1 /* Paravirtualization interfaces
2 Copyright (C) 2006 Rusty Russell IBM Corporation
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software
16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
18 2007 - x86_64 support added by Glauber de Oliveira Costa, Red Hat Inc
21 #include <linux/errno.h>
22 #include <linux/module.h>
23 #include <linux/efi.h>
24 #include <linux/bcd.h>
25 #include <linux/highmem.h>
28 #include <asm/paravirt.h>
29 #include <asm/debugreg.h>
31 #include <asm/setup.h>
32 #include <asm/pgtable.h>
34 #include <asm/pgalloc.h>
36 #include <asm/delay.h>
37 #include <asm/fixmap.h>
39 #include <asm/tlbflush.h>
40 #include <asm/timer.h>
43 void _paravirt_nop(void)
47 /* identity function, which can be inlined */
48 u32
_paravirt_ident_32(u32 x
)
53 u64
_paravirt_ident_64(u64 x
)
58 void __init
default_banner(void)
60 printk(KERN_INFO
"Booting paravirtualized kernel on %s\n",
64 /* Simple instruction patching code. */
65 #define DEF_NATIVE(ops, name, code) \
66 extern const char start_##ops##_##name[], end_##ops##_##name[]; \
67 asm("start_" #ops "_" #name ": " code "; end_" #ops "_" #name ":")
69 /* Undefined instruction for dealing with missing ops pointers. */
70 static const unsigned char ud2a
[] = { 0x0f, 0x0b };
72 unsigned paravirt_patch_nop(void)
77 unsigned paravirt_patch_ignore(unsigned len
)
85 } __attribute__((packed
));
87 unsigned paravirt_patch_call(void *insnbuf
,
88 const void *target
, u16 tgt_clobbers
,
89 unsigned long addr
, u16 site_clobbers
,
92 struct branch
*b
= insnbuf
;
93 unsigned long delta
= (unsigned long)target
- (addr
+5);
95 if (tgt_clobbers
& ~site_clobbers
)
96 return len
; /* target would clobber too much for this site */
98 return len
; /* call too long for patch site */
100 b
->opcode
= 0xe8; /* call */
102 BUILD_BUG_ON(sizeof(*b
) != 5);
107 unsigned paravirt_patch_jmp(void *insnbuf
, const void *target
,
108 unsigned long addr
, unsigned len
)
110 struct branch
*b
= insnbuf
;
111 unsigned long delta
= (unsigned long)target
- (addr
+5);
114 return len
; /* call too long for patch site */
116 b
->opcode
= 0xe9; /* jmp */
122 /* Neat trick to map patch type back to the call within the
123 * corresponding structure. */
124 static void *get_call_destination(u8 type
)
126 struct paravirt_patch_template tmpl
= {
127 .pv_init_ops
= pv_init_ops
,
128 .pv_time_ops
= pv_time_ops
,
129 .pv_cpu_ops
= pv_cpu_ops
,
130 .pv_irq_ops
= pv_irq_ops
,
131 .pv_apic_ops
= pv_apic_ops
,
132 .pv_mmu_ops
= pv_mmu_ops
,
133 #ifdef CONFIG_PARAVIRT_SPINLOCKS
134 .pv_lock_ops
= pv_lock_ops
,
137 return *((void **)&tmpl
+ type
);
140 unsigned paravirt_patch_default(u8 type
, u16 clobbers
, void *insnbuf
,
141 unsigned long addr
, unsigned len
)
143 void *opfunc
= get_call_destination(type
);
147 /* If there's no function, patch it with a ud2a (BUG) */
148 ret
= paravirt_patch_insns(insnbuf
, len
, ud2a
, ud2a
+sizeof(ud2a
));
149 else if (opfunc
== _paravirt_nop
)
150 /* If the operation is a nop, then nop the callsite */
151 ret
= paravirt_patch_nop();
153 /* identity functions just return their single argument */
154 else if (opfunc
== _paravirt_ident_32
)
155 ret
= paravirt_patch_ident_32(insnbuf
, len
);
156 else if (opfunc
== _paravirt_ident_64
)
157 ret
= paravirt_patch_ident_64(insnbuf
, len
);
159 else if (type
== PARAVIRT_PATCH(pv_cpu_ops
.iret
) ||
160 type
== PARAVIRT_PATCH(pv_cpu_ops
.irq_enable_sysexit
) ||
161 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret32
) ||
162 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret64
))
163 /* If operation requires a jmp, then jmp */
164 ret
= paravirt_patch_jmp(insnbuf
, opfunc
, addr
, len
);
166 /* Otherwise call the function; assume target could
167 clobber any caller-save reg */
168 ret
= paravirt_patch_call(insnbuf
, opfunc
, CLBR_ANY
,
169 addr
, clobbers
, len
);
174 unsigned paravirt_patch_insns(void *insnbuf
, unsigned len
,
175 const char *start
, const char *end
)
177 unsigned insn_len
= end
- start
;
179 if (insn_len
> len
|| start
== NULL
)
182 memcpy(insnbuf
, start
, insn_len
);
187 static void native_flush_tlb(void)
189 __native_flush_tlb();
193 * Global pages have to be flushed a bit differently. Not a real
194 * performance problem because this does not happen often.
196 static void native_flush_tlb_global(void)
198 __native_flush_tlb_global();
201 static void native_flush_tlb_single(unsigned long addr
)
203 __native_flush_tlb_single(addr
);
206 struct jump_label_key paravirt_steal_enabled
;
207 struct jump_label_key paravirt_steal_rq_enabled
;
209 static u64
native_steal_clock(int cpu
)
214 /* These are in entry.S */
215 extern void native_iret(void);
216 extern void native_irq_enable_sysexit(void);
217 extern void native_usergs_sysret32(void);
218 extern void native_usergs_sysret64(void);
220 static struct resource reserve_ioports
= {
222 .end
= IO_SPACE_LIMIT
,
223 .name
= "paravirt-ioport",
224 .flags
= IORESOURCE_IO
| IORESOURCE_BUSY
,
228 * Reserve the whole legacy IO space to prevent any legacy drivers
229 * from wasting time probing for their hardware. This is a fairly
230 * brute-force approach to disabling all non-virtual drivers.
232 * Note that this must be called very early to have any effect.
234 int paravirt_disable_iospace(void)
236 return request_resource(&ioport_resource
, &reserve_ioports
);
239 static DEFINE_PER_CPU(enum paravirt_lazy_mode
, paravirt_lazy_mode
) = PARAVIRT_LAZY_NONE
;
241 static inline void enter_lazy(enum paravirt_lazy_mode mode
)
243 BUG_ON(percpu_read(paravirt_lazy_mode
) != PARAVIRT_LAZY_NONE
);
245 percpu_write(paravirt_lazy_mode
, mode
);
248 static void leave_lazy(enum paravirt_lazy_mode mode
)
250 BUG_ON(percpu_read(paravirt_lazy_mode
) != mode
);
252 percpu_write(paravirt_lazy_mode
, PARAVIRT_LAZY_NONE
);
255 void paravirt_enter_lazy_mmu(void)
257 enter_lazy(PARAVIRT_LAZY_MMU
);
260 void paravirt_leave_lazy_mmu(void)
262 leave_lazy(PARAVIRT_LAZY_MMU
);
265 void paravirt_start_context_switch(struct task_struct
*prev
)
267 BUG_ON(preemptible());
269 if (percpu_read(paravirt_lazy_mode
) == PARAVIRT_LAZY_MMU
) {
270 arch_leave_lazy_mmu_mode();
271 set_ti_thread_flag(task_thread_info(prev
), TIF_LAZY_MMU_UPDATES
);
273 enter_lazy(PARAVIRT_LAZY_CPU
);
276 void paravirt_end_context_switch(struct task_struct
*next
)
278 BUG_ON(preemptible());
280 leave_lazy(PARAVIRT_LAZY_CPU
);
282 if (test_and_clear_ti_thread_flag(task_thread_info(next
), TIF_LAZY_MMU_UPDATES
))
283 arch_enter_lazy_mmu_mode();
286 enum paravirt_lazy_mode
paravirt_get_lazy_mode(void)
289 return PARAVIRT_LAZY_NONE
;
291 return percpu_read(paravirt_lazy_mode
);
294 void arch_flush_lazy_mmu_mode(void)
298 if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_MMU
) {
299 arch_leave_lazy_mmu_mode();
300 arch_enter_lazy_mmu_mode();
306 struct pv_info pv_info
= {
307 .name
= "bare hardware",
308 .paravirt_enabled
= 0,
310 .shared_kernel_pmd
= 1, /* Only used when CONFIG_X86_PAE is set */
313 .extra_user_64bit_cs
= __USER_CS
,
317 struct pv_init_ops pv_init_ops
= {
318 .patch
= native_patch
,
321 struct pv_time_ops pv_time_ops
= {
322 .sched_clock
= native_sched_clock
,
323 .steal_clock
= native_steal_clock
,
326 struct pv_irq_ops pv_irq_ops
= {
327 .save_fl
= __PV_IS_CALLEE_SAVE(native_save_fl
),
328 .restore_fl
= __PV_IS_CALLEE_SAVE(native_restore_fl
),
329 .irq_disable
= __PV_IS_CALLEE_SAVE(native_irq_disable
),
330 .irq_enable
= __PV_IS_CALLEE_SAVE(native_irq_enable
),
331 .safe_halt
= native_safe_halt
,
334 .adjust_exception_frame
= paravirt_nop
,
338 struct pv_cpu_ops pv_cpu_ops
= {
339 .cpuid
= native_cpuid
,
340 .get_debugreg
= native_get_debugreg
,
341 .set_debugreg
= native_set_debugreg
,
343 .read_cr0
= native_read_cr0
,
344 .write_cr0
= native_write_cr0
,
345 .read_cr4
= native_read_cr4
,
346 .read_cr4_safe
= native_read_cr4_safe
,
347 .write_cr4
= native_write_cr4
,
349 .read_cr8
= native_read_cr8
,
350 .write_cr8
= native_write_cr8
,
352 .wbinvd
= native_wbinvd
,
353 .read_msr
= native_read_msr_safe
,
354 .rdmsr_regs
= native_rdmsr_safe_regs
,
355 .write_msr
= native_write_msr_safe
,
356 .wrmsr_regs
= native_wrmsr_safe_regs
,
357 .read_tsc
= native_read_tsc
,
358 .read_pmc
= native_read_pmc
,
359 .read_tscp
= native_read_tscp
,
360 .load_tr_desc
= native_load_tr_desc
,
361 .set_ldt
= native_set_ldt
,
362 .load_gdt
= native_load_gdt
,
363 .load_idt
= native_load_idt
,
364 .store_gdt
= native_store_gdt
,
365 .store_idt
= native_store_idt
,
366 .store_tr
= native_store_tr
,
367 .load_tls
= native_load_tls
,
369 .load_gs_index
= native_load_gs_index
,
371 .write_ldt_entry
= native_write_ldt_entry
,
372 .write_gdt_entry
= native_write_gdt_entry
,
373 .write_idt_entry
= native_write_idt_entry
,
375 .alloc_ldt
= paravirt_nop
,
376 .free_ldt
= paravirt_nop
,
378 .load_sp0
= native_load_sp0
,
380 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
381 .irq_enable_sysexit
= native_irq_enable_sysexit
,
384 #ifdef CONFIG_IA32_EMULATION
385 .usergs_sysret32
= native_usergs_sysret32
,
387 .usergs_sysret64
= native_usergs_sysret64
,
390 .swapgs
= native_swapgs
,
392 .set_iopl_mask
= native_set_iopl_mask
,
393 .io_delay
= native_io_delay
,
395 .start_context_switch
= paravirt_nop
,
396 .end_context_switch
= paravirt_nop
,
399 struct pv_apic_ops pv_apic_ops
= {
400 #ifdef CONFIG_X86_LOCAL_APIC
401 .startup_ipi_hook
= paravirt_nop
,
405 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
406 /* 32-bit pagetable entries */
407 #define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_32)
409 /* 64-bit pagetable entries */
410 #define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_64)
413 struct pv_mmu_ops pv_mmu_ops
= {
415 .read_cr2
= native_read_cr2
,
416 .write_cr2
= native_write_cr2
,
417 .read_cr3
= native_read_cr3
,
418 .write_cr3
= native_write_cr3
,
420 .flush_tlb_user
= native_flush_tlb
,
421 .flush_tlb_kernel
= native_flush_tlb_global
,
422 .flush_tlb_single
= native_flush_tlb_single
,
423 .flush_tlb_others
= native_flush_tlb_others
,
425 .pgd_alloc
= __paravirt_pgd_alloc
,
426 .pgd_free
= paravirt_nop
,
428 .alloc_pte
= paravirt_nop
,
429 .alloc_pmd
= paravirt_nop
,
430 .alloc_pud
= paravirt_nop
,
431 .release_pte
= paravirt_nop
,
432 .release_pmd
= paravirt_nop
,
433 .release_pud
= paravirt_nop
,
435 .set_pte
= native_set_pte
,
436 .set_pte_at
= native_set_pte_at
,
437 .set_pmd
= native_set_pmd
,
438 .set_pmd_at
= native_set_pmd_at
,
439 .pte_update
= paravirt_nop
,
440 .pte_update_defer
= paravirt_nop
,
441 .pmd_update
= paravirt_nop
,
442 .pmd_update_defer
= paravirt_nop
,
444 .ptep_modify_prot_start
= __ptep_modify_prot_start
,
445 .ptep_modify_prot_commit
= __ptep_modify_prot_commit
,
447 #if PAGETABLE_LEVELS >= 3
448 #ifdef CONFIG_X86_PAE
449 .set_pte_atomic
= native_set_pte_atomic
,
450 .pte_clear
= native_pte_clear
,
451 .pmd_clear
= native_pmd_clear
,
453 .set_pud
= native_set_pud
,
455 .pmd_val
= PTE_IDENT
,
456 .make_pmd
= PTE_IDENT
,
458 #if PAGETABLE_LEVELS == 4
459 .pud_val
= PTE_IDENT
,
460 .make_pud
= PTE_IDENT
,
462 .set_pgd
= native_set_pgd
,
464 #endif /* PAGETABLE_LEVELS >= 3 */
466 .pte_val
= PTE_IDENT
,
467 .pgd_val
= PTE_IDENT
,
469 .make_pte
= PTE_IDENT
,
470 .make_pgd
= PTE_IDENT
,
472 .dup_mmap
= paravirt_nop
,
473 .exit_mmap
= paravirt_nop
,
474 .activate_mm
= paravirt_nop
,
477 .enter
= paravirt_nop
,
478 .leave
= paravirt_nop
,
481 .set_fixmap
= native_set_fixmap
,
484 EXPORT_SYMBOL_GPL(pv_time_ops
);
485 EXPORT_SYMBOL (pv_cpu_ops
);
486 EXPORT_SYMBOL (pv_mmu_ops
);
487 EXPORT_SYMBOL_GPL(pv_apic_ops
);
488 EXPORT_SYMBOL_GPL(pv_info
);
489 EXPORT_SYMBOL (pv_irq_ops
);