vmmapi.c revision 239025 
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#define BIOS_ROM_BASE		(0xf0000)
53#define BIOS_ROM_SIZE		(0x10000)
54
55struct vmctx {
56	int	fd;
57	char	*name;
58};
59
60#define	CREATE(x)  sysctlbyname("hw.vmm.create", NULL, NULL, (x), strlen((x)))
61#define	DESTROY(x) sysctlbyname("hw.vmm.destroy", NULL, NULL, (x), strlen((x)))
62
63static int
64vm_device_open(const char *name)
65{
66        int fd, len;
67        char *vmfile;
68
69	len = strlen("/dev/vmm/") + strlen(name) + 1;
70	vmfile = malloc(len);
71	assert(vmfile != NULL);
72	snprintf(vmfile, len, "/dev/vmm/%s", name);
73
74        /* Open the device file */
75        fd = open(vmfile, O_RDWR, 0);
76
77	free(vmfile);
78        return (fd);
79}
80
81int
82vm_create(const char *name)
83{
84
85	return (CREATE((char *)name));
86}
87
88struct vmctx *
89vm_open(const char *name)
90{
91	struct vmctx *vm;
92
93	vm = malloc(sizeof(struct vmctx) + strlen(name) + 1);
94	assert(vm != NULL);
95
96	vm->fd = -1;
97	vm->name = (char *)(vm + 1);
98	strcpy(vm->name, name);
99
100	if ((vm->fd = vm_device_open(vm->name)) < 0)
101		goto err;
102
103	return (vm);
104err:
105	vm_destroy(vm);
106	return (NULL);
107}
108
109void
110vm_destroy(struct vmctx *vm)
111{
112	assert(vm != NULL);
113
114	DESTROY(vm->name);
115	if (vm->fd >= 0)
116		close(vm->fd);
117	free(vm);
118}
119
120int
121vm_get_memory_seg(struct vmctx *ctx, vm_paddr_t gpa,
122		  vm_paddr_t *ret_hpa, size_t *ret_len)
123{
124	int error;
125	struct vm_memory_segment seg;
126
127	bzero(&seg, sizeof(seg));
128	seg.gpa = gpa;
129	error = ioctl(ctx->fd, VM_GET_MEMORY_SEG, &seg);
130	*ret_hpa = seg.hpa;
131	*ret_len = seg.len;
132	return (error);
133}
134
135int
136vm_setup_memory(struct vmctx *ctx, vm_paddr_t gpa, size_t len, char **mapaddr)
137{
138	int error;
139	struct vm_memory_segment seg;
140
141	/*
142	 * Create and optionally map 'len' bytes of memory at guest
143	 * physical address 'gpa'
144	 */
145	bzero(&seg, sizeof(seg));
146	seg.gpa = gpa;
147	seg.len = len;
148	error = ioctl(ctx->fd, VM_MAP_MEMORY, &seg);
149	if (error == 0 && mapaddr != NULL) {
150		*mapaddr = mmap(NULL, len, PROT_READ | PROT_WRITE, MAP_SHARED,
151				ctx->fd, gpa);
152	}
153	return (error);
154}
155
156char *
157vm_map_memory(struct vmctx *ctx, vm_paddr_t gpa, size_t len)
158{
159
160	/* Map 'len' bytes of memory at guest physical address 'gpa' */
161	return ((char *)mmap(NULL, len, PROT_READ | PROT_WRITE, MAP_SHARED,
162		     ctx->fd, gpa));
163}
164
165int
166vm_set_desc(struct vmctx *ctx, int vcpu, int reg,
167	    uint64_t base, uint32_t limit, uint32_t access)
168{
169	int error;
170	struct vm_seg_desc vmsegdesc;
171
172	bzero(&vmsegdesc, sizeof(vmsegdesc));
173	vmsegdesc.cpuid = vcpu;
174	vmsegdesc.regnum = reg;
175	vmsegdesc.desc.base = base;
176	vmsegdesc.desc.limit = limit;
177	vmsegdesc.desc.access = access;
178
179	error = ioctl(ctx->fd, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc);
180	return (error);
181}
182
183int
184vm_get_desc(struct vmctx *ctx, int vcpu, int reg,
185	    uint64_t *base, uint32_t *limit, uint32_t *access)
186{
187	int error;
188	struct vm_seg_desc vmsegdesc;
189
190	bzero(&vmsegdesc, sizeof(vmsegdesc));
191	vmsegdesc.cpuid = vcpu;
192	vmsegdesc.regnum = reg;
193
194	error = ioctl(ctx->fd, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc);
195	if (error == 0) {
196		*base = vmsegdesc.desc.base;
197		*limit = vmsegdesc.desc.limit;
198		*access = vmsegdesc.desc.access;
199	}
200	return (error);
201}
202
203int
204vm_set_register(struct vmctx *ctx, int vcpu, int reg, uint64_t val)
205{
206	int error;
207	struct vm_register vmreg;
208
209	bzero(&vmreg, sizeof(vmreg));
210	vmreg.cpuid = vcpu;
211	vmreg.regnum = reg;
212	vmreg.regval = val;
213
214	error = ioctl(ctx->fd, VM_SET_REGISTER, &vmreg);
215	return (error);
216}
217
218int
219vm_get_register(struct vmctx *ctx, int vcpu, int reg, uint64_t *ret_val)
220{
221	int error;
222	struct vm_register vmreg;
223
224	bzero(&vmreg, sizeof(vmreg));
225	vmreg.cpuid = vcpu;
226	vmreg.regnum = reg;
227
228	error = ioctl(ctx->fd, VM_GET_REGISTER, &vmreg);
229	*ret_val = vmreg.regval;
230	return (error);
231}
232
233int
234vm_get_pinning(struct vmctx *ctx, int vcpu, int *host_cpuid)
235{
236	int error;
237	struct vm_pin vmpin;
238
239	bzero(&vmpin, sizeof(vmpin));
240	vmpin.vm_cpuid = vcpu;
241
242	error = ioctl(ctx->fd, VM_GET_PINNING, &vmpin);
243	*host_cpuid = vmpin.host_cpuid;
244	return (error);
245}
246
247int
248vm_set_pinning(struct vmctx *ctx, int vcpu, int host_cpuid)
249{
250	int error;
251	struct vm_pin vmpin;
252
253	bzero(&vmpin, sizeof(vmpin));
254	vmpin.vm_cpuid = vcpu;
255	vmpin.host_cpuid = host_cpuid;
256
257	error = ioctl(ctx->fd, VM_SET_PINNING, &vmpin);
258	return (error);
259}
260
261int
262vm_run(struct vmctx *ctx, int vcpu, uint64_t rip, struct vm_exit *vmexit)
263{
264	int error;
265	struct vm_run vmrun;
266
267	bzero(&vmrun, sizeof(vmrun));
268	vmrun.cpuid = vcpu;
269	vmrun.rip = rip;
270
271	error = ioctl(ctx->fd, VM_RUN, &vmrun);
272	bcopy(&vmrun.vm_exit, vmexit, sizeof(struct vm_exit));
273	return (error);
274}
275
276static int
277vm_inject_event_real(struct vmctx *ctx, int vcpu, enum vm_event_type type,
278		     int vector, int error_code, int error_code_valid)
279{
280	struct vm_event ev;
281
282	bzero(&ev, sizeof(ev));
283	ev.cpuid = vcpu;
284	ev.type = type;
285	ev.vector = vector;
286	ev.error_code = error_code;
287	ev.error_code_valid = error_code_valid;
288
289	return (ioctl(ctx->fd, VM_INJECT_EVENT, &ev));
290}
291
292int
293vm_inject_event(struct vmctx *ctx, int vcpu, enum vm_event_type type,
294		int vector)
295{
296
297	return (vm_inject_event_real(ctx, vcpu, type, vector, 0, 0));
298}
299
300int
301vm_inject_event2(struct vmctx *ctx, int vcpu, enum vm_event_type type,
302		 int vector, int error_code)
303{
304
305	return (vm_inject_event_real(ctx, vcpu, type, vector, error_code, 1));
306}
307
308int
309vm_build_tables(struct vmctx *ctxt, int ncpu, void *oemtbl, int oemtblsz)
310{
311
312	return (vm_build_mptable(ctxt, BIOS_ROM_BASE, BIOS_ROM_SIZE, ncpu,
313				 oemtbl, oemtblsz));
314}
315
316int
317vm_lapic_irq(struct vmctx *ctx, int vcpu, int vector)
318{
319	struct vm_lapic_irq vmirq;
320
321	bzero(&vmirq, sizeof(vmirq));
322	vmirq.cpuid = vcpu;
323	vmirq.vector = vector;
324
325	return (ioctl(ctx->fd, VM_LAPIC_IRQ, &vmirq));
326}
327
328int
329vm_inject_nmi(struct vmctx *ctx, int vcpu)
330{
331	struct vm_nmi vmnmi;
332
333	bzero(&vmnmi, sizeof(vmnmi));
334	vmnmi.cpuid = vcpu;
335
336	return (ioctl(ctx->fd, VM_INJECT_NMI, &vmnmi));
337}
338
339int
340vm_capability_name2type(const char *capname)
341{
342	int i;
343
344	static struct {
345		const char	*name;
346		int		type;
347	} capstrmap[] = {
348		{ "hlt_exit",		VM_CAP_HALT_EXIT },
349		{ "mtrap_exit",		VM_CAP_MTRAP_EXIT },
350		{ "pause_exit",		VM_CAP_PAUSE_EXIT },
351		{ "unrestricted_guest",	VM_CAP_UNRESTRICTED_GUEST },
352		{ 0 }
353	};
354
355	for (i = 0; capstrmap[i].name != NULL && capname != NULL; i++) {
356		if (strcmp(capstrmap[i].name, capname) == 0)
357			return (capstrmap[i].type);
358	}
359
360	return (-1);
361}
362
363int
364vm_get_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap,
365		  int *retval)
366{
367	int error;
368	struct vm_capability vmcap;
369
370	bzero(&vmcap, sizeof(vmcap));
371	vmcap.cpuid = vcpu;
372	vmcap.captype = cap;
373
374	error = ioctl(ctx->fd, VM_GET_CAPABILITY, &vmcap);
375	*retval = vmcap.capval;
376	return (error);
377}
378
379int
380vm_set_capability(struct vmctx *ctx, int vcpu, enum vm_cap_type cap, int val)
381{
382	struct vm_capability vmcap;
383
384	bzero(&vmcap, sizeof(vmcap));
385	vmcap.cpuid = vcpu;
386	vmcap.captype = cap;
387	vmcap.capval = val;
388
389	return (ioctl(ctx->fd, VM_SET_CAPABILITY, &vmcap));
390}
391
392int
393vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
394{
395	struct vm_pptdev pptdev;
396
397	bzero(&pptdev, sizeof(pptdev));
398	pptdev.bus = bus;
399	pptdev.slot = slot;
400	pptdev.func = func;
401
402	return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev));
403}
404
405int
406vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func)
407{
408	struct vm_pptdev pptdev;
409
410	bzero(&pptdev, sizeof(pptdev));
411	pptdev.bus = bus;
412	pptdev.slot = slot;
413	pptdev.func = func;
414
415	return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev));
416}
417
418int
419vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func,
420		   vm_paddr_t gpa, size_t len, vm_paddr_t hpa)
421{
422	struct vm_pptdev_mmio pptmmio;
423
424	bzero(&pptmmio, sizeof(pptmmio));
425	pptmmio.bus = bus;
426	pptmmio.slot = slot;
427	pptmmio.func = func;
428	pptmmio.gpa = gpa;
429	pptmmio.len = len;
430	pptmmio.hpa = hpa;
431
432	return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio));
433}
434
435int
436vm_setup_msi(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
437	     int destcpu, int vector, int numvec)
438{
439	struct vm_pptdev_msi pptmsi;
440
441	bzero(&pptmsi, sizeof(pptmsi));
442	pptmsi.vcpu = vcpu;
443	pptmsi.bus = bus;
444	pptmsi.slot = slot;
445	pptmsi.func = func;
446	pptmsi.destcpu = destcpu;
447	pptmsi.vector = vector;
448	pptmsi.numvec = numvec;
449
450	return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi));
451}
452
453int
454vm_setup_msix(struct vmctx *ctx, int vcpu, int bus, int slot, int func,
455	      int idx, uint32_t msg, uint32_t vector_control, uint64_t addr)
456{
457	struct vm_pptdev_msix pptmsix;
458
459	bzero(&pptmsix, sizeof(pptmsix));
460	pptmsix.vcpu = vcpu;
461	pptmsix.bus = bus;
462	pptmsix.slot = slot;
463	pptmsix.func = func;
464	pptmsix.idx = idx;
465	pptmsix.msg = msg;
466	pptmsix.addr = addr;
467	pptmsix.vector_control = vector_control;
468
469	return ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix);
470}
471
472uint64_t *
473vm_get_stats(struct vmctx *ctx, int vcpu, struct timeval *ret_tv,
474	     int *ret_entries)
475{
476	int error;
477
478	static struct vm_stats vmstats;
479
480	vmstats.cpuid = vcpu;
481
482	error = ioctl(ctx->fd, VM_STATS, &vmstats);
483	if (error == 0) {
484		if (ret_entries)
485			*ret_entries = vmstats.num_entries;
486		if (ret_tv)
487			*ret_tv = vmstats.tv;
488		return (vmstats.statbuf);
489	} else
490		return (NULL);
491}
492
493const char *
494vm_get_stat_desc(struct vmctx *ctx, int index)
495{
496	static struct vm_stat_desc statdesc;
497
498	statdesc.index = index;
499	if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0)
500		return (statdesc.desc);
501	else
502		return (NULL);
503}
504
505/*
506 * From Intel Vol 3a:
507 * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT
508 */
509int
510vcpu_reset(struct vmctx *vmctx, int vcpu)
511{
512	int error;
513	uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx;
514	uint32_t desc_access, desc_limit;
515	uint16_t sel;
516
517	zero = 0;
518
519	rflags = 0x2;
520	error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RFLAGS, rflags);
521	if (error)
522		goto done;
523
524	rip = 0xfff0;
525	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RIP, rip)) != 0)
526		goto done;
527
528	cr0 = CR0_NE;
529	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR0, cr0)) != 0)
530		goto done;
531
532	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR3, zero)) != 0)
533		goto done;
534
535	cr4 = 0;
536	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CR4, cr4)) != 0)
537		goto done;
538
539	/*
540	 * CS: present, r/w, accessed, 16-bit, byte granularity, usable
541	 */
542	desc_base = 0xffff0000;
543	desc_limit = 0xffff;
544	desc_access = 0x0093;
545	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_CS,
546			    desc_base, desc_limit, desc_access);
547	if (error)
548		goto done;
549
550	sel = 0xf000;
551	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_CS, sel)) != 0)
552		goto done;
553
554	/*
555	 * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity
556	 */
557	desc_base = 0;
558	desc_limit = 0xffff;
559	desc_access = 0x0093;
560	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_SS,
561			    desc_base, desc_limit, desc_access);
562	if (error)
563		goto done;
564
565	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_DS,
566			    desc_base, desc_limit, desc_access);
567	if (error)
568		goto done;
569
570	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_ES,
571			    desc_base, desc_limit, desc_access);
572	if (error)
573		goto done;
574
575	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_FS,
576			    desc_base, desc_limit, desc_access);
577	if (error)
578		goto done;
579
580	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GS,
581			    desc_base, desc_limit, desc_access);
582	if (error)
583		goto done;
584
585	sel = 0;
586	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_SS, sel)) != 0)
587		goto done;
588	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_DS, sel)) != 0)
589		goto done;
590	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_ES, sel)) != 0)
591		goto done;
592	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_FS, sel)) != 0)
593		goto done;
594	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_GS, sel)) != 0)
595		goto done;
596
597	/* General purpose registers */
598	rdx = 0xf00;
599	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RAX, zero)) != 0)
600		goto done;
601	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBX, zero)) != 0)
602		goto done;
603	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RCX, zero)) != 0)
604		goto done;
605	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDX, rdx)) != 0)
606		goto done;
607	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSI, zero)) != 0)
608		goto done;
609	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RDI, zero)) != 0)
610		goto done;
611	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RBP, zero)) != 0)
612		goto done;
613	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_RSP, zero)) != 0)
614		goto done;
615
616	/* GDTR, IDTR */
617	desc_base = 0;
618	desc_limit = 0xffff;
619	desc_access = 0;
620	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_GDTR,
621			    desc_base, desc_limit, desc_access);
622	if (error != 0)
623		goto done;
624
625	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_IDTR,
626			    desc_base, desc_limit, desc_access);
627	if (error != 0)
628		goto done;
629
630	/* TR */
631	desc_base = 0;
632	desc_limit = 0xffff;
633	desc_access = 0x0000008b;
634	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_TR, 0, 0, desc_access);
635	if (error)
636		goto done;
637
638	sel = 0;
639	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_TR, sel)) != 0)
640		goto done;
641
642	/* LDTR */
643	desc_base = 0;
644	desc_limit = 0xffff;
645	desc_access = 0x00000082;
646	error = vm_set_desc(vmctx, vcpu, VM_REG_GUEST_LDTR, desc_base,
647			    desc_limit, desc_access);
648	if (error)
649		goto done;
650
651	sel = 0;
652	if ((error = vm_set_register(vmctx, vcpu, VM_REG_GUEST_LDTR, 0)) != 0)
653		goto done;
654
655	/* XXX cr2, debug registers */
656
657	error = 0;
658done:
659	return (error);
660}
661