]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/i386/i386/vm86.c
Fix bhyve SVM guest escape.
[FreeBSD/FreeBSD.git] / sys / i386 / i386 / vm86.c
1 /*-
2  * Copyright (c) 1997 Jonathan Lemon
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  */
26
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
29
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/priv.h>
33 #include <sys/proc.h>
34 #include <sys/lock.h>
35 #include <sys/malloc.h>
36 #include <sys/mutex.h>
37
38 #include <vm/vm.h>
39 #include <vm/pmap.h>
40 #include <vm/vm_map.h>
41 #include <vm/vm_page.h>
42
43 #include <machine/md_var.h>
44 #include <machine/pcb.h>
45 #include <machine/pcb_ext.h>
46 #include <machine/psl.h>
47 #include <machine/specialreg.h>
48 #include <machine/sysarch.h>
49
50 extern int vm86pa;
51 extern struct pcb *vm86pcb;
52
53 static struct mtx vm86_lock;
54
55 extern int vm86_bioscall(struct vm86frame *);
56 extern void vm86_biosret(struct vm86frame *);
57
58 void vm86_prepcall(struct vm86frame *);
59
60 struct system_map {
61         int             type;
62         vm_offset_t     start;
63         vm_offset_t     end;
64 };
65
66 #define HLT     0xf4
67 #define CLI     0xfa
68 #define STI     0xfb
69 #define PUSHF   0x9c
70 #define POPF    0x9d
71 #define INTn    0xcd
72 #define IRET    0xcf
73 #define CALLm   0xff
74 #define OPERAND_SIZE_PREFIX     0x66
75 #define ADDRESS_SIZE_PREFIX     0x67
76 #define PUSH_MASK       ~(PSL_VM | PSL_RF | PSL_I)
77 #define POP_MASK        ~(PSL_VIP | PSL_VIF | PSL_VM | PSL_RF | PSL_IOPL)
78
79 static __inline caddr_t
80 MAKE_ADDR(u_short sel, u_short off)
81 {
82         return ((caddr_t)((sel << 4) + off));
83 }
84
85 static __inline void
86 GET_VEC(u_int vec, u_short *sel, u_short *off)
87 {
88         *sel = vec >> 16;
89         *off = vec & 0xffff;
90 }
91
92 static __inline u_int
93 MAKE_VEC(u_short sel, u_short off)
94 {
95         return ((sel << 16) | off);
96 }
97
98 static __inline void
99 PUSH(u_short x, struct vm86frame *vmf)
100 {
101         vmf->vmf_sp -= 2;
102         suword16(MAKE_ADDR(vmf->vmf_ss, vmf->vmf_sp), x);
103 }
104
105 static __inline void
106 PUSHL(u_int x, struct vm86frame *vmf)
107 {
108         vmf->vmf_sp -= 4;
109         suword(MAKE_ADDR(vmf->vmf_ss, vmf->vmf_sp), x);
110 }
111
112 static __inline u_short
113 POP(struct vm86frame *vmf)
114 {
115         u_short x = fuword16(MAKE_ADDR(vmf->vmf_ss, vmf->vmf_sp));
116
117         vmf->vmf_sp += 2;
118         return (x);
119 }
120
121 static __inline u_int
122 POPL(struct vm86frame *vmf)
123 {
124         u_int x = fuword(MAKE_ADDR(vmf->vmf_ss, vmf->vmf_sp));
125
126         vmf->vmf_sp += 4;
127         return (x);
128 }
129
130 int
131 vm86_emulate(struct vm86frame *vmf)
132 {
133         struct vm86_kernel *vm86;
134         caddr_t addr;
135         u_char i_byte;
136         u_int temp_flags;
137         int inc_ip = 1;
138         int retcode = 0;
139
140         /*
141          * pcb_ext contains the address of the extension area, or zero if
142          * the extension is not present.  (This check should not be needed,
143          * as we can't enter vm86 mode until we set up an extension area)
144          */
145         if (curpcb->pcb_ext == 0)
146                 return (SIGBUS);
147         vm86 = &curpcb->pcb_ext->ext_vm86;
148
149         if (vmf->vmf_eflags & PSL_T)
150                 retcode = SIGTRAP;
151
152         addr = MAKE_ADDR(vmf->vmf_cs, vmf->vmf_ip);
153         i_byte = fubyte(addr);
154         if (i_byte == ADDRESS_SIZE_PREFIX) {
155                 i_byte = fubyte(++addr);
156                 inc_ip++;
157         }
158
159         if (vm86->vm86_has_vme) {
160                 switch (i_byte) {
161                 case OPERAND_SIZE_PREFIX:
162                         i_byte = fubyte(++addr);
163                         inc_ip++;
164                         switch (i_byte) {
165                         case PUSHF:
166                                 if (vmf->vmf_eflags & PSL_VIF)
167                                         PUSHL((vmf->vmf_eflags & PUSH_MASK)
168                                             | PSL_IOPL | PSL_I, vmf);
169                                 else
170                                         PUSHL((vmf->vmf_eflags & PUSH_MASK)
171                                             | PSL_IOPL, vmf);
172                                 vmf->vmf_ip += inc_ip;
173                                 return (retcode);
174
175                         case POPF:
176                                 temp_flags = POPL(vmf) & POP_MASK;
177                                 vmf->vmf_eflags = (vmf->vmf_eflags & ~POP_MASK)
178                                     | temp_flags | PSL_VM | PSL_I;
179                                 vmf->vmf_ip += inc_ip;
180                                 if (temp_flags & PSL_I) {
181                                         vmf->vmf_eflags |= PSL_VIF;
182                                         if (vmf->vmf_eflags & PSL_VIP)
183                                                 break;
184                                 } else {
185                                         vmf->vmf_eflags &= ~PSL_VIF;
186                                 }
187                                 return (retcode);
188                         }
189                         break;
190
191                 /* VME faults here if VIP is set, but does not set VIF. */
192                 case STI:
193                         vmf->vmf_eflags |= PSL_VIF;
194                         vmf->vmf_ip += inc_ip;
195                         if ((vmf->vmf_eflags & PSL_VIP) == 0) {
196                                 uprintf("fatal sti\n");
197                                 return (SIGKILL);
198                         }
199                         break;
200
201                 /* VME if no redirection support */
202                 case INTn:
203                         break;
204
205                 /* VME if trying to set PSL_T, or PSL_I when VIP is set */
206                 case POPF:
207                         temp_flags = POP(vmf) & POP_MASK;
208                         vmf->vmf_flags = (vmf->vmf_flags & ~POP_MASK)
209                             | temp_flags | PSL_VM | PSL_I;
210                         vmf->vmf_ip += inc_ip;
211                         if (temp_flags & PSL_I) {
212                                 vmf->vmf_eflags |= PSL_VIF;
213                                 if (vmf->vmf_eflags & PSL_VIP)
214                                         break;
215                         } else {
216                                 vmf->vmf_eflags &= ~PSL_VIF;
217                         }
218                         return (retcode);
219
220                 /* VME if trying to set PSL_T, or PSL_I when VIP is set */
221                 case IRET:
222                         vmf->vmf_ip = POP(vmf);
223                         vmf->vmf_cs = POP(vmf);
224                         temp_flags = POP(vmf) & POP_MASK;
225                         vmf->vmf_flags = (vmf->vmf_flags & ~POP_MASK)
226                             | temp_flags | PSL_VM | PSL_I;
227                         if (temp_flags & PSL_I) {
228                                 vmf->vmf_eflags |= PSL_VIF;
229                                 if (vmf->vmf_eflags & PSL_VIP)
230                                         break;
231                         } else {
232                                 vmf->vmf_eflags &= ~PSL_VIF;
233                         }
234                         return (retcode);
235
236                 }
237                 return (SIGBUS);
238         }
239
240         switch (i_byte) {
241         case OPERAND_SIZE_PREFIX:
242                 i_byte = fubyte(++addr);
243                 inc_ip++;
244                 switch (i_byte) {
245                 case PUSHF:
246                         if (vm86->vm86_eflags & PSL_VIF)
247                                 PUSHL((vmf->vmf_flags & PUSH_MASK)
248                                     | PSL_IOPL | PSL_I, vmf);
249                         else
250                                 PUSHL((vmf->vmf_flags & PUSH_MASK)
251                                     | PSL_IOPL, vmf);
252                         vmf->vmf_ip += inc_ip;
253                         return (retcode);
254
255                 case POPF:
256                         temp_flags = POPL(vmf) & POP_MASK;
257                         vmf->vmf_eflags = (vmf->vmf_eflags & ~POP_MASK)
258                             | temp_flags | PSL_VM | PSL_I;
259                         vmf->vmf_ip += inc_ip;
260                         if (temp_flags & PSL_I) {
261                                 vm86->vm86_eflags |= PSL_VIF;
262                                 if (vm86->vm86_eflags & PSL_VIP)
263                                         break;
264                         } else {
265                                 vm86->vm86_eflags &= ~PSL_VIF;
266                         }
267                         return (retcode);
268                 }
269                 return (SIGBUS);
270
271         case CLI:
272                 vm86->vm86_eflags &= ~PSL_VIF;
273                 vmf->vmf_ip += inc_ip;
274                 return (retcode);
275
276         case STI:
277                 /* if there is a pending interrupt, go to the emulator */
278                 vm86->vm86_eflags |= PSL_VIF;
279                 vmf->vmf_ip += inc_ip;
280                 if (vm86->vm86_eflags & PSL_VIP)
281                         break;
282                 return (retcode);
283
284         case PUSHF:
285                 if (vm86->vm86_eflags & PSL_VIF)
286                         PUSH((vmf->vmf_flags & PUSH_MASK)
287                             | PSL_IOPL | PSL_I, vmf);
288                 else
289                         PUSH((vmf->vmf_flags & PUSH_MASK) | PSL_IOPL, vmf);
290                 vmf->vmf_ip += inc_ip;
291                 return (retcode);
292
293         case INTn:
294                 i_byte = fubyte(addr + 1);
295                 if ((vm86->vm86_intmap[i_byte >> 3] & (1 << (i_byte & 7))) != 0)
296                         break;
297                 if (vm86->vm86_eflags & PSL_VIF)
298                         PUSH((vmf->vmf_flags & PUSH_MASK)
299                             | PSL_IOPL | PSL_I, vmf);
300                 else
301                         PUSH((vmf->vmf_flags & PUSH_MASK) | PSL_IOPL, vmf);
302                 PUSH(vmf->vmf_cs, vmf);
303                 PUSH(vmf->vmf_ip + inc_ip + 1, vmf);    /* increment IP */
304                 GET_VEC(fuword((caddr_t)(i_byte * 4)),
305                      &vmf->vmf_cs, &vmf->vmf_ip);
306                 vmf->vmf_flags &= ~PSL_T;
307                 vm86->vm86_eflags &= ~PSL_VIF;
308                 return (retcode);
309
310         case IRET:
311                 vmf->vmf_ip = POP(vmf);
312                 vmf->vmf_cs = POP(vmf);
313                 temp_flags = POP(vmf) & POP_MASK;
314                 vmf->vmf_flags = (vmf->vmf_flags & ~POP_MASK)
315                     | temp_flags | PSL_VM | PSL_I;
316                 if (temp_flags & PSL_I) {
317                         vm86->vm86_eflags |= PSL_VIF;
318                         if (vm86->vm86_eflags & PSL_VIP)
319                                 break;
320                 } else {
321                         vm86->vm86_eflags &= ~PSL_VIF;
322                 }
323                 return (retcode);
324
325         case POPF:
326                 temp_flags = POP(vmf) & POP_MASK;
327                 vmf->vmf_flags = (vmf->vmf_flags & ~POP_MASK)
328                     | temp_flags | PSL_VM | PSL_I;
329                 vmf->vmf_ip += inc_ip;
330                 if (temp_flags & PSL_I) {
331                         vm86->vm86_eflags |= PSL_VIF;
332                         if (vm86->vm86_eflags & PSL_VIP)
333                                 break;
334                 } else {
335                         vm86->vm86_eflags &= ~PSL_VIF;
336                 }
337                 return (retcode);
338         }
339         return (SIGBUS);
340 }
341
342 #define PGTABLE_SIZE    ((1024 + 64) * 1024 / PAGE_SIZE)
343 #define INTMAP_SIZE     32
344 #define IOMAP_SIZE      ctob(IOPAGES)
345 #define TSS_SIZE \
346         (sizeof(struct pcb_ext) - sizeof(struct segment_descriptor) + \
347          INTMAP_SIZE + IOMAP_SIZE + 1)
348
349 struct vm86_layout {
350         pt_entry_t      vml_pgtbl[PGTABLE_SIZE];
351         struct  pcb vml_pcb;
352         struct  pcb_ext vml_ext;
353         char    vml_intmap[INTMAP_SIZE];
354         char    vml_iomap[IOMAP_SIZE];
355         char    vml_iomap_trailer;
356 };
357
358 void
359 vm86_initialize(void)
360 {
361         int i;
362         u_int *addr;
363         struct vm86_layout *vml = (struct vm86_layout *)vm86paddr;
364         struct pcb *pcb;
365         struct pcb_ext *ext;
366         struct soft_segment_descriptor ssd = {
367                 0,                      /* segment base address (overwritten) */
368                 0,                      /* length (overwritten) */
369                 SDT_SYS386TSS,          /* segment type */
370                 0,                      /* priority level */
371                 1,                      /* descriptor present */
372                 0, 0,
373                 0,                      /* default 16 size */
374                 0                       /* granularity */
375         };
376
377         /*
378          * this should be a compile time error, but cpp doesn't grok sizeof().
379          */
380         if (sizeof(struct vm86_layout) > ctob(3))
381                 panic("struct vm86_layout exceeds space allocated in locore.s");
382
383         /*
384          * Below is the memory layout that we use for the vm86 region.
385          *
386          * +--------+
387          * |        | 
388          * |        |
389          * | page 0 |       
390          * |        | +--------+
391          * |        | | stack  |
392          * +--------+ +--------+ <--------- vm86paddr
393          * |        | |Page Tbl| 1M + 64K = 272 entries = 1088 bytes
394          * |        | +--------+
395          * |        | |  PCB   | size: ~240 bytes
396          * | page 1 | |PCB Ext | size: ~140 bytes (includes TSS)
397          * |        | +--------+
398          * |        | |int map |
399          * |        | +--------+
400          * +--------+ |        |
401          * | page 2 | |  I/O   |
402          * +--------+ | bitmap |
403          * | page 3 | |        |
404          * |        | +--------+
405          * +--------+ 
406          */
407
408         /*
409          * A rudimentary PCB must be installed, in order to get to the
410          * PCB extension area.  We use the PCB area as a scratchpad for
411          * data storage, the layout of which is shown below.
412          *
413          * pcb_esi      = new PTD entry 0
414          * pcb_ebp      = pointer to frame on vm86 stack
415          * pcb_esp      =    stack frame pointer at time of switch
416          * pcb_ebx      = va of vm86 page table
417          * pcb_eip      =    argument pointer to initial call
418          * pcb_vm86[0]  =    saved TSS descriptor, word 0
419          * pcb_vm86[1]  =    saved TSS descriptor, word 1
420          */
421 #define new_ptd         pcb_esi
422 #define vm86_frame      pcb_ebp
423 #define pgtable_va      pcb_ebx
424
425         pcb = &vml->vml_pcb;
426         ext = &vml->vml_ext;
427
428         mtx_init(&vm86_lock, "vm86 lock", NULL, MTX_DEF);
429
430         bzero(pcb, sizeof(struct pcb));
431         pcb->new_ptd = vm86pa | PG_V | PG_RW | PG_U;
432         pcb->vm86_frame = vm86paddr - sizeof(struct vm86frame);
433         pcb->pgtable_va = vm86paddr;
434         pcb->pcb_flags = PCB_VM86CALL; 
435         pcb->pcb_ext = ext;
436
437         bzero(ext, sizeof(struct pcb_ext)); 
438         ext->ext_tss.tss_esp0 = vm86paddr;
439         ext->ext_tss.tss_ss0 = GSEL(GDATA_SEL, SEL_KPL);
440         ext->ext_tss.tss_ioopt = 
441                 ((u_int)vml->vml_iomap - (u_int)&ext->ext_tss) << 16;
442         ext->ext_iomap = vml->vml_iomap;
443         ext->ext_vm86.vm86_intmap = vml->vml_intmap;
444
445         if (cpu_feature & CPUID_VME)
446                 ext->ext_vm86.vm86_has_vme = (rcr4() & CR4_VME ? 1 : 0);
447
448         addr = (u_int *)ext->ext_vm86.vm86_intmap;
449         for (i = 0; i < (INTMAP_SIZE + IOMAP_SIZE) / sizeof(u_int); i++)
450                 *addr++ = 0;
451         vml->vml_iomap_trailer = 0xff;
452
453         ssd.ssd_base = (u_int)&ext->ext_tss;
454         ssd.ssd_limit = TSS_SIZE - 1; 
455         ssdtosd(&ssd, &ext->ext_tssd);
456
457         vm86pcb = pcb;
458
459 #if 0
460         /*
461          * use whatever is leftover of the vm86 page layout as a
462          * message buffer so we can capture early output.
463          */
464         msgbufinit((vm_offset_t)vm86paddr + sizeof(struct vm86_layout),
465             ctob(3) - sizeof(struct vm86_layout));
466 #endif
467 }
468
469 vm_offset_t
470 vm86_getpage(struct vm86context *vmc, int pagenum)
471 {
472         int i;
473
474         for (i = 0; i < vmc->npages; i++)
475                 if (vmc->pmap[i].pte_num == pagenum)
476                         return (vmc->pmap[i].kva);
477         return (0);
478 }
479
480 vm_offset_t
481 vm86_addpage(struct vm86context *vmc, int pagenum, vm_offset_t kva)
482 {
483         int i, flags = 0;
484
485         for (i = 0; i < vmc->npages; i++)
486                 if (vmc->pmap[i].pte_num == pagenum)
487                         goto overlap;
488
489         if (vmc->npages == VM86_PMAPSIZE)
490                 goto full;                      /* XXX grow map? */
491
492         if (kva == 0) {
493                 kva = (vm_offset_t)malloc(PAGE_SIZE, M_TEMP, M_WAITOK);
494                 flags = VMAP_MALLOC;
495         }
496
497         i = vmc->npages++;
498         vmc->pmap[i].flags = flags;
499         vmc->pmap[i].kva = kva;
500         vmc->pmap[i].pte_num = pagenum;
501         return (kva);
502 overlap:
503         panic("vm86_addpage: overlap");
504 full:
505         panic("vm86_addpage: not enough room");
506 }
507
508 /*
509  * called from vm86_bioscall, while in vm86 address space, to finalize setup.
510  */
511 void
512 vm86_prepcall(struct vm86frame *vmf)
513 {
514         struct vm86_kernel *vm86;
515         uint32_t *stack;
516         uint8_t *code;
517
518         code = (void *)0xa00;
519         stack = (void *)(0x1000 - 2);   /* keep aligned */
520         if ((vmf->vmf_trapno & PAGE_MASK) <= 0xff) {
521                 /* interrupt call requested */
522                 code[0] = INTn;
523                 code[1] = vmf->vmf_trapno & 0xff;
524                 code[2] = HLT;
525                 vmf->vmf_ip = (uintptr_t)code;
526                 vmf->vmf_cs = 0;
527         } else {
528                 code[0] = HLT;
529                 stack--;
530                 stack[0] = MAKE_VEC(0, (uintptr_t)code);
531         }
532         vmf->vmf_sp = (uintptr_t)stack;
533         vmf->vmf_ss = 0;
534         vmf->kernel_fs = vmf->kernel_es = vmf->kernel_ds = 0;
535         vmf->vmf_eflags = PSL_VIF | PSL_VM | PSL_USER;
536
537         vm86 = &curpcb->pcb_ext->ext_vm86;
538         if (!vm86->vm86_has_vme) 
539                 vm86->vm86_eflags = vmf->vmf_eflags;  /* save VIF, VIP */
540 }
541
542 /*
543  * vm86 trap handler; determines whether routine succeeded or not.
544  * Called while in vm86 space, returns to calling process.
545  */
546 void
547 vm86_trap(struct vm86frame *vmf)
548 {
549         caddr_t addr;
550
551         /* "should not happen" */
552         if ((vmf->vmf_eflags & PSL_VM) == 0)
553                 panic("vm86_trap called, but not in vm86 mode");
554
555         addr = MAKE_ADDR(vmf->vmf_cs, vmf->vmf_ip);
556         if (*(u_char *)addr == HLT)
557                 vmf->vmf_trapno = vmf->vmf_eflags & PSL_C;
558         else
559                 vmf->vmf_trapno = vmf->vmf_trapno << 16;
560
561         vm86_biosret(vmf);
562 }
563
564 int
565 vm86_intcall(int intnum, struct vm86frame *vmf)
566 {
567         int retval;
568
569         if (intnum < 0 || intnum > 0xff)
570                 return (EINVAL);
571
572         vmf->vmf_trapno = intnum;
573         mtx_lock(&vm86_lock);
574         critical_enter();
575         retval = vm86_bioscall(vmf);
576         critical_exit();
577         mtx_unlock(&vm86_lock);
578         return (retval);
579 }
580
581 /*
582  * struct vm86context contains the page table to use when making
583  * vm86 calls.  If intnum is a valid interrupt number (0-255), then
584  * the "interrupt trampoline" will be used, otherwise we use the
585  * caller's cs:ip routine.  
586  */
587 int
588 vm86_datacall(int intnum, struct vm86frame *vmf, struct vm86context *vmc)
589 {
590         pt_entry_t *pte = (pt_entry_t *)vm86paddr;
591         vm_paddr_t page;
592         int i, entry, retval;
593
594         mtx_lock(&vm86_lock);
595         for (i = 0; i < vmc->npages; i++) {
596                 page = vtophys(vmc->pmap[i].kva & PG_FRAME);
597                 entry = vmc->pmap[i].pte_num; 
598                 vmc->pmap[i].old_pte = pte[entry];
599                 pte[entry] = page | PG_V | PG_RW | PG_U;
600                 pmap_invalidate_page(kernel_pmap, vmc->pmap[i].kva);
601         }
602
603         vmf->vmf_trapno = intnum;
604         critical_enter();
605         retval = vm86_bioscall(vmf);
606         critical_exit();
607
608         for (i = 0; i < vmc->npages; i++) {
609                 entry = vmc->pmap[i].pte_num;
610                 pte[entry] = vmc->pmap[i].old_pte;
611                 pmap_invalidate_page(kernel_pmap, vmc->pmap[i].kva);
612         }
613         mtx_unlock(&vm86_lock);
614
615         return (retval);
616 }
617
618 vm_offset_t
619 vm86_getaddr(struct vm86context *vmc, u_short sel, u_short off)
620 {
621         int i, page;
622         vm_offset_t addr;
623
624         addr = (vm_offset_t)MAKE_ADDR(sel, off);
625         page = addr >> PAGE_SHIFT;
626         for (i = 0; i < vmc->npages; i++)
627                 if (page == vmc->pmap[i].pte_num)
628                         return (vmc->pmap[i].kva + (addr & PAGE_MASK));
629         return (0);
630 }
631
632 int
633 vm86_getptr(struct vm86context *vmc, vm_offset_t kva, u_short *sel,
634      u_short *off)
635 {
636         int i;
637
638         for (i = 0; i < vmc->npages; i++)
639                 if (kva >= vmc->pmap[i].kva &&
640                     kva < vmc->pmap[i].kva + PAGE_SIZE) {
641                         *off = kva - vmc->pmap[i].kva;
642                         *sel = vmc->pmap[i].pte_num << 8;
643                         return (1);
644                 }
645         return (0);
646 }
647         
648 int
649 vm86_sysarch(struct thread *td, char *args)
650 {
651         int error = 0;
652         struct i386_vm86_args ua;
653         struct vm86_kernel *vm86;
654
655         if ((error = copyin(args, &ua, sizeof(struct i386_vm86_args))) != 0)
656                 return (error);
657
658         if (td->td_pcb->pcb_ext == 0)
659                 if ((error = i386_extend_pcb(td)) != 0)
660                         return (error);
661         vm86 = &td->td_pcb->pcb_ext->ext_vm86;
662
663         switch (ua.sub_op) {
664         case VM86_INIT: {
665                 struct vm86_init_args sa;
666
667                 if ((error = copyin(ua.sub_args, &sa, sizeof(sa))) != 0)
668                         return (error);
669                 if (cpu_feature & CPUID_VME)
670                         vm86->vm86_has_vme = (rcr4() & CR4_VME ? 1 : 0);
671                 else
672                         vm86->vm86_has_vme = 0;
673                 vm86->vm86_inited = 1;
674                 vm86->vm86_debug = sa.debug;
675                 bcopy(&sa.int_map, vm86->vm86_intmap, 32);
676                 }
677                 break;
678
679 #if 0
680         case VM86_SET_VME: {
681                 struct vm86_vme_args sa;
682         
683                 if ((cpu_feature & CPUID_VME) == 0)
684                         return (ENODEV);
685
686                 if (error = copyin(ua.sub_args, &sa, sizeof(sa)))
687                         return (error);
688                 if (sa.state)
689                         load_cr4(rcr4() | CR4_VME);
690                 else
691                         load_cr4(rcr4() & ~CR4_VME);
692                 }
693                 break;
694 #endif
695
696         case VM86_GET_VME: {
697                 struct vm86_vme_args sa;
698
699                 sa.state = (rcr4() & CR4_VME ? 1 : 0);
700                 error = copyout(&sa, ua.sub_args, sizeof(sa));
701                 }
702                 break;
703
704         case VM86_INTCALL: {
705                 struct vm86_intcall_args sa;
706
707                 if ((error = priv_check(td, PRIV_VM86_INTCALL)))
708                         return (error);
709                 if ((error = copyin(ua.sub_args, &sa, sizeof(sa))))
710                         return (error);
711                 if ((error = vm86_intcall(sa.intnum, &sa.vmf)))
712                         return (error);
713                 error = copyout(&sa, ua.sub_args, sizeof(sa));
714                 }
715                 break;
716
717         default:
718                 error = EINVAL;
719         }
720         return (error);
721 }