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>
30 #include <asm/setup.h>
31 #include <asm/arch_hooks.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 static void __init
default_banner(void)
60 printk(KERN_INFO
"Booting paravirtualized kernel on %s\n",
64 char *memory_setup(void)
66 return pv_init_ops
.memory_setup();
69 /* Simple instruction patching code. */
70 #define DEF_NATIVE(ops, name, code) \
71 extern const char start_##ops##_##name[], end_##ops##_##name[]; \
72 asm("start_" #ops "_" #name ": " code "; end_" #ops "_" #name ":")
74 /* Undefined instruction for dealing with missing ops pointers. */
75 static const unsigned char ud2a
[] = { 0x0f, 0x0b };
77 unsigned paravirt_patch_nop(void)
82 unsigned paravirt_patch_ignore(unsigned len
)
90 } __attribute__((packed
));
92 unsigned paravirt_patch_call(void *insnbuf
,
93 const void *target
, u16 tgt_clobbers
,
94 unsigned long addr
, u16 site_clobbers
,
97 struct branch
*b
= insnbuf
;
98 unsigned long delta
= (unsigned long)target
- (addr
+5);
100 if (tgt_clobbers
& ~site_clobbers
)
101 return len
; /* target would clobber too much for this site */
103 return len
; /* call too long for patch site */
105 b
->opcode
= 0xe8; /* call */
107 BUILD_BUG_ON(sizeof(*b
) != 5);
112 unsigned paravirt_patch_jmp(void *insnbuf
, const void *target
,
113 unsigned long addr
, unsigned len
)
115 struct branch
*b
= insnbuf
;
116 unsigned long delta
= (unsigned long)target
- (addr
+5);
119 return len
; /* call too long for patch site */
121 b
->opcode
= 0xe9; /* jmp */
127 /* Neat trick to map patch type back to the call within the
128 * corresponding structure. */
129 static void *get_call_destination(u8 type
)
131 struct paravirt_patch_template tmpl
= {
132 .pv_init_ops
= pv_init_ops
,
133 .pv_time_ops
= pv_time_ops
,
134 .pv_cpu_ops
= pv_cpu_ops
,
135 .pv_irq_ops
= pv_irq_ops
,
136 .pv_apic_ops
= pv_apic_ops
,
137 .pv_mmu_ops
= pv_mmu_ops
,
138 .pv_lock_ops
= pv_lock_ops
,
140 return *((void **)&tmpl
+ type
);
143 unsigned paravirt_patch_default(u8 type
, u16 clobbers
, void *insnbuf
,
144 unsigned long addr
, unsigned len
)
146 void *opfunc
= get_call_destination(type
);
150 /* If there's no function, patch it with a ud2a (BUG) */
151 ret
= paravirt_patch_insns(insnbuf
, len
, ud2a
, ud2a
+sizeof(ud2a
));
152 else if (opfunc
== _paravirt_nop
)
153 /* If the operation is a nop, then nop the callsite */
154 ret
= paravirt_patch_nop();
156 /* identity functions just return their single argument */
157 else if (opfunc
== _paravirt_ident_32
)
158 ret
= paravirt_patch_ident_32(insnbuf
, len
);
159 else if (opfunc
== _paravirt_ident_64
)
160 ret
= paravirt_patch_ident_64(insnbuf
, len
);
162 else if (type
== PARAVIRT_PATCH(pv_cpu_ops
.iret
) ||
163 type
== PARAVIRT_PATCH(pv_cpu_ops
.irq_enable_sysexit
) ||
164 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret32
) ||
165 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret64
))
166 /* If operation requires a jmp, then jmp */
167 ret
= paravirt_patch_jmp(insnbuf
, opfunc
, addr
, len
);
169 /* Otherwise call the function; assume target could
170 clobber any caller-save reg */
171 ret
= paravirt_patch_call(insnbuf
, opfunc
, CLBR_ANY
,
172 addr
, clobbers
, len
);
177 unsigned paravirt_patch_insns(void *insnbuf
, unsigned len
,
178 const char *start
, const char *end
)
180 unsigned insn_len
= end
- start
;
182 if (insn_len
> len
|| start
== NULL
)
185 memcpy(insnbuf
, start
, insn_len
);
192 pv_irq_ops
.init_IRQ();
195 static void native_flush_tlb(void)
197 __native_flush_tlb();
201 * Global pages have to be flushed a bit differently. Not a real
202 * performance problem because this does not happen often.
204 static void native_flush_tlb_global(void)
206 __native_flush_tlb_global();
209 static void native_flush_tlb_single(unsigned long addr
)
211 __native_flush_tlb_single(addr
);
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 int __init
print_banner(void)
222 pv_init_ops
.banner();
225 core_initcall(print_banner
);
227 static struct resource reserve_ioports
= {
229 .end
= IO_SPACE_LIMIT
,
230 .name
= "paravirt-ioport",
231 .flags
= IORESOURCE_IO
| IORESOURCE_BUSY
,
235 * Reserve the whole legacy IO space to prevent any legacy drivers
236 * from wasting time probing for their hardware. This is a fairly
237 * brute-force approach to disabling all non-virtual drivers.
239 * Note that this must be called very early to have any effect.
241 int paravirt_disable_iospace(void)
243 return request_resource(&ioport_resource
, &reserve_ioports
);
246 static DEFINE_PER_CPU(enum paravirt_lazy_mode
, paravirt_lazy_mode
) = PARAVIRT_LAZY_NONE
;
248 static inline void enter_lazy(enum paravirt_lazy_mode mode
)
250 BUG_ON(__get_cpu_var(paravirt_lazy_mode
) != PARAVIRT_LAZY_NONE
);
251 BUG_ON(preemptible());
253 __get_cpu_var(paravirt_lazy_mode
) = mode
;
256 void paravirt_leave_lazy(enum paravirt_lazy_mode mode
)
258 BUG_ON(__get_cpu_var(paravirt_lazy_mode
) != mode
);
259 BUG_ON(preemptible());
261 __get_cpu_var(paravirt_lazy_mode
) = PARAVIRT_LAZY_NONE
;
264 void paravirt_enter_lazy_mmu(void)
266 enter_lazy(PARAVIRT_LAZY_MMU
);
269 void paravirt_leave_lazy_mmu(void)
271 paravirt_leave_lazy(PARAVIRT_LAZY_MMU
);
274 void paravirt_enter_lazy_cpu(void)
276 enter_lazy(PARAVIRT_LAZY_CPU
);
279 void paravirt_leave_lazy_cpu(void)
281 paravirt_leave_lazy(PARAVIRT_LAZY_CPU
);
284 enum paravirt_lazy_mode
paravirt_get_lazy_mode(void)
286 return __get_cpu_var(paravirt_lazy_mode
);
289 struct pv_info pv_info
= {
290 .name
= "bare hardware",
291 .paravirt_enabled
= 0,
293 .shared_kernel_pmd
= 1, /* Only used when CONFIG_X86_PAE is set */
296 struct pv_init_ops pv_init_ops
= {
297 .patch
= native_patch
,
298 .banner
= default_banner
,
299 .arch_setup
= paravirt_nop
,
300 .memory_setup
= machine_specific_memory_setup
,
303 struct pv_time_ops pv_time_ops
= {
304 .time_init
= hpet_time_init
,
305 .get_wallclock
= native_get_wallclock
,
306 .set_wallclock
= native_set_wallclock
,
307 .sched_clock
= native_sched_clock
,
308 .get_tsc_khz
= native_calibrate_tsc
,
311 struct pv_irq_ops pv_irq_ops
= {
312 .init_IRQ
= native_init_IRQ
,
313 .save_fl
= native_save_fl
,
314 .restore_fl
= native_restore_fl
,
315 .irq_disable
= native_irq_disable
,
316 .irq_enable
= native_irq_enable
,
317 .safe_halt
= native_safe_halt
,
320 .adjust_exception_frame
= paravirt_nop
,
324 struct pv_cpu_ops pv_cpu_ops
= {
325 .cpuid
= native_cpuid
,
326 .get_debugreg
= native_get_debugreg
,
327 .set_debugreg
= native_set_debugreg
,
329 .read_cr0
= native_read_cr0
,
330 .write_cr0
= native_write_cr0
,
331 .read_cr4
= native_read_cr4
,
332 .read_cr4_safe
= native_read_cr4_safe
,
333 .write_cr4
= native_write_cr4
,
335 .read_cr8
= native_read_cr8
,
336 .write_cr8
= native_write_cr8
,
338 .wbinvd
= native_wbinvd
,
339 .read_msr
= native_read_msr_safe
,
340 .read_msr_amd
= native_read_msr_amd_safe
,
341 .write_msr
= native_write_msr_safe
,
342 .read_tsc
= native_read_tsc
,
343 .read_pmc
= native_read_pmc
,
344 .read_tscp
= native_read_tscp
,
345 .load_tr_desc
= native_load_tr_desc
,
346 .set_ldt
= native_set_ldt
,
347 .load_gdt
= native_load_gdt
,
348 .load_idt
= native_load_idt
,
349 .store_gdt
= native_store_gdt
,
350 .store_idt
= native_store_idt
,
351 .store_tr
= native_store_tr
,
352 .load_tls
= native_load_tls
,
354 .load_gs_index
= native_load_gs_index
,
356 .write_ldt_entry
= native_write_ldt_entry
,
357 .write_gdt_entry
= native_write_gdt_entry
,
358 .write_idt_entry
= native_write_idt_entry
,
360 .alloc_ldt
= paravirt_nop
,
361 .free_ldt
= paravirt_nop
,
363 .load_sp0
= native_load_sp0
,
365 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
366 .irq_enable_sysexit
= native_irq_enable_sysexit
,
369 #ifdef CONFIG_IA32_EMULATION
370 .usergs_sysret32
= native_usergs_sysret32
,
372 .usergs_sysret64
= native_usergs_sysret64
,
375 .swapgs
= native_swapgs
,
377 .set_iopl_mask
= native_set_iopl_mask
,
378 .io_delay
= native_io_delay
,
381 .enter
= paravirt_nop
,
382 .leave
= paravirt_nop
,
386 struct pv_apic_ops pv_apic_ops
= {
387 #ifdef CONFIG_X86_LOCAL_APIC
388 .setup_boot_clock
= setup_boot_APIC_clock
,
389 .setup_secondary_clock
= setup_secondary_APIC_clock
,
390 .startup_ipi_hook
= paravirt_nop
,
394 typedef pte_t
make_pte_t(pteval_t
);
395 typedef pmd_t
make_pmd_t(pmdval_t
);
396 typedef pud_t
make_pud_t(pudval_t
);
397 typedef pgd_t
make_pgd_t(pgdval_t
);
399 typedef pteval_t
pte_val_t(pte_t
);
400 typedef pmdval_t
pmd_val_t(pmd_t
);
401 typedef pudval_t
pud_val_t(pud_t
);
402 typedef pgdval_t
pgd_val_t(pgd_t
);
405 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
406 /* 32-bit pagetable entries */
407 #define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_32
408 #define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_32
410 #define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_32
411 #define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_32
413 #define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_32
414 #define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_32
416 #define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_32
417 #define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_32
419 /* 64-bit pagetable entries */
420 #define paravirt_native_make_pte (make_pte_t *)_paravirt_ident_64
421 #define paravirt_native_pte_val (pte_val_t *)_paravirt_ident_64
423 #define paravirt_native_make_pmd (make_pmd_t *)_paravirt_ident_64
424 #define paravirt_native_pmd_val (pmd_val_t *)_paravirt_ident_64
426 #define paravirt_native_make_pud (make_pud_t *)_paravirt_ident_64
427 #define paravirt_native_pud_val (pud_val_t *)_paravirt_ident_64
429 #define paravirt_native_make_pgd (make_pgd_t *)_paravirt_ident_64
430 #define paravirt_native_pgd_val (pgd_val_t *)_paravirt_ident_64
433 struct pv_mmu_ops pv_mmu_ops
= {
434 #ifndef CONFIG_X86_64
435 .pagetable_setup_start
= native_pagetable_setup_start
,
436 .pagetable_setup_done
= native_pagetable_setup_done
,
438 .pagetable_setup_start
= paravirt_nop
,
439 .pagetable_setup_done
= paravirt_nop
,
442 .read_cr2
= native_read_cr2
,
443 .write_cr2
= native_write_cr2
,
444 .read_cr3
= native_read_cr3
,
445 .write_cr3
= native_write_cr3
,
447 .flush_tlb_user
= native_flush_tlb
,
448 .flush_tlb_kernel
= native_flush_tlb_global
,
449 .flush_tlb_single
= native_flush_tlb_single
,
450 .flush_tlb_others
= native_flush_tlb_others
,
452 .pgd_alloc
= __paravirt_pgd_alloc
,
453 .pgd_free
= paravirt_nop
,
455 .alloc_pte
= paravirt_nop
,
456 .alloc_pmd
= paravirt_nop
,
457 .alloc_pmd_clone
= paravirt_nop
,
458 .alloc_pud
= paravirt_nop
,
459 .release_pte
= paravirt_nop
,
460 .release_pmd
= paravirt_nop
,
461 .release_pud
= paravirt_nop
,
463 .set_pte
= native_set_pte
,
464 .set_pte_at
= native_set_pte_at
,
465 .set_pmd
= native_set_pmd
,
466 .pte_update
= paravirt_nop
,
467 .pte_update_defer
= paravirt_nop
,
469 .ptep_modify_prot_start
= __ptep_modify_prot_start
,
470 .ptep_modify_prot_commit
= __ptep_modify_prot_commit
,
472 #ifdef CONFIG_HIGHPTE
473 .kmap_atomic_pte
= kmap_atomic
,
476 #if PAGETABLE_LEVELS >= 3
477 #ifdef CONFIG_X86_PAE
478 .set_pte_atomic
= native_set_pte_atomic
,
479 .set_pte_present
= native_set_pte_present
,
480 .pte_clear
= native_pte_clear
,
481 .pmd_clear
= native_pmd_clear
,
483 .set_pud
= native_set_pud
,
484 .pmd_val
= paravirt_native_pmd_val
,
485 .make_pmd
= paravirt_native_make_pmd
,
487 #if PAGETABLE_LEVELS == 4
488 .pud_val
= paravirt_native_pud_val
,
489 .make_pud
= paravirt_native_make_pud
,
490 .set_pgd
= native_set_pgd
,
492 #endif /* PAGETABLE_LEVELS >= 3 */
494 .pte_val
= paravirt_native_pte_val
,
495 .pgd_val
= paravirt_native_pgd_val
,
497 .make_pte
= paravirt_native_make_pte
,
498 .make_pgd
= paravirt_native_make_pgd
,
500 .dup_mmap
= paravirt_nop
,
501 .exit_mmap
= paravirt_nop
,
502 .activate_mm
= paravirt_nop
,
505 .enter
= paravirt_nop
,
506 .leave
= paravirt_nop
,
509 .set_fixmap
= native_set_fixmap
,
512 EXPORT_SYMBOL_GPL(pv_time_ops
);
513 EXPORT_SYMBOL (pv_cpu_ops
);
514 EXPORT_SYMBOL (pv_mmu_ops
);
515 EXPORT_SYMBOL_GPL(pv_apic_ops
);
516 EXPORT_SYMBOL_GPL(pv_info
);
517 EXPORT_SYMBOL (pv_irq_ops
);