x86: change write_ldt_entry signature
[deliverable/linux.git] / include / asm-x86 / paravirt.h
1 #ifndef __ASM_PARAVIRT_H
2 #define __ASM_PARAVIRT_H
3 /* Various instructions on x86 need to be replaced for
4 * para-virtualization: those hooks are defined here. */
5
6 #ifdef CONFIG_PARAVIRT
7 #include <asm/page.h>
8
9 /* Bitmask of what can be clobbered: usually at least eax. */
10 #define CLBR_NONE 0x0
11 #define CLBR_EAX 0x1
12 #define CLBR_ECX 0x2
13 #define CLBR_EDX 0x4
14 #define CLBR_ANY 0x7
15
16 #ifndef __ASSEMBLY__
17 #include <linux/types.h>
18 #include <linux/cpumask.h>
19 #include <asm/kmap_types.h>
20 #include <asm/desc_defs.h>
21
22 struct page;
23 struct thread_struct;
24 struct desc_ptr;
25 struct tss_struct;
26 struct mm_struct;
27 struct desc_struct;
28
29 /* general info */
30 struct pv_info {
31 unsigned int kernel_rpl;
32 int shared_kernel_pmd;
33 int paravirt_enabled;
34 const char *name;
35 };
36
37 struct pv_init_ops {
38 /*
39 * Patch may replace one of the defined code sequences with
40 * arbitrary code, subject to the same register constraints.
41 * This generally means the code is not free to clobber any
42 * registers other than EAX. The patch function should return
43 * the number of bytes of code generated, as we nop pad the
44 * rest in generic code.
45 */
46 unsigned (*patch)(u8 type, u16 clobber, void *insnbuf,
47 unsigned long addr, unsigned len);
48
49 /* Basic arch-specific setup */
50 void (*arch_setup)(void);
51 char *(*memory_setup)(void);
52 void (*post_allocator_init)(void);
53
54 /* Print a banner to identify the environment */
55 void (*banner)(void);
56 };
57
58
59 struct pv_lazy_ops {
60 /* Set deferred update mode, used for batching operations. */
61 void (*enter)(void);
62 void (*leave)(void);
63 };
64
65 struct pv_time_ops {
66 void (*time_init)(void);
67
68 /* Set and set time of day */
69 unsigned long (*get_wallclock)(void);
70 int (*set_wallclock)(unsigned long);
71
72 unsigned long long (*sched_clock)(void);
73 unsigned long (*get_cpu_khz)(void);
74 };
75
76 struct pv_cpu_ops {
77 /* hooks for various privileged instructions */
78 unsigned long (*get_debugreg)(int regno);
79 void (*set_debugreg)(int regno, unsigned long value);
80
81 void (*clts)(void);
82
83 unsigned long (*read_cr0)(void);
84 void (*write_cr0)(unsigned long);
85
86 unsigned long (*read_cr4_safe)(void);
87 unsigned long (*read_cr4)(void);
88 void (*write_cr4)(unsigned long);
89
90 /* Segment descriptor handling */
91 void (*load_tr_desc)(void);
92 void (*load_gdt)(const struct desc_ptr *);
93 void (*load_idt)(const struct desc_ptr *);
94 void (*store_gdt)(struct desc_ptr *);
95 void (*store_idt)(struct desc_ptr *);
96 void (*set_ldt)(const void *desc, unsigned entries);
97 unsigned long (*store_tr)(void);
98 void (*load_tls)(struct thread_struct *t, unsigned int cpu);
99 void (*write_ldt_entry)(struct desc_struct *ldt, int entrynum,
100 const void *desc);
101 void (*write_gdt_entry)(struct desc_struct *,
102 int entrynum, const void *desc, int size);
103 void (*write_idt_entry)(gate_desc *,
104 int entrynum, const gate_desc *gate);
105 void (*load_sp0)(struct tss_struct *tss, struct thread_struct *t);
106
107 void (*set_iopl_mask)(unsigned mask);
108
109 void (*wbinvd)(void);
110 void (*io_delay)(void);
111
112 /* cpuid emulation, mostly so that caps bits can be disabled */
113 void (*cpuid)(unsigned int *eax, unsigned int *ebx,
114 unsigned int *ecx, unsigned int *edx);
115
116 /* MSR, PMC and TSR operations.
117 err = 0/-EFAULT. wrmsr returns 0/-EFAULT. */
118 u64 (*read_msr)(unsigned int msr, int *err);
119 int (*write_msr)(unsigned int msr, unsigned low, unsigned high);
120
121 u64 (*read_tsc)(void);
122 u64 (*read_pmc)(int counter);
123
124 /* These two are jmp to, not actually called. */
125 void (*irq_enable_syscall_ret)(void);
126 void (*iret)(void);
127
128 struct pv_lazy_ops lazy_mode;
129 };
130
131 struct pv_irq_ops {
132 void (*init_IRQ)(void);
133
134 /*
135 * Get/set interrupt state. save_fl and restore_fl are only
136 * expected to use X86_EFLAGS_IF; all other bits
137 * returned from save_fl are undefined, and may be ignored by
138 * restore_fl.
139 */
140 unsigned long (*save_fl)(void);
141 void (*restore_fl)(unsigned long);
142 void (*irq_disable)(void);
143 void (*irq_enable)(void);
144 void (*safe_halt)(void);
145 void (*halt)(void);
146 };
147
148 struct pv_apic_ops {
149 #ifdef CONFIG_X86_LOCAL_APIC
150 /*
151 * Direct APIC operations, principally for VMI. Ideally
152 * these shouldn't be in this interface.
153 */
154 void (*apic_write)(unsigned long reg, u32 v);
155 void (*apic_write_atomic)(unsigned long reg, u32 v);
156 u32 (*apic_read)(unsigned long reg);
157 void (*setup_boot_clock)(void);
158 void (*setup_secondary_clock)(void);
159
160 void (*startup_ipi_hook)(int phys_apicid,
161 unsigned long start_eip,
162 unsigned long start_esp);
163 #endif
164 };
165
166 struct pv_mmu_ops {
167 /*
168 * Called before/after init_mm pagetable setup. setup_start
169 * may reset %cr3, and may pre-install parts of the pagetable;
170 * pagetable setup is expected to preserve any existing
171 * mapping.
172 */
173 void (*pagetable_setup_start)(pgd_t *pgd_base);
174 void (*pagetable_setup_done)(pgd_t *pgd_base);
175
176 unsigned long (*read_cr2)(void);
177 void (*write_cr2)(unsigned long);
178
179 unsigned long (*read_cr3)(void);
180 void (*write_cr3)(unsigned long);
181
182 /*
183 * Hooks for intercepting the creation/use/destruction of an
184 * mm_struct.
185 */
186 void (*activate_mm)(struct mm_struct *prev,
187 struct mm_struct *next);
188 void (*dup_mmap)(struct mm_struct *oldmm,
189 struct mm_struct *mm);
190 void (*exit_mmap)(struct mm_struct *mm);
191
192
193 /* TLB operations */
194 void (*flush_tlb_user)(void);
195 void (*flush_tlb_kernel)(void);
196 void (*flush_tlb_single)(unsigned long addr);
197 void (*flush_tlb_others)(const cpumask_t *cpus, struct mm_struct *mm,
198 unsigned long va);
199
200 /* Hooks for allocating/releasing pagetable pages */
201 void (*alloc_pt)(struct mm_struct *mm, u32 pfn);
202 void (*alloc_pd)(u32 pfn);
203 void (*alloc_pd_clone)(u32 pfn, u32 clonepfn, u32 start, u32 count);
204 void (*release_pt)(u32 pfn);
205 void (*release_pd)(u32 pfn);
206
207 /* Pagetable manipulation functions */
208 void (*set_pte)(pte_t *ptep, pte_t pteval);
209 void (*set_pte_at)(struct mm_struct *mm, unsigned long addr,
210 pte_t *ptep, pte_t pteval);
211 void (*set_pmd)(pmd_t *pmdp, pmd_t pmdval);
212 void (*pte_update)(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
213 void (*pte_update_defer)(struct mm_struct *mm,
214 unsigned long addr, pte_t *ptep);
215
216 #ifdef CONFIG_X86_PAE
217 void (*set_pte_atomic)(pte_t *ptep, pte_t pteval);
218 void (*set_pte_present)(struct mm_struct *mm, unsigned long addr,
219 pte_t *ptep, pte_t pte);
220 void (*set_pud)(pud_t *pudp, pud_t pudval);
221 void (*pte_clear)(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
222 void (*pmd_clear)(pmd_t *pmdp);
223
224 unsigned long long (*pte_val)(pte_t);
225 unsigned long long (*pmd_val)(pmd_t);
226 unsigned long long (*pgd_val)(pgd_t);
227
228 pte_t (*make_pte)(unsigned long long pte);
229 pmd_t (*make_pmd)(unsigned long long pmd);
230 pgd_t (*make_pgd)(unsigned long long pgd);
231 #else
232 unsigned long (*pte_val)(pte_t);
233 unsigned long (*pgd_val)(pgd_t);
234
235 pte_t (*make_pte)(unsigned long pte);
236 pgd_t (*make_pgd)(unsigned long pgd);
237 #endif
238
239 #ifdef CONFIG_HIGHPTE
240 void *(*kmap_atomic_pte)(struct page *page, enum km_type type);
241 #endif
242
243 struct pv_lazy_ops lazy_mode;
244 };
245
246 /* This contains all the paravirt structures: we get a convenient
247 * number for each function using the offset which we use to indicate
248 * what to patch. */
249 struct paravirt_patch_template
250 {
251 struct pv_init_ops pv_init_ops;
252 struct pv_time_ops pv_time_ops;
253 struct pv_cpu_ops pv_cpu_ops;
254 struct pv_irq_ops pv_irq_ops;
255 struct pv_apic_ops pv_apic_ops;
256 struct pv_mmu_ops pv_mmu_ops;
257 };
258
259 extern struct pv_info pv_info;
260 extern struct pv_init_ops pv_init_ops;
261 extern struct pv_time_ops pv_time_ops;
262 extern struct pv_cpu_ops pv_cpu_ops;
263 extern struct pv_irq_ops pv_irq_ops;
264 extern struct pv_apic_ops pv_apic_ops;
265 extern struct pv_mmu_ops pv_mmu_ops;
266
267 #define PARAVIRT_PATCH(x) \
268 (offsetof(struct paravirt_patch_template, x) / sizeof(void *))
269
270 #define paravirt_type(op) \
271 [paravirt_typenum] "i" (PARAVIRT_PATCH(op)), \
272 [paravirt_opptr] "m" (op)
273 #define paravirt_clobber(clobber) \
274 [paravirt_clobber] "i" (clobber)
275
276 /*
277 * Generate some code, and mark it as patchable by the
278 * apply_paravirt() alternate instruction patcher.
279 */
280 #define _paravirt_alt(insn_string, type, clobber) \
281 "771:\n\t" insn_string "\n" "772:\n" \
282 ".pushsection .parainstructions,\"a\"\n" \
283 " .long 771b\n" \
284 " .byte " type "\n" \
285 " .byte 772b-771b\n" \
286 " .short " clobber "\n" \
287 ".popsection\n"
288
289 /* Generate patchable code, with the default asm parameters. */
290 #define paravirt_alt(insn_string) \
291 _paravirt_alt(insn_string, "%c[paravirt_typenum]", "%c[paravirt_clobber]")
292
293 unsigned paravirt_patch_nop(void);
294 unsigned paravirt_patch_ignore(unsigned len);
295 unsigned paravirt_patch_call(void *insnbuf,
296 const void *target, u16 tgt_clobbers,
297 unsigned long addr, u16 site_clobbers,
298 unsigned len);
299 unsigned paravirt_patch_jmp(void *insnbuf, const void *target,
300 unsigned long addr, unsigned len);
301 unsigned paravirt_patch_default(u8 type, u16 clobbers, void *insnbuf,
302 unsigned long addr, unsigned len);
303
304 unsigned paravirt_patch_insns(void *insnbuf, unsigned len,
305 const char *start, const char *end);
306
307 int paravirt_disable_iospace(void);
308
309 /*
310 * This generates an indirect call based on the operation type number.
311 * The type number, computed in PARAVIRT_PATCH, is derived from the
312 * offset into the paravirt_patch_template structure, and can therefore be
313 * freely converted back into a structure offset.
314 */
315 #define PARAVIRT_CALL "call *%[paravirt_opptr];"
316
317 /*
318 * These macros are intended to wrap calls through one of the paravirt
319 * ops structs, so that they can be later identified and patched at
320 * runtime.
321 *
322 * Normally, a call to a pv_op function is a simple indirect call:
323 * (paravirt_ops.operations)(args...).
324 *
325 * Unfortunately, this is a relatively slow operation for modern CPUs,
326 * because it cannot necessarily determine what the destination
327 * address is. In this case, the address is a runtime constant, so at
328 * the very least we can patch the call to e a simple direct call, or
329 * ideally, patch an inline implementation into the callsite. (Direct
330 * calls are essentially free, because the call and return addresses
331 * are completely predictable.)
332 *
333 * These macros rely on the standard gcc "regparm(3)" calling
334 * convention, in which the first three arguments are placed in %eax,
335 * %edx, %ecx (in that order), and the remaining arguments are placed
336 * on the stack. All caller-save registers (eax,edx,ecx) are expected
337 * to be modified (either clobbered or used for return values).
338 *
339 * The call instruction itself is marked by placing its start address
340 * and size into the .parainstructions section, so that
341 * apply_paravirt() in arch/i386/kernel/alternative.c can do the
342 * appropriate patching under the control of the backend pv_init_ops
343 * implementation.
344 *
345 * Unfortunately there's no way to get gcc to generate the args setup
346 * for the call, and then allow the call itself to be generated by an
347 * inline asm. Because of this, we must do the complete arg setup and
348 * return value handling from within these macros. This is fairly
349 * cumbersome.
350 *
351 * There are 5 sets of PVOP_* macros for dealing with 0-4 arguments.
352 * It could be extended to more arguments, but there would be little
353 * to be gained from that. For each number of arguments, there are
354 * the two VCALL and CALL variants for void and non-void functions.
355 *
356 * When there is a return value, the invoker of the macro must specify
357 * the return type. The macro then uses sizeof() on that type to
358 * determine whether its a 32 or 64 bit value, and places the return
359 * in the right register(s) (just %eax for 32-bit, and %edx:%eax for
360 * 64-bit).
361 *
362 * 64-bit arguments are passed as a pair of adjacent 32-bit arguments
363 * in low,high order.
364 *
365 * Small structures are passed and returned in registers. The macro
366 * calling convention can't directly deal with this, so the wrapper
367 * functions must do this.
368 *
369 * These PVOP_* macros are only defined within this header. This
370 * means that all uses must be wrapped in inline functions. This also
371 * makes sure the incoming and outgoing types are always correct.
372 */
373 #define __PVOP_CALL(rettype, op, pre, post, ...) \
374 ({ \
375 rettype __ret; \
376 unsigned long __eax, __edx, __ecx; \
377 if (sizeof(rettype) > sizeof(unsigned long)) { \
378 asm volatile(pre \
379 paravirt_alt(PARAVIRT_CALL) \
380 post \
381 : "=a" (__eax), "=d" (__edx), \
382 "=c" (__ecx) \
383 : paravirt_type(op), \
384 paravirt_clobber(CLBR_ANY), \
385 ##__VA_ARGS__ \
386 : "memory", "cc"); \
387 __ret = (rettype)((((u64)__edx) << 32) | __eax); \
388 } else { \
389 asm volatile(pre \
390 paravirt_alt(PARAVIRT_CALL) \
391 post \
392 : "=a" (__eax), "=d" (__edx), \
393 "=c" (__ecx) \
394 : paravirt_type(op), \
395 paravirt_clobber(CLBR_ANY), \
396 ##__VA_ARGS__ \
397 : "memory", "cc"); \
398 __ret = (rettype)__eax; \
399 } \
400 __ret; \
401 })
402 #define __PVOP_VCALL(op, pre, post, ...) \
403 ({ \
404 unsigned long __eax, __edx, __ecx; \
405 asm volatile(pre \
406 paravirt_alt(PARAVIRT_CALL) \
407 post \
408 : "=a" (__eax), "=d" (__edx), "=c" (__ecx) \
409 : paravirt_type(op), \
410 paravirt_clobber(CLBR_ANY), \
411 ##__VA_ARGS__ \
412 : "memory", "cc"); \
413 })
414
415 #define PVOP_CALL0(rettype, op) \
416 __PVOP_CALL(rettype, op, "", "")
417 #define PVOP_VCALL0(op) \
418 __PVOP_VCALL(op, "", "")
419
420 #define PVOP_CALL1(rettype, op, arg1) \
421 __PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)))
422 #define PVOP_VCALL1(op, arg1) \
423 __PVOP_VCALL(op, "", "", "0" ((u32)(arg1)))
424
425 #define PVOP_CALL2(rettype, op, arg1, arg2) \
426 __PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)), "1" ((u32)(arg2)))
427 #define PVOP_VCALL2(op, arg1, arg2) \
428 __PVOP_VCALL(op, "", "", "0" ((u32)(arg1)), "1" ((u32)(arg2)))
429
430 #define PVOP_CALL3(rettype, op, arg1, arg2, arg3) \
431 __PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)), \
432 "1"((u32)(arg2)), "2"((u32)(arg3)))
433 #define PVOP_VCALL3(op, arg1, arg2, arg3) \
434 __PVOP_VCALL(op, "", "", "0" ((u32)(arg1)), "1"((u32)(arg2)), \
435 "2"((u32)(arg3)))
436
437 #define PVOP_CALL4(rettype, op, arg1, arg2, arg3, arg4) \
438 __PVOP_CALL(rettype, op, \
439 "push %[_arg4];", "lea 4(%%esp),%%esp;", \
440 "0" ((u32)(arg1)), "1" ((u32)(arg2)), \
441 "2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
442 #define PVOP_VCALL4(op, arg1, arg2, arg3, arg4) \
443 __PVOP_VCALL(op, \
444 "push %[_arg4];", "lea 4(%%esp),%%esp;", \
445 "0" ((u32)(arg1)), "1" ((u32)(arg2)), \
446 "2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
447
448 static inline int paravirt_enabled(void)
449 {
450 return pv_info.paravirt_enabled;
451 }
452
453 static inline void load_sp0(struct tss_struct *tss,
454 struct thread_struct *thread)
455 {
456 PVOP_VCALL2(pv_cpu_ops.load_sp0, tss, thread);
457 }
458
459 #define ARCH_SETUP pv_init_ops.arch_setup();
460 static inline unsigned long get_wallclock(void)
461 {
462 return PVOP_CALL0(unsigned long, pv_time_ops.get_wallclock);
463 }
464
465 static inline int set_wallclock(unsigned long nowtime)
466 {
467 return PVOP_CALL1(int, pv_time_ops.set_wallclock, nowtime);
468 }
469
470 static inline void (*choose_time_init(void))(void)
471 {
472 return pv_time_ops.time_init;
473 }
474
475 /* The paravirtualized CPUID instruction. */
476 static inline void __cpuid(unsigned int *eax, unsigned int *ebx,
477 unsigned int *ecx, unsigned int *edx)
478 {
479 PVOP_VCALL4(pv_cpu_ops.cpuid, eax, ebx, ecx, edx);
480 }
481
482 /*
483 * These special macros can be used to get or set a debugging register
484 */
485 static inline unsigned long paravirt_get_debugreg(int reg)
486 {
487 return PVOP_CALL1(unsigned long, pv_cpu_ops.get_debugreg, reg);
488 }
489 #define get_debugreg(var, reg) var = paravirt_get_debugreg(reg)
490 static inline void set_debugreg(unsigned long val, int reg)
491 {
492 PVOP_VCALL2(pv_cpu_ops.set_debugreg, reg, val);
493 }
494
495 static inline void clts(void)
496 {
497 PVOP_VCALL0(pv_cpu_ops.clts);
498 }
499
500 static inline unsigned long read_cr0(void)
501 {
502 return PVOP_CALL0(unsigned long, pv_cpu_ops.read_cr0);
503 }
504
505 static inline void write_cr0(unsigned long x)
506 {
507 PVOP_VCALL1(pv_cpu_ops.write_cr0, x);
508 }
509
510 static inline unsigned long read_cr2(void)
511 {
512 return PVOP_CALL0(unsigned long, pv_mmu_ops.read_cr2);
513 }
514
515 static inline void write_cr2(unsigned long x)
516 {
517 PVOP_VCALL1(pv_mmu_ops.write_cr2, x);
518 }
519
520 static inline unsigned long read_cr3(void)
521 {
522 return PVOP_CALL0(unsigned long, pv_mmu_ops.read_cr3);
523 }
524
525 static inline void write_cr3(unsigned long x)
526 {
527 PVOP_VCALL1(pv_mmu_ops.write_cr3, x);
528 }
529
530 static inline unsigned long read_cr4(void)
531 {
532 return PVOP_CALL0(unsigned long, pv_cpu_ops.read_cr4);
533 }
534 static inline unsigned long read_cr4_safe(void)
535 {
536 return PVOP_CALL0(unsigned long, pv_cpu_ops.read_cr4_safe);
537 }
538
539 static inline void write_cr4(unsigned long x)
540 {
541 PVOP_VCALL1(pv_cpu_ops.write_cr4, x);
542 }
543
544 static inline void raw_safe_halt(void)
545 {
546 PVOP_VCALL0(pv_irq_ops.safe_halt);
547 }
548
549 static inline void halt(void)
550 {
551 PVOP_VCALL0(pv_irq_ops.safe_halt);
552 }
553
554 static inline void wbinvd(void)
555 {
556 PVOP_VCALL0(pv_cpu_ops.wbinvd);
557 }
558
559 #define get_kernel_rpl() (pv_info.kernel_rpl)
560
561 static inline u64 paravirt_read_msr(unsigned msr, int *err)
562 {
563 return PVOP_CALL2(u64, pv_cpu_ops.read_msr, msr, err);
564 }
565 static inline int paravirt_write_msr(unsigned msr, unsigned low, unsigned high)
566 {
567 return PVOP_CALL3(int, pv_cpu_ops.write_msr, msr, low, high);
568 }
569
570 /* These should all do BUG_ON(_err), but our headers are too tangled. */
571 #define rdmsr(msr,val1,val2) do { \
572 int _err; \
573 u64 _l = paravirt_read_msr(msr, &_err); \
574 val1 = (u32)_l; \
575 val2 = _l >> 32; \
576 } while(0)
577
578 #define wrmsr(msr,val1,val2) do { \
579 paravirt_write_msr(msr, val1, val2); \
580 } while(0)
581
582 #define rdmsrl(msr,val) do { \
583 int _err; \
584 val = paravirt_read_msr(msr, &_err); \
585 } while(0)
586
587 #define wrmsrl(msr,val) wrmsr(msr, (u32)((u64)(val)), ((u64)(val))>>32)
588 #define wrmsr_safe(msr,a,b) paravirt_write_msr(msr, a, b)
589
590 /* rdmsr with exception handling */
591 #define rdmsr_safe(msr,a,b) ({ \
592 int _err; \
593 u64 _l = paravirt_read_msr(msr, &_err); \
594 (*a) = (u32)_l; \
595 (*b) = _l >> 32; \
596 _err; })
597
598
599 static inline u64 paravirt_read_tsc(void)
600 {
601 return PVOP_CALL0(u64, pv_cpu_ops.read_tsc);
602 }
603
604 #define rdtscl(low) do { \
605 u64 _l = paravirt_read_tsc(); \
606 low = (int)_l; \
607 } while(0)
608
609 #define rdtscll(val) (val = paravirt_read_tsc())
610
611 static inline unsigned long long paravirt_sched_clock(void)
612 {
613 return PVOP_CALL0(unsigned long long, pv_time_ops.sched_clock);
614 }
615 #define calculate_cpu_khz() (pv_time_ops.get_cpu_khz())
616
617 #define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
618
619 static inline unsigned long long paravirt_read_pmc(int counter)
620 {
621 return PVOP_CALL1(u64, pv_cpu_ops.read_pmc, counter);
622 }
623
624 #define rdpmc(counter,low,high) do { \
625 u64 _l = paravirt_read_pmc(counter); \
626 low = (u32)_l; \
627 high = _l >> 32; \
628 } while(0)
629
630 static inline void load_TR_desc(void)
631 {
632 PVOP_VCALL0(pv_cpu_ops.load_tr_desc);
633 }
634 static inline void load_gdt(const struct desc_ptr *dtr)
635 {
636 PVOP_VCALL1(pv_cpu_ops.load_gdt, dtr);
637 }
638 static inline void load_idt(const struct desc_ptr *dtr)
639 {
640 PVOP_VCALL1(pv_cpu_ops.load_idt, dtr);
641 }
642 static inline void set_ldt(const void *addr, unsigned entries)
643 {
644 PVOP_VCALL2(pv_cpu_ops.set_ldt, addr, entries);
645 }
646 static inline void store_gdt(struct desc_ptr *dtr)
647 {
648 PVOP_VCALL1(pv_cpu_ops.store_gdt, dtr);
649 }
650 static inline void store_idt(struct desc_ptr *dtr)
651 {
652 PVOP_VCALL1(pv_cpu_ops.store_idt, dtr);
653 }
654 static inline unsigned long paravirt_store_tr(void)
655 {
656 return PVOP_CALL0(unsigned long, pv_cpu_ops.store_tr);
657 }
658 #define store_tr(tr) ((tr) = paravirt_store_tr())
659 static inline void load_TLS(struct thread_struct *t, unsigned cpu)
660 {
661 PVOP_VCALL2(pv_cpu_ops.load_tls, t, cpu);
662 }
663
664 static inline void write_ldt_entry(struct desc_struct *dt, int entry,
665 const void *desc)
666 {
667 PVOP_VCALL3(pv_cpu_ops.write_ldt_entry, dt, entry, desc);
668 }
669
670 static inline void write_gdt_entry(struct desc_struct *dt, int entry,
671 void *desc, int type)
672 {
673 PVOP_VCALL4(pv_cpu_ops.write_gdt_entry, dt, entry, desc, type);
674 }
675
676 static inline void write_idt_entry(gate_desc *dt, int entry, const gate_desc *g)
677 {
678 PVOP_VCALL3(pv_cpu_ops.write_idt_entry, dt, entry, g);
679 }
680 static inline void set_iopl_mask(unsigned mask)
681 {
682 PVOP_VCALL1(pv_cpu_ops.set_iopl_mask, mask);
683 }
684
685 /* The paravirtualized I/O functions */
686 static inline void slow_down_io(void) {
687 pv_cpu_ops.io_delay();
688 #ifdef REALLY_SLOW_IO
689 pv_cpu_ops.io_delay();
690 pv_cpu_ops.io_delay();
691 pv_cpu_ops.io_delay();
692 #endif
693 }
694
695 #ifdef CONFIG_X86_LOCAL_APIC
696 /*
697 * Basic functions accessing APICs.
698 */
699 static inline void apic_write(unsigned long reg, u32 v)
700 {
701 PVOP_VCALL2(pv_apic_ops.apic_write, reg, v);
702 }
703
704 static inline void apic_write_atomic(unsigned long reg, u32 v)
705 {
706 PVOP_VCALL2(pv_apic_ops.apic_write_atomic, reg, v);
707 }
708
709 static inline u32 apic_read(unsigned long reg)
710 {
711 return PVOP_CALL1(unsigned long, pv_apic_ops.apic_read, reg);
712 }
713
714 static inline void setup_boot_clock(void)
715 {
716 PVOP_VCALL0(pv_apic_ops.setup_boot_clock);
717 }
718
719 static inline void setup_secondary_clock(void)
720 {
721 PVOP_VCALL0(pv_apic_ops.setup_secondary_clock);
722 }
723 #endif
724
725 static inline void paravirt_post_allocator_init(void)
726 {
727 if (pv_init_ops.post_allocator_init)
728 (*pv_init_ops.post_allocator_init)();
729 }
730
731 static inline void paravirt_pagetable_setup_start(pgd_t *base)
732 {
733 (*pv_mmu_ops.pagetable_setup_start)(base);
734 }
735
736 static inline void paravirt_pagetable_setup_done(pgd_t *base)
737 {
738 (*pv_mmu_ops.pagetable_setup_done)(base);
739 }
740
741 #ifdef CONFIG_SMP
742 static inline void startup_ipi_hook(int phys_apicid, unsigned long start_eip,
743 unsigned long start_esp)
744 {
745 PVOP_VCALL3(pv_apic_ops.startup_ipi_hook,
746 phys_apicid, start_eip, start_esp);
747 }
748 #endif
749
750 static inline void paravirt_activate_mm(struct mm_struct *prev,
751 struct mm_struct *next)
752 {
753 PVOP_VCALL2(pv_mmu_ops.activate_mm, prev, next);
754 }
755
756 static inline void arch_dup_mmap(struct mm_struct *oldmm,
757 struct mm_struct *mm)
758 {
759 PVOP_VCALL2(pv_mmu_ops.dup_mmap, oldmm, mm);
760 }
761
762 static inline void arch_exit_mmap(struct mm_struct *mm)
763 {
764 PVOP_VCALL1(pv_mmu_ops.exit_mmap, mm);
765 }
766
767 static inline void __flush_tlb(void)
768 {
769 PVOP_VCALL0(pv_mmu_ops.flush_tlb_user);
770 }
771 static inline void __flush_tlb_global(void)
772 {
773 PVOP_VCALL0(pv_mmu_ops.flush_tlb_kernel);
774 }
775 static inline void __flush_tlb_single(unsigned long addr)
776 {
777 PVOP_VCALL1(pv_mmu_ops.flush_tlb_single, addr);
778 }
779
780 static inline void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm,
781 unsigned long va)
782 {
783 PVOP_VCALL3(pv_mmu_ops.flush_tlb_others, &cpumask, mm, va);
784 }
785
786 static inline void paravirt_alloc_pt(struct mm_struct *mm, unsigned pfn)
787 {
788 PVOP_VCALL2(pv_mmu_ops.alloc_pt, mm, pfn);
789 }
790 static inline void paravirt_release_pt(unsigned pfn)
791 {
792 PVOP_VCALL1(pv_mmu_ops.release_pt, pfn);
793 }
794
795 static inline void paravirt_alloc_pd(unsigned pfn)
796 {
797 PVOP_VCALL1(pv_mmu_ops.alloc_pd, pfn);
798 }
799
800 static inline void paravirt_alloc_pd_clone(unsigned pfn, unsigned clonepfn,
801 unsigned start, unsigned count)
802 {
803 PVOP_VCALL4(pv_mmu_ops.alloc_pd_clone, pfn, clonepfn, start, count);
804 }
805 static inline void paravirt_release_pd(unsigned pfn)
806 {
807 PVOP_VCALL1(pv_mmu_ops.release_pd, pfn);
808 }
809
810 #ifdef CONFIG_HIGHPTE
811 static inline void *kmap_atomic_pte(struct page *page, enum km_type type)
812 {
813 unsigned long ret;
814 ret = PVOP_CALL2(unsigned long, pv_mmu_ops.kmap_atomic_pte, page, type);
815 return (void *)ret;
816 }
817 #endif
818
819 static inline void pte_update(struct mm_struct *mm, unsigned long addr,
820 pte_t *ptep)
821 {
822 PVOP_VCALL3(pv_mmu_ops.pte_update, mm, addr, ptep);
823 }
824
825 static inline void pte_update_defer(struct mm_struct *mm, unsigned long addr,
826 pte_t *ptep)
827 {
828 PVOP_VCALL3(pv_mmu_ops.pte_update_defer, mm, addr, ptep);
829 }
830
831 #ifdef CONFIG_X86_PAE
832 static inline pte_t __pte(unsigned long long val)
833 {
834 unsigned long long ret = PVOP_CALL2(unsigned long long,
835 pv_mmu_ops.make_pte,
836 val, val >> 32);
837 return (pte_t) { ret, ret >> 32 };
838 }
839
840 static inline pmd_t __pmd(unsigned long long val)
841 {
842 return (pmd_t) { PVOP_CALL2(unsigned long long, pv_mmu_ops.make_pmd,
843 val, val >> 32) };
844 }
845
846 static inline pgd_t __pgd(unsigned long long val)
847 {
848 return (pgd_t) { PVOP_CALL2(unsigned long long, pv_mmu_ops.make_pgd,
849 val, val >> 32) };
850 }
851
852 static inline unsigned long long pte_val(pte_t x)
853 {
854 return PVOP_CALL2(unsigned long long, pv_mmu_ops.pte_val,
855 x.pte_low, x.pte_high);
856 }
857
858 static inline unsigned long long pmd_val(pmd_t x)
859 {
860 return PVOP_CALL2(unsigned long long, pv_mmu_ops.pmd_val,
861 x.pmd, x.pmd >> 32);
862 }
863
864 static inline unsigned long long pgd_val(pgd_t x)
865 {
866 return PVOP_CALL2(unsigned long long, pv_mmu_ops.pgd_val,
867 x.pgd, x.pgd >> 32);
868 }
869
870 static inline void set_pte(pte_t *ptep, pte_t pteval)
871 {
872 PVOP_VCALL3(pv_mmu_ops.set_pte, ptep, pteval.pte_low, pteval.pte_high);
873 }
874
875 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
876 pte_t *ptep, pte_t pteval)
877 {
878 /* 5 arg words */
879 pv_mmu_ops.set_pte_at(mm, addr, ptep, pteval);
880 }
881
882 static inline void set_pte_atomic(pte_t *ptep, pte_t pteval)
883 {
884 PVOP_VCALL3(pv_mmu_ops.set_pte_atomic, ptep,
885 pteval.pte_low, pteval.pte_high);
886 }
887
888 static inline void set_pte_present(struct mm_struct *mm, unsigned long addr,
889 pte_t *ptep, pte_t pte)
890 {
891 /* 5 arg words */
892 pv_mmu_ops.set_pte_present(mm, addr, ptep, pte);
893 }
894
895 static inline void set_pmd(pmd_t *pmdp, pmd_t pmdval)
896 {
897 PVOP_VCALL3(pv_mmu_ops.set_pmd, pmdp,
898 pmdval.pmd, pmdval.pmd >> 32);
899 }
900
901 static inline void set_pud(pud_t *pudp, pud_t pudval)
902 {
903 PVOP_VCALL3(pv_mmu_ops.set_pud, pudp,
904 pudval.pgd.pgd, pudval.pgd.pgd >> 32);
905 }
906
907 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
908 {
909 PVOP_VCALL3(pv_mmu_ops.pte_clear, mm, addr, ptep);
910 }
911
912 static inline void pmd_clear(pmd_t *pmdp)
913 {
914 PVOP_VCALL1(pv_mmu_ops.pmd_clear, pmdp);
915 }
916
917 #else /* !CONFIG_X86_PAE */
918
919 static inline pte_t __pte(unsigned long val)
920 {
921 return (pte_t) { PVOP_CALL1(unsigned long, pv_mmu_ops.make_pte, val) };
922 }
923
924 static inline pgd_t __pgd(unsigned long val)
925 {
926 return (pgd_t) { PVOP_CALL1(unsigned long, pv_mmu_ops.make_pgd, val) };
927 }
928
929 static inline unsigned long pte_val(pte_t x)
930 {
931 return PVOP_CALL1(unsigned long, pv_mmu_ops.pte_val, x.pte_low);
932 }
933
934 static inline unsigned long pgd_val(pgd_t x)
935 {
936 return PVOP_CALL1(unsigned long, pv_mmu_ops.pgd_val, x.pgd);
937 }
938
939 static inline void set_pte(pte_t *ptep, pte_t pteval)
940 {
941 PVOP_VCALL2(pv_mmu_ops.set_pte, ptep, pteval.pte_low);
942 }
943
944 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
945 pte_t *ptep, pte_t pteval)
946 {
947 PVOP_VCALL4(pv_mmu_ops.set_pte_at, mm, addr, ptep, pteval.pte_low);
948 }
949
950 static inline void set_pmd(pmd_t *pmdp, pmd_t pmdval)
951 {
952 PVOP_VCALL2(pv_mmu_ops.set_pmd, pmdp, pmdval.pud.pgd.pgd);
953 }
954 #endif /* CONFIG_X86_PAE */
955
956 /* Lazy mode for batching updates / context switch */
957 enum paravirt_lazy_mode {
958 PARAVIRT_LAZY_NONE,
959 PARAVIRT_LAZY_MMU,
960 PARAVIRT_LAZY_CPU,
961 };
962
963 enum paravirt_lazy_mode paravirt_get_lazy_mode(void);
964 void paravirt_enter_lazy_cpu(void);
965 void paravirt_leave_lazy_cpu(void);
966 void paravirt_enter_lazy_mmu(void);
967 void paravirt_leave_lazy_mmu(void);
968 void paravirt_leave_lazy(enum paravirt_lazy_mode mode);
969
970 #define __HAVE_ARCH_ENTER_LAZY_CPU_MODE
971 static inline void arch_enter_lazy_cpu_mode(void)
972 {
973 PVOP_VCALL0(pv_cpu_ops.lazy_mode.enter);
974 }
975
976 static inline void arch_leave_lazy_cpu_mode(void)
977 {
978 PVOP_VCALL0(pv_cpu_ops.lazy_mode.leave);
979 }
980
981 static inline void arch_flush_lazy_cpu_mode(void)
982 {
983 if (unlikely(paravirt_get_lazy_mode() == PARAVIRT_LAZY_CPU)) {
984 arch_leave_lazy_cpu_mode();
985 arch_enter_lazy_cpu_mode();
986 }
987 }
988
989
990 #define __HAVE_ARCH_ENTER_LAZY_MMU_MODE
991 static inline void arch_enter_lazy_mmu_mode(void)
992 {
993 PVOP_VCALL0(pv_mmu_ops.lazy_mode.enter);
994 }
995
996 static inline void arch_leave_lazy_mmu_mode(void)
997 {
998 PVOP_VCALL0(pv_mmu_ops.lazy_mode.leave);
999 }
1000
1001 static inline void arch_flush_lazy_mmu_mode(void)
1002 {
1003 if (unlikely(paravirt_get_lazy_mode() == PARAVIRT_LAZY_MMU)) {
1004 arch_leave_lazy_mmu_mode();
1005 arch_enter_lazy_mmu_mode();
1006 }
1007 }
1008
1009 void _paravirt_nop(void);
1010 #define paravirt_nop ((void *)_paravirt_nop)
1011
1012 /* These all sit in the .parainstructions section to tell us what to patch. */
1013 struct paravirt_patch_site {
1014 u8 *instr; /* original instructions */
1015 u8 instrtype; /* type of this instruction */
1016 u8 len; /* length of original instruction */
1017 u16 clobbers; /* what registers you may clobber */
1018 };
1019
1020 extern struct paravirt_patch_site __parainstructions[],
1021 __parainstructions_end[];
1022
1023 static inline unsigned long __raw_local_save_flags(void)
1024 {
1025 unsigned long f;
1026
1027 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
1028 PARAVIRT_CALL
1029 "popl %%edx; popl %%ecx")
1030 : "=a"(f)
1031 : paravirt_type(pv_irq_ops.save_fl),
1032 paravirt_clobber(CLBR_EAX)
1033 : "memory", "cc");
1034 return f;
1035 }
1036
1037 static inline void raw_local_irq_restore(unsigned long f)
1038 {
1039 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
1040 PARAVIRT_CALL
1041 "popl %%edx; popl %%ecx")
1042 : "=a"(f)
1043 : "0"(f),
1044 paravirt_type(pv_irq_ops.restore_fl),
1045 paravirt_clobber(CLBR_EAX)
1046 : "memory", "cc");
1047 }
1048
1049 static inline void raw_local_irq_disable(void)
1050 {
1051 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
1052 PARAVIRT_CALL
1053 "popl %%edx; popl %%ecx")
1054 :
1055 : paravirt_type(pv_irq_ops.irq_disable),
1056 paravirt_clobber(CLBR_EAX)
1057 : "memory", "eax", "cc");
1058 }
1059
1060 static inline void raw_local_irq_enable(void)
1061 {
1062 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
1063 PARAVIRT_CALL
1064 "popl %%edx; popl %%ecx")
1065 :
1066 : paravirt_type(pv_irq_ops.irq_enable),
1067 paravirt_clobber(CLBR_EAX)
1068 : "memory", "eax", "cc");
1069 }
1070
1071 static inline unsigned long __raw_local_irq_save(void)
1072 {
1073 unsigned long f;
1074
1075 f = __raw_local_save_flags();
1076 raw_local_irq_disable();
1077 return f;
1078 }
1079
1080 #define CLI_STRING \
1081 _paravirt_alt("pushl %%ecx; pushl %%edx;" \
1082 "call *%[paravirt_cli_opptr];" \
1083 "popl %%edx; popl %%ecx", \
1084 "%c[paravirt_cli_type]", "%c[paravirt_clobber]")
1085
1086 #define STI_STRING \
1087 _paravirt_alt("pushl %%ecx; pushl %%edx;" \
1088 "call *%[paravirt_sti_opptr];" \
1089 "popl %%edx; popl %%ecx", \
1090 "%c[paravirt_sti_type]", "%c[paravirt_clobber]")
1091
1092 #define CLI_STI_CLOBBERS , "%eax"
1093 #define CLI_STI_INPUT_ARGS \
1094 , \
1095 [paravirt_cli_type] "i" (PARAVIRT_PATCH(pv_irq_ops.irq_disable)), \
1096 [paravirt_cli_opptr] "m" (pv_irq_ops.irq_disable), \
1097 [paravirt_sti_type] "i" (PARAVIRT_PATCH(pv_irq_ops.irq_enable)), \
1098 [paravirt_sti_opptr] "m" (pv_irq_ops.irq_enable), \
1099 paravirt_clobber(CLBR_EAX)
1100
1101 /* Make sure as little as possible of this mess escapes. */
1102 #undef PARAVIRT_CALL
1103 #undef __PVOP_CALL
1104 #undef __PVOP_VCALL
1105 #undef PVOP_VCALL0
1106 #undef PVOP_CALL0
1107 #undef PVOP_VCALL1
1108 #undef PVOP_CALL1
1109 #undef PVOP_VCALL2
1110 #undef PVOP_CALL2
1111 #undef PVOP_VCALL3
1112 #undef PVOP_CALL3
1113 #undef PVOP_VCALL4
1114 #undef PVOP_CALL4
1115
1116 #else /* __ASSEMBLY__ */
1117
1118 #define PARA_PATCH(struct, off) ((PARAVIRT_PATCH_##struct + (off)) / 4)
1119
1120 #define PARA_SITE(ptype, clobbers, ops) \
1121 771:; \
1122 ops; \
1123 772:; \
1124 .pushsection .parainstructions,"a"; \
1125 .long 771b; \
1126 .byte ptype; \
1127 .byte 772b-771b; \
1128 .short clobbers; \
1129 .popsection
1130
1131 #define INTERRUPT_RETURN \
1132 PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_iret), CLBR_NONE, \
1133 jmp *%cs:pv_cpu_ops+PV_CPU_iret)
1134
1135 #define DISABLE_INTERRUPTS(clobbers) \
1136 PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_disable), clobbers, \
1137 pushl %eax; pushl %ecx; pushl %edx; \
1138 call *%cs:pv_irq_ops+PV_IRQ_irq_disable; \
1139 popl %edx; popl %ecx; popl %eax) \
1140
1141 #define ENABLE_INTERRUPTS(clobbers) \
1142 PARA_SITE(PARA_PATCH(pv_irq_ops, PV_IRQ_irq_enable), clobbers, \
1143 pushl %eax; pushl %ecx; pushl %edx; \
1144 call *%cs:pv_irq_ops+PV_IRQ_irq_enable; \
1145 popl %edx; popl %ecx; popl %eax)
1146
1147 #define ENABLE_INTERRUPTS_SYSCALL_RET \
1148 PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_irq_enable_syscall_ret),\
1149 CLBR_NONE, \
1150 jmp *%cs:pv_cpu_ops+PV_CPU_irq_enable_syscall_ret)
1151
1152 #define GET_CR0_INTO_EAX \
1153 push %ecx; push %edx; \
1154 call *pv_cpu_ops+PV_CPU_read_cr0; \
1155 pop %edx; pop %ecx
1156
1157 #endif /* __ASSEMBLY__ */
1158 #endif /* CONFIG_PARAVIRT */
1159 #endif /* __ASM_PARAVIRT_H */
This page took 0.065681 seconds and 6 git commands to generate.