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/pgtable.h>
33 #include <asm/pgalloc.h>
35 #include <asm/delay.h>
36 #include <asm/fixmap.h>
38 #include <asm/tlbflush.h>
39 #include <asm/timer.h>
42 void _paravirt_nop(void)
46 /* identity function, which can be inlined */
47 u32
_paravirt_ident_32(u32 x
)
52 u64
_paravirt_ident_64(u64 x
)
57 void __init
default_banner(void)
59 printk(KERN_INFO
"Booting paravirtualized kernel on %s\n",
63 /* Simple instruction patching code. */
64 #define DEF_NATIVE(ops, name, code) \
65 extern const char start_##ops##_##name[], end_##ops##_##name[]; \
66 asm("start_" #ops "_" #name ": " code "; end_" #ops "_" #name ":")
68 /* Undefined instruction for dealing with missing ops pointers. */
69 static const unsigned char ud2a
[] = { 0x0f, 0x0b };
71 unsigned paravirt_patch_nop(void)
76 unsigned paravirt_patch_ignore(unsigned len
)
84 } __attribute__((packed
));
86 unsigned paravirt_patch_call(void *insnbuf
,
87 const void *target
, u16 tgt_clobbers
,
88 unsigned long addr
, u16 site_clobbers
,
91 struct branch
*b
= insnbuf
;
92 unsigned long delta
= (unsigned long)target
- (addr
+5);
94 if (tgt_clobbers
& ~site_clobbers
)
95 return len
; /* target would clobber too much for this site */
97 return len
; /* call too long for patch site */
99 b
->opcode
= 0xe8; /* call */
101 BUILD_BUG_ON(sizeof(*b
) != 5);
106 unsigned paravirt_patch_jmp(void *insnbuf
, const void *target
,
107 unsigned long addr
, unsigned len
)
109 struct branch
*b
= insnbuf
;
110 unsigned long delta
= (unsigned long)target
- (addr
+5);
113 return len
; /* call too long for patch site */
115 b
->opcode
= 0xe9; /* jmp */
121 /* Neat trick to map patch type back to the call within the
122 * corresponding structure. */
123 static void *get_call_destination(u8 type
)
125 struct paravirt_patch_template tmpl
= {
126 .pv_init_ops
= pv_init_ops
,
127 .pv_time_ops
= pv_time_ops
,
128 .pv_cpu_ops
= pv_cpu_ops
,
129 .pv_irq_ops
= pv_irq_ops
,
130 .pv_apic_ops
= pv_apic_ops
,
131 .pv_mmu_ops
= pv_mmu_ops
,
132 #ifdef CONFIG_PARAVIRT_SPINLOCKS
133 .pv_lock_ops
= pv_lock_ops
,
136 return *((void **)&tmpl
+ type
);
139 unsigned paravirt_patch_default(u8 type
, u16 clobbers
, void *insnbuf
,
140 unsigned long addr
, unsigned len
)
142 void *opfunc
= get_call_destination(type
);
146 /* If there's no function, patch it with a ud2a (BUG) */
147 ret
= paravirt_patch_insns(insnbuf
, len
, ud2a
, ud2a
+sizeof(ud2a
));
148 else if (opfunc
== _paravirt_nop
)
149 /* If the operation is a nop, then nop the callsite */
150 ret
= paravirt_patch_nop();
152 /* identity functions just return their single argument */
153 else if (opfunc
== _paravirt_ident_32
)
154 ret
= paravirt_patch_ident_32(insnbuf
, len
);
155 else if (opfunc
== _paravirt_ident_64
)
156 ret
= paravirt_patch_ident_64(insnbuf
, len
);
158 else if (type
== PARAVIRT_PATCH(pv_cpu_ops
.iret
) ||
159 type
== PARAVIRT_PATCH(pv_cpu_ops
.irq_enable_sysexit
) ||
160 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret32
) ||
161 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret64
))
162 /* If operation requires a jmp, then jmp */
163 ret
= paravirt_patch_jmp(insnbuf
, opfunc
, addr
, len
);
165 /* Otherwise call the function; assume target could
166 clobber any caller-save reg */
167 ret
= paravirt_patch_call(insnbuf
, opfunc
, CLBR_ANY
,
168 addr
, clobbers
, len
);
173 unsigned paravirt_patch_insns(void *insnbuf
, unsigned len
,
174 const char *start
, const char *end
)
176 unsigned insn_len
= end
- start
;
178 if (insn_len
> len
|| start
== NULL
)
181 memcpy(insnbuf
, start
, insn_len
);
186 static void native_flush_tlb(void)
188 __native_flush_tlb();
192 * Global pages have to be flushed a bit differently. Not a real
193 * performance problem because this does not happen often.
195 static void native_flush_tlb_global(void)
197 __native_flush_tlb_global();
200 static void native_flush_tlb_single(unsigned long addr
)
202 __native_flush_tlb_single(addr
);
205 struct jump_label_key paravirt_steal_enabled
;
206 struct jump_label_key paravirt_steal_rq_enabled
;
208 static u64
native_steal_clock(int cpu
)
213 /* These are in entry.S */
214 extern void native_iret(void);
215 extern void native_irq_enable_sysexit(void);
216 extern void native_usergs_sysret32(void);
217 extern void native_usergs_sysret64(void);
219 static struct resource reserve_ioports
= {
221 .end
= IO_SPACE_LIMIT
,
222 .name
= "paravirt-ioport",
223 .flags
= IORESOURCE_IO
| IORESOURCE_BUSY
,
227 * Reserve the whole legacy IO space to prevent any legacy drivers
228 * from wasting time probing for their hardware. This is a fairly
229 * brute-force approach to disabling all non-virtual drivers.
231 * Note that this must be called very early to have any effect.
233 int paravirt_disable_iospace(void)
235 return request_resource(&ioport_resource
, &reserve_ioports
);
238 static DEFINE_PER_CPU(enum paravirt_lazy_mode
, paravirt_lazy_mode
) = PARAVIRT_LAZY_NONE
;
240 static inline void enter_lazy(enum paravirt_lazy_mode mode
)
242 BUG_ON(percpu_read(paravirt_lazy_mode
) != PARAVIRT_LAZY_NONE
);
244 percpu_write(paravirt_lazy_mode
, mode
);
247 static void leave_lazy(enum paravirt_lazy_mode mode
)
249 BUG_ON(percpu_read(paravirt_lazy_mode
) != mode
);
251 percpu_write(paravirt_lazy_mode
, PARAVIRT_LAZY_NONE
);
254 void paravirt_enter_lazy_mmu(void)
256 enter_lazy(PARAVIRT_LAZY_MMU
);
259 void paravirt_leave_lazy_mmu(void)
261 leave_lazy(PARAVIRT_LAZY_MMU
);
264 void paravirt_start_context_switch(struct task_struct
*prev
)
266 BUG_ON(preemptible());
268 if (percpu_read(paravirt_lazy_mode
) == PARAVIRT_LAZY_MMU
) {
269 arch_leave_lazy_mmu_mode();
270 set_ti_thread_flag(task_thread_info(prev
), TIF_LAZY_MMU_UPDATES
);
272 enter_lazy(PARAVIRT_LAZY_CPU
);
275 void paravirt_end_context_switch(struct task_struct
*next
)
277 BUG_ON(preemptible());
279 leave_lazy(PARAVIRT_LAZY_CPU
);
281 if (test_and_clear_ti_thread_flag(task_thread_info(next
), TIF_LAZY_MMU_UPDATES
))
282 arch_enter_lazy_mmu_mode();
285 enum paravirt_lazy_mode
paravirt_get_lazy_mode(void)
288 return PARAVIRT_LAZY_NONE
;
290 return percpu_read(paravirt_lazy_mode
);
293 void arch_flush_lazy_mmu_mode(void)
297 if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_MMU
) {
298 arch_leave_lazy_mmu_mode();
299 arch_enter_lazy_mmu_mode();
305 struct pv_info pv_info
= {
306 .name
= "bare hardware",
307 .paravirt_enabled
= 0,
309 .shared_kernel_pmd
= 1, /* Only used when CONFIG_X86_PAE is set */
312 .extra_user_64bit_cs
= __USER_CS
,
316 struct pv_init_ops pv_init_ops
= {
317 .patch
= native_patch
,
320 struct pv_time_ops pv_time_ops
= {
321 .sched_clock
= native_sched_clock
,
322 .steal_clock
= native_steal_clock
,
325 struct pv_irq_ops pv_irq_ops
= {
326 .save_fl
= __PV_IS_CALLEE_SAVE(native_save_fl
),
327 .restore_fl
= __PV_IS_CALLEE_SAVE(native_restore_fl
),
328 .irq_disable
= __PV_IS_CALLEE_SAVE(native_irq_disable
),
329 .irq_enable
= __PV_IS_CALLEE_SAVE(native_irq_enable
),
330 .safe_halt
= native_safe_halt
,
333 .adjust_exception_frame
= paravirt_nop
,
337 struct pv_cpu_ops pv_cpu_ops
= {
338 .cpuid
= native_cpuid
,
339 .get_debugreg
= native_get_debugreg
,
340 .set_debugreg
= native_set_debugreg
,
342 .read_cr0
= native_read_cr0
,
343 .write_cr0
= native_write_cr0
,
344 .read_cr4
= native_read_cr4
,
345 .read_cr4_safe
= native_read_cr4_safe
,
346 .write_cr4
= native_write_cr4
,
348 .read_cr8
= native_read_cr8
,
349 .write_cr8
= native_write_cr8
,
351 .wbinvd
= native_wbinvd
,
352 .read_msr
= native_read_msr_safe
,
353 .rdmsr_regs
= native_rdmsr_safe_regs
,
354 .write_msr
= native_write_msr_safe
,
355 .wrmsr_regs
= native_wrmsr_safe_regs
,
356 .read_tsc
= native_read_tsc
,
357 .read_pmc
= native_read_pmc
,
358 .read_tscp
= native_read_tscp
,
359 .load_tr_desc
= native_load_tr_desc
,
360 .set_ldt
= native_set_ldt
,
361 .load_gdt
= native_load_gdt
,
362 .load_idt
= native_load_idt
,
363 .store_gdt
= native_store_gdt
,
364 .store_idt
= native_store_idt
,
365 .store_tr
= native_store_tr
,
366 .load_tls
= native_load_tls
,
368 .load_gs_index
= native_load_gs_index
,
370 .write_ldt_entry
= native_write_ldt_entry
,
371 .write_gdt_entry
= native_write_gdt_entry
,
372 .write_idt_entry
= native_write_idt_entry
,
374 .alloc_ldt
= paravirt_nop
,
375 .free_ldt
= paravirt_nop
,
377 .load_sp0
= native_load_sp0
,
379 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
380 .irq_enable_sysexit
= native_irq_enable_sysexit
,
383 #ifdef CONFIG_IA32_EMULATION
384 .usergs_sysret32
= native_usergs_sysret32
,
386 .usergs_sysret64
= native_usergs_sysret64
,
389 .swapgs
= native_swapgs
,
391 .set_iopl_mask
= native_set_iopl_mask
,
392 .io_delay
= native_io_delay
,
394 .start_context_switch
= paravirt_nop
,
395 .end_context_switch
= paravirt_nop
,
398 struct pv_apic_ops pv_apic_ops
= {
399 #ifdef CONFIG_X86_LOCAL_APIC
400 .startup_ipi_hook
= paravirt_nop
,
404 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
405 /* 32-bit pagetable entries */
406 #define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_32)
408 /* 64-bit pagetable entries */
409 #define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_64)
412 struct pv_mmu_ops pv_mmu_ops
= {
414 .read_cr2
= native_read_cr2
,
415 .write_cr2
= native_write_cr2
,
416 .read_cr3
= native_read_cr3
,
417 .write_cr3
= native_write_cr3
,
419 .flush_tlb_user
= native_flush_tlb
,
420 .flush_tlb_kernel
= native_flush_tlb_global
,
421 .flush_tlb_single
= native_flush_tlb_single
,
422 .flush_tlb_others
= native_flush_tlb_others
,
424 .pgd_alloc
= __paravirt_pgd_alloc
,
425 .pgd_free
= paravirt_nop
,
427 .alloc_pte
= paravirt_nop
,
428 .alloc_pmd
= paravirt_nop
,
429 .alloc_pud
= paravirt_nop
,
430 .release_pte
= paravirt_nop
,
431 .release_pmd
= paravirt_nop
,
432 .release_pud
= paravirt_nop
,
434 .set_pte
= native_set_pte
,
435 .set_pte_at
= native_set_pte_at
,
436 .set_pmd
= native_set_pmd
,
437 .set_pmd_at
= native_set_pmd_at
,
438 .pte_update
= paravirt_nop
,
439 .pte_update_defer
= paravirt_nop
,
440 .pmd_update
= paravirt_nop
,
441 .pmd_update_defer
= paravirt_nop
,
443 .ptep_modify_prot_start
= __ptep_modify_prot_start
,
444 .ptep_modify_prot_commit
= __ptep_modify_prot_commit
,
446 #if PAGETABLE_LEVELS >= 3
447 #ifdef CONFIG_X86_PAE
448 .set_pte_atomic
= native_set_pte_atomic
,
449 .pte_clear
= native_pte_clear
,
450 .pmd_clear
= native_pmd_clear
,
452 .set_pud
= native_set_pud
,
454 .pmd_val
= PTE_IDENT
,
455 .make_pmd
= PTE_IDENT
,
457 #if PAGETABLE_LEVELS == 4
458 .pud_val
= PTE_IDENT
,
459 .make_pud
= PTE_IDENT
,
461 .set_pgd
= native_set_pgd
,
463 #endif /* PAGETABLE_LEVELS >= 3 */
465 .pte_val
= PTE_IDENT
,
466 .pgd_val
= PTE_IDENT
,
468 .make_pte
= PTE_IDENT
,
469 .make_pgd
= PTE_IDENT
,
471 .dup_mmap
= paravirt_nop
,
472 .exit_mmap
= paravirt_nop
,
473 .activate_mm
= paravirt_nop
,
476 .enter
= paravirt_nop
,
477 .leave
= paravirt_nop
,
480 .set_fixmap
= native_set_fixmap
,
483 EXPORT_SYMBOL_GPL(pv_time_ops
);
484 EXPORT_SYMBOL (pv_cpu_ops
);
485 EXPORT_SYMBOL (pv_mmu_ops
);
486 EXPORT_SYMBOL_GPL(pv_apic_ops
);
487 EXPORT_SYMBOL_GPL(pv_info
);
488 EXPORT_SYMBOL (pv_irq_ops
);