vmmapi.c revision 221828 
1/*-
2 * Copyright (c) 2011 NetApp, Inc.
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 NETAPP, INC ``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 NETAPP, INC 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 * $FreeBSD$
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD$");
31
32#include <sys/types.h>
33#include <sys/sysctl.h>
34#include <sys/ioctl.h>
35#include <sys/mman.h>
36
37#include <machine/specialreg.h>
38
39#include <stdio.h>
40#include <stdlib.h>
41#include <assert.h>
42#include <string.h>
43#include <fcntl.h>
44#include <unistd.h>
45
46#include <machine/vmm.h>
47#include <machine/vmm_dev.h>
48
49#include "vmmapi.h"
50#include "mptable.h"
51
52#ifndef	CR4_VMXE
53#define	CR4_VMXE	(1UL << 13)
54#endif
55
56#define BIOS_ROM_BASE		(0xf0000)
57#define BIOS_ROM_SIZE		(0x10000)
58
59struct vmctx {
60	int	fd;
61	char	*name;
62};
63
64#define	CREATE(x)  sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
65#define	DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
66
67static int
68vm_device_open(const char *name)
69{
70        int fd, len;
71        char *vmfile;
72
73	len = strlen("/dev/vmm/") + strlen(name) + 1;
74	vmfile = malloc(len);
75	assert(vmfile != NULL);
76	snprintf(vmfile, len, "/dev/vmm/%s", name);
77
78        /* Open the device file */
79        fd = open(vmfile, O_RDWR, 0);
80
81	free(vmfile);
82        return (fd);
83}
84
85int
86vm_create(const char *name)
87{
88
89	return (CREATE((char *)name));
90}
91
92struct vmctx *
93vm_open(const char *name)
94{
95	struct vmctx *vm;
96
97	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
98	assert(vm != NULL);
99
100	vm->fd = -1;
101	vm->name = (char *)(vm + 1);
102	strcpy(vm->name, name);
103
104	if ((vm->fd = vm_device_open(vm->name)) < 0)
105		goto err;
106
107	return (vm);
108err:
109	vm_destroy(vm);
110	return (NULL);
111}
112
113void
114vm_destroy(struct vmctx *vm)
115{
116	assert(vm != NULL);
117
118	DESTROY(vm->name);
119	if (vm->fd >= 0)
120		close(vm->fd);
121	free(vm);
122}
123
124int
125vm_get_memory_seg(struct vmctx *ctx, vm_paddr_t gpa,
126		  vm_paddr_t *ret_hpa, size_t *ret_len)
127{
128	int error;
129	struct vm_memory_segment seg;
130
131	bzero(&seg, sizeof(seg));
132	seg.gpa = gpa;
133	error = ioctl(ctx->fd, VM_GET_MEMORY_SEG, &seg);
134	*ret_hpa = seg.hpa;
135	*ret_len = seg.len;
136	return (error);
137}
138
139int
140vm_setup_memory(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char **mapaddr)
141{
142	int error;
143	struct vm_memory_segment seg;
144
145	/*
146	 * Create and optionally map 'len' bytes of memory at guest
147	 * physical address 'gpa'
148	 */
149	bzero(&seg, sizeof(seg));
150	seg.gpa = gpa;
151	seg.len = len;
152	error = ioctl(ctx->fd, VM_MAP_MEMORY, &seg);
153	if (error == 0 && mapaddr != NULL) {
154		*mapaddr = mmap(NULL, len, PROT_READ | PROT_WRITE, MAP_SHARED,
155				ctx->fd, gpa);
156	}
157	return (error);
158}
159
160char *
161vm_map_memory(struct vmctx *ctx, vm_paddr_t gpa, size_t len)
162{
163
164	/* Map 'len' bytes of memory at guest physical address 'gpa' */
165	return ((char *)mmap(NULL, len, PROT_READ | PROT_WRITE, MAP_SHARED,
166		     ctx->fd, gpa));
167}
168
169int
170vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
171	    uint64_t base, uint32_t limit, uint32_t access)
172{
173	int error;
174	struct vm_seg_desc vmsegdesc;
175
176	bzero(&vmsegdesc, sizeof(vmsegdesc));
177	vmsegdesc.cpuid = vcpu;
178	vmsegdesc.regnum = reg;
179	vmsegdesc.desc.base = base;
180	vmsegdesc.desc.limit = limit;
181	vmsegdesc.desc.access = access;
182
183	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
184	return (error);
185}
186
187int
188vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
189	    uint64_t *base, uint32_t *limit, uint32_t *access)
190{
191	int error;
192	struct vm_seg_desc vmsegdesc;
193
194	bzero(&vmsegdesc, sizeof(vmsegdesc));
195	vmsegdesc.cpuid = vcpu;
196	vmsegdesc.regnum = reg;
197
198	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
199	if (error == 0) {
200		*base = vmsegdesc.desc.base;
201		*limit = vmsegdesc.desc.limit;
202		*access = vmsegdesc.desc.access;
203	}
204	return (error);
205}
206
207int
208vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
209{
210	int error;
211	struct vm_register vmreg;
212
213	bzero(&vmreg, sizeof(vmreg));
214	vmreg.cpuid = vcpu;
215	vmreg.regnum = reg;
216	vmreg.regval = val;
217
218	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
219	return (error);
220}
221
222int
223vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
224{
225	int error;
226	struct vm_register vmreg;
227
228	bzero(&vmreg, sizeof(vmreg));
229	vmreg.cpuid = vcpu;
230	vmreg.regnum = reg;
231
232	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
233	*ret_val = vmreg.regval;
234	return (error);
235}
236
237int
238vm_get_pinning(struct vmctx *ctx, int vcpu, int *host_cpuid)
239{
240	int error;
241	struct vm_pin vmpin;
242
243	bzero(&vmpin, sizeof(vmpin));
244	vmpin.vm_cpuid = vcpu;
245
246	error = ioctl(ctx->fd, VM_GET_PINNING, &vmpin);
247	*host_cpuid = vmpin.host_cpuid;
248	return (error);
249}
250
251int
252vm_set_pinning(struct vmctx *ctx, int vcpu, int host_cpuid)
253{
254	int error;
255	struct vm_pin vmpin;
256
257	bzero(&vmpin, sizeof(vmpin));
258	vmpin.vm_cpuid = vcpu;
259	vmpin.host_cpuid = host_cpuid;
260
261	error = ioctl(ctx->fd, VM_SET_PINNING, &vmpin);
262	return (error);
263}
264
265int
266vm_run(struct vmctx *ctx, int vcpu, uint64_t rip, struct vm_exit *vmexit)
267{
268	int error;
269	struct vm_run vmrun;
270
271	bzero(&vmrun, sizeof(vmrun));
272	vmrun.cpuid = vcpu;
273	vmrun.rip = rip;
274
275	error = ioctl(ctx->fd, VM_RUN, &vmrun);
276	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
277	return (error);
278}
279
280static int
281vm_inject_event_real(struct vmctx *ctx, int vcpu, enum vm_event_type type,
282		     int vector, int error_code, int error_code_valid)
283{
284	struct vm_event ev;
285
286	bzero(&ev, sizeof(ev));
287	ev.cpuid = vcpu;
288	ev.type = type;
289	ev.vector = vector;
290	ev.error_code = error_code;
291	ev.error_code_valid = error_code_valid;
292
293	return (ioctl(ctx->fd, VM_INJECT_EVENT, &ev));
294}
295
296int
297vm_inject_event(struct vmctx *ctx, int vcpu, enum vm_event_type type,
298		int vector)
299{
300
301	return (vm_inject_event_real(ctx, vcpu, type, vector, 0, 0));
302}
303
304int
305vm_inject_event2(struct vmctx *ctx, int vcpu, enum vm_event_type type,
306		 int vector, int error_code)
307{
308
309	return (vm_inject_event_real(ctx, vcpu, type, vector, error_code, 1));
310}
311
312int
313vm_build_tables(struct vmctx *ctxt, int ncpu, void *oemtbl, int oemtblsz)
314{
315
316	return (vm_build_mptable(ctxt, BIOS_ROM_BASE, BIOS_ROM_SIZE, ncpu,
317				 oemtbl, oemtblsz));
318}
319
320int
321vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
322{
323	struct vm_lapic_irq vmirq;
324
325	bzero(&vmirq, sizeof(vmirq));
326	vmirq.cpuid = vcpu;
327	vmirq.vector = vector;
328
329	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
330}
331
332int
333vm_inject_nmi(struct vmctx *ctx, int vcpu)
334{
335	struct vm_nmi vmnmi;
336
337	bzero(&vmnmi, sizeof(vmnmi));
338	vmnmi.cpuid = vcpu;
339
340	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
341}
342
343int
344vm_capability_name2type(const char *capname)
345{
346	int i;
347
348	static struct {
349		const char	*name;
350		int		type;
351	} capstrmap[] = {
352		{ "hlt_exit",		VM_CAP_HALT_EXIT },
353		{ "mtrap_exit",		VM_CAP_MTRAP_EXIT },
354		{ "pause_exit",		VM_CAP_PAUSE_EXIT },
355		{ "unrestricted_guest",	VM_CAP_UNRESTRICTED_GUEST },
356		{ 0 }
357	};
358
359	for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) {
360		if (strcmp(capstrmap[i].name, capname) == 0)
361			return (capstrmap[i].type);
362	}
363
364	return (-1);
365}
366
367int
368vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
369		  int *retval)
370{
371	int error;
372	struct vm_capability vmcap;
373
374	bzero(&vmcap, sizeof(vmcap));
375	vmcap.cpuid = vcpu;
376	vmcap.captype = cap;
377
378	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
379	*retval = vmcap.capval;
380	return (error);
381}
382
383int
384vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
385{
386	struct vm_capability vmcap;
387
388	bzero(&vmcap, sizeof(vmcap));
389	vmcap.cpuid = vcpu;
390	vmcap.captype = cap;
391	vmcap.capval = val;
392
393	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
394}
395
396int
397vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
398{
399	struct vm_pptdev pptdev;
400
401	bzero(&pptdev, sizeof(pptdev));
402	pptdev.bus = bus;
403	pptdev.slot = slot;
404	pptdev.func = func;
405
406	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
407}
408
409int
410vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
411{
412	struct vm_pptdev pptdev;
413
414	bzero(&pptdev, sizeof(pptdev));
415	pptdev.bus = bus;
416	pptdev.slot = slot;
417	pptdev.func = func;
418
419	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
420}
421
422int
423vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
424		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
425{
426	struct vm_pptdev_mmio pptmmio;
427
428	bzero(&pptmmio, sizeof(pptmmio));
429	pptmmio.bus = bus;
430	pptmmio.slot = slot;
431	pptmmio.func = func;
432	pptmmio.gpa = gpa;
433	pptmmio.len = len;
434	pptmmio.hpa = hpa;
435
436	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
437}
438
439int
440vm_setup_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
441	     int destcpu, int vector, int numvec)
442{
443	struct vm_pptdev_msi pptmsi;
444
445	bzero(&pptmsi, sizeof(pptmsi));
446	pptmsi.vcpu = vcpu;
447	pptmsi.bus = bus;
448	pptmsi.slot = slot;
449	pptmsi.func = func;
450	pptmsi.destcpu = destcpu;
451	pptmsi.vector = vector;
452	pptmsi.numvec = numvec;
453
454	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
455}
456
457uint64_t *
458vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
459	     int *ret_entries)
460{
461	int error;
462
463	static struct vm_stats vmstats;
464
465	vmstats.cpuid = vcpu;
466
467	error = ioctl(ctx->fd, VM_STATS, &vmstats);
468	if (error == 0) {
469		if (ret_entries)
470			*ret_entries = vmstats.num_entries;
471		if (ret_tv)
472			*ret_tv = vmstats.tv;
473		return (vmstats.statbuf);
474	} else
475		return (NULL);
476}
477
478const char *
479vm_get_stat_desc(struct vmctx *ctx, int index)
480{
481	int error;
482
483	static struct vm_stat_desc statdesc;
484
485	statdesc.index = index;
486	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
487		return (statdesc.desc);
488	else
489		return (NULL);
490}
491
492/*
493 * From Intel Vol 3a:
494 * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
495 */
496int
497vcpu_reset(struct vmctx *vmctx, int vcpu)
498{
499	int error;
500	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
501	uint32_t desc_access, desc_limit;
502	uint16_t sel;
503
504	zero = 0;
505
506	rflags = 0x2;
507	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
508	if (error)
509		goto done;
510
511	rip = 0xfff0;
512	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
513		goto done;
514
515	cr0 = CR0_NE;
516	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
517		goto done;
518
519	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
520		goto done;
521
522	cr4 = CR4_VMXE;
523	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
524		goto done;
525
526	/*
527	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
528	 */
529	desc_base = 0xffff0000;
530	desc_limit = 0xffff;
531	desc_access = 0x0093;
532	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
533			    desc_base, desc_limit, desc_access);
534	if (error)
535		goto done;
536
537	sel = 0xf000;
538	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
539		goto done;
540
541	/*
542	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
543	 */
544	desc_base = 0;
545	desc_limit = 0xffff;
546	desc_access = 0x0093;
547	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
548			    desc_base, desc_limit, desc_access);
549	if (error)
550		goto done;
551
552	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
553			    desc_base, desc_limit, desc_access);
554	if (error)
555		goto done;
556
557	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
558			    desc_base, desc_limit, desc_access);
559	if (error)
560		goto done;
561
562	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
563			    desc_base, desc_limit, desc_access);
564	if (error)
565		goto done;
566
567	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
568			    desc_base, desc_limit, desc_access);
569	if (error)
570		goto done;
571
572	sel = 0;
573	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
574		goto done;
575	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
576		goto done;
577	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
578		goto done;
579	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
580		goto done;
581	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
582		goto done;
583
584	/* General purpose registers */
585	rdx = 0xf00;
586	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
587		goto done;
588	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
589		goto done;
590	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
591		goto done;
592	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
593		goto done;
594	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
595		goto done;
596	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
597		goto done;
598	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
599		goto done;
600	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
601		goto done;
602
603	/* GDTR, IDTR */
604	desc_base = 0;
605	desc_limit = 0xffff;
606	desc_access = 0;
607	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
608			    desc_base, desc_limit, desc_access);
609	if (error != 0)
610		goto done;
611
612	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
613			    desc_base, desc_limit, desc_access);
614	if (error != 0)
615		goto done;
616
617	/* TR */
618	desc_base = 0;
619	desc_limit = 0xffff;
620	desc_access = 0x0000008b;
621	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
622	if (error)
623		goto done;
624
625	sel = 0;
626	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
627		goto done;
628
629	/* LDTR */
630	desc_base = 0;
631	desc_limit = 0xffff;
632	desc_access = 0x00000082;
633	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
634			    desc_limit, desc_access);
635	if (error)
636		goto done;
637
638	sel = 0;
639	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
640		goto done;
641
642	/* XXX cr2, debug registers */
643
644	error = 0;
645done:
646	return (error);
647}
648