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,
128 return *((void **)&tmpl + type);
131 unsigned paravirt_patch_default(u8 type, u16 clobbers, void *insnbuf,
132 unsigned long addr, unsigned len)
134 void *opfunc = get_call_destination(type);
138 /* If there's no function, patch it with a ud2a (BUG) */
139 ret = paravirt_patch_insns(insnbuf, len, ud2a, ud2a+sizeof(ud2a));
140 else if (opfunc == paravirt_nop)
141 /* If the operation is a nop, then nop the callsite */
142 ret = paravirt_patch_nop();
143 else if (type == PARAVIRT_PATCH(pv_cpu_ops.iret) ||
144 type == PARAVIRT_PATCH(pv_cpu_ops.irq_enable_sysexit) ||
145 type == PARAVIRT_PATCH(pv_cpu_ops.usergs_sysret32) ||
146 type == PARAVIRT_PATCH(pv_cpu_ops.usergs_sysret64))
147 /* If operation requires a jmp, then jmp */
148 ret = paravirt_patch_jmp(insnbuf, opfunc, addr, len);
150 /* Otherwise call the function; assume target could
151 clobber any caller-save reg */
152 ret = paravirt_patch_call(insnbuf, opfunc, CLBR_ANY,
153 addr, clobbers, len);
158 unsigned paravirt_patch_insns(void *insnbuf, unsigned len,
159 const char *start, const char *end)
161 unsigned insn_len = end - start;
163 if (insn_len > len || start == NULL)
166 memcpy(insnbuf, start, insn_len);
173 pv_irq_ops.init_IRQ();
176 static void native_flush_tlb(void)
178 __native_flush_tlb();
182 * Global pages have to be flushed a bit differently. Not a real
183 * performance problem because this does not happen often.
185 static void native_flush_tlb_global(void)
187 __native_flush_tlb_global();
190 static void native_flush_tlb_single(unsigned long addr)
192 __native_flush_tlb_single(addr);
195 /* These are in entry.S */
196 extern void native_iret(void);
197 extern void native_irq_enable_sysexit(void);
198 extern void native_usergs_sysret32(void);
199 extern void native_usergs_sysret64(void);
201 static int __init print_banner(void)
203 pv_init_ops.banner();
206 core_initcall(print_banner);
208 static struct resource reserve_ioports = {
210 .end = IO_SPACE_LIMIT,
211 .name = "paravirt-ioport",
212 .flags = IORESOURCE_IO | IORESOURCE_BUSY,
216 * Reserve the whole legacy IO space to prevent any legacy drivers
217 * from wasting time probing for their hardware. This is a fairly
218 * brute-force approach to disabling all non-virtual drivers.
220 * Note that this must be called very early to have any effect.
222 int paravirt_disable_iospace(void)
224 return request_resource(&ioport_resource, &reserve_ioports);
227 static DEFINE_PER_CPU(enum paravirt_lazy_mode, paravirt_lazy_mode) = PARAVIRT_LAZY_NONE;
229 static inline void enter_lazy(enum paravirt_lazy_mode mode)
231 BUG_ON(__get_cpu_var(paravirt_lazy_mode) != PARAVIRT_LAZY_NONE);
232 BUG_ON(preemptible());
234 __get_cpu_var(paravirt_lazy_mode) = mode;
237 void paravirt_leave_lazy(enum paravirt_lazy_mode mode)
239 BUG_ON(__get_cpu_var(paravirt_lazy_mode) != mode);
240 BUG_ON(preemptible());
242 __get_cpu_var(paravirt_lazy_mode) = PARAVIRT_LAZY_NONE;
245 void paravirt_enter_lazy_mmu(void)
247 enter_lazy(PARAVIRT_LAZY_MMU);
250 void paravirt_leave_lazy_mmu(void)
252 paravirt_leave_lazy(PARAVIRT_LAZY_MMU);
255 void paravirt_enter_lazy_cpu(void)
257 enter_lazy(PARAVIRT_LAZY_CPU);
260 void paravirt_leave_lazy_cpu(void)
262 paravirt_leave_lazy(PARAVIRT_LAZY_CPU);
265 enum paravirt_lazy_mode paravirt_get_lazy_mode(void)
267 return __get_cpu_var(paravirt_lazy_mode);
270 struct pv_info pv_info = {
271 .name = "bare hardware",
272 .paravirt_enabled = 0,
274 .shared_kernel_pmd = 1, /* Only used when CONFIG_X86_PAE is set */
277 struct pv_init_ops pv_init_ops = {
278 .patch = native_patch,
279 .banner = default_banner,
280 .arch_setup = paravirt_nop,
281 .memory_setup = machine_specific_memory_setup,
284 struct pv_time_ops pv_time_ops = {
285 .time_init = hpet_time_init,
286 .get_wallclock = native_get_wallclock,
287 .set_wallclock = native_set_wallclock,
288 .sched_clock = native_sched_clock,
289 .get_tsc_khz = native_calibrate_tsc,
292 struct pv_irq_ops pv_irq_ops = {
293 .init_IRQ = native_init_IRQ,
294 .save_fl = native_save_fl,
295 .restore_fl = native_restore_fl,
296 .irq_disable = native_irq_disable,
297 .irq_enable = native_irq_enable,
298 .safe_halt = native_safe_halt,
301 .adjust_exception_frame = paravirt_nop,
305 struct pv_cpu_ops pv_cpu_ops = {
306 .cpuid = native_cpuid,
307 .get_debugreg = native_get_debugreg,
308 .set_debugreg = native_set_debugreg,
310 .read_cr0 = native_read_cr0,
311 .write_cr0 = native_write_cr0,
312 .read_cr4 = native_read_cr4,
313 .read_cr4_safe = native_read_cr4_safe,
314 .write_cr4 = native_write_cr4,
316 .read_cr8 = native_read_cr8,
317 .write_cr8 = native_write_cr8,
319 .wbinvd = native_wbinvd,
320 .read_msr = native_read_msr_safe,
321 .write_msr = native_write_msr_safe,
322 .read_tsc = native_read_tsc,
323 .read_pmc = native_read_pmc,
324 .read_tscp = native_read_tscp,
325 .load_tr_desc = native_load_tr_desc,
326 .set_ldt = native_set_ldt,
327 .load_gdt = native_load_gdt,
328 .load_idt = native_load_idt,
329 .store_gdt = native_store_gdt,
330 .store_idt = native_store_idt,
331 .store_tr = native_store_tr,
332 .load_tls = native_load_tls,
334 .load_gs_index = native_load_gs_index,
336 .write_ldt_entry = native_write_ldt_entry,
337 .write_gdt_entry = native_write_gdt_entry,
338 .write_idt_entry = native_write_idt_entry,
339 .load_sp0 = native_load_sp0,
341 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
342 .irq_enable_sysexit = native_irq_enable_sysexit,
345 #ifdef CONFIG_IA32_EMULATION
346 .usergs_sysret32 = native_usergs_sysret32,
348 .usergs_sysret64 = native_usergs_sysret64,
351 .swapgs = native_swapgs,
353 .set_iopl_mask = native_set_iopl_mask,
354 .io_delay = native_io_delay,
357 .enter = paravirt_nop,
358 .leave = paravirt_nop,
362 struct pv_apic_ops pv_apic_ops = {
363 #ifdef CONFIG_X86_LOCAL_APIC
364 .apic_write = native_apic_write,
365 .apic_write_atomic = native_apic_write_atomic,
366 .apic_read = native_apic_read,
367 .setup_boot_clock = setup_boot_APIC_clock,
368 .setup_secondary_clock = setup_secondary_APIC_clock,
369 .startup_ipi_hook = paravirt_nop,
373 struct pv_mmu_ops pv_mmu_ops = {
374 #ifndef CONFIG_X86_64
375 .pagetable_setup_start = native_pagetable_setup_start,
376 .pagetable_setup_done = native_pagetable_setup_done,
378 .pagetable_setup_start = paravirt_nop,
379 .pagetable_setup_done = paravirt_nop,
382 .read_cr2 = native_read_cr2,
383 .write_cr2 = native_write_cr2,
384 .read_cr3 = native_read_cr3,
385 .write_cr3 = native_write_cr3,
387 .flush_tlb_user = native_flush_tlb,
388 .flush_tlb_kernel = native_flush_tlb_global,
389 .flush_tlb_single = native_flush_tlb_single,
390 .flush_tlb_others = native_flush_tlb_others,
392 .pgd_alloc = __paravirt_pgd_alloc,
393 .pgd_free = paravirt_nop,
395 .alloc_pte = paravirt_nop,
396 .alloc_pmd = paravirt_nop,
397 .alloc_pmd_clone = paravirt_nop,
398 .alloc_pud = paravirt_nop,
399 .release_pte = paravirt_nop,
400 .release_pmd = paravirt_nop,
401 .release_pud = paravirt_nop,
403 .set_pte = native_set_pte,
404 .set_pte_at = native_set_pte_at,
405 .set_pmd = native_set_pmd,
406 .pte_update = paravirt_nop,
407 .pte_update_defer = paravirt_nop,
409 .ptep_modify_prot_start = __ptep_modify_prot_start,
410 .ptep_modify_prot_commit = __ptep_modify_prot_commit,
412 #ifdef CONFIG_HIGHPTE
413 .kmap_atomic_pte = kmap_atomic,
416 #if PAGETABLE_LEVELS >= 3
417 #ifdef CONFIG_X86_PAE
418 .set_pte_atomic = native_set_pte_atomic,
419 .set_pte_present = native_set_pte_present,
420 .pte_clear = native_pte_clear,
421 .pmd_clear = native_pmd_clear,
423 .set_pud = native_set_pud,
424 .pmd_val = native_pmd_val,
425 .make_pmd = native_make_pmd,
427 #if PAGETABLE_LEVELS == 4
428 .pud_val = native_pud_val,
429 .make_pud = native_make_pud,
430 .set_pgd = native_set_pgd,
432 #endif /* PAGETABLE_LEVELS >= 3 */
434 .pte_val = native_pte_val,
435 .pte_flags = native_pte_val,
436 .pgd_val = native_pgd_val,
438 .make_pte = native_make_pte,
439 .make_pgd = native_make_pgd,
441 .dup_mmap = paravirt_nop,
442 .exit_mmap = paravirt_nop,
443 .activate_mm = paravirt_nop,
446 .enter = paravirt_nop,
447 .leave = paravirt_nop,
450 .set_fixmap = native_set_fixmap,
453 EXPORT_SYMBOL_GPL(pv_time_ops);
454 EXPORT_SYMBOL (pv_cpu_ops);
455 EXPORT_SYMBOL (pv_mmu_ops);
456 EXPORT_SYMBOL_GPL(pv_apic_ops);
457 EXPORT_SYMBOL_GPL(pv_info);
458 EXPORT_SYMBOL (pv_irq_ops);