E820MAX) { | |
364 | printk(KERN_ERR "Ooops! Too many entries in the memory map!\n"); | |
365 | return; | |
366 | } | |
367 | ||
368 | e820.map[x].addr = start; | |
369 | e820.map[x].size = size; | |
370 | e820.map[x].type = type; | |
371 | e820.nr_map++; | |
372 | } | |
373 | } /* add_memory_region */ | |
374 | ||
375 | /* | |
376 | * Sanitize the BIOS e820 map. | |
377 | * | |
378 | * Some e820 responses include overlapping entries. The following | |
379 | * replaces the original e820 map with a new one, removing overlaps. | |
380 | * | |
381 | */ | |
382 | int __init sanitize_e820_map(struct e820entry * biosmap, char * pnr_map) | |
383 | { | |
384 | struct change_member *change_tmp; | |
385 | unsigned long current_type, last_type; | |
386 | unsigned long long last_addr; | |
387 | int chgidx, still_changing; | |
388 | int overlap_entries; | |
389 | int new_bios_entry; | |
390 | int old_nr, new_nr, chg_nr; | |
391 | int i; | |
392 | ||
393 | /* | |
394 | Visually we're performing the following (1,2,3,4 = memory types)... | |
395 | ||
396 | Sample memory map (w/overlaps): | |
397 | ____22__________________ | |
398 | ______________________4_ | |
399 | ____1111________________ | |
400 | _44_____________________ | |
401 | 11111111________________ | |
402 | ____________________33__ | |
403 | ___________44___________ | |
404 | __________33333_________ | |
405 | ______________22________ | |
406 | ___________________2222_ | |
407 | _________111111111______ | |
408 | _____________________11_ | |
409 | _________________4______ | |
410 | ||
411 | Sanitized equivalent (no overlap): | |
412 | 1_______________________ | |
413 | _44_____________________ | |
414 | ___1____________________ | |
415 | ____22__________________ | |
416 | ______11________________ | |
417 | _________1______________ | |
418 | __________3_____________ | |
419 | ___________44___________ | |
420 | _____________33_________ | |
421 | _______________2________ | |
422 | ________________1_______ | |
423 | _________________4______ | |
424 | ___________________2____ | |
425 | ____________________33__ | |
426 | ______________________4_ | |
427 | */ | |
8e3342f7 | 428 | /* if there's only one memory region, don't bother */ |
429 | if (*pnr_map < 2) { | |
8e3342f7 | 430 | return -1; |
431 | } | |
432 | ||
433 | old_nr = *pnr_map; | |
434 | ||
435 | /* bail out if we find any unreasonable addresses in bios map */ | |
436 | for (i=0; i<old_nr; i++) | |
437 | if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr) { | |
8e3342f7 | 438 | return -1; |
439 | } | |
440 | ||
441 | /* create pointers for initial change-point information (for sorting) */ | |
442 | for (i=0; i < 2*old_nr; i++) | |
443 | change_point[i] = &change_point_list[i]; | |
444 | ||
445 | /* record all known change-points (starting and ending addresses), | |
446 | omitting those that are for empty memory regions */ | |
447 | chgidx = 0; | |
448 | for (i=0; i < old_nr; i++) { | |
449 | if (biosmap[i].size != 0) { | |
450 | change_point[chgidx]->addr = biosmap[i].addr; | |
451 | change_point[chgidx++]->pbios = &biosmap[i]; | |
452 | change_point[chgidx]->addr = biosmap[i].addr + biosmap[i].size; | |
453 | change_point[chgidx++]->pbios = &biosmap[i]; | |
454 | } | |
455 | } | |
456 | chg_nr = chgidx; /* true number of change-points */ | |
457 | ||
458 | /* sort change-point list by memory addresses (low -> high) */ | |
459 | still_changing = 1; | |
460 | while (still_changing) { | |
461 | still_changing = 0; | |
462 | for (i=1; i < chg_nr; i++) { | |
463 | /* if <current_addr> > <last_addr>, swap */ | |
464 | /* or, if current=<start_addr> & last=<end_addr>, swap */ | |
465 | if ((change_point[i]->addr < change_point[i-1]->addr) || | |
466 | ((change_point[i]->addr == change_point[i-1]->addr) && | |
467 | (change_point[i]->addr == change_point[i]->pbios->addr) && | |
468 | (change_point[i-1]->addr != change_point[i-1]->pbios->addr)) | |
469 | ) | |
470 | { | |
471 | change_tmp = change_point[i]; | |
472 | change_point[i] = change_point[i-1]; | |
473 | change_point[i-1] = change_tmp; | |
474 | still_changing=1; | |
475 | } | |
476 | } | |
477 | } | |
478 | ||
479 | /* create a new bios memory map, removing overlaps */ | |
480 | overlap_entries=0; /* number of entries in the overlap table */ | |
481 | new_bios_entry=0; /* index for creating new bios map entries */ | |
482 | last_type = 0; /* start with undefined memory type */ | |
483 | last_addr = 0; /* start with 0 as last starting address */ | |
484 | /* loop through change-points, determining affect on the new bios map */ | |
485 | for (chgidx=0; chgidx < chg_nr; chgidx++) | |
486 | { | |
487 | /* keep track of all overlapping bios entries */ | |
488 | if (change_point[chgidx]->addr == change_point[chgidx]->pbios->addr) | |
489 | { | |
490 | /* add map entry to overlap list (> 1 entry implies an overlap) */ | |
491 | overlap_list[overlap_entries++]=change_point[chgidx]->pbios; | |
492 | } | |
493 | else | |
494 | { | |
495 | /* remove entry from list (order independent, so swap with last) */ | |
496 | for (i=0; i<overlap_entries; i++) | |
497 | { | |
498 | if (overlap_list[i] == change_point[chgidx]->pbios) | |
499 | overlap_list[i] = overlap_list[overlap_entries-1]; | |
500 | } | |
501 | overlap_entries--; | |
502 | } | |
503 | /* if there are overlapping entries, decide which "type" to use */ | |
504 | /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */ | |
505 | current_type = 0; | |
506 | for (i=0; i<overlap_entries; i++) | |
507 | if (overlap_list[i]->type > current_type) | |
508 | current_type = overlap_list[i]->type; | |
509 | /* continue building up new bios map based on this information */ | |
510 | if (current_type != last_type) { | |
511 | if (last_type != 0) { | |
512 | new_bios[new_bios_entry].size = | |
513 | change_point[chgidx]->addr - last_addr; | |
514 | /* move forward only if the new size was non-zero */ | |
515 | if (new_bios[new_bios_entry].size != 0) | |
516 | if (++new_bios_entry >= E820MAX) | |
517 | break; /* no more space left for new bios entries */ | |
518 | } | |
519 | if (current_type != 0) { | |
520 | new_bios[new_bios_entry].addr = change_point[chgidx]->addr; | |
521 | new_bios[new_bios_entry].type = current_type; | |
522 | last_addr=change_point[chgidx]->addr; | |
523 | } | |
524 | last_type = current_type; | |
525 | } | |
526 | } | |
527 | new_nr = new_bios_entry; /* retain count for new bios entries */ | |
528 | ||
529 | /* copy new bios mapping into original location */ | |
530 | memcpy(biosmap, new_bios, new_nr*sizeof(struct e820entry)); | |
531 | *pnr_map = new_nr; | |
532 | ||
8e3342f7 | 533 | return 0; |
534 | } | |
535 | ||
536 | /* | |
537 | * Copy the BIOS e820 map into a safe place. | |
538 | * | |
539 | * Sanity-check it while we're at it.. | |
540 | * | |
541 | * If we're lucky and live on a modern system, the setup code | |
542 | * will have given us a memory map that we can use to properly | |
543 | * set up memory. If we aren't, we'll fake a memory map. | |
544 | * | |
545 | * We check to see that the memory map contains at least 2 elements | |
546 | * before we'll use it, because the detection code in setup.S may | |
547 | * not be perfect and most every PC known to man has two memory | |
548 | * regions: one from 0 to 640k, and one from 1mb up. (The IBM | |
549 | * thinkpad 560x, for example, does not cooperate with the memory | |
550 | * detection code.) | |
551 | */ | |
552 | int __init copy_e820_map(struct e820entry * biosmap, int nr_map) | |
553 | { | |
554 | /* Only one memory region (or negative)? Ignore it */ | |
555 | if (nr_map < 2) | |
556 | return -1; | |
557 | ||
558 | do { | |
559 | unsigned long long start = biosmap->addr; | |
560 | unsigned long long size = biosmap->size; | |
561 | unsigned long long end = start + size; | |
562 | unsigned long type = biosmap->type; | |
8e3342f7 | 563 | |
564 | /* Overflow in 64 bits? Ignore the memory map. */ | |
565 | if (start > end) | |
566 | return -1; | |
567 | ||
568 | /* | |
569 | * Some BIOSes claim RAM in the 640k - 1M region. | |
570 | * Not right. Fix it up. | |
571 | */ | |
572 | if (type == E820_RAM) { | |
8e3342f7 | 573 | if (start < 0x100000ULL && end > 0xA0000ULL) { |
13063832 | 574 | if (start < 0xA0000ULL) |
8e3342f7 | 575 | add_memory_region(start, 0xA0000ULL-start, type); |
13063832 | 576 | if (end <= 0x100000ULL) |
8e3342f7 | 577 | continue; |
8e3342f7 | 578 | start = 0x100000ULL; |
579 | size = end - start; | |
580 | } | |
581 | } | |
582 | add_memory_region(start, size, type); | |
583 | } while (biosmap++,--nr_map); | |
584 | return 0; | |
585 | } | |
586 | ||
b2dff6a8 | 587 | /* |
588 | * Callback for efi_memory_walk. | |
589 | */ | |
590 | static int __init | |
591 | efi_find_max_pfn(unsigned long start, unsigned long end, void *arg) | |
592 | { | |
593 | unsigned long *max_pfn = arg, pfn; | |
594 | ||
595 | if (start < end) { | |
596 | pfn = PFN_UP(end -1); | |
597 | if (pfn > *max_pfn) | |
598 | *max_pfn = pfn; | |
599 | } | |
600 | return 0; | |
601 | } | |
602 | ||
603 | static int __init | |
604 | efi_memory_present_wrapper(unsigned long start, unsigned long end, void *arg) | |
605 | { | |
606 | memory_present(0, PFN_UP(start), PFN_DOWN(end)); | |
607 | return 0; | |
608 | } | |
609 | ||
610 | /* | |
611 | * Find the highest page frame number we have available | |
612 | */ | |
613 | void __init find_max_pfn(void) | |
614 | { | |
615 | int i; | |
616 | ||
617 | max_pfn = 0; | |
618 | if (efi_enabled) { | |
619 | efi_memmap_walk(efi_find_max_pfn, &max_pfn); | |
620 | efi_memmap_walk(efi_memory_present_wrapper, NULL); | |
621 | return; | |
622 | } | |
623 | ||
624 | for (i = 0; i < e820.nr_map; i++) { | |
625 | unsigned long start, end; | |
626 | /* RAM? */ | |
627 | if (e820.map[i].type != E820_RAM) | |
628 | continue; | |
629 | start = PFN_UP(e820.map[i].addr); | |
630 | end = PFN_DOWN(e820.map[i].addr + e820.map[i].size); | |
631 | if (start >= end) | |
632 | continue; | |
633 | if (end > max_pfn) | |
634 | max_pfn = end; | |
635 | memory_present(0, start, end); | |
636 | } | |
637 | } | |
b5b24057 | 638 | |
639 | /* | |
640 | * Free all available memory for boot time allocation. Used | |
641 | * as a callback function by efi_memory_walk() | |
642 | */ | |
643 | ||
644 | static int __init | |
645 | free_available_memory(unsigned long start, unsigned long end, void *arg) | |
646 | { | |
647 | /* check max_low_pfn */ | |
648 | if (start >= (max_low_pfn << PAGE_SHIFT)) | |
649 | return 0; | |
650 | if (end >= (max_low_pfn << PAGE_SHIFT)) | |
651 | end = max_low_pfn << PAGE_SHIFT; | |
652 | if (start < end) | |
653 | free_bootmem(start, end - start); | |
654 | ||
655 | return 0; | |
656 | } | |
657 | /* | |
658 | * Register fully available low RAM pages with the bootmem allocator. | |
659 | */ | |
660 | void __init register_bootmem_low_pages(unsigned long max_low_pfn) | |
661 | { | |
662 | int i; | |
663 | ||
664 | if (efi_enabled) { | |
665 | efi_memmap_walk(free_available_memory, NULL); | |
666 | return; | |
667 | } | |
668 | for (i = 0; i < e820.nr_map; i++) { | |
669 | unsigned long curr_pfn, last_pfn, size; | |
670 | /* | |
671 | * Reserve usable low memory | |
672 | */ | |
673 | if (e820.map[i].type != E820_RAM) | |
674 | continue; | |
675 | /* | |
676 | * We are rounding up the start address of usable memory: | |
677 | */ | |
678 | curr_pfn = PFN_UP(e820.map[i].addr); | |
679 | if (curr_pfn >= max_low_pfn) | |
680 | continue; | |
681 | /* | |
682 | * ... and at the end of the usable range downwards: | |
683 | */ | |
684 | last_pfn = PFN_DOWN(e820.map[i].addr + e820.map[i].size); | |
685 | ||
686 | if (last_pfn > max_low_pfn) | |
687 | last_pfn = max_low_pfn; | |
688 | ||
689 | /* | |
690 | * .. finally, did all the rounding and playing | |
691 | * around just make the area go away? | |
692 | */ | |
693 | if (last_pfn <= curr_pfn) | |
694 | continue; | |
695 | ||
696 | size = last_pfn - curr_pfn; | |
697 | free_bootmem(PFN_PHYS(curr_pfn), PFN_PHYS(size)); | |
698 | } | |
699 | } | |
700 | ||
5c95da9f | 701 | void __init e820_register_memory(void) |
b5b24057 | 702 | { |
703 | unsigned long gapstart, gapsize, round; | |
704 | unsigned long long last; | |
705 | int i; | |
706 | ||
707 | /* | |
708 | * Search for the bigest gap in the low 32 bits of the e820 | |
709 | * memory space. | |
710 | */ | |
711 | last = 0x100000000ull; | |
712 | gapstart = 0x10000000; | |
713 | gapsize = 0x400000; | |
714 | i = e820.nr_map; | |
715 | while (--i >= 0) { | |
716 | unsigned long long start = e820.map[i].addr; | |
717 | unsigned long long end = start + e820.map[i].size; | |
718 | ||
719 | /* | |
720 | * Since "last" is at most 4GB, we know we'll | |
721 | * fit in 32 bits if this condition is true | |
722 | */ | |
723 | if (last > end) { | |
724 | unsigned long gap = last - end; | |
725 | ||
726 | if (gap > gapsize) { | |
727 | gapsize = gap; | |
728 | gapstart = end; | |
729 | } | |
730 | } | |
731 | if (start < last) | |
732 | last = start; | |
733 | } | |
734 | ||
735 | /* | |
736 | * See how much we want to round up: start off with | |
737 | * rounding to the next 1MB area. | |
738 | */ | |
739 | round = 0x100000; | |
740 | while ((gapsize >> 4) > round) | |
741 | round += round; | |
742 | /* Fun with two's complement */ | |
743 | pci_mem_start = (gapstart + round) & -round; | |
744 | ||
745 | printk("Allocating PCI resources starting at %08lx (gap: %08lx:%08lx)\n", | |
746 | pci_mem_start, gapstart, gapsize); | |
747 | } | |
cef518e8 | 748 | |
749 | void __init print_memory_map(char *who) | |
750 | { | |
751 | int i; | |
752 | ||
753 | for (i = 0; i < e820.nr_map; i++) { | |
754 | printk(" %s: %016Lx - %016Lx ", who, | |
755 | e820.map[i].addr, | |
756 | e820.map[i].addr + e820.map[i].size); | |
757 | switch (e820.map[i].type) { | |
758 | case E820_RAM: printk("(usable)\n"); | |
759 | break; | |
760 | case E820_RESERVED: | |
761 | printk("(reserved)\n"); | |
762 | break; | |
763 | case E820_ACPI: | |
764 | printk("(ACPI data)\n"); | |
765 | break; | |
766 | case E820_NVS: | |
767 | printk("(ACPI NVS)\n"); | |
768 | break; | |
9c25d134 | 769 | default: printk("type %u\n", e820.map[i].type); |
cef518e8 | 770 | break; |
771 | } | |
772 | } | |
773 | } | |
774 | ||
bf7e6a19 | 775 | static __init __always_inline void efi_limit_regions(unsigned long long size) |
cef518e8 | 776 | { |
777 | unsigned long long current_addr = 0; | |
bf7e6a19 AM |
778 | efi_memory_desc_t *md, *next_md; |
779 | void *p, *p1; | |
780 | int i, j; | |
781 | ||
782 | j = 0; | |
783 | p1 = memmap.map; | |
784 | for (p = p1, i = 0; p < memmap.map_end; p += memmap.desc_size, i++) { | |
785 | md = p; | |
786 | next_md = p1; | |
787 | current_addr = md->phys_addr + | |
788 | PFN_PHYS(md->num_pages); | |
789 | if (is_available_memory(md)) { | |
790 | if (md->phys_addr >= size) continue; | |
791 | memcpy(next_md, md, memmap.desc_size); | |
792 | if (current_addr >= size) { | |
793 | next_md->num_pages -= | |
794 | PFN_UP(current_addr-size); | |
795 | } | |
796 | p1 += memmap.desc_size; | |
797 | next_md = p1; | |
798 | j++; | |
799 | } else if ((md->attribute & EFI_MEMORY_RUNTIME) == | |
800 | EFI_MEMORY_RUNTIME) { | |
801 | /* In order to make runtime services | |
802 | * available we have to include runtime | |
803 | * memory regions in memory map */ | |
804 | memcpy(next_md, md, memmap.desc_size); | |
805 | p1 += memmap.desc_size; | |
806 | next_md = p1; | |
807 | j++; | |
808 | } | |
809 | } | |
810 | memmap.nr_map = j; | |
811 | memmap.map_end = memmap.map + | |
812 | (memmap.nr_map * memmap.desc_size); | |
813 | } | |
814 | ||
815 | void __init limit_regions(unsigned long long size) | |
816 | { | |
817 | unsigned long long current_addr; | |
cef518e8 | 818 | int i; |
819 | ||
820 | print_memory_map("limit_regions start"); | |
821 | if (efi_enabled) { | |
bf7e6a19 AM |
822 | efi_limit_regions(size); |
823 | return; | |
cef518e8 | 824 | } |
825 | for (i = 0; i < e820.nr_map; i++) { | |
826 | current_addr = e820.map[i].addr + e820.map[i].size; | |
827 | if (current_addr < size) | |
828 | continue; | |
829 | ||
830 | if (e820.map[i].type != E820_RAM) | |
831 | continue; | |
832 | ||
833 | if (e820.map[i].addr >= size) { | |
834 | /* | |
835 | * This region starts past the end of the | |
836 | * requested size, skip it completely. | |
837 | */ | |
838 | e820.nr_map = i; | |
839 | } else { | |
840 | e820.nr_map = i + 1; | |
841 | e820.map[i].size -= current_addr - size; | |
842 | } | |
843 | print_memory_map("limit_regions endfor"); | |
844 | return; | |
845 | } | |
846 | print_memory_map("limit_regions endfunc"); | |
847 | } | |
848 | ||
b92e9fac JB |
849 | /* |
850 | * This function checks if any part of the range <start,end> is mapped | |
851 | * with type. | |
852 | */ | |
853 | int | |
854 | e820_any_mapped(u64 start, u64 end, unsigned type) | |
855 | { | |
856 | int i; | |
857 | for (i = 0; i < e820.nr_map; i++) { | |
858 | const struct e820entry *ei = &e820.map[i]; | |
859 | if (type && ei->type != type) | |
860 | continue; | |
861 | if (ei->addr >= end || ei->addr + ei->size <= start) | |
862 | continue; | |
863 | return 1; | |
864 | } | |
865 | return 0; | |
866 | } | |
867 | EXPORT_SYMBOL_GPL(e820_any_mapped); | |
868 | ||
cef518e8 | 869 | /* |
870 | * This function checks if the entire range <start,end> is mapped with type. | |
871 | * | |
872 | * Note: this function only works correct if the e820 table is sorted and | |
873 | * not-overlapping, which is the case | |
874 | */ | |
875 | int __init | |
876 | e820_all_mapped(unsigned long s, unsigned long e, unsigned type) | |
877 | { | |
878 | u64 start = s; | |
879 | u64 end = e; | |
880 | int i; | |
881 | for (i = 0; i < e820.nr_map; i++) { | |
882 | struct e820entry *ei = &e820.map[i]; | |
883 | if (type && ei->type != type) | |
884 | continue; | |
885 | /* is the region (part) in overlap with the current region ?*/ | |
886 | if (ei->addr >= end || ei->addr + ei->size <= start) | |
887 | continue; | |
888 | /* if the region is at the beginning of <start,end> we move | |
889 | * start to the end of the region since it's ok until there | |
890 | */ | |
891 | if (ei->addr <= start) | |
892 | start = ei->addr + ei->size; | |
893 | /* if start is now at or beyond end, we're done, full | |
894 | * coverage */ | |
895 | if (start >= end) | |
896 | return 1; /* we're done */ | |
897 | } | |
898 | return 0; | |
899 | } | |
900 | ||
901 | static int __init parse_memmap(char *arg) | |
902 | { | |
903 | if (!arg) | |
904 | return -EINVAL; | |
905 | ||
906 | if (strcmp(arg, "exactmap") == 0) { | |
907 | #ifdef CONFIG_CRASH_DUMP | |
908 | /* If we are doing a crash dump, we | |
909 | * still need to know the real mem | |
910 | * size before original memory map is | |
911 | * reset. | |
912 | */ | |
913 | find_max_pfn(); | |
914 | saved_max_pfn = max_pfn; | |
915 | #endif | |
916 | e820.nr_map = 0; | |
917 | user_defined_memmap = 1; | |
918 | } else { | |
919 | /* If the user specifies memory size, we | |
920 | * limit the BIOS-provided memory map to | |
921 | * that size. exactmap can be used to specify | |
922 | * the exact map. mem=number can be used to | |
923 | * trim the existing memory map. | |
924 | */ | |
925 | unsigned long long start_at, mem_size; | |
926 | ||
927 | mem_size = memparse(arg, &arg); | |
928 | if (*arg == '@') { | |
929 | start_at = memparse(arg+1, &arg); | |
930 | add_memory_region(start_at, mem_size, E820_RAM); | |
931 | } else if (*arg == '#') { | |
932 | start_at = memparse(arg+1, &arg); | |
933 | add_memory_region(start_at, mem_size, E820_ACPI); | |
934 | } else if (*arg == '$') { | |
935 | start_at = memparse(arg+1, &arg); | |
936 | add_memory_region(start_at, mem_size, E820_RESERVED); | |
937 | } else { | |
938 | limit_regions(mem_size); | |
939 | user_defined_memmap = 1; | |
940 | } | |
941 | } | |
942 | return 0; | |
943 | } | |
944 | early_param("memmap", parse_memmap); |