Commit | Line | Data |
---|---|---|
5ead97c8 JF |
1 | /* |
2 | * Core of Xen paravirt_ops implementation. | |
3 | * | |
4 | * This file contains the xen_paravirt_ops structure itself, and the | |
5 | * implementations for: | |
6 | * - privileged instructions | |
7 | * - interrupt flags | |
8 | * - segment operations | |
9 | * - booting and setup | |
10 | * | |
11 | * Jeremy Fitzhardinge <jeremy@xensource.com>, XenSource Inc, 2007 | |
12 | */ | |
13 | ||
14 | #include <linux/kernel.h> | |
15 | #include <linux/init.h> | |
16 | #include <linux/smp.h> | |
17 | #include <linux/preempt.h> | |
18 | #include <linux/percpu.h> | |
19 | #include <linux/delay.h> | |
20 | #include <linux/start_kernel.h> | |
21 | #include <linux/sched.h> | |
22 | #include <linux/bootmem.h> | |
23 | #include <linux/module.h> | |
24 | ||
25 | #include <xen/interface/xen.h> | |
26 | #include <xen/interface/physdev.h> | |
27 | #include <xen/interface/vcpu.h> | |
28 | #include <xen/features.h> | |
29 | #include <xen/page.h> | |
30 | ||
31 | #include <asm/paravirt.h> | |
32 | #include <asm/page.h> | |
33 | #include <asm/xen/hypercall.h> | |
34 | #include <asm/xen/hypervisor.h> | |
35 | #include <asm/fixmap.h> | |
36 | #include <asm/processor.h> | |
37 | #include <asm/setup.h> | |
38 | #include <asm/desc.h> | |
39 | #include <asm/pgtable.h> | |
40 | ||
41 | #include "xen-ops.h" | |
3b827c1b | 42 | #include "mmu.h" |
5ead97c8 JF |
43 | #include "multicalls.h" |
44 | ||
45 | EXPORT_SYMBOL_GPL(hypercall_page); | |
46 | ||
47 | DEFINE_PER_CPU(enum paravirt_lazy_mode, xen_lazy_mode); | |
48 | ||
49 | DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu); | |
50 | DEFINE_PER_CPU(struct vcpu_info, xen_vcpu_info); | |
51 | DEFINE_PER_CPU(unsigned long, xen_cr3); | |
52 | ||
53 | struct start_info *xen_start_info; | |
54 | EXPORT_SYMBOL_GPL(xen_start_info); | |
55 | ||
56 | static void xen_vcpu_setup(int cpu) | |
57 | { | |
58 | per_cpu(xen_vcpu, cpu) = &HYPERVISOR_shared_info->vcpu_info[cpu]; | |
59 | } | |
60 | ||
61 | static void __init xen_banner(void) | |
62 | { | |
63 | printk(KERN_INFO "Booting paravirtualized kernel on %s\n", | |
64 | paravirt_ops.name); | |
65 | printk(KERN_INFO "Hypervisor signature: %s\n", xen_start_info->magic); | |
66 | } | |
67 | ||
68 | static void xen_cpuid(unsigned int *eax, unsigned int *ebx, | |
69 | unsigned int *ecx, unsigned int *edx) | |
70 | { | |
71 | unsigned maskedx = ~0; | |
72 | ||
73 | /* | |
74 | * Mask out inconvenient features, to try and disable as many | |
75 | * unsupported kernel subsystems as possible. | |
76 | */ | |
77 | if (*eax == 1) | |
78 | maskedx = ~((1 << X86_FEATURE_APIC) | /* disable APIC */ | |
79 | (1 << X86_FEATURE_ACPI) | /* disable ACPI */ | |
80 | (1 << X86_FEATURE_ACC)); /* thermal monitoring */ | |
81 | ||
82 | asm(XEN_EMULATE_PREFIX "cpuid" | |
83 | : "=a" (*eax), | |
84 | "=b" (*ebx), | |
85 | "=c" (*ecx), | |
86 | "=d" (*edx) | |
87 | : "0" (*eax), "2" (*ecx)); | |
88 | *edx &= maskedx; | |
89 | } | |
90 | ||
91 | static void xen_set_debugreg(int reg, unsigned long val) | |
92 | { | |
93 | HYPERVISOR_set_debugreg(reg, val); | |
94 | } | |
95 | ||
96 | static unsigned long xen_get_debugreg(int reg) | |
97 | { | |
98 | return HYPERVISOR_get_debugreg(reg); | |
99 | } | |
100 | ||
101 | static unsigned long xen_save_fl(void) | |
102 | { | |
103 | struct vcpu_info *vcpu; | |
104 | unsigned long flags; | |
105 | ||
106 | preempt_disable(); | |
107 | vcpu = x86_read_percpu(xen_vcpu); | |
108 | /* flag has opposite sense of mask */ | |
109 | flags = !vcpu->evtchn_upcall_mask; | |
110 | preempt_enable(); | |
111 | ||
112 | /* convert to IF type flag | |
113 | -0 -> 0x00000000 | |
114 | -1 -> 0xffffffff | |
115 | */ | |
116 | return (-flags) & X86_EFLAGS_IF; | |
117 | } | |
118 | ||
119 | static void xen_restore_fl(unsigned long flags) | |
120 | { | |
121 | struct vcpu_info *vcpu; | |
122 | ||
123 | preempt_disable(); | |
124 | ||
125 | /* convert from IF type flag */ | |
126 | flags = !(flags & X86_EFLAGS_IF); | |
127 | vcpu = x86_read_percpu(xen_vcpu); | |
128 | vcpu->evtchn_upcall_mask = flags; | |
129 | ||
130 | if (flags == 0) { | |
131 | /* Unmask then check (avoid races). We're only protecting | |
132 | against updates by this CPU, so there's no need for | |
133 | anything stronger. */ | |
134 | barrier(); | |
135 | ||
136 | if (unlikely(vcpu->evtchn_upcall_pending)) | |
137 | force_evtchn_callback(); | |
138 | preempt_enable(); | |
139 | } else | |
140 | preempt_enable_no_resched(); | |
141 | } | |
142 | ||
143 | static void xen_irq_disable(void) | |
144 | { | |
145 | struct vcpu_info *vcpu; | |
146 | preempt_disable(); | |
147 | vcpu = x86_read_percpu(xen_vcpu); | |
148 | vcpu->evtchn_upcall_mask = 1; | |
149 | preempt_enable_no_resched(); | |
150 | } | |
151 | ||
152 | static void xen_irq_enable(void) | |
153 | { | |
154 | struct vcpu_info *vcpu; | |
155 | ||
156 | preempt_disable(); | |
157 | vcpu = x86_read_percpu(xen_vcpu); | |
158 | vcpu->evtchn_upcall_mask = 0; | |
159 | ||
160 | /* Unmask then check (avoid races). We're only protecting | |
161 | against updates by this CPU, so there's no need for | |
162 | anything stronger. */ | |
163 | barrier(); | |
164 | ||
165 | if (unlikely(vcpu->evtchn_upcall_pending)) | |
166 | force_evtchn_callback(); | |
167 | preempt_enable(); | |
168 | } | |
169 | ||
170 | static void xen_safe_halt(void) | |
171 | { | |
172 | /* Blocking includes an implicit local_irq_enable(). */ | |
173 | if (HYPERVISOR_sched_op(SCHEDOP_block, 0) != 0) | |
174 | BUG(); | |
175 | } | |
176 | ||
177 | static void xen_halt(void) | |
178 | { | |
179 | if (irqs_disabled()) | |
180 | HYPERVISOR_vcpu_op(VCPUOP_down, smp_processor_id(), NULL); | |
181 | else | |
182 | xen_safe_halt(); | |
183 | } | |
184 | ||
185 | static void xen_set_lazy_mode(enum paravirt_lazy_mode mode) | |
186 | { | |
187 | switch (mode) { | |
188 | case PARAVIRT_LAZY_NONE: | |
189 | BUG_ON(x86_read_percpu(xen_lazy_mode) == PARAVIRT_LAZY_NONE); | |
190 | break; | |
191 | ||
192 | case PARAVIRT_LAZY_MMU: | |
193 | case PARAVIRT_LAZY_CPU: | |
194 | BUG_ON(x86_read_percpu(xen_lazy_mode) != PARAVIRT_LAZY_NONE); | |
195 | break; | |
196 | ||
197 | case PARAVIRT_LAZY_FLUSH: | |
198 | /* flush if necessary, but don't change state */ | |
199 | if (x86_read_percpu(xen_lazy_mode) != PARAVIRT_LAZY_NONE) | |
200 | xen_mc_flush(); | |
201 | return; | |
202 | } | |
203 | ||
204 | xen_mc_flush(); | |
205 | x86_write_percpu(xen_lazy_mode, mode); | |
206 | } | |
207 | ||
208 | static unsigned long xen_store_tr(void) | |
209 | { | |
210 | return 0; | |
211 | } | |
212 | ||
213 | static void xen_set_ldt(const void *addr, unsigned entries) | |
214 | { | |
215 | unsigned long linear_addr = (unsigned long)addr; | |
216 | struct mmuext_op *op; | |
217 | struct multicall_space mcs = xen_mc_entry(sizeof(*op)); | |
218 | ||
219 | op = mcs.args; | |
220 | op->cmd = MMUEXT_SET_LDT; | |
221 | if (linear_addr) { | |
222 | /* ldt my be vmalloced, use arbitrary_virt_to_machine */ | |
223 | xmaddr_t maddr; | |
224 | maddr = arbitrary_virt_to_machine((unsigned long)addr); | |
225 | linear_addr = (unsigned long)maddr.maddr; | |
226 | } | |
227 | op->arg1.linear_addr = linear_addr; | |
228 | op->arg2.nr_ents = entries; | |
229 | ||
230 | MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); | |
231 | ||
232 | xen_mc_issue(PARAVIRT_LAZY_CPU); | |
233 | } | |
234 | ||
235 | static void xen_load_gdt(const struct Xgt_desc_struct *dtr) | |
236 | { | |
237 | unsigned long *frames; | |
238 | unsigned long va = dtr->address; | |
239 | unsigned int size = dtr->size + 1; | |
240 | unsigned pages = (size + PAGE_SIZE - 1) / PAGE_SIZE; | |
241 | int f; | |
242 | struct multicall_space mcs; | |
243 | ||
244 | /* A GDT can be up to 64k in size, which corresponds to 8192 | |
245 | 8-byte entries, or 16 4k pages.. */ | |
246 | ||
247 | BUG_ON(size > 65536); | |
248 | BUG_ON(va & ~PAGE_MASK); | |
249 | ||
250 | mcs = xen_mc_entry(sizeof(*frames) * pages); | |
251 | frames = mcs.args; | |
252 | ||
253 | for (f = 0; va < dtr->address + size; va += PAGE_SIZE, f++) { | |
254 | frames[f] = virt_to_mfn(va); | |
255 | make_lowmem_page_readonly((void *)va); | |
256 | } | |
257 | ||
258 | MULTI_set_gdt(mcs.mc, frames, size / sizeof(struct desc_struct)); | |
259 | ||
260 | xen_mc_issue(PARAVIRT_LAZY_CPU); | |
261 | } | |
262 | ||
263 | static void load_TLS_descriptor(struct thread_struct *t, | |
264 | unsigned int cpu, unsigned int i) | |
265 | { | |
266 | struct desc_struct *gdt = get_cpu_gdt_table(cpu); | |
267 | xmaddr_t maddr = virt_to_machine(&gdt[GDT_ENTRY_TLS_MIN+i]); | |
268 | struct multicall_space mc = __xen_mc_entry(0); | |
269 | ||
270 | MULTI_update_descriptor(mc.mc, maddr.maddr, t->tls_array[i]); | |
271 | } | |
272 | ||
273 | static void xen_load_tls(struct thread_struct *t, unsigned int cpu) | |
274 | { | |
275 | xen_mc_batch(); | |
276 | ||
277 | load_TLS_descriptor(t, cpu, 0); | |
278 | load_TLS_descriptor(t, cpu, 1); | |
279 | load_TLS_descriptor(t, cpu, 2); | |
280 | ||
281 | xen_mc_issue(PARAVIRT_LAZY_CPU); | |
282 | } | |
283 | ||
284 | static void xen_write_ldt_entry(struct desc_struct *dt, int entrynum, | |
285 | u32 low, u32 high) | |
286 | { | |
287 | unsigned long lp = (unsigned long)&dt[entrynum]; | |
288 | xmaddr_t mach_lp = virt_to_machine(lp); | |
289 | u64 entry = (u64)high << 32 | low; | |
290 | ||
291 | xen_mc_flush(); | |
292 | if (HYPERVISOR_update_descriptor(mach_lp.maddr, entry)) | |
293 | BUG(); | |
294 | } | |
295 | ||
296 | static int cvt_gate_to_trap(int vector, u32 low, u32 high, | |
297 | struct trap_info *info) | |
298 | { | |
299 | u8 type, dpl; | |
300 | ||
301 | type = (high >> 8) & 0x1f; | |
302 | dpl = (high >> 13) & 3; | |
303 | ||
304 | if (type != 0xf && type != 0xe) | |
305 | return 0; | |
306 | ||
307 | info->vector = vector; | |
308 | info->address = (high & 0xffff0000) | (low & 0x0000ffff); | |
309 | info->cs = low >> 16; | |
310 | info->flags = dpl; | |
311 | /* interrupt gates clear IF */ | |
312 | if (type == 0xe) | |
313 | info->flags |= 4; | |
314 | ||
315 | return 1; | |
316 | } | |
317 | ||
318 | /* Locations of each CPU's IDT */ | |
319 | static DEFINE_PER_CPU(struct Xgt_desc_struct, idt_desc); | |
320 | ||
321 | /* Set an IDT entry. If the entry is part of the current IDT, then | |
322 | also update Xen. */ | |
323 | static void xen_write_idt_entry(struct desc_struct *dt, int entrynum, | |
324 | u32 low, u32 high) | |
325 | { | |
326 | ||
327 | int cpu = smp_processor_id(); | |
328 | unsigned long p = (unsigned long)&dt[entrynum]; | |
329 | unsigned long start = per_cpu(idt_desc, cpu).address; | |
330 | unsigned long end = start + per_cpu(idt_desc, cpu).size + 1; | |
331 | ||
332 | xen_mc_flush(); | |
333 | ||
334 | write_dt_entry(dt, entrynum, low, high); | |
335 | ||
336 | if (p >= start && (p + 8) <= end) { | |
337 | struct trap_info info[2]; | |
338 | ||
339 | info[1].address = 0; | |
340 | ||
341 | if (cvt_gate_to_trap(entrynum, low, high, &info[0])) | |
342 | if (HYPERVISOR_set_trap_table(info)) | |
343 | BUG(); | |
344 | } | |
345 | } | |
346 | ||
347 | /* Load a new IDT into Xen. In principle this can be per-CPU, so we | |
348 | hold a spinlock to protect the static traps[] array (static because | |
349 | it avoids allocation, and saves stack space). */ | |
350 | static void xen_load_idt(const struct Xgt_desc_struct *desc) | |
351 | { | |
352 | static DEFINE_SPINLOCK(lock); | |
353 | static struct trap_info traps[257]; | |
354 | ||
355 | int cpu = smp_processor_id(); | |
356 | unsigned in, out, count; | |
357 | ||
358 | per_cpu(idt_desc, cpu) = *desc; | |
359 | ||
360 | count = (desc->size+1) / 8; | |
361 | BUG_ON(count > 256); | |
362 | ||
363 | spin_lock(&lock); | |
364 | for (in = out = 0; in < count; in++) { | |
365 | const u32 *entry = (u32 *)(desc->address + in * 8); | |
366 | ||
367 | if (cvt_gate_to_trap(in, entry[0], entry[1], &traps[out])) | |
368 | out++; | |
369 | } | |
370 | traps[out].address = 0; | |
371 | ||
372 | xen_mc_flush(); | |
373 | if (HYPERVISOR_set_trap_table(traps)) | |
374 | BUG(); | |
375 | ||
376 | spin_unlock(&lock); | |
377 | } | |
378 | ||
379 | /* Write a GDT descriptor entry. Ignore LDT descriptors, since | |
380 | they're handled differently. */ | |
381 | static void xen_write_gdt_entry(struct desc_struct *dt, int entry, | |
382 | u32 low, u32 high) | |
383 | { | |
384 | switch ((high >> 8) & 0xff) { | |
385 | case DESCTYPE_LDT: | |
386 | case DESCTYPE_TSS: | |
387 | /* ignore */ | |
388 | break; | |
389 | ||
390 | default: { | |
391 | xmaddr_t maddr = virt_to_machine(&dt[entry]); | |
392 | u64 desc = (u64)high << 32 | low; | |
393 | ||
394 | xen_mc_flush(); | |
395 | if (HYPERVISOR_update_descriptor(maddr.maddr, desc)) | |
396 | BUG(); | |
397 | } | |
398 | ||
399 | } | |
400 | } | |
401 | ||
402 | static void xen_load_esp0(struct tss_struct *tss, | |
403 | struct thread_struct *thread) | |
404 | { | |
405 | struct multicall_space mcs = xen_mc_entry(0); | |
406 | MULTI_stack_switch(mcs.mc, __KERNEL_DS, thread->esp0); | |
407 | xen_mc_issue(PARAVIRT_LAZY_CPU); | |
408 | } | |
409 | ||
410 | static void xen_set_iopl_mask(unsigned mask) | |
411 | { | |
412 | struct physdev_set_iopl set_iopl; | |
413 | ||
414 | /* Force the change at ring 0. */ | |
415 | set_iopl.iopl = (mask == 0) ? 1 : (mask >> 12) & 3; | |
416 | HYPERVISOR_physdev_op(PHYSDEVOP_set_iopl, &set_iopl); | |
417 | } | |
418 | ||
419 | static void xen_io_delay(void) | |
420 | { | |
421 | } | |
422 | ||
423 | #ifdef CONFIG_X86_LOCAL_APIC | |
424 | static unsigned long xen_apic_read(unsigned long reg) | |
425 | { | |
426 | return 0; | |
427 | } | |
428 | #endif | |
429 | ||
430 | static void xen_flush_tlb(void) | |
431 | { | |
432 | struct mmuext_op op; | |
433 | ||
434 | op.cmd = MMUEXT_TLB_FLUSH_LOCAL; | |
435 | if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF)) | |
436 | BUG(); | |
437 | } | |
438 | ||
439 | static void xen_flush_tlb_single(unsigned long addr) | |
440 | { | |
441 | struct mmuext_op op; | |
442 | ||
443 | op.cmd = MMUEXT_INVLPG_LOCAL; | |
444 | op.arg1.linear_addr = addr & PAGE_MASK; | |
445 | if (HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF)) | |
446 | BUG(); | |
447 | } | |
448 | ||
449 | static unsigned long xen_read_cr2(void) | |
450 | { | |
451 | return x86_read_percpu(xen_vcpu)->arch.cr2; | |
452 | } | |
453 | ||
454 | static void xen_write_cr4(unsigned long cr4) | |
455 | { | |
456 | /* never allow TSC to be disabled */ | |
457 | native_write_cr4(cr4 & ~X86_CR4_TSD); | |
458 | } | |
459 | ||
460 | /* | |
461 | * Page-directory addresses above 4GB do not fit into architectural %cr3. | |
462 | * When accessing %cr3, or equivalent field in vcpu_guest_context, guests | |
463 | * must use the following accessor macros to pack/unpack valid MFNs. | |
464 | * | |
465 | * Note that Xen is using the fact that the pagetable base is always | |
466 | * page-aligned, and putting the 12 MSB of the address into the 12 LSB | |
467 | * of cr3. | |
468 | */ | |
469 | #define xen_pfn_to_cr3(pfn) (((unsigned)(pfn) << 12) | ((unsigned)(pfn) >> 20)) | |
470 | #define xen_cr3_to_pfn(cr3) (((unsigned)(cr3) >> 12) | ((unsigned)(cr3) << 20)) | |
471 | ||
472 | static unsigned long xen_read_cr3(void) | |
473 | { | |
474 | return x86_read_percpu(xen_cr3); | |
475 | } | |
476 | ||
477 | static void xen_write_cr3(unsigned long cr3) | |
478 | { | |
479 | if (cr3 == x86_read_percpu(xen_cr3)) { | |
480 | /* just a simple tlb flush */ | |
481 | xen_flush_tlb(); | |
482 | return; | |
483 | } | |
484 | ||
485 | x86_write_percpu(xen_cr3, cr3); | |
486 | ||
487 | ||
488 | { | |
489 | struct mmuext_op *op; | |
490 | struct multicall_space mcs = xen_mc_entry(sizeof(*op)); | |
491 | unsigned long mfn = pfn_to_mfn(PFN_DOWN(cr3)); | |
492 | ||
493 | op = mcs.args; | |
494 | op->cmd = MMUEXT_NEW_BASEPTR; | |
495 | op->arg1.mfn = mfn; | |
496 | ||
497 | MULTI_mmuext_op(mcs.mc, op, 1, NULL, DOMID_SELF); | |
498 | ||
499 | xen_mc_issue(PARAVIRT_LAZY_CPU); | |
500 | } | |
501 | } | |
502 | ||
503 | static void xen_alloc_pt(struct mm_struct *mm, u32 pfn) | |
504 | { | |
505 | /* XXX pfn isn't necessarily a lowmem page */ | |
506 | make_lowmem_page_readonly(__va(PFN_PHYS(pfn))); | |
507 | } | |
508 | ||
509 | static void xen_alloc_pd(u32 pfn) | |
510 | { | |
511 | make_lowmem_page_readonly(__va(PFN_PHYS(pfn))); | |
512 | } | |
513 | ||
514 | static void xen_release_pd(u32 pfn) | |
515 | { | |
516 | make_lowmem_page_readwrite(__va(PFN_PHYS(pfn))); | |
517 | } | |
518 | ||
519 | static void xen_release_pt(u32 pfn) | |
520 | { | |
521 | make_lowmem_page_readwrite(__va(PFN_PHYS(pfn))); | |
522 | } | |
523 | ||
524 | static void xen_alloc_pd_clone(u32 pfn, u32 clonepfn, | |
525 | u32 start, u32 count) | |
526 | { | |
527 | xen_alloc_pd(pfn); | |
528 | } | |
529 | ||
530 | static __init void xen_pagetable_setup_start(pgd_t *base) | |
531 | { | |
532 | pgd_t *xen_pgd = (pgd_t *)xen_start_info->pt_base; | |
533 | ||
534 | init_mm.pgd = base; | |
535 | /* | |
536 | * copy top-level of Xen-supplied pagetable into place. For | |
537 | * !PAE we can use this as-is, but for PAE it is a stand-in | |
538 | * while we copy the pmd pages. | |
539 | */ | |
540 | memcpy(base, xen_pgd, PTRS_PER_PGD * sizeof(pgd_t)); | |
541 | ||
542 | if (PTRS_PER_PMD > 1) { | |
543 | int i; | |
544 | /* | |
545 | * For PAE, need to allocate new pmds, rather than | |
546 | * share Xen's, since Xen doesn't like pmd's being | |
547 | * shared between address spaces. | |
548 | */ | |
549 | for (i = 0; i < PTRS_PER_PGD; i++) { | |
550 | if (pgd_val_ma(xen_pgd[i]) & _PAGE_PRESENT) { | |
551 | pmd_t *pmd = (pmd_t *)alloc_bootmem_low_pages(PAGE_SIZE); | |
552 | ||
553 | memcpy(pmd, (void *)pgd_page_vaddr(xen_pgd[i]), | |
554 | PAGE_SIZE); | |
555 | ||
556 | xen_alloc_pd(PFN_DOWN(__pa(pmd))); | |
557 | ||
558 | set_pgd(&base[i], __pgd(1 + __pa(pmd))); | |
559 | } else | |
560 | pgd_clear(&base[i]); | |
561 | } | |
562 | } | |
563 | ||
564 | /* make sure zero_page is mapped RO so we can use it in pagetables */ | |
565 | make_lowmem_page_readonly(empty_zero_page); | |
566 | make_lowmem_page_readonly(base); | |
567 | /* | |
568 | * Switch to new pagetable. This is done before | |
569 | * pagetable_init has done anything so that the new pages | |
570 | * added to the table can be prepared properly for Xen. | |
571 | */ | |
572 | xen_write_cr3(__pa(base)); | |
573 | } | |
574 | ||
575 | static __init void xen_pagetable_setup_done(pgd_t *base) | |
576 | { | |
577 | if (!xen_feature(XENFEAT_auto_translated_physmap)) { | |
578 | /* | |
579 | * Create a mapping for the shared info page. | |
580 | * Should be set_fixmap(), but shared_info is a machine | |
581 | * address with no corresponding pseudo-phys address. | |
582 | */ | |
5ead97c8 JF |
583 | set_pte_mfn(fix_to_virt(FIX_PARAVIRT_BOOTMAP), |
584 | PFN_DOWN(xen_start_info->shared_info), | |
585 | PAGE_KERNEL); | |
5ead97c8 JF |
586 | |
587 | HYPERVISOR_shared_info = | |
588 | (struct shared_info *)fix_to_virt(FIX_PARAVIRT_BOOTMAP); | |
589 | ||
590 | } else | |
591 | HYPERVISOR_shared_info = | |
592 | (struct shared_info *)__va(xen_start_info->shared_info); | |
593 | ||
5ead97c8 | 594 | xen_pgd_pin(base); |
5ead97c8 JF |
595 | |
596 | xen_vcpu_setup(smp_processor_id()); | |
597 | } | |
598 | ||
599 | static const struct paravirt_ops xen_paravirt_ops __initdata = { | |
600 | .paravirt_enabled = 1, | |
601 | .shared_kernel_pmd = 0, | |
602 | ||
603 | .name = "Xen", | |
604 | .banner = xen_banner, | |
605 | ||
606 | .patch = paravirt_patch_default, | |
607 | ||
608 | .memory_setup = xen_memory_setup, | |
609 | .arch_setup = xen_arch_setup, | |
e46cdb66 | 610 | .init_IRQ = xen_init_IRQ, |
5ead97c8 | 611 | |
15c84731 JF |
612 | .time_init = xen_time_init, |
613 | .set_wallclock = xen_set_wallclock, | |
614 | .get_wallclock = xen_get_wallclock, | |
615 | .get_cpu_khz = xen_cpu_khz, | |
616 | .sched_clock = xen_clocksource_read, | |
617 | ||
5ead97c8 JF |
618 | .cpuid = xen_cpuid, |
619 | ||
620 | .set_debugreg = xen_set_debugreg, | |
621 | .get_debugreg = xen_get_debugreg, | |
622 | ||
623 | .clts = native_clts, | |
624 | ||
625 | .read_cr0 = native_read_cr0, | |
626 | .write_cr0 = native_write_cr0, | |
627 | ||
628 | .read_cr2 = xen_read_cr2, | |
629 | .write_cr2 = native_write_cr2, | |
630 | ||
631 | .read_cr3 = xen_read_cr3, | |
632 | .write_cr3 = xen_write_cr3, | |
633 | ||
634 | .read_cr4 = native_read_cr4, | |
635 | .read_cr4_safe = native_read_cr4_safe, | |
636 | .write_cr4 = xen_write_cr4, | |
637 | ||
638 | .save_fl = xen_save_fl, | |
639 | .restore_fl = xen_restore_fl, | |
640 | .irq_disable = xen_irq_disable, | |
641 | .irq_enable = xen_irq_enable, | |
642 | .safe_halt = xen_safe_halt, | |
643 | .halt = xen_halt, | |
644 | .wbinvd = native_wbinvd, | |
645 | ||
646 | .read_msr = native_read_msr_safe, | |
647 | .write_msr = native_write_msr_safe, | |
648 | .read_tsc = native_read_tsc, | |
649 | .read_pmc = native_read_pmc, | |
650 | ||
651 | .iret = (void *)&hypercall_page[__HYPERVISOR_iret], | |
652 | .irq_enable_sysexit = NULL, /* never called */ | |
653 | ||
654 | .load_tr_desc = paravirt_nop, | |
655 | .set_ldt = xen_set_ldt, | |
656 | .load_gdt = xen_load_gdt, | |
657 | .load_idt = xen_load_idt, | |
658 | .load_tls = xen_load_tls, | |
659 | ||
660 | .store_gdt = native_store_gdt, | |
661 | .store_idt = native_store_idt, | |
662 | .store_tr = xen_store_tr, | |
663 | ||
664 | .write_ldt_entry = xen_write_ldt_entry, | |
665 | .write_gdt_entry = xen_write_gdt_entry, | |
666 | .write_idt_entry = xen_write_idt_entry, | |
667 | .load_esp0 = xen_load_esp0, | |
668 | ||
669 | .set_iopl_mask = xen_set_iopl_mask, | |
670 | .io_delay = xen_io_delay, | |
671 | ||
672 | #ifdef CONFIG_X86_LOCAL_APIC | |
673 | .apic_write = paravirt_nop, | |
674 | .apic_write_atomic = paravirt_nop, | |
675 | .apic_read = xen_apic_read, | |
676 | .setup_boot_clock = paravirt_nop, | |
677 | .setup_secondary_clock = paravirt_nop, | |
678 | .startup_ipi_hook = paravirt_nop, | |
679 | #endif | |
680 | ||
681 | .flush_tlb_user = xen_flush_tlb, | |
682 | .flush_tlb_kernel = xen_flush_tlb, | |
683 | .flush_tlb_single = xen_flush_tlb_single, | |
684 | ||
685 | .pte_update = paravirt_nop, | |
686 | .pte_update_defer = paravirt_nop, | |
687 | ||
688 | .pagetable_setup_start = xen_pagetable_setup_start, | |
689 | .pagetable_setup_done = xen_pagetable_setup_done, | |
690 | ||
691 | .alloc_pt = xen_alloc_pt, | |
692 | .alloc_pd = xen_alloc_pd, | |
693 | .alloc_pd_clone = xen_alloc_pd_clone, | |
694 | .release_pd = xen_release_pd, | |
695 | .release_pt = xen_release_pt, | |
696 | ||
3b827c1b JF |
697 | .set_pte = xen_set_pte, |
698 | .set_pte_at = xen_set_pte_at, | |
699 | .set_pmd = xen_set_pmd, | |
700 | ||
701 | .pte_val = xen_pte_val, | |
702 | .pgd_val = xen_pgd_val, | |
703 | ||
704 | .make_pte = xen_make_pte, | |
705 | .make_pgd = xen_make_pgd, | |
706 | ||
707 | #ifdef CONFIG_X86_PAE | |
708 | .set_pte_atomic = xen_set_pte_atomic, | |
709 | .set_pte_present = xen_set_pte_at, | |
710 | .set_pud = xen_set_pud, | |
711 | .pte_clear = xen_pte_clear, | |
712 | .pmd_clear = xen_pmd_clear, | |
713 | ||
714 | .make_pmd = xen_make_pmd, | |
715 | .pmd_val = xen_pmd_val, | |
716 | #endif /* PAE */ | |
717 | ||
718 | .activate_mm = xen_activate_mm, | |
719 | .dup_mmap = xen_dup_mmap, | |
720 | .exit_mmap = xen_exit_mmap, | |
721 | ||
5ead97c8 JF |
722 | .set_lazy_mode = xen_set_lazy_mode, |
723 | }; | |
724 | ||
725 | /* First C function to be called on Xen boot */ | |
726 | asmlinkage void __init xen_start_kernel(void) | |
727 | { | |
728 | pgd_t *pgd; | |
729 | ||
730 | if (!xen_start_info) | |
731 | return; | |
732 | ||
733 | BUG_ON(memcmp(xen_start_info->magic, "xen-3.0", 7) != 0); | |
734 | ||
735 | /* Install Xen paravirt ops */ | |
736 | paravirt_ops = xen_paravirt_ops; | |
737 | ||
738 | xen_setup_features(); | |
739 | ||
740 | /* Get mfn list */ | |
741 | if (!xen_feature(XENFEAT_auto_translated_physmap)) | |
742 | phys_to_machine_mapping = (unsigned long *)xen_start_info->mfn_list; | |
743 | ||
744 | pgd = (pgd_t *)xen_start_info->pt_base; | |
745 | ||
746 | init_pg_tables_end = __pa(pgd) + xen_start_info->nr_pt_frames*PAGE_SIZE; | |
747 | ||
748 | init_mm.pgd = pgd; /* use the Xen pagetables to start */ | |
749 | ||
750 | /* keep using Xen gdt for now; no urgent need to change it */ | |
751 | ||
752 | x86_write_percpu(xen_cr3, __pa(pgd)); | |
753 | xen_vcpu_setup(0); | |
754 | ||
755 | paravirt_ops.kernel_rpl = 1; | |
756 | if (xen_feature(XENFEAT_supervisor_mode_kernel)) | |
757 | paravirt_ops.kernel_rpl = 0; | |
758 | ||
759 | /* set the limit of our address space */ | |
760 | reserve_top_address(-HYPERVISOR_VIRT_START + 2 * PAGE_SIZE); | |
761 | ||
762 | /* set up basic CPUID stuff */ | |
763 | cpu_detect(&new_cpu_data); | |
764 | new_cpu_data.hard_math = 1; | |
765 | new_cpu_data.x86_capability[0] = cpuid_edx(1); | |
766 | ||
767 | /* Poke various useful things into boot_params */ | |
768 | LOADER_TYPE = (9 << 4) | 0; | |
769 | INITRD_START = xen_start_info->mod_start ? __pa(xen_start_info->mod_start) : 0; | |
770 | INITRD_SIZE = xen_start_info->mod_len; | |
771 | ||
772 | /* Start the world */ | |
773 | start_kernel(); | |
774 | } |