xref: /kvmtool/builtin-run.c (revision af7b08685968ff8f6ffdf5cac1ef067688bce8c6)
1 #include "kvm/builtin-run.h"
2 
3 #include "kvm/builtin-setup.h"
4 #include "kvm/virtio-balloon.h"
5 #include "kvm/virtio-console.h"
6 #include "kvm/parse-options.h"
7 #include "kvm/8250-serial.h"
8 #include "kvm/framebuffer.h"
9 #include "kvm/disk-image.h"
10 #include "kvm/threadpool.h"
11 #include "kvm/virtio-blk.h"
12 #include "kvm/virtio-net.h"
13 #include "kvm/virtio-rng.h"
14 #include "kvm/ioeventfd.h"
15 #include "kvm/virtio-9p.h"
16 #include "kvm/barrier.h"
17 #include "kvm/kvm-cpu.h"
18 #include "kvm/ioport.h"
19 #include "kvm/symbol.h"
20 #include "kvm/i8042.h"
21 #include "kvm/mutex.h"
22 #include "kvm/term.h"
23 #include "kvm/util.h"
24 #include "kvm/vesa.h"
25 #include "kvm/irq.h"
26 #include "kvm/kvm.h"
27 #include "kvm/pci.h"
28 #include "kvm/rtc.h"
29 #include "kvm/sdl.h"
30 #include "kvm/vnc.h"
31 #include "kvm/guest_compat.h"
32 #include "kvm/pci-shmem.h"
33 #include "kvm/kvm-ipc.h"
34 
35 #include <linux/types.h>
36 
37 #include <sys/utsname.h>
38 #include <sys/types.h>
39 #include <sys/stat.h>
40 #include <termios.h>
41 #include <signal.h>
42 #include <stdlib.h>
43 #include <string.h>
44 #include <unistd.h>
45 #include <ctype.h>
46 #include <stdio.h>
47 
48 #define DEFAULT_KVM_DEV		"/dev/kvm"
49 #define DEFAULT_CONSOLE		"serial"
50 #define DEFAULT_NETWORK		"user"
51 #define DEFAULT_HOST_ADDR	"192.168.33.1"
52 #define DEFAULT_GUEST_ADDR	"192.168.33.15"
53 #define DEFAULT_GUEST_MAC	"02:15:15:15:15:15"
54 #define DEFAULT_HOST_MAC	"02:01:01:01:01:01"
55 #define DEFAULT_SCRIPT		"none"
56 
57 #define MB_SHIFT		(20)
58 #define KB_SHIFT		(10)
59 #define GB_SHIFT		(30)
60 #define MIN_RAM_SIZE_MB		(64ULL)
61 #define MIN_RAM_SIZE_BYTE	(MIN_RAM_SIZE_MB << MB_SHIFT)
62 
63 struct kvm *kvm;
64 struct kvm_cpu *kvm_cpus[KVM_NR_CPUS];
65 __thread struct kvm_cpu *current_kvm_cpu;
66 
67 static u64 ram_size;
68 static u8  image_count;
69 static u8 num_net_devices;
70 static bool virtio_rng;
71 static const char *kernel_cmdline;
72 static const char *kernel_filename;
73 static const char *vmlinux_filename;
74 static const char *initrd_filename;
75 static const char *image_filename[MAX_DISK_IMAGES];
76 static const char *console;
77 static const char *dev;
78 static const char *network;
79 static const char *host_ip;
80 static const char *guest_ip;
81 static const char *guest_mac;
82 static const char *host_mac;
83 static const char *script;
84 static const char *guest_name;
85 static struct virtio_net_params *net_params;
86 static bool single_step;
87 static bool readonly_image[MAX_DISK_IMAGES];
88 static bool vnc;
89 static bool sdl;
90 static bool balloon;
91 static bool using_rootfs;
92 static bool custom_rootfs;
93 static bool no_net;
94 static bool no_dhcp;
95 extern bool ioport_debug;
96 extern int  active_console;
97 extern int  debug_iodelay;
98 
99 bool do_debug_print = false;
100 
101 static int nrcpus;
102 static int vidmode = -1;
103 
104 static const char * const run_usage[] = {
105 	"kvm run [<options>] [<kernel image>]",
106 	NULL
107 };
108 
109 static int img_name_parser(const struct option *opt, const char *arg, int unset)
110 {
111 	char *sep;
112 	struct stat st;
113 	char path[PATH_MAX];
114 
115 	if (stat(arg, &st) == 0 &&
116 	    S_ISDIR(st.st_mode)) {
117 		char tmp[PATH_MAX];
118 
119 		if (realpath(arg, tmp) == 0 ||
120 		    virtio_9p__register(kvm, tmp, "/dev/root") < 0)
121 			die("Unable to initialize virtio 9p");
122 		using_rootfs = 1;
123 		return 0;
124 	}
125 
126 	snprintf(path, PATH_MAX, "%s%s", kvm__get_dir(), arg);
127 
128 	if (stat(path, &st) == 0 &&
129 	    S_ISDIR(st.st_mode)) {
130 		char tmp[PATH_MAX];
131 
132 		if (realpath(path, tmp) == 0 ||
133 		    virtio_9p__register(kvm, tmp, "/dev/root") < 0)
134 			die("Unable to initialize virtio 9p");
135 		if (virtio_9p__register(kvm, "/", "hostfs") < 0)
136 			die("Unable to initialize virtio 9p");
137 		kvm_setup_resolv(arg);
138 		using_rootfs = custom_rootfs = 1;
139 		return 0;
140 	}
141 
142 	if (image_count >= MAX_DISK_IMAGES)
143 		die("Currently only 4 images are supported");
144 
145 	image_filename[image_count] = arg;
146 	sep = strstr(arg, ",");
147 	if (sep) {
148 		if (strcmp(sep + 1, "ro") == 0)
149 			readonly_image[image_count] = 1;
150 		*sep = 0;
151 	}
152 
153 	image_count++;
154 
155 	return 0;
156 }
157 
158 static int virtio_9p_rootdir_parser(const struct option *opt, const char *arg, int unset)
159 {
160 	char *tag_name;
161 	char tmp[PATH_MAX];
162 
163 	/*
164 	 * 9p dir can be of the form dirname,tag_name or
165 	 * just dirname. In the later case we use the
166 	 * default tag name
167 	 */
168 	tag_name = strstr(arg, ",");
169 	if (tag_name) {
170 		*tag_name = '\0';
171 		tag_name++;
172 	}
173 	if (realpath(arg, tmp)) {
174 		if (virtio_9p__register(kvm, tmp, tag_name) < 0)
175 			die("Unable to initialize virtio 9p");
176 	} else
177 		die("Failed resolving 9p path");
178 	return 0;
179 }
180 
181 static int tty_parser(const struct option *opt, const char *arg, int unset)
182 {
183 	int tty = atoi(arg);
184 
185 	term_set_tty(tty);
186 
187 	return 0;
188 }
189 
190 static inline void str_to_mac(const char *str, char *mac)
191 {
192 	sscanf(str, "%hhx:%hhx:%hhx:%hhx:%hhx:%hhx",
193 		mac, mac+1, mac+2, mac+3, mac+4, mac+5);
194 }
195 static int set_net_param(struct virtio_net_params *p, const char *param,
196 				const char *val)
197 {
198 	if (strcmp(param, "guest_mac") == 0) {
199 		str_to_mac(val, p->guest_mac);
200 	} else if (strcmp(param, "mode") == 0) {
201 		if (!strncmp(val, "user", 4)) {
202 			int i;
203 
204 			for (i = 0; i < num_net_devices; i++)
205 				if (net_params[i].mode == NET_MODE_USER)
206 					die("Only one usermode network device allowed at a time");
207 			p->mode = NET_MODE_USER;
208 		} else if (!strncmp(val, "tap", 3)) {
209 			p->mode = NET_MODE_TAP;
210 		} else if (!strncmp(val, "none", 4)) {
211 			no_net = 1;
212 			return -1;
213 		} else
214 			die("Unkown network mode %s, please use user, tap or none", network);
215 	} else if (strcmp(param, "script") == 0) {
216 		p->script = strdup(val);
217 	} else if (strcmp(param, "guest_ip") == 0) {
218 		p->guest_ip = strdup(val);
219 	} else if (strcmp(param, "host_ip") == 0) {
220 		p->host_ip = strdup(val);
221 	} else if (strcmp(param, "vhost") == 0) {
222 		p->vhost = atoi(val);
223 	}
224 
225 	return 0;
226 }
227 
228 static int netdev_parser(const struct option *opt, const char *arg, int unset)
229 {
230 	struct virtio_net_params p;
231 	char *buf = NULL, *cmd = NULL, *cur = NULL;
232 	bool on_cmd = true;
233 
234 	if (arg) {
235 		buf = strdup(arg);
236 		if (buf == NULL)
237 			die("Failed allocating new net buffer");
238 		cur = strtok(buf, ",=");
239 	}
240 
241 	p = (struct virtio_net_params) {
242 		.guest_ip	= DEFAULT_GUEST_ADDR,
243 		.host_ip	= DEFAULT_HOST_ADDR,
244 		.script		= DEFAULT_SCRIPT,
245 		.mode		= NET_MODE_TAP,
246 	};
247 
248 	str_to_mac(DEFAULT_GUEST_MAC, p.guest_mac);
249 	p.guest_mac[5] += num_net_devices;
250 
251 	while (cur) {
252 		if (on_cmd) {
253 			cmd = cur;
254 		} else {
255 			if (set_net_param(&p, cmd, cur) < 0)
256 				goto done;
257 		}
258 		on_cmd = !on_cmd;
259 
260 		cur = strtok(NULL, ",=");
261 	};
262 
263 	num_net_devices++;
264 
265 	net_params = realloc(net_params, num_net_devices * sizeof(*net_params));
266 	if (net_params == NULL)
267 		die("Failed adding new network device");
268 
269 	net_params[num_net_devices - 1] = p;
270 
271 done:
272 	free(buf);
273 	return 0;
274 }
275 
276 static int shmem_parser(const struct option *opt, const char *arg, int unset)
277 {
278 	const u64 default_size = SHMEM_DEFAULT_SIZE;
279 	const u64 default_phys_addr = SHMEM_DEFAULT_ADDR;
280 	const char *default_handle = SHMEM_DEFAULT_HANDLE;
281 	struct shmem_info *si = malloc(sizeof(struct shmem_info));
282 	u64 phys_addr;
283 	u64 size;
284 	char *handle = NULL;
285 	int create = 0;
286 	const char *p = arg;
287 	char *next;
288 	int base = 10;
289 	int verbose = 0;
290 
291 	const int skip_pci = strlen("pci:");
292 	if (verbose)
293 		pr_info("shmem_parser(%p,%s,%d)", opt, arg, unset);
294 	/* parse out optional addr family */
295 	if (strcasestr(p, "pci:")) {
296 		p += skip_pci;
297 	} else if (strcasestr(p, "mem:")) {
298 		die("I can't add to E820 map yet.\n");
299 	}
300 	/* parse out physical addr */
301 	base = 10;
302 	if (strcasestr(p, "0x"))
303 		base = 16;
304 	phys_addr = strtoll(p, &next, base);
305 	if (next == p && phys_addr == 0) {
306 		pr_info("shmem: no physical addr specified, using default.");
307 		phys_addr = default_phys_addr;
308 	}
309 	if (*next != ':' && *next != '\0')
310 		die("shmem: unexpected chars after phys addr.\n");
311 	if (*next == '\0')
312 		p = next;
313 	else
314 		p = next + 1;
315 	/* parse out size */
316 	base = 10;
317 	if (strcasestr(p, "0x"))
318 		base = 16;
319 	size = strtoll(p, &next, base);
320 	if (next == p && size == 0) {
321 		pr_info("shmem: no size specified, using default.");
322 		size = default_size;
323 	}
324 	/* look for [KMGkmg][Bb]*  uses base 2. */
325 	int skip_B = 0;
326 	if (strspn(next, "KMGkmg")) {	/* might have a prefix */
327 		if (*(next + 1) == 'B' || *(next + 1) == 'b')
328 			skip_B = 1;
329 		switch (*next) {
330 		case 'K':
331 		case 'k':
332 			size = size << KB_SHIFT;
333 			break;
334 		case 'M':
335 		case 'm':
336 			size = size << MB_SHIFT;
337 			break;
338 		case 'G':
339 		case 'g':
340 			size = size << GB_SHIFT;
341 			break;
342 		default:
343 			die("shmem: bug in detecting size prefix.");
344 			break;
345 		}
346 		next += 1 + skip_B;
347 	}
348 	if (*next != ':' && *next != '\0') {
349 		die("shmem: unexpected chars after phys size. <%c><%c>\n",
350 		    *next, *p);
351 	}
352 	if (*next == '\0')
353 		p = next;
354 	else
355 		p = next + 1;
356 	/* parse out optional shmem handle */
357 	const int skip_handle = strlen("handle=");
358 	next = strcasestr(p, "handle=");
359 	if (*p && next) {
360 		if (p != next)
361 			die("unexpected chars before handle\n");
362 		p += skip_handle;
363 		next = strchrnul(p, ':');
364 		if (next - p) {
365 			handle = malloc(next - p + 1);
366 			strncpy(handle, p, next - p);
367 			handle[next - p] = '\0';	/* just in case. */
368 		}
369 		if (*next == '\0')
370 			p = next;
371 		else
372 			p = next + 1;
373 	}
374 	/* parse optional create flag to see if we should create shm seg. */
375 	if (*p && strcasestr(p, "create")) {
376 		create = 1;
377 		p += strlen("create");
378 	}
379 	if (*p != '\0')
380 		die("shmem: unexpected trailing chars\n");
381 	if (handle == NULL) {
382 		handle = malloc(strlen(default_handle) + 1);
383 		strcpy(handle, default_handle);
384 	}
385 	if (verbose) {
386 		pr_info("shmem: phys_addr = %llx", phys_addr);
387 		pr_info("shmem: size      = %llx", size);
388 		pr_info("shmem: handle    = %s", handle);
389 		pr_info("shmem: create    = %d", create);
390 	}
391 
392 	si->phys_addr = phys_addr;
393 	si->size = size;
394 	si->handle = handle;
395 	si->create = create;
396 	pci_shmem__register_mem(si);	/* ownership of si, etc. passed on. */
397 	return 0;
398 }
399 
400 static const struct option options[] = {
401 	OPT_GROUP("Basic options:"),
402 	OPT_STRING('\0', "name", &guest_name, "guest name",
403 			"A name for the guest"),
404 	OPT_INTEGER('c', "cpus", &nrcpus, "Number of CPUs"),
405 	OPT_U64('m', "mem", &ram_size, "Virtual machine memory size in MiB."),
406 	OPT_CALLBACK('\0', "shmem", NULL,
407 		     "[pci:]<addr>:<size>[:handle=<handle>][:create]",
408 		     "Share host shmem with guest via pci device",
409 		     shmem_parser),
410 	OPT_CALLBACK('d', "disk", NULL, "image or rootfs_dir", "Disk image or rootfs directory", img_name_parser),
411 	OPT_BOOLEAN('\0', "balloon", &balloon, "Enable virtio balloon"),
412 	OPT_BOOLEAN('\0', "vnc", &vnc, "Enable VNC framebuffer"),
413 	OPT_BOOLEAN('\0', "sdl", &sdl, "Enable SDL framebuffer"),
414 	OPT_BOOLEAN('\0', "rng", &virtio_rng, "Enable virtio Random Number Generator"),
415 	OPT_CALLBACK('\0', "9p", NULL, "dir_to_share,tag_name",
416 		     "Enable virtio 9p to share files between host and guest", virtio_9p_rootdir_parser),
417 	OPT_STRING('\0', "console", &console, "serial or virtio",
418 			"Console to use"),
419 	OPT_STRING('\0', "dev", &dev, "device_file", "KVM device file"),
420 	OPT_CALLBACK('\0', "tty", NULL, "tty id",
421 		     "Remap guest TTY into a pty on the host",
422 		     tty_parser),
423 
424 	OPT_GROUP("Kernel options:"),
425 	OPT_STRING('k', "kernel", &kernel_filename, "kernel",
426 			"Kernel to boot in virtual machine"),
427 	OPT_STRING('i', "initrd", &initrd_filename, "initrd",
428 			"Initial RAM disk image"),
429 	OPT_STRING('p', "params", &kernel_cmdline, "params",
430 			"Kernel command line arguments"),
431 
432 	OPT_GROUP("Networking options:"),
433 	OPT_CALLBACK_DEFAULT('n', "network", NULL, "network params",
434 		     "Create a new guest NIC",
435 		     netdev_parser, NULL),
436 	OPT_BOOLEAN('\0', "no-dhcp", &no_dhcp, "Disable kernel DHCP in rootfs mode"),
437 
438 	OPT_GROUP("BIOS options:"),
439 	OPT_INTEGER('\0', "vidmode", &vidmode,
440 		    "Video mode"),
441 
442 	OPT_GROUP("Debug options:"),
443 	OPT_BOOLEAN('\0', "debug", &do_debug_print,
444 			"Enable debug messages"),
445 	OPT_BOOLEAN('\0', "debug-single-step", &single_step,
446 			"Enable single stepping"),
447 	OPT_BOOLEAN('\0', "debug-ioport", &ioport_debug,
448 			"Enable ioport debugging"),
449 	OPT_INTEGER('\0', "debug-iodelay", &debug_iodelay,
450 			"Delay IO by millisecond"),
451 	OPT_END()
452 };
453 
454 /*
455  * Serialize debug printout so that the output of multiple vcpus does not
456  * get mixed up:
457  */
458 static int printout_done;
459 
460 static void handle_sigusr1(int sig)
461 {
462 	struct kvm_cpu *cpu = current_kvm_cpu;
463 	int fd = kvm_cpu__get_debug_fd();
464 
465 	if (!cpu)
466 		return;
467 
468 	dprintf(fd, "\n #\n # vCPU #%ld's dump:\n #\n", cpu->cpu_id);
469 	kvm_cpu__show_registers(cpu);
470 	kvm_cpu__show_code(cpu);
471 	kvm_cpu__show_page_tables(cpu);
472 	fflush(stdout);
473 	printout_done = 1;
474 	mb();
475 }
476 
477 /* Pause/resume the guest using SIGUSR2 */
478 static int is_paused;
479 
480 static void handle_pause(int fd, u32 type, u32 len, u8 *msg)
481 {
482 	if (type == KVM_IPC_RESUME && is_paused)
483 		kvm__continue();
484 	else if (type == KVM_IPC_PAUSE && !is_paused)
485 		kvm__pause();
486 	else
487 		return;
488 
489 	is_paused = !is_paused;
490 	pr_info("Guest %s\n", is_paused ? "paused" : "resumed");
491 }
492 
493 static void handle_debug(int fd, u32 type, u32 len, u8 *msg)
494 {
495 	int i;
496 
497 	for (i = 0; i < nrcpus; i++) {
498 		struct kvm_cpu *cpu = kvm_cpus[i];
499 
500 		if (!cpu)
501 			continue;
502 
503 		printout_done = 0;
504 
505 		kvm_cpu__set_debug_fd(fd);
506 		pthread_kill(cpu->thread, SIGUSR1);
507 		/*
508 		 * Wait for the vCPU to dump state before signalling
509 		 * the next thread. Since this is debug code it does
510 		 * not matter that we are burning CPU time a bit:
511 		 */
512 		while (!printout_done)
513 			mb();
514 	}
515 
516 	close(fd);
517 
518 	serial8250__inject_sysrq(kvm);
519 }
520 
521 static void handle_sigalrm(int sig)
522 {
523 	serial8250__inject_interrupt(kvm);
524 	virtio_console__inject_interrupt(kvm);
525 }
526 
527 static void handle_stop(int fd, u32 type, u32 len, u8 *msg)
528 {
529 	kvm_cpu__reboot();
530 }
531 
532 static void *kvm_cpu_thread(void *arg)
533 {
534 	current_kvm_cpu		= arg;
535 
536 	if (kvm_cpu__start(current_kvm_cpu))
537 		goto panic_kvm;
538 
539 	kvm_cpu__delete(current_kvm_cpu);
540 
541 	return (void *) (intptr_t) 0;
542 
543 panic_kvm:
544 	fprintf(stderr, "KVM exit reason: %u (\"%s\")\n",
545 		current_kvm_cpu->kvm_run->exit_reason,
546 		kvm_exit_reasons[current_kvm_cpu->kvm_run->exit_reason]);
547 	if (current_kvm_cpu->kvm_run->exit_reason == KVM_EXIT_UNKNOWN)
548 		fprintf(stderr, "KVM exit code: 0x%Lu\n",
549 			current_kvm_cpu->kvm_run->hw.hardware_exit_reason);
550 
551 	kvm_cpu__set_debug_fd(STDOUT_FILENO);
552 	kvm_cpu__show_registers(current_kvm_cpu);
553 	kvm_cpu__show_code(current_kvm_cpu);
554 	kvm_cpu__show_page_tables(current_kvm_cpu);
555 
556 	kvm_cpu__delete(current_kvm_cpu);
557 
558 	return (void *) (intptr_t) 1;
559 }
560 
561 static char kernel[PATH_MAX];
562 
563 static const char *host_kernels[] = {
564 	"/boot/vmlinuz",
565 	"/boot/bzImage",
566 	NULL
567 };
568 
569 static const char *default_kernels[] = {
570 	"./bzImage",
571 	"../../arch/" BUILD_ARCH "/boot/bzImage",
572 	NULL
573 };
574 
575 static const char *default_vmlinux[] = {
576 	"../../../vmlinux",
577 	"../../vmlinux",
578 	NULL
579 };
580 
581 static void kernel_usage_with_options(void)
582 {
583 	const char **k;
584 	struct utsname uts;
585 
586 	fprintf(stderr, "Fatal: could not find default kernel image in:\n");
587 	k = &default_kernels[0];
588 	while (*k) {
589 		fprintf(stderr, "\t%s\n", *k);
590 		k++;
591 	}
592 
593 	if (uname(&uts) < 0)
594 		return;
595 
596 	k = &host_kernels[0];
597 	while (*k) {
598 		if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0)
599 			return;
600 		fprintf(stderr, "\t%s\n", kernel);
601 		k++;
602 	}
603 	fprintf(stderr, "\nPlease see 'kvm run --help' for more options.\n\n");
604 }
605 
606 static u64 host_ram_size(void)
607 {
608 	long page_size;
609 	long nr_pages;
610 
611 	nr_pages	= sysconf(_SC_PHYS_PAGES);
612 	if (nr_pages < 0) {
613 		pr_warning("sysconf(_SC_PHYS_PAGES) failed");
614 		return 0;
615 	}
616 
617 	page_size	= sysconf(_SC_PAGE_SIZE);
618 	if (page_size < 0) {
619 		pr_warning("sysconf(_SC_PAGE_SIZE) failed");
620 		return 0;
621 	}
622 
623 	return (nr_pages * page_size) >> MB_SHIFT;
624 }
625 
626 /*
627  * If user didn't specify how much memory it wants to allocate for the guest,
628  * avoid filling the whole host RAM.
629  */
630 #define RAM_SIZE_RATIO		0.8
631 
632 static u64 get_ram_size(int nr_cpus)
633 {
634 	u64 available;
635 	u64 ram_size;
636 
637 	ram_size	= 64 * (nr_cpus + 3);
638 
639 	available	= host_ram_size() * RAM_SIZE_RATIO;
640 	if (!available)
641 		available = MIN_RAM_SIZE_MB;
642 
643 	if (ram_size > available)
644 		ram_size	= available;
645 
646 	return ram_size;
647 }
648 
649 static const char *find_kernel(void)
650 {
651 	const char **k;
652 	struct stat st;
653 	struct utsname uts;
654 
655 	k = &default_kernels[0];
656 	while (*k) {
657 		if (stat(*k, &st) < 0 || !S_ISREG(st.st_mode)) {
658 			k++;
659 			continue;
660 		}
661 		strncpy(kernel, *k, PATH_MAX);
662 		return kernel;
663 	}
664 
665 	if (uname(&uts) < 0)
666 		return NULL;
667 
668 	k = &host_kernels[0];
669 	while (*k) {
670 		if (snprintf(kernel, PATH_MAX, "%s-%s", *k, uts.release) < 0)
671 			return NULL;
672 
673 		if (stat(kernel, &st) < 0 || !S_ISREG(st.st_mode)) {
674 			k++;
675 			continue;
676 		}
677 		return kernel;
678 
679 	}
680 	return NULL;
681 }
682 
683 static const char *find_vmlinux(void)
684 {
685 	const char **vmlinux;
686 
687 	vmlinux = &default_vmlinux[0];
688 	while (*vmlinux) {
689 		struct stat st;
690 
691 		if (stat(*vmlinux, &st) < 0 || !S_ISREG(st.st_mode)) {
692 			vmlinux++;
693 			continue;
694 		}
695 		return *vmlinux;
696 	}
697 	return NULL;
698 }
699 
700 void kvm_run_help(void)
701 {
702 	usage_with_options(run_usage, options);
703 }
704 
705 int kvm_cmd_run(int argc, const char **argv, const char *prefix)
706 {
707 	static char real_cmdline[2048], default_name[20];
708 	struct framebuffer *fb = NULL;
709 	unsigned int nr_online_cpus;
710 	int exit_code = 0;
711 	int max_cpus, recommended_cpus;
712 	int i;
713 	void *ret;
714 
715 	signal(SIGALRM, handle_sigalrm);
716 	kvm_ipc__register_handler(KVM_IPC_DEBUG, handle_debug);
717 	signal(SIGUSR1, handle_sigusr1);
718 	kvm_ipc__register_handler(KVM_IPC_PAUSE, handle_pause);
719 	kvm_ipc__register_handler(KVM_IPC_RESUME, handle_pause);
720 	kvm_ipc__register_handler(KVM_IPC_STOP, handle_stop);
721 
722 	nr_online_cpus = sysconf(_SC_NPROCESSORS_ONLN);
723 
724 	while (argc != 0) {
725 		argc = parse_options(argc, argv, options, run_usage,
726 				PARSE_OPT_STOP_AT_NON_OPTION);
727 		if (argc != 0) {
728 			if (kernel_filename) {
729 				fprintf(stderr, "Cannot handle parameter: "
730 						"%s\n", argv[0]);
731 				usage_with_options(run_usage, options);
732 				return EINVAL;
733 			}
734 			/* first unhandled parameter is treated as a kernel
735 			   image
736 			 */
737 			kernel_filename = argv[0];
738 			argv++;
739 			argc--;
740 		}
741 
742 	}
743 
744 	if (!kernel_filename)
745 		kernel_filename = find_kernel();
746 
747 	if (!kernel_filename) {
748 		kernel_usage_with_options();
749 		return EINVAL;
750 	}
751 
752 	vmlinux_filename = find_vmlinux();
753 
754 	if (nrcpus == 0)
755 		nrcpus = nr_online_cpus;
756 	else if (nrcpus < 1 || nrcpus > KVM_NR_CPUS)
757 		die("Number of CPUs %d is out of [1;%d] range", nrcpus, KVM_NR_CPUS);
758 
759 	if (!ram_size)
760 		ram_size	= get_ram_size(nrcpus);
761 
762 	if (ram_size < MIN_RAM_SIZE_MB)
763 		die("Not enough memory specified: %lluMB (min %lluMB)", ram_size, MIN_RAM_SIZE_MB);
764 
765 	if (ram_size > host_ram_size())
766 		pr_warning("Guest memory size %lluMB exceeds host physical RAM size %lluMB", ram_size, host_ram_size());
767 
768 	ram_size <<= MB_SHIFT;
769 
770 	if (!dev)
771 		dev = DEFAULT_KVM_DEV;
772 
773 	if (!console)
774 		console = DEFAULT_CONSOLE;
775 
776 	if (!strncmp(console, "virtio", 6))
777 		active_console  = CONSOLE_VIRTIO;
778 	else
779 		active_console  = CONSOLE_8250;
780 
781 	if (!host_ip)
782 		host_ip = DEFAULT_HOST_ADDR;
783 
784 	if (!guest_ip)
785 		guest_ip = DEFAULT_GUEST_ADDR;
786 
787 	if (!guest_mac)
788 		guest_mac = DEFAULT_GUEST_MAC;
789 
790 	if (!host_mac)
791 		host_mac = DEFAULT_HOST_MAC;
792 
793 	if (!script)
794 		script = DEFAULT_SCRIPT;
795 
796 	symbol__init(vmlinux_filename);
797 
798 	term_init();
799 
800 	if (!guest_name) {
801 		sprintf(default_name, "guest-%u", getpid());
802 		guest_name = default_name;
803 	}
804 
805 	kvm = kvm__init(dev, ram_size, guest_name);
806 
807 	irq__init(kvm);
808 
809 	kvm->single_step = single_step;
810 
811 	ioeventfd__init();
812 
813 	max_cpus = kvm__max_cpus(kvm);
814 	recommended_cpus = kvm__recommended_cpus(kvm);
815 
816 	if (nrcpus > max_cpus) {
817 		printf("  # Limit the number of CPUs to %d\n", max_cpus);
818 		kvm->nrcpus	= max_cpus;
819 	} else if (nrcpus > recommended_cpus) {
820 		printf("  # Warning: The maximum recommended amount of VCPUs"
821 			" is %d\n", recommended_cpus);
822 	}
823 
824 	kvm->nrcpus = nrcpus;
825 
826 	/*
827 	 * vidmode should be either specified
828 	 * either set by default
829 	 */
830 	if (vnc || sdl) {
831 		if (vidmode == -1)
832 			vidmode = 0x312;
833 	} else
834 		vidmode = 0;
835 
836 	memset(real_cmdline, 0, sizeof(real_cmdline));
837 	strcpy(real_cmdline, "noapic noacpi pci=conf1 reboot=k panic=1 i8042.direct=1 "
838 				"i8042.dumbkbd=1 i8042.nopnp=1");
839 	if (vnc || sdl) {
840 		strcat(real_cmdline, " video=vesafb console=tty0");
841 	} else
842 		strcat(real_cmdline, " console=ttyS0 earlyprintk=serial i8042.noaux=1");
843 	strcat(real_cmdline, " ");
844 	if (kernel_cmdline)
845 		strlcat(real_cmdline, kernel_cmdline, sizeof(real_cmdline));
846 
847 	if (!using_rootfs && !image_filename[0]) {
848 		char tmp[PATH_MAX];
849 
850 		kvm_setup_create_new("default");
851 		kvm_setup_resolv("default");
852 
853 		snprintf(tmp, PATH_MAX, "%s%s", kvm__get_dir(), "default");
854 		if (virtio_9p__register(kvm, tmp, "/dev/root") < 0)
855 			die("Unable to initialize virtio 9p");
856 		if (virtio_9p__register(kvm, "/", "hostfs") < 0)
857 			die("Unable to initialize virtio 9p");
858 		using_rootfs = custom_rootfs = 1;
859 	}
860 
861 	if (using_rootfs) {
862 		strcat(real_cmdline, " root=/dev/root rw rootflags=rw,trans=virtio,version=9p2000.L rootfstype=9p");
863 		if (custom_rootfs) {
864 			strcat(real_cmdline, " init=/virt/init");
865 			if (!no_dhcp)
866 				strcat(real_cmdline, "  ip=dhcp");
867 		}
868 	} else if (!strstr(real_cmdline, "root=")) {
869 		strlcat(real_cmdline, " root=/dev/vda rw ", sizeof(real_cmdline));
870 	}
871 
872 	if (image_count) {
873 		kvm->nr_disks = image_count;
874 		kvm->disks    = disk_image__open_all(image_filename, readonly_image, image_count);
875 		if (!kvm->disks)
876 			die("Unable to load all disk images.");
877 
878 		virtio_blk__init_all(kvm);
879 	}
880 
881 	printf("  # kvm run -k %s -m %Lu -c %d --name %s\n", kernel_filename, ram_size / 1024 / 1024, nrcpus, guest_name);
882 
883 	if (!kvm__load_kernel(kvm, kernel_filename, initrd_filename,
884 				real_cmdline, vidmode))
885 		die("unable to load kernel %s", kernel_filename);
886 
887 	kvm->vmlinux		= vmlinux_filename;
888 
889 	ioport__setup_arch();
890 
891 	rtc__init();
892 
893 	serial8250__init(kvm);
894 
895 	pci__init();
896 
897 	if (active_console == CONSOLE_VIRTIO)
898 		virtio_console__init(kvm);
899 
900 	if (virtio_rng)
901 		virtio_rng__init(kvm);
902 
903 	if (balloon)
904 		virtio_bln__init(kvm);
905 
906 	if (!network)
907 		network = DEFAULT_NETWORK;
908 
909 	virtio_9p__init(kvm);
910 
911 	for (i = 0; i < num_net_devices; i++) {
912 		net_params[i].kvm = kvm;
913 		virtio_net__init(&net_params[i]);
914 	}
915 
916 	if (num_net_devices == 0 && no_net == 0) {
917 		struct virtio_net_params net_params;
918 
919 		net_params = (struct virtio_net_params) {
920 			.guest_ip	= guest_ip,
921 			.host_ip	= host_ip,
922 			.kvm		= kvm,
923 			.script		= script,
924 			.mode		= NET_MODE_USER,
925 		};
926 		str_to_mac(guest_mac, net_params.guest_mac);
927 		str_to_mac(host_mac, net_params.host_mac);
928 
929 		virtio_net__init(&net_params);
930 	}
931 
932 	kvm__start_timer(kvm);
933 
934 	kvm__arch_setup_firmware(kvm);
935 
936 	for (i = 0; i < nrcpus; i++) {
937 		kvm_cpus[i] = kvm_cpu__init(kvm, i);
938 		if (!kvm_cpus[i])
939 			die("unable to initialize KVM VCPU");
940 	}
941 
942 	kvm__init_ram(kvm);
943 
944 	kbd__init(kvm);
945 
946 	pci_shmem__init(kvm);
947 
948 	if (vnc || sdl)
949 		fb = vesa__init(kvm);
950 
951 	if (vnc) {
952 		if (fb)
953 			vnc__init(fb);
954 	}
955 
956 	if (sdl) {
957 		if (fb)
958 			sdl__init(fb);
959 	}
960 
961 	fb__start();
962 
963 	thread_pool__init(nr_online_cpus);
964 	ioeventfd__start();
965 
966 	for (i = 0; i < nrcpus; i++) {
967 		if (pthread_create(&kvm_cpus[i]->thread, NULL, kvm_cpu_thread, kvm_cpus[i]) != 0)
968 			die("unable to create KVM VCPU thread");
969 	}
970 
971 	/* Only VCPU #0 is going to exit by itself when shutting down */
972 	if (pthread_join(kvm_cpus[0]->thread, &ret) != 0)
973 		exit_code = 1;
974 
975 	for (i = 1; i < nrcpus; i++) {
976 		if (kvm_cpus[i]->is_running) {
977 			pthread_kill(kvm_cpus[i]->thread, SIGKVMEXIT);
978 			if (pthread_join(kvm_cpus[i]->thread, &ret) != 0)
979 				die("pthread_join");
980 		}
981 		if (ret != NULL)
982 			exit_code = 1;
983 	}
984 
985 	compat__print_all_messages();
986 
987 	fb__stop();
988 
989 	virtio_blk__delete_all(kvm);
990 	virtio_rng__delete_all(kvm);
991 
992 	disk_image__close_all(kvm->disks, image_count);
993 	kvm__delete(kvm);
994 
995 	if (!exit_code)
996 		printf("\n  # KVM session ended normally.\n");
997 
998 	return exit_code;
999 }
1000