xref: /kvmtool/builtin-run.c (revision ff42603fdb21fd75af54b58d3b7f448e3ea0319f)
1 #include "kvm/builtin-run.h"
2 
3 #include "kvm/virtio-balloon.h"
4 #include "kvm/virtio-console.h"
5 #include "kvm/parse-options.h"
6 #include "kvm/8250-serial.h"
7 #include "kvm/framebuffer.h"
8 #include "kvm/disk-image.h"
9 #include "kvm/threadpool.h"
10 #include "kvm/virtio-blk.h"
11 #include "kvm/virtio-net.h"
12 #include "kvm/virtio-rng.h"
13 #include "kvm/ioeventfd.h"
14 #include "kvm/virtio-9p.h"
15 #include "kvm/barrier.h"
16 #include "kvm/kvm-cpu.h"
17 #include "kvm/ioport.h"
18 #include "kvm/symbol.h"
19 #include "kvm/i8042.h"
20 #include "kvm/mutex.h"
21 #include "kvm/term.h"
22 #include "kvm/util.h"
23 #include "kvm/vesa.h"
24 #include "kvm/irq.h"
25 #include "kvm/kvm.h"
26 #include "kvm/pci.h"
27 #include "kvm/rtc.h"
28 #include "kvm/sdl.h"
29 #include "kvm/vnc.h"
30 #include "kvm/guest_compat.h"
31 #include "kvm/pci-shmem.h"
32 
33 #include <linux/types.h>
34 
35 #include <sys/utsname.h>
36 #include <sys/types.h>
37 #include <sys/stat.h>
38 #include <termios.h>
39 #include <signal.h>
40 #include <stdlib.h>
41 #include <string.h>
42 #include <unistd.h>
43 #include <ctype.h>
44 #include <stdio.h>
45 
46 #define DEFAULT_KVM_DEV		"/dev/kvm"
47 #define DEFAULT_CONSOLE		"serial"
48 #define DEFAULT_NETWORK		"user"
49 #define DEFAULT_HOST_ADDR	"192.168.33.1"
50 #define DEFAULT_GUEST_ADDR	"192.168.33.15"
51 #define DEFAULT_GUEST_MAC	"02:15:15:15:15:15"
52 #define DEFAULT_HOST_MAC	"02:01:01:01:01:01"
53 #define DEFAULT_SCRIPT		"none"
54 
55 #define MB_SHIFT		(20)
56 #define KB_SHIFT		(10)
57 #define GB_SHIFT		(30)
58 #define MIN_RAM_SIZE_MB		(64ULL)
59 #define MIN_RAM_SIZE_BYTE	(MIN_RAM_SIZE_MB << MB_SHIFT)
60 
61 struct kvm *kvm;
62 struct kvm_cpu *kvm_cpus[KVM_NR_CPUS];
63 __thread struct kvm_cpu *current_kvm_cpu;
64 
65 static u64 ram_size;
66 static u8  image_count;
67 static bool virtio_rng;
68 static const char *kernel_cmdline;
69 static const char *kernel_filename;
70 static const char *vmlinux_filename;
71 static const char *initrd_filename;
72 static const char *image_filename[MAX_DISK_IMAGES];
73 static const char *console;
74 static const char *dev;
75 static const char *network;
76 static const char *host_ip;
77 static const char *guest_ip;
78 static const char *guest_mac;
79 static const char *host_mac;
80 static const char *script;
81 static const char *guest_name;
82 static bool single_step;
83 static bool readonly_image[MAX_DISK_IMAGES];
84 static bool vnc;
85 static bool sdl;
86 static bool balloon;
87 static bool using_rootfs;
88 extern bool ioport_debug;
89 extern int  active_console;
90 extern int  debug_iodelay;
91 
92 bool do_debug_print = false;
93 
94 static int nrcpus;
95 static int vidmode = -1;
96 
97 static const char * const run_usage[] = {
98 	"kvm run [<options>] [<kernel image>]",
99 	NULL
100 };
101 
102 static int img_name_parser(const struct option *opt, const char *arg, int unset)
103 {
104 	char *sep;
105 	struct stat st;
106 
107 	if (stat(arg, &st) == 0 &&
108 	    S_ISDIR(st.st_mode)) {
109 		char tmp[PATH_MAX];
110 
111 		if (realpath(arg, tmp) == 0 ||
112 		    virtio_9p__register(kvm, tmp, "/dev/root") < 0)
113 			die("Unable to initialize virtio 9p");
114 		using_rootfs = 1;
115 		return 0;
116 	}
117 
118 	if (image_count >= MAX_DISK_IMAGES)
119 		die("Currently only 4 images are supported");
120 
121 	image_filename[image_count] = arg;
122 	sep = strstr(arg, ",");
123 	if (sep) {
124 		if (strcmp(sep + 1, "ro") == 0)
125 			readonly_image[image_count] = 1;
126 		*sep = 0;
127 	}
128 
129 	image_count++;
130 
131 	return 0;
132 }
133 
134 static int virtio_9p_rootdir_parser(const struct option *opt, const char *arg, int unset)
135 {
136 	char *tag_name;
137 	char tmp[PATH_MAX];
138 
139 	/*
140 	 * 9p dir can be of the form dirname,tag_name or
141 	 * just dirname. In the later case we use the
142 	 * default tag name
143 	 */
144 	tag_name = strstr(arg, ",");
145 	if (tag_name) {
146 		*tag_name = '\0';
147 		tag_name++;
148 	}
149 	if (realpath(arg, tmp)) {
150 		if (virtio_9p__register(kvm, tmp, tag_name) < 0)
151 			die("Unable to initialize virtio 9p");
152 	} else
153 		die("Failed resolving 9p path");
154 	return 0;
155 }
156 
157 static int shmem_parser(const struct option *opt, const char *arg, int unset)
158 {
159 	const uint64_t default_size = SHMEM_DEFAULT_SIZE;
160 	const uint64_t default_phys_addr = SHMEM_DEFAULT_ADDR;
161 	const char *default_handle = SHMEM_DEFAULT_HANDLE;
162 	struct shmem_info *si = malloc(sizeof(struct shmem_info));
163 	enum { PCI, UNK } addr_type = PCI;
164 	uint64_t phys_addr;
165 	uint64_t size;
166 	char *handle = NULL;
167 	int create = 0;
168 	const char *p = arg;
169 	char *next;
170 	int base = 10;
171 	int verbose = 0;
172 
173 	const int skip_pci = strlen("pci:");
174 	if (verbose)
175 		pr_info("shmem_parser(%p,%s,%d)", opt, arg, unset);
176 	/* parse out optional addr family */
177 	if (strcasestr(p, "pci:")) {
178 		p += skip_pci;
179 		addr_type = PCI;
180 	} else if (strcasestr(p, "mem:")) {
181 		die("I can't add to E820 map yet.\n");
182 	}
183 	/* parse out physical addr */
184 	base = 10;
185 	if (strcasestr(p, "0x"))
186 		base = 16;
187 	phys_addr = strtoll(p, &next, base);
188 	if (next == p && phys_addr == 0) {
189 		pr_info("shmem: no physical addr specified, using default.");
190 		phys_addr = default_phys_addr;
191 	}
192 	if (*next != ':' && *next != '\0')
193 		die("shmem: unexpected chars after phys addr.\n");
194 	if (*next == '\0')
195 		p = next;
196 	else
197 		p = next + 1;
198 	/* parse out size */
199 	base = 10;
200 	if (strcasestr(p, "0x"))
201 		base = 16;
202 	size = strtoll(p, &next, base);
203 	if (next == p && size == 0) {
204 		pr_info("shmem: no size specified, using default.");
205 		size = default_size;
206 	}
207 	/* look for [KMGkmg][Bb]*  uses base 2. */
208 	int skip_B = 0;
209 	if (strspn(next, "KMGkmg")) {	/* might have a prefix */
210 		if (*(next + 1) == 'B' || *(next + 1) == 'b')
211 			skip_B = 1;
212 		switch (*next) {
213 		case 'K':
214 		case 'k':
215 			size = size << KB_SHIFT;
216 			break;
217 		case 'M':
218 		case 'm':
219 			size = size << MB_SHIFT;
220 			break;
221 		case 'G':
222 		case 'g':
223 			size = size << GB_SHIFT;
224 			break;
225 		default:
226 			die("shmem: bug in detecting size prefix.");
227 			break;
228 		}
229 		next += 1 + skip_B;
230 	}
231 	if (*next != ':' && *next != '\0') {
232 		die("shmem: unexpected chars after phys size. <%c><%c>\n",
233 		    *next, *p);
234 	}
235 	if (*next == '\0')
236 		p = next;
237 	else
238 		p = next + 1;
239 	/* parse out optional shmem handle */
240 	const int skip_handle = strlen("handle=");
241 	next = strcasestr(p, "handle=");
242 	if (*p && next) {
243 		if (p != next)
244 			die("unexpected chars before handle\n");
245 		p += skip_handle;
246 		next = strchrnul(p, ':');
247 		if (next - p) {
248 			handle = malloc(next - p + 1);
249 			strncpy(handle, p, next - p);
250 			handle[next - p] = '\0';	/* just in case. */
251 		}
252 		if (*next == '\0')
253 			p = next;
254 		else
255 			p = next + 1;
256 	}
257 	/* parse optional create flag to see if we should create shm seg. */
258 	if (*p && strcasestr(p, "create")) {
259 		create = 1;
260 		p += strlen("create");
261 	}
262 	if (*p != '\0')
263 		die("shmem: unexpected trailing chars\n");
264 	if (handle == NULL) {
265 		handle = malloc(strlen(default_handle) + 1);
266 		strcpy(handle, default_handle);
267 	}
268 	if (verbose) {
269 		pr_info("shmem: phys_addr = %lx", phys_addr);
270 		pr_info("shmem: size      = %lx", size);
271 		pr_info("shmem: handle    = %s", handle);
272 		pr_info("shmem: create    = %d", create);
273 	}
274 
275 	si->phys_addr = phys_addr;
276 	si->size = size;
277 	si->handle = handle;
278 	si->create = create;
279 	pci_shmem__register_mem(si);	/* ownership of si, etc. passed on. */
280 	return 0;
281 }
282 
283 static const struct option options[] = {
284 	OPT_GROUP("Basic options:"),
285 	OPT_STRING('\0', "name", &guest_name, "guest name",
286 			"A name for the guest"),
287 	OPT_INTEGER('c', "cpus", &nrcpus, "Number of CPUs"),
288 	OPT_U64('m', "mem", &ram_size, "Virtual machine memory size in MiB."),
289 	OPT_CALLBACK('\0', "shmem", NULL,
290 		     "[pci:]<addr>:<size>[:handle=<handle>][:create]",
291 		     "Share host shmem with guest via pci device",
292 		     shmem_parser),
293 	OPT_CALLBACK('d', "disk", NULL, "image or rootfs_dir", "Disk image or rootfs directory", img_name_parser),
294 	OPT_BOOLEAN('\0', "balloon", &balloon, "Enable virtio balloon"),
295 	OPT_BOOLEAN('\0', "vnc", &vnc, "Enable VNC framebuffer"),
296 	OPT_BOOLEAN('\0', "sdl", &sdl, "Enable SDL framebuffer"),
297 	OPT_BOOLEAN('\0', "rng", &virtio_rng, "Enable virtio Random Number Generator"),
298 	OPT_CALLBACK('\0', "9p", NULL, "dir_to_share,tag_name",
299 		     "Enable virtio 9p to share files between host and guest", virtio_9p_rootdir_parser),
300 	OPT_STRING('\0', "console", &console, "serial or virtio",
301 			"Console to use"),
302 	OPT_STRING('\0', "dev", &dev, "device_file", "KVM device file"),
303 
304 	OPT_GROUP("Kernel options:"),
305 	OPT_STRING('k', "kernel", &kernel_filename, "kernel",
306 			"Kernel to boot in virtual machine"),
307 	OPT_STRING('i', "initrd", &initrd_filename, "initrd",
308 			"Initial RAM disk image"),
309 	OPT_STRING('p', "params", &kernel_cmdline, "params",
310 			"Kernel command line arguments"),
311 
312 	OPT_GROUP("Networking options:"),
313 	OPT_STRING('n', "network", &network, "user, tap, none",
314 			"Network to use"),
315 	OPT_STRING('\0', "host-ip", &host_ip, "a.b.c.d",
316 			"Assign this address to the host side networking"),
317 	OPT_STRING('\0', "guest-ip", &guest_ip, "a.b.c.d",
318 			"Assign this address to the guest side networking"),
319 	OPT_STRING('\0', "host-mac", &host_mac, "aa:bb:cc:dd:ee:ff",
320 			"Assign this address to the host side NIC"),
321 	OPT_STRING('\0', "guest-mac", &guest_mac, "aa:bb:cc:dd:ee:ff",
322 			"Assign this address to the guest side NIC"),
323 	OPT_STRING('\0', "tapscript", &script, "Script path",
324 			 "Assign a script to process created tap device"),
325 
326 	OPT_GROUP("BIOS options:"),
327 	OPT_INTEGER('\0', "vidmode", &vidmode,
328 		    "Video mode"),
329 
330 	OPT_GROUP("Debug options:"),
331 	OPT_BOOLEAN('\0', "debug", &do_debug_print,
332 			"Enable debug messages"),
333 	OPT_BOOLEAN('\0', "debug-single-step", &single_step,
334 			"Enable single stepping"),
335 	OPT_BOOLEAN('\0', "debug-ioport", &ioport_debug,
336 			"Enable ioport debugging"),
337 	OPT_INTEGER('\0', "debug-iodelay", &debug_iodelay,
338 			"Delay IO by millisecond"),
339 	OPT_END()
340 };
341 
342 /*
343  * Serialize debug printout so that the output of multiple vcpus does not
344  * get mixed up:
345  */
346 static int printout_done;
347 
348 static void handle_sigusr1(int sig)
349 {
350 	struct kvm_cpu *cpu = current_kvm_cpu;
351 
352 	if (!cpu)
353 		return;
354 
355 	printf("\n #\n # vCPU #%ld's dump:\n #\n", cpu->cpu_id);
356 	kvm_cpu__show_registers(cpu);
357 	kvm_cpu__show_code(cpu);
358 	kvm_cpu__show_page_tables(cpu);
359 	fflush(stdout);
360 	printout_done = 1;
361 	mb();
362 }
363 
364 /* Pause/resume the guest using SIGUSR2 */
365 static int is_paused;
366 
367 static void handle_sigusr2(int sig)
368 {
369 	if (sig == SIGKVMRESUME && is_paused)
370 		kvm__continue();
371 	else if (sig == SIGUSR2 && !is_paused)
372 		kvm__pause();
373 	else
374 		return;
375 
376 	is_paused = !is_paused;
377 	pr_info("Guest %s\n", is_paused ? "paused" : "resumed");
378 }
379 
380 static void handle_sigquit(int sig)
381 {
382 	int i;
383 
384 	for (i = 0; i < nrcpus; i++) {
385 		struct kvm_cpu *cpu = kvm_cpus[i];
386 
387 		if (!cpu)
388 			continue;
389 
390 		printout_done = 0;
391 		pthread_kill(cpu->thread, SIGUSR1);
392 		/*
393 		 * Wait for the vCPU to dump state before signalling
394 		 * the next thread. Since this is debug code it does
395 		 * not matter that we are burning CPU time a bit:
396 		 */
397 		while (!printout_done)
398 			mb();
399 	}
400 
401 	serial8250__inject_sysrq(kvm);
402 }
403 
404 static void handle_sigalrm(int sig)
405 {
406 	serial8250__inject_interrupt(kvm);
407 	virtio_console__inject_interrupt(kvm);
408 }
409 
410 static void handle_sigstop(int sig)
411 {
412 	kvm_cpu__reboot();
413 }
414 
415 static void *kvm_cpu_thread(void *arg)
416 {
417 	current_kvm_cpu		= arg;
418 
419 	if (kvm_cpu__start(current_kvm_cpu))
420 		goto panic_kvm;
421 
422 	kvm_cpu__delete(current_kvm_cpu);
423 
424 	return (void *) (intptr_t) 0;
425 
426 panic_kvm:
427 	fprintf(stderr, "KVM exit reason: %u (\"%s\")\n",
428 		current_kvm_cpu->kvm_run->exit_reason,
429 		kvm_exit_reasons[current_kvm_cpu->kvm_run->exit_reason]);
430 	if (current_kvm_cpu->kvm_run->exit_reason == KVM_EXIT_UNKNOWN)
431 		fprintf(stderr, "KVM exit code: 0x%Lu\n",
432 			current_kvm_cpu->kvm_run->hw.hardware_exit_reason);
433 
434 	kvm_cpu__show_registers(current_kvm_cpu);
435 	kvm_cpu__show_code(current_kvm_cpu);
436 	kvm_cpu__show_page_tables(current_kvm_cpu);
437 
438 	kvm_cpu__delete(current_kvm_cpu);
439 
440 	return (void *) (intptr_t) 1;
441 }
442 
443 static char kernel[PATH_MAX];
444 
445 static const char *host_kernels[] = {
446 	"/boot/vmlinuz",
447 	"/boot/bzImage",
448 	NULL
449 };
450 
451 static const char *default_kernels[] = {
452 	"./bzImage",
453 	"../../arch/x86/boot/bzImage",
454 	NULL
455 };
456 
457 static const char *default_vmlinux[] = {
458 	"../../../vmlinux",
459 	"../../vmlinux",
460 	NULL
461 };
462 
463 static void kernel_usage_with_options(void)
464 {
465 	const char **k;
466 	struct utsname uts;
467 
468 	fprintf(stderr, "Fatal: could not find default kernel image in:\n");
469 	k = &default_kernels[0];
470 	while (*k) {
471 		fprintf(stderr, "\t%s\n", *k);
472 		k++;
473 	}
474 
475 	if (uname(&uts) < 0)
476 		return;
477 
478 	k = &host_kernels[0];
479 	while (*k) {
480 		if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0)
481 			return;
482 		fprintf(stderr, "\t%s\n", kernel);
483 		k++;
484 	}
485 	fprintf(stderr, "\nPlease see 'kvm run --help' for more options.\n\n");
486 }
487 
488 static u64 host_ram_size(void)
489 {
490 	long page_size;
491 	long nr_pages;
492 
493 	nr_pages	= sysconf(_SC_PHYS_PAGES);
494 	if (nr_pages < 0) {
495 		pr_warning("sysconf(_SC_PHYS_PAGES) failed");
496 		return 0;
497 	}
498 
499 	page_size	= sysconf(_SC_PAGE_SIZE);
500 	if (page_size < 0) {
501 		pr_warning("sysconf(_SC_PAGE_SIZE) failed");
502 		return 0;
503 	}
504 
505 	return (nr_pages * page_size) >> MB_SHIFT;
506 }
507 
508 /*
509  * If user didn't specify how much memory it wants to allocate for the guest,
510  * avoid filling the whole host RAM.
511  */
512 #define RAM_SIZE_RATIO		0.8
513 
514 static u64 get_ram_size(int nr_cpus)
515 {
516 	u64 available;
517 	u64 ram_size;
518 
519 	ram_size	= 64 * (nr_cpus + 3);
520 
521 	available	= host_ram_size() * RAM_SIZE_RATIO;
522 	if (!available)
523 		available = MIN_RAM_SIZE_MB;
524 
525 	if (ram_size > available)
526 		ram_size	= available;
527 
528 	return ram_size;
529 }
530 
531 static const char *find_kernel(void)
532 {
533 	const char **k;
534 	struct stat st;
535 	struct utsname uts;
536 
537 	k = &default_kernels[0];
538 	while (*k) {
539 		if (stat(*k, &st) < 0 || !S_ISREG(st.st_mode)) {
540 			k++;
541 			continue;
542 		}
543 		strncpy(kernel, *k, PATH_MAX);
544 		return kernel;
545 	}
546 
547 	if (uname(&uts) < 0)
548 		return NULL;
549 
550 	k = &host_kernels[0];
551 	while (*k) {
552 		if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0)
553 			return NULL;
554 
555 		if (stat(kernel, &st) < 0 || !S_ISREG(st.st_mode)) {
556 			k++;
557 			continue;
558 		}
559 		return kernel;
560 
561 	}
562 	return NULL;
563 }
564 
565 static const char *find_vmlinux(void)
566 {
567 	const char **vmlinux;
568 
569 	vmlinux = &default_vmlinux[0];
570 	while (*vmlinux) {
571 		struct stat st;
572 
573 		if (stat(*vmlinux, &st) < 0 || !S_ISREG(st.st_mode)) {
574 			vmlinux++;
575 			continue;
576 		}
577 		return *vmlinux;
578 	}
579 	return NULL;
580 }
581 
582 void kvm_run_help(void)
583 {
584 	usage_with_options(run_usage, options);
585 }
586 
587 int kvm_cmd_run(int argc, const char **argv, const char *prefix)
588 {
589 	struct virtio_net_parameters net_params;
590 	static char real_cmdline[2048], default_name[20];
591 	struct framebuffer *fb = NULL;
592 	unsigned int nr_online_cpus;
593 	int exit_code = 0;
594 	int max_cpus, recommended_cpus;
595 	int i;
596 	void *ret;
597 
598 	signal(SIGALRM, handle_sigalrm);
599 	signal(SIGQUIT, handle_sigquit);
600 	signal(SIGUSR1, handle_sigusr1);
601 	signal(SIGUSR2, handle_sigusr2);
602 	signal(SIGKVMSTOP, handle_sigstop);
603 	signal(SIGKVMRESUME, handle_sigusr2);
604 	/* ignore balloon signal by default if not enable balloon optiion */
605 	signal(SIGKVMADDMEM, SIG_IGN);
606 	signal(SIGKVMDELMEM, SIG_IGN);
607 
608 	nr_online_cpus = sysconf(_SC_NPROCESSORS_ONLN);
609 
610 	while (argc != 0) {
611 		argc = parse_options(argc, argv, options, run_usage,
612 				PARSE_OPT_STOP_AT_NON_OPTION);
613 		if (argc != 0) {
614 			if (kernel_filename) {
615 				fprintf(stderr, "Cannot handle parameter: "
616 						"%s\n", argv[0]);
617 				usage_with_options(run_usage, options);
618 				return EINVAL;
619 			}
620 			/* first unhandled parameter is treated as a kernel
621 			   image
622 			 */
623 			kernel_filename = argv[0];
624 			argv++;
625 			argc--;
626 		}
627 
628 	}
629 
630 	if (!kernel_filename)
631 		kernel_filename = find_kernel();
632 
633 	if (!kernel_filename) {
634 		kernel_usage_with_options();
635 		return EINVAL;
636 	}
637 
638 	vmlinux_filename = find_vmlinux();
639 
640 	if (nrcpus == 0)
641 		nrcpus = nr_online_cpus;
642 	else if (nrcpus < 1 || nrcpus > KVM_NR_CPUS)
643 		die("Number of CPUs %d is out of [1;%d] range", nrcpus, KVM_NR_CPUS);
644 
645 	if (!ram_size)
646 		ram_size	= get_ram_size(nrcpus);
647 
648 	if (ram_size < MIN_RAM_SIZE_MB)
649 		die("Not enough memory specified: %lluMB (min %lluMB)", ram_size, MIN_RAM_SIZE_MB);
650 
651 	if (ram_size > host_ram_size())
652 		pr_warning("Guest memory size %lluMB exceeds host physical RAM size %lluMB", ram_size, host_ram_size());
653 
654 	ram_size <<= MB_SHIFT;
655 
656 	if (!dev)
657 		dev = DEFAULT_KVM_DEV;
658 
659 	if (!console)
660 		console = DEFAULT_CONSOLE;
661 
662 	if (!strncmp(console, "virtio", 6))
663 		active_console  = CONSOLE_VIRTIO;
664 	else
665 		active_console  = CONSOLE_8250;
666 
667 	if (!host_ip)
668 		host_ip = DEFAULT_HOST_ADDR;
669 
670 	if (!guest_ip)
671 		guest_ip = DEFAULT_GUEST_ADDR;
672 
673 	if (!guest_mac)
674 		guest_mac = DEFAULT_GUEST_MAC;
675 
676 	if (!host_mac)
677 		host_mac = DEFAULT_HOST_MAC;
678 
679 	if (!script)
680 		script = DEFAULT_SCRIPT;
681 
682 	symbol__init(vmlinux_filename);
683 
684 	term_init();
685 
686 	if (!guest_name) {
687 		sprintf(default_name, "guest-%u", getpid());
688 		guest_name = default_name;
689 	}
690 
691 	kvm = kvm__init(dev, ram_size, guest_name);
692 
693 	irq__init(kvm);
694 
695 	kvm->single_step = single_step;
696 
697 	ioeventfd__init();
698 
699 	max_cpus = kvm__max_cpus(kvm);
700 	recommended_cpus = kvm__recommended_cpus(kvm);
701 
702 	if (nrcpus > max_cpus) {
703 		printf("  # Limit the number of CPUs to %d\n", max_cpus);
704 		kvm->nrcpus	= max_cpus;
705 	} else if (nrcpus > recommended_cpus) {
706 		printf("  # Warning: The maximum recommended amount of VCPUs"
707 			" is %d\n", recommended_cpus);
708 	}
709 
710 	kvm->nrcpus = nrcpus;
711 
712 	/*
713 	 * vidmode should be either specified
714 	 * either set by default
715 	 */
716 	if (vnc || sdl) {
717 		if (vidmode == -1)
718 			vidmode = 0x312;
719 	} else
720 		vidmode = 0;
721 
722 	memset(real_cmdline, 0, sizeof(real_cmdline));
723 	strcpy(real_cmdline, "notsc noapic noacpi pci=conf1 reboot=k panic=1 i8042.direct=1 i8042.dumbkbd=1 i8042.nopnp=1");
724 	if (vnc || sdl) {
725 		strcat(real_cmdline, " video=vesafb console=tty0");
726 	} else
727 		strcat(real_cmdline, " console=ttyS0 earlyprintk=serial");
728 	strcat(real_cmdline, " ");
729 	if (kernel_cmdline)
730 		strlcat(real_cmdline, kernel_cmdline, sizeof(real_cmdline));
731 
732 	if (!using_rootfs && !image_filename[0]) {
733 		if (virtio_9p__register(kvm, "/", "/dev/root") < 0)
734 			die("Unable to initialize virtio 9p");
735 
736 		using_rootfs = 1;
737 
738 		if (!strstr(real_cmdline, "init="))
739 			strlcat(real_cmdline, " init=/bin/sh ", sizeof(real_cmdline));
740 	}
741 
742 	if (using_rootfs)
743 		strcat(real_cmdline, " root=/dev/root rw rootflags=rw,trans=virtio,version=9p2000.L rootfstype=9p");
744 	else if (!strstr(real_cmdline, "root="))
745 		strlcat(real_cmdline, " root=/dev/vda rw ", sizeof(real_cmdline));
746 
747 	if (image_count) {
748 		kvm->nr_disks = image_count;
749 		kvm->disks    = disk_image__open_all(image_filename, readonly_image, image_count);
750 		if (!kvm->disks)
751 			die("Unable to load all disk images.");
752 
753 		virtio_blk__init_all(kvm);
754 	}
755 
756 	printf("  # kvm run -k %s -m %Lu -c %d --name %s\n", kernel_filename, ram_size / 1024 / 1024, nrcpus, guest_name);
757 
758 	if (!kvm__load_kernel(kvm, kernel_filename, initrd_filename,
759 				real_cmdline, vidmode))
760 		die("unable to load kernel %s", kernel_filename);
761 
762 	kvm->vmlinux		= vmlinux_filename;
763 
764 	ioport__setup_legacy();
765 
766 	rtc__init();
767 
768 	serial8250__init(kvm);
769 
770 	pci__init();
771 
772 	if (active_console == CONSOLE_VIRTIO)
773 		virtio_console__init(kvm);
774 
775 	if (virtio_rng)
776 		virtio_rng__init(kvm);
777 
778 	if (balloon)
779 		virtio_bln__init(kvm);
780 
781 	if (!network)
782 		network = DEFAULT_NETWORK;
783 
784 	virtio_9p__init(kvm);
785 
786 	if (strncmp(network, "none", 4)) {
787 		net_params.guest_ip = guest_ip;
788 		net_params.host_ip = host_ip;
789 		net_params.kvm = kvm;
790 		net_params.script = script;
791 		sscanf(guest_mac, "%hhx:%hhx:%hhx:%hhx:%hhx:%hhx",
792 			net_params.guest_mac,
793 			net_params.guest_mac+1,
794 			net_params.guest_mac+2,
795 			net_params.guest_mac+3,
796 			net_params.guest_mac+4,
797 			net_params.guest_mac+5);
798 		sscanf(host_mac, "%hhx:%hhx:%hhx:%hhx:%hhx:%hhx",
799 			net_params.host_mac,
800 			net_params.host_mac+1,
801 			net_params.host_mac+2,
802 			net_params.host_mac+3,
803 			net_params.host_mac+4,
804 			net_params.host_mac+5);
805 
806 		if (!strncmp(network, "user", 4))
807 			net_params.mode = NET_MODE_USER;
808 		else if (!strncmp(network, "tap", 3))
809 			net_params.mode = NET_MODE_TAP;
810 		else
811 			die("Unkown network mode %s, please use -network user, tap, none", network);
812 		virtio_net__init(&net_params);
813 	}
814 
815 	kvm__start_timer(kvm);
816 
817 	kvm__setup_bios(kvm);
818 
819 	for (i = 0; i < nrcpus; i++) {
820 		kvm_cpus[i] = kvm_cpu__init(kvm, i);
821 		if (!kvm_cpus[i])
822 			die("unable to initialize KVM VCPU");
823 	}
824 
825 	kvm__init_ram(kvm);
826 
827 	kbd__init(kvm);
828 
829 	pci_shmem__init(kvm);
830 
831 	if (vnc || sdl)
832 		fb = vesa__init(kvm);
833 
834 	if (vnc) {
835 		if (fb)
836 			vnc__init(fb);
837 	}
838 
839 	if (sdl) {
840 		if (fb)
841 			sdl__init(fb);
842 	}
843 
844 	fb__start();
845 
846 	thread_pool__init(nr_online_cpus);
847 	ioeventfd__start();
848 
849 	for (i = 0; i < nrcpus; i++) {
850 		if (pthread_create(&kvm_cpus[i]->thread, NULL, kvm_cpu_thread, kvm_cpus[i]) != 0)
851 			die("unable to create KVM VCPU thread");
852 	}
853 
854 	/* Only VCPU #0 is going to exit by itself when shutting down */
855 	if (pthread_join(kvm_cpus[0]->thread, &ret) != 0)
856 		exit_code = 1;
857 
858 	for (i = 1; i < nrcpus; i++) {
859 		if (kvm_cpus[i]->is_running) {
860 			pthread_kill(kvm_cpus[i]->thread, SIGKVMEXIT);
861 			if (pthread_join(kvm_cpus[i]->thread, &ret) != 0)
862 				die("pthread_join");
863 		}
864 		if (ret != NULL)
865 			exit_code = 1;
866 	}
867 
868 	compat__print_all_messages();
869 
870 	fb__stop();
871 
872 	virtio_blk__delete_all(kvm);
873 	virtio_rng__delete_all(kvm);
874 
875 	disk_image__close_all(kvm->disks, image_count);
876 	kvm__delete(kvm);
877 
878 	if (!exit_code)
879 		printf("\n  # KVM session ended normally.\n");
880 
881 	return exit_code;
882 }
883