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
21struct page;
22struct thread_struct;
23struct Xgt_desc_struct;
24struct tss_struct;
25struct mm_struct;
26struct desc_struct;
27
28/* Lazy mode for batching updates / context switch */
29enum 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
36struct 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
225extern struct paravirt_ops paravirt_ops;
226
227#define PARAVIRT_PATCH(x)					\
228	(offsetof(struct paravirt_ops, x) / sizeof(void *))
229
230#define paravirt_type(type)					\
231	[paravirt_typenum] "i" (PARAVIRT_PATCH(type))
232#define paravirt_clobber(clobber)		\
233	[paravirt_clobber] "i" (clobber)
234
235/*
236 * Generate some code, and mark it as patchable by the
237 * apply_paravirt() alternate instruction patcher.
238 */
239#define _paravirt_alt(insn_string, type, clobber)	\
240	"771:\n\t" insn_string "\n" "772:\n"		\
241	".pushsection .parainstructions,\"a\"\n"	\
242	"  .long 771b\n"				\
243	"  .byte " type "\n"				\
244	"  .byte 772b-771b\n"				\
245	"  .short " clobber "\n"			\
246	".popsection\n"
247
248/* Generate patchable code, with the default asm parameters. */
249#define paravirt_alt(insn_string)					\
250	_paravirt_alt(insn_string, "%c[paravirt_typenum]", "%c[paravirt_clobber]")
251
252unsigned paravirt_patch_nop(void);
253unsigned paravirt_patch_ignore(unsigned len);
254unsigned paravirt_patch_call(void *target, u16 tgt_clobbers,
255			     void *site, u16 site_clobbers,
256			     unsigned len);
257unsigned paravirt_patch_jmp(void *target, void *site, unsigned len);
258unsigned paravirt_patch_default(u8 type, u16 clobbers, void *site, unsigned len);
259
260unsigned paravirt_patch_insns(void *site, unsigned len,
261			      const char *start, const char *end);
262
263
264/*
265 * This generates an indirect call based on the operation type number.
266 * The type number, computed in PARAVIRT_PATCH, is derived from the
267 * offset into the paravirt_ops structure, and can therefore be freely
268 * converted back into a structure offset.
269 */
270#define PARAVIRT_CALL	"call *(paravirt_ops+%c[paravirt_typenum]*4);"
271
272/*
273 * These macros are intended to wrap calls into a paravirt_ops
274 * operation, so that they can be later identified and patched at
275 * runtime.
276 *
277 * Normally, a call to a pv_op function is a simple indirect call:
278 * (paravirt_ops.operations)(args...).
279 *
280 * Unfortunately, this is a relatively slow operation for modern CPUs,
281 * because it cannot necessarily determine what the destination
282 * address is.  In this case, the address is a runtime constant, so at
283 * the very least we can patch the call to e a simple direct call, or
284 * ideally, patch an inline implementation into the callsite.  (Direct
285 * calls are essentially free, because the call and return addresses
286 * are completely predictable.)
287 *
288 * These macros rely on the standard gcc "regparm(3)" calling
289 * convention, in which the first three arguments are placed in %eax,
290 * %edx, %ecx (in that order), and the remaining arguments are placed
291 * on the stack.  All caller-save registers (eax,edx,ecx) are expected
292 * to be modified (either clobbered or used for return values).
293 *
294 * The call instruction itself is marked by placing its start address
295 * and size into the .parainstructions section, so that
296 * apply_paravirt() in arch/i386/kernel/alternative.c can do the
297 * appropriate patching under the control of the backend paravirt_ops
298 * implementation.
299 *
300 * Unfortunately there's no way to get gcc to generate the args setup
301 * for the call, and then allow the call itself to be generated by an
302 * inline asm.  Because of this, we must do the complete arg setup and
303 * return value handling from within these macros.  This is fairly
304 * cumbersome.
305 *
306 * There are 5 sets of PVOP_* macros for dealing with 0-4 arguments.
307 * It could be extended to more arguments, but there would be little
308 * to be gained from that.  For each number of arguments, there are
309 * the two VCALL and CALL variants for void and non-void functions.
310 *
311 * When there is a return value, the invoker of the macro must specify
312 * the return type.  The macro then uses sizeof() on that type to
313 * determine whether its a 32 or 64 bit value, and places the return
314 * in the right register(s) (just %eax for 32-bit, and %edx:%eax for
315 * 64-bit).
316 *
317 * 64-bit arguments are passed as a pair of adjacent 32-bit arguments
318 * in low,high order.
319 *
320 * Small structures are passed and returned in registers.  The macro
321 * calling convention can't directly deal with this, so the wrapper
322 * functions must do this.
323 *
324 * These PVOP_* macros are only defined within this header.  This
325 * means that all uses must be wrapped in inline functions.  This also
326 * makes sure the incoming and outgoing types are always correct.
327 */
328#define __PVOP_CALL(rettype, op, pre, post, ...)			\
329	({								\
330		rettype __ret;						\
331		unsigned long __eax, __edx, __ecx;			\
332		if (sizeof(rettype) > sizeof(unsigned long)) {		\
333			asm volatile(pre				\
334				     paravirt_alt(PARAVIRT_CALL)	\
335				     post				\
336				     : "=a" (__eax), "=d" (__edx),	\
337				       "=c" (__ecx)			\
338				     : paravirt_type(op),		\
339				       paravirt_clobber(CLBR_ANY),	\
340				       ##__VA_ARGS__			\
341				     : "memory", "cc");			\
342			__ret = (rettype)((((u64)__edx) << 32) | __eax); \
343		} else {						\
344			asm volatile(pre				\
345				     paravirt_alt(PARAVIRT_CALL)	\
346				     post				\
347				     : "=a" (__eax), "=d" (__edx),	\
348				       "=c" (__ecx)			\
349				     : paravirt_type(op),		\
350				       paravirt_clobber(CLBR_ANY),	\
351				       ##__VA_ARGS__			\
352				     : "memory", "cc");			\
353			__ret = (rettype)__eax;				\
354		}							\
355		__ret;							\
356	})
357#define __PVOP_VCALL(op, pre, post, ...)				\
358	({								\
359		unsigned long __eax, __edx, __ecx;			\
360		asm volatile(pre					\
361			     paravirt_alt(PARAVIRT_CALL)		\
362			     post					\
363			     : "=a" (__eax), "=d" (__edx), "=c" (__ecx) \
364			     : paravirt_type(op),			\
365			       paravirt_clobber(CLBR_ANY),		\
366			       ##__VA_ARGS__				\
367			     : "memory", "cc");				\
368	})
369
370#define PVOP_CALL0(rettype, op)						\
371	__PVOP_CALL(rettype, op, "", "")
372#define PVOP_VCALL0(op)							\
373	__PVOP_VCALL(op, "", "")
374
375#define PVOP_CALL1(rettype, op, arg1)					\
376	__PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)))
377#define PVOP_VCALL1(op, arg1)						\
378	__PVOP_VCALL(op, "", "", "0" ((u32)(arg1)))
379
380#define PVOP_CALL2(rettype, op, arg1, arg2)				\
381	__PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)), "1" ((u32)(arg2)))
382#define PVOP_VCALL2(op, arg1, arg2)					\
383	__PVOP_VCALL(op, "", "", "0" ((u32)(arg1)), "1" ((u32)(arg2)))
384
385#define PVOP_CALL3(rettype, op, arg1, arg2, arg3)			\
386	__PVOP_CALL(rettype, op, "", "", "0" ((u32)(arg1)),		\
387		    "1"((u32)(arg2)), "2"((u32)(arg3)))
388#define PVOP_VCALL3(op, arg1, arg2, arg3)				\
389	__PVOP_VCALL(op, "", "", "0" ((u32)(arg1)), "1"((u32)(arg2)),	\
390		     "2"((u32)(arg3)))
391
392#define PVOP_CALL4(rettype, op, arg1, arg2, arg3, arg4)			\
393	__PVOP_CALL(rettype, op,					\
394		    "push %[_arg4];", "lea 4(%%esp),%%esp;",		\
395		    "0" ((u32)(arg1)), "1" ((u32)(arg2)),		\
396		    "2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
397#define PVOP_VCALL4(op, arg1, arg2, arg3, arg4)				\
398	__PVOP_VCALL(op,						\
399		    "push %[_arg4];", "lea 4(%%esp),%%esp;",		\
400		    "0" ((u32)(arg1)), "1" ((u32)(arg2)),		\
401		    "2" ((u32)(arg3)), [_arg4] "mr" ((u32)(arg4)))
402
403static inline int paravirt_enabled(void)
404{
405	return paravirt_ops.paravirt_enabled;
406}
407
408static inline void load_esp0(struct tss_struct *tss,
409			     struct thread_struct *thread)
410{
411	PVOP_VCALL2(load_esp0, tss, thread);
412}
413
414#define ARCH_SETUP			paravirt_ops.arch_setup();
415static inline unsigned long get_wallclock(void)
416{
417	return PVOP_CALL0(unsigned long, get_wallclock);
418}
419
420static inline int set_wallclock(unsigned long nowtime)
421{
422	return PVOP_CALL1(int, set_wallclock, nowtime);
423}
424
425static inline void (*choose_time_init(void))(void)
426{
427	return paravirt_ops.time_init;
428}
429
430/* The paravirtualized CPUID instruction. */
431static inline void __cpuid(unsigned int *eax, unsigned int *ebx,
432			   unsigned int *ecx, unsigned int *edx)
433{
434	PVOP_VCALL4(cpuid, eax, ebx, ecx, edx);
435}
436
437/*
438 * These special macros can be used to get or set a debugging register
439 */
440static inline unsigned long paravirt_get_debugreg(int reg)
441{
442	return PVOP_CALL1(unsigned long, get_debugreg, reg);
443}
444#define get_debugreg(var, reg) var = paravirt_get_debugreg(reg)
445static inline void set_debugreg(unsigned long val, int reg)
446{
447	PVOP_VCALL2(set_debugreg, reg, val);
448}
449
450static inline void clts(void)
451{
452	PVOP_VCALL0(clts);
453}
454
455static inline unsigned long read_cr0(void)
456{
457	return PVOP_CALL0(unsigned long, read_cr0);
458}
459
460static inline void write_cr0(unsigned long x)
461{
462	PVOP_VCALL1(write_cr0, x);
463}
464
465static inline unsigned long read_cr2(void)
466{
467	return PVOP_CALL0(unsigned long, read_cr2);
468}
469
470static inline void write_cr2(unsigned long x)
471{
472	PVOP_VCALL1(write_cr2, x);
473}
474
475static inline unsigned long read_cr3(void)
476{
477	return PVOP_CALL0(unsigned long, read_cr3);
478}
479
480static inline void write_cr3(unsigned long x)
481{
482	PVOP_VCALL1(write_cr3, x);
483}
484
485static inline unsigned long read_cr4(void)
486{
487	return PVOP_CALL0(unsigned long, read_cr4);
488}
489static inline unsigned long read_cr4_safe(void)
490{
491	return PVOP_CALL0(unsigned long, read_cr4_safe);
492}
493
494static inline void write_cr4(unsigned long x)
495{
496	PVOP_VCALL1(write_cr4, x);
497}
498
499static inline void raw_safe_halt(void)
500{
501	PVOP_VCALL0(safe_halt);
502}
503
504static inline void halt(void)
505{
506	PVOP_VCALL0(safe_halt);
507}
508
509static inline void wbinvd(void)
510{
511	PVOP_VCALL0(wbinvd);
512}
513
514#define get_kernel_rpl()  (paravirt_ops.kernel_rpl)
515
516static inline u64 paravirt_read_msr(unsigned msr, int *err)
517{
518	return PVOP_CALL2(u64, read_msr, msr, err);
519}
520static inline int paravirt_write_msr(unsigned msr, unsigned low, unsigned high)
521{
522	return PVOP_CALL3(int, write_msr, msr, low, high);
523}
524
525/* These should all do BUG_ON(_err), but our headers are too tangled. */
526#define rdmsr(msr,val1,val2) do {		\
527	int _err;				\
528	u64 _l = paravirt_read_msr(msr, &_err);	\
529	val1 = (u32)_l;				\
530	val2 = _l >> 32;			\
531} while(0)
532
533#define wrmsr(msr,val1,val2) do {		\
534	paravirt_write_msr(msr, val1, val2);	\
535} while(0)
536
537#define rdmsrl(msr,val) do {			\
538	int _err;				\
539	val = paravirt_read_msr(msr, &_err);	\
540} while(0)
541
542#define wrmsrl(msr,val)		wrmsr(msr, (u32)((u64)(val)), ((u64)(val))>>32)
543#define wrmsr_safe(msr,a,b)	paravirt_write_msr(msr, a, b)
544
545/* rdmsr with exception handling */
546#define rdmsr_safe(msr,a,b) ({			\
547	int _err;				\
548	u64 _l = paravirt_read_msr(msr, &_err);	\
549	(*a) = (u32)_l;				\
550	(*b) = _l >> 32;			\
551	_err; })
552
553
554static inline u64 paravirt_read_tsc(void)
555{
556	return PVOP_CALL0(u64, read_tsc);
557}
558
559#define rdtscl(low) do {			\
560	u64 _l = paravirt_read_tsc();		\
561	low = (int)_l;				\
562} while(0)
563
564#define rdtscll(val) (val = paravirt_read_tsc())
565
566#define get_scheduled_cycles(val) (val = paravirt_ops.get_scheduled_cycles())
567#define calculate_cpu_khz() (paravirt_ops.get_cpu_khz())
568
569#define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
570
571static inline unsigned long long paravirt_read_pmc(int counter)
572{
573	return PVOP_CALL1(u64, read_pmc, counter);
574}
575
576#define rdpmc(counter,low,high) do {		\
577	u64 _l = paravirt_read_pmc(counter);	\
578	low = (u32)_l;				\
579	high = _l >> 32;			\
580} while(0)
581
582static inline void load_TR_desc(void)
583{
584	PVOP_VCALL0(load_tr_desc);
585}
586static inline void load_gdt(const struct Xgt_desc_struct *dtr)
587{
588	PVOP_VCALL1(load_gdt, dtr);
589}
590static inline void load_idt(const struct Xgt_desc_struct *dtr)
591{
592	PVOP_VCALL1(load_idt, dtr);
593}
594static inline void set_ldt(const void *addr, unsigned entries)
595{
596	PVOP_VCALL2(set_ldt, addr, entries);
597}
598static inline void store_gdt(struct Xgt_desc_struct *dtr)
599{
600	PVOP_VCALL1(store_gdt, dtr);
601}
602static inline void store_idt(struct Xgt_desc_struct *dtr)
603{
604	PVOP_VCALL1(store_idt, dtr);
605}
606static inline unsigned long paravirt_store_tr(void)
607{
608	return PVOP_CALL0(unsigned long, store_tr);
609}
610#define store_tr(tr)	((tr) = paravirt_store_tr())
611static inline void load_TLS(struct thread_struct *t, unsigned cpu)
612{
613	PVOP_VCALL2(load_tls, t, cpu);
614}
615static inline void write_ldt_entry(void *dt, int entry, u32 low, u32 high)
616{
617	PVOP_VCALL4(write_ldt_entry, dt, entry, low, high);
618}
619static inline void write_gdt_entry(void *dt, int entry, u32 low, u32 high)
620{
621	PVOP_VCALL4(write_gdt_entry, dt, entry, low, high);
622}
623static inline void write_idt_entry(void *dt, int entry, u32 low, u32 high)
624{
625	PVOP_VCALL4(write_idt_entry, dt, entry, low, high);
626}
627static inline void set_iopl_mask(unsigned mask)
628{
629	PVOP_VCALL1(set_iopl_mask, mask);
630}
631
632/* The paravirtualized I/O functions */
633static inline void slow_down_io(void) {
634	paravirt_ops.io_delay();
635#ifdef REALLY_SLOW_IO
636	paravirt_ops.io_delay();
637	paravirt_ops.io_delay();
638	paravirt_ops.io_delay();
639#endif
640}
641
642#ifdef CONFIG_X86_LOCAL_APIC
643/*
644 * Basic functions accessing APICs.
645 */
646static inline void apic_write(unsigned long reg, unsigned long v)
647{
648	PVOP_VCALL2(apic_write, reg, v);
649}
650
651static inline void apic_write_atomic(unsigned long reg, unsigned long v)
652{
653	PVOP_VCALL2(apic_write_atomic, reg, v);
654}
655
656static inline unsigned long apic_read(unsigned long reg)
657{
658	return PVOP_CALL1(unsigned long, apic_read, reg);
659}
660
661static inline void setup_boot_clock(void)
662{
663	PVOP_VCALL0(setup_boot_clock);
664}
665
666static inline void setup_secondary_clock(void)
667{
668	PVOP_VCALL0(setup_secondary_clock);
669}
670#endif
671
672static inline void paravirt_pagetable_setup_start(pgd_t *base)
673{
674	if (paravirt_ops.pagetable_setup_start)
675		(*paravirt_ops.pagetable_setup_start)(base);
676}
677
678static inline void paravirt_pagetable_setup_done(pgd_t *base)
679{
680	if (paravirt_ops.pagetable_setup_done)
681		(*paravirt_ops.pagetable_setup_done)(base);
682}
683
684#ifdef CONFIG_SMP
685static inline void startup_ipi_hook(int phys_apicid, unsigned long start_eip,
686				    unsigned long start_esp)
687{
688	PVOP_VCALL3(startup_ipi_hook, phys_apicid, start_eip, start_esp);
689}
690#endif
691
692static inline void paravirt_activate_mm(struct mm_struct *prev,
693					struct mm_struct *next)
694{
695	PVOP_VCALL2(activate_mm, prev, next);
696}
697
698static inline void arch_dup_mmap(struct mm_struct *oldmm,
699				 struct mm_struct *mm)
700{
701	PVOP_VCALL2(dup_mmap, oldmm, mm);
702}
703
704static inline void arch_exit_mmap(struct mm_struct *mm)
705{
706	PVOP_VCALL1(exit_mmap, mm);
707}
708
709static inline void __flush_tlb(void)
710{
711	PVOP_VCALL0(flush_tlb_user);
712}
713static inline void __flush_tlb_global(void)
714{
715	PVOP_VCALL0(flush_tlb_kernel);
716}
717static inline void __flush_tlb_single(unsigned long addr)
718{
719	PVOP_VCALL1(flush_tlb_single, addr);
720}
721
722static inline void flush_tlb_others(cpumask_t cpumask, struct mm_struct *mm,
723				    unsigned long va)
724{
725	PVOP_VCALL3(flush_tlb_others, &cpumask, mm, va);
726}
727
728static inline void paravirt_alloc_pt(unsigned pfn)
729{
730	PVOP_VCALL1(alloc_pt, pfn);
731}
732static inline void paravirt_release_pt(unsigned pfn)
733{
734	PVOP_VCALL1(release_pt, pfn);
735}
736
737static inline void paravirt_alloc_pd(unsigned pfn)
738{
739	PVOP_VCALL1(alloc_pd, pfn);
740}
741
742static inline void paravirt_alloc_pd_clone(unsigned pfn, unsigned clonepfn,
743					   unsigned start, unsigned count)
744{
745	PVOP_VCALL4(alloc_pd_clone, pfn, clonepfn, start, count);
746}
747static inline void paravirt_release_pd(unsigned pfn)
748{
749	PVOP_VCALL1(release_pd, pfn);
750}
751
752#ifdef CONFIG_HIGHPTE
753static inline void *kmap_atomic_pte(struct page *page, enum km_type type)
754{
755	unsigned long ret;
756	ret = PVOP_CALL2(unsigned long, kmap_atomic_pte, page, type);
757	return (void *)ret;
758}
759#endif
760
761static inline void pte_update(struct mm_struct *mm, unsigned long addr,
762			      pte_t *ptep)
763{
764	PVOP_VCALL3(pte_update, mm, addr, ptep);
765}
766
767static inline void pte_update_defer(struct mm_struct *mm, unsigned long addr,
768				    pte_t *ptep)
769{
770	PVOP_VCALL3(pte_update_defer, mm, addr, ptep);
771}
772
773#ifdef CONFIG_X86_PAE
774static inline pte_t __pte(unsigned long long val)
775{
776	unsigned long long ret = PVOP_CALL2(unsigned long long, make_pte,
777					    val, val >> 32);
778	return (pte_t) { ret, ret >> 32 };
779}
780
781static inline pmd_t __pmd(unsigned long long val)
782{
783	return (pmd_t) { PVOP_CALL2(unsigned long long, make_pmd, val, val >> 32) };
784}
785
786static inline pgd_t __pgd(unsigned long long val)
787{
788	return (pgd_t) { PVOP_CALL2(unsigned long long, make_pgd, val, val >> 32) };
789}
790
791static inline unsigned long long pte_val(pte_t x)
792{
793	return PVOP_CALL2(unsigned long long, pte_val, x.pte_low, x.pte_high);
794}
795
796static inline unsigned long long pmd_val(pmd_t x)
797{
798	return PVOP_CALL2(unsigned long long, pmd_val, x.pmd, x.pmd >> 32);
799}
800
801static inline unsigned long long pgd_val(pgd_t x)
802{
803	return PVOP_CALL2(unsigned long long, pgd_val, x.pgd, x.pgd >> 32);
804}
805
806static inline void set_pte(pte_t *ptep, pte_t pteval)
807{
808	PVOP_VCALL3(set_pte, ptep, pteval.pte_low, pteval.pte_high);
809}
810
811static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
812			      pte_t *ptep, pte_t pteval)
813{
814	/* 5 arg words */
815	paravirt_ops.set_pte_at(mm, addr, ptep, pteval);
816}
817
818static inline void set_pte_atomic(pte_t *ptep, pte_t pteval)
819{
820	PVOP_VCALL3(set_pte_atomic, ptep, pteval.pte_low, pteval.pte_high);
821}
822
823static inline void set_pte_present(struct mm_struct *mm, unsigned long addr,
824				   pte_t *ptep, pte_t pte)
825{
826	/* 5 arg words */
827	paravirt_ops.set_pte_present(mm, addr, ptep, pte);
828}
829
830static inline void set_pmd(pmd_t *pmdp, pmd_t pmdval)
831{
832	PVOP_VCALL3(set_pmd, pmdp, pmdval.pmd, pmdval.pmd >> 32);
833}
834
835static inline void set_pud(pud_t *pudp, pud_t pudval)
836{
837	PVOP_VCALL3(set_pud, pudp, pudval.pgd.pgd, pudval.pgd.pgd >> 32);
838}
839
840static inline void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
841{
842	PVOP_VCALL3(pte_clear, mm, addr, ptep);
843}
844
845static inline void pmd_clear(pmd_t *pmdp)
846{
847	PVOP_VCALL1(pmd_clear, pmdp);
848}
849
850#else  /* !CONFIG_X86_PAE */
851
852static inline pte_t __pte(unsigned long val)
853{
854	return (pte_t) { PVOP_CALL1(unsigned long, make_pte, val) };
855}
856
857static inline pgd_t __pgd(unsigned long val)
858{
859	return (pgd_t) { PVOP_CALL1(unsigned long, make_pgd, val) };
860}
861
862static inline unsigned long pte_val(pte_t x)
863{
864	return PVOP_CALL1(unsigned long, pte_val, x.pte_low);
865}
866
867static inline unsigned long pgd_val(pgd_t x)
868{
869	return PVOP_CALL1(unsigned long, pgd_val, x.pgd);
870}
871
872static inline void set_pte(pte_t *ptep, pte_t pteval)
873{
874	PVOP_VCALL2(set_pte, ptep, pteval.pte_low);
875}
876
877static inline void set_pte_at(struct mm_struct *mm, unsigned long addr,
878			      pte_t *ptep, pte_t pteval)
879{
880	PVOP_VCALL4(set_pte_at, mm, addr, ptep, pteval.pte_low);
881}
882
883static inline void set_pmd(pmd_t *pmdp, pmd_t pmdval)
884{
885	PVOP_VCALL2(set_pmd, pmdp, pmdval.pud.pgd.pgd);
886}
887#endif	/* CONFIG_X86_PAE */
888
889#define  __HAVE_ARCH_ENTER_LAZY_CPU_MODE
890static inline void arch_enter_lazy_cpu_mode(void)
891{
892	PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_CPU);
893}
894
895static inline void arch_leave_lazy_cpu_mode(void)
896{
897	PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_NONE);
898}
899
900static inline void arch_flush_lazy_cpu_mode(void)
901{
902	PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_FLUSH);
903}
904
905
906#define  __HAVE_ARCH_ENTER_LAZY_MMU_MODE
907static inline void arch_enter_lazy_mmu_mode(void)
908{
909	PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_MMU);
910}
911
912static inline void arch_leave_lazy_mmu_mode(void)
913{
914	PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_NONE);
915}
916
917static inline void arch_flush_lazy_mmu_mode(void)
918{
919	PVOP_VCALL1(set_lazy_mode, PARAVIRT_LAZY_FLUSH);
920}
921
922void _paravirt_nop(void);
923#define paravirt_nop	((void *)_paravirt_nop)
924
925/* These all sit in the .parainstructions section to tell us what to patch. */
926struct paravirt_patch_site {
927	u8 *instr; 		/* original instructions */
928	u8 instrtype;		/* type of this instruction */
929	u8 len;			/* length of original instruction */
930	u16 clobbers;		/* what registers you may clobber */
931};
932
933extern struct paravirt_patch_site __parainstructions[],
934	__parainstructions_end[];
935
936static inline unsigned long __raw_local_save_flags(void)
937{
938	unsigned long f;
939
940	asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
941				  PARAVIRT_CALL
942				  "popl %%edx; popl %%ecx")
943		     : "=a"(f)
944		     : paravirt_type(save_fl),
945		       paravirt_clobber(CLBR_EAX)
946		     : "memory", "cc");
947	return f;
948}
949
950static inline void raw_local_irq_restore(unsigned long f)
951{
952	asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
953				  PARAVIRT_CALL
954				  "popl %%edx; popl %%ecx")
955		     : "=a"(f)
956		     : "0"(f),
957		       paravirt_type(restore_fl),
958		       paravirt_clobber(CLBR_EAX)
959		     : "memory", "cc");
960}
961
962static inline void raw_local_irq_disable(void)
963{
964	asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
965				  PARAVIRT_CALL
966				  "popl %%edx; popl %%ecx")
967		     :
968		     : paravirt_type(irq_disable),
969		       paravirt_clobber(CLBR_EAX)
970		     : "memory", "eax", "cc");
971}
972
973static inline void raw_local_irq_enable(void)
974{
975	asm volatile(paravirt_alt("pushl %%ecx; pushl %%edx;"
976				  PARAVIRT_CALL
977				  "popl %%edx; popl %%ecx")
978		     :
979		     : paravirt_type(irq_enable),
980		       paravirt_clobber(CLBR_EAX)
981		     : "memory", "eax", "cc");
982}
983
984static inline unsigned long __raw_local_irq_save(void)
985{
986	unsigned long f;
987
988	f = __raw_local_save_flags();
989	raw_local_irq_disable();
990	return f;
991}
992
993#define CLI_STRING							\
994	_paravirt_alt("pushl %%ecx; pushl %%edx;"			\
995		      "call *paravirt_ops+%c[paravirt_cli_type]*4;"	\
996		      "popl %%edx; popl %%ecx",				\
997		      "%c[paravirt_cli_type]", "%c[paravirt_clobber]")
998
999#define STI_STRING							\
1000	_paravirt_alt("pushl %%ecx; pushl %%edx;"			\
1001		      "call *paravirt_ops+%c[paravirt_sti_type]*4;"	\
1002		      "popl %%edx; popl %%ecx",				\
1003		      "%c[paravirt_sti_type]", "%c[paravirt_clobber]")
1004
1005#define CLI_STI_CLOBBERS , "%eax"
1006#define CLI_STI_INPUT_ARGS						\
1007	,								\
1008	[paravirt_cli_type] "i" (PARAVIRT_PATCH(irq_disable)),		\
1009	[paravirt_sti_type] "i" (PARAVIRT_PATCH(irq_enable)),		\
1010	paravirt_clobber(CLBR_EAX)
1011
1012/* Make sure as little as possible of this mess escapes. */
1013#undef PARAVIRT_CALL
1014#undef __PVOP_CALL
1015#undef __PVOP_VCALL
1016#undef PVOP_VCALL0
1017#undef PVOP_CALL0
1018#undef PVOP_VCALL1
1019#undef PVOP_CALL1
1020#undef PVOP_VCALL2
1021#undef PVOP_CALL2
1022#undef PVOP_VCALL3
1023#undef PVOP_CALL3
1024#undef PVOP_VCALL4
1025#undef PVOP_CALL4
1026
1027#else  /* __ASSEMBLY__ */
1028
1029#define PARA_PATCH(off)	((off) / 4)
1030
1031#define PARA_SITE(ptype, clobbers, ops)		\
1032771:;						\
1033	ops;					\
1034772:;						\
1035	.pushsection .parainstructions,"a";	\
1036	 .long 771b;				\
1037	 .byte ptype;				\
1038	 .byte 772b-771b;			\
1039	 .short clobbers;			\
1040	.popsection
1041
1042#define INTERRUPT_RETURN					\
1043	PARA_SITE(PARA_PATCH(PARAVIRT_iret), CLBR_NONE,		\
1044		  jmp *%cs:paravirt_ops+PARAVIRT_iret)
1045
1046#define DISABLE_INTERRUPTS(clobbers)					\
1047	PARA_SITE(PARA_PATCH(PARAVIRT_irq_disable), clobbers,		\
1048		  pushl %eax; pushl %ecx; pushl %edx;			\
1049		  call *%cs:paravirt_ops+PARAVIRT_irq_disable;		\
1050		  popl %edx; popl %ecx; popl %eax)			\
1051
1052#define ENABLE_INTERRUPTS(clobbers)					\
1053	PARA_SITE(PARA_PATCH(PARAVIRT_irq_enable), clobbers,		\
1054		  pushl %eax; pushl %ecx; pushl %edx;			\
1055		  call *%cs:paravirt_ops+PARAVIRT_irq_enable;		\
1056		  popl %edx; popl %ecx; popl %eax)
1057
1058#define ENABLE_INTERRUPTS_SYSEXIT					\
1059	PARA_SITE(PARA_PATCH(PARAVIRT_irq_enable_sysexit), CLBR_NONE,	\
1060		  jmp *%cs:paravirt_ops+PARAVIRT_irq_enable_sysexit)
1061
1062#define GET_CR0_INTO_EAX			\
1063	push %ecx; push %edx;			\
1064	call *paravirt_ops+PARAVIRT_read_cr0;	\
1065	pop %edx; pop %ecx
1066
1067#endif /* __ASSEMBLY__ */
1068#endif /* CONFIG_PARAVIRT */
1069#endif	/* __ASM_PARAVIRT_H */
1070