bhyverun.c revision 256062 
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: head/usr.sbin/bhyve/bhyverun.c 256062 2013-10-04 23:29:07Z grehan $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: head/usr.sbin/bhyve/bhyverun.c 256062 2013-10-04 23:29:07Z grehan $");
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 <libgen.h>
41#include <unistd.h>
42#include <assert.h>
43#include <errno.h>
44#include <pthread.h>
45#include <pthread_np.h>
46
47#include <machine/vmm.h>
48#include <vmmapi.h>
49
50#include "bhyverun.h"
51#include "acpi.h"
52#include "inout.h"
53#include "dbgport.h"
54#include "mem.h"
55#include "mevent.h"
56#include "mptbl.h"
57#include "pci_emul.h"
58#include "xmsr.h"
59#include "ioapic.h"
60#include "spinup_ap.h"
61#include "rtc.h"
62
63#define GUEST_NIO_PORT		0x488	/* guest upcalls via i/o port */
64
65#define	VMEXIT_SWITCH		0	/* force vcpu switch in mux mode */
66#define	VMEXIT_CONTINUE		1	/* continue from next instruction */
67#define	VMEXIT_RESTART		2	/* restart current instruction */
68#define	VMEXIT_ABORT		3	/* abort the vm run loop */
69#define	VMEXIT_RESET		4	/* guest machine has reset */
70
71#define MB		(1024UL * 1024)
72#define GB		(1024UL * MB)
73
74typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu);
75
76char *vmname;
77
78int guest_ncpus;
79
80static int pincpu = -1;
81static int guest_vmexit_on_hlt, guest_vmexit_on_pause, disable_x2apic;
82
83static int foundcpus;
84
85static int strictio;
86
87static int acpi;
88
89static char *progname;
90static const int BSP = 0;
91
92static int cpumask;
93
94static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip);
95
96struct vm_exit vmexit[VM_MAXCPU];
97
98struct bhyvestats {
99        uint64_t        vmexit_bogus;
100        uint64_t        vmexit_bogus_switch;
101        uint64_t        vmexit_hlt;
102        uint64_t        vmexit_pause;
103        uint64_t        vmexit_mtrap;
104        uint64_t        vmexit_paging;
105        uint64_t        cpu_switch_rotate;
106        uint64_t        cpu_switch_direct;
107        int             io_reset;
108} stats;
109
110struct mt_vmm_info {
111	pthread_t	mt_thr;
112	struct vmctx	*mt_ctx;
113	int		mt_vcpu;
114} mt_vmm_info[VM_MAXCPU];
115
116static void
117usage(int code)
118{
119
120        fprintf(stderr,
121                "Usage: %s [-aehAHIP][-g <gdb port>][-s <pci>][-S <pci>]"
122		"[-c vcpus][-p pincpu][-m mem]"
123		" <vmname>\n"
124		"       -a: local apic is in XAPIC mode (default is X2APIC)\n"
125		"       -A: create an ACPI table\n"
126		"       -g: gdb port (default is %d and 0 means don't open)\n"
127		"       -c: # cpus (default 1)\n"
128		"       -p: pin vcpu 'n' to host cpu 'pincpu + n'\n"
129		"       -H: vmexit from the guest on hlt\n"
130		"       -I: present an ioapic to the guest\n"
131		"       -P: vmexit from the guest on pause\n"
132		"	-e: exit on unhandled i/o access\n"
133		"       -h: help\n"
134		"       -s: <slot,driver,configinfo> PCI slot config\n"
135		"       -S: <slot,driver,configinfo> legacy PCI slot config\n"
136		"       -m: memory size in MB\n",
137		progname, DEFAULT_GDB_PORT);
138
139	exit(code);
140}
141
142void *
143paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len)
144{
145
146	return (vm_map_gpa(ctx, gaddr, len));
147}
148
149int
150fbsdrun_disable_x2apic(void)
151{
152
153	return (disable_x2apic);
154}
155
156int
157fbsdrun_vmexit_on_pause(void)
158{
159
160	return (guest_vmexit_on_pause);
161}
162
163int
164fbsdrun_vmexit_on_hlt(void)
165{
166
167	return (guest_vmexit_on_hlt);
168}
169
170static void *
171fbsdrun_start_thread(void *param)
172{
173	char tname[MAXCOMLEN + 1];
174	struct mt_vmm_info *mtp;
175	int vcpu;
176
177	mtp = param;
178	vcpu = mtp->mt_vcpu;
179
180	snprintf(tname, sizeof(tname), "%s vcpu %d", vmname, vcpu);
181	pthread_set_name_np(mtp->mt_thr, tname);
182
183	vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip);
184
185	/* not reached */
186	exit(1);
187	return (NULL);
188}
189
190void
191fbsdrun_addcpu(struct vmctx *ctx, int vcpu, uint64_t rip)
192{
193	int error;
194
195	if (cpumask & (1 << vcpu)) {
196		fprintf(stderr, "addcpu: attempting to add existing cpu %d\n",
197		    vcpu);
198		exit(1);
199	}
200
201	cpumask |= 1 << vcpu;
202	foundcpus++;
203
204	/*
205	 * Set up the vmexit struct to allow execution to start
206	 * at the given RIP
207	 */
208	vmexit[vcpu].rip = rip;
209	vmexit[vcpu].inst_length = 0;
210
211	if (vcpu == BSP) {
212		mt_vmm_info[vcpu].mt_ctx = ctx;
213		mt_vmm_info[vcpu].mt_vcpu = vcpu;
214
215		error = pthread_create(&mt_vmm_info[vcpu].mt_thr, NULL,
216				fbsdrun_start_thread, &mt_vmm_info[vcpu]);
217		assert(error == 0);
218	}
219}
220
221static int
222fbsdrun_get_next_cpu(int curcpu)
223{
224
225	/*
226	 * Get the next available CPU. Assumes they arrive
227	 * in ascending order with no gaps.
228	 */
229	return ((curcpu + 1) % foundcpus);
230}
231
232static int
233vmexit_catch_reset(void)
234{
235        stats.io_reset++;
236        return (VMEXIT_RESET);
237}
238
239static int
240vmexit_catch_inout(void)
241{
242	return (VMEXIT_ABORT);
243}
244
245static int
246vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu,
247		     uint32_t eax)
248{
249#if BHYVE_DEBUG
250	/*
251	 * put guest-driven debug here
252	 */
253#endif
254        return (VMEXIT_CONTINUE);
255}
256
257static int
258vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
259{
260	int error;
261	int bytes, port, in, out;
262	uint32_t eax;
263	int vcpu;
264
265	vcpu = *pvcpu;
266
267	port = vme->u.inout.port;
268	bytes = vme->u.inout.bytes;
269	eax = vme->u.inout.eax;
270	in = vme->u.inout.in;
271	out = !in;
272
273	/* We don't deal with these */
274	if (vme->u.inout.string || vme->u.inout.rep)
275		return (VMEXIT_ABORT);
276
277	/* Special case of guest reset */
278	if (out && port == 0x64 && (uint8_t)eax == 0xFE)
279		return (vmexit_catch_reset());
280
281        /* Extra-special case of host notifications */
282        if (out && port == GUEST_NIO_PORT)
283                return (vmexit_handle_notify(ctx, vme, pvcpu, eax));
284
285	error = emulate_inout(ctx, vcpu, in, port, bytes, &eax, strictio);
286	if (error == 0 && in)
287		error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, eax);
288
289	if (error == 0)
290		return (VMEXIT_CONTINUE);
291	else {
292		fprintf(stderr, "Unhandled %s%c 0x%04x\n",
293			in ? "in" : "out",
294			bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port);
295		return (vmexit_catch_inout());
296	}
297}
298
299static int
300vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
301{
302	fprintf(stderr, "vm exit rdmsr 0x%x, cpu %d\n", vme->u.msr.code,
303	    *pvcpu);
304	return (VMEXIT_ABORT);
305}
306
307static int
308vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
309{
310	int newcpu;
311	int retval = VMEXIT_CONTINUE;
312
313	newcpu = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code,vme->u.msr.wval);
314
315        return (retval);
316}
317
318static int
319vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu)
320{
321	int newcpu;
322	int retval = VMEXIT_CONTINUE;
323
324	newcpu = spinup_ap(ctx, *pvcpu,
325			   vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip);
326
327	return (retval);
328}
329
330static int
331vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
332{
333
334	fprintf(stderr, "vm exit[%d]\n", *pvcpu);
335	fprintf(stderr, "\treason\t\tVMX\n");
336	fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip);
337	fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length);
338	fprintf(stderr, "\terror\t\t%d\n", vmexit->u.vmx.error);
339	fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason);
340	fprintf(stderr, "\tqualification\t0x%016lx\n",
341	    vmexit->u.vmx.exit_qualification);
342
343	return (VMEXIT_ABORT);
344}
345
346static int
347vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
348{
349
350	stats.vmexit_bogus++;
351
352	return (VMEXIT_RESTART);
353}
354
355static int
356vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
357{
358
359	stats.vmexit_hlt++;
360
361	/*
362	 * Just continue execution with the next instruction. We use
363	 * the HLT VM exit as a way to be friendly with the host
364	 * scheduler.
365	 */
366	return (VMEXIT_CONTINUE);
367}
368
369static int
370vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
371{
372
373	stats.vmexit_pause++;
374
375	return (VMEXIT_CONTINUE);
376}
377
378static int
379vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
380{
381
382	stats.vmexit_mtrap++;
383
384	return (VMEXIT_RESTART);
385}
386
387static int
388vmexit_paging(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu)
389{
390	int err;
391	stats.vmexit_paging++;
392
393	err = emulate_mem(ctx, *pvcpu, vmexit->u.paging.gpa,
394			  &vmexit->u.paging.vie);
395
396	if (err) {
397		if (err == EINVAL) {
398			fprintf(stderr,
399			    "Failed to emulate instruction at 0x%lx\n",
400			    vmexit->rip);
401		} else if (err == ESRCH) {
402			fprintf(stderr, "Unhandled memory access to 0x%lx\n",
403			    vmexit->u.paging.gpa);
404		}
405
406		return (VMEXIT_ABORT);
407	}
408
409	return (VMEXIT_CONTINUE);
410}
411
412static vmexit_handler_t handler[VM_EXITCODE_MAX] = {
413	[VM_EXITCODE_INOUT]  = vmexit_inout,
414	[VM_EXITCODE_VMX]    = vmexit_vmx,
415	[VM_EXITCODE_BOGUS]  = vmexit_bogus,
416	[VM_EXITCODE_RDMSR]  = vmexit_rdmsr,
417	[VM_EXITCODE_WRMSR]  = vmexit_wrmsr,
418	[VM_EXITCODE_MTRAP]  = vmexit_mtrap,
419	[VM_EXITCODE_PAGING] = vmexit_paging,
420	[VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap,
421};
422
423static void
424vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip)
425{
426	cpuset_t mask;
427	int error, rc, prevcpu;
428	enum vm_exitcode exitcode;
429
430	if (pincpu >= 0) {
431		CPU_ZERO(&mask);
432		CPU_SET(pincpu + vcpu, &mask);
433		error = pthread_setaffinity_np(pthread_self(),
434					       sizeof(mask), &mask);
435		assert(error == 0);
436	}
437
438	while (1) {
439		error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]);
440		if (error != 0) {
441			/*
442			 * It is possible that 'vmmctl' or some other process
443			 * has transitioned the vcpu to CANNOT_RUN state right
444			 * before we tried to transition it to RUNNING.
445			 *
446			 * This is expected to be temporary so just retry.
447			 */
448			if (errno == EBUSY)
449				continue;
450			else
451				break;
452		}
453
454		prevcpu = vcpu;
455
456		exitcode = vmexit[vcpu].exitcode;
457		if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) {
458			fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n",
459			    exitcode);
460			exit(1);
461		}
462
463                rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu);
464
465		switch (rc) {
466		case VMEXIT_CONTINUE:
467                        rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length;
468			break;
469		case VMEXIT_RESTART:
470                        rip = vmexit[vcpu].rip;
471			break;
472		case VMEXIT_RESET:
473			exit(0);
474		default:
475			exit(1);
476		}
477	}
478	fprintf(stderr, "vm_run error %d, errno %d\n", error, errno);
479}
480
481static int
482num_vcpus_allowed(struct vmctx *ctx)
483{
484	int tmp, error;
485
486	error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp);
487
488	/*
489	 * The guest is allowed to spinup more than one processor only if the
490	 * UNRESTRICTED_GUEST capability is available.
491	 */
492	if (error == 0)
493		return (VM_MAXCPU);
494	else
495		return (1);
496}
497
498int
499main(int argc, char *argv[])
500{
501	int c, error, gdb_port, tmp, err, ioapic, bvmcons;
502	int max_vcpus;
503	struct vmctx *ctx;
504	uint64_t rip;
505	size_t memsize;
506
507	bvmcons = 0;
508	progname = basename(argv[0]);
509	gdb_port = DEFAULT_GDB_PORT;
510	guest_ncpus = 1;
511	ioapic = 0;
512	memsize = 256 * MB;
513
514	while ((c = getopt(argc, argv, "abehAHIPp:g:c:s:S:m:")) != -1) {
515		switch (c) {
516		case 'a':
517			disable_x2apic = 1;
518			break;
519		case 'A':
520			acpi = 1;
521			break;
522		case 'b':
523			bvmcons = 1;
524			break;
525		case 'p':
526			pincpu = atoi(optarg);
527			break;
528                case 'c':
529			guest_ncpus = atoi(optarg);
530			break;
531		case 'g':
532			gdb_port = atoi(optarg);
533			break;
534		case 's':
535			if (pci_parse_slot(optarg, 0) != 0)
536				exit(1);
537			else
538				break;
539		case 'S':
540			if (pci_parse_slot(optarg, 1) != 0)
541				exit(1);
542			else
543				break;
544                case 'm':
545			memsize = strtoul(optarg, NULL, 0) * MB;
546			break;
547		case 'H':
548			guest_vmexit_on_hlt = 1;
549			break;
550		case 'I':
551			ioapic = 1;
552			break;
553		case 'P':
554			guest_vmexit_on_pause = 1;
555			break;
556		case 'e':
557			strictio = 1;
558			break;
559		case 'h':
560			usage(0);
561		default:
562			usage(1);
563		}
564	}
565	argc -= optind;
566	argv += optind;
567
568	if (argc != 1)
569		usage(1);
570
571	vmname = argv[0];
572
573	ctx = vm_open(vmname);
574	if (ctx == NULL) {
575		perror("vm_open");
576		exit(1);
577	}
578
579	max_vcpus = num_vcpus_allowed(ctx);
580	if (guest_ncpus > max_vcpus) {
581		fprintf(stderr, "%d vCPUs requested but only %d available\n",
582			guest_ncpus, max_vcpus);
583		exit(1);
584	}
585
586	if (fbsdrun_vmexit_on_hlt()) {
587		err = vm_get_capability(ctx, BSP, VM_CAP_HALT_EXIT, &tmp);
588		if (err < 0) {
589			fprintf(stderr, "VM exit on HLT not supported\n");
590			exit(1);
591		}
592		vm_set_capability(ctx, BSP, VM_CAP_HALT_EXIT, 1);
593		handler[VM_EXITCODE_HLT] = vmexit_hlt;
594	}
595
596        if (fbsdrun_vmexit_on_pause()) {
597		/*
598		 * pause exit support required for this mode
599		 */
600		err = vm_get_capability(ctx, BSP, VM_CAP_PAUSE_EXIT, &tmp);
601		if (err < 0) {
602			fprintf(stderr,
603			    "SMP mux requested, no pause support\n");
604			exit(1);
605		}
606		vm_set_capability(ctx, BSP, VM_CAP_PAUSE_EXIT, 1);
607		handler[VM_EXITCODE_PAUSE] = vmexit_pause;
608        }
609
610	if (fbsdrun_disable_x2apic())
611		err = vm_set_x2apic_state(ctx, BSP, X2APIC_DISABLED);
612	else
613		err = vm_set_x2apic_state(ctx, BSP, X2APIC_ENABLED);
614
615	if (err) {
616		fprintf(stderr, "Unable to set x2apic state (%d)\n", err);
617		exit(1);
618	}
619
620	err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL);
621	if (err) {
622		fprintf(stderr, "Unable to setup memory (%d)\n", err);
623		exit(1);
624	}
625
626	init_mem();
627	init_inout();
628
629	rtc_init(ctx);
630
631	/*
632	 * Exit if a device emulation finds an error in it's initilization
633	 */
634	if (init_pci(ctx) != 0)
635		exit(1);
636
637	if (ioapic)
638		ioapic_init(0);
639
640	if (gdb_port != 0)
641		init_dbgport(gdb_port);
642
643	if (bvmcons)
644		init_bvmcons();
645
646	error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip);
647	assert(error == 0);
648
649	/*
650	 * build the guest tables, MP etc.
651	 */
652	mptable_build(ctx, guest_ncpus, ioapic);
653
654	if (acpi) {
655		error = acpi_build(ctx, guest_ncpus, ioapic);
656		assert(error == 0);
657	}
658
659	/*
660	 * Add CPU 0
661	 */
662	fbsdrun_addcpu(ctx, BSP, rip);
663
664	/*
665	 * Head off to the main event dispatch loop
666	 */
667	mevent_dispatch();
668
669	exit(1);
670}
671