]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
5ead97c8 JF |
2 | /* |
3 | * Machine specific setup for xen | |
4 | * | |
5 | * Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007 | |
6 | */ | |
7 | ||
7a2463dc | 8 | #include <linux/init.h> |
5ead97c8 JF |
9 | #include <linux/sched.h> |
10 | #include <linux/mm.h> | |
11 | #include <linux/pm.h> | |
a9ce6bc1 | 12 | #include <linux/memblock.h> |
d91ee586 | 13 | #include <linux/cpuidle.h> |
48cdd828 | 14 | #include <linux/cpufreq.h> |
1d988ed4 | 15 | #include <linux/memory_hotplug.h> |
5ead97c8 JF |
16 | |
17 | #include <asm/elf.h> | |
6c3652ef | 18 | #include <asm/vdso.h> |
66441bd3 | 19 | #include <asm/e820/api.h> |
5ead97c8 | 20 | #include <asm/setup.h> |
b792c755 | 21 | #include <asm/acpi.h> |
8d54db79 | 22 | #include <asm/numa.h> |
2f6474e4 | 23 | #include <asm/idtentry.h> |
5ead97c8 JF |
24 | #include <asm/xen/hypervisor.h> |
25 | #include <asm/xen/hypercall.h> | |
26 | ||
45263cb0 | 27 | #include <xen/xen.h> |
8006ec3e | 28 | #include <xen/page.h> |
e2a81baf | 29 | #include <xen/interface/callback.h> |
35ae11fd | 30 | #include <xen/interface/memory.h> |
5ead97c8 JF |
31 | #include <xen/interface/physdev.h> |
32 | #include <xen/features.h> | |
808fdb71 | 33 | #include <xen/hvc-console.h> |
5ead97c8 | 34 | #include "xen-ops.h" |
1f3ac86b | 35 | #include "mmu.h" |
5ead97c8 | 36 | |
c70727a5 JG |
37 | #define GB(x) ((uint64_t)(x) * 1024 * 1024 * 1024) |
38 | ||
42ee1471 | 39 | /* Amount of extra memory space we add to the e820 ranges */ |
8b5d44a5 | 40 | struct xen_memory_region xen_extra_mem[XEN_EXTRA_MEM_MAX_REGIONS] __initdata; |
42ee1471 | 41 | |
aa24411b DV |
42 | /* Number of pages released from the initial allocation. */ |
43 | unsigned long xen_released_pages; | |
44 | ||
69632ecf | 45 | /* E820 map used during setting up memory. */ |
e7dbf7ad | 46 | static struct e820_table xen_e820_table __initdata; |
69632ecf | 47 | |
1f3ac86b JG |
48 | /* |
49 | * Buffer used to remap identity mapped pages. We only need the virtual space. | |
50 | * The physical page behind this address is remapped as needed to different | |
51 | * buffer pages. | |
52 | */ | |
53 | #define REMAP_SIZE (P2M_PER_PAGE - 3) | |
54 | static struct { | |
55 | unsigned long next_area_mfn; | |
56 | unsigned long target_pfn; | |
57 | unsigned long size; | |
58 | unsigned long mfns[REMAP_SIZE]; | |
59 | } xen_remap_buf __initdata __aligned(PAGE_SIZE); | |
60 | static unsigned long xen_remap_mfn __initdata = INVALID_P2M_ENTRY; | |
4fbb67e3 | 61 | |
58e0c804 RPM |
62 | /* |
63 | * The maximum amount of extra memory compared to the base size. The | |
64 | * main scaling factor is the size of struct page. At extreme ratios | |
65 | * of base:extra, all the base memory can be filled with page | |
66 | * structures for the extra memory, leaving no space for anything | |
67 | * else. | |
68 | * | |
69 | * 10x seems like a reasonable balance between scaling flexibility and | |
70 | * leaving a practically usable system. | |
71 | */ | |
72 | #define EXTRA_MEM_RATIO (10) | |
73 | ||
c70727a5 JG |
74 | static bool xen_512gb_limit __initdata = IS_ENABLED(CONFIG_XEN_512GB); |
75 | ||
76 | static void __init xen_parse_512gb(void) | |
77 | { | |
78 | bool val = false; | |
79 | char *arg; | |
80 | ||
81 | arg = strstr(xen_start_info->cmd_line, "xen_512gb_limit"); | |
82 | if (!arg) | |
83 | return; | |
84 | ||
85 | arg = strstr(xen_start_info->cmd_line, "xen_512gb_limit="); | |
86 | if (!arg) | |
87 | val = true; | |
88 | else if (strtobool(arg + strlen("xen_512gb_limit="), &val)) | |
89 | return; | |
90 | ||
91 | xen_512gb_limit = val; | |
92 | } | |
93 | ||
626d7508 JG |
94 | static void __init xen_add_extra_mem(unsigned long start_pfn, |
95 | unsigned long n_pfns) | |
42ee1471 | 96 | { |
dc91c728 | 97 | int i; |
6eaa412f | 98 | |
626d7508 JG |
99 | /* |
100 | * No need to check for zero size, should happen rarely and will only | |
101 | * write a new entry regarded to be unused due to zero size. | |
102 | */ | |
dc91c728 DV |
103 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { |
104 | /* Add new region. */ | |
626d7508 JG |
105 | if (xen_extra_mem[i].n_pfns == 0) { |
106 | xen_extra_mem[i].start_pfn = start_pfn; | |
107 | xen_extra_mem[i].n_pfns = n_pfns; | |
dc91c728 DV |
108 | break; |
109 | } | |
110 | /* Append to existing region. */ | |
626d7508 JG |
111 | if (xen_extra_mem[i].start_pfn + xen_extra_mem[i].n_pfns == |
112 | start_pfn) { | |
113 | xen_extra_mem[i].n_pfns += n_pfns; | |
dc91c728 DV |
114 | break; |
115 | } | |
116 | } | |
117 | if (i == XEN_EXTRA_MEM_MAX_REGIONS) | |
118 | printk(KERN_WARNING "Warning: not enough extra memory regions\n"); | |
42ee1471 | 119 | |
626d7508 | 120 | memblock_reserve(PFN_PHYS(start_pfn), PFN_PHYS(n_pfns)); |
5b8e7d80 | 121 | } |
2f7acb20 | 122 | |
626d7508 JG |
123 | static void __init xen_del_extra_mem(unsigned long start_pfn, |
124 | unsigned long n_pfns) | |
5b8e7d80 JG |
125 | { |
126 | int i; | |
626d7508 | 127 | unsigned long start_r, size_r; |
c96aae1f | 128 | |
5b8e7d80 | 129 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { |
626d7508 JG |
130 | start_r = xen_extra_mem[i].start_pfn; |
131 | size_r = xen_extra_mem[i].n_pfns; | |
5b8e7d80 JG |
132 | |
133 | /* Start of region. */ | |
626d7508 JG |
134 | if (start_r == start_pfn) { |
135 | BUG_ON(n_pfns > size_r); | |
136 | xen_extra_mem[i].start_pfn += n_pfns; | |
137 | xen_extra_mem[i].n_pfns -= n_pfns; | |
5b8e7d80 JG |
138 | break; |
139 | } | |
140 | /* End of region. */ | |
626d7508 JG |
141 | if (start_r + size_r == start_pfn + n_pfns) { |
142 | BUG_ON(n_pfns > size_r); | |
143 | xen_extra_mem[i].n_pfns -= n_pfns; | |
5b8e7d80 JG |
144 | break; |
145 | } | |
146 | /* Mid of region. */ | |
626d7508 JG |
147 | if (start_pfn > start_r && start_pfn < start_r + size_r) { |
148 | BUG_ON(start_pfn + n_pfns > start_r + size_r); | |
149 | xen_extra_mem[i].n_pfns = start_pfn - start_r; | |
5b8e7d80 | 150 | /* Calling memblock_reserve() again is okay. */ |
626d7508 JG |
151 | xen_add_extra_mem(start_pfn + n_pfns, start_r + size_r - |
152 | (start_pfn + n_pfns)); | |
5b8e7d80 JG |
153 | break; |
154 | } | |
155 | } | |
626d7508 | 156 | memblock_free(PFN_PHYS(start_pfn), PFN_PHYS(n_pfns)); |
5b8e7d80 JG |
157 | } |
158 | ||
159 | /* | |
160 | * Called during boot before the p2m list can take entries beyond the | |
161 | * hypervisor supplied p2m list. Entries in extra mem are to be regarded as | |
162 | * invalid. | |
163 | */ | |
164 | unsigned long __ref xen_chk_extra_mem(unsigned long pfn) | |
165 | { | |
166 | int i; | |
6eaa412f | 167 | |
5b8e7d80 | 168 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { |
626d7508 JG |
169 | if (pfn >= xen_extra_mem[i].start_pfn && |
170 | pfn < xen_extra_mem[i].start_pfn + xen_extra_mem[i].n_pfns) | |
5b8e7d80 JG |
171 | return INVALID_P2M_ENTRY; |
172 | } | |
173 | ||
174 | return IDENTITY_FRAME(pfn); | |
175 | } | |
176 | ||
177 | /* | |
178 | * Mark all pfns of extra mem as invalid in p2m list. | |
179 | */ | |
180 | void __init xen_inv_extra_mem(void) | |
181 | { | |
182 | unsigned long pfn, pfn_s, pfn_e; | |
183 | int i; | |
184 | ||
185 | for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) { | |
626d7508 | 186 | if (!xen_extra_mem[i].n_pfns) |
9a17ad7f | 187 | continue; |
626d7508 JG |
188 | pfn_s = xen_extra_mem[i].start_pfn; |
189 | pfn_e = pfn_s + xen_extra_mem[i].n_pfns; | |
5b8e7d80 JG |
190 | for (pfn = pfn_s; pfn < pfn_e; pfn++) |
191 | set_phys_to_machine(pfn, INVALID_P2M_ENTRY); | |
c96aae1f | 192 | } |
42ee1471 JF |
193 | } |
194 | ||
4fbb67e3 MR |
195 | /* |
196 | * Finds the next RAM pfn available in the E820 map after min_pfn. | |
197 | * This function updates min_pfn with the pfn found and returns | |
198 | * the size of that range or zero if not found. | |
199 | */ | |
69632ecf | 200 | static unsigned long __init xen_find_pfn_range(unsigned long *min_pfn) |
2e2fb754 | 201 | { |
e7dbf7ad | 202 | const struct e820_entry *entry = xen_e820_table.entries; |
2e2fb754 KRW |
203 | unsigned int i; |
204 | unsigned long done = 0; | |
2e2fb754 | 205 | |
e7dbf7ad | 206 | for (i = 0; i < xen_e820_table.nr_entries; i++, entry++) { |
2e2fb754 KRW |
207 | unsigned long s_pfn; |
208 | unsigned long e_pfn; | |
2e2fb754 | 209 | |
09821ff1 | 210 | if (entry->type != E820_TYPE_RAM) |
2e2fb754 KRW |
211 | continue; |
212 | ||
c3d93f88 | 213 | e_pfn = PFN_DOWN(entry->addr + entry->size); |
2e2fb754 | 214 | |
4fbb67e3 | 215 | /* We only care about E820 after this */ |
abed7d07 | 216 | if (e_pfn <= *min_pfn) |
2e2fb754 KRW |
217 | continue; |
218 | ||
c3d93f88 | 219 | s_pfn = PFN_UP(entry->addr); |
4fbb67e3 MR |
220 | |
221 | /* If min_pfn falls within the E820 entry, we want to start | |
222 | * at the min_pfn PFN. | |
2e2fb754 | 223 | */ |
4fbb67e3 MR |
224 | if (s_pfn <= *min_pfn) { |
225 | done = e_pfn - *min_pfn; | |
2e2fb754 | 226 | } else { |
4fbb67e3 MR |
227 | done = e_pfn - s_pfn; |
228 | *min_pfn = s_pfn; | |
2e2fb754 | 229 | } |
4fbb67e3 MR |
230 | break; |
231 | } | |
2e2fb754 | 232 | |
4fbb67e3 MR |
233 | return done; |
234 | } | |
2e2fb754 | 235 | |
1f3ac86b JG |
236 | static int __init xen_free_mfn(unsigned long mfn) |
237 | { | |
238 | struct xen_memory_reservation reservation = { | |
239 | .address_bits = 0, | |
240 | .extent_order = 0, | |
241 | .domid = DOMID_SELF | |
242 | }; | |
243 | ||
244 | set_xen_guest_handle(reservation.extent_start, &mfn); | |
245 | reservation.nr_extents = 1; | |
246 | ||
247 | return HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation); | |
248 | } | |
249 | ||
4fbb67e3 | 250 | /* |
1f3ac86b | 251 | * This releases a chunk of memory and then does the identity map. It's used |
4fbb67e3 MR |
252 | * as a fallback if the remapping fails. |
253 | */ | |
254 | static void __init xen_set_identity_and_release_chunk(unsigned long start_pfn, | |
5097cdf6 | 255 | unsigned long end_pfn, unsigned long nr_pages) |
4fbb67e3 | 256 | { |
1f3ac86b JG |
257 | unsigned long pfn, end; |
258 | int ret; | |
259 | ||
4fbb67e3 MR |
260 | WARN_ON(start_pfn > end_pfn); |
261 | ||
bc7142cf | 262 | /* Release pages first. */ |
1f3ac86b JG |
263 | end = min(end_pfn, nr_pages); |
264 | for (pfn = start_pfn; pfn < end; pfn++) { | |
265 | unsigned long mfn = pfn_to_mfn(pfn); | |
266 | ||
267 | /* Make sure pfn exists to start with */ | |
268 | if (mfn == INVALID_P2M_ENTRY || mfn_to_pfn(mfn) != pfn) | |
269 | continue; | |
270 | ||
271 | ret = xen_free_mfn(mfn); | |
272 | WARN(ret != 1, "Failed to release pfn %lx err=%d\n", pfn, ret); | |
273 | ||
274 | if (ret == 1) { | |
5097cdf6 | 275 | xen_released_pages++; |
1f3ac86b JG |
276 | if (!__set_phys_to_machine(pfn, INVALID_P2M_ENTRY)) |
277 | break; | |
1f3ac86b JG |
278 | } else |
279 | break; | |
280 | } | |
281 | ||
bc7142cf | 282 | set_phys_range_identity(start_pfn, end_pfn); |
4fbb67e3 MR |
283 | } |
284 | ||
285 | /* | |
1f3ac86b | 286 | * Helper function to update the p2m and m2p tables and kernel mapping. |
4fbb67e3 | 287 | */ |
1f3ac86b | 288 | static void __init xen_update_mem_tables(unsigned long pfn, unsigned long mfn) |
4fbb67e3 MR |
289 | { |
290 | struct mmu_update update = { | |
3ba5c867 | 291 | .ptr = ((uint64_t)mfn << PAGE_SHIFT) | MMU_MACHPHYS_UPDATE, |
4fbb67e3 MR |
292 | .val = pfn |
293 | }; | |
294 | ||
295 | /* Update p2m */ | |
1f3ac86b | 296 | if (!set_phys_to_machine(pfn, mfn)) { |
4fbb67e3 MR |
297 | WARN(1, "Failed to set p2m mapping for pfn=%ld mfn=%ld\n", |
298 | pfn, mfn); | |
1f3ac86b | 299 | BUG(); |
2e2fb754 | 300 | } |
4fbb67e3 MR |
301 | |
302 | /* Update m2p */ | |
303 | if (HYPERVISOR_mmu_update(&update, 1, NULL, DOMID_SELF) < 0) { | |
304 | WARN(1, "Failed to set m2p mapping for mfn=%ld pfn=%ld\n", | |
305 | mfn, pfn); | |
1f3ac86b | 306 | BUG(); |
4fbb67e3 MR |
307 | } |
308 | ||
1f3ac86b | 309 | /* Update kernel mapping, but not for highmem. */ |
e86f9496 | 310 | if (pfn >= PFN_UP(__pa(high_memory - 1))) |
1f3ac86b JG |
311 | return; |
312 | ||
313 | if (HYPERVISOR_update_va_mapping((unsigned long)__va(pfn << PAGE_SHIFT), | |
314 | mfn_pte(mfn, PAGE_KERNEL), 0)) { | |
315 | WARN(1, "Failed to update kernel mapping for mfn=%ld pfn=%ld\n", | |
316 | mfn, pfn); | |
317 | BUG(); | |
318 | } | |
2e2fb754 | 319 | } |
83d51ab4 | 320 | |
4fbb67e3 MR |
321 | /* |
322 | * This function updates the p2m and m2p tables with an identity map from | |
1f3ac86b JG |
323 | * start_pfn to start_pfn+size and prepares remapping the underlying RAM of the |
324 | * original allocation at remap_pfn. The information needed for remapping is | |
325 | * saved in the memory itself to avoid the need for allocating buffers. The | |
326 | * complete remap information is contained in a list of MFNs each containing | |
327 | * up to REMAP_SIZE MFNs and the start target PFN for doing the remap. | |
328 | * This enables us to preserve the original mfn sequence while doing the | |
329 | * remapping at a time when the memory management is capable of allocating | |
330 | * virtual and physical memory in arbitrary amounts, see 'xen_remap_memory' and | |
331 | * its callers. | |
4fbb67e3 | 332 | */ |
1f3ac86b | 333 | static void __init xen_do_set_identity_and_remap_chunk( |
4fbb67e3 | 334 | unsigned long start_pfn, unsigned long size, unsigned long remap_pfn) |
83d51ab4 | 335 | { |
1f3ac86b JG |
336 | unsigned long buf = (unsigned long)&xen_remap_buf; |
337 | unsigned long mfn_save, mfn; | |
4fbb67e3 | 338 | unsigned long ident_pfn_iter, remap_pfn_iter; |
1f3ac86b | 339 | unsigned long ident_end_pfn = start_pfn + size; |
4fbb67e3 | 340 | unsigned long left = size; |
1f3ac86b | 341 | unsigned int i, chunk; |
4fbb67e3 MR |
342 | |
343 | WARN_ON(size == 0); | |
344 | ||
1f3ac86b | 345 | mfn_save = virt_to_mfn(buf); |
e201bfcc | 346 | |
1f3ac86b JG |
347 | for (ident_pfn_iter = start_pfn, remap_pfn_iter = remap_pfn; |
348 | ident_pfn_iter < ident_end_pfn; | |
349 | ident_pfn_iter += REMAP_SIZE, remap_pfn_iter += REMAP_SIZE) { | |
350 | chunk = (left < REMAP_SIZE) ? left : REMAP_SIZE; | |
4fbb67e3 | 351 | |
1f3ac86b JG |
352 | /* Map first pfn to xen_remap_buf */ |
353 | mfn = pfn_to_mfn(ident_pfn_iter); | |
354 | set_pte_mfn(buf, mfn, PAGE_KERNEL); | |
4fbb67e3 | 355 | |
1f3ac86b JG |
356 | /* Save mapping information in page */ |
357 | xen_remap_buf.next_area_mfn = xen_remap_mfn; | |
358 | xen_remap_buf.target_pfn = remap_pfn_iter; | |
359 | xen_remap_buf.size = chunk; | |
360 | for (i = 0; i < chunk; i++) | |
361 | xen_remap_buf.mfns[i] = pfn_to_mfn(ident_pfn_iter + i); | |
4fbb67e3 | 362 | |
1f3ac86b JG |
363 | /* Put remap buf into list. */ |
364 | xen_remap_mfn = mfn; | |
4fbb67e3 | 365 | |
1f3ac86b | 366 | /* Set identity map */ |
bc7142cf | 367 | set_phys_range_identity(ident_pfn_iter, ident_pfn_iter + chunk); |
83d51ab4 | 368 | |
1f3ac86b | 369 | left -= chunk; |
4fbb67e3 | 370 | } |
83d51ab4 | 371 | |
1f3ac86b JG |
372 | /* Restore old xen_remap_buf mapping */ |
373 | set_pte_mfn(buf, mfn_save, PAGE_KERNEL); | |
83d51ab4 DV |
374 | } |
375 | ||
4fbb67e3 MR |
376 | /* |
377 | * This function takes a contiguous pfn range that needs to be identity mapped | |
378 | * and: | |
379 | * | |
380 | * 1) Finds a new range of pfns to use to remap based on E820 and remap_pfn. | |
381 | * 2) Calls the do_ function to actually do the mapping/remapping work. | |
382 | * | |
383 | * The goal is to not allocate additional memory but to remap the existing | |
384 | * pages. In the case of an error the underlying memory is simply released back | |
385 | * to Xen and not remapped. | |
386 | */ | |
76f0a486 | 387 | static unsigned long __init xen_set_identity_and_remap_chunk( |
69632ecf | 388 | unsigned long start_pfn, unsigned long end_pfn, unsigned long nr_pages, |
5097cdf6 | 389 | unsigned long remap_pfn) |
4fbb67e3 MR |
390 | { |
391 | unsigned long pfn; | |
392 | unsigned long i = 0; | |
393 | unsigned long n = end_pfn - start_pfn; | |
394 | ||
dd14be92 JG |
395 | if (remap_pfn == 0) |
396 | remap_pfn = nr_pages; | |
397 | ||
4fbb67e3 MR |
398 | while (i < n) { |
399 | unsigned long cur_pfn = start_pfn + i; | |
400 | unsigned long left = n - i; | |
401 | unsigned long size = left; | |
402 | unsigned long remap_range_size; | |
403 | ||
404 | /* Do not remap pages beyond the current allocation */ | |
405 | if (cur_pfn >= nr_pages) { | |
406 | /* Identity map remaining pages */ | |
bc7142cf | 407 | set_phys_range_identity(cur_pfn, cur_pfn + size); |
4fbb67e3 MR |
408 | break; |
409 | } | |
410 | if (cur_pfn + size > nr_pages) | |
411 | size = nr_pages - cur_pfn; | |
412 | ||
69632ecf | 413 | remap_range_size = xen_find_pfn_range(&remap_pfn); |
4fbb67e3 | 414 | if (!remap_range_size) { |
8d3bcc44 | 415 | pr_warn("Unable to find available pfn range, not remapping identity pages\n"); |
4fbb67e3 | 416 | xen_set_identity_and_release_chunk(cur_pfn, |
5097cdf6 | 417 | cur_pfn + left, nr_pages); |
4fbb67e3 MR |
418 | break; |
419 | } | |
420 | /* Adjust size to fit in current e820 RAM region */ | |
421 | if (size > remap_range_size) | |
422 | size = remap_range_size; | |
423 | ||
1f3ac86b | 424 | xen_do_set_identity_and_remap_chunk(cur_pfn, size, remap_pfn); |
4fbb67e3 MR |
425 | |
426 | /* Update variables to reflect new mappings. */ | |
427 | i += size; | |
428 | remap_pfn += size; | |
4fbb67e3 MR |
429 | } |
430 | ||
431 | /* | |
432 | * If the PFNs are currently mapped, the VA mapping also needs | |
433 | * to be updated to be 1:1. | |
434 | */ | |
435 | for (pfn = start_pfn; pfn <= max_pfn_mapped && pfn < end_pfn; pfn++) | |
436 | (void)HYPERVISOR_update_va_mapping( | |
437 | (unsigned long)__va(pfn << PAGE_SHIFT), | |
438 | mfn_pte(pfn, PAGE_KERNEL_IO), 0); | |
439 | ||
440 | return remap_pfn; | |
441 | } | |
442 | ||
dd14be92 JG |
443 | static unsigned long __init xen_count_remap_pages( |
444 | unsigned long start_pfn, unsigned long end_pfn, unsigned long nr_pages, | |
445 | unsigned long remap_pages) | |
446 | { | |
447 | if (start_pfn >= nr_pages) | |
448 | return remap_pages; | |
449 | ||
450 | return remap_pages + min(end_pfn, nr_pages) - start_pfn; | |
451 | } | |
452 | ||
453 | static unsigned long __init xen_foreach_remap_area(unsigned long nr_pages, | |
454 | unsigned long (*func)(unsigned long start_pfn, unsigned long end_pfn, | |
455 | unsigned long nr_pages, unsigned long last_val)) | |
093d7b46 | 456 | { |
f3f436e3 | 457 | phys_addr_t start = 0; |
dd14be92 | 458 | unsigned long ret_val = 0; |
e7dbf7ad | 459 | const struct e820_entry *entry = xen_e820_table.entries; |
68df0da7 KRW |
460 | int i; |
461 | ||
f3f436e3 DV |
462 | /* |
463 | * Combine non-RAM regions and gaps until a RAM region (or the | |
dd14be92 JG |
464 | * end of the map) is reached, then call the provided function |
465 | * to perform its duty on the non-RAM region. | |
f3f436e3 DV |
466 | * |
467 | * The combined non-RAM regions are rounded to a whole number | |
468 | * of pages so any partial pages are accessible via the 1:1 | |
469 | * mapping. This is needed for some BIOSes that put (for | |
470 | * example) the DMI tables in a reserved region that begins on | |
471 | * a non-page boundary. | |
472 | */ | |
e7dbf7ad | 473 | for (i = 0; i < xen_e820_table.nr_entries; i++, entry++) { |
f3f436e3 | 474 | phys_addr_t end = entry->addr + entry->size; |
e7dbf7ad | 475 | if (entry->type == E820_TYPE_RAM || i == xen_e820_table.nr_entries - 1) { |
f3f436e3 DV |
476 | unsigned long start_pfn = PFN_DOWN(start); |
477 | unsigned long end_pfn = PFN_UP(end); | |
68df0da7 | 478 | |
09821ff1 | 479 | if (entry->type == E820_TYPE_RAM) |
f3f436e3 | 480 | end_pfn = PFN_UP(entry->addr); |
68df0da7 | 481 | |
83d51ab4 | 482 | if (start_pfn < end_pfn) |
dd14be92 JG |
483 | ret_val = func(start_pfn, end_pfn, nr_pages, |
484 | ret_val); | |
f3f436e3 | 485 | start = end; |
68df0da7 | 486 | } |
68df0da7 | 487 | } |
f3f436e3 | 488 | |
dd14be92 | 489 | return ret_val; |
4fbb67e3 | 490 | } |
1f3ac86b JG |
491 | |
492 | /* | |
493 | * Remap the memory prepared in xen_do_set_identity_and_remap_chunk(). | |
494 | * The remap information (which mfn remap to which pfn) is contained in the | |
495 | * to be remapped memory itself in a linked list anchored at xen_remap_mfn. | |
496 | * This scheme allows to remap the different chunks in arbitrary order while | |
a97673a1 | 497 | * the resulting mapping will be independent from the order. |
1f3ac86b JG |
498 | */ |
499 | void __init xen_remap_memory(void) | |
500 | { | |
501 | unsigned long buf = (unsigned long)&xen_remap_buf; | |
bf1b9ddf | 502 | unsigned long mfn_save, pfn; |
1f3ac86b JG |
503 | unsigned long remapped = 0; |
504 | unsigned int i; | |
505 | unsigned long pfn_s = ~0UL; | |
506 | unsigned long len = 0; | |
507 | ||
508 | mfn_save = virt_to_mfn(buf); | |
509 | ||
510 | while (xen_remap_mfn != INVALID_P2M_ENTRY) { | |
511 | /* Map the remap information */ | |
512 | set_pte_mfn(buf, xen_remap_mfn, PAGE_KERNEL); | |
513 | ||
514 | BUG_ON(xen_remap_mfn != xen_remap_buf.mfns[0]); | |
515 | ||
516 | pfn = xen_remap_buf.target_pfn; | |
517 | for (i = 0; i < xen_remap_buf.size; i++) { | |
bf1b9ddf | 518 | xen_update_mem_tables(pfn, xen_remap_buf.mfns[i]); |
1f3ac86b JG |
519 | remapped++; |
520 | pfn++; | |
521 | } | |
522 | if (pfn_s == ~0UL || pfn == pfn_s) { | |
523 | pfn_s = xen_remap_buf.target_pfn; | |
524 | len += xen_remap_buf.size; | |
525 | } else if (pfn_s + len == xen_remap_buf.target_pfn) { | |
526 | len += xen_remap_buf.size; | |
527 | } else { | |
626d7508 | 528 | xen_del_extra_mem(pfn_s, len); |
1f3ac86b JG |
529 | pfn_s = xen_remap_buf.target_pfn; |
530 | len = xen_remap_buf.size; | |
531 | } | |
1f3ac86b JG |
532 | xen_remap_mfn = xen_remap_buf.next_area_mfn; |
533 | } | |
534 | ||
535 | if (pfn_s != ~0UL && len) | |
626d7508 | 536 | xen_del_extra_mem(pfn_s, len); |
1f3ac86b JG |
537 | |
538 | set_pte_mfn(buf, mfn_save, PAGE_KERNEL); | |
539 | ||
540 | pr_info("Remapped %ld page(s)\n", remapped); | |
541 | } | |
542 | ||
c70727a5 JG |
543 | static unsigned long __init xen_get_pages_limit(void) |
544 | { | |
545 | unsigned long limit; | |
546 | ||
cb9e444b | 547 | limit = MAXMEM / PAGE_SIZE; |
c70727a5 JG |
548 | if (!xen_initial_domain() && xen_512gb_limit) |
549 | limit = GB(512) / PAGE_SIZE; | |
a13f2ef1 | 550 | |
c70727a5 JG |
551 | return limit; |
552 | } | |
553 | ||
d312ae87 DV |
554 | static unsigned long __init xen_get_max_pages(void) |
555 | { | |
c70727a5 | 556 | unsigned long max_pages, limit; |
d312ae87 | 557 | domid_t domid = DOMID_SELF; |
24f775a6 | 558 | long ret; |
d312ae87 | 559 | |
c70727a5 JG |
560 | limit = xen_get_pages_limit(); |
561 | max_pages = limit; | |
562 | ||
d3db7281 IC |
563 | /* |
564 | * For the initial domain we use the maximum reservation as | |
565 | * the maximum page. | |
566 | * | |
567 | * For guest domains the current maximum reservation reflects | |
568 | * the current maximum rather than the static maximum. In this | |
569 | * case the e820 map provided to us will cover the static | |
570 | * maximum region. | |
571 | */ | |
572 | if (xen_initial_domain()) { | |
573 | ret = HYPERVISOR_memory_op(XENMEM_maximum_reservation, &domid); | |
574 | if (ret > 0) | |
575 | max_pages = ret; | |
576 | } | |
577 | ||
c70727a5 | 578 | return min(max_pages, limit); |
d312ae87 DV |
579 | } |
580 | ||
a3f52396 JG |
581 | static void __init xen_align_and_add_e820_region(phys_addr_t start, |
582 | phys_addr_t size, int type) | |
dc91c728 | 583 | { |
3ba5c867 | 584 | phys_addr_t end = start + size; |
dc91c728 DV |
585 | |
586 | /* Align RAM regions to page boundaries. */ | |
09821ff1 | 587 | if (type == E820_TYPE_RAM) { |
dc91c728 | 588 | start = PAGE_ALIGN(start); |
3ba5c867 | 589 | end &= ~((phys_addr_t)PAGE_SIZE - 1); |
1d988ed4 JG |
590 | #ifdef CONFIG_MEMORY_HOTPLUG |
591 | /* | |
592 | * Don't allow adding memory not in E820 map while booting the | |
593 | * system. Once the balloon driver is up it will remove that | |
594 | * restriction again. | |
595 | */ | |
596 | max_mem_size = end; | |
597 | #endif | |
dc91c728 DV |
598 | } |
599 | ||
ab6bc04c | 600 | e820__range_add(start, end - start, type); |
dc91c728 DV |
601 | } |
602 | ||
69632ecf | 603 | static void __init xen_ignore_unusable(void) |
3bc38cbc | 604 | { |
e7dbf7ad | 605 | struct e820_entry *entry = xen_e820_table.entries; |
3bc38cbc DV |
606 | unsigned int i; |
607 | ||
e7dbf7ad | 608 | for (i = 0; i < xen_e820_table.nr_entries; i++, entry++) { |
09821ff1 IM |
609 | if (entry->type == E820_TYPE_UNUSABLE) |
610 | entry->type = E820_TYPE_RAM; | |
3bc38cbc DV |
611 | } |
612 | } | |
613 | ||
e612b4a7 JG |
614 | bool __init xen_is_e820_reserved(phys_addr_t start, phys_addr_t size) |
615 | { | |
8ec67d97 | 616 | struct e820_entry *entry; |
e612b4a7 JG |
617 | unsigned mapcnt; |
618 | phys_addr_t end; | |
619 | ||
620 | if (!size) | |
621 | return false; | |
622 | ||
623 | end = start + size; | |
e7dbf7ad | 624 | entry = xen_e820_table.entries; |
e612b4a7 | 625 | |
e7dbf7ad | 626 | for (mapcnt = 0; mapcnt < xen_e820_table.nr_entries; mapcnt++) { |
09821ff1 | 627 | if (entry->type == E820_TYPE_RAM && entry->addr <= start && |
e612b4a7 JG |
628 | (entry->addr + entry->size) >= end) |
629 | return false; | |
630 | ||
631 | entry++; | |
632 | } | |
633 | ||
634 | return true; | |
635 | } | |
636 | ||
9ddac5b7 JG |
637 | /* |
638 | * Find a free area in physical memory not yet reserved and compliant with | |
639 | * E820 map. | |
640 | * Used to relocate pre-allocated areas like initrd or p2m list which are in | |
641 | * conflict with the to be used E820 map. | |
642 | * In case no area is found, return 0. Otherwise return the physical address | |
643 | * of the area which is already reserved for convenience. | |
644 | */ | |
645 | phys_addr_t __init xen_find_free_area(phys_addr_t size) | |
646 | { | |
647 | unsigned mapcnt; | |
648 | phys_addr_t addr, start; | |
e7dbf7ad | 649 | struct e820_entry *entry = xen_e820_table.entries; |
9ddac5b7 | 650 | |
e7dbf7ad | 651 | for (mapcnt = 0; mapcnt < xen_e820_table.nr_entries; mapcnt++, entry++) { |
09821ff1 | 652 | if (entry->type != E820_TYPE_RAM || entry->size < size) |
9ddac5b7 JG |
653 | continue; |
654 | start = entry->addr; | |
655 | for (addr = start; addr < start + size; addr += PAGE_SIZE) { | |
656 | if (!memblock_is_reserved(addr)) | |
657 | continue; | |
658 | start = addr + PAGE_SIZE; | |
659 | if (start + size > entry->addr + entry->size) | |
660 | break; | |
661 | } | |
662 | if (addr >= start + size) { | |
663 | memblock_reserve(start, size); | |
664 | return start; | |
665 | } | |
666 | } | |
667 | ||
668 | return 0; | |
669 | } | |
670 | ||
4b9c1537 JG |
671 | /* |
672 | * Like memcpy, but with physical addresses for dest and src. | |
673 | */ | |
674 | static void __init xen_phys_memcpy(phys_addr_t dest, phys_addr_t src, | |
675 | phys_addr_t n) | |
676 | { | |
677 | phys_addr_t dest_off, src_off, dest_len, src_len, len; | |
678 | void *from, *to; | |
679 | ||
680 | while (n) { | |
681 | dest_off = dest & ~PAGE_MASK; | |
682 | src_off = src & ~PAGE_MASK; | |
683 | dest_len = n; | |
684 | if (dest_len > (NR_FIX_BTMAPS << PAGE_SHIFT) - dest_off) | |
685 | dest_len = (NR_FIX_BTMAPS << PAGE_SHIFT) - dest_off; | |
686 | src_len = n; | |
687 | if (src_len > (NR_FIX_BTMAPS << PAGE_SHIFT) - src_off) | |
688 | src_len = (NR_FIX_BTMAPS << PAGE_SHIFT) - src_off; | |
689 | len = min(dest_len, src_len); | |
690 | to = early_memremap(dest - dest_off, dest_len + dest_off); | |
691 | from = early_memremap(src - src_off, src_len + src_off); | |
692 | memcpy(to, from, len); | |
693 | early_memunmap(to, dest_len + dest_off); | |
694 | early_memunmap(from, src_len + src_off); | |
695 | n -= len; | |
696 | dest += len; | |
697 | src += len; | |
698 | } | |
699 | } | |
700 | ||
8f5b0c63 JG |
701 | /* |
702 | * Reserve Xen mfn_list. | |
8f5b0c63 JG |
703 | */ |
704 | static void __init xen_reserve_xen_mfnlist(void) | |
705 | { | |
70e61199 JG |
706 | phys_addr_t start, size; |
707 | ||
8f5b0c63 | 708 | if (xen_start_info->mfn_list >= __START_KERNEL_map) { |
70e61199 JG |
709 | start = __pa(xen_start_info->mfn_list); |
710 | size = PFN_ALIGN(xen_start_info->nr_pages * | |
711 | sizeof(unsigned long)); | |
712 | } else { | |
713 | start = PFN_PHYS(xen_start_info->first_p2m_pfn); | |
714 | size = PFN_PHYS(xen_start_info->nr_p2m_frames); | |
715 | } | |
716 | ||
7ecec850 RL |
717 | memblock_reserve(start, size); |
718 | if (!xen_is_e820_reserved(start, size)) | |
8f5b0c63 | 719 | return; |
8f5b0c63 | 720 | |
70e61199 | 721 | xen_relocate_p2m(); |
7ecec850 | 722 | memblock_free(start, size); |
8f5b0c63 JG |
723 | } |
724 | ||
5ead97c8 JF |
725 | /** |
726 | * machine_specific_memory_setup - Hook for machine specific memory setup. | |
727 | **/ | |
5ead97c8 JF |
728 | char * __init xen_memory_setup(void) |
729 | { | |
626d7508 | 730 | unsigned long max_pfn, pfn_s, n_pfns; |
5097cdf6 JG |
731 | phys_addr_t mem_end, addr, size, chunk_size; |
732 | u32 type; | |
35ae11fd IC |
733 | int rc; |
734 | struct xen_memory_map memmap; | |
dc91c728 | 735 | unsigned long max_pages; |
42ee1471 | 736 | unsigned long extra_pages = 0; |
35ae11fd | 737 | int i; |
9e9a5fcb | 738 | int op; |
5ead97c8 | 739 | |
c70727a5 JG |
740 | xen_parse_512gb(); |
741 | max_pfn = xen_get_pages_limit(); | |
742 | max_pfn = min(max_pfn, xen_start_info->nr_pages); | |
35ae11fd IC |
743 | mem_end = PFN_PHYS(max_pfn); |
744 | ||
e7dbf7ad IM |
745 | memmap.nr_entries = ARRAY_SIZE(xen_e820_table.entries); |
746 | set_xen_guest_handle(memmap.buffer, xen_e820_table.entries); | |
35ae11fd | 747 | |
1d988ed4 JG |
748 | #if defined(CONFIG_MEMORY_HOTPLUG) && defined(CONFIG_XEN_BALLOON) |
749 | xen_saved_max_mem_size = max_mem_size; | |
750 | #endif | |
751 | ||
9e9a5fcb IC |
752 | op = xen_initial_domain() ? |
753 | XENMEM_machine_memory_map : | |
754 | XENMEM_memory_map; | |
755 | rc = HYPERVISOR_memory_op(op, &memmap); | |
35ae11fd | 756 | if (rc == -ENOSYS) { |
9ec23a7f | 757 | BUG_ON(xen_initial_domain()); |
35ae11fd | 758 | memmap.nr_entries = 1; |
e7dbf7ad IM |
759 | xen_e820_table.entries[0].addr = 0ULL; |
760 | xen_e820_table.entries[0].size = mem_end; | |
35ae11fd | 761 | /* 8MB slack (to balance backend allocations). */ |
e7dbf7ad IM |
762 | xen_e820_table.entries[0].size += 8ULL << 20; |
763 | xen_e820_table.entries[0].type = E820_TYPE_RAM; | |
35ae11fd IC |
764 | rc = 0; |
765 | } | |
766 | BUG_ON(rc); | |
1ea644c8 | 767 | BUG_ON(memmap.nr_entries == 0); |
e7dbf7ad | 768 | xen_e820_table.nr_entries = memmap.nr_entries; |
8006ec3e | 769 | |
3bc38cbc DV |
770 | /* |
771 | * Xen won't allow a 1:1 mapping to be created to UNUSABLE | |
772 | * regions, so if we're using the machine memory map leave the | |
773 | * region as RAM as it is in the pseudo-physical map. | |
774 | * | |
775 | * UNUSABLE regions in domUs are not handled and will need | |
776 | * a patch in the future. | |
777 | */ | |
778 | if (xen_initial_domain()) | |
69632ecf | 779 | xen_ignore_unusable(); |
3bc38cbc | 780 | |
dc91c728 | 781 | /* Make sure the Xen-supplied memory map is well-ordered. */ |
f9748fa0 | 782 | e820__update_table(&xen_e820_table); |
dc91c728 DV |
783 | |
784 | max_pages = xen_get_max_pages(); | |
dc91c728 | 785 | |
5097cdf6 | 786 | /* How many extra pages do we need due to remapping? */ |
dd14be92 | 787 | max_pages += xen_foreach_remap_area(max_pfn, xen_count_remap_pages); |
eafd72e0 JG |
788 | |
789 | if (max_pages > max_pfn) | |
790 | extra_pages += max_pages - max_pfn; | |
2e2fb754 | 791 | |
dc91c728 | 792 | /* |
58e0c804 | 793 | * Clamp the amount of extra memory to a EXTRA_MEM_RATIO |
415de44f | 794 | * factor the base size. |
dc91c728 | 795 | * |
c70727a5 JG |
796 | * Make sure we have no memory above max_pages, as this area |
797 | * isn't handled by the p2m management. | |
dc91c728 | 798 | */ |
58e0c804 | 799 | extra_pages = min3(EXTRA_MEM_RATIO * min(max_pfn, PFN_DOWN(MAXMEM)), |
c70727a5 | 800 | extra_pages, max_pages - max_pfn); |
dc91c728 | 801 | i = 0; |
e7dbf7ad IM |
802 | addr = xen_e820_table.entries[0].addr; |
803 | size = xen_e820_table.entries[0].size; | |
804 | while (i < xen_e820_table.nr_entries) { | |
12366410 | 805 | bool discard = false; |
f5775e0b | 806 | |
5097cdf6 | 807 | chunk_size = size; |
e7dbf7ad | 808 | type = xen_e820_table.entries[i].type; |
dc91c728 | 809 | |
09821ff1 | 810 | if (type == E820_TYPE_RAM) { |
dc91c728 | 811 | if (addr < mem_end) { |
5097cdf6 | 812 | chunk_size = min(size, mem_end - addr); |
dc91c728 | 813 | } else if (extra_pages) { |
5097cdf6 | 814 | chunk_size = min(size, PFN_PHYS(extra_pages)); |
626d7508 JG |
815 | pfn_s = PFN_UP(addr); |
816 | n_pfns = PFN_DOWN(addr + chunk_size) - pfn_s; | |
817 | extra_pages -= n_pfns; | |
818 | xen_add_extra_mem(pfn_s, n_pfns); | |
819 | xen_max_p2m_pfn = pfn_s + n_pfns; | |
dc91c728 | 820 | } else |
12366410 | 821 | discard = true; |
3654581e JF |
822 | } |
823 | ||
12366410 ID |
824 | if (!discard) |
825 | xen_align_and_add_e820_region(addr, chunk_size, type); | |
b5b43ced | 826 | |
5097cdf6 JG |
827 | addr += chunk_size; |
828 | size -= chunk_size; | |
829 | if (size == 0) { | |
dc91c728 | 830 | i++; |
e7dbf7ad IM |
831 | if (i < xen_e820_table.nr_entries) { |
832 | addr = xen_e820_table.entries[i].addr; | |
833 | size = xen_e820_table.entries[i].size; | |
5097cdf6 JG |
834 | } |
835 | } | |
35ae11fd | 836 | } |
b792c755 | 837 | |
25b884a8 DV |
838 | /* |
839 | * Set the rest as identity mapped, in case PCI BARs are | |
840 | * located here. | |
25b884a8 | 841 | */ |
5097cdf6 | 842 | set_phys_range_identity(addr / PAGE_SIZE, ~0ul); |
25b884a8 | 843 | |
b792c755 | 844 | /* |
9ec23a7f IC |
845 | * In domU, the ISA region is normal, usable memory, but we |
846 | * reserve ISA memory anyway because too many things poke | |
b792c755 JF |
847 | * about in there. |
848 | */ | |
f9748fa0 | 849 | e820__range_add(ISA_START_ADDRESS, ISA_END_ADDRESS - ISA_START_ADDRESS, E820_TYPE_RESERVED); |
5ead97c8 | 850 | |
f9748fa0 | 851 | e820__update_table(e820_table); |
be5bf9fa | 852 | |
808fdb71 JG |
853 | /* |
854 | * Check whether the kernel itself conflicts with the target E820 map. | |
855 | * Failing now is better than running into weird problems later due | |
856 | * to relocating (and even reusing) pages with kernel text or data. | |
857 | */ | |
858 | if (xen_is_e820_reserved(__pa_symbol(_text), | |
859 | __pa_symbol(__bss_stop) - __pa_symbol(_text))) { | |
860 | xen_raw_console_write("Xen hypervisor allocated kernel memory conflicts with E820 map\n"); | |
861 | BUG(); | |
862 | } | |
863 | ||
04414baa JG |
864 | /* |
865 | * Check for a conflict of the hypervisor supplied page tables with | |
866 | * the target E820 map. | |
867 | */ | |
868 | xen_pt_check_e820(); | |
869 | ||
8f5b0c63 JG |
870 | xen_reserve_xen_mfnlist(); |
871 | ||
4b9c1537 JG |
872 | /* Check for a conflict of the initrd with the target E820 map. */ |
873 | if (xen_is_e820_reserved(boot_params.hdr.ramdisk_image, | |
874 | boot_params.hdr.ramdisk_size)) { | |
875 | phys_addr_t new_area, start, size; | |
876 | ||
877 | new_area = xen_find_free_area(boot_params.hdr.ramdisk_size); | |
878 | if (!new_area) { | |
879 | xen_raw_console_write("Can't find new memory area for initrd needed due to E820 map conflict\n"); | |
880 | BUG(); | |
881 | } | |
882 | ||
883 | start = boot_params.hdr.ramdisk_image; | |
884 | size = boot_params.hdr.ramdisk_size; | |
885 | xen_phys_memcpy(new_area, start, size); | |
886 | pr_info("initrd moved from [mem %#010llx-%#010llx] to [mem %#010llx-%#010llx]\n", | |
887 | start, start + size, new_area, new_area + size); | |
888 | memblock_free(start, size); | |
889 | boot_params.hdr.ramdisk_image = new_area; | |
890 | boot_params.ext_ramdisk_image = new_area >> 32; | |
891 | } | |
892 | ||
5097cdf6 JG |
893 | /* |
894 | * Set identity map on non-RAM pages and prepare remapping the | |
895 | * underlying RAM. | |
896 | */ | |
dd14be92 JG |
897 | xen_foreach_remap_area(max_pfn, xen_set_identity_and_remap_chunk); |
898 | ||
899 | pr_info("Released %ld page(s)\n", xen_released_pages); | |
5097cdf6 | 900 | |
5ead97c8 JF |
901 | return "Xen"; |
902 | } | |
903 | ||
148f9bb8 | 904 | static int register_callback(unsigned type, const void *func) |
e2a81baf | 905 | { |
88459d4c JF |
906 | struct callback_register callback = { |
907 | .type = type, | |
908 | .address = XEN_CALLBACK(__KERNEL_CS, func), | |
e2a81baf JF |
909 | .flags = CALLBACKF_mask_events, |
910 | }; | |
911 | ||
88459d4c JF |
912 | return HYPERVISOR_callback_op(CALLBACKOP_register, &callback); |
913 | } | |
914 | ||
148f9bb8 | 915 | void xen_enable_sysenter(void) |
88459d4c | 916 | { |
6fcac6d3 | 917 | int ret; |
62541c37 | 918 | unsigned sysenter_feature; |
6fcac6d3 | 919 | |
62541c37 | 920 | sysenter_feature = X86_FEATURE_SYSENTER32; |
88459d4c | 921 | |
62541c37 JF |
922 | if (!boot_cpu_has(sysenter_feature)) |
923 | return; | |
924 | ||
6fcac6d3 | 925 | ret = register_callback(CALLBACKTYPE_sysenter, xen_sysenter_target); |
62541c37 JF |
926 | if(ret != 0) |
927 | setup_clear_cpu_cap(sysenter_feature); | |
e2a81baf JF |
928 | } |
929 | ||
148f9bb8 | 930 | void xen_enable_syscall(void) |
6fcac6d3 | 931 | { |
6fcac6d3 | 932 | int ret; |
6fcac6d3 JF |
933 | |
934 | ret = register_callback(CALLBACKTYPE_syscall, xen_syscall_target); | |
935 | if (ret != 0) { | |
d5303b81 | 936 | printk(KERN_ERR "Failed to set syscall callback: %d\n", ret); |
62541c37 JF |
937 | /* Pretty fatal; 64-bit userspace has no other |
938 | mechanism for syscalls. */ | |
939 | } | |
940 | ||
941 | if (boot_cpu_has(X86_FEATURE_SYSCALL32)) { | |
6fcac6d3 JF |
942 | ret = register_callback(CALLBACKTYPE_syscall32, |
943 | xen_syscall32_target); | |
d5303b81 | 944 | if (ret != 0) |
62541c37 | 945 | setup_clear_cpu_cap(X86_FEATURE_SYSCALL32); |
6fcac6d3 | 946 | } |
6fcac6d3 | 947 | } |
ea9f9274 | 948 | |
0e1b4271 | 949 | static void __init xen_pvmmu_arch_setup(void) |
5ead97c8 | 950 | { |
5ead97c8 JF |
951 | HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments); |
952 | HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_writable_pagetables); | |
953 | ||
d285d683 MR |
954 | HYPERVISOR_vm_assist(VMASST_CMD_enable, |
955 | VMASST_TYPE_pae_extended_cr3); | |
5ead97c8 | 956 | |
2f6474e4 TG |
957 | if (register_callback(CALLBACKTYPE_event, |
958 | xen_asm_exc_xen_hypervisor_callback) || | |
88459d4c JF |
959 | register_callback(CALLBACKTYPE_failsafe, xen_failsafe_callback)) |
960 | BUG(); | |
5ead97c8 | 961 | |
e2a81baf | 962 | xen_enable_sysenter(); |
6fcac6d3 | 963 | xen_enable_syscall(); |
d285d683 MR |
964 | } |
965 | ||
966 | /* This function is not called for HVM domains */ | |
967 | void __init xen_arch_setup(void) | |
968 | { | |
969 | xen_panic_handler_init(); | |
82616f95 | 970 | xen_pvmmu_arch_setup(); |
d285d683 | 971 | |
5ead97c8 JF |
972 | #ifdef CONFIG_ACPI |
973 | if (!(xen_start_info->flags & SIF_INITDOMAIN)) { | |
974 | printk(KERN_INFO "ACPI in unprivileged domain disabled\n"); | |
975 | disable_acpi(); | |
976 | } | |
977 | #endif | |
978 | ||
979 | memcpy(boot_command_line, xen_start_info->cmd_line, | |
980 | MAX_GUEST_CMDLINE > COMMAND_LINE_SIZE ? | |
981 | COMMAND_LINE_SIZE : MAX_GUEST_CMDLINE); | |
982 | ||
bc15fde7 | 983 | /* Set up idle, making sure it calls safe_halt() pvop */ |
d91ee586 | 984 | disable_cpuidle(); |
48cdd828 | 985 | disable_cpufreq(); |
6a377ddc | 986 | WARN_ON(xen_set_default_idle()); |
8d54db79 KRW |
987 | #ifdef CONFIG_NUMA |
988 | numa_off = 1; | |
989 | #endif | |
5ead97c8 | 990 | } |