]>
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> | |
808fdb71 | 30 | #include <xen/hvc-console.h> |
5ead97c8 | 31 | #include "xen-ops.h" |
d2eea68e | 32 | #include "vdso.h" |
1f3ac86b | 33 | #include "mmu.h" |
5ead97c8 | 34 | |
c70727a5 JG |
35 | #define GB(x) ((uint64_t)(x) * 1024 * 1024 * 1024) |
36 | ||
42ee1471 | 37 | /* Amount of extra memory space we add to the e820 ranges */ |
8b5d44a5 | 38 | struct xen_memory_region xen_extra_mem[XEN_EXTRA_MEM_MAX_REGIONS] __initdata; |
42ee1471 | 39 | |
aa24411b DV |
40 | /* Number of pages released from the initial allocation. */ |
41 | unsigned long xen_released_pages; | |
42 | ||
69632ecf JG |
43 | /* E820 map used during setting up memory. */ |
44 | static struct e820entry xen_e820_map[E820MAX] __initdata; | |
45 | static u32 xen_e820_map_entries __initdata; | |
46 | ||
1f3ac86b JG |
47 | /* |
48 | * Buffer used to remap identity mapped pages. We only need the virtual space. | |
49 | * The physical page behind this address is remapped as needed to different | |
50 | * buffer pages. | |
51 | */ | |
52 | #define REMAP_SIZE (P2M_PER_PAGE - 3) | |
53 | static struct { | |
54 | unsigned long next_area_mfn; | |
55 | unsigned long target_pfn; | |
56 | unsigned long size; | |
57 | unsigned long mfns[REMAP_SIZE]; | |
58 | } xen_remap_buf __initdata __aligned(PAGE_SIZE); | |
59 | static unsigned long xen_remap_mfn __initdata = INVALID_P2M_ENTRY; | |
4fbb67e3 | 60 | |
698bb8d1 JF |
61 | /* |
62 | * The maximum amount of extra memory compared to the base size. The | |
63 | * main scaling factor is the size of struct page. At extreme ratios | |
64 | * of base:extra, all the base memory can be filled with page | |
65 | * structures for the extra memory, leaving no space for anything | |
66 | * else. | |
67 | * | |
68 | * 10x seems like a reasonable balance between scaling flexibility and | |
69 | * leaving a practically usable system. | |
70 | */ | |
71 | #define EXTRA_MEM_RATIO (10) | |
72 | ||
c70727a5 JG |
73 | static bool xen_512gb_limit __initdata = IS_ENABLED(CONFIG_XEN_512GB); |
74 | ||
75 | static void __init xen_parse_512gb(void) | |
76 | { | |
77 | bool val = false; | |
78 | char *arg; | |
79 | ||
80 | arg = strstr(xen_start_info->cmd_line, "xen_512gb_limit"); | |
81 | if (!arg) | |
82 | return; | |
83 | ||
84 | arg = strstr(xen_start_info->cmd_line, "xen_512gb_limit="); | |
85 | if (!arg) | |
86 | val = true; | |
87 | else if (strtobool(arg + strlen("xen_512gb_limit="), &val)) | |
88 | return; | |
89 | ||
90 | xen_512gb_limit = val; | |
91 | } | |
92 | ||
3ba5c867 | 93 | static void __init xen_add_extra_mem(phys_addr_t start, phys_addr_t size) |
42ee1471 | 94 | { |
dc91c728 | 95 | int i; |
6eaa412f | 96 | |
dc91c728 DV |
97 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { |
98 | /* Add new region. */ | |
99 | if (xen_extra_mem[i].size == 0) { | |
100 | xen_extra_mem[i].start = start; | |
101 | xen_extra_mem[i].size = size; | |
102 | break; | |
103 | } | |
104 | /* Append to existing region. */ | |
105 | if (xen_extra_mem[i].start + xen_extra_mem[i].size == start) { | |
106 | xen_extra_mem[i].size += size; | |
107 | break; | |
108 | } | |
109 | } | |
110 | if (i == XEN_EXTRA_MEM_MAX_REGIONS) | |
111 | printk(KERN_WARNING "Warning: not enough extra memory regions\n"); | |
42ee1471 | 112 | |
d4bbf7e7 | 113 | memblock_reserve(start, size); |
5b8e7d80 | 114 | } |
2f7acb20 | 115 | |
3ba5c867 | 116 | static void __init xen_del_extra_mem(phys_addr_t start, phys_addr_t size) |
5b8e7d80 JG |
117 | { |
118 | int i; | |
3ba5c867 | 119 | phys_addr_t start_r, size_r; |
c96aae1f | 120 | |
5b8e7d80 JG |
121 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { |
122 | start_r = xen_extra_mem[i].start; | |
123 | size_r = xen_extra_mem[i].size; | |
124 | ||
125 | /* Start of region. */ | |
126 | if (start_r == start) { | |
127 | BUG_ON(size > size_r); | |
128 | xen_extra_mem[i].start += size; | |
129 | xen_extra_mem[i].size -= size; | |
130 | break; | |
131 | } | |
132 | /* End of region. */ | |
133 | if (start_r + size_r == start + size) { | |
134 | BUG_ON(size > size_r); | |
135 | xen_extra_mem[i].size -= size; | |
136 | break; | |
137 | } | |
138 | /* Mid of region. */ | |
139 | if (start > start_r && start < start_r + size_r) { | |
140 | BUG_ON(start + size > start_r + size_r); | |
141 | xen_extra_mem[i].size = start - start_r; | |
142 | /* Calling memblock_reserve() again is okay. */ | |
143 | xen_add_extra_mem(start + size, start_r + size_r - | |
144 | (start + size)); | |
145 | break; | |
146 | } | |
147 | } | |
148 | memblock_free(start, size); | |
149 | } | |
150 | ||
151 | /* | |
152 | * Called during boot before the p2m list can take entries beyond the | |
153 | * hypervisor supplied p2m list. Entries in extra mem are to be regarded as | |
154 | * invalid. | |
155 | */ | |
156 | unsigned long __ref xen_chk_extra_mem(unsigned long pfn) | |
157 | { | |
158 | int i; | |
e86f9496 | 159 | phys_addr_t addr = PFN_PHYS(pfn); |
6eaa412f | 160 | |
5b8e7d80 JG |
161 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { |
162 | if (addr >= xen_extra_mem[i].start && | |
163 | addr < xen_extra_mem[i].start + xen_extra_mem[i].size) | |
164 | return INVALID_P2M_ENTRY; | |
165 | } | |
166 | ||
167 | return IDENTITY_FRAME(pfn); | |
168 | } | |
169 | ||
170 | /* | |
171 | * Mark all pfns of extra mem as invalid in p2m list. | |
172 | */ | |
173 | void __init xen_inv_extra_mem(void) | |
174 | { | |
175 | unsigned long pfn, pfn_s, pfn_e; | |
176 | int i; | |
177 | ||
178 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { | |
9a17ad7f JG |
179 | if (!xen_extra_mem[i].size) |
180 | continue; | |
5b8e7d80 JG |
181 | pfn_s = PFN_DOWN(xen_extra_mem[i].start); |
182 | pfn_e = PFN_UP(xen_extra_mem[i].start + xen_extra_mem[i].size); | |
183 | for (pfn = pfn_s; pfn < pfn_e; pfn++) | |
184 | set_phys_to_machine(pfn, INVALID_P2M_ENTRY); | |
c96aae1f | 185 | } |
42ee1471 JF |
186 | } |
187 | ||
4fbb67e3 MR |
188 | /* |
189 | * Finds the next RAM pfn available in the E820 map after min_pfn. | |
190 | * This function updates min_pfn with the pfn found and returns | |
191 | * the size of that range or zero if not found. | |
192 | */ | |
69632ecf | 193 | static unsigned long __init xen_find_pfn_range(unsigned long *min_pfn) |
2e2fb754 | 194 | { |
69632ecf | 195 | const struct e820entry *entry = xen_e820_map; |
2e2fb754 KRW |
196 | unsigned int i; |
197 | unsigned long done = 0; | |
2e2fb754 | 198 | |
69632ecf | 199 | for (i = 0; i < xen_e820_map_entries; i++, entry++) { |
2e2fb754 KRW |
200 | unsigned long s_pfn; |
201 | unsigned long e_pfn; | |
2e2fb754 KRW |
202 | |
203 | if (entry->type != E820_RAM) | |
204 | continue; | |
205 | ||
c3d93f88 | 206 | e_pfn = PFN_DOWN(entry->addr + entry->size); |
2e2fb754 | 207 | |
4fbb67e3 MR |
208 | /* We only care about E820 after this */ |
209 | if (e_pfn < *min_pfn) | |
2e2fb754 KRW |
210 | continue; |
211 | ||
c3d93f88 | 212 | s_pfn = PFN_UP(entry->addr); |
4fbb67e3 MR |
213 | |
214 | /* If min_pfn falls within the E820 entry, we want to start | |
215 | * at the min_pfn PFN. | |
2e2fb754 | 216 | */ |
4fbb67e3 MR |
217 | if (s_pfn <= *min_pfn) { |
218 | done = e_pfn - *min_pfn; | |
2e2fb754 | 219 | } else { |
4fbb67e3 MR |
220 | done = e_pfn - s_pfn; |
221 | *min_pfn = s_pfn; | |
2e2fb754 | 222 | } |
4fbb67e3 MR |
223 | break; |
224 | } | |
2e2fb754 | 225 | |
4fbb67e3 MR |
226 | return done; |
227 | } | |
2e2fb754 | 228 | |
1f3ac86b JG |
229 | static int __init xen_free_mfn(unsigned long mfn) |
230 | { | |
231 | struct xen_memory_reservation reservation = { | |
232 | .address_bits = 0, | |
233 | .extent_order = 0, | |
234 | .domid = DOMID_SELF | |
235 | }; | |
236 | ||
237 | set_xen_guest_handle(reservation.extent_start, &mfn); | |
238 | reservation.nr_extents = 1; | |
239 | ||
240 | return HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation); | |
241 | } | |
242 | ||
4fbb67e3 | 243 | /* |
1f3ac86b | 244 | * This releases a chunk of memory and then does the identity map. It's used |
4fbb67e3 MR |
245 | * as a fallback if the remapping fails. |
246 | */ | |
247 | static void __init xen_set_identity_and_release_chunk(unsigned long start_pfn, | |
5097cdf6 | 248 | unsigned long end_pfn, unsigned long nr_pages) |
4fbb67e3 | 249 | { |
1f3ac86b JG |
250 | unsigned long pfn, end; |
251 | int ret; | |
252 | ||
4fbb67e3 MR |
253 | WARN_ON(start_pfn > end_pfn); |
254 | ||
bc7142cf | 255 | /* Release pages first. */ |
1f3ac86b JG |
256 | end = min(end_pfn, nr_pages); |
257 | for (pfn = start_pfn; pfn < end; pfn++) { | |
258 | unsigned long mfn = pfn_to_mfn(pfn); | |
259 | ||
260 | /* Make sure pfn exists to start with */ | |
261 | if (mfn == INVALID_P2M_ENTRY || mfn_to_pfn(mfn) != pfn) | |
262 | continue; | |
263 | ||
264 | ret = xen_free_mfn(mfn); | |
265 | WARN(ret != 1, "Failed to release pfn %lx err=%d\n", pfn, ret); | |
266 | ||
267 | if (ret == 1) { | |
5097cdf6 | 268 | xen_released_pages++; |
1f3ac86b JG |
269 | if (!__set_phys_to_machine(pfn, INVALID_P2M_ENTRY)) |
270 | break; | |
1f3ac86b JG |
271 | } else |
272 | break; | |
273 | } | |
274 | ||
bc7142cf | 275 | set_phys_range_identity(start_pfn, end_pfn); |
4fbb67e3 MR |
276 | } |
277 | ||
278 | /* | |
1f3ac86b | 279 | * Helper function to update the p2m and m2p tables and kernel mapping. |
4fbb67e3 | 280 | */ |
1f3ac86b | 281 | static void __init xen_update_mem_tables(unsigned long pfn, unsigned long mfn) |
4fbb67e3 MR |
282 | { |
283 | struct mmu_update update = { | |
3ba5c867 | 284 | .ptr = ((uint64_t)mfn << PAGE_SHIFT) | MMU_MACHPHYS_UPDATE, |
4fbb67e3 MR |
285 | .val = pfn |
286 | }; | |
287 | ||
288 | /* Update p2m */ | |
1f3ac86b | 289 | if (!set_phys_to_machine(pfn, mfn)) { |
4fbb67e3 MR |
290 | WARN(1, "Failed to set p2m mapping for pfn=%ld mfn=%ld\n", |
291 | pfn, mfn); | |
1f3ac86b | 292 | BUG(); |
2e2fb754 | 293 | } |
4fbb67e3 MR |
294 | |
295 | /* Update m2p */ | |
296 | if (HYPERVISOR_mmu_update(&update, 1, NULL, DOMID_SELF) < 0) { | |
297 | WARN(1, "Failed to set m2p mapping for mfn=%ld pfn=%ld\n", | |
298 | mfn, pfn); | |
1f3ac86b | 299 | BUG(); |
4fbb67e3 MR |
300 | } |
301 | ||
1f3ac86b | 302 | /* Update kernel mapping, but not for highmem. */ |
e86f9496 | 303 | if (pfn >= PFN_UP(__pa(high_memory - 1))) |
1f3ac86b JG |
304 | return; |
305 | ||
306 | if (HYPERVISOR_update_va_mapping((unsigned long)__va(pfn << PAGE_SHIFT), | |
307 | mfn_pte(mfn, PAGE_KERNEL), 0)) { | |
308 | WARN(1, "Failed to update kernel mapping for mfn=%ld pfn=%ld\n", | |
309 | mfn, pfn); | |
310 | BUG(); | |
311 | } | |
2e2fb754 | 312 | } |
83d51ab4 | 313 | |
4fbb67e3 MR |
314 | /* |
315 | * This function updates the p2m and m2p tables with an identity map from | |
1f3ac86b JG |
316 | * start_pfn to start_pfn+size and prepares remapping the underlying RAM of the |
317 | * original allocation at remap_pfn. The information needed for remapping is | |
318 | * saved in the memory itself to avoid the need for allocating buffers. The | |
319 | * complete remap information is contained in a list of MFNs each containing | |
320 | * up to REMAP_SIZE MFNs and the start target PFN for doing the remap. | |
321 | * This enables us to preserve the original mfn sequence while doing the | |
322 | * remapping at a time when the memory management is capable of allocating | |
323 | * virtual and physical memory in arbitrary amounts, see 'xen_remap_memory' and | |
324 | * its callers. | |
4fbb67e3 | 325 | */ |
1f3ac86b | 326 | static void __init xen_do_set_identity_and_remap_chunk( |
4fbb67e3 | 327 | unsigned long start_pfn, unsigned long size, unsigned long remap_pfn) |
83d51ab4 | 328 | { |
1f3ac86b JG |
329 | unsigned long buf = (unsigned long)&xen_remap_buf; |
330 | unsigned long mfn_save, mfn; | |
4fbb67e3 | 331 | unsigned long ident_pfn_iter, remap_pfn_iter; |
1f3ac86b | 332 | unsigned long ident_end_pfn = start_pfn + size; |
4fbb67e3 | 333 | unsigned long left = size; |
1f3ac86b | 334 | unsigned int i, chunk; |
4fbb67e3 MR |
335 | |
336 | WARN_ON(size == 0); | |
337 | ||
338 | BUG_ON(xen_feature(XENFEAT_auto_translated_physmap)); | |
83d51ab4 | 339 | |
1f3ac86b | 340 | mfn_save = virt_to_mfn(buf); |
e201bfcc | 341 | |
1f3ac86b JG |
342 | for (ident_pfn_iter = start_pfn, remap_pfn_iter = remap_pfn; |
343 | ident_pfn_iter < ident_end_pfn; | |
344 | ident_pfn_iter += REMAP_SIZE, remap_pfn_iter += REMAP_SIZE) { | |
345 | chunk = (left < REMAP_SIZE) ? left : REMAP_SIZE; | |
4fbb67e3 | 346 | |
1f3ac86b JG |
347 | /* Map first pfn to xen_remap_buf */ |
348 | mfn = pfn_to_mfn(ident_pfn_iter); | |
349 | set_pte_mfn(buf, mfn, PAGE_KERNEL); | |
4fbb67e3 | 350 | |
1f3ac86b JG |
351 | /* Save mapping information in page */ |
352 | xen_remap_buf.next_area_mfn = xen_remap_mfn; | |
353 | xen_remap_buf.target_pfn = remap_pfn_iter; | |
354 | xen_remap_buf.size = chunk; | |
355 | for (i = 0; i < chunk; i++) | |
356 | xen_remap_buf.mfns[i] = pfn_to_mfn(ident_pfn_iter + i); | |
4fbb67e3 | 357 | |
1f3ac86b JG |
358 | /* Put remap buf into list. */ |
359 | xen_remap_mfn = mfn; | |
4fbb67e3 | 360 | |
1f3ac86b | 361 | /* Set identity map */ |
bc7142cf | 362 | set_phys_range_identity(ident_pfn_iter, ident_pfn_iter + chunk); |
83d51ab4 | 363 | |
1f3ac86b | 364 | left -= chunk; |
4fbb67e3 | 365 | } |
83d51ab4 | 366 | |
1f3ac86b JG |
367 | /* Restore old xen_remap_buf mapping */ |
368 | set_pte_mfn(buf, mfn_save, PAGE_KERNEL); | |
83d51ab4 DV |
369 | } |
370 | ||
4fbb67e3 MR |
371 | /* |
372 | * This function takes a contiguous pfn range that needs to be identity mapped | |
373 | * and: | |
374 | * | |
375 | * 1) Finds a new range of pfns to use to remap based on E820 and remap_pfn. | |
376 | * 2) Calls the do_ function to actually do the mapping/remapping work. | |
377 | * | |
378 | * The goal is to not allocate additional memory but to remap the existing | |
379 | * pages. In the case of an error the underlying memory is simply released back | |
380 | * to Xen and not remapped. | |
381 | */ | |
76f0a486 | 382 | static unsigned long __init xen_set_identity_and_remap_chunk( |
69632ecf | 383 | unsigned long start_pfn, unsigned long end_pfn, unsigned long nr_pages, |
5097cdf6 | 384 | unsigned long remap_pfn) |
4fbb67e3 MR |
385 | { |
386 | unsigned long pfn; | |
387 | unsigned long i = 0; | |
388 | unsigned long n = end_pfn - start_pfn; | |
389 | ||
390 | while (i < n) { | |
391 | unsigned long cur_pfn = start_pfn + i; | |
392 | unsigned long left = n - i; | |
393 | unsigned long size = left; | |
394 | unsigned long remap_range_size; | |
395 | ||
396 | /* Do not remap pages beyond the current allocation */ | |
397 | if (cur_pfn >= nr_pages) { | |
398 | /* Identity map remaining pages */ | |
bc7142cf | 399 | set_phys_range_identity(cur_pfn, cur_pfn + size); |
4fbb67e3 MR |
400 | break; |
401 | } | |
402 | if (cur_pfn + size > nr_pages) | |
403 | size = nr_pages - cur_pfn; | |
404 | ||
69632ecf | 405 | remap_range_size = xen_find_pfn_range(&remap_pfn); |
4fbb67e3 MR |
406 | if (!remap_range_size) { |
407 | pr_warning("Unable to find available pfn range, not remapping identity pages\n"); | |
408 | xen_set_identity_and_release_chunk(cur_pfn, | |
5097cdf6 | 409 | cur_pfn + left, nr_pages); |
4fbb67e3 MR |
410 | break; |
411 | } | |
412 | /* Adjust size to fit in current e820 RAM region */ | |
413 | if (size > remap_range_size) | |
414 | size = remap_range_size; | |
415 | ||
1f3ac86b | 416 | xen_do_set_identity_and_remap_chunk(cur_pfn, size, remap_pfn); |
4fbb67e3 MR |
417 | |
418 | /* Update variables to reflect new mappings. */ | |
419 | i += size; | |
420 | remap_pfn += size; | |
4fbb67e3 MR |
421 | } |
422 | ||
423 | /* | |
424 | * If the PFNs are currently mapped, the VA mapping also needs | |
425 | * to be updated to be 1:1. | |
426 | */ | |
427 | for (pfn = start_pfn; pfn <= max_pfn_mapped && pfn < end_pfn; pfn++) | |
428 | (void)HYPERVISOR_update_va_mapping( | |
429 | (unsigned long)__va(pfn << PAGE_SHIFT), | |
430 | mfn_pte(pfn, PAGE_KERNEL_IO), 0); | |
431 | ||
432 | return remap_pfn; | |
433 | } | |
434 | ||
5097cdf6 | 435 | static void __init xen_set_identity_and_remap(unsigned long nr_pages) |
093d7b46 | 436 | { |
f3f436e3 | 437 | phys_addr_t start = 0; |
4fbb67e3 | 438 | unsigned long last_pfn = nr_pages; |
69632ecf | 439 | const struct e820entry *entry = xen_e820_map; |
68df0da7 KRW |
440 | int i; |
441 | ||
f3f436e3 DV |
442 | /* |
443 | * Combine non-RAM regions and gaps until a RAM region (or the | |
444 | * end of the map) is reached, then set the 1:1 map and | |
4fbb67e3 | 445 | * remap the memory in those non-RAM regions. |
f3f436e3 DV |
446 | * |
447 | * The combined non-RAM regions are rounded to a whole number | |
448 | * of pages so any partial pages are accessible via the 1:1 | |
449 | * mapping. This is needed for some BIOSes that put (for | |
450 | * example) the DMI tables in a reserved region that begins on | |
451 | * a non-page boundary. | |
452 | */ | |
69632ecf | 453 | for (i = 0; i < xen_e820_map_entries; i++, entry++) { |
f3f436e3 | 454 | phys_addr_t end = entry->addr + entry->size; |
69632ecf | 455 | if (entry->type == E820_RAM || i == xen_e820_map_entries - 1) { |
f3f436e3 DV |
456 | unsigned long start_pfn = PFN_DOWN(start); |
457 | unsigned long end_pfn = PFN_UP(end); | |
68df0da7 | 458 | |
f3f436e3 DV |
459 | if (entry->type == E820_RAM) |
460 | end_pfn = PFN_UP(entry->addr); | |
68df0da7 | 461 | |
83d51ab4 | 462 | if (start_pfn < end_pfn) |
4fbb67e3 | 463 | last_pfn = xen_set_identity_and_remap_chunk( |
69632ecf | 464 | start_pfn, end_pfn, nr_pages, |
5097cdf6 | 465 | last_pfn); |
f3f436e3 | 466 | start = end; |
68df0da7 | 467 | } |
68df0da7 | 468 | } |
f3f436e3 | 469 | |
5097cdf6 | 470 | pr_info("Released %ld page(s)\n", xen_released_pages); |
4fbb67e3 | 471 | } |
1f3ac86b JG |
472 | |
473 | /* | |
474 | * Remap the memory prepared in xen_do_set_identity_and_remap_chunk(). | |
475 | * The remap information (which mfn remap to which pfn) is contained in the | |
476 | * to be remapped memory itself in a linked list anchored at xen_remap_mfn. | |
477 | * This scheme allows to remap the different chunks in arbitrary order while | |
478 | * the resulting mapping will be independant from the order. | |
479 | */ | |
480 | void __init xen_remap_memory(void) | |
481 | { | |
482 | unsigned long buf = (unsigned long)&xen_remap_buf; | |
483 | unsigned long mfn_save, mfn, pfn; | |
484 | unsigned long remapped = 0; | |
485 | unsigned int i; | |
486 | unsigned long pfn_s = ~0UL; | |
487 | unsigned long len = 0; | |
488 | ||
489 | mfn_save = virt_to_mfn(buf); | |
490 | ||
491 | while (xen_remap_mfn != INVALID_P2M_ENTRY) { | |
492 | /* Map the remap information */ | |
493 | set_pte_mfn(buf, xen_remap_mfn, PAGE_KERNEL); | |
494 | ||
495 | BUG_ON(xen_remap_mfn != xen_remap_buf.mfns[0]); | |
496 | ||
497 | pfn = xen_remap_buf.target_pfn; | |
498 | for (i = 0; i < xen_remap_buf.size; i++) { | |
499 | mfn = xen_remap_buf.mfns[i]; | |
500 | xen_update_mem_tables(pfn, mfn); | |
501 | remapped++; | |
502 | pfn++; | |
503 | } | |
504 | if (pfn_s == ~0UL || pfn == pfn_s) { | |
505 | pfn_s = xen_remap_buf.target_pfn; | |
506 | len += xen_remap_buf.size; | |
507 | } else if (pfn_s + len == xen_remap_buf.target_pfn) { | |
508 | len += xen_remap_buf.size; | |
509 | } else { | |
5b8e7d80 | 510 | xen_del_extra_mem(PFN_PHYS(pfn_s), PFN_PHYS(len)); |
1f3ac86b JG |
511 | pfn_s = xen_remap_buf.target_pfn; |
512 | len = xen_remap_buf.size; | |
513 | } | |
514 | ||
515 | mfn = xen_remap_mfn; | |
516 | xen_remap_mfn = xen_remap_buf.next_area_mfn; | |
517 | } | |
518 | ||
519 | if (pfn_s != ~0UL && len) | |
5b8e7d80 | 520 | xen_del_extra_mem(PFN_PHYS(pfn_s), PFN_PHYS(len)); |
1f3ac86b JG |
521 | |
522 | set_pte_mfn(buf, mfn_save, PAGE_KERNEL); | |
523 | ||
524 | pr_info("Remapped %ld page(s)\n", remapped); | |
525 | } | |
526 | ||
c70727a5 JG |
527 | static unsigned long __init xen_get_pages_limit(void) |
528 | { | |
529 | unsigned long limit; | |
530 | ||
531 | #ifdef CONFIG_X86_32 | |
532 | limit = GB(64) / PAGE_SIZE; | |
533 | #else | |
534 | limit = ~0ul; | |
535 | if (!xen_initial_domain() && xen_512gb_limit) | |
536 | limit = GB(512) / PAGE_SIZE; | |
537 | #endif | |
538 | return limit; | |
539 | } | |
540 | ||
d312ae87 DV |
541 | static unsigned long __init xen_get_max_pages(void) |
542 | { | |
c70727a5 | 543 | unsigned long max_pages, limit; |
d312ae87 DV |
544 | domid_t domid = DOMID_SELF; |
545 | int ret; | |
546 | ||
c70727a5 JG |
547 | limit = xen_get_pages_limit(); |
548 | max_pages = limit; | |
549 | ||
d3db7281 IC |
550 | /* |
551 | * For the initial domain we use the maximum reservation as | |
552 | * the maximum page. | |
553 | * | |
554 | * For guest domains the current maximum reservation reflects | |
555 | * the current maximum rather than the static maximum. In this | |
556 | * case the e820 map provided to us will cover the static | |
557 | * maximum region. | |
558 | */ | |
559 | if (xen_initial_domain()) { | |
560 | ret = HYPERVISOR_memory_op(XENMEM_maximum_reservation, &domid); | |
561 | if (ret > 0) | |
562 | max_pages = ret; | |
563 | } | |
564 | ||
c70727a5 | 565 | return min(max_pages, limit); |
d312ae87 DV |
566 | } |
567 | ||
a3f52396 JG |
568 | static void __init xen_align_and_add_e820_region(phys_addr_t start, |
569 | phys_addr_t size, int type) | |
dc91c728 | 570 | { |
3ba5c867 | 571 | phys_addr_t end = start + size; |
dc91c728 DV |
572 | |
573 | /* Align RAM regions to page boundaries. */ | |
574 | if (type == E820_RAM) { | |
575 | start = PAGE_ALIGN(start); | |
3ba5c867 | 576 | end &= ~((phys_addr_t)PAGE_SIZE - 1); |
dc91c728 DV |
577 | } |
578 | ||
579 | e820_add_region(start, end - start, type); | |
580 | } | |
581 | ||
69632ecf | 582 | static void __init xen_ignore_unusable(void) |
3bc38cbc | 583 | { |
69632ecf | 584 | struct e820entry *entry = xen_e820_map; |
3bc38cbc DV |
585 | unsigned int i; |
586 | ||
69632ecf | 587 | for (i = 0; i < xen_e820_map_entries; i++, entry++) { |
3bc38cbc DV |
588 | if (entry->type == E820_UNUSABLE) |
589 | entry->type = E820_RAM; | |
590 | } | |
591 | } | |
592 | ||
5097cdf6 JG |
593 | static unsigned long __init xen_count_remap_pages(unsigned long max_pfn) |
594 | { | |
595 | unsigned long extra = 0; | |
596 | const struct e820entry *entry = xen_e820_map; | |
597 | int i; | |
598 | ||
599 | for (i = 0; i < xen_e820_map_entries; i++, entry++) { | |
600 | unsigned long start_pfn = PFN_DOWN(entry->addr); | |
601 | unsigned long end_pfn = PFN_UP(entry->addr + entry->size); | |
602 | ||
603 | if (start_pfn >= max_pfn) | |
604 | break; | |
605 | if (entry->type == E820_RAM) | |
606 | continue; | |
607 | if (end_pfn >= max_pfn) | |
608 | end_pfn = max_pfn; | |
609 | extra += end_pfn - start_pfn; | |
610 | } | |
611 | ||
612 | return extra; | |
613 | } | |
614 | ||
e612b4a7 JG |
615 | bool __init xen_is_e820_reserved(phys_addr_t start, phys_addr_t size) |
616 | { | |
617 | struct e820entry *entry; | |
618 | unsigned mapcnt; | |
619 | phys_addr_t end; | |
620 | ||
621 | if (!size) | |
622 | return false; | |
623 | ||
624 | end = start + size; | |
625 | entry = xen_e820_map; | |
626 | ||
627 | for (mapcnt = 0; mapcnt < xen_e820_map_entries; mapcnt++) { | |
628 | if (entry->type == E820_RAM && entry->addr <= start && | |
629 | (entry->addr + entry->size) >= end) | |
630 | return false; | |
631 | ||
632 | entry++; | |
633 | } | |
634 | ||
635 | return true; | |
636 | } | |
637 | ||
9ddac5b7 JG |
638 | /* |
639 | * Find a free area in physical memory not yet reserved and compliant with | |
640 | * E820 map. | |
641 | * Used to relocate pre-allocated areas like initrd or p2m list which are in | |
642 | * conflict with the to be used E820 map. | |
643 | * In case no area is found, return 0. Otherwise return the physical address | |
644 | * of the area which is already reserved for convenience. | |
645 | */ | |
646 | phys_addr_t __init xen_find_free_area(phys_addr_t size) | |
647 | { | |
648 | unsigned mapcnt; | |
649 | phys_addr_t addr, start; | |
650 | struct e820entry *entry = xen_e820_map; | |
651 | ||
652 | for (mapcnt = 0; mapcnt < xen_e820_map_entries; mapcnt++, entry++) { | |
653 | if (entry->type != E820_RAM || entry->size < size) | |
654 | continue; | |
655 | start = entry->addr; | |
656 | for (addr = start; addr < start + size; addr += PAGE_SIZE) { | |
657 | if (!memblock_is_reserved(addr)) | |
658 | continue; | |
659 | start = addr + PAGE_SIZE; | |
660 | if (start + size > entry->addr + entry->size) | |
661 | break; | |
662 | } | |
663 | if (addr >= start + size) { | |
664 | memblock_reserve(start, size); | |
665 | return start; | |
666 | } | |
667 | } | |
668 | ||
669 | return 0; | |
670 | } | |
671 | ||
4b9c1537 JG |
672 | /* |
673 | * Like memcpy, but with physical addresses for dest and src. | |
674 | */ | |
675 | static void __init xen_phys_memcpy(phys_addr_t dest, phys_addr_t src, | |
676 | phys_addr_t n) | |
677 | { | |
678 | phys_addr_t dest_off, src_off, dest_len, src_len, len; | |
679 | void *from, *to; | |
680 | ||
681 | while (n) { | |
682 | dest_off = dest & ~PAGE_MASK; | |
683 | src_off = src & ~PAGE_MASK; | |
684 | dest_len = n; | |
685 | if (dest_len > (NR_FIX_BTMAPS << PAGE_SHIFT) - dest_off) | |
686 | dest_len = (NR_FIX_BTMAPS << PAGE_SHIFT) - dest_off; | |
687 | src_len = n; | |
688 | if (src_len > (NR_FIX_BTMAPS << PAGE_SHIFT) - src_off) | |
689 | src_len = (NR_FIX_BTMAPS << PAGE_SHIFT) - src_off; | |
690 | len = min(dest_len, src_len); | |
691 | to = early_memremap(dest - dest_off, dest_len + dest_off); | |
692 | from = early_memremap(src - src_off, src_len + src_off); | |
693 | memcpy(to, from, len); | |
694 | early_memunmap(to, dest_len + dest_off); | |
695 | early_memunmap(from, src_len + src_off); | |
696 | n -= len; | |
697 | dest += len; | |
698 | src += len; | |
699 | } | |
700 | } | |
701 | ||
8f5b0c63 JG |
702 | /* |
703 | * Reserve Xen mfn_list. | |
8f5b0c63 JG |
704 | */ |
705 | static void __init xen_reserve_xen_mfnlist(void) | |
706 | { | |
70e61199 JG |
707 | phys_addr_t start, size; |
708 | ||
8f5b0c63 | 709 | if (xen_start_info->mfn_list >= __START_KERNEL_map) { |
70e61199 JG |
710 | start = __pa(xen_start_info->mfn_list); |
711 | size = PFN_ALIGN(xen_start_info->nr_pages * | |
712 | sizeof(unsigned long)); | |
713 | } else { | |
714 | start = PFN_PHYS(xen_start_info->first_p2m_pfn); | |
715 | size = PFN_PHYS(xen_start_info->nr_p2m_frames); | |
716 | } | |
717 | ||
718 | if (!xen_is_e820_reserved(start, size)) { | |
719 | memblock_reserve(start, size); | |
8f5b0c63 JG |
720 | return; |
721 | } | |
722 | ||
70e61199 JG |
723 | #ifdef CONFIG_X86_32 |
724 | /* | |
725 | * Relocating the p2m on 32 bit system to an arbitrary virtual address | |
726 | * is not supported, so just give up. | |
727 | */ | |
728 | xen_raw_console_write("Xen hypervisor allocated p2m list conflicts with E820 map\n"); | |
729 | BUG(); | |
730 | #else | |
731 | xen_relocate_p2m(); | |
732 | #endif | |
8f5b0c63 JG |
733 | } |
734 | ||
5ead97c8 JF |
735 | /** |
736 | * machine_specific_memory_setup - Hook for machine specific memory setup. | |
737 | **/ | |
5ead97c8 JF |
738 | char * __init xen_memory_setup(void) |
739 | { | |
c70727a5 | 740 | unsigned long max_pfn; |
5097cdf6 JG |
741 | phys_addr_t mem_end, addr, size, chunk_size; |
742 | u32 type; | |
35ae11fd IC |
743 | int rc; |
744 | struct xen_memory_map memmap; | |
dc91c728 | 745 | unsigned long max_pages; |
42ee1471 | 746 | unsigned long extra_pages = 0; |
35ae11fd | 747 | int i; |
9e9a5fcb | 748 | int op; |
5ead97c8 | 749 | |
c70727a5 JG |
750 | xen_parse_512gb(); |
751 | max_pfn = xen_get_pages_limit(); | |
752 | max_pfn = min(max_pfn, xen_start_info->nr_pages); | |
35ae11fd IC |
753 | mem_end = PFN_PHYS(max_pfn); |
754 | ||
755 | memmap.nr_entries = E820MAX; | |
69632ecf | 756 | set_xen_guest_handle(memmap.buffer, xen_e820_map); |
35ae11fd | 757 | |
9e9a5fcb IC |
758 | op = xen_initial_domain() ? |
759 | XENMEM_machine_memory_map : | |
760 | XENMEM_memory_map; | |
761 | rc = HYPERVISOR_memory_op(op, &memmap); | |
35ae11fd | 762 | if (rc == -ENOSYS) { |
9ec23a7f | 763 | BUG_ON(xen_initial_domain()); |
35ae11fd | 764 | memmap.nr_entries = 1; |
69632ecf JG |
765 | xen_e820_map[0].addr = 0ULL; |
766 | xen_e820_map[0].size = mem_end; | |
35ae11fd | 767 | /* 8MB slack (to balance backend allocations). */ |
69632ecf JG |
768 | xen_e820_map[0].size += 8ULL << 20; |
769 | xen_e820_map[0].type = E820_RAM; | |
35ae11fd IC |
770 | rc = 0; |
771 | } | |
772 | BUG_ON(rc); | |
1ea644c8 | 773 | BUG_ON(memmap.nr_entries == 0); |
69632ecf | 774 | xen_e820_map_entries = memmap.nr_entries; |
8006ec3e | 775 | |
3bc38cbc DV |
776 | /* |
777 | * Xen won't allow a 1:1 mapping to be created to UNUSABLE | |
778 | * regions, so if we're using the machine memory map leave the | |
779 | * region as RAM as it is in the pseudo-physical map. | |
780 | * | |
781 | * UNUSABLE regions in domUs are not handled and will need | |
782 | * a patch in the future. | |
783 | */ | |
784 | if (xen_initial_domain()) | |
69632ecf | 785 | xen_ignore_unusable(); |
3bc38cbc | 786 | |
dc91c728 | 787 | /* Make sure the Xen-supplied memory map is well-ordered. */ |
69632ecf JG |
788 | sanitize_e820_map(xen_e820_map, xen_e820_map_entries, |
789 | &xen_e820_map_entries); | |
dc91c728 DV |
790 | |
791 | max_pages = xen_get_max_pages(); | |
792 | if (max_pages > max_pfn) | |
793 | extra_pages += max_pages - max_pfn; | |
794 | ||
5097cdf6 JG |
795 | /* How many extra pages do we need due to remapping? */ |
796 | extra_pages += xen_count_remap_pages(max_pfn); | |
2e2fb754 | 797 | |
dc91c728 DV |
798 | /* |
799 | * Clamp the amount of extra memory to a EXTRA_MEM_RATIO | |
800 | * factor the base size. On non-highmem systems, the base | |
801 | * size is the full initial memory allocation; on highmem it | |
802 | * is limited to the max size of lowmem, so that it doesn't | |
803 | * get completely filled. | |
804 | * | |
c70727a5 JG |
805 | * Make sure we have no memory above max_pages, as this area |
806 | * isn't handled by the p2m management. | |
807 | * | |
dc91c728 DV |
808 | * In principle there could be a problem in lowmem systems if |
809 | * the initial memory is also very large with respect to | |
810 | * lowmem, but we won't try to deal with that here. | |
811 | */ | |
c70727a5 JG |
812 | extra_pages = min3(EXTRA_MEM_RATIO * min(max_pfn, PFN_DOWN(MAXMEM)), |
813 | extra_pages, max_pages - max_pfn); | |
dc91c728 | 814 | i = 0; |
5097cdf6 JG |
815 | addr = xen_e820_map[0].addr; |
816 | size = xen_e820_map[0].size; | |
69632ecf | 817 | while (i < xen_e820_map_entries) { |
5097cdf6 JG |
818 | chunk_size = size; |
819 | type = xen_e820_map[i].type; | |
dc91c728 DV |
820 | |
821 | if (type == E820_RAM) { | |
822 | if (addr < mem_end) { | |
5097cdf6 | 823 | chunk_size = min(size, mem_end - addr); |
dc91c728 | 824 | } else if (extra_pages) { |
5097cdf6 JG |
825 | chunk_size = min(size, PFN_PHYS(extra_pages)); |
826 | extra_pages -= PFN_DOWN(chunk_size); | |
827 | xen_add_extra_mem(addr, chunk_size); | |
828 | xen_max_p2m_pfn = PFN_DOWN(addr + chunk_size); | |
dc91c728 DV |
829 | } else |
830 | type = E820_UNUSABLE; | |
3654581e JF |
831 | } |
832 | ||
5097cdf6 | 833 | xen_align_and_add_e820_region(addr, chunk_size, type); |
b5b43ced | 834 | |
5097cdf6 JG |
835 | addr += chunk_size; |
836 | size -= chunk_size; | |
837 | if (size == 0) { | |
dc91c728 | 838 | i++; |
5097cdf6 JG |
839 | if (i < xen_e820_map_entries) { |
840 | addr = xen_e820_map[i].addr; | |
841 | size = xen_e820_map[i].size; | |
842 | } | |
843 | } | |
35ae11fd | 844 | } |
b792c755 | 845 | |
25b884a8 DV |
846 | /* |
847 | * Set the rest as identity mapped, in case PCI BARs are | |
848 | * located here. | |
25b884a8 | 849 | */ |
5097cdf6 | 850 | set_phys_range_identity(addr / PAGE_SIZE, ~0ul); |
25b884a8 | 851 | |
b792c755 | 852 | /* |
9ec23a7f IC |
853 | * In domU, the ISA region is normal, usable memory, but we |
854 | * reserve ISA memory anyway because too many things poke | |
b792c755 JF |
855 | * about in there. |
856 | */ | |
857 | e820_add_region(ISA_START_ADDRESS, ISA_END_ADDRESS - ISA_START_ADDRESS, | |
858 | E820_RESERVED); | |
5ead97c8 | 859 | |
be5bf9fa JF |
860 | sanitize_e820_map(e820.map, ARRAY_SIZE(e820.map), &e820.nr_map); |
861 | ||
808fdb71 JG |
862 | /* |
863 | * Check whether the kernel itself conflicts with the target E820 map. | |
864 | * Failing now is better than running into weird problems later due | |
865 | * to relocating (and even reusing) pages with kernel text or data. | |
866 | */ | |
867 | if (xen_is_e820_reserved(__pa_symbol(_text), | |
868 | __pa_symbol(__bss_stop) - __pa_symbol(_text))) { | |
869 | xen_raw_console_write("Xen hypervisor allocated kernel memory conflicts with E820 map\n"); | |
870 | BUG(); | |
871 | } | |
872 | ||
04414baa JG |
873 | /* |
874 | * Check for a conflict of the hypervisor supplied page tables with | |
875 | * the target E820 map. | |
876 | */ | |
877 | xen_pt_check_e820(); | |
878 | ||
8f5b0c63 JG |
879 | xen_reserve_xen_mfnlist(); |
880 | ||
4b9c1537 JG |
881 | /* Check for a conflict of the initrd with the target E820 map. */ |
882 | if (xen_is_e820_reserved(boot_params.hdr.ramdisk_image, | |
883 | boot_params.hdr.ramdisk_size)) { | |
884 | phys_addr_t new_area, start, size; | |
885 | ||
886 | new_area = xen_find_free_area(boot_params.hdr.ramdisk_size); | |
887 | if (!new_area) { | |
888 | xen_raw_console_write("Can't find new memory area for initrd needed due to E820 map conflict\n"); | |
889 | BUG(); | |
890 | } | |
891 | ||
892 | start = boot_params.hdr.ramdisk_image; | |
893 | size = boot_params.hdr.ramdisk_size; | |
894 | xen_phys_memcpy(new_area, start, size); | |
895 | pr_info("initrd moved from [mem %#010llx-%#010llx] to [mem %#010llx-%#010llx]\n", | |
896 | start, start + size, new_area, new_area + size); | |
897 | memblock_free(start, size); | |
898 | boot_params.hdr.ramdisk_image = new_area; | |
899 | boot_params.ext_ramdisk_image = new_area >> 32; | |
900 | } | |
901 | ||
5097cdf6 JG |
902 | /* |
903 | * Set identity map on non-RAM pages and prepare remapping the | |
904 | * underlying RAM. | |
905 | */ | |
906 | xen_set_identity_and_remap(max_pfn); | |
907 | ||
5ead97c8 JF |
908 | return "Xen"; |
909 | } | |
910 | ||
abacaadc DV |
911 | /* |
912 | * Machine specific memory setup for auto-translated guests. | |
913 | */ | |
914 | char * __init xen_auto_xlated_memory_setup(void) | |
915 | { | |
abacaadc DV |
916 | struct xen_memory_map memmap; |
917 | int i; | |
918 | int rc; | |
919 | ||
920 | memmap.nr_entries = E820MAX; | |
69632ecf | 921 | set_xen_guest_handle(memmap.buffer, xen_e820_map); |
abacaadc DV |
922 | |
923 | rc = HYPERVISOR_memory_op(XENMEM_memory_map, &memmap); | |
924 | if (rc < 0) | |
925 | panic("No memory map (%d)\n", rc); | |
926 | ||
69632ecf JG |
927 | xen_e820_map_entries = memmap.nr_entries; |
928 | ||
929 | sanitize_e820_map(xen_e820_map, ARRAY_SIZE(xen_e820_map), | |
930 | &xen_e820_map_entries); | |
abacaadc | 931 | |
69632ecf JG |
932 | for (i = 0; i < xen_e820_map_entries; i++) |
933 | e820_add_region(xen_e820_map[i].addr, xen_e820_map[i].size, | |
934 | xen_e820_map[i].type); | |
abacaadc | 935 | |
70e61199 JG |
936 | /* Remove p2m info, it is not needed. */ |
937 | xen_start_info->mfn_list = 0; | |
938 | xen_start_info->first_p2m_pfn = 0; | |
939 | xen_start_info->nr_p2m_frames = 0; | |
abacaadc DV |
940 | |
941 | return "Xen"; | |
942 | } | |
943 | ||
d2eea68e RM |
944 | /* |
945 | * Set the bit indicating "nosegneg" library variants should be used. | |
6a52e4b1 JF |
946 | * We only need to bother in pure 32-bit mode; compat 32-bit processes |
947 | * can have un-truncated segments, so wrapping around is allowed. | |
d2eea68e | 948 | */ |
08b6d290 | 949 | static void __init fiddle_vdso(void) |
d2eea68e | 950 | { |
6a52e4b1 | 951 | #ifdef CONFIG_X86_32 |
6f121e54 AL |
952 | /* |
953 | * This could be called before selected_vdso32 is initialized, so | |
954 | * just fiddle with both possible images. vdso_image_32_syscall | |
955 | * can't be selected, since it only exists on 64-bit systems. | |
956 | */ | |
6a52e4b1 | 957 | u32 *mask; |
6f121e54 AL |
958 | mask = vdso_image_32_int80.data + |
959 | vdso_image_32_int80.sym_VDSO32_NOTE_MASK; | |
6a52e4b1 | 960 | *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; |
6f121e54 AL |
961 | mask = vdso_image_32_sysenter.data + |
962 | vdso_image_32_sysenter.sym_VDSO32_NOTE_MASK; | |
d2eea68e | 963 | *mask |= 1 << VDSO_NOTE_NONEGSEG_BIT; |
6fcac6d3 | 964 | #endif |
d2eea68e RM |
965 | } |
966 | ||
148f9bb8 | 967 | static int register_callback(unsigned type, const void *func) |
e2a81baf | 968 | { |
88459d4c JF |
969 | struct callback_register callback = { |
970 | .type = type, | |
971 | .address = XEN_CALLBACK(__KERNEL_CS, func), | |
e2a81baf JF |
972 | .flags = CALLBACKF_mask_events, |
973 | }; | |
974 | ||
88459d4c JF |
975 | return HYPERVISOR_callback_op(CALLBACKOP_register, &callback); |
976 | } | |
977 | ||
148f9bb8 | 978 | void xen_enable_sysenter(void) |
88459d4c | 979 | { |
6fcac6d3 | 980 | int ret; |
62541c37 | 981 | unsigned sysenter_feature; |
6fcac6d3 JF |
982 | |
983 | #ifdef CONFIG_X86_32 | |
62541c37 | 984 | sysenter_feature = X86_FEATURE_SEP; |
6fcac6d3 | 985 | #else |
62541c37 | 986 | sysenter_feature = X86_FEATURE_SYSENTER32; |
6fcac6d3 | 987 | #endif |
88459d4c | 988 | |
62541c37 JF |
989 | if (!boot_cpu_has(sysenter_feature)) |
990 | return; | |
991 | ||
6fcac6d3 | 992 | ret = register_callback(CALLBACKTYPE_sysenter, xen_sysenter_target); |
62541c37 JF |
993 | if(ret != 0) |
994 | setup_clear_cpu_cap(sysenter_feature); | |
e2a81baf JF |
995 | } |
996 | ||
148f9bb8 | 997 | void xen_enable_syscall(void) |
6fcac6d3 JF |
998 | { |
999 | #ifdef CONFIG_X86_64 | |
6fcac6d3 | 1000 | int ret; |
6fcac6d3 JF |
1001 | |
1002 | ret = register_callback(CALLBACKTYPE_syscall, xen_syscall_target); | |
1003 | if (ret != 0) { | |
d5303b81 | 1004 | printk(KERN_ERR "Failed to set syscall callback: %d\n", ret); |
62541c37 JF |
1005 | /* Pretty fatal; 64-bit userspace has no other |
1006 | mechanism for syscalls. */ | |
1007 | } | |
1008 | ||
1009 | if (boot_cpu_has(X86_FEATURE_SYSCALL32)) { | |
6fcac6d3 JF |
1010 | ret = register_callback(CALLBACKTYPE_syscall32, |
1011 | xen_syscall32_target); | |
d5303b81 | 1012 | if (ret != 0) |
62541c37 | 1013 | setup_clear_cpu_cap(X86_FEATURE_SYSCALL32); |
6fcac6d3 JF |
1014 | } |
1015 | #endif /* CONFIG_X86_64 */ | |
1016 | } | |
ea9f9274 | 1017 | |
d285d683 | 1018 | void __init xen_pvmmu_arch_setup(void) |
5ead97c8 | 1019 | { |
5ead97c8 JF |
1020 | HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments); |
1021 | HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_writable_pagetables); | |
1022 | ||
d285d683 MR |
1023 | HYPERVISOR_vm_assist(VMASST_CMD_enable, |
1024 | VMASST_TYPE_pae_extended_cr3); | |
5ead97c8 | 1025 | |
88459d4c JF |
1026 | if (register_callback(CALLBACKTYPE_event, xen_hypervisor_callback) || |
1027 | register_callback(CALLBACKTYPE_failsafe, xen_failsafe_callback)) | |
1028 | BUG(); | |
5ead97c8 | 1029 | |
e2a81baf | 1030 | xen_enable_sysenter(); |
6fcac6d3 | 1031 | xen_enable_syscall(); |
d285d683 MR |
1032 | } |
1033 | ||
1034 | /* This function is not called for HVM domains */ | |
1035 | void __init xen_arch_setup(void) | |
1036 | { | |
1037 | xen_panic_handler_init(); | |
1038 | if (!xen_feature(XENFEAT_auto_translated_physmap)) | |
1039 | xen_pvmmu_arch_setup(); | |
1040 | ||
5ead97c8 JF |
1041 | #ifdef CONFIG_ACPI |
1042 | if (!(xen_start_info->flags & SIF_INITDOMAIN)) { | |
1043 | printk(KERN_INFO "ACPI in unprivileged domain disabled\n"); | |
1044 | disable_acpi(); | |
1045 | } | |
1046 | #endif | |
1047 | ||
1048 | memcpy(boot_command_line, xen_start_info->cmd_line, | |
1049 | MAX_GUEST_CMDLINE > COMMAND_LINE_SIZE ? | |
1050 | COMMAND_LINE_SIZE : MAX_GUEST_CMDLINE); | |
1051 | ||
bc15fde7 | 1052 | /* Set up idle, making sure it calls safe_halt() pvop */ |
d91ee586 | 1053 | disable_cpuidle(); |
48cdd828 | 1054 | disable_cpufreq(); |
6a377ddc | 1055 | WARN_ON(xen_set_default_idle()); |
d2eea68e | 1056 | fiddle_vdso(); |
8d54db79 KRW |
1057 | #ifdef CONFIG_NUMA |
1058 | numa_off = 1; | |
1059 | #endif | |
5ead97c8 | 1060 | } |