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 static void __init
default_banner(void)
49 printk(KERN_INFO
"Booting paravirtualized kernel on %s\n",
53 char *memory_setup(void)
55 return pv_init_ops
.memory_setup();
58 /* Simple instruction patching code. */
59 #define DEF_NATIVE(ops, name, code) \
60 extern const char start_##ops##_##name[], end_##ops##_##name[]; \
61 asm("start_" #ops "_" #name ": " code "; end_" #ops "_" #name ":")
63 /* Undefined instruction for dealing with missing ops pointers. */
64 static const unsigned char ud2a
[] = { 0x0f, 0x0b };
66 unsigned paravirt_patch_nop(void)
71 unsigned paravirt_patch_ignore(unsigned len
)
79 } __attribute__((packed
));
81 unsigned paravirt_patch_call(void *insnbuf
,
82 const void *target
, u16 tgt_clobbers
,
83 unsigned long addr
, u16 site_clobbers
,
86 struct branch
*b
= insnbuf
;
87 unsigned long delta
= (unsigned long)target
- (addr
+5);
89 if (tgt_clobbers
& ~site_clobbers
)
90 return len
; /* target would clobber too much for this site */
92 return len
; /* call too long for patch site */
94 b
->opcode
= 0xe8; /* call */
96 BUILD_BUG_ON(sizeof(*b
) != 5);
101 unsigned paravirt_patch_jmp(void *insnbuf
, const void *target
,
102 unsigned long addr
, unsigned len
)
104 struct branch
*b
= insnbuf
;
105 unsigned long delta
= (unsigned long)target
- (addr
+5);
108 return len
; /* call too long for patch site */
110 b
->opcode
= 0xe9; /* jmp */
116 /* Neat trick to map patch type back to the call within the
117 * corresponding structure. */
118 static void *get_call_destination(u8 type
)
120 struct paravirt_patch_template tmpl
= {
121 .pv_init_ops
= pv_init_ops
,
122 .pv_time_ops
= pv_time_ops
,
123 .pv_cpu_ops
= pv_cpu_ops
,
124 .pv_irq_ops
= pv_irq_ops
,
125 .pv_apic_ops
= pv_apic_ops
,
126 .pv_mmu_ops
= pv_mmu_ops
,
127 .pv_lock_ops
= pv_lock_ops
,
129 return *((void **)&tmpl
+ type
);
132 unsigned paravirt_patch_default(u8 type
, u16 clobbers
, void *insnbuf
,
133 unsigned long addr
, unsigned len
)
135 void *opfunc
= get_call_destination(type
);
139 /* If there's no function, patch it with a ud2a (BUG) */
140 ret
= paravirt_patch_insns(insnbuf
, len
, ud2a
, ud2a
+sizeof(ud2a
));
141 else if (opfunc
== paravirt_nop
)
142 /* If the operation is a nop, then nop the callsite */
143 ret
= paravirt_patch_nop();
144 else if (type
== PARAVIRT_PATCH(pv_cpu_ops
.iret
) ||
145 type
== PARAVIRT_PATCH(pv_cpu_ops
.irq_enable_sysexit
) ||
146 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret32
) ||
147 type
== PARAVIRT_PATCH(pv_cpu_ops
.usergs_sysret64
))
148 /* If operation requires a jmp, then jmp */
149 ret
= paravirt_patch_jmp(insnbuf
, opfunc
, addr
, len
);
151 /* Otherwise call the function; assume target could
152 clobber any caller-save reg */
153 ret
= paravirt_patch_call(insnbuf
, opfunc
, CLBR_ANY
,
154 addr
, clobbers
, len
);
159 unsigned paravirt_patch_insns(void *insnbuf
, unsigned len
,
160 const char *start
, const char *end
)
162 unsigned insn_len
= end
- start
;
164 if (insn_len
> len
|| start
== NULL
)
167 memcpy(insnbuf
, start
, insn_len
);
174 pv_irq_ops
.init_IRQ();
177 static void native_flush_tlb(void)
179 __native_flush_tlb();
183 * Global pages have to be flushed a bit differently. Not a real
184 * performance problem because this does not happen often.
186 static void native_flush_tlb_global(void)
188 __native_flush_tlb_global();
191 static void native_flush_tlb_single(unsigned long addr
)
193 __native_flush_tlb_single(addr
);
196 /* These are in entry.S */
197 extern void native_iret(void);
198 extern void native_irq_enable_sysexit(void);
199 extern void native_usergs_sysret32(void);
200 extern void native_usergs_sysret64(void);
202 static int __init
print_banner(void)
204 pv_init_ops
.banner();
207 core_initcall(print_banner
);
209 static struct resource reserve_ioports
= {
211 .end
= IO_SPACE_LIMIT
,
212 .name
= "paravirt-ioport",
213 .flags
= IORESOURCE_IO
| IORESOURCE_BUSY
,
217 * Reserve the whole legacy IO space to prevent any legacy drivers
218 * from wasting time probing for their hardware. This is a fairly
219 * brute-force approach to disabling all non-virtual drivers.
221 * Note that this must be called very early to have any effect.
223 int paravirt_disable_iospace(void)
225 return request_resource(&ioport_resource
, &reserve_ioports
);
228 static DEFINE_PER_CPU(enum paravirt_lazy_mode
, paravirt_lazy_mode
) = PARAVIRT_LAZY_NONE
;
230 static inline void enter_lazy(enum paravirt_lazy_mode mode
)
232 BUG_ON(__get_cpu_var(paravirt_lazy_mode
) != PARAVIRT_LAZY_NONE
);
233 BUG_ON(preemptible());
235 __get_cpu_var(paravirt_lazy_mode
) = mode
;
238 void paravirt_leave_lazy(enum paravirt_lazy_mode mode
)
240 BUG_ON(__get_cpu_var(paravirt_lazy_mode
) != mode
);
241 BUG_ON(preemptible());
243 __get_cpu_var(paravirt_lazy_mode
) = PARAVIRT_LAZY_NONE
;
246 void paravirt_enter_lazy_mmu(void)
248 enter_lazy(PARAVIRT_LAZY_MMU
);
251 void paravirt_leave_lazy_mmu(void)
253 paravirt_leave_lazy(PARAVIRT_LAZY_MMU
);
256 void paravirt_enter_lazy_cpu(void)
258 enter_lazy(PARAVIRT_LAZY_CPU
);
261 void paravirt_leave_lazy_cpu(void)
263 paravirt_leave_lazy(PARAVIRT_LAZY_CPU
);
266 enum paravirt_lazy_mode
paravirt_get_lazy_mode(void)
268 return __get_cpu_var(paravirt_lazy_mode
);
271 void __init
paravirt_use_bytelocks(void)
274 pv_lock_ops
.spin_is_locked
= __byte_spin_is_locked
;
275 pv_lock_ops
.spin_is_contended
= __byte_spin_is_contended
;
276 pv_lock_ops
.spin_lock
= __byte_spin_lock
;
277 pv_lock_ops
.spin_trylock
= __byte_spin_trylock
;
278 pv_lock_ops
.spin_unlock
= __byte_spin_unlock
;
282 struct pv_info pv_info
= {
283 .name
= "bare hardware",
284 .paravirt_enabled
= 0,
286 .shared_kernel_pmd
= 1, /* Only used when CONFIG_X86_PAE is set */
289 struct pv_init_ops pv_init_ops
= {
290 .patch
= native_patch
,
291 .banner
= default_banner
,
292 .arch_setup
= paravirt_nop
,
293 .memory_setup
= machine_specific_memory_setup
,
296 struct pv_time_ops pv_time_ops
= {
297 .time_init
= hpet_time_init
,
298 .get_wallclock
= native_get_wallclock
,
299 .set_wallclock
= native_set_wallclock
,
300 .sched_clock
= native_sched_clock
,
301 .get_tsc_khz
= native_calibrate_tsc
,
304 struct pv_irq_ops pv_irq_ops
= {
305 .init_IRQ
= native_init_IRQ
,
306 .save_fl
= native_save_fl
,
307 .restore_fl
= native_restore_fl
,
308 .irq_disable
= native_irq_disable
,
309 .irq_enable
= native_irq_enable
,
310 .safe_halt
= native_safe_halt
,
313 .adjust_exception_frame
= paravirt_nop
,
317 struct pv_cpu_ops pv_cpu_ops
= {
318 .cpuid
= native_cpuid
,
319 .get_debugreg
= native_get_debugreg
,
320 .set_debugreg
= native_set_debugreg
,
322 .read_cr0
= native_read_cr0
,
323 .write_cr0
= native_write_cr0
,
324 .read_cr4
= native_read_cr4
,
325 .read_cr4_safe
= native_read_cr4_safe
,
326 .write_cr4
= native_write_cr4
,
328 .read_cr8
= native_read_cr8
,
329 .write_cr8
= native_write_cr8
,
331 .wbinvd
= native_wbinvd
,
332 .read_msr
= native_read_msr_safe
,
333 .write_msr
= native_write_msr_safe
,
334 .read_tsc
= native_read_tsc
,
335 .read_pmc
= native_read_pmc
,
336 .read_tscp
= native_read_tscp
,
337 .load_tr_desc
= native_load_tr_desc
,
338 .set_ldt
= native_set_ldt
,
339 .load_gdt
= native_load_gdt
,
340 .load_idt
= native_load_idt
,
341 .store_gdt
= native_store_gdt
,
342 .store_idt
= native_store_idt
,
343 .store_tr
= native_store_tr
,
344 .load_tls
= native_load_tls
,
346 .load_gs_index
= native_load_gs_index
,
348 .write_ldt_entry
= native_write_ldt_entry
,
349 .write_gdt_entry
= native_write_gdt_entry
,
350 .write_idt_entry
= native_write_idt_entry
,
351 .load_sp0
= native_load_sp0
,
353 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
354 .irq_enable_sysexit
= native_irq_enable_sysexit
,
357 #ifdef CONFIG_IA32_EMULATION
358 .usergs_sysret32
= native_usergs_sysret32
,
360 .usergs_sysret64
= native_usergs_sysret64
,
363 .swapgs
= native_swapgs
,
365 .set_iopl_mask
= native_set_iopl_mask
,
366 .io_delay
= native_io_delay
,
369 .enter
= paravirt_nop
,
370 .leave
= paravirt_nop
,
374 struct pv_apic_ops pv_apic_ops
= {
375 #ifdef CONFIG_X86_LOCAL_APIC
376 .setup_boot_clock
= setup_boot_APIC_clock
,
377 .setup_secondary_clock
= setup_secondary_APIC_clock
,
378 .startup_ipi_hook
= paravirt_nop
,
382 struct pv_mmu_ops pv_mmu_ops
= {
383 #ifndef CONFIG_X86_64
384 .pagetable_setup_start
= native_pagetable_setup_start
,
385 .pagetable_setup_done
= native_pagetable_setup_done
,
387 .pagetable_setup_start
= paravirt_nop
,
388 .pagetable_setup_done
= paravirt_nop
,
391 .read_cr2
= native_read_cr2
,
392 .write_cr2
= native_write_cr2
,
393 .read_cr3
= native_read_cr3
,
394 .write_cr3
= native_write_cr3
,
396 .flush_tlb_user
= native_flush_tlb
,
397 .flush_tlb_kernel
= native_flush_tlb_global
,
398 .flush_tlb_single
= native_flush_tlb_single
,
399 .flush_tlb_others
= native_flush_tlb_others
,
401 .pgd_alloc
= __paravirt_pgd_alloc
,
402 .pgd_free
= paravirt_nop
,
404 .alloc_pte
= paravirt_nop
,
405 .alloc_pmd
= paravirt_nop
,
406 .alloc_pmd_clone
= paravirt_nop
,
407 .alloc_pud
= paravirt_nop
,
408 .release_pte
= paravirt_nop
,
409 .release_pmd
= paravirt_nop
,
410 .release_pud
= paravirt_nop
,
412 .set_pte
= native_set_pte
,
413 .set_pte_at
= native_set_pte_at
,
414 .set_pmd
= native_set_pmd
,
415 .pte_update
= paravirt_nop
,
416 .pte_update_defer
= paravirt_nop
,
418 .ptep_modify_prot_start
= __ptep_modify_prot_start
,
419 .ptep_modify_prot_commit
= __ptep_modify_prot_commit
,
421 #ifdef CONFIG_HIGHPTE
422 .kmap_atomic_pte
= kmap_atomic
,
425 #if PAGETABLE_LEVELS >= 3
426 #ifdef CONFIG_X86_PAE
427 .set_pte_atomic
= native_set_pte_atomic
,
428 .set_pte_present
= native_set_pte_present
,
429 .pte_clear
= native_pte_clear
,
430 .pmd_clear
= native_pmd_clear
,
432 .set_pud
= native_set_pud
,
433 .pmd_val
= native_pmd_val
,
434 .make_pmd
= native_make_pmd
,
436 #if PAGETABLE_LEVELS == 4
437 .pud_val
= native_pud_val
,
438 .make_pud
= native_make_pud
,
439 .set_pgd
= native_set_pgd
,
441 #endif /* PAGETABLE_LEVELS >= 3 */
443 .pte_val
= native_pte_val
,
444 .pte_flags
= native_pte_flags
,
445 .pgd_val
= native_pgd_val
,
447 .make_pte
= native_make_pte
,
448 .make_pgd
= native_make_pgd
,
450 .dup_mmap
= paravirt_nop
,
451 .exit_mmap
= paravirt_nop
,
452 .activate_mm
= paravirt_nop
,
455 .enter
= paravirt_nop
,
456 .leave
= paravirt_nop
,
459 .set_fixmap
= native_set_fixmap
,
462 struct pv_lock_ops pv_lock_ops
= {
464 .spin_is_locked
= __ticket_spin_is_locked
,
465 .spin_is_contended
= __ticket_spin_is_contended
,
467 .spin_lock
= __ticket_spin_lock
,
468 .spin_trylock
= __ticket_spin_trylock
,
469 .spin_unlock
= __ticket_spin_unlock
,
472 EXPORT_SYMBOL(pv_lock_ops
);
474 EXPORT_SYMBOL_GPL(pv_time_ops
);
475 EXPORT_SYMBOL (pv_cpu_ops
);
476 EXPORT_SYMBOL (pv_mmu_ops
);
477 EXPORT_SYMBOL_GPL(pv_apic_ops
);
478 EXPORT_SYMBOL_GPL(pv_info
);
479 EXPORT_SYMBOL (pv_irq_ops
);