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>
26 #include <linux/kprobes.h>
29 #include <asm/paravirt.h>
30 #include <asm/debugreg.h>
32 #include <asm/setup.h>
33 #include <asm/pgtable.h>
35 #include <asm/pgalloc.h>
37 #include <asm/delay.h>
38 #include <asm/fixmap.h>
40 #include <asm/tlbflush.h>
41 #include <asm/timer.h>
42 #include <asm/special_insns.h>
45 * nop stub, which must not clobber anything *including the stack* to
46 * avoid confusing the entry prologues.
48 extern void _paravirt_nop(void);
49 asm (".pushsection .entry.text, \"ax\"\n"
50 ".global _paravirt_nop\n"
53 ".size _paravirt_nop, . - _paravirt_nop\n\t"
54 ".type _paravirt_nop, @function\n\t"
57 /* identity function, which can be inlined */
58 u32
_paravirt_ident_32(u32 x
)
63 u64
_paravirt_ident_64(u64 x
)
68 void __init
default_banner(void)
70 printk(KERN_INFO
"Booting paravirtualized kernel on %s\n",
74 /* Undefined instruction for dealing with missing ops pointers. */
75 static const unsigned char ud2a
[] = { 0x0f, 0x0b };
80 } __attribute__((packed
));
82 unsigned paravirt_patch_call(void *insnbuf
,
83 const void *target
, u16 tgt_clobbers
,
84 unsigned long addr
, u16 site_clobbers
,
87 struct branch
*b
= insnbuf
;
88 unsigned long delta
= (unsigned long)target
- (addr
+5);
90 if (tgt_clobbers
& ~site_clobbers
)
91 return len
; /* target would clobber too much for this site */
93 return len
; /* call too long for patch site */
95 b
->opcode
= 0xe8; /* call */
97 BUILD_BUG_ON(sizeof(*b
) != 5);
102 unsigned paravirt_patch_jmp(void *insnbuf
, const void *target
,
103 unsigned long addr
, unsigned len
)
105 struct branch
*b
= insnbuf
;
106 unsigned long delta
= (unsigned long)target
- (addr
+5);
109 return len
; /* call too long for patch site */
111 b
->opcode
= 0xe9; /* jmp */
117 /* Neat trick to map patch type back to the call within the
118 * corresponding structure. */
119 static void *get_call_destination(u8 type
)
121 struct paravirt_patch_template tmpl
= {
122 .pv_init_ops
= pv_init_ops
,
123 .pv_time_ops
= pv_time_ops
,
124 .pv_cpu_ops
= pv_cpu_ops
,
125 .pv_irq_ops
= pv_irq_ops
,
126 .pv_apic_ops
= pv_apic_ops
,
127 .pv_mmu_ops
= pv_mmu_ops
,
128 #ifdef CONFIG_PARAVIRT_SPINLOCKS
129 .pv_lock_ops
= pv_lock_ops
,
132 return *((void **)&tmpl
+ type
);
135 unsigned paravirt_patch_default(u8 type
, u16 clobbers
, void *insnbuf
,
136 unsigned long addr
, unsigned len
)
138 void *opfunc
= get_call_destination(type
);
142 /* If there's no function, patch it with a ud2a (BUG) */
143 ret
= paravirt_patch_insns(insnbuf
, len
, ud2a
, ud2a
+sizeof(ud2a
));
144 else if (opfunc
== _paravirt_nop
)
147 /* identity functions just return their single argument */
148 else if (opfunc
== _paravirt_ident_32
)
149 ret
= paravirt_patch_ident_32(insnbuf
, len
);
150 else if (opfunc
== _paravirt_ident_64
)
151 ret
= paravirt_patch_ident_64(insnbuf
, len
);
153 else if (type
== PARAVIRT_PATCH(pv_cpu_ops
.iret
) ||
155 type
== PARAVIRT_PATCH(pv_cpu_ops
.irq_enable_sysexit
) ||
157 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret32
) ||
158 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret64
))
159 /* If operation requires a jmp, then jmp */
160 ret
= paravirt_patch_jmp(insnbuf
, opfunc
, addr
, len
);
162 /* Otherwise call the function; assume target could
163 clobber any caller-save reg */
164 ret
= paravirt_patch_call(insnbuf
, opfunc
, CLBR_ANY
,
165 addr
, clobbers
, len
);
170 unsigned paravirt_patch_insns(void *insnbuf
, unsigned len
,
171 const char *start
, const char *end
)
173 unsigned insn_len
= end
- start
;
175 if (insn_len
> len
|| start
== NULL
)
178 memcpy(insnbuf
, start
, insn_len
);
183 static void native_flush_tlb(void)
185 __native_flush_tlb();
189 * Global pages have to be flushed a bit differently. Not a real
190 * performance problem because this does not happen often.
192 static void native_flush_tlb_global(void)
194 __native_flush_tlb_global();
197 static void native_flush_tlb_single(unsigned long addr
)
199 __native_flush_tlb_single(addr
);
202 struct static_key paravirt_steal_enabled
;
203 struct static_key paravirt_steal_rq_enabled
;
205 static u64
native_steal_clock(int cpu
)
210 /* These are in entry.S */
211 extern void native_iret(void);
212 extern void native_irq_enable_sysexit(void);
213 extern void native_usergs_sysret32(void);
214 extern void native_usergs_sysret64(void);
216 static struct resource reserve_ioports
= {
218 .end
= IO_SPACE_LIMIT
,
219 .name
= "paravirt-ioport",
220 .flags
= IORESOURCE_IO
| IORESOURCE_BUSY
,
224 * Reserve the whole legacy IO space to prevent any legacy drivers
225 * from wasting time probing for their hardware. This is a fairly
226 * brute-force approach to disabling all non-virtual drivers.
228 * Note that this must be called very early to have any effect.
230 int paravirt_disable_iospace(void)
232 return request_resource(&ioport_resource
, &reserve_ioports
);
235 static DEFINE_PER_CPU(enum paravirt_lazy_mode
, paravirt_lazy_mode
) = PARAVIRT_LAZY_NONE
;
237 static inline void enter_lazy(enum paravirt_lazy_mode mode
)
239 BUG_ON(this_cpu_read(paravirt_lazy_mode
) != PARAVIRT_LAZY_NONE
);
241 this_cpu_write(paravirt_lazy_mode
, mode
);
244 static void leave_lazy(enum paravirt_lazy_mode mode
)
246 BUG_ON(this_cpu_read(paravirt_lazy_mode
) != mode
);
248 this_cpu_write(paravirt_lazy_mode
, PARAVIRT_LAZY_NONE
);
251 void paravirt_enter_lazy_mmu(void)
253 enter_lazy(PARAVIRT_LAZY_MMU
);
256 void paravirt_leave_lazy_mmu(void)
258 leave_lazy(PARAVIRT_LAZY_MMU
);
261 void paravirt_flush_lazy_mmu(void)
265 if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_MMU
) {
266 arch_leave_lazy_mmu_mode();
267 arch_enter_lazy_mmu_mode();
273 void paravirt_start_context_switch(struct task_struct
*prev
)
275 BUG_ON(preemptible());
277 if (this_cpu_read(paravirt_lazy_mode
) == PARAVIRT_LAZY_MMU
) {
278 arch_leave_lazy_mmu_mode();
279 set_ti_thread_flag(task_thread_info(prev
), TIF_LAZY_MMU_UPDATES
);
281 enter_lazy(PARAVIRT_LAZY_CPU
);
284 void paravirt_end_context_switch(struct task_struct
*next
)
286 BUG_ON(preemptible());
288 leave_lazy(PARAVIRT_LAZY_CPU
);
290 if (test_and_clear_ti_thread_flag(task_thread_info(next
), TIF_LAZY_MMU_UPDATES
))
291 arch_enter_lazy_mmu_mode();
294 enum paravirt_lazy_mode
paravirt_get_lazy_mode(void)
297 return PARAVIRT_LAZY_NONE
;
299 return this_cpu_read(paravirt_lazy_mode
);
302 struct pv_info pv_info
= {
303 .name
= "bare hardware",
304 .paravirt_enabled
= 0,
306 .shared_kernel_pmd
= 1, /* Only used when CONFIG_X86_PAE is set */
309 .extra_user_64bit_cs
= __USER_CS
,
313 struct pv_init_ops pv_init_ops
= {
314 .patch
= native_patch
,
317 struct pv_time_ops pv_time_ops
= {
318 .sched_clock
= native_sched_clock
,
319 .steal_clock
= native_steal_clock
,
322 __visible
struct pv_irq_ops pv_irq_ops
= {
323 .save_fl
= __PV_IS_CALLEE_SAVE(native_save_fl
),
324 .restore_fl
= __PV_IS_CALLEE_SAVE(native_restore_fl
),
325 .irq_disable
= __PV_IS_CALLEE_SAVE(native_irq_disable
),
326 .irq_enable
= __PV_IS_CALLEE_SAVE(native_irq_enable
),
327 .safe_halt
= native_safe_halt
,
330 .adjust_exception_frame
= paravirt_nop
,
334 __visible
struct pv_cpu_ops pv_cpu_ops
= {
335 .cpuid
= native_cpuid
,
336 .get_debugreg
= native_get_debugreg
,
337 .set_debugreg
= native_set_debugreg
,
339 .read_cr0
= native_read_cr0
,
340 .write_cr0
= native_write_cr0
,
341 .read_cr4
= native_read_cr4
,
342 .read_cr4_safe
= native_read_cr4_safe
,
343 .write_cr4
= native_write_cr4
,
345 .read_cr8
= native_read_cr8
,
346 .write_cr8
= native_write_cr8
,
348 .wbinvd
= native_wbinvd
,
349 .read_msr
= native_read_msr_safe
,
350 .write_msr
= native_write_msr_safe
,
351 .read_pmc
= native_read_pmc
,
352 .load_tr_desc
= native_load_tr_desc
,
353 .set_ldt
= native_set_ldt
,
354 .load_gdt
= native_load_gdt
,
355 .load_idt
= native_load_idt
,
356 .store_idt
= native_store_idt
,
357 .store_tr
= native_store_tr
,
358 .load_tls
= native_load_tls
,
360 .load_gs_index
= native_load_gs_index
,
362 .write_ldt_entry
= native_write_ldt_entry
,
363 .write_gdt_entry
= native_write_gdt_entry
,
364 .write_idt_entry
= native_write_idt_entry
,
366 .alloc_ldt
= paravirt_nop
,
367 .free_ldt
= paravirt_nop
,
369 .load_sp0
= native_load_sp0
,
371 #if defined(CONFIG_X86_32)
372 .irq_enable_sysexit
= native_irq_enable_sysexit
,
375 #ifdef CONFIG_IA32_EMULATION
376 .usergs_sysret32
= native_usergs_sysret32
,
378 .usergs_sysret64
= native_usergs_sysret64
,
381 .swapgs
= native_swapgs
,
383 .set_iopl_mask
= native_set_iopl_mask
,
384 .io_delay
= native_io_delay
,
386 .start_context_switch
= paravirt_nop
,
387 .end_context_switch
= paravirt_nop
,
390 /* At this point, native_get/set_debugreg has real function entries */
391 NOKPROBE_SYMBOL(native_get_debugreg
);
392 NOKPROBE_SYMBOL(native_set_debugreg
);
393 NOKPROBE_SYMBOL(native_load_idt
);
395 struct pv_apic_ops pv_apic_ops
= {
396 #ifdef CONFIG_X86_LOCAL_APIC
397 .startup_ipi_hook
= paravirt_nop
,
401 #if defined(CONFIG_X86_32) && !defined(CONFIG_X86_PAE)
402 /* 32-bit pagetable entries */
403 #define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_32)
405 /* 64-bit pagetable entries */
406 #define PTE_IDENT __PV_IS_CALLEE_SAVE(_paravirt_ident_64)
409 struct pv_mmu_ops pv_mmu_ops
= {
411 .read_cr2
= native_read_cr2
,
412 .write_cr2
= native_write_cr2
,
413 .read_cr3
= native_read_cr3
,
414 .write_cr3
= native_write_cr3
,
416 .flush_tlb_user
= native_flush_tlb
,
417 .flush_tlb_kernel
= native_flush_tlb_global
,
418 .flush_tlb_single
= native_flush_tlb_single
,
419 .flush_tlb_others
= native_flush_tlb_others
,
421 .pgd_alloc
= __paravirt_pgd_alloc
,
422 .pgd_free
= paravirt_nop
,
424 .alloc_pte
= paravirt_nop
,
425 .alloc_pmd
= paravirt_nop
,
426 .alloc_pud
= paravirt_nop
,
427 .release_pte
= paravirt_nop
,
428 .release_pmd
= paravirt_nop
,
429 .release_pud
= paravirt_nop
,
431 .set_pte
= native_set_pte
,
432 .set_pte_at
= native_set_pte_at
,
433 .set_pmd
= native_set_pmd
,
434 .set_pmd_at
= native_set_pmd_at
,
435 .pte_update
= paravirt_nop
,
436 .pte_update_defer
= paravirt_nop
,
437 .pmd_update
= paravirt_nop
,
438 .pmd_update_defer
= paravirt_nop
,
440 .ptep_modify_prot_start
= __ptep_modify_prot_start
,
441 .ptep_modify_prot_commit
= __ptep_modify_prot_commit
,
443 #if CONFIG_PGTABLE_LEVELS >= 3
444 #ifdef CONFIG_X86_PAE
445 .set_pte_atomic
= native_set_pte_atomic
,
446 .pte_clear
= native_pte_clear
,
447 .pmd_clear
= native_pmd_clear
,
449 .set_pud
= native_set_pud
,
451 .pmd_val
= PTE_IDENT
,
452 .make_pmd
= PTE_IDENT
,
454 #if CONFIG_PGTABLE_LEVELS == 4
455 .pud_val
= PTE_IDENT
,
456 .make_pud
= PTE_IDENT
,
458 .set_pgd
= native_set_pgd
,
460 #endif /* CONFIG_PGTABLE_LEVELS >= 3 */
462 .pte_val
= PTE_IDENT
,
463 .pgd_val
= PTE_IDENT
,
465 .make_pte
= PTE_IDENT
,
466 .make_pgd
= PTE_IDENT
,
468 .dup_mmap
= paravirt_nop
,
469 .exit_mmap
= paravirt_nop
,
470 .activate_mm
= paravirt_nop
,
473 .enter
= paravirt_nop
,
474 .leave
= paravirt_nop
,
475 .flush
= paravirt_nop
,
478 .set_fixmap
= native_set_fixmap
,
481 EXPORT_SYMBOL_GPL(pv_time_ops
);
482 EXPORT_SYMBOL (pv_cpu_ops
);
483 EXPORT_SYMBOL (pv_mmu_ops
);
484 EXPORT_SYMBOL_GPL(pv_apic_ops
);
485 EXPORT_SYMBOL_GPL(pv_info
);
486 EXPORT_SYMBOL (pv_irq_ops
);