Merge branch 'server-cluster-locking-api' of git://linux-nfs.org/~bfields/linux
[deliverable/linux.git] / include / asm-i386 / 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
21 struct page;
22 struct thread_struct;
23 struct Xgt_desc_struct;
24 struct tss_struct;
25 struct mm_struct;
26 struct desc_struct;
27
28 /* Lazy mode for batching updates / context switch */
29 enum paravirt_lazy_mode {
30 PARAVIRT_LAZY_NONE = 0,
31 PARAVIRT_LAZY_MMU = 1,
32 PARAVIRT_LAZY_CPU = 2,
33 PARAVIRT_LAZY_FLUSH = 3,
34 };
35
36 struct paravirt_ops
37 {
38 unsigned int kernel_rpl;
39 int shared_kernel_pmd;
40 int paravirt_enabled;
41 const char *name;
42
43 /*
44 * Patch may replace one of the defined code sequences with arbitrary
45 * code, subject to the same register constraints. This generally
46 * means the code is not free to clobber any registers other than EAX.
47 * The patch function should return the number of bytes of code
48 * generated, as we nop pad the rest in generic code.
49 */
50 unsigned (*patch)(u8 type, u16 clobber, void *firstinsn, unsigned len);
51
52 /* Basic arch-specific setup */
53 void (*arch_setup)(void);
54 char *(*memory_setup)(void);
55 void (*init_IRQ)(void);
56 void (*time_init)(void);
57
58 /*
59 * Called before/after init_mm pagetable setup. setup_start
60 * may reset %cr3, and may pre-install parts of the pagetable;
61 * pagetable setup is expected to preserve any existing
62 * mapping.
63 */
64 void (*pagetable_setup_start)(pgd_t *pgd_base);
65 void (*pagetable_setup_done)(pgd_t *pgd_base);
66
67 /* Print a banner to identify the environment */
68 void (*banner)(void);
69
70 /* Set and set time of day */
71 unsigned long (*get_wallclock)(void);
72 int (*set_wallclock)(unsigned long);
73
74 /* cpuid emulation, mostly so that caps bits can be disabled */
75 void (*cpuid)(unsigned int *eax, unsigned int *ebx,
76 unsigned int *ecx, unsigned int *edx);
77
78 /* hooks for various privileged instructions */
79 unsigned long (*get_debugreg)(int regno);
80 void (*set_debugreg)(int regno, unsigned long value);
81
82 void (*clts)(void);
83
84 unsigned long (*read_cr0)(void);
85 void (*write_cr0)(unsigned long);
86
87 unsigned long (*read_cr2)(void);
88 void (*write_cr2)(unsigned long);
89
90 unsigned long (*read_cr3)(void);
91 void (*write_cr3)(unsigned long);
92
93 unsigned long (*read_cr4_safe)(void);
94 unsigned long (*read_cr4)(void);
95 void (*write_cr4)(unsigned long);
96
97 /*
98 * Get/set interrupt state. save_fl and restore_fl are only
99 * expected to use X86_EFLAGS_IF; all other bits
100 * returned from save_fl are undefined, and may be ignored by
101 * restore_fl.
102 */
103 unsigned long (*save_fl)(void);
104 void (*restore_fl)(unsigned long);
105 void (*irq_disable)(void);
106 void (*irq_enable)(void);
107 void (*safe_halt)(void);
108 void (*halt)(void);
109
110 void (*wbinvd)(void);
111
112 /* MSR, PMC and TSR operations.
113 err = 0/-EFAULT. wrmsr returns 0/-EFAULT. */
114 u64 (*read_msr)(unsigned int msr, int *err);
115 int (*write_msr)(unsigned int msr, u64 val);
116
117 u64 (*read_tsc)(void);
118 u64 (*read_pmc)(void);
119 u64 (*get_scheduled_cycles)(void);
120 unsigned long (*get_cpu_khz)(void);
121
122 /* Segment descriptor handling */
123 void (*load_tr_desc)(void);
124 void (*load_gdt)(const struct Xgt_desc_struct *);
125 void (*load_idt)(const struct Xgt_desc_struct *);
126 void (*store_gdt)(struct Xgt_desc_struct *);
127 void (*store_idt)(struct Xgt_desc_struct *);
128 void (*set_ldt)(const void *desc, unsigned entries);
129 unsigned long (*store_tr)(void);
130 void (*load_tls)(struct thread_struct *t, unsigned int cpu);
131 void (*write_ldt_entry)(struct desc_struct *,
132 int entrynum, u32 low, u32 high);
133 void (*write_gdt_entry)(struct desc_struct *,
134 int entrynum, u32 low, u32 high);
135 void (*write_idt_entry)(struct desc_struct *,
136 int entrynum, u32 low, u32 high);
137 void (*load_esp0)(struct tss_struct *tss, struct thread_struct *t);
138
139 void (*set_iopl_mask)(unsigned mask);
140 void (*io_delay)(void);
141
142 /*
143 * Hooks for intercepting the creation/use/destruction of an
144 * mm_struct.
145 */
146 void (*activate_mm)(struct mm_struct *prev,
147 struct mm_struct *next);
148 void (*dup_mmap)(struct mm_struct *oldmm,
149 struct mm_struct *mm);
150 void (*exit_mmap)(struct mm_struct *mm);
151
152 #ifdef CONFIG_X86_LOCAL_APIC
153 /*
154 * Direct APIC operations, principally for VMI. Ideally
155 * these shouldn't be in this interface.
156 */
157 void (*apic_write)(unsigned long reg, unsigned long v);
158 void (*apic_write_atomic)(unsigned long reg, unsigned long v);
159 unsigned long (*apic_read)(unsigned long reg);
160 void (*setup_boot_clock)(void);
161 void (*setup_secondary_clock)(void);
162
163 void (*startup_ipi_hook)(int phys_apicid,
164 unsigned long start_eip,
165 unsigned long start_esp);
166 #endif
167
168 /* TLB operations */
169 void (*flush_tlb_user)(void);
170 void (*flush_tlb_kernel)(void);
171 void (*flush_tlb_single)(unsigned long addr);
172 void (*flush_tlb_others)(const cpumask_t *cpus, struct mm_struct *mm,
173 unsigned long va);
174
175 /* Hooks for allocating/releasing pagetable pages */
176 void (*alloc_pt)(u32 pfn);
177 void (*alloc_pd)(u32 pfn);
178 void (*alloc_pd_clone)(u32 pfn, u32 clonepfn, u32 start, u32 count);
179 void (*release_pt)(u32 pfn);
180 void (*release_pd)(u32 pfn);
181
182 /* Pagetable manipulation functions */
183 void (*set_pte)(pte_t *ptep, pte_t pteval);
184 void (*set_pte_at)(struct mm_struct *mm, unsigned long addr,
185 pte_t *ptep, pte_t pteval);
186 void (*set_pmd)(pmd_t *pmdp, pmd_t pmdval);
187 void (*pte_update)(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
188 void (*pte_update_defer)(struct mm_struct *mm,
189 unsigned long addr, pte_t *ptep);
190
191 #ifdef CONFIG_HIGHPTE
192 void *(*kmap_atomic_pte)(struct page *page, enum km_type type);
193 #endif
194
195 #ifdef CONFIG_X86_PAE
196 void (*set_pte_atomic)(pte_t *ptep, pte_t pteval);
197 void (*set_pte_present)(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte);
198 void (*set_pud)(pud_t *pudp, pud_t pudval);
199 void (*pte_clear)(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
200 void (*pmd_clear)(pmd_t *pmdp);
201
202 unsigned long long (*pte_val)(pte_t);
203 unsigned long long (*pmd_val)(pmd_t);
204 unsigned long long (*pgd_val)(pgd_t);
205
206 pte_t (*make_pte)(unsigned long long pte);
207 pmd_t (*make_pmd)(unsigned long long pmd);
208 pgd_t (*make_pgd)(unsigned long long pgd);
209 #else
210 unsigned long (*pte_val)(pte_t);
211 unsigned long (*pgd_val)(pgd_t);
212
213 pte_t (*make_pte)(unsigned long pte);
214 pgd_t (*make_pgd)(unsigned long pgd);
215 #endif
216
217 /* Set deferred update mode, used for batching operations. */
218 void (*set_lazy_mode)(enum paravirt_lazy_mode mode);
219
220 /* These two are jmp to, not actually called. */
221 void (*irq_enable_sysexit)(void);
222 void (*iret)(void);
223 };
224
225 /* Mark a paravirt probe function. */
226 #define paravirt_probe(fn) \
227 static asmlinkage void (*__paravirtprobe_##fn)(void) __attribute_used__ \
228 __attribute__((__section__(".paravirtprobe"))) = fn
229
230 extern struct paravirt_ops paravirt_ops;
231
232 #define PARAVIRT_PATCH(x) \
233 (offsetof(struct paravirt_ops, x) / sizeof(void *))
234
235 #define paravirt_type(type) \
236 [paravirt_typenum] "i" (PARAVIRT_PATCH(type))
237 #define paravirt_clobber(clobber) \
238 [paravirt_clobber] "i" (clobber)
239
240 /*
241 * Generate some code, and mark it as patchable by the
242 * apply_paravirt() alternate instruction patcher.
243 */
244 #define _paravirt_alt(insn_string, type, clobber) \
245 "771:\n\t" insn_string "\n" "772:\n" \
246 ".pushsection .parainstructions,\"a\"\n" \
247 " .long 771b\n" \
248 " .byte " type "\n" \
249 " .byte 772b-771b\n" \
250 " .short " clobber "\n" \
251 ".popsection\n"
252
253 /* Generate patchable code, with the default asm parameters. */
254 #define paravirt_alt(insn_string) \
255 _paravirt_alt(insn_string, "%c[paravirt_typenum]", "%c[paravirt_clobber]")
256
257 unsigned paravirt_patch_nop(void);
258 unsigned paravirt_patch_ignore(unsigned len);
259 unsigned paravirt_patch_call(void *target, u16 tgt_clobbers,
260 void *site, u16 site_clobbers,
261 unsigned len);
262 unsigned paravirt_patch_jmp(void *target, void *site, unsigned len);
263 unsigned paravirt_patch_default(u8 type, u16 clobbers, void *site, unsigned len);
264
265 unsigned paravirt_patch_insns(void *site, unsigned len,
266 const char *start, const char *end);
267
268
269 /*
270 * This generates an indirect call based on the operation type number.
271 * The type number, computed in PARAVIRT_PATCH, is derived from the
272 * offset into the paravirt_ops structure, and can therefore be freely
273 * converted back into a structure offset.
274 */
275 #define PARAVIRT_CALL "call *(paravirt_ops+%c[paravirt_typenum]*4);"
276
277 /*
278 * These macros are intended to wrap calls into a paravirt_ops
279 * operation, so that they can be later identified and patched at
280 * runtime.
281 *
282 * Normally, a call to a pv_op function is a simple indirect call:
283 * (paravirt_ops.operations)(args...).
284 *
285 * Unfortunately, this is a relatively slow operation for modern CPUs,
286 * because it cannot necessarily determine what the destination
287 * address is. In this case, the address is a runtime constant, so at
288 * the very least we can patch the call to e a simple direct call, or
289 * ideally, patch an inline implementation into the callsite. (Direct
290 * calls are essentially free, because the call and return addresses
291 * are completely predictable.)
292 *
293 * These macros rely on the standard gcc "regparm(3)" calling
294 * convention, in which the first three arguments are placed in %eax,
295 * %edx, %ecx (in that order), and the remaining arguments are placed
296 * on the stack. All caller-save registers (eax,edx,ecx) are expected
297 * to be modified (either clobbered or used for return values).
298 *
299 * The call instruction itself is marked by placing its start address
300 * and size into the .parainstructions section, so that
301 * apply_paravirt() in arch/i386/kernel/alternative.c can do the
302 * appropriate patching under the control of the backend paravirt_ops
303 * implementation.
304 *
305 * Unfortunately there's no way to get gcc to generate the args setup
306 * for the call, and then allow the call itself to be generated by an
307 * inline asm. Because of this, we must do the complete arg setup and
308 * return value handling from within these macros. This is fairly
309 * cumbersome.
310 *
311 * There are 5 sets of PVOP_* macros for dealing with 0-4 arguments.
312 * It could be extended to more arguments, but there would be little
313 * to be gained from that. For each number of arguments, there are
314 * the two VCALL and CALL variants for void and non-void functions.
315 *
316 * When there is a return value, the invoker of the macro must specify
317 * the return type. The macro then uses sizeof() on that type to
318 * determine whether its a 32 or 64 bit value, and places the return
319 * in the right register(s) (just %eax for 32-bit, and %edx:%eax for
320 * 64-bit).
321 *
322 * 64-bit arguments are passed as a pair of adjacent 32-bit arguments
323 * in low,high order.
324 *
325 * Small structures are passed and returned in registers. The macro
326 * calling convention can't directly deal with this, so the wrapper
327 * functions must do this.
328 *
329 * These PVOP_* macros are only defined within this header. This
330 * means that all uses must be wrapped in inline functions. This also
331 * makes sure the incoming and outgoing types are always correct.
332 */
333 #define __PVOP_CALL(rettype, op, pre, post, ...) \
334 ({ \
335 rettype __ret; \
336 unsigned long __eax, __edx, __ecx; \
337 if (sizeof(rettype) > sizeof(unsigned long)) { \
338 asm volatile(pre \
339 paravirt_alt(PARAVIRT_CALL) \
340 post \
341 : "=a" (__eax), "=d" (__edx), \
342 "=c" (__ecx) \
343 : paravirt_type(op), \
344 paravirt_clobber(CLBR_ANY), \
345 ##__VA_ARGS__ \
346 : "memory", "cc"); \
347 __ret = (rettype)((((u64)__edx) << 32) | __eax); \
348 } else { \
349 asm volatile(pre \
350 paravirt_alt(PARAVIRT_CALL) \
351 post \
352 : "=a" (__eax), "=d" (__edx), \
353 "=c" (__ecx) \
354 : paravirt_type(op), \
355 paravirt_clobber(CLBR_ANY), \
356 ##__VA_ARGS__ \
357 : "memory", "cc"); \
358 __ret = (rettype)__eax; \
359 } \
360 __ret; \
361 })
362 #define __PVOP_VCALL(op, pre, post, ...) \
363 ({ \
364 unsigned long __eax, __edx, __ecx; \
365 asm volatile(pre \
366 paravirt_alt(PARAVIRT_CALL) \
367 post \
368 : "=a" (__eax), "=d" (__edx), "=c" (__ecx) \
369 : paravirt_type(op), \
370 paravirt_clobber(CLBR_ANY), \
371 ##__VA_ARGS__ \
372 : "memory", "cc"); \
373 })
374
375 #define PVOP_CALL0(rettype, op) \
376 __PVOP_CALL(rettype, op, "", "")
377 #define PVOP_VCALL0(op) \
378 __PVOP_VCALL(op, "", "")
379
380 #define PVOP_CALL1(rettype, op, arg1) \
381 __PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)))
382 #define PVOP_VCALL1(op, arg1) \
383 __PVOP_VCALL(op, "", "", "0" ((u32)(arg1)))
384
385 #define PVOP_CALL2(rettype, op, arg1, arg2) \
386 __PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)), "1" ((u32)(arg2)))
387 #define PVOP_VCALL2(op, arg1, arg2) \
388 __PVOP_VCALL(op, "", "", "0" ((u32)(arg1)), "1" ((u32)(arg2)))
389
390 #define PVOP_CALL3(rettype, op, arg1, arg2, arg3) \
391 __PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)), \
392 "1"((u32)(arg2)), "2"((u32)(arg3)))
393 #define PVOP_VCALL3(op, arg1, arg2, arg3) \
394 __PVOP_VCALL(op, "", "", "0" ((u32)(arg1)), "1"((u32)(arg2)), \
395 "2"((u32)(arg3)))
396
397 #define PVOP_CALL4(rettype, op, arg1, arg2, arg3, arg4) \
398 __PVOP_CALL(rettype, op, \
399 "push %[_arg4];", "lea 4(%%esp),%%esp;", \
400 "0" ((u32)(arg1)), "1" ((u32)(arg2)), \
401 "2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
402 #define PVOP_VCALL4(op, arg1, arg2, arg3, arg4) \
403 __PVOP_VCALL(op, \
404 "push %[_arg4];", "lea 4(%%esp),%%esp;", \
405 "0" ((u32)(arg1)), "1" ((u32)(arg2)), \
406 "2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
407
408 static inline int paravirt_enabled(void)
409 {
410 return paravirt_ops.paravirt_enabled;
411 }
412
413 static inline void load_esp0(struct tss_struct *tss,
414 struct thread_struct *thread)
415 {
416 PVOP_VCALL2(load_esp0, tss, thread);
417 }
418
419 #define ARCH_SETUP paravirt_ops.arch_setup();
420 static inline unsigned long get_wallclock(void)
421 {
422 return PVOP_CALL0(unsigned long, get_wallclock);
423 }
424
425 static inline int set_wallclock(unsigned long nowtime)
426 {
427 return PVOP_CALL1(int, set_wallclock, nowtime);
428 }
429
430 static inline void (*choose_time_init(void))(void)
431 {
432 return paravirt_ops.time_init;
433 }
434
435 /* The paravirtualized CPUID instruction. */
436 static inline void __cpuid(unsigned int *eax, unsigned int *ebx,
437 unsigned int *ecx, unsigned int *edx)
438 {
439 PVOP_VCALL4(cpuid, eax, ebx, ecx, edx);
440 }
441
442 /*
443 * These special macros can be used to get or set a debugging register
444 */
445 static inline unsigned long paravirt_get_debugreg(int reg)
446 {
447 return PVOP_CALL1(unsigned long, get_debugreg, reg);
448 }
449 #define get_debugreg(var, reg) var = paravirt_get_debugreg(reg)
450 static inline void set_debugreg(unsigned long val, int reg)
451 {
452 PVOP_VCALL2(set_debugreg, reg, val);
453 }
454
455 static inline void clts(void)
456 {
457 PVOP_VCALL0(clts);
458 }
459
460 static inline unsigned long read_cr0(void)
461 {
462 return PVOP_CALL0(unsigned long, read_cr0);
463 }
464
465 static inline void write_cr0(unsigned long x)
466 {
467 PVOP_VCALL1(write_cr0, x);
468 }
469
470 static inline unsigned long read_cr2(void)
471 {
472 return PVOP_CALL0(unsigned long, read_cr2);
473 }
474
475 static inline void write_cr2(unsigned long x)
476 {
477 PVOP_VCALL1(write_cr2, x);
478 }
479
480 static inline unsigned long read_cr3(void)
481 {
482 return PVOP_CALL0(unsigned long, read_cr3);
483 }
484
485 static inline void write_cr3(unsigned long x)
486 {
487 PVOP_VCALL1(write_cr3, x);
488 }
489
490 static inline unsigned long read_cr4(void)
491 {
492 return PVOP_CALL0(unsigned long, read_cr4);
493 }
494 static inline unsigned long read_cr4_safe(void)
495 {
496 return PVOP_CALL0(unsigned long, read_cr4_safe);
497 }
498
499 static inline void write_cr4(unsigned long x)
500 {
501 PVOP_VCALL1(write_cr4, x);
502 }
503
504 static inline void raw_safe_halt(void)
505 {
506 PVOP_VCALL0(safe_halt);
507 }
508
509 static inline void halt(void)
510 {
511 PVOP_VCALL0(safe_halt);
512 }
513
514 static inline void wbinvd(void)
515 {
516 PVOP_VCALL0(wbinvd);
517 }
518
519 #define get_kernel_rpl() (paravirt_ops.kernel_rpl)
520
521 static inline u64 paravirt_read_msr(unsigned msr, int *err)
522 {
523 return PVOP_CALL2(u64, read_msr, msr, err);
524 }
525 static inline int paravirt_write_msr(unsigned msr, unsigned low, unsigned high)
526 {
527 return PVOP_CALL3(int, write_msr, msr, low, high);
528 }
529
530 /* These should all do BUG_ON(_err), but our headers are too tangled. */
531 #define rdmsr(msr,val1,val2) do { \
532 int _err; \
533 u64 _l = paravirt_read_msr(msr, &_err); \
534 val1 = (u32)_l; \
535 val2 = _l >> 32; \
536 } while(0)
537
538 #define wrmsr(msr,val1,val2) do { \
539 paravirt_write_msr(msr, val1, val2); \
540 } while(0)
541
542 #define rdmsrl(msr,val) do { \
543 int _err; \
544 val = paravirt_read_msr(msr, &_err); \
545 } while(0)
546
547 #define wrmsrl(msr,val) ((void)paravirt_write_msr(msr, val, 0))
548 #define wrmsr_safe(msr,a,b) paravirt_write_msr(msr, a, b)
549
550 /* rdmsr with exception handling */
551 #define rdmsr_safe(msr,a,b) ({ \
552 int _err; \
553 u64 _l = paravirt_read_msr(msr, &_err); \
554 (*a) = (u32)_l; \
555 (*b) = _l >> 32; \
556 _err; })
557
558
559 static inline u64 paravirt_read_tsc(void)
560 {
561 return PVOP_CALL0(u64, read_tsc);
562 }
563 #define rdtsc(low,high) do { \
564 u64 _l = paravirt_read_tsc(); \
565 low = (u32)_l; \
566 high = _l >> 32; \
567 } while(0)
568
569 #define rdtscl(low) do { \
570 u64 _l = paravirt_read_tsc(); \
571 low = (int)_l; \
572 } while(0)
573
574 #define rdtscll(val) (val = paravirt_read_tsc())
575
576 #define get_scheduled_cycles(val) (val = paravirt_ops.get_scheduled_cycles())
577 #define calculate_cpu_khz() (paravirt_ops.get_cpu_khz())
578
579 #define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
580
581 static inline unsigned long long paravirt_read_pmc(int counter)
582 {
583 return PVOP_CALL1(u64, read_pmc, counter);
584 }
585
586 #define rdpmc(counter,low,high) do { \
587 u64 _l = paravirt_read_pmc(counter); \
588 low = (u32)_l; \
589 high = _l >> 32; \
590 } while(0)
591
592 static inline void load_TR_desc(void)
593 {
594 PVOP_VCALL0(load_tr_desc);
595 }
596 static inline void load_gdt(const struct Xgt_desc_struct *dtr)
597 {
598 PVOP_VCALL1(load_gdt, dtr);
599 }
600 static inline void load_idt(const struct Xgt_desc_struct *dtr)
601 {
602 PVOP_VCALL1(load_idt, dtr);
603 }
604 static inline void set_ldt(const void *addr, unsigned entries)
605 {
606 PVOP_VCALL2(set_ldt, addr, entries);
607 }
608 static inline void store_gdt(struct Xgt_desc_struct *dtr)
609 {
610 PVOP_VCALL1(store_gdt, dtr);
611 }
612 static inline void store_idt(struct Xgt_desc_struct *dtr)
613 {
614 PVOP_VCALL1(store_idt, dtr);
615 }
616 static inline unsigned long paravirt_store_tr(void)
617 {
618 return PVOP_CALL0(unsigned long, store_tr);
619 }
620 #define store_tr(tr) ((tr) = paravirt_store_tr())
621 static inline void load_TLS(struct thread_struct *t, unsigned cpu)
622 {
623 PVOP_VCALL2(load_tls, t, cpu);
624 }
625 static inline void write_ldt_entry(void *dt, int entry, u32 low, u32 high)
626 {
627 PVOP_VCALL4(write_ldt_entry, dt, entry, low, high);
628 }
629 static inline void write_gdt_entry(void *dt, int entry, u32 low, u32 high)
630 {
631 PVOP_VCALL4(write_gdt_entry, dt, entry, low, high);
632 }
633 static inline void write_idt_entry(void *dt, int entry, u32 low, u32 high)
634 {
635 PVOP_VCALL4(write_idt_entry, dt, entry, low, high);
636 }
637 static inline void set_iopl_mask(unsigned mask)
638 {
639 PVOP_VCALL1(set_iopl_mask, mask);
640 }
641
642 /* The paravirtualized I/O functions */
643 static inline void slow_down_io(void) {
644 paravirt_ops.io_delay();
645 #ifdef REALLY_SLOW_IO
646 paravirt_ops.io_delay();
647 paravirt_ops.io_delay();
648 paravirt_ops.io_delay();
649 #endif
650 }
651
652 #ifdef CONFIG_X86_LOCAL_APIC
653 /*
654 * Basic functions accessing APICs.
655 */
656 static inline void apic_write(unsigned long reg, unsigned long v)
657 {
658 PVOP_VCALL2(apic_write, reg, v);
659 }
660
661 static inline void apic_write_atomic(unsigned long reg, unsigned long v)
662 {
663 PVOP_VCALL2(apic_write_atomic, reg, v);
664 }
665
666 static inline unsigned long apic_read(unsigned long reg)
667 {
668 return PVOP_CALL1(unsigned long, apic_read, reg);
669 }
670
671 static inline void setup_boot_clock(void)
672 {
673 PVOP_VCALL0(setup_boot_clock);
674 }
675
676 static inline void setup_secondary_clock(void)
677 {
678 PVOP_VCALL0(setup_secondary_clock);
679 }
680 #endif
681
682 static inline void paravirt_pagetable_setup_start(pgd_t *base)
683 {
684 if (paravirt_ops.pagetable_setup_start)
685 (*paravirt_ops.pagetable_setup_start)(base);
686 }
687
688 static inline void paravirt_pagetable_setup_done(pgd_t *base)
689 {
690 if (paravirt_ops.pagetable_setup_done)
691 (*paravirt_ops.pagetable_setup_done)(base);
692 }
693
694 #ifdef CONFIG_SMP
695 static inline void startup_ipi_hook(int phys_apicid, unsigned long start_eip,
696 unsigned long start_esp)
697 {
698 PVOP_VCALL3(startup_ipi_hook, phys_apicid, start_eip, start_esp);
699 }
700 #endif
701
702 static inline void paravirt_activate_mm(struct mm_struct *prev,
703 struct mm_struct *next)
704 {
705 PVOP_VCALL2(activate_mm, prev, next);
706 }
707
708 static inline void arch_dup_mmap(struct mm_struct *oldmm,
709 struct mm_struct *mm)
710 {
711 PVOP_VCALL2(dup_mmap, oldmm, mm);
712 }
713
714 static inline void arch_exit_mmap(struct mm_struct *mm)
715 {
716 PVOP_VCALL1(exit_mmap, mm);
717 }
718
719 static inline void __flush_tlb(void)
720 {
721 PVOP_VCALL0(flush_tlb_user);
722 }
723 static inline void __flush_tlb_global(void)
724 {
725 PVOP_VCALL0(flush_tlb_kernel);
726 }
727 static inline void __flush_tlb_single(unsigned long addr)
728 {
729 PVOP_VCALL1(flush_tlb_single, addr);
730 }
731
732 static inline void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm,
733 unsigned long va)
734 {
735 PVOP_VCALL3(flush_tlb_others, &cpumask, mm, va);
736 }
737
738 static inline void paravirt_alloc_pt(unsigned pfn)
739 {
740 PVOP_VCALL1(alloc_pt, pfn);
741 }
742 static inline void paravirt_release_pt(unsigned pfn)
743 {
744 PVOP_VCALL1(release_pt, pfn);
745 }
746
747 static inline void paravirt_alloc_pd(unsigned pfn)
748 {
749 PVOP_VCALL1(alloc_pd, pfn);
750 }
751
752 static inline void paravirt_alloc_pd_clone(unsigned pfn, unsigned clonepfn,
753 unsigned start, unsigned count)
754 {
755 PVOP_VCALL4(alloc_pd_clone, pfn, clonepfn, start, count);
756 }
757 static inline void paravirt_release_pd(unsigned pfn)
758 {
759 PVOP_VCALL1(release_pd, pfn);
760 }
761
762 #ifdef CONFIG_HIGHPTE
763 static inline void *kmap_atomic_pte(struct page *page, enum km_type type)
764 {
765 unsigned long ret;
766 ret = PVOP_CALL2(unsigned long, kmap_atomic_pte, page, type);
767 return (void *)ret;
768 }
769 #endif
770
771 static inline void pte_update(struct mm_struct *mm, unsigned long addr,
772 pte_t *ptep)
773 {
774 PVOP_VCALL3(pte_update, mm, addr, ptep);
775 }
776
777 static inline void pte_update_defer(struct mm_struct *mm, unsigned long addr,
778 pte_t *ptep)
779 {
780 PVOP_VCALL3(pte_update_defer, mm, addr, ptep);
781 }
782
783 #ifdef CONFIG_X86_PAE
784 static inline pte_t __pte(unsigned long long val)
785 {
786 unsigned long long ret = PVOP_CALL2(unsigned long long, make_pte,
787 val, val >> 32);
788 return (pte_t) { ret, ret >> 32 };
789 }
790
791 static inline pmd_t __pmd(unsigned long long val)
792 {
793 return (pmd_t) { PVOP_CALL2(unsigned long long, make_pmd, val, val >> 32) };
794 }
795
796 static inline pgd_t __pgd(unsigned long long val)
797 {
798 return (pgd_t) { PVOP_CALL2(unsigned long long, make_pgd, val, val >> 32) };
799 }
800
801 static inline unsigned long long pte_val(pte_t x)
802 {
803 return PVOP_CALL2(unsigned long long, pte_val, x.pte_low, x.pte_high);
804 }
805
806 static inline unsigned long long pmd_val(pmd_t x)
807 {
808 return PVOP_CALL2(unsigned long long, pmd_val, x.pmd, x.pmd >> 32);
809 }
810
811 static inline unsigned long long pgd_val(pgd_t x)
812 {
813 return PVOP_CALL2(unsigned long long, pgd_val, x.pgd, x.pgd >> 32);
814 }
815
816 static inline void set_pte(pte_t *ptep, pte_t pteval)
817 {
818 PVOP_VCALL3(set_pte, ptep, pteval.pte_low, pteval.pte_high);
819 }
820
821 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
822 pte_t *ptep, pte_t pteval)
823 {
824 /* 5 arg words */
825 paravirt_ops.set_pte_at(mm, addr, ptep, pteval);
826 }
827
828 static inline void set_pte_atomic(pte_t *ptep, pte_t pteval)
829 {
830 PVOP_VCALL3(set_pte_atomic, ptep, pteval.pte_low, pteval.pte_high);
831 }
832
833 static inline void set_pte_present(struct mm_struct *mm, unsigned long addr,
834 pte_t *ptep, pte_t pte)
835 {
836 /* 5 arg words */
837 paravirt_ops.set_pte_present(mm, addr, ptep, pte);
838 }
839
840 static inline void set_pmd(pmd_t *pmdp, pmd_t pmdval)
841 {
842 PVOP_VCALL3(set_pmd, pmdp, pmdval.pmd, pmdval.pmd >> 32);
843 }
844
845 static inline void set_pud(pud_t *pudp, pud_t pudval)
846 {
847 PVOP_VCALL3(set_pud, pudp, pudval.pgd.pgd, pudval.pgd.pgd >> 32);
848 }
849
850 static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
851 {
852 PVOP_VCALL3(pte_clear, mm, addr, ptep);
853 }
854
855 static inline void pmd_clear(pmd_t *pmdp)
856 {
857 PVOP_VCALL1(pmd_clear, pmdp);
858 }
859
860 #else /* !CONFIG_X86_PAE */
861
862 static inline pte_t __pte(unsigned long val)
863 {
864 return (pte_t) { PVOP_CALL1(unsigned long, make_pte, val) };
865 }
866
867 static inline pgd_t __pgd(unsigned long val)
868 {
869 return (pgd_t) { PVOP_CALL1(unsigned long, make_pgd, val) };
870 }
871
872 static inline unsigned long pte_val(pte_t x)
873 {
874 return PVOP_CALL1(unsigned long, pte_val, x.pte_low);
875 }
876
877 static inline unsigned long pgd_val(pgd_t x)
878 {
879 return PVOP_CALL1(unsigned long, pgd_val, x.pgd);
880 }
881
882 static inline void set_pte(pte_t *ptep, pte_t pteval)
883 {
884 PVOP_VCALL2(set_pte, ptep, pteval.pte_low);
885 }
886
887 static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
888 pte_t *ptep, pte_t pteval)
889 {
890 PVOP_VCALL4(set_pte_at, mm, addr, ptep, pteval.pte_low);
891 }
892
893 static inline void set_pmd(pmd_t *pmdp, pmd_t pmdval)
894 {
895 PVOP_VCALL2(set_pmd, pmdp, pmdval.pud.pgd.pgd);
896 }
897 #endif /* CONFIG_X86_PAE */
898
899 #define __HAVE_ARCH_ENTER_LAZY_CPU_MODE
900 static inline void arch_enter_lazy_cpu_mode(void)
901 {
902 PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_CPU);
903 }
904
905 static inline void arch_leave_lazy_cpu_mode(void)
906 {
907 PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_NONE);
908 }
909
910 static inline void arch_flush_lazy_cpu_mode(void)
911 {
912 PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_FLUSH);
913 }
914
915
916 #define __HAVE_ARCH_ENTER_LAZY_MMU_MODE
917 static inline void arch_enter_lazy_mmu_mode(void)
918 {
919 PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_MMU);
920 }
921
922 static inline void arch_leave_lazy_mmu_mode(void)
923 {
924 PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_NONE);
925 }
926
927 static inline void arch_flush_lazy_mmu_mode(void)
928 {
929 PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_FLUSH);
930 }
931
932 void _paravirt_nop(void);
933 #define paravirt_nop ((void *)_paravirt_nop)
934
935 /* These all sit in the .parainstructions section to tell us what to patch. */
936 struct paravirt_patch_site {
937 u8 *instr; /* original instructions */
938 u8 instrtype; /* type of this instruction */
939 u8 len; /* length of original instruction */
940 u16 clobbers; /* what registers you may clobber */
941 };
942
943 extern struct paravirt_patch_site __parainstructions[],
944 __parainstructions_end[];
945
946 static inline unsigned long __raw_local_save_flags(void)
947 {
948 unsigned long f;
949
950 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
951 PARAVIRT_CALL
952 "popl %%edx; popl %%ecx")
953 : "=a"(f)
954 : paravirt_type(save_fl),
955 paravirt_clobber(CLBR_EAX)
956 : "memory", "cc");
957 return f;
958 }
959
960 static inline void raw_local_irq_restore(unsigned long f)
961 {
962 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
963 PARAVIRT_CALL
964 "popl %%edx; popl %%ecx")
965 : "=a"(f)
966 : "0"(f),
967 paravirt_type(restore_fl),
968 paravirt_clobber(CLBR_EAX)
969 : "memory", "cc");
970 }
971
972 static inline void raw_local_irq_disable(void)
973 {
974 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
975 PARAVIRT_CALL
976 "popl %%edx; popl %%ecx")
977 :
978 : paravirt_type(irq_disable),
979 paravirt_clobber(CLBR_EAX)
980 : "memory", "eax", "cc");
981 }
982
983 static inline void raw_local_irq_enable(void)
984 {
985 asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
986 PARAVIRT_CALL
987 "popl %%edx; popl %%ecx")
988 :
989 : paravirt_type(irq_enable),
990 paravirt_clobber(CLBR_EAX)
991 : "memory", "eax", "cc");
992 }
993
994 static inline unsigned long __raw_local_irq_save(void)
995 {
996 unsigned long f;
997
998 f = __raw_local_save_flags();
999 raw_local_irq_disable();
1000 return f;
1001 }
1002
1003 #define CLI_STRING \
1004 _paravirt_alt("pushl %%ecx; pushl %%edx;" \
1005 "call *paravirt_ops+%c[paravirt_cli_type]*4;" \
1006 "popl %%edx; popl %%ecx", \
1007 "%c[paravirt_cli_type]", "%c[paravirt_clobber]")
1008
1009 #define STI_STRING \
1010 _paravirt_alt("pushl %%ecx; pushl %%edx;" \
1011 "call *paravirt_ops+%c[paravirt_sti_type]*4;" \
1012 "popl %%edx; popl %%ecx", \
1013 "%c[paravirt_sti_type]", "%c[paravirt_clobber]")
1014
1015 #define CLI_STI_CLOBBERS , "%eax"
1016 #define CLI_STI_INPUT_ARGS \
1017 , \
1018 [paravirt_cli_type] "i" (PARAVIRT_PATCH(irq_disable)), \
1019 [paravirt_sti_type] "i" (PARAVIRT_PATCH(irq_enable)), \
1020 paravirt_clobber(CLBR_EAX)
1021
1022 /* Make sure as little as possible of this mess escapes. */
1023 #undef PARAVIRT_CALL
1024 #undef __PVOP_CALL
1025 #undef __PVOP_VCALL
1026 #undef PVOP_VCALL0
1027 #undef PVOP_CALL0
1028 #undef PVOP_VCALL1
1029 #undef PVOP_CALL1
1030 #undef PVOP_VCALL2
1031 #undef PVOP_CALL2
1032 #undef PVOP_VCALL3
1033 #undef PVOP_CALL3
1034 #undef PVOP_VCALL4
1035 #undef PVOP_CALL4
1036
1037 #else /* __ASSEMBLY__ */
1038
1039 #define PARA_PATCH(off) ((off) / 4)
1040
1041 #define PARA_SITE(ptype, clobbers, ops) \
1042 771:; \
1043 ops; \
1044 772:; \
1045 .pushsection .parainstructions,"a"; \
1046 .long 771b; \
1047 .byte ptype; \
1048 .byte 772b-771b; \
1049 .short clobbers; \
1050 .popsection
1051
1052 #define INTERRUPT_RETURN \
1053 PARA_SITE(PARA_PATCH(PARAVIRT_iret), CLBR_NONE, \
1054 jmp *%cs:paravirt_ops+PARAVIRT_iret)
1055
1056 #define DISABLE_INTERRUPTS(clobbers) \
1057 PARA_SITE(PARA_PATCH(PARAVIRT_irq_disable), clobbers, \
1058 pushl %eax; pushl %ecx; pushl %edx; \
1059 call *%cs:paravirt_ops+PARAVIRT_irq_disable; \
1060 popl %edx; popl %ecx; popl %eax) \
1061
1062 #define ENABLE_INTERRUPTS(clobbers) \
1063 PARA_SITE(PARA_PATCH(PARAVIRT_irq_enable), clobbers, \
1064 pushl %eax; pushl %ecx; pushl %edx; \
1065 call *%cs:paravirt_ops+PARAVIRT_irq_enable; \
1066 popl %edx; popl %ecx; popl %eax)
1067
1068 #define ENABLE_INTERRUPTS_SYSEXIT \
1069 PARA_SITE(PARA_PATCH(PARAVIRT_irq_enable_sysexit), CLBR_NONE, \
1070 jmp *%cs:paravirt_ops+PARAVIRT_irq_enable_sysexit)
1071
1072 #define GET_CR0_INTO_EAX \
1073 push %ecx; push %edx; \
1074 call *paravirt_ops+PARAVIRT_read_cr0; \
1075 pop %edx; pop %ecx
1076
1077 #endif /* __ASSEMBLY__ */
1078 #endif /* CONFIG_PARAVIRT */
1079 #endif /* __ASM_PARAVIRT_H */
This page took 0.05312 seconds and 5 git commands to generate.