/root/src/xen/xen/arch/x86/e820.c
Line | Count | Source (jump to first uncovered line) |
1 | | #include <xen/init.h> |
2 | | #include <xen/lib.h> |
3 | | #include <xen/mm.h> |
4 | | #include <xen/compat.h> |
5 | | #include <xen/dmi.h> |
6 | | #include <xen/pfn.h> |
7 | | #include <asm/e820.h> |
8 | | #include <asm/page.h> |
9 | | #include <asm/processor.h> |
10 | | #include <asm/mtrr.h> |
11 | | #include <asm/msr.h> |
12 | | |
13 | | /* |
14 | | * opt_mem: Limit maximum address of physical RAM. |
15 | | * Any RAM beyond this address limit is ignored. |
16 | | */ |
17 | | static unsigned long long __initdata opt_mem; |
18 | | size_param("mem", opt_mem); |
19 | | |
20 | | /* |
21 | | * opt_availmem: Limit maximum usable amount of physical RAM. |
22 | | * Any RAM beyond this limited amount is ignored. |
23 | | */ |
24 | | static unsigned long long __initdata opt_availmem; |
25 | | size_param("availmem", opt_availmem); |
26 | | |
27 | | /* opt_nomtrr_check: Don't clip ram to highest cacheable MTRR. */ |
28 | | static s8 __initdata e820_mtrr_clip = -1; |
29 | | boolean_param("e820-mtrr-clip", e820_mtrr_clip); |
30 | | |
31 | | /* opt_e820_verbose: Be verbose about clipping, the original e820, &c */ |
32 | | static bool __initdata e820_verbose; |
33 | | boolean_param("e820-verbose", e820_verbose); |
34 | | |
35 | | struct e820map e820; |
36 | | struct e820map __initdata e820_raw; |
37 | | |
38 | | /* |
39 | | * This function checks if the entire range <start,end> is mapped with type. |
40 | | * |
41 | | * Note: this function only works correct if the e820 table is sorted and |
42 | | * not-overlapping, which is the case |
43 | | */ |
44 | | int __init e820_all_mapped(u64 start, u64 end, unsigned type) |
45 | 1 | { |
46 | 1 | int i; |
47 | 1 | |
48 | 16 | for (i = 0; i < e820.nr_map; i++) { |
49 | 16 | struct e820entry *ei = &e820.map[i]; |
50 | 16 | |
51 | 16 | if (type && ei->type != type) |
52 | 10 | continue; |
53 | 16 | /* is the region (part) in overlap with the current region ?*/ |
54 | 6 | if (ei->addr >= end || ei->addr + ei->size <= start) |
55 | 5 | continue; |
56 | 6 | |
57 | 6 | /* if the region is at the beginning of <start,end> we move |
58 | 6 | * start to the end of the region since it's ok until there |
59 | 6 | */ |
60 | 1 | if (ei->addr <= start) |
61 | 1 | start = ei->addr + ei->size; |
62 | 1 | /* |
63 | 1 | * if start is now at or beyond end, we're done, full |
64 | 1 | * coverage |
65 | 1 | */ |
66 | 1 | if (start >= end) |
67 | 1 | return 1; |
68 | 1 | } |
69 | 0 | return 0; |
70 | 1 | } |
71 | | |
72 | | static void __init add_memory_region(unsigned long long start, |
73 | | unsigned long long size, int type) |
74 | 19 | { |
75 | 19 | int x; |
76 | 19 | |
77 | 19 | x = e820.nr_map; |
78 | 19 | |
79 | 19 | if (x == ARRAY_SIZE(e820.map)) { |
80 | 0 | printk(KERN_ERR "Ooops! Too many entries in the memory map!\n"); |
81 | 0 | return; |
82 | 0 | } |
83 | 19 | |
84 | 19 | e820.map[x].addr = start; |
85 | 19 | e820.map[x].size = size; |
86 | 19 | e820.map[x].type = type; |
87 | 19 | e820.nr_map++; |
88 | 19 | } |
89 | | |
90 | | static void __init print_e820_memory_map(struct e820entry *map, unsigned int entries) |
91 | 1 | { |
92 | 1 | unsigned int i; |
93 | 1 | |
94 | 20 | for (i = 0; i < entries; i++) { |
95 | 19 | printk(" %016Lx - %016Lx ", |
96 | 19 | (unsigned long long)(map[i].addr), |
97 | 19 | (unsigned long long)(map[i].addr + map[i].size)); |
98 | 19 | switch (map[i].type) { |
99 | 8 | case E820_RAM: |
100 | 8 | printk("(usable)\n"); |
101 | 8 | break; |
102 | 8 | case E820_RESERVED: |
103 | 8 | printk("(reserved)\n"); |
104 | 8 | break; |
105 | 1 | case E820_ACPI: |
106 | 1 | printk("(ACPI data)\n"); |
107 | 1 | break; |
108 | 2 | case E820_NVS: |
109 | 2 | printk("(ACPI NVS)\n"); |
110 | 2 | break; |
111 | 0 | case E820_UNUSABLE: |
112 | 0 | printk("(unusable)\n"); |
113 | 0 | break; |
114 | 0 | default: |
115 | 0 | printk("type %u\n", map[i].type); |
116 | 0 | break; |
117 | 19 | } |
118 | 19 | } |
119 | 1 | } |
120 | | |
121 | | /* |
122 | | * Sanitize the BIOS e820 map. |
123 | | * |
124 | | * Some e820 responses include overlapping entries. The following |
125 | | * replaces the original e820 map with a new one, removing overlaps. |
126 | | * |
127 | | */ |
128 | | struct change_member { |
129 | | struct e820entry *pbios; /* pointer to original bios entry */ |
130 | | unsigned long long addr; /* address for this change point */ |
131 | | }; |
132 | | static struct change_member change_point_list[2*E820MAX] __initdata; |
133 | | static struct change_member *change_point[2*E820MAX] __initdata; |
134 | | static struct e820entry *overlap_list[E820MAX] __initdata; |
135 | | static struct e820entry new_bios[E820MAX] __initdata; |
136 | | |
137 | | static int __init sanitize_e820_map(struct e820entry *biosmap, |
138 | | unsigned int *pnr_map) |
139 | 1 | { |
140 | 1 | struct change_member *change_tmp; |
141 | 1 | unsigned long current_type, last_type; |
142 | 1 | unsigned long long last_addr; |
143 | 1 | int chgidx, still_changing; |
144 | 1 | int overlap_entries; |
145 | 1 | int new_bios_entry; |
146 | 1 | int old_nr, new_nr, chg_nr; |
147 | 1 | int i; |
148 | 1 | |
149 | 1 | /* |
150 | 1 | Visually we're performing the following (1,2,3,4 = memory types)... |
151 | 1 | |
152 | 1 | Sample memory map (w/overlaps): |
153 | 1 | ____22__________________ |
154 | 1 | ______________________4_ |
155 | 1 | ____1111________________ |
156 | 1 | _44_____________________ |
157 | 1 | 11111111________________ |
158 | 1 | ____________________33__ |
159 | 1 | ___________44___________ |
160 | 1 | __________33333_________ |
161 | 1 | ______________22________ |
162 | 1 | ___________________2222_ |
163 | 1 | _________111111111______ |
164 | 1 | _____________________11_ |
165 | 1 | _________________4______ |
166 | 1 | |
167 | 1 | Sanitized equivalent (no overlap): |
168 | 1 | 1_______________________ |
169 | 1 | _44_____________________ |
170 | 1 | ___1____________________ |
171 | 1 | ____22__________________ |
172 | 1 | ______11________________ |
173 | 1 | _________1______________ |
174 | 1 | __________3_____________ |
175 | 1 | ___________44___________ |
176 | 1 | _____________33_________ |
177 | 1 | _______________2________ |
178 | 1 | ________________1_______ |
179 | 1 | _________________4______ |
180 | 1 | ___________________2____ |
181 | 1 | ____________________33__ |
182 | 1 | ______________________4_ |
183 | 1 | */ |
184 | 1 | |
185 | 1 | /* if there's only one memory region, don't bother */ |
186 | 1 | if (*pnr_map < 2) |
187 | 0 | return -1; |
188 | 1 | |
189 | 1 | old_nr = *pnr_map; |
190 | 1 | |
191 | 1 | /* bail out if we find any unreasonable addresses in bios map */ |
192 | 20 | for (i=0; i<old_nr; i++) |
193 | 19 | if (biosmap[i].addr + biosmap[i].size < biosmap[i].addr) |
194 | 0 | return -1; |
195 | 1 | |
196 | 1 | /* create pointers for initial change-point information (for sorting) */ |
197 | 39 | for (i=0; i < 2*old_nr; i++) |
198 | 38 | change_point[i] = &change_point_list[i]; |
199 | 1 | |
200 | 1 | /* record all known change-points (starting and ending addresses), |
201 | 1 | omitting those that are for empty memory regions */ |
202 | 1 | chgidx = 0; |
203 | 20 | for (i=0; i < old_nr; i++) { |
204 | 19 | if (biosmap[i].size != 0) { |
205 | 19 | change_point[chgidx]->addr = biosmap[i].addr; |
206 | 19 | change_point[chgidx++]->pbios = &biosmap[i]; |
207 | 19 | change_point[chgidx]->addr = biosmap[i].addr + biosmap[i].size; |
208 | 19 | change_point[chgidx++]->pbios = &biosmap[i]; |
209 | 19 | } |
210 | 19 | } |
211 | 1 | chg_nr = chgidx; /* true number of change-points */ |
212 | 1 | |
213 | 1 | /* sort change-point list by memory addresses (low -> high) */ |
214 | 1 | still_changing = 1; |
215 | 3 | while (still_changing) { |
216 | 2 | still_changing = 0; |
217 | 76 | for (i=1; i < chg_nr; i++) { |
218 | 74 | /* if <current_addr> > <last_addr>, swap */ |
219 | 74 | /* or, if current=<start_addr> & last=<end_addr>, swap */ |
220 | 74 | if ((change_point[i]->addr < change_point[i-1]->addr) || |
221 | 74 | ((change_point[i]->addr == change_point[i-1]->addr) && |
222 | 28 | (change_point[i]->addr == change_point[i]->pbios->addr) && |
223 | 14 | (change_point[i-1]->addr != change_point[i-1]->pbios->addr)) |
224 | 74 | ) |
225 | 14 | { |
226 | 14 | change_tmp = change_point[i]; |
227 | 14 | change_point[i] = change_point[i-1]; |
228 | 14 | change_point[i-1] = change_tmp; |
229 | 14 | still_changing=1; |
230 | 14 | } |
231 | 74 | } |
232 | 2 | } |
233 | 1 | |
234 | 1 | /* create a new bios memory map, removing overlaps */ |
235 | 1 | overlap_entries=0; /* number of entries in the overlap table */ |
236 | 1 | new_bios_entry=0; /* index for creating new bios map entries */ |
237 | 1 | last_type = 0; /* start with undefined memory type */ |
238 | 1 | last_addr = 0; /* start with 0 as last starting address */ |
239 | 1 | /* loop through change-points, determining affect on the new bios map */ |
240 | 39 | for (chgidx=0; chgidx < chg_nr; chgidx++) |
241 | 38 | { |
242 | 38 | /* keep track of all overlapping bios entries */ |
243 | 38 | if (change_point[chgidx]->addr == change_point[chgidx]->pbios->addr) |
244 | 19 | { |
245 | 19 | /* add map entry to overlap list (> 1 entry implies an overlap) */ |
246 | 19 | overlap_list[overlap_entries++]=change_point[chgidx]->pbios; |
247 | 19 | } |
248 | 38 | else |
249 | 19 | { |
250 | 19 | /* remove entry from list (order independent, so swap with last) */ |
251 | 52 | for (i=0; i<overlap_entries; i++) |
252 | 33 | { |
253 | 33 | if (overlap_list[i] == change_point[chgidx]->pbios) |
254 | 19 | overlap_list[i] = overlap_list[overlap_entries-1]; |
255 | 33 | } |
256 | 19 | overlap_entries--; |
257 | 19 | } |
258 | 38 | /* if there are overlapping entries, decide which "type" to use */ |
259 | 38 | /* (larger value takes precedence -- 1=usable, 2,3,4,4+=unusable) */ |
260 | 38 | current_type = 0; |
261 | 85 | for (i=0; i<overlap_entries; i++) |
262 | 47 | if (overlap_list[i]->type > current_type) |
263 | 40 | current_type = overlap_list[i]->type; |
264 | 38 | /* continue building up new bios map based on this information */ |
265 | 38 | if (current_type != last_type) { |
266 | 24 | if (last_type != 0) { |
267 | 19 | new_bios[new_bios_entry].size = |
268 | 19 | change_point[chgidx]->addr - last_addr; |
269 | 19 | /* move forward only if the new size was non-zero */ |
270 | 19 | if (new_bios[new_bios_entry].size != 0) |
271 | 19 | if (++new_bios_entry >= ARRAY_SIZE(new_bios)) |
272 | 0 | break; /* no more space left for new bios entries */ |
273 | 19 | } |
274 | 24 | if (current_type != 0) { |
275 | 19 | new_bios[new_bios_entry].addr = change_point[chgidx]->addr; |
276 | 19 | new_bios[new_bios_entry].type = current_type; |
277 | 19 | last_addr=change_point[chgidx]->addr; |
278 | 19 | } |
279 | 24 | last_type = current_type; |
280 | 24 | } |
281 | 38 | } |
282 | 1 | new_nr = new_bios_entry; /* retain count for new bios entries */ |
283 | 1 | |
284 | 1 | /* copy new bios mapping into original location */ |
285 | 1 | memcpy(biosmap, new_bios, new_nr*sizeof(struct e820entry)); |
286 | 1 | *pnr_map = new_nr; |
287 | 1 | |
288 | 1 | return 0; |
289 | 1 | } |
290 | | |
291 | | /* |
292 | | * Copy the BIOS e820 map into a safe place. |
293 | | * |
294 | | * Sanity-check it while we're at it.. |
295 | | * |
296 | | * If we're lucky and live on a modern system, the setup code |
297 | | * will have given us a memory map that we can use to properly |
298 | | * set up memory. If we aren't, we'll fake a memory map. |
299 | | * |
300 | | * We check to see that the memory map contains at least 2 elements |
301 | | * before we'll use it, because the detection code in setup.S may |
302 | | * not be perfect and most every PC known to man has two memory |
303 | | * regions: one from 0 to 640k, and one from 1mb up. (The IBM |
304 | | * thinkpad 560x, for example, does not cooperate with the memory |
305 | | * detection code.) |
306 | | */ |
307 | | static int __init copy_e820_map(struct e820entry * biosmap, int nr_map) |
308 | 1 | { |
309 | 1 | /* Only one memory region (or negative)? Ignore it */ |
310 | 1 | if (nr_map < 2) |
311 | 0 | return -1; |
312 | 1 | |
313 | 19 | do { |
314 | 19 | unsigned long long start = biosmap->addr; |
315 | 19 | unsigned long long size = biosmap->size; |
316 | 19 | unsigned long long end = start + size; |
317 | 19 | unsigned long type = biosmap->type; |
318 | 19 | |
319 | 19 | /* Overflow in 64 bits? Ignore the memory map. */ |
320 | 19 | if (start > end) |
321 | 0 | return -1; |
322 | 19 | |
323 | 19 | /* |
324 | 19 | * Some BIOSes claim RAM in the 640k - 1M region. |
325 | 19 | * Not right. Fix it up. |
326 | 19 | */ |
327 | 19 | if (type == E820_RAM) { |
328 | 8 | if (start < 0x100000ULL && end > 0xA0000ULL) { |
329 | 0 | if (start < 0xA0000ULL) |
330 | 0 | add_memory_region(start, 0xA0000ULL-start, type); |
331 | 0 | if (end <= 0x100000ULL) |
332 | 0 | continue; |
333 | 0 | start = 0x100000ULL; |
334 | 0 | size = end - start; |
335 | 0 | } |
336 | 8 | } |
337 | 19 | add_memory_region(start, size, type); |
338 | 19 | } while (biosmap++,--nr_map); |
339 | 1 | return 0; |
340 | 1 | } |
341 | | |
342 | | |
343 | | /* |
344 | | * Find the highest page frame number we have available |
345 | | */ |
346 | | static unsigned long __init find_max_pfn(void) |
347 | 1 | { |
348 | 1 | int i; |
349 | 1 | unsigned long max_pfn = 0; |
350 | 1 | |
351 | 20 | for (i = 0; i < e820.nr_map; i++) { |
352 | 19 | unsigned long start, end; |
353 | 19 | /* RAM? */ |
354 | 19 | if (e820.map[i].type != E820_RAM) |
355 | 11 | continue; |
356 | 8 | start = PFN_UP(e820.map[i].addr); |
357 | 8 | end = PFN_DOWN(e820.map[i].addr + e820.map[i].size); |
358 | 8 | if (start >= end) |
359 | 0 | continue; |
360 | 8 | if (end > max_pfn) |
361 | 8 | max_pfn = end; |
362 | 8 | } |
363 | 1 | |
364 | 1 | return max_pfn; |
365 | 1 | } |
366 | | |
367 | | static void __init clip_to_limit(uint64_t limit, char *warnmsg) |
368 | 2 | { |
369 | 2 | int i; |
370 | 2 | char _warnmsg[160]; |
371 | 2 | uint64_t old_limit = 0; |
372 | 2 | |
373 | 2 | for ( ; ; ) |
374 | 2 | { |
375 | 2 | /* Find a RAM region needing clipping. */ |
376 | 40 | for ( i = 0; i < e820.nr_map; i++ ) |
377 | 38 | if ( (e820.map[i].type == E820_RAM) && |
378 | 16 | ((e820.map[i].addr + e820.map[i].size) > limit) ) |
379 | 0 | break; |
380 | 2 | |
381 | 2 | /* If none found, we are done. */ |
382 | 2 | if ( i == e820.nr_map ) |
383 | 2 | break; |
384 | 2 | |
385 | 0 | old_limit = max_t( |
386 | 0 | uint64_t, old_limit, e820.map[i].addr + e820.map[i].size); |
387 | 0 |
|
388 | 0 | /* We try to convert clipped RAM areas to E820_UNUSABLE. */ |
389 | 0 | if ( e820_change_range_type(&e820, max(e820.map[i].addr, limit), |
390 | 0 | e820.map[i].addr + e820.map[i].size, |
391 | 0 | E820_RAM, E820_UNUSABLE) ) |
392 | 0 | continue; |
393 | 0 |
|
394 | 0 | /* |
395 | 0 | * If the type change fails (e.g., not space in table) then we clip or |
396 | 0 | * delete the region as appropriate. |
397 | 0 | */ |
398 | 0 | if ( e820.map[i].addr < limit ) |
399 | 0 | { |
400 | 0 | e820.map[i].size = limit - e820.map[i].addr; |
401 | 0 | } |
402 | 0 | else |
403 | 0 | { |
404 | 0 | memmove(&e820.map[i], &e820.map[i+1], |
405 | 0 | (e820.nr_map - i - 1) * sizeof(struct e820entry)); |
406 | 0 | e820.nr_map--; |
407 | 0 | } |
408 | 0 | } |
409 | 2 | |
410 | 2 | if ( old_limit ) |
411 | 0 | { |
412 | 0 | if ( warnmsg ) |
413 | 0 | { |
414 | 0 | snprintf(_warnmsg, sizeof(_warnmsg), warnmsg, (long)(limit>>30)); |
415 | 0 | printk("WARNING: %s\n", _warnmsg); |
416 | 0 | } |
417 | 0 | printk("Truncating RAM from %lukB to %lukB\n", |
418 | 0 | (unsigned long)(old_limit >> 10), (unsigned long)(limit >> 10)); |
419 | 0 | } |
420 | 2 | } |
421 | | |
422 | | /* Conservative estimate of top-of-RAM by looking for MTRR WB regions. */ |
423 | | static uint64_t __init mtrr_top_of_ram(void) |
424 | 1 | { |
425 | 1 | uint32_t eax, ebx, ecx, edx; |
426 | 1 | uint64_t mtrr_cap, mtrr_def, addr_mask, base, mask, top; |
427 | 1 | unsigned int i, phys_bits = 36; |
428 | 1 | |
429 | 1 | /* By default we check only Intel systems. */ |
430 | 1 | if ( e820_mtrr_clip == -1 ) |
431 | 1 | { |
432 | 1 | char vendor[13]; |
433 | 1 | cpuid(0x00000000, &eax, |
434 | 1 | (uint32_t *)&vendor[0], |
435 | 1 | (uint32_t *)&vendor[8], |
436 | 1 | (uint32_t *)&vendor[4]); |
437 | 1 | vendor[12] = '\0'; |
438 | 1 | e820_mtrr_clip = !strcmp(vendor, "GenuineIntel"); |
439 | 1 | } |
440 | 1 | |
441 | 1 | if ( !e820_mtrr_clip ) |
442 | 0 | return 0; |
443 | 1 | |
444 | 1 | if ( e820_verbose ) |
445 | 0 | printk("Checking MTRR ranges...\n"); |
446 | 1 | |
447 | 1 | /* Does the CPU support architectural MTRRs? */ |
448 | 1 | cpuid(0x00000001, &eax, &ebx, &ecx, &edx); |
449 | 1 | if ( !test_bit(X86_FEATURE_MTRR & 31, &edx) ) |
450 | 0 | return 0; |
451 | 1 | |
452 | 1 | /* Find the physical address size for this CPU. */ |
453 | 1 | eax = cpuid_eax(0x80000000); |
454 | 1 | if ( (eax >> 16) == 0x8000 && eax >= 0x80000008 ) |
455 | 1 | { |
456 | 1 | phys_bits = (uint8_t)cpuid_eax(0x80000008); |
457 | 1 | if ( phys_bits > PADDR_BITS ) |
458 | 0 | phys_bits = PADDR_BITS; |
459 | 1 | } |
460 | 1 | addr_mask = ((1ull << phys_bits) - 1) & ~((1ull << 12) - 1); |
461 | 1 | |
462 | 1 | rdmsrl(MSR_MTRRcap, mtrr_cap); |
463 | 1 | rdmsrl(MSR_MTRRdefType, mtrr_def); |
464 | 1 | |
465 | 1 | if ( e820_verbose ) |
466 | 0 | printk(" MTRR cap: %"PRIx64" type: %"PRIx64"\n", mtrr_cap, mtrr_def); |
467 | 1 | |
468 | 1 | /* MTRRs enabled, and default memory type is not writeback? */ |
469 | 1 | if ( !test_bit(11, &mtrr_def) || ((uint8_t)mtrr_def == MTRR_TYPE_WRBACK) ) |
470 | 0 | return 0; |
471 | 1 | |
472 | 1 | /* |
473 | 1 | * Find end of highest WB-type range. This is a conservative estimate |
474 | 1 | * of the highest WB address since overlapping UC/WT ranges dominate. |
475 | 1 | */ |
476 | 1 | top = 0; |
477 | 11 | for ( i = 0; i < (uint8_t)mtrr_cap; i++ ) |
478 | 10 | { |
479 | 10 | rdmsrl(MSR_IA32_MTRR_PHYSBASE(i), base); |
480 | 10 | rdmsrl(MSR_IA32_MTRR_PHYSMASK(i), mask); |
481 | 10 | |
482 | 10 | if ( e820_verbose ) |
483 | 0 | printk(" MTRR[%d]: base %"PRIx64" mask %"PRIx64"\n", |
484 | 0 | i, base, mask); |
485 | 10 | |
486 | 10 | if ( !test_bit(11, &mask) || ((uint8_t)base != MTRR_TYPE_WRBACK) ) |
487 | 8 | continue; |
488 | 2 | base &= addr_mask; |
489 | 2 | mask &= addr_mask; |
490 | 2 | top = max_t(uint64_t, top, ((base | ~mask) & addr_mask) + PAGE_SIZE); |
491 | 2 | } |
492 | 1 | |
493 | 1 | return top; |
494 | 1 | } |
495 | | |
496 | | static void __init reserve_dmi_region(void) |
497 | 1 | { |
498 | 1 | for ( ; ; ) |
499 | 2 | { |
500 | 2 | paddr_t base; |
501 | 2 | u32 len; |
502 | 2 | const char *what = dmi_get_table(&base, &len); |
503 | 2 | |
504 | 2 | if ( !what ) |
505 | 1 | break; |
506 | 1 | if ( ((base + len) > base) && |
507 | 1 | reserve_e820_ram(&e820, base, base + len) ) |
508 | 0 | printk("WARNING: %s table located in E820 RAM %"PRIpaddr"-%"PRIpaddr". Fixed.\n", |
509 | 0 | what, base, base + len); |
510 | 1 | } |
511 | 1 | } |
512 | | |
513 | | static void __init machine_specific_memory_setup(struct e820map *raw) |
514 | 1 | { |
515 | 1 | unsigned long mpt_limit, ro_mpt_limit; |
516 | 1 | uint64_t top_of_ram, size; |
517 | 1 | int i; |
518 | 1 | |
519 | 1 | sanitize_e820_map(raw->map, &raw->nr_map); |
520 | 1 | copy_e820_map(raw->map, raw->nr_map); |
521 | 1 | |
522 | 1 | if ( opt_mem ) |
523 | 0 | clip_to_limit(opt_mem, NULL); |
524 | 1 | |
525 | 1 | if ( opt_availmem ) |
526 | 0 | { |
527 | 0 | for ( i = size = 0; (i < e820.nr_map) && (size <= opt_availmem); i++ ) |
528 | 0 | if ( e820.map[i].type == E820_RAM ) |
529 | 0 | size += e820.map[i].size; |
530 | 0 | if ( size > opt_availmem ) |
531 | 0 | clip_to_limit( |
532 | 0 | e820.map[i-1].addr + e820.map[i-1].size - (size-opt_availmem), |
533 | 0 | NULL); |
534 | 0 | } |
535 | 1 | |
536 | 1 | mpt_limit = ((RDWR_MPT_VIRT_END - RDWR_MPT_VIRT_START) |
537 | 1 | / sizeof(unsigned long)) << PAGE_SHIFT; |
538 | 1 | ro_mpt_limit = ((RO_MPT_VIRT_END - RO_MPT_VIRT_START) |
539 | 1 | / sizeof(unsigned long)) << PAGE_SHIFT; |
540 | 1 | if ( mpt_limit > ro_mpt_limit ) |
541 | 0 | mpt_limit = ro_mpt_limit; |
542 | 1 | clip_to_limit(mpt_limit, |
543 | 1 | "Only the first %lu GB of the physical " |
544 | 1 | "memory map can be accessed by Xen."); |
545 | 1 | |
546 | 1 | reserve_dmi_region(); |
547 | 1 | |
548 | 1 | top_of_ram = mtrr_top_of_ram(); |
549 | 1 | if ( top_of_ram ) |
550 | 1 | clip_to_limit(top_of_ram, "MTRRs do not cover all of memory."); |
551 | 1 | } |
552 | | |
553 | | /* This function relies on the passed in e820->map[] being sorted. */ |
554 | | int __init e820_add_range( |
555 | | struct e820map *e820, uint64_t s, uint64_t e, uint32_t type) |
556 | 0 | { |
557 | 0 | unsigned int i; |
558 | 0 |
|
559 | 0 | for ( i = 0; i < e820->nr_map; ++i ) |
560 | 0 | { |
561 | 0 | uint64_t rs = e820->map[i].addr; |
562 | 0 | uint64_t re = rs + e820->map[i].size; |
563 | 0 |
|
564 | 0 | if ( rs == e && e820->map[i].type == type ) |
565 | 0 | { |
566 | 0 | e820->map[i].addr = s; |
567 | 0 | return 1; |
568 | 0 | } |
569 | 0 |
|
570 | 0 | if ( re == s && e820->map[i].type == type && |
571 | 0 | (i + 1 == e820->nr_map || e820->map[i + 1].addr >= e) ) |
572 | 0 | { |
573 | 0 | e820->map[i].size += e - s; |
574 | 0 | return 1; |
575 | 0 | } |
576 | 0 |
|
577 | 0 | if ( rs >= e ) |
578 | 0 | break; |
579 | 0 |
|
580 | 0 | if ( re > s ) |
581 | 0 | return 0; |
582 | 0 | } |
583 | 0 |
|
584 | 0 | if ( e820->nr_map >= ARRAY_SIZE(e820->map) ) |
585 | 0 | { |
586 | 0 | printk(XENLOG_WARNING "E820: overflow while adding region" |
587 | 0 | " %"PRIx64"-%"PRIx64"\n", s, e); |
588 | 0 | return 0; |
589 | 0 | } |
590 | 0 |
|
591 | 0 | memmove(e820->map + i + 1, e820->map + i, |
592 | 0 | (e820->nr_map - i) * sizeof(*e820->map)); |
593 | 0 |
|
594 | 0 | e820->nr_map++; |
595 | 0 | e820->map[i].addr = s; |
596 | 0 | e820->map[i].size = e - s; |
597 | 0 | e820->map[i].type = type; |
598 | 0 |
|
599 | 0 | return 1; |
600 | 0 | } |
601 | | |
602 | | int __init e820_change_range_type( |
603 | | struct e820map *e820, uint64_t s, uint64_t e, |
604 | | uint32_t orig_type, uint32_t new_type) |
605 | 4 | { |
606 | 4 | uint64_t rs = 0, re = 0; |
607 | 4 | int i; |
608 | 4 | |
609 | 55 | for ( i = 0; i < e820->nr_map; i++ ) |
610 | 55 | { |
611 | 55 | /* Have we found the e820 region that includes the specified range? */ |
612 | 55 | rs = e820->map[i].addr; |
613 | 55 | re = rs + e820->map[i].size; |
614 | 55 | if ( (s >= rs) && (e <= re) ) |
615 | 4 | break; |
616 | 55 | } |
617 | 4 | |
618 | 4 | if ( (i == e820->nr_map) || (e820->map[i].type != orig_type) ) |
619 | 1 | return 0; |
620 | 4 | |
621 | 3 | if ( (s == rs) && (e == re) ) |
622 | 1 | { |
623 | 1 | e820->map[i].type = new_type; |
624 | 1 | } |
625 | 2 | else if ( (s == rs) || (e == re) ) |
626 | 0 | { |
627 | 0 | if ( (e820->nr_map + 1) > ARRAY_SIZE(e820->map) ) |
628 | 0 | goto overflow; |
629 | 0 |
|
630 | 0 | memmove(&e820->map[i+1], &e820->map[i], |
631 | 0 | (e820->nr_map-i) * sizeof(e820->map[0])); |
632 | 0 | e820->nr_map++; |
633 | 0 |
|
634 | 0 | if ( s == rs ) |
635 | 0 | { |
636 | 0 | e820->map[i].size = e - s; |
637 | 0 | e820->map[i].type = new_type; |
638 | 0 | e820->map[i+1].addr = e; |
639 | 0 | e820->map[i+1].size = re - e; |
640 | 0 | } |
641 | 0 | else |
642 | 0 | { |
643 | 0 | e820->map[i].size = s - rs; |
644 | 0 | e820->map[i+1].addr = s; |
645 | 0 | e820->map[i+1].size = e - s; |
646 | 0 | e820->map[i+1].type = new_type; |
647 | 0 | } |
648 | 0 | } |
649 | 2 | else |
650 | 2 | { |
651 | 2 | if ( (e820->nr_map + 2) > ARRAY_SIZE(e820->map) ) |
652 | 0 | goto overflow; |
653 | 2 | |
654 | 2 | memmove(&e820->map[i+2], &e820->map[i], |
655 | 2 | (e820->nr_map-i) * sizeof(e820->map[0])); |
656 | 2 | e820->nr_map += 2; |
657 | 2 | |
658 | 2 | e820->map[i].size = s - rs; |
659 | 2 | e820->map[i+1].addr = s; |
660 | 2 | e820->map[i+1].size = e - s; |
661 | 2 | e820->map[i+1].type = new_type; |
662 | 2 | e820->map[i+2].addr = e; |
663 | 2 | e820->map[i+2].size = re - e; |
664 | 2 | } |
665 | 3 | |
666 | 3 | /* Finally, look for any opportunities to merge adjacent e820 entries. */ |
667 | 64 | for ( i = 0; i < (e820->nr_map - 1); i++ ) |
668 | 61 | { |
669 | 61 | if ( (e820->map[i].type != e820->map[i+1].type) || |
670 | 10 | ((e820->map[i].addr + e820->map[i].size) != e820->map[i+1].addr) ) |
671 | 60 | continue; |
672 | 1 | e820->map[i].size += e820->map[i+1].size; |
673 | 1 | memmove(&e820->map[i+1], &e820->map[i+2], |
674 | 1 | (e820->nr_map-i-2) * sizeof(e820->map[0])); |
675 | 1 | e820->nr_map--; |
676 | 1 | i--; |
677 | 1 | } |
678 | 3 | |
679 | 3 | return 1; |
680 | 3 | |
681 | 0 | overflow: |
682 | 0 | printk("Overflow in e820 while reserving region %"PRIx64"-%"PRIx64"\n", |
683 | 0 | s, e); |
684 | 0 | return 0; |
685 | 3 | } |
686 | | |
687 | | /* Set E820_RAM area (@s,@e) as RESERVED in specified e820 map. */ |
688 | | int __init reserve_e820_ram(struct e820map *e820, uint64_t s, uint64_t e) |
689 | 4 | { |
690 | 4 | return e820_change_range_type(e820, s, e, E820_RAM, E820_RESERVED); |
691 | 4 | } |
692 | | |
693 | | unsigned long __init init_e820(const char *str, struct e820map *raw) |
694 | 1 | { |
695 | 1 | if ( e820_verbose ) |
696 | 0 | { |
697 | 0 | printk("Initial %s RAM map:\n", str); |
698 | 0 | print_e820_memory_map(raw->map, raw->nr_map); |
699 | 0 | } |
700 | 1 | |
701 | 1 | machine_specific_memory_setup(raw); |
702 | 1 | |
703 | 1 | printk("%s RAM map:\n", str); |
704 | 1 | print_e820_memory_map(e820.map, e820.nr_map); |
705 | 1 | |
706 | 1 | return find_max_pfn(); |
707 | 1 | } |