bhyverun.c revision 261088 
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: stable/10/usr.sbin/bhyve/bhyverun.c 261088 2014-01-23 20:21:39Z jhb $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: stable/10/usr.sbin/bhyve/bhyverun.c 261088 2014-01-23 20:21:39Z jhb $");
31
32#include <sys/types.h>
33#include <sys/mman.h>
34#include <sys/time.h>
35
36#include <machine/segments.h>
37
38#include <stdio.h>
39#include <stdlib.h>
40#include <string.h>
41#include <err.h>
42#include <libgen.h>
43#include <unistd.h>
44#include <assert.h>
45#include <errno.h>
46#include <pthread.h>
47#include <pthread_np.h>
48#include <sysexits.h>
49
50#include <machine/vmm.h>
51#include <vmmapi.h>
52
53#include "bhyverun.h"
54#include "acpi.h"
55#include "inout.h"
56#include "dbgport.h"
57#include "legacy_irq.h"
58#include "mem.h"
59#include "mevent.h"
60#include "mptbl.h"
61#include "pci_emul.h"
62#include "pci_lpc.h"
63#include "xmsr.h"
64#include "spinup_ap.h"
65#include "rtc.h"
66
67#define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
68
69#define	VMEXIT_SWITCH		0	/* force vcpu switch in mux mode */
70#define	VMEXIT_CONTINUE		1	/* continue from next instruction */
71#define	VMEXIT_RESTART		2	/* restart current instruction */
72#define	VMEXIT_ABORT		3	/* abort the vm run loop */
73#define	VMEXIT_RESET		4	/* guest machine has reset */
74
75#define MB		(1024UL * 1024)
76#define GB		(1024UL * MB)
77
78typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
79
80char *vmname;
81
82int guest_ncpus;
83
84static int pincpu = -1;
85static int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic;
86static int virtio_msix = 1;
87
88static int foundcpus;
89
90static int strictio;
91
92static int acpi;
93
94static char *progname;
95static const int BSP = 0;
96
97static int cpumask;
98
99static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
100
101struct vm_exit vmexit[VM_MAXCPU];
102
103struct bhyvestats {
104        uint64_t        vmexit_bogus;
105        uint64_t        vmexit_bogus_switch;
106        uint64_t        vmexit_hlt;
107        uint64_t        vmexit_pause;
108        uint64_t        vmexit_mtrap;
109        uint64_t        vmexit_inst_emul;
110        uint64_t        cpu_switch_rotate;
111        uint64_t        cpu_switch_direct;
112        int             io_reset;
113} stats;
114
115struct mt_vmm_info {
116	pthread_t	mt_thr;
117	struct vmctx	*mt_ctx;
118	int		mt_vcpu;
119} mt_vmm_info[VM_MAXCPU];
120
121static void
122usage(int code)
123{
124
125        fprintf(stderr,
126                "Usage: %s [-aehAHIPW] [-g <gdb port>] [-s <pci>] [-S <pci>]\n"
127		"       %*s [-c vcpus] [-p pincpu] [-m mem] [-l <lpc>] <vm>\n"
128		"       -a: local apic is in XAPIC mode (default is X2APIC)\n"
129		"       -A: create an ACPI table\n"
130		"       -g: gdb port\n"
131		"       -c: # cpus (default 1)\n"
132		"       -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
133		"       -H: vmexit from the guest on hlt\n"
134		"       -P: vmexit from the guest on pause\n"
135		"       -W: force virtio to use single-vector MSI\n"
136		"       -e: exit on unhandled I/O access\n"
137		"       -h: help\n"
138		"       -s: <slot,driver,configinfo> PCI slot config\n"
139		"       -S: <slot,driver,configinfo> legacy PCI slot config\n"
140		"       -l: LPC device configuration\n"
141		"       -m: memory size in MB\n",
142		progname, (int)strlen(progname), "");
143
144	exit(code);
145}
146
147void *
148paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
149{
150
151	return (vm_map_gpa(ctx, gaddr, len));
152}
153
154int
155fbsdrun_disable_x2apic(void)
156{
157
158	return (disable_x2apic);
159}
160
161int
162fbsdrun_vmexit_on_pause(void)
163{
164
165	return (guest_vmexit_on_pause);
166}
167
168int
169fbsdrun_vmexit_on_hlt(void)
170{
171
172	return (guest_vmexit_on_hlt);
173}
174
175int
176fbsdrun_virtio_msix(void)
177{
178
179	return (virtio_msix);
180}
181
182static void *
183fbsdrun_start_thread(void *param)
184{
185	char tname[MAXCOMLEN + 1];
186	struct mt_vmm_info *mtp;
187	int vcpu;
188
189	mtp = param;
190	vcpu = mtp->mt_vcpu;
191
192	snprintf(tname, sizeof(tname), "vcpu %d", vcpu);
193	pthread_set_name_np(mtp->mt_thr, tname);
194
195	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
196
197	/* not reached */
198	exit(1);
199	return (NULL);
200}
201
202void
203fbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip)
204{
205	int error;
206
207	if (cpumask & (1 << vcpu)) {
208		fprintf(stderr, "addcpu: attempting to add existing cpu %d\n",
209		    vcpu);
210		exit(1);
211	}
212
213	cpumask |= 1 << vcpu;
214	foundcpus++;
215
216	/*
217	 * Set up the vmexit struct to allow execution to start
218	 * at the given RIP
219	 */
220	vmexit[vcpu].rip = rip;
221	vmexit[vcpu].inst_length = 0;
222
223	mt_vmm_info[vcpu].mt_ctx = ctx;
224	mt_vmm_info[vcpu].mt_vcpu = vcpu;
225
226	error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL,
227	    fbsdrun_start_thread, &mt_vmm_info[vcpu]);
228	assert(error == 0);
229}
230
231static int
232vmexit_catch_reset(void)
233{
234        stats.io_reset++;
235        return (VMEXIT_RESET);
236}
237
238static int
239vmexit_catch_inout(void)
240{
241	return (VMEXIT_ABORT);
242}
243
244static int
245vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
246		     uint32_t eax)
247{
248#if BHYVE_DEBUG
249	/*
250	 * put guest-driven debug here
251	 */
252#endif
253        return (VMEXIT_CONTINUE);
254}
255
256static int
257vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
258{
259	int error;
260	int bytes, port, in, out;
261	uint32_t eax;
262	int vcpu;
263
264	vcpu = *pvcpu;
265
266	port = vme->u.inout.port;
267	bytes = vme->u.inout.bytes;
268	eax = vme->u.inout.eax;
269	in = vme->u.inout.in;
270	out = !in;
271
272	/* We don't deal with these */
273	if (vme->u.inout.string || vme->u.inout.rep)
274		return (VMEXIT_ABORT);
275
276	/* Special case of guest reset */
277	if (out && port == 0x64 && (uint8_t)eax == 0xFE)
278		return (vmexit_catch_reset());
279
280        /* Extra-special case of host notifications */
281        if (out && port == GUEST_NIO_PORT)
282                return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
283
284	error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
285	if (error == 0 && in)
286		error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
287
288	if (error == 0)
289		return (VMEXIT_CONTINUE);
290	else {
291		fprintf(stderr, "Unhandled %s%c 0x%04x\n",
292			in ? "in" : "out",
293			bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
294		return (vmexit_catch_inout());
295	}
296}
297
298static int
299vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
300{
301	fprintf(stderr, "vm exit rdmsr 0x%x, cpu %d\n", vme->u.msr.code,
302	    *pvcpu);
303	return (VMEXIT_ABORT);
304}
305
306static int
307vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
308{
309	int newcpu;
310	int retval = VMEXIT_CONTINUE;
311
312	newcpu = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code,vme->u.msr.wval);
313
314        return (retval);
315}
316
317static int
318vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
319{
320	int newcpu;
321	int retval = VMEXIT_CONTINUE;
322
323	newcpu = spinup_ap(ctx, *pvcpu,
324			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
325
326	return (retval);
327}
328
329static int
330vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
331{
332
333	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
334	fprintf(stderr, "\treason\t\tVMX\n");
335	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
336	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
337	fprintf(stderr, "\terror\t\t%d\n", vmexit->u.vmx.error);
338	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
339	fprintf(stderr, "\tqualification\t0x%016lx\n",
340	    vmexit->u.vmx.exit_qualification);
341
342	return (VMEXIT_ABORT);
343}
344
345static int
346vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
347{
348
349	stats.vmexit_bogus++;
350
351	return (VMEXIT_RESTART);
352}
353
354static int
355vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
356{
357
358	stats.vmexit_hlt++;
359
360	/*
361	 * Just continue execution with the next instruction. We use
362	 * the HLT VM exit as a way to be friendly with the host
363	 * scheduler.
364	 */
365	return (VMEXIT_CONTINUE);
366}
367
368static int
369vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
370{
371
372	stats.vmexit_pause++;
373
374	return (VMEXIT_CONTINUE);
375}
376
377static int
378vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
379{
380
381	stats.vmexit_mtrap++;
382
383	return (VMEXIT_RESTART);
384}
385
386static int
387vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
388{
389	int err;
390	stats.vmexit_inst_emul++;
391
392	err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa,
393			  &vmexit->u.inst_emul.vie);
394
395	if (err) {
396		if (err == EINVAL) {
397			fprintf(stderr,
398			    "Failed to emulate instruction at 0x%lx\n",
399			    vmexit->rip);
400		} else if (err == ESRCH) {
401			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
402			    vmexit->u.inst_emul.gpa);
403		}
404
405		return (VMEXIT_ABORT);
406	}
407
408	return (VMEXIT_CONTINUE);
409}
410
411static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
412	[VM_EXITCODE_INOUT]  = vmexit_inout,
413	[VM_EXITCODE_VMX]    = vmexit_vmx,
414	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
415	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
416	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
417	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
418	[VM_EXITCODE_INST_EMUL] = vmexit_inst_emul,
419	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
420};
421
422static void
423vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
424{
425	cpuset_t mask;
426	int error, rc, prevcpu;
427	enum vm_exitcode exitcode;
428
429	if (pincpu >= 0) {
430		CPU_ZERO(&mask);
431		CPU_SET(pincpu + vcpu, &mask);
432		error = pthread_setaffinity_np(pthread_self(),
433					       sizeof(mask), &mask);
434		assert(error == 0);
435	}
436
437	while (1) {
438		error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
439		if (error != 0) {
440			/*
441			 * It is possible that 'vmmctl' or some other process
442			 * has transitioned the vcpu to CANNOT_RUN state right
443			 * before we tried to transition it to RUNNING.
444			 *
445			 * This is expected to be temporary so just retry.
446			 */
447			if (errno == EBUSY)
448				continue;
449			else
450				break;
451		}
452
453		prevcpu = vcpu;
454
455		exitcode = vmexit[vcpu].exitcode;
456		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
457			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
458			    exitcode);
459			exit(1);
460		}
461
462                rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
463
464		switch (rc) {
465		case VMEXIT_CONTINUE:
466                        rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
467			break;
468		case VMEXIT_RESTART:
469                        rip = vmexit[vcpu].rip;
470			break;
471		case VMEXIT_RESET:
472			exit(0);
473		default:
474			exit(1);
475		}
476	}
477	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
478}
479
480static int
481num_vcpus_allowed(struct vmctx *ctx)
482{
483	int tmp, error;
484
485	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
486
487	/*
488	 * The guest is allowed to spinup more than one processor only if the
489	 * UNRESTRICTED_GUEST capability is available.
490	 */
491	if (error == 0)
492		return (VM_MAXCPU);
493	else
494		return (1);
495}
496
497void
498fbsdrun_set_capabilities(struct vmctx *ctx, int cpu)
499{
500	int err, tmp;
501
502	if (fbsdrun_vmexit_on_hlt()) {
503		err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp);
504		if (err < 0) {
505			fprintf(stderr, "VM exit on HLT not supported\n");
506			exit(1);
507		}
508		vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1);
509		if (cpu == BSP)
510			handler[VM_EXITCODE_HLT] = vmexit_hlt;
511	}
512
513        if (fbsdrun_vmexit_on_pause()) {
514		/*
515		 * pause exit support required for this mode
516		 */
517		err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp);
518		if (err < 0) {
519			fprintf(stderr,
520			    "SMP mux requested, no pause support\n");
521			exit(1);
522		}
523		vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1);
524		if (cpu == BSP)
525			handler[VM_EXITCODE_PAUSE] = vmexit_pause;
526        }
527
528	if (fbsdrun_disable_x2apic())
529		err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED);
530	else
531		err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED);
532
533	if (err) {
534		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
535		exit(1);
536	}
537
538	vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1);
539}
540
541int
542main(int argc, char *argv[])
543{
544	int c, error, gdb_port, err, bvmcons;
545	int max_vcpus;
546	struct vmctx *ctx;
547	uint64_t rip;
548	size_t memsize;
549
550	bvmcons = 0;
551	progname = basename(argv[0]);
552	gdb_port = 0;
553	guest_ncpus = 1;
554	memsize = 256 * MB;
555
556	while ((c = getopt(argc, argv, "abehAHIPWp:g:c:s:S:m:l:")) != -1) {
557		switch (c) {
558		case 'a':
559			disable_x2apic = 1;
560			break;
561		case 'A':
562			acpi = 1;
563			break;
564		case 'b':
565			bvmcons = 1;
566			break;
567		case 'p':
568			pincpu = atoi(optarg);
569			break;
570                case 'c':
571			guest_ncpus = atoi(optarg);
572			break;
573		case 'g':
574			gdb_port = atoi(optarg);
575			break;
576		case 'l':
577			if (lpc_device_parse(optarg) != 0) {
578				errx(EX_USAGE, "invalid lpc device "
579				    "configuration '%s'", optarg);
580			}
581			break;
582		case 's':
583			if (pci_parse_slot(optarg, 0) != 0)
584				exit(1);
585			else
586				break;
587		case 'S':
588			if (pci_parse_slot(optarg, 1) != 0)
589				exit(1);
590			else
591				break;
592                case 'm':
593			error = vm_parse_memsize(optarg, &memsize);
594			if (error)
595				errx(EX_USAGE, "invalid memsize '%s'", optarg);
596			break;
597		case 'H':
598			guest_vmexit_on_hlt = 1;
599			break;
600		case 'I':
601			/*
602			 * The "-I" option was used to add an ioapic to the
603			 * virtual machine.
604			 *
605			 * An ioapic is now provided unconditionally for each
606			 * virtual machine and this option is now deprecated.
607			 */
608			break;
609		case 'P':
610			guest_vmexit_on_pause = 1;
611			break;
612		case 'e':
613			strictio = 1;
614			break;
615		case 'W':
616			virtio_msix = 0;
617			break;
618		case 'h':
619			usage(0);
620		default:
621			usage(1);
622		}
623	}
624	argc -= optind;
625	argv += optind;
626
627	if (argc != 1)
628		usage(1);
629
630	vmname = argv[0];
631
632	ctx = vm_open(vmname);
633	if (ctx == NULL) {
634		perror("vm_open");
635		exit(1);
636	}
637
638	max_vcpus = num_vcpus_allowed(ctx);
639	if (guest_ncpus > max_vcpus) {
640		fprintf(stderr, "%d vCPUs requested but only %d available\n",
641			guest_ncpus, max_vcpus);
642		exit(1);
643	}
644
645	fbsdrun_set_capabilities(ctx, BSP);
646
647	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
648	if (err) {
649		fprintf(stderr, "Unable to setup memory (%d)\n", err);
650		exit(1);
651	}
652
653	init_mem();
654	init_inout();
655	legacy_irq_init();
656
657	rtc_init(ctx);
658
659	/*
660	 * Exit if a device emulation finds an error in it's initilization
661	 */
662	if (init_pci(ctx) != 0)
663		exit(1);
664
665	if (gdb_port != 0)
666		init_dbgport(gdb_port);
667
668	if (bvmcons)
669		init_bvmcons();
670
671	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
672	assert(error == 0);
673
674	/*
675	 * build the guest tables, MP etc.
676	 */
677	mptable_build(ctx, guest_ncpus);
678
679	if (acpi) {
680		error = acpi_build(ctx, guest_ncpus);
681		assert(error == 0);
682	}
683
684	/*
685	 * Change the proc title to include the VM name.
686	 */
687	setproctitle("%s", vmname);
688
689	/*
690	 * Add CPU 0
691	 */
692	fbsdrun_addcpu(ctx, BSP, rip);
693
694	/*
695	 * Head off to the main event dispatch loop
696	 */
697	mevent_dispatch();
698
699	exit(1);
700}
701