Commit | Line | Data |
---|---|---|
5ead97c8 JF |
1 | /* |
2 | * Machine specific setup for xen | |
3 | * | |
4 | * Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007 | |
5 | */ | |
6 | ||
7 | #include <linux/module.h> | |
8 | #include <linux/sched.h> | |
9 | #include <linux/mm.h> | |
10 | #include <linux/pm.h> | |
a9ce6bc1 | 11 | #include <linux/memblock.h> |
d91ee586 | 12 | #include <linux/cpuidle.h> |
48cdd828 | 13 | #include <linux/cpufreq.h> |
5ead97c8 JF |
14 | |
15 | #include <asm/elf.h> | |
6c3652ef | 16 | #include <asm/vdso.h> |
5ead97c8 JF |
17 | #include <asm/e820.h> |
18 | #include <asm/setup.h> | |
b792c755 | 19 | #include <asm/acpi.h> |
8d54db79 | 20 | #include <asm/numa.h> |
5ead97c8 JF |
21 | #include <asm/xen/hypervisor.h> |
22 | #include <asm/xen/hypercall.h> | |
23 | ||
45263cb0 | 24 | #include <xen/xen.h> |
8006ec3e | 25 | #include <xen/page.h> |
e2a81baf | 26 | #include <xen/interface/callback.h> |
35ae11fd | 27 | #include <xen/interface/memory.h> |
5ead97c8 JF |
28 | #include <xen/interface/physdev.h> |
29 | #include <xen/features.h> | |
9103bb0f | 30 | #include "mmu.h" |
5ead97c8 | 31 | #include "xen-ops.h" |
d2eea68e | 32 | #include "vdso.h" |
5ead97c8 JF |
33 | |
34 | /* These are code, but not functions. Defined in entry.S */ | |
35 | extern const char xen_hypervisor_callback[]; | |
36 | extern const char xen_failsafe_callback[]; | |
6efa20e4 | 37 | #ifdef CONFIG_X86_64 |
07ba06d9 | 38 | extern asmlinkage void nmi(void); |
6efa20e4 | 39 | #endif |
f63c2f24 T |
40 | extern void xen_sysenter_target(void); |
41 | extern void xen_syscall_target(void); | |
42 | extern void xen_syscall32_target(void); | |
5ead97c8 | 43 | |
42ee1471 | 44 | /* Amount of extra memory space we add to the e820 ranges */ |
8b5d44a5 | 45 | struct xen_memory_region xen_extra_mem[XEN_EXTRA_MEM_MAX_REGIONS] __initdata; |
42ee1471 | 46 | |
aa24411b DV |
47 | /* Number of pages released from the initial allocation. */ |
48 | unsigned long xen_released_pages; | |
49 | ||
698bb8d1 JF |
50 | /* |
51 | * The maximum amount of extra memory compared to the base size. The | |
52 | * main scaling factor is the size of struct page. At extreme ratios | |
53 | * of base:extra, all the base memory can be filled with page | |
54 | * structures for the extra memory, leaving no space for anything | |
55 | * else. | |
56 | * | |
57 | * 10x seems like a reasonable balance between scaling flexibility and | |
58 | * leaving a practically usable system. | |
59 | */ | |
60 | #define EXTRA_MEM_RATIO (10) | |
61 | ||
dc91c728 | 62 | static void __init xen_add_extra_mem(u64 start, u64 size) |
42ee1471 | 63 | { |
6eaa412f | 64 | unsigned long pfn; |
dc91c728 | 65 | int i; |
6eaa412f | 66 | |
dc91c728 DV |
67 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { |
68 | /* Add new region. */ | |
69 | if (xen_extra_mem[i].size == 0) { | |
70 | xen_extra_mem[i].start = start; | |
71 | xen_extra_mem[i].size = size; | |
72 | break; | |
73 | } | |
74 | /* Append to existing region. */ | |
75 | if (xen_extra_mem[i].start + xen_extra_mem[i].size == start) { | |
76 | xen_extra_mem[i].size += size; | |
77 | break; | |
78 | } | |
79 | } | |
80 | if (i == XEN_EXTRA_MEM_MAX_REGIONS) | |
81 | printk(KERN_WARNING "Warning: not enough extra memory regions\n"); | |
42ee1471 | 82 | |
d4bbf7e7 | 83 | memblock_reserve(start, size); |
2f7acb20 | 84 | |
9103bb0f MR |
85 | if (xen_feature(XENFEAT_auto_translated_physmap)) |
86 | return; | |
87 | ||
dc91c728 | 88 | xen_max_p2m_pfn = PFN_DOWN(start + size); |
c96aae1f KRW |
89 | for (pfn = PFN_DOWN(start); pfn < xen_max_p2m_pfn; pfn++) { |
90 | unsigned long mfn = pfn_to_mfn(pfn); | |
91 | ||
2dcc9a3d | 92 | if (WARN_ONCE(mfn == pfn, "Trying to over-write 1-1 mapping (pfn: %lx)\n", pfn)) |
c96aae1f | 93 | continue; |
2dcc9a3d DV |
94 | WARN_ONCE(mfn != INVALID_P2M_ENTRY, "Trying to remove %lx which has %lx mfn!\n", |
95 | pfn, mfn); | |
6eaa412f | 96 | |
6eaa412f | 97 | __set_phys_to_machine(pfn, INVALID_P2M_ENTRY); |
c96aae1f | 98 | } |
42ee1471 JF |
99 | } |
100 | ||
96dc08b3 KRW |
101 | static unsigned long __init xen_do_chunk(unsigned long start, |
102 | unsigned long end, bool release) | |
093d7b46 MR |
103 | { |
104 | struct xen_memory_reservation reservation = { | |
105 | .address_bits = 0, | |
106 | .extent_order = 0, | |
107 | .domid = DOMID_SELF | |
108 | }; | |
f89e048e | 109 | unsigned long len = 0; |
9103bb0f | 110 | int xlated_phys = xen_feature(XENFEAT_auto_translated_physmap); |
093d7b46 MR |
111 | unsigned long pfn; |
112 | int ret; | |
113 | ||
2e2fb754 KRW |
114 | for (pfn = start; pfn < end; pfn++) { |
115 | unsigned long frame; | |
96dc08b3 | 116 | unsigned long mfn = pfn_to_mfn(pfn); |
2e2fb754 | 117 | |
96dc08b3 KRW |
118 | if (release) { |
119 | /* Make sure pfn exists to start with */ | |
120 | if (mfn == INVALID_P2M_ENTRY || mfn_to_pfn(mfn) != pfn) | |
121 | continue; | |
122 | frame = mfn; | |
123 | } else { | |
9103bb0f | 124 | if (!xlated_phys && mfn != INVALID_P2M_ENTRY) |
96dc08b3 KRW |
125 | continue; |
126 | frame = pfn; | |
127 | } | |
2e2fb754 KRW |
128 | set_xen_guest_handle(reservation.extent_start, &frame); |
129 | reservation.nr_extents = 1; | |
130 | ||
96dc08b3 KRW |
131 | ret = HYPERVISOR_memory_op(release ? XENMEM_decrease_reservation : XENMEM_populate_physmap, |
132 | &reservation); | |
133 | WARN(ret != 1, "Failed to %s pfn %lx err=%d\n", | |
134 | release ? "release" : "populate", pfn, ret); | |
135 | ||
2e2fb754 | 136 | if (ret == 1) { |
96dc08b3 KRW |
137 | if (!early_set_phys_to_machine(pfn, release ? INVALID_P2M_ENTRY : frame)) { |
138 | if (release) | |
139 | break; | |
2e2fb754 KRW |
140 | set_xen_guest_handle(reservation.extent_start, &frame); |
141 | reservation.nr_extents = 1; | |
142 | ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation, | |
96dc08b3 | 143 | &reservation); |
2e2fb754 KRW |
144 | break; |
145 | } | |
146 | len++; | |
147 | } else | |
148 | break; | |
149 | } | |
150 | if (len) | |
96dc08b3 KRW |
151 | printk(KERN_INFO "%s %lx-%lx pfn range: %lu pages %s\n", |
152 | release ? "Freeing" : "Populating", | |
153 | start, end, len, | |
154 | release ? "freed" : "added"); | |
155 | ||
2e2fb754 KRW |
156 | return len; |
157 | } | |
83d51ab4 DV |
158 | |
159 | static unsigned long __init xen_release_chunk(unsigned long start, | |
160 | unsigned long end) | |
161 | { | |
9103bb0f MR |
162 | /* |
163 | * Xen already ballooned out the E820 non RAM regions for us | |
164 | * and set them up properly in EPT. | |
165 | */ | |
166 | if (xen_feature(XENFEAT_auto_translated_physmap)) | |
167 | return end - start; | |
168 | ||
83d51ab4 DV |
169 | return xen_do_chunk(start, end, true); |
170 | } | |
171 | ||
2e2fb754 KRW |
172 | static unsigned long __init xen_populate_chunk( |
173 | const struct e820entry *list, size_t map_size, | |
174 | unsigned long max_pfn, unsigned long *last_pfn, | |
175 | unsigned long credits_left) | |
176 | { | |
177 | const struct e820entry *entry; | |
178 | unsigned int i; | |
179 | unsigned long done = 0; | |
180 | unsigned long dest_pfn; | |
181 | ||
182 | for (i = 0, entry = list; i < map_size; i++, entry++) { | |
2e2fb754 KRW |
183 | unsigned long s_pfn; |
184 | unsigned long e_pfn; | |
185 | unsigned long pfns; | |
186 | long capacity; | |
187 | ||
c3d93f88 | 188 | if (credits_left <= 0) |
2e2fb754 KRW |
189 | break; |
190 | ||
191 | if (entry->type != E820_RAM) | |
192 | continue; | |
193 | ||
c3d93f88 | 194 | e_pfn = PFN_DOWN(entry->addr + entry->size); |
2e2fb754 KRW |
195 | |
196 | /* We only care about E820 after the xen_start_info->nr_pages */ | |
197 | if (e_pfn <= max_pfn) | |
198 | continue; | |
199 | ||
c3d93f88 | 200 | s_pfn = PFN_UP(entry->addr); |
2e2fb754 KRW |
201 | /* If the E820 falls within the nr_pages, we want to start |
202 | * at the nr_pages PFN. | |
203 | * If that would mean going past the E820 entry, skip it | |
204 | */ | |
205 | if (s_pfn <= max_pfn) { | |
206 | capacity = e_pfn - max_pfn; | |
207 | dest_pfn = max_pfn; | |
208 | } else { | |
2e2fb754 KRW |
209 | capacity = e_pfn - s_pfn; |
210 | dest_pfn = s_pfn; | |
211 | } | |
2e2fb754 | 212 | |
c3d93f88 | 213 | if (credits_left < capacity) |
214 | capacity = credits_left; | |
2e2fb754 | 215 | |
c3d93f88 | 216 | pfns = xen_do_chunk(dest_pfn, dest_pfn + capacity, false); |
2e2fb754 | 217 | done += pfns; |
2e2fb754 | 218 | *last_pfn = (dest_pfn + pfns); |
c3d93f88 | 219 | if (pfns < capacity) |
220 | break; | |
221 | credits_left -= pfns; | |
2e2fb754 KRW |
222 | } |
223 | return done; | |
224 | } | |
83d51ab4 DV |
225 | |
226 | static void __init xen_set_identity_and_release_chunk( | |
227 | unsigned long start_pfn, unsigned long end_pfn, unsigned long nr_pages, | |
228 | unsigned long *released, unsigned long *identity) | |
229 | { | |
230 | unsigned long pfn; | |
231 | ||
232 | /* | |
e201bfcc DV |
233 | * If the PFNs are currently mapped, clear the mappings |
234 | * (except for the ISA region which must be 1:1 mapped) to | |
235 | * release the refcounts (in Xen) on the original frames. | |
83d51ab4 | 236 | */ |
9103bb0f MR |
237 | |
238 | /* | |
239 | * PVH E820 matches the hypervisor's P2M which means we need to | |
240 | * account for the proper values of *release and *identity. | |
241 | */ | |
242 | for (pfn = start_pfn; !xen_feature(XENFEAT_auto_translated_physmap) && | |
243 | pfn <= max_pfn_mapped && pfn < end_pfn; pfn++) { | |
e201bfcc DV |
244 | pte_t pte = __pte_ma(0); |
245 | ||
246 | if (pfn < PFN_UP(ISA_END_ADDRESS)) | |
247 | pte = mfn_pte(pfn, PAGE_KERNEL_IO); | |
248 | ||
83d51ab4 | 249 | (void)HYPERVISOR_update_va_mapping( |
e201bfcc DV |
250 | (unsigned long)__va(pfn << PAGE_SHIFT), pte, 0); |
251 | } | |
83d51ab4 DV |
252 | |
253 | if (start_pfn < nr_pages) | |
254 | *released += xen_release_chunk( | |
255 | start_pfn, min(end_pfn, nr_pages)); | |
256 | ||
257 | *identity += set_phys_range_identity(start_pfn, end_pfn); | |
258 | } | |
259 | ||
f3f436e3 DV |
260 | static unsigned long __init xen_set_identity_and_release( |
261 | const struct e820entry *list, size_t map_size, unsigned long nr_pages) | |
093d7b46 | 262 | { |
f3f436e3 | 263 | phys_addr_t start = 0; |
093d7b46 | 264 | unsigned long released = 0; |
68df0da7 | 265 | unsigned long identity = 0; |
f3f436e3 | 266 | const struct e820entry *entry; |
68df0da7 KRW |
267 | int i; |
268 | ||
f3f436e3 DV |
269 | /* |
270 | * Combine non-RAM regions and gaps until a RAM region (or the | |
271 | * end of the map) is reached, then set the 1:1 map and | |
272 | * release the pages (if available) in those non-RAM regions. | |
273 | * | |
274 | * The combined non-RAM regions are rounded to a whole number | |
275 | * of pages so any partial pages are accessible via the 1:1 | |
276 | * mapping. This is needed for some BIOSes that put (for | |
277 | * example) the DMI tables in a reserved region that begins on | |
278 | * a non-page boundary. | |
279 | */ | |
68df0da7 | 280 | for (i = 0, entry = list; i < map_size; i++, entry++) { |
f3f436e3 | 281 | phys_addr_t end = entry->addr + entry->size; |
f3f436e3 DV |
282 | if (entry->type == E820_RAM || i == map_size - 1) { |
283 | unsigned long start_pfn = PFN_DOWN(start); | |
284 | unsigned long end_pfn = PFN_UP(end); | |
68df0da7 | 285 | |
f3f436e3 DV |
286 | if (entry->type == E820_RAM) |
287 | end_pfn = PFN_UP(entry->addr); | |
68df0da7 | 288 | |
83d51ab4 DV |
289 | if (start_pfn < end_pfn) |
290 | xen_set_identity_and_release_chunk( | |
291 | start_pfn, end_pfn, nr_pages, | |
292 | &released, &identity); | |
68df0da7 | 293 | |
f3f436e3 | 294 | start = end; |
68df0da7 | 295 | } |
68df0da7 | 296 | } |
f3f436e3 | 297 | |
ca118238 KRW |
298 | if (released) |
299 | printk(KERN_INFO "Released %lu pages of unused memory\n", released); | |
300 | if (identity) | |
301 | printk(KERN_INFO "Set %ld page(s) to 1-1 mapping\n", identity); | |
f3f436e3 DV |
302 | |
303 | return released; | |
68df0da7 | 304 | } |
d312ae87 DV |
305 | |
306 | static unsigned long __init xen_get_max_pages(void) | |
307 | { | |
308 | unsigned long max_pages = MAX_DOMAIN_PAGES; | |
309 | domid_t domid = DOMID_SELF; | |
310 | int ret; | |
311 | ||
d3db7281 IC |
312 | /* |
313 | * For the initial domain we use the maximum reservation as | |
314 | * the maximum page. | |
315 | * | |
316 | * For guest domains the current maximum reservation reflects | |
317 | * the current maximum rather than the static maximum. In this | |
318 | * case the e820 map provided to us will cover the static | |
319 | * maximum region. | |
320 | */ | |
321 | if (xen_initial_domain()) { | |
322 | ret = HYPERVISOR_memory_op(XENMEM_maximum_reservation, &domid); | |
323 | if (ret > 0) | |
324 | max_pages = ret; | |
325 | } | |
326 | ||
d312ae87 DV |
327 | return min(max_pages, MAX_DOMAIN_PAGES); |
328 | } | |
329 | ||
dc91c728 DV |
330 | static void xen_align_and_add_e820_region(u64 start, u64 size, int type) |
331 | { | |
332 | u64 end = start + size; | |
333 | ||
334 | /* Align RAM regions to page boundaries. */ | |
335 | if (type == E820_RAM) { | |
336 | start = PAGE_ALIGN(start); | |
337 | end &= ~((u64)PAGE_SIZE - 1); | |
338 | } | |
339 | ||
340 | e820_add_region(start, end - start, type); | |
341 | } | |
342 | ||
3bc38cbc DV |
343 | void xen_ignore_unusable(struct e820entry *list, size_t map_size) |
344 | { | |
345 | struct e820entry *entry; | |
346 | unsigned int i; | |
347 | ||
348 | for (i = 0, entry = list; i < map_size; i++, entry++) { | |
349 | if (entry->type == E820_UNUSABLE) | |
350 | entry->type = E820_RAM; | |
351 | } | |
352 | } | |
353 | ||
5ead97c8 JF |
354 | /** |
355 | * machine_specific_memory_setup - Hook for machine specific memory setup. | |
356 | **/ | |
5ead97c8 JF |
357 | char * __init xen_memory_setup(void) |
358 | { | |
35ae11fd IC |
359 | static struct e820entry map[E820MAX] __initdata; |
360 | ||
5ead97c8 | 361 | unsigned long max_pfn = xen_start_info->nr_pages; |
35ae11fd IC |
362 | unsigned long long mem_end; |
363 | int rc; | |
364 | struct xen_memory_map memmap; | |
dc91c728 | 365 | unsigned long max_pages; |
2e2fb754 | 366 | unsigned long last_pfn = 0; |
42ee1471 | 367 | unsigned long extra_pages = 0; |
2e2fb754 | 368 | unsigned long populated; |
35ae11fd | 369 | int i; |
9e9a5fcb | 370 | int op; |
5ead97c8 | 371 | |
8006ec3e | 372 | max_pfn = min(MAX_DOMAIN_PAGES, max_pfn); |
35ae11fd IC |
373 | mem_end = PFN_PHYS(max_pfn); |
374 | ||
375 | memmap.nr_entries = E820MAX; | |
376 | set_xen_guest_handle(memmap.buffer, map); | |
377 | ||
9e9a5fcb IC |
378 | op = xen_initial_domain() ? |
379 | XENMEM_machine_memory_map : | |
380 | XENMEM_memory_map; | |
381 | rc = HYPERVISOR_memory_op(op, &memmap); | |
35ae11fd | 382 | if (rc == -ENOSYS) { |
9ec23a7f | 383 | BUG_ON(xen_initial_domain()); |
35ae11fd IC |
384 | memmap.nr_entries = 1; |
385 | map[0].addr = 0ULL; | |
386 | map[0].size = mem_end; | |
387 | /* 8MB slack (to balance backend allocations). */ | |
388 | map[0].size += 8ULL << 20; | |
389 | map[0].type = E820_RAM; | |
390 | rc = 0; | |
391 | } | |
392 | BUG_ON(rc); | |
8006ec3e | 393 | |
3bc38cbc DV |
394 | /* |
395 | * Xen won't allow a 1:1 mapping to be created to UNUSABLE | |
396 | * regions, so if we're using the machine memory map leave the | |
397 | * region as RAM as it is in the pseudo-physical map. | |
398 | * | |
399 | * UNUSABLE regions in domUs are not handled and will need | |
400 | * a patch in the future. | |
401 | */ | |
402 | if (xen_initial_domain()) | |
403 | xen_ignore_unusable(map, memmap.nr_entries); | |
404 | ||
dc91c728 DV |
405 | /* Make sure the Xen-supplied memory map is well-ordered. */ |
406 | sanitize_e820_map(map, memmap.nr_entries, &memmap.nr_entries); | |
407 | ||
408 | max_pages = xen_get_max_pages(); | |
409 | if (max_pages > max_pfn) | |
410 | extra_pages += max_pages - max_pfn; | |
411 | ||
f3f436e3 DV |
412 | /* |
413 | * Set P2M for all non-RAM pages and E820 gaps to be identity | |
414 | * type PFNs. Any RAM pages that would be made inaccesible by | |
415 | * this are first released. | |
416 | */ | |
417 | xen_released_pages = xen_set_identity_and_release( | |
418 | map, memmap.nr_entries, max_pfn); | |
dc91c728 | 419 | |
2e2fb754 KRW |
420 | /* |
421 | * Populate back the non-RAM pages and E820 gaps that had been | |
422 | * released. */ | |
423 | populated = xen_populate_chunk(map, memmap.nr_entries, | |
424 | max_pfn, &last_pfn, xen_released_pages); | |
425 | ||
58b7b53a KRW |
426 | xen_released_pages -= populated; |
427 | extra_pages += xen_released_pages; | |
2e2fb754 KRW |
428 | |
429 | if (last_pfn > max_pfn) { | |
430 | max_pfn = min(MAX_DOMAIN_PAGES, last_pfn); | |
431 | mem_end = PFN_PHYS(max_pfn); | |
432 | } | |
dc91c728 DV |
433 | /* |
434 | * Clamp the amount of extra memory to a EXTRA_MEM_RATIO | |
435 | * factor the base size. On non-highmem systems, the base | |
436 | * size is the full initial memory allocation; on highmem it | |
437 | * is limited to the max size of lowmem, so that it doesn't | |
438 | * get completely filled. | |
439 | * | |
440 | * In principle there could be a problem in lowmem systems if | |
441 | * the initial memory is also very large with respect to | |
442 | * lowmem, but we won't try to deal with that here. | |
443 | */ | |
444 | extra_pages = min(EXTRA_MEM_RATIO * min(max_pfn, PFN_DOWN(MAXMEM)), | |
445 | extra_pages); | |
dc91c728 DV |
446 | i = 0; |
447 | while (i < memmap.nr_entries) { | |
448 | u64 addr = map[i].addr; | |
449 | u64 size = map[i].size; | |
450 | u32 type = map[i].type; | |
451 | ||
452 | if (type == E820_RAM) { | |
453 | if (addr < mem_end) { | |
454 | size = min(size, mem_end - addr); | |
455 | } else if (extra_pages) { | |
456 | size = min(size, (u64)extra_pages * PAGE_SIZE); | |
457 | extra_pages -= size / PAGE_SIZE; | |
458 | xen_add_extra_mem(addr, size); | |
459 | } else | |
460 | type = E820_UNUSABLE; | |
3654581e JF |
461 | } |
462 | ||
dc91c728 | 463 | xen_align_and_add_e820_region(addr, size, type); |
b5b43ced | 464 | |
dc91c728 DV |
465 | map[i].addr += size; |
466 | map[i].size -= size; | |
467 | if (map[i].size == 0) | |
468 | i++; | |
35ae11fd | 469 | } |
b792c755 | 470 | |
25b884a8 DV |
471 | /* |
472 | * Set the rest as identity mapped, in case PCI BARs are | |
473 | * located here. | |
474 | * | |
475 | * PFNs above MAX_P2M_PFN are considered identity mapped as | |
476 | * well. | |
477 | */ | |
478 | set_phys_range_identity(map[i-1].addr / PAGE_SIZE, ~0ul); | |
479 | ||
b792c755 | 480 | /* |
9ec23a7f IC |
481 | * In domU, the ISA region is normal, usable memory, but we |
482 | * reserve ISA memory anyway because too many things poke | |
b792c755 JF |
483 | * about in there. |
484 | */ | |
485 | e820_add_region(ISA_START_ADDRESS, ISA_END_ADDRESS - ISA_START_ADDRESS, | |
486 | E820_RESERVED); | |
5ead97c8 | 487 | |
be5bf9fa JF |
488 | /* |
489 | * Reserve Xen bits: | |
490 | * - mfn_list | |
491 | * - xen_start_info | |
492 | * See comment above "struct start_info" in <xen/interface/xen.h> | |
51faaf2b KRW |
493 | * We tried to make the the memblock_reserve more selective so |
494 | * that it would be clear what region is reserved. Sadly we ran | |
495 | * in the problem wherein on a 64-bit hypervisor with a 32-bit | |
496 | * initial domain, the pt_base has the cr3 value which is not | |
497 | * neccessarily where the pagetable starts! As Jan put it: " | |
498 | * Actually, the adjustment turns out to be correct: The page | |
499 | * tables for a 32-on-64 dom0 get allocated in the order "first L1", | |
500 | * "first L2", "first L3", so the offset to the page table base is | |
501 | * indeed 2. When reading xen/include/public/xen.h's comment | |
502 | * very strictly, this is not a violation (since there nothing is said | |
503 | * that the first thing in the page table space is pointed to by | |
504 | * pt_base; I admit that this seems to be implied though, namely | |
505 | * do I think that it is implied that the page table space is the | |
506 | * range [pt_base, pt_base + nt_pt_frames), whereas that | |
507 | * range here indeed is [pt_base - 2, pt_base - 2 + nt_pt_frames), | |
508 | * which - without a priori knowledge - the kernel would have | |
509 | * difficulty to figure out)." - so lets just fall back to the | |
510 | * easy way and reserve the whole region. | |
be5bf9fa | 511 | */ |
24aa0788 TH |
512 | memblock_reserve(__pa(xen_start_info->mfn_list), |
513 | xen_start_info->pt_base - xen_start_info->mfn_list); | |
be5bf9fa JF |
514 | |
515 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); | |
516 | ||
5ead97c8 JF |
517 | return "Xen"; |
518 | } | |
519 | ||
d2eea68e RM |
520 | /* |
521 | * Set the bit indicating "nosegneg" library variants should be used. | |
6a52e4b1 JF |
522 | * We only need to bother in pure 32-bit mode; compat 32-bit processes |
523 | * can have un-truncated segments, so wrapping around is allowed. | |
d2eea68e | 524 | */ |
08b6d290 | 525 | static void __init fiddle_vdso(void) |
d2eea68e | 526 | { |
6a52e4b1 JF |
527 | #ifdef CONFIG_X86_32 |
528 | u32 *mask; | |
529 | mask = VDSO32_SYMBOL(&vdso32_int80_start, NOTE_MASK); | |
530 | *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; | |
531 | mask = VDSO32_SYMBOL(&vdso32_sysenter_start, NOTE_MASK); | |
d2eea68e | 532 | *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; |
6fcac6d3 | 533 | #endif |
d2eea68e RM |
534 | } |
535 | ||
148f9bb8 | 536 | static int register_callback(unsigned type, const void *func) |
e2a81baf | 537 | { |
88459d4c JF |
538 | struct callback_register callback = { |
539 | .type = type, | |
540 | .address = XEN_CALLBACK(__KERNEL_CS, func), | |
e2a81baf JF |
541 | .flags = CALLBACKF_mask_events, |
542 | }; | |
543 | ||
88459d4c JF |
544 | return HYPERVISOR_callback_op(CALLBACKOP_register, &callback); |
545 | } | |
546 | ||
148f9bb8 | 547 | void xen_enable_sysenter(void) |
88459d4c | 548 | { |
6fcac6d3 | 549 | int ret; |
62541c37 | 550 | unsigned sysenter_feature; |
6fcac6d3 JF |
551 | |
552 | #ifdef CONFIG_X86_32 | |
62541c37 | 553 | sysenter_feature = X86_FEATURE_SEP; |
6fcac6d3 | 554 | #else |
62541c37 | 555 | sysenter_feature = X86_FEATURE_SYSENTER32; |
6fcac6d3 | 556 | #endif |
88459d4c | 557 | |
62541c37 JF |
558 | if (!boot_cpu_has(sysenter_feature)) |
559 | return; | |
560 | ||
6fcac6d3 | 561 | ret = register_callback(CALLBACKTYPE_sysenter, xen_sysenter_target); |
62541c37 JF |
562 | if(ret != 0) |
563 | setup_clear_cpu_cap(sysenter_feature); | |
e2a81baf JF |
564 | } |
565 | ||
148f9bb8 | 566 | void xen_enable_syscall(void) |
6fcac6d3 JF |
567 | { |
568 | #ifdef CONFIG_X86_64 | |
6fcac6d3 | 569 | int ret; |
6fcac6d3 JF |
570 | |
571 | ret = register_callback(CALLBACKTYPE_syscall, xen_syscall_target); | |
572 | if (ret != 0) { | |
d5303b81 | 573 | printk(KERN_ERR "Failed to set syscall callback: %d\n", ret); |
62541c37 JF |
574 | /* Pretty fatal; 64-bit userspace has no other |
575 | mechanism for syscalls. */ | |
576 | } | |
577 | ||
578 | if (boot_cpu_has(X86_FEATURE_SYSCALL32)) { | |
6fcac6d3 JF |
579 | ret = register_callback(CALLBACKTYPE_syscall32, |
580 | xen_syscall32_target); | |
d5303b81 | 581 | if (ret != 0) |
62541c37 | 582 | setup_clear_cpu_cap(X86_FEATURE_SYSCALL32); |
6fcac6d3 JF |
583 | } |
584 | #endif /* CONFIG_X86_64 */ | |
585 | } | |
3b284bde | 586 | void xen_enable_nmi(void) |
6efa20e4 KRW |
587 | { |
588 | #ifdef CONFIG_X86_64 | |
07ba06d9 | 589 | if (register_callback(CALLBACKTYPE_nmi, (char *)nmi)) |
6efa20e4 KRW |
590 | BUG(); |
591 | #endif | |
592 | } | |
d285d683 | 593 | void __init xen_pvmmu_arch_setup(void) |
5ead97c8 | 594 | { |
5ead97c8 JF |
595 | HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments); |
596 | HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_writable_pagetables); | |
597 | ||
d285d683 MR |
598 | HYPERVISOR_vm_assist(VMASST_CMD_enable, |
599 | VMASST_TYPE_pae_extended_cr3); | |
5ead97c8 | 600 | |
88459d4c JF |
601 | if (register_callback(CALLBACKTYPE_event, xen_hypervisor_callback) || |
602 | register_callback(CALLBACKTYPE_failsafe, xen_failsafe_callback)) | |
603 | BUG(); | |
5ead97c8 | 604 | |
e2a81baf | 605 | xen_enable_sysenter(); |
6fcac6d3 | 606 | xen_enable_syscall(); |
6efa20e4 | 607 | xen_enable_nmi(); |
d285d683 MR |
608 | } |
609 | ||
610 | /* This function is not called for HVM domains */ | |
611 | void __init xen_arch_setup(void) | |
612 | { | |
613 | xen_panic_handler_init(); | |
614 | if (!xen_feature(XENFEAT_auto_translated_physmap)) | |
615 | xen_pvmmu_arch_setup(); | |
616 | ||
5ead97c8 JF |
617 | #ifdef CONFIG_ACPI |
618 | if (!(xen_start_info->flags & SIF_INITDOMAIN)) { | |
619 | printk(KERN_INFO "ACPI in unprivileged domain disabled\n"); | |
620 | disable_acpi(); | |
621 | } | |
622 | #endif | |
623 | ||
624 | memcpy(boot_command_line, xen_start_info->cmd_line, | |
625 | MAX_GUEST_CMDLINE > COMMAND_LINE_SIZE ? | |
626 | COMMAND_LINE_SIZE : MAX_GUEST_CMDLINE); | |
627 | ||
bc15fde7 | 628 | /* Set up idle, making sure it calls safe_halt() pvop */ |
d91ee586 | 629 | disable_cpuidle(); |
48cdd828 | 630 | disable_cpufreq(); |
6a377ddc | 631 | WARN_ON(xen_set_default_idle()); |
d2eea68e | 632 | fiddle_vdso(); |
8d54db79 KRW |
633 | #ifdef CONFIG_NUMA |
634 | numa_off = 1; | |
635 | #endif | |
5ead97c8 | 636 | } |