?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
introduce and use qemu_clock_enable
[qemu-kvm/amd-iommu.git] / vl.c
bloba9edff6ff3244b4599e90d2e0d2dc7b30a188d9c
1 /*
2 * QEMU System Emulator
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24 #include <unistd.h>
25 #include <fcntl.h>
26 #include <signal.h>
27 #include <time.h>
28 #include <errno.h>
29 #include <sys/time.h>
30 #include <zlib.h>
31
32 /* Needed early for CONFIG_BSD etc. */
33 #include "config-host.h"
34
35 #ifndef _WIN32
36 #include <libgen.h>
37 #include <pwd.h>
38 #include <sys/times.h>
39 #include <sys/wait.h>
40 #include <termios.h>
41 #include <sys/mman.h>
42 #include <sys/ioctl.h>
43 #include <sys/resource.h>
44 #include <sys/socket.h>
45 #include <netinet/in.h>
46 #include <net/if.h>
47 #include <arpa/inet.h>
48 #include <dirent.h>
49 #include <netdb.h>
50 #include <sys/select.h>
51 #ifdef CONFIG_BSD
52 #include <sys/stat.h>
53 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) || defined(__DragonFly__)
54 #include <libutil.h>
55 #else
56 #include <util.h>
57 #endif
58 #else
59 #ifdef __linux__
60 #include <pty.h>
61 #include <malloc.h>
62 #include <linux/rtc.h>
63 #include <sys/prctl.h>
64
65 /* For the benefit of older linux systems which don't supply it,
66 we use a local copy of hpet.h. */
67 /* #include <linux/hpet.h> */
68 #include "hpet.h"
69
70 #include <linux/ppdev.h>
71 #include <linux/parport.h>
72 #endif
73 #ifdef __sun__
74 #include <sys/stat.h>
75 #include <sys/ethernet.h>
76 #include <sys/sockio.h>
77 #include <netinet/arp.h>
78 #include <netinet/in.h>
79 #include <netinet/in_systm.h>
80 #include <netinet/ip.h>
81 #include <netinet/ip_icmp.h> // must come after ip.h
82 #include <netinet/udp.h>
83 #include <netinet/tcp.h>
84 #include <net/if.h>
85 #include <syslog.h>
86 #include <stropts.h>
87 /* See MySQL bug #7156 (http://bugs.mysql.com/bug.php?id=7156) for
88 discussion about Solaris header problems */
89 extern int madvise(caddr_t, size_t, int);
90 #endif
91 #endif
92 #endif
93
94 #if defined(__OpenBSD__)
95 #include <util.h>
96 #endif
97
98 #if defined(CONFIG_VDE)
99 #include <libvdeplug.h>
100 #endif
101
102 #ifdef _WIN32
103 #include <windows.h>
104 #include <mmsystem.h>
105 #endif
106
107 #ifdef CONFIG_SDL
108 #if defined(__APPLE__) || defined(main)
109 #include <SDL.h>
110 int qemu_main(int argc, char **argv, char **envp);
111 int main(int argc, char **argv)
112 {
113 return qemu_main(argc, argv, NULL);
114 }
115 #undef main
116 #define main qemu_main
117 #endif
118 #endif /* CONFIG_SDL */
119
120 #ifdef CONFIG_COCOA
121 #undef main
122 #define main qemu_main
123 #endif /* CONFIG_COCOA */
124
125 #include "hw/hw.h"
126 #include "hw/boards.h"
127 #include "hw/usb.h"
128 #include "hw/pcmcia.h"
129 #include "hw/pc.h"
130 #include "hw/audiodev.h"
131 #include "hw/isa.h"
132 #include "hw/baum.h"
133 #include "hw/bt.h"
134 #include "hw/watchdog.h"
135 #include "hw/smbios.h"
136 #include "hw/xen.h"
137 #include "hw/qdev.h"
138 #include "hw/loader.h"
139 #include "bt-host.h"
140 #include "net.h"
141 #include "net/slirp.h"
142 #include "monitor.h"
143 #include "console.h"
144 #include "sysemu.h"
145 #include "gdbstub.h"
146 #include "qemu-timer.h"
147 #include "qemu-char.h"
148 #include "cache-utils.h"
149 #include "block.h"
150 #include "block_int.h"
151 #include "block-migration.h"
152 #include "dma.h"
153 #include "audio/audio.h"
154 #include "migration.h"
155 #include "kvm.h"
156 #include "balloon.h"
157 #include "qemu-option.h"
158 #include "qemu-config.h"
159 #include "qemu-objects.h"
160
161 #include "disas.h"
162
163 #include "exec-all.h"
164
165 #include "qemu_socket.h"
166
167 #include "slirp/libslirp.h"
168
169 #include "qemu-queue.h"
170
171 //#define DEBUG_NET
172 //#define DEBUG_SLIRP
173
174 #define DEFAULT_RAM_SIZE 128
175
176 #define MAX_VIRTIO_CONSOLES 1
177
178 static const char *data_dir;
179 const char *bios_name = NULL;
180 /* Note: drives_table[MAX_DRIVES] is a dummy block driver if none available
181 to store the VM snapshots */
182 struct drivelist drives = QTAILQ_HEAD_INITIALIZER(drives);
183 struct driveoptlist driveopts = QTAILQ_HEAD_INITIALIZER(driveopts);
184 enum vga_retrace_method vga_retrace_method = VGA_RETRACE_DUMB;
185 DisplayType display_type = DT_DEFAULT;
186 const char* keyboard_layout = NULL;
187 ram_addr_t ram_size;
188 const char *mem_path = NULL;
189 #ifdef MAP_POPULATE
190 int mem_prealloc = 0; /* force preallocation of physical target memory */
191 #endif
192 int nb_nics;
193 NICInfo nd_table[MAX_NICS];
194 int vm_running;
195 int autostart;
196 static int rtc_utc = 1;
197 static int rtc_date_offset = -1; /* -1 means no change */
198 QEMUClock *rtc_clock;
199 int vga_interface_type = VGA_NONE;
200 #ifdef TARGET_SPARC
201 int graphic_width = 1024;
202 int graphic_height = 768;
203 int graphic_depth = 8;
204 #else
205 int graphic_width = 800;
206 int graphic_height = 600;
207 int graphic_depth = 15;
208 #endif
209 static int full_screen = 0;
210 #ifdef CONFIG_SDL
211 static int no_frame = 0;
212 #endif
213 int no_quit = 0;
214 CharDriverState *serial_hds[MAX_SERIAL_PORTS];
215 CharDriverState *parallel_hds[MAX_PARALLEL_PORTS];
216 CharDriverState *virtcon_hds[MAX_VIRTIO_CONSOLES];
217 #ifdef TARGET_I386
218 int win2k_install_hack = 0;
219 int rtc_td_hack = 0;
220 #endif
221 int usb_enabled = 0;
222 int singlestep = 0;
223 int smp_cpus = 1;
224 int max_cpus = 0;
225 int smp_cores = 1;
226 int smp_threads = 1;
227 const char *vnc_display;
228 int acpi_enabled = 1;
229 int no_hpet = 0;
230 int fd_bootchk = 1;
231 int no_reboot = 0;
232 int no_shutdown = 0;
233 int cursor_hide = 1;
234 int graphic_rotate = 0;
235 uint8_t irq0override = 1;
236 #ifndef _WIN32
237 int daemonize = 0;
238 #endif
239 const char *watchdog;
240 const char *option_rom[MAX_OPTION_ROMS];
241 int nb_option_roms;
242 int semihosting_enabled = 0;
243 #ifdef TARGET_ARM
244 int old_param = 0;
245 #endif
246 const char *qemu_name;
247 int alt_grab = 0;
248 int ctrl_grab = 0;
249 #if defined(TARGET_SPARC) || defined(TARGET_PPC)
250 unsigned int nb_prom_envs = 0;
251 const char *prom_envs[MAX_PROM_ENVS];
252 #endif
253 int boot_menu;
254
255 int nb_numa_nodes;
256 uint64_t node_mem[MAX_NODES];
257 uint64_t node_cpumask[MAX_NODES];
258
259 static CPUState *cur_cpu;
260 static CPUState *next_cpu;
261 /* Conversion factor from emulated instructions to virtual clock ticks. */
262 static int icount_time_shift;
263 /* Arbitrarily pick 1MIPS as the minimum allowable speed. */
264 #define MAX_ICOUNT_SHIFT 10
265 /* Compensate for varying guest execution speed. */
266 static int64_t qemu_icount_bias;
267 static QEMUTimer *icount_rt_timer;
268 static QEMUTimer *icount_vm_timer;
269 static QEMUTimer *nographic_timer;
270
271 uint8_t qemu_uuid[16];
272
273 static QEMUBootSetHandler *boot_set_handler;
274 static void *boot_set_opaque;
275
276 #ifdef SIGRTMIN
277 #define SIG_IPI (SIGRTMIN+4)
278 #else
279 #define SIG_IPI SIGUSR1
280 #endif
281
282 static int default_serial = 1;
283 static int default_parallel = 1;
284 static int default_virtcon = 1;
285 static int default_monitor = 1;
286 static int default_vga = 1;
287 static int default_floppy = 1;
288 static int default_cdrom = 1;
289 static int default_sdcard = 1;
290
291 static struct {
292 const char *driver;
293 int *flag;
294 } default_list[] = {
295 { .driver = "isa-serial", .flag = &default_serial },
296 { .driver = "isa-parallel", .flag = &default_parallel },
297 { .driver = "isa-fdc", .flag = &default_floppy },
298 { .driver = "ide-drive", .flag = &default_cdrom },
299 { .driver = "virtio-serial-pci", .flag = &default_virtcon },
300 { .driver = "virtio-serial-s390", .flag = &default_virtcon },
301 { .driver = "virtio-serial", .flag = &default_virtcon },
302 { .driver = "VGA", .flag = &default_vga },
303 { .driver = "cirrus-vga", .flag = &default_vga },
304 { .driver = "vmware-svga", .flag = &default_vga },
305 };
306
307 static int default_driver_check(QemuOpts *opts, void *opaque)
308 {
309 const char *driver = qemu_opt_get(opts, "driver");
310 int i;
311
312 if (!driver)
313 return 0;
314 for (i = 0; i < ARRAY_SIZE(default_list); i++) {
315 if (strcmp(default_list[i].driver, driver) != 0)
316 continue;
317 *(default_list[i].flag) = 0;
318 }
319 return 0;
320 }
321
322 /***********************************************************/
323 /* x86 ISA bus support */
324
325 target_phys_addr_t isa_mem_base = 0;
326 PicState2 *isa_pic;
327
328 /***********************************************************/
329 void hw_error(const char *fmt, ...)
330 {
331 va_list ap;
332 CPUState *env;
333
334 va_start(ap, fmt);
335 fprintf(stderr, "qemu: hardware error: ");
336 vfprintf(stderr, fmt, ap);
337 fprintf(stderr, "\n");
338 for(env = first_cpu; env != NULL; env = env->next_cpu) {
339 fprintf(stderr, "CPU #%d:\n", env->cpu_index);
340 #ifdef TARGET_I386
341 cpu_dump_state(env, stderr, fprintf, X86_DUMP_FPU);
342 #else
343 cpu_dump_state(env, stderr, fprintf, 0);
344 #endif
345 }
346 va_end(ap);
347 abort();
348 }
349
350 static void set_proc_name(const char *s)
351 {
352 #if defined(__linux__) && defined(PR_SET_NAME)
353 char name[16];
354 if (!s)
355 return;
356 name[sizeof(name) - 1] = 0;
357 strncpy(name, s, sizeof(name));
358 /* Could rewrite argv[0] too, but that's a bit more complicated.
359 This simple way is enough for `top'. */
360 prctl(PR_SET_NAME, name);
361 #endif
362 }
363
364 /***************/
365 /* ballooning */
366
367 static QEMUBalloonEvent *qemu_balloon_event;
368 void *qemu_balloon_event_opaque;
369
370 void qemu_add_balloon_handler(QEMUBalloonEvent *func, void *opaque)
371 {
372 qemu_balloon_event = func;
373 qemu_balloon_event_opaque = opaque;
374 }
375
376 int qemu_balloon(ram_addr_t target, MonitorCompletion cb, void *opaque)
377 {
378 if (qemu_balloon_event) {
379 qemu_balloon_event(qemu_balloon_event_opaque, target, cb, opaque);
380 return 1;
381 } else {
382 return 0;
383 }
384 }
385
386 int qemu_balloon_status(MonitorCompletion cb, void *opaque)
387 {
388 if (qemu_balloon_event) {
389 qemu_balloon_event(qemu_balloon_event_opaque, 0, cb, opaque);
390 return 1;
391 } else {
392 return 0;
393 }
394 }
395
396
397 /***********************************************************/
398 /* real time host monotonic timer */
399
400 /* compute with 96 bit intermediate result: (a*b)/c */
401 uint64_t muldiv64(uint64_t a, uint32_t b, uint32_t c)
402 {
403 union {
404 uint64_t ll;
405 struct {
406 #ifdef HOST_WORDS_BIGENDIAN
407 uint32_t high, low;
408 #else
409 uint32_t low, high;
410 #endif
411 } l;
412 } u, res;
413 uint64_t rl, rh;
414
415 u.ll = a;
416 rl = (uint64_t)u.l.low * (uint64_t)b;
417 rh = (uint64_t)u.l.high * (uint64_t)b;
418 rh += (rl >> 32);
419 res.l.high = rh / c;
420 res.l.low = (((rh % c) << 32) + (rl & 0xffffffff)) / c;
421 return res.ll;
422 }
423
424 static int64_t get_clock_realtime(void)
425 {
426 struct timeval tv;
427
428 gettimeofday(&tv, NULL);
429 return tv.tv_sec * 1000000000LL + (tv.tv_usec * 1000);
430 }
431
432 #ifdef WIN32
433
434 static int64_t clock_freq;
435
436 static void init_get_clock(void)
437 {
438 LARGE_INTEGER freq;
439 int ret;
440 ret = QueryPerformanceFrequency(&freq);
441 if (ret == 0) {
442 fprintf(stderr, "Could not calibrate ticks\n");
443 exit(1);
444 }
445 clock_freq = freq.QuadPart;
446 }
447
448 static int64_t get_clock(void)
449 {
450 LARGE_INTEGER ti;
451 QueryPerformanceCounter(&ti);
452 return muldiv64(ti.QuadPart, get_ticks_per_sec(), clock_freq);
453 }
454
455 #else
456
457 static int use_rt_clock;
458
459 static void init_get_clock(void)
460 {
461 use_rt_clock = 0;
462 #if defined(__linux__) || (defined(__FreeBSD__) && __FreeBSD_version >= 500000) \
463 || defined(__DragonFly__) || defined(__FreeBSD_kernel__)
464 {
465 struct timespec ts;
466 if (clock_gettime(CLOCK_MONOTONIC, &ts) == 0) {
467 use_rt_clock = 1;
468 }
469 }
470 #endif
471 }
472
473 static int64_t get_clock(void)
474 {
475 #if defined(__linux__) || (defined(__FreeBSD__) && __FreeBSD_version >= 500000) \
476 || defined(__DragonFly__) || defined(__FreeBSD_kernel__)
477 if (use_rt_clock) {
478 struct timespec ts;
479 clock_gettime(CLOCK_MONOTONIC, &ts);
480 return ts.tv_sec * 1000000000LL + ts.tv_nsec;
481 } else
482 #endif
483 {
484 /* XXX: using gettimeofday leads to problems if the date
485 changes, so it should be avoided. */
486 return get_clock_realtime();
487 }
488 }
489 #endif
490
491 /* Return the virtual CPU time, based on the instruction counter. */
492 static int64_t cpu_get_icount(void)
493 {
494 int64_t icount;
495 CPUState *env = cpu_single_env;;
496 icount = qemu_icount;
497 if (env) {
498 if (!can_do_io(env))
499 fprintf(stderr, "Bad clock read\n");
500 icount -= (env->icount_decr.u16.low + env->icount_extra);
501 }
502 return qemu_icount_bias + (icount << icount_time_shift);
503 }
504
505 /***********************************************************/
506 /* guest cycle counter */
507
508 typedef struct TimersState {
509 int64_t cpu_ticks_prev;
510 int64_t cpu_ticks_offset;
511 int64_t cpu_clock_offset;
512 int32_t cpu_ticks_enabled;
513 int64_t dummy;
514 } TimersState;
515
516 TimersState timers_state;
517
518 /* return the host CPU cycle counter and handle stop/restart */
519 int64_t cpu_get_ticks(void)
520 {
521 if (use_icount) {
522 return cpu_get_icount();
523 }
524 if (!timers_state.cpu_ticks_enabled) {
525 return timers_state.cpu_ticks_offset;
526 } else {
527 int64_t ticks;
528 ticks = cpu_get_real_ticks();
529 if (timers_state.cpu_ticks_prev > ticks) {
530 /* Note: non increasing ticks may happen if the host uses
531 software suspend */
532 timers_state.cpu_ticks_offset += timers_state.cpu_ticks_prev - ticks;
533 }
534 timers_state.cpu_ticks_prev = ticks;
535 return ticks + timers_state.cpu_ticks_offset;
536 }
537 }
538
539 /* return the host CPU monotonic timer and handle stop/restart */
540 static int64_t cpu_get_clock(void)
541 {
542 int64_t ti;
543 if (!timers_state.cpu_ticks_enabled) {
544 return timers_state.cpu_clock_offset;
545 } else {
546 ti = get_clock();
547 return ti + timers_state.cpu_clock_offset;
548 }
549 }
550
551 /* enable cpu_get_ticks() */
552 void cpu_enable_ticks(void)
553 {
554 if (!timers_state.cpu_ticks_enabled) {
555 timers_state.cpu_ticks_offset -= cpu_get_real_ticks();
556 timers_state.cpu_clock_offset -= get_clock();
557 timers_state.cpu_ticks_enabled = 1;
558 }
559 }
560
561 /* disable cpu_get_ticks() : the clock is stopped. You must not call
562 cpu_get_ticks() after that. */
563 void cpu_disable_ticks(void)
564 {
565 if (timers_state.cpu_ticks_enabled) {
566 timers_state.cpu_ticks_offset = cpu_get_ticks();
567 timers_state.cpu_clock_offset = cpu_get_clock();
568 timers_state.cpu_ticks_enabled = 0;
569 }
570 }
571
572 /***********************************************************/
573 /* timers */
574
575 #define QEMU_CLOCK_REALTIME 0
576 #define QEMU_CLOCK_VIRTUAL 1
577 #define QEMU_CLOCK_HOST 2
578
579 struct QEMUClock {
580 int type;
581 int enabled;
582 /* XXX: add frequency */
583 };
584
585 struct QEMUTimer {
586 QEMUClock *clock;
587 int64_t expire_time;
588 QEMUTimerCB *cb;
589 void *opaque;
590 struct QEMUTimer *next;
591 };
592
593 struct qemu_alarm_timer {
594 char const *name;
595 int (*start)(struct qemu_alarm_timer *t);
596 void (*stop)(struct qemu_alarm_timer *t);
597 void (*rearm)(struct qemu_alarm_timer *t);
598 void *priv;
599
600 char expired;
601 char pending;
602 };
603
604 static inline int alarm_has_dynticks(struct qemu_alarm_timer *t)
605 {
606 return !!t->rearm;
607 }
608
609 static void qemu_rearm_alarm_timer(struct qemu_alarm_timer *t)
610 {
611 if (!alarm_has_dynticks(t))
612 return;
613
614 t->rearm(t);
615 }
616
617 /* TODO: MIN_TIMER_REARM_US should be optimized */
618 #define MIN_TIMER_REARM_US 250
619
620 static struct qemu_alarm_timer *alarm_timer;
621
622 #ifdef _WIN32
623
624 struct qemu_alarm_win32 {
625 MMRESULT timerId;
626 unsigned int period;
627 } alarm_win32_data = {0, 0};
628
629 static int win32_start_timer(struct qemu_alarm_timer *t);
630 static void win32_stop_timer(struct qemu_alarm_timer *t);
631 static void win32_rearm_timer(struct qemu_alarm_timer *t);
632
633 #else
634
635 static int unix_start_timer(struct qemu_alarm_timer *t);
636 static void unix_stop_timer(struct qemu_alarm_timer *t);
637
638 #ifdef __linux__
639
640 static int dynticks_start_timer(struct qemu_alarm_timer *t);
641 static void dynticks_stop_timer(struct qemu_alarm_timer *t);
642 static void dynticks_rearm_timer(struct qemu_alarm_timer *t);
643
644 static int hpet_start_timer(struct qemu_alarm_timer *t);
645 static void hpet_stop_timer(struct qemu_alarm_timer *t);
646
647 static int rtc_start_timer(struct qemu_alarm_timer *t);
648 static void rtc_stop_timer(struct qemu_alarm_timer *t);
649
650 #endif /* __linux__ */
651
652 #endif /* _WIN32 */
653
654 /* Correlation between real and virtual time is always going to be
655 fairly approximate, so ignore small variation.
656 When the guest is idle real and virtual time will be aligned in
657 the IO wait loop. */
658 #define ICOUNT_WOBBLE (get_ticks_per_sec() / 10)
659
660 static void icount_adjust(void)
661 {
662 int64_t cur_time;
663 int64_t cur_icount;
664 int64_t delta;
665 static int64_t last_delta;
666 /* If the VM is not running, then do nothing. */
667 if (!vm_running)
668 return;
669
670 cur_time = cpu_get_clock();
671 cur_icount = qemu_get_clock(vm_clock);
672 delta = cur_icount - cur_time;
673 /* FIXME: This is a very crude algorithm, somewhat prone to oscillation. */
674 if (delta > 0
675 && last_delta + ICOUNT_WOBBLE < delta * 2
676 && icount_time_shift > 0) {
677 /* The guest is getting too far ahead. Slow time down. */
678 icount_time_shift--;
679 }
680 if (delta < 0
681 && last_delta - ICOUNT_WOBBLE > delta * 2
682 && icount_time_shift < MAX_ICOUNT_SHIFT) {
683 /* The guest is getting too far behind. Speed time up. */
684 icount_time_shift++;
685 }
686 last_delta = delta;
687 qemu_icount_bias = cur_icount - (qemu_icount << icount_time_shift);
688 }
689
690 static void icount_adjust_rt(void * opaque)
691 {
692 qemu_mod_timer(icount_rt_timer,
693 qemu_get_clock(rt_clock) + 1000);
694 icount_adjust();
695 }
696
697 static void icount_adjust_vm(void * opaque)
698 {
699 qemu_mod_timer(icount_vm_timer,
700 qemu_get_clock(vm_clock) + get_ticks_per_sec() / 10);
701 icount_adjust();
702 }
703
704 static void init_icount_adjust(void)
705 {
706 /* Have both realtime and virtual time triggers for speed adjustment.
707 The realtime trigger catches emulated time passing too slowly,
708 the virtual time trigger catches emulated time passing too fast.
709 Realtime triggers occur even when idle, so use them less frequently
710 than VM triggers. */
711 icount_rt_timer = qemu_new_timer(rt_clock, icount_adjust_rt, NULL);
712 qemu_mod_timer(icount_rt_timer,
713 qemu_get_clock(rt_clock) + 1000);
714 icount_vm_timer = qemu_new_timer(vm_clock, icount_adjust_vm, NULL);
715 qemu_mod_timer(icount_vm_timer,
716 qemu_get_clock(vm_clock) + get_ticks_per_sec() / 10);
717 }
718
719 static struct qemu_alarm_timer alarm_timers[] = {
720 #ifndef _WIN32
721 #ifdef __linux__
722 {"dynticks", dynticks_start_timer,
723 dynticks_stop_timer, dynticks_rearm_timer, NULL},
724 /* HPET - if available - is preferred */
725 {"hpet", hpet_start_timer, hpet_stop_timer, NULL, NULL},
726 /* ...otherwise try RTC */
727 {"rtc", rtc_start_timer, rtc_stop_timer, NULL, NULL},
728 #endif
729 {"unix", unix_start_timer, unix_stop_timer, NULL, NULL},
730 #else
731 {"dynticks", win32_start_timer,
732 win32_stop_timer, win32_rearm_timer, &alarm_win32_data},
733 {"win32", win32_start_timer,
734 win32_stop_timer, NULL, &alarm_win32_data},
735 #endif
736 {NULL, }
737 };
738
739 static void show_available_alarms(void)
740 {
741 int i;
742
743 printf("Available alarm timers, in order of precedence:\n");
744 for (i = 0; alarm_timers[i].name; i++)
745 printf("%s\n", alarm_timers[i].name);
746 }
747
748 static void configure_alarms(char const *opt)
749 {
750 int i;
751 int cur = 0;
752 int count = ARRAY_SIZE(alarm_timers) - 1;
753 char *arg;
754 char *name;
755 struct qemu_alarm_timer tmp;
756
757 if (!strcmp(opt, "?")) {
758 show_available_alarms();
759 exit(0);
760 }
761
762 arg = qemu_strdup(opt);
763
764 /* Reorder the array */
765 name = strtok(arg, ",");
766 while (name) {
767 for (i = 0; i < count && alarm_timers[i].name; i++) {
768 if (!strcmp(alarm_timers[i].name, name))
769 break;
770 }
771
772 if (i == count) {
773 fprintf(stderr, "Unknown clock %s\n", name);
774 goto next;
775 }
776
777 if (i < cur)
778 /* Ignore */
779 goto next;
780
781 /* Swap */
782 tmp = alarm_timers[i];
783 alarm_timers[i] = alarm_timers[cur];
784 alarm_timers[cur] = tmp;
785
786 cur++;
787 next:
788 name = strtok(NULL, ",");
789 }
790
791 qemu_free(arg);
792
793 if (cur) {
794 /* Disable remaining timers */
795 for (i = cur; i < count; i++)
796 alarm_timers[i].name = NULL;
797 } else {
798 show_available_alarms();
799 exit(1);
800 }
801 }
802
803 #define QEMU_NUM_CLOCKS 3
804
805 QEMUClock *rt_clock;
806 QEMUClock *vm_clock;
807 QEMUClock *host_clock;
808
809 static QEMUTimer *active_timers[QEMU_NUM_CLOCKS];
810
811 static QEMUClock *qemu_new_clock(int type)
812 {
813 QEMUClock *clock;
814 clock = qemu_mallocz(sizeof(QEMUClock));
815 clock->type = type;
816 clock->enabled = 1;
817 return clock;
818 }
819
820 static void qemu_clock_enable(QEMUClock *clock, int enabled)
821 {
822 clock->enabled = enabled;
823 }
824
825 QEMUTimer *qemu_new_timer(QEMUClock *clock, QEMUTimerCB *cb, void *opaque)
826 {
827 QEMUTimer *ts;
828
829 ts = qemu_mallocz(sizeof(QEMUTimer));
830 ts->clock = clock;
831 ts->cb = cb;
832 ts->opaque = opaque;
833 return ts;
834 }
835
836 void qemu_free_timer(QEMUTimer *ts)
837 {
838 qemu_free(ts);
839 }
840
841 /* stop a timer, but do not dealloc it */
842 void qemu_del_timer(QEMUTimer *ts)
843 {
844 QEMUTimer **pt, *t;
845
846 /* NOTE: this code must be signal safe because
847 qemu_timer_expired() can be called from a signal. */
848 pt = &active_timers[ts->clock->type];
849 for(;;) {
850 t = *pt;
851 if (!t)
852 break;
853 if (t == ts) {
854 *pt = t->next;
855 break;
856 }
857 pt = &t->next;
858 }
859 }
860
861 /* modify the current timer so that it will be fired when current_time
862 >= expire_time. The corresponding callback will be called. */
863 void qemu_mod_timer(QEMUTimer *ts, int64_t expire_time)
864 {
865 QEMUTimer **pt, *t;
866
867 qemu_del_timer(ts);
868
869 /* add the timer in the sorted list */
870 /* NOTE: this code must be signal safe because
871 qemu_timer_expired() can be called from a signal. */
872 pt = &active_timers[ts->clock->type];
873 for(;;) {
874 t = *pt;
875 if (!t)
876 break;
877 if (t->expire_time > expire_time)
878 break;
879 pt = &t->next;
880 }
881 ts->expire_time = expire_time;
882 ts->next = *pt;
883 *pt = ts;
884
885 /* Rearm if necessary */
886 if (pt == &active_timers[ts->clock->type]) {
887 if (!alarm_timer->pending) {
888 qemu_rearm_alarm_timer(alarm_timer);
889 }
890 /* Interrupt execution to force deadline recalculation. */
891 if (use_icount)
892 qemu_notify_event();
893 }
894 }
895
896 int qemu_timer_pending(QEMUTimer *ts)
897 {
898 QEMUTimer *t;
899 for(t = active_timers[ts->clock->type]; t != NULL; t = t->next) {
900 if (t == ts)
901 return 1;
902 }
903 return 0;
904 }
905
906 int qemu_timer_expired(QEMUTimer *timer_head, int64_t current_time)
907 {
908 if (!timer_head)
909 return 0;
910 return (timer_head->expire_time <= current_time);
911 }
912
913 static void qemu_run_timers(QEMUClock *clock)
914 {
915 QEMUTimer **ptimer_head, *ts;
916 int64_t current_time;
917
918 if (!clock->enabled)
919 return;
920
921 current_time = qemu_get_clock (clock);
922 ptimer_head = &active_timers[clock->type];
923 for(;;) {
924 ts = *ptimer_head;
925 if (!ts || ts->expire_time > current_time)
926 break;
927 /* remove timer from the list before calling the callback */
928 *ptimer_head = ts->next;
929 ts->next = NULL;
930
931 /* run the callback (the timer list can be modified) */
932 ts->cb(ts->opaque);
933 }
934 }
935
936 int64_t qemu_get_clock(QEMUClock *clock)
937 {
938 switch(clock->type) {
939 case QEMU_CLOCK_REALTIME:
940 return get_clock() / 1000000;
941 default:
942 case QEMU_CLOCK_VIRTUAL:
943 if (use_icount) {
944 return cpu_get_icount();
945 } else {
946 return cpu_get_clock();
947 }
948 case QEMU_CLOCK_HOST:
949 return get_clock_realtime();
950 }
951 }
952
953 int64_t qemu_get_clock_ns(QEMUClock *clock)
954 {
955 switch(clock->type) {
956 case QEMU_CLOCK_REALTIME:
957 return get_clock();
958 default:
959 case QEMU_CLOCK_VIRTUAL:
960 if (use_icount) {
961 return cpu_get_icount();
962 } else {
963 return cpu_get_clock();
964 }
965 case QEMU_CLOCK_HOST:
966 return get_clock_realtime();
967 }
968 }
969
970 static void init_clocks(void)
971 {
972 init_get_clock();
973 rt_clock = qemu_new_clock(QEMU_CLOCK_REALTIME);
974 vm_clock = qemu_new_clock(QEMU_CLOCK_VIRTUAL);
975 host_clock = qemu_new_clock(QEMU_CLOCK_HOST);
976
977 rtc_clock = host_clock;
978 }
979
980 /* save a timer */
981 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
982 {
983 uint64_t expire_time;
984
985 if (qemu_timer_pending(ts)) {
986 expire_time = ts->expire_time;
987 } else {
988 expire_time = -1;
989 }
990 qemu_put_be64(f, expire_time);
991 }
992
993 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
994 {
995 uint64_t expire_time;
996
997 expire_time = qemu_get_be64(f);
998 if (expire_time != -1) {
999 qemu_mod_timer(ts, expire_time);
1000 } else {
1001 qemu_del_timer(ts);
1002 }
1003 }
1004
1005 static const VMStateDescription vmstate_timers = {
1006 .name = "timer",
1007 .version_id = 2,
1008 .minimum_version_id = 1,
1009 .minimum_version_id_old = 1,
1010 .fields = (VMStateField []) {
1011 VMSTATE_INT64(cpu_ticks_offset, TimersState),
1012 VMSTATE_INT64(dummy, TimersState),
1013 VMSTATE_INT64_V(cpu_clock_offset, TimersState, 2),
1014 VMSTATE_END_OF_LIST()
1015 }
1016 };
1017
1018 static void qemu_run_all_timers(void)
1019 {
1020 /* rearm timer, if not periodic */
1021 if (alarm_timer->expired) {
1022 alarm_timer->expired = 0;
1023 qemu_rearm_alarm_timer(alarm_timer);
1024 }
1025
1026 alarm_timer->pending = 0;
1027
1028 /* vm time timers */
1029 if (vm_running) {
1030 qemu_run_timers(vm_clock);
1031 }
1032
1033 qemu_run_timers(rt_clock);
1034 qemu_run_timers(host_clock);
1035 }
1036
1037 #ifdef _WIN32
1038 static void CALLBACK host_alarm_handler(UINT uTimerID, UINT uMsg,
1039 DWORD_PTR dwUser, DWORD_PTR dw1,
1040 DWORD_PTR dw2)
1041 #else
1042 static void host_alarm_handler(int host_signum)
1043 #endif
1044 {
1045 struct qemu_alarm_timer *t = alarm_timer;
1046 if (!t)
1047 return;
1048
1049 #if 0
1050 #define DISP_FREQ 1000
1051 {
1052 static int64_t delta_min = INT64_MAX;
1053 static int64_t delta_max, delta_cum, last_clock, delta, ti;
1054 static int count;
1055 ti = qemu_get_clock(vm_clock);
1056 if (last_clock != 0) {
1057 delta = ti - last_clock;
1058 if (delta < delta_min)
1059 delta_min = delta;
1060 if (delta > delta_max)
1061 delta_max = delta;
1062 delta_cum += delta;
1063 if (++count == DISP_FREQ) {
1064 printf("timer: min=%" PRId64 " us max=%" PRId64 " us avg=%" PRId64 " us avg_freq=%0.3f Hz\n",
1065 muldiv64(delta_min, 1000000, get_ticks_per_sec()),
1066 muldiv64(delta_max, 1000000, get_ticks_per_sec()),
1067 muldiv64(delta_cum, 1000000 / DISP_FREQ, get_ticks_per_sec()),
1068 (double)get_ticks_per_sec() / ((double)delta_cum / DISP_FREQ));
1069 count = 0;
1070 delta_min = INT64_MAX;
1071 delta_max = 0;
1072 delta_cum = 0;
1073 }
1074 }
1075 last_clock = ti;
1076 }
1077 #endif
1078 if (alarm_has_dynticks(t) ||
1079 (!use_icount &&
1080 qemu_timer_expired(active_timers[QEMU_CLOCK_VIRTUAL],
1081 qemu_get_clock(vm_clock))) ||
1082 qemu_timer_expired(active_timers[QEMU_CLOCK_REALTIME],
1083 qemu_get_clock(rt_clock)) ||
1084 qemu_timer_expired(active_timers[QEMU_CLOCK_HOST],
1085 qemu_get_clock(host_clock))) {
1086
1087 t->expired = alarm_has_dynticks(t);
1088 t->pending = 1;
1089 qemu_notify_event();
1090 }
1091 }
1092
1093 static int64_t qemu_next_deadline(void)
1094 {
1095 /* To avoid problems with overflow limit this to 2^32. */
1096 int64_t delta = INT32_MAX;
1097
1098 if (active_timers[QEMU_CLOCK_VIRTUAL]) {
1099 delta = active_timers[QEMU_CLOCK_VIRTUAL]->expire_time -
1100 qemu_get_clock(vm_clock);
1101 }
1102 if (active_timers[QEMU_CLOCK_HOST]) {
1103 int64_t hdelta = active_timers[QEMU_CLOCK_HOST]->expire_time -
1104 qemu_get_clock(host_clock);
1105 if (hdelta < delta)
1106 delta = hdelta;
1107 }
1108
1109 if (delta < 0)
1110 delta = 0;
1111
1112 return delta;
1113 }
1114
1115 #if defined(__linux__)
1116 static uint64_t qemu_next_deadline_dyntick(void)
1117 {
1118 int64_t delta;
1119 int64_t rtdelta;
1120
1121 if (use_icount)
1122 delta = INT32_MAX;
1123 else
1124 delta = (qemu_next_deadline() + 999) / 1000;
1125
1126 if (active_timers[QEMU_CLOCK_REALTIME]) {
1127 rtdelta = (active_timers[QEMU_CLOCK_REALTIME]->expire_time -
1128 qemu_get_clock(rt_clock))*1000;
1129 if (rtdelta < delta)
1130 delta = rtdelta;
1131 }
1132
1133 if (delta < MIN_TIMER_REARM_US)
1134 delta = MIN_TIMER_REARM_US;
1135
1136 return delta;
1137 }
1138 #endif
1139
1140 #ifndef _WIN32
1141
1142 /* Sets a specific flag */
1143 static int fcntl_setfl(int fd, int flag)
1144 {
1145 int flags;
1146
1147 flags = fcntl(fd, F_GETFL);
1148 if (flags == -1)
1149 return -errno;
1150
1151 if (fcntl(fd, F_SETFL, flags | flag) == -1)
1152 return -errno;
1153
1154 return 0;
1155 }
1156
1157 #if defined(__linux__)
1158
1159 #define RTC_FREQ 1024
1160
1161 static void enable_sigio_timer(int fd)
1162 {
1163 struct sigaction act;
1164
1165 /* timer signal */
1166 sigfillset(&act.sa_mask);
1167 act.sa_flags = 0;
1168 act.sa_handler = host_alarm_handler;
1169
1170 sigaction(SIGIO, &act, NULL);
1171 fcntl_setfl(fd, O_ASYNC);
1172 fcntl(fd, F_SETOWN, getpid());
1173 }
1174
1175 static int hpet_start_timer(struct qemu_alarm_timer *t)
1176 {
1177 struct hpet_info info;
1178 int r, fd;
1179
1180 fd = qemu_open("/dev/hpet", O_RDONLY);
1181 if (fd < 0)
1182 return -1;
1183
1184 /* Set frequency */
1185 r = ioctl(fd, HPET_IRQFREQ, RTC_FREQ);
1186 if (r < 0) {
1187 fprintf(stderr, "Could not configure '/dev/hpet' to have a 1024Hz timer. This is not a fatal\n"
1188 "error, but for better emulation accuracy type:\n"
1189 "'echo 1024 > /proc/sys/dev/hpet/max-user-freq' as root.\n");
1190 goto fail;
1191 }
1192
1193 /* Check capabilities */
1194 r = ioctl(fd, HPET_INFO, &info);
1195 if (r < 0)
1196 goto fail;
1197
1198 /* Enable periodic mode */
1199 r = ioctl(fd, HPET_EPI, 0);
1200 if (info.hi_flags && (r < 0))
1201 goto fail;
1202
1203 /* Enable interrupt */
1204 r = ioctl(fd, HPET_IE_ON, 0);
1205 if (r < 0)
1206 goto fail;
1207
1208 enable_sigio_timer(fd);
1209 t->priv = (void *)(long)fd;
1210
1211 return 0;
1212 fail:
1213 close(fd);
1214 return -1;
1215 }
1216
1217 static void hpet_stop_timer(struct qemu_alarm_timer *t)
1218 {
1219 int fd = (long)t->priv;
1220
1221 close(fd);
1222 }
1223
1224 static int rtc_start_timer(struct qemu_alarm_timer *t)
1225 {
1226 int rtc_fd;
1227 unsigned long current_rtc_freq = 0;
1228
1229 TFR(rtc_fd = qemu_open("/dev/rtc", O_RDONLY));
1230 if (rtc_fd < 0)
1231 return -1;
1232 ioctl(rtc_fd, RTC_IRQP_READ, &current_rtc_freq);
1233 if (current_rtc_freq != RTC_FREQ &&
1234 ioctl(rtc_fd, RTC_IRQP_SET, RTC_FREQ) < 0) {
1235 fprintf(stderr, "Could not configure '/dev/rtc' to have a 1024 Hz timer. This is not a fatal\n"
1236 "error, but for better emulation accuracy either use a 2.6 host Linux kernel or\n"
1237 "type 'echo 1024 > /proc/sys/dev/rtc/max-user-freq' as root.\n");
1238 goto fail;
1239 }
1240 if (ioctl(rtc_fd, RTC_PIE_ON, 0) < 0) {
1241 fail:
1242 close(rtc_fd);
1243 return -1;
1244 }
1245
1246 enable_sigio_timer(rtc_fd);
1247
1248 t->priv = (void *)(long)rtc_fd;
1249
1250 return 0;
1251 }
1252
1253 static void rtc_stop_timer(struct qemu_alarm_timer *t)
1254 {
1255 int rtc_fd = (long)t->priv;
1256
1257 close(rtc_fd);
1258 }
1259
1260 static int dynticks_start_timer(struct qemu_alarm_timer *t)
1261 {
1262 struct sigevent ev;
1263 timer_t host_timer;
1264 struct sigaction act;
1265
1266 sigfillset(&act.sa_mask);
1267 act.sa_flags = 0;
1268 act.sa_handler = host_alarm_handler;
1269
1270 sigaction(SIGALRM, &act, NULL);
1271
1272 /*
1273 * Initialize ev struct to 0 to avoid valgrind complaining
1274 * about uninitialized data in timer_create call
1275 */
1276 memset(&ev, 0, sizeof(ev));
1277 ev.sigev_value.sival_int = 0;
1278 ev.sigev_notify = SIGEV_SIGNAL;
1279 ev.sigev_signo = SIGALRM;
1280
1281 if (timer_create(CLOCK_REALTIME, &ev, &host_timer)) {
1282 perror("timer_create");
1283
1284 /* disable dynticks */
1285 fprintf(stderr, "Dynamic Ticks disabled\n");
1286
1287 return -1;
1288 }
1289
1290 t->priv = (void *)(long)host_timer;
1291
1292 return 0;
1293 }
1294
1295 static void dynticks_stop_timer(struct qemu_alarm_timer *t)
1296 {
1297 timer_t host_timer = (timer_t)(long)t->priv;
1298
1299 timer_delete(host_timer);
1300 }
1301
1302 static void dynticks_rearm_timer(struct qemu_alarm_timer *t)
1303 {
1304 timer_t host_timer = (timer_t)(long)t->priv;
1305 struct itimerspec timeout;
1306 int64_t nearest_delta_us = INT64_MAX;
1307 int64_t current_us;
1308
1309 assert(alarm_has_dynticks(t));
1310 if (!active_timers[QEMU_CLOCK_REALTIME] &&
1311 !active_timers[QEMU_CLOCK_VIRTUAL] &&
1312 !active_timers[QEMU_CLOCK_HOST])
1313 return;
1314
1315 nearest_delta_us = qemu_next_deadline_dyntick();
1316
1317 /* check whether a timer is already running */
1318 if (timer_gettime(host_timer, &timeout)) {
1319 perror("gettime");
1320 fprintf(stderr, "Internal timer error: aborting\n");
1321 exit(1);
1322 }
1323 current_us = timeout.it_value.tv_sec * 1000000 + timeout.it_value.tv_nsec/1000;
1324 if (current_us && current_us <= nearest_delta_us)
1325 return;
1326
1327 timeout.it_interval.tv_sec = 0;
1328 timeout.it_interval.tv_nsec = 0; /* 0 for one-shot timer */
1329 timeout.it_value.tv_sec = nearest_delta_us / 1000000;
1330 timeout.it_value.tv_nsec = (nearest_delta_us % 1000000) * 1000;
1331 if (timer_settime(host_timer, 0 /* RELATIVE */, &timeout, NULL)) {
1332 perror("settime");
1333 fprintf(stderr, "Internal timer error: aborting\n");
1334 exit(1);
1335 }
1336 }
1337
1338 #endif /* defined(__linux__) */
1339
1340 static int unix_start_timer(struct qemu_alarm_timer *t)
1341 {
1342 struct sigaction act;
1343 struct itimerval itv;
1344 int err;
1345
1346 /* timer signal */
1347 sigfillset(&act.sa_mask);
1348 act.sa_flags = 0;
1349 act.sa_handler = host_alarm_handler;
1350
1351 sigaction(SIGALRM, &act, NULL);
1352
1353 itv.it_interval.tv_sec = 0;
1354 /* for i386 kernel 2.6 to get 1 ms */
1355 itv.it_interval.tv_usec = 999;
1356 itv.it_value.tv_sec = 0;
1357 itv.it_value.tv_usec = 10 * 1000;
1358
1359 err = setitimer(ITIMER_REAL, &itv, NULL);
1360 if (err)
1361 return -1;
1362
1363 return 0;
1364 }
1365
1366 static void unix_stop_timer(struct qemu_alarm_timer *t)
1367 {
1368 struct itimerval itv;
1369
1370 memset(&itv, 0, sizeof(itv));
1371 setitimer(ITIMER_REAL, &itv, NULL);
1372 }
1373
1374 #endif /* !defined(_WIN32) */
1375
1376
1377 #ifdef _WIN32
1378
1379 static int win32_start_timer(struct qemu_alarm_timer *t)
1380 {
1381 TIMECAPS tc;
1382 struct qemu_alarm_win32 *data = t->priv;
1383 UINT flags;
1384
1385 memset(&tc, 0, sizeof(tc));
1386 timeGetDevCaps(&tc, sizeof(tc));
1387
1388 data->period = tc.wPeriodMin;
1389 timeBeginPeriod(data->period);
1390
1391 flags = TIME_CALLBACK_FUNCTION;
1392 if (alarm_has_dynticks(t))
1393 flags |= TIME_ONESHOT;
1394 else
1395 flags |= TIME_PERIODIC;
1396
1397 data->timerId = timeSetEvent(1, // interval (ms)
1398 data->period, // resolution
1399 host_alarm_handler, // function
1400 (DWORD)t, // parameter
1401 flags);
1402
1403 if (!data->timerId) {
1404 fprintf(stderr, "Failed to initialize win32 alarm timer: %ld\n",
1405 GetLastError());
1406 timeEndPeriod(data->period);
1407 return -1;
1408 }
1409
1410 return 0;
1411 }
1412
1413 static void win32_stop_timer(struct qemu_alarm_timer *t)
1414 {
1415 struct qemu_alarm_win32 *data = t->priv;
1416
1417 timeKillEvent(data->timerId);
1418 timeEndPeriod(data->period);
1419 }
1420
1421 static void win32_rearm_timer(struct qemu_alarm_timer *t)
1422 {
1423 struct qemu_alarm_win32 *data = t->priv;
1424
1425 assert(alarm_has_dynticks(t));
1426 if (!active_timers[QEMU_CLOCK_REALTIME] &&
1427 !active_timers[QEMU_CLOCK_VIRTUAL] &&
1428 !active_timers[QEMU_CLOCK_HOST])
1429 return;
1430
1431 timeKillEvent(data->timerId);
1432
1433 data->timerId = timeSetEvent(1,
1434 data->period,
1435 host_alarm_handler,
1436 (DWORD)t,
1437 TIME_ONESHOT | TIME_CALLBACK_FUNCTION);
1438
1439 if (!data->timerId) {
1440 fprintf(stderr, "Failed to re-arm win32 alarm timer %ld\n",
1441 GetLastError());
1442
1443 timeEndPeriod(data->period);
1444 exit(1);
1445 }
1446 }
1447
1448 #endif /* _WIN32 */
1449
1450 static void alarm_timer_on_change_state_rearm(void *opaque, int running, int reason)
1451 {
1452 if (running)
1453 qemu_rearm_alarm_timer((struct qemu_alarm_timer *) opaque);
1454 }
1455
1456 static int init_timer_alarm(void)
1457 {
1458 struct qemu_alarm_timer *t = NULL;
1459 int i, err = -1;
1460
1461 for (i = 0; alarm_timers[i].name; i++) {
1462 t = &alarm_timers[i];
1463
1464 err = t->start(t);
1465 if (!err)
1466 break;
1467 }
1468
1469 if (err) {
1470 err = -ENOENT;
1471 goto fail;
1472 }
1473
1474 /* first event is at time 0 */
1475 t->pending = 1;
1476 alarm_timer = t;
1477 qemu_add_vm_change_state_handler(alarm_timer_on_change_state_rearm, t);
1478
1479 return 0;
1480
1481 fail:
1482 return err;
1483 }
1484
1485 static void quit_timers(void)
1486 {
1487 struct qemu_alarm_timer *t = alarm_timer;
1488 alarm_timer = NULL;
1489 t->stop(t);
1490 }
1491
1492 /***********************************************************/
1493 /* host time/date access */
1494 void qemu_get_timedate(struct tm *tm, int offset)
1495 {
1496 time_t ti;
1497 struct tm *ret;
1498
1499 time(&ti);
1500 ti += offset;
1501 if (rtc_date_offset == -1) {
1502 if (rtc_utc)
1503 ret = gmtime(&ti);
1504 else
1505 ret = localtime(&ti);
1506 } else {
1507 ti -= rtc_date_offset;
1508 ret = gmtime(&ti);
1509 }
1510
1511 memcpy(tm, ret, sizeof(struct tm));
1512 }
1513
1514 int qemu_timedate_diff(struct tm *tm)
1515 {
1516 time_t seconds;
1517
1518 if (rtc_date_offset == -1)
1519 if (rtc_utc)
1520 seconds = mktimegm(tm);
1521 else
1522 seconds = mktime(tm);
1523 else
1524 seconds = mktimegm(tm) + rtc_date_offset;
1525
1526 return seconds - time(NULL);
1527 }
1528
1529 void rtc_change_mon_event(struct tm *tm)
1530 {
1531 QObject *data;
1532
1533 data = qobject_from_jsonf("{ 'offset': %d }", qemu_timedate_diff(tm));
1534 monitor_protocol_event(QEVENT_RTC_CHANGE, data);
1535 qobject_decref(data);
1536 }
1537
1538 static void configure_rtc_date_offset(const char *startdate, int legacy)
1539 {
1540 time_t rtc_start_date;
1541 struct tm tm;
1542
1543 if (!strcmp(startdate, "now") && legacy) {
1544 rtc_date_offset = -1;
1545 } else {
1546 if (sscanf(startdate, "%d-%d-%dT%d:%d:%d",
1547 &tm.tm_year,
1548 &tm.tm_mon,
1549 &tm.tm_mday,
1550 &tm.tm_hour,
1551 &tm.tm_min,
1552 &tm.tm_sec) == 6) {
1553 /* OK */
1554 } else if (sscanf(startdate, "%d-%d-%d",
1555 &tm.tm_year,
1556 &tm.tm_mon,
1557 &tm.tm_mday) == 3) {
1558 tm.tm_hour = 0;
1559 tm.tm_min = 0;
1560 tm.tm_sec = 0;
1561 } else {
1562 goto date_fail;
1563 }
1564 tm.tm_year -= 1900;
1565 tm.tm_mon--;
1566 rtc_start_date = mktimegm(&tm);
1567 if (rtc_start_date == -1) {
1568 date_fail:
1569 fprintf(stderr, "Invalid date format. Valid formats are:\n"
1570 "'2006-06-17T16:01:21' or '2006-06-17'\n");
1571 exit(1);
1572 }
1573 rtc_date_offset = time(NULL) - rtc_start_date;
1574 }
1575 }
1576
1577 static void configure_rtc(QemuOpts *opts)
1578 {
1579 const char *value;
1580
1581 value = qemu_opt_get(opts, "base");
1582 if (value) {
1583 if (!strcmp(value, "utc")) {
1584 rtc_utc = 1;
1585 } else if (!strcmp(value, "localtime")) {
1586 rtc_utc = 0;
1587 } else {
1588 configure_rtc_date_offset(value, 0);
1589 }
1590 }
1591 value = qemu_opt_get(opts, "clock");
1592 if (value) {
1593 if (!strcmp(value, "host")) {
1594 rtc_clock = host_clock;
1595 } else if (!strcmp(value, "vm")) {
1596 rtc_clock = vm_clock;
1597 } else {
1598 fprintf(stderr, "qemu: invalid option value '%s'\n", value);
1599 exit(1);
1600 }
1601 }
1602 #ifdef CONFIG_TARGET_I386
1603 value = qemu_opt_get(opts, "driftfix");
1604 if (value) {
1605 if (!strcmp(buf, "slew")) {
1606 rtc_td_hack = 1;
1607 } else if (!strcmp(buf, "none")) {
1608 rtc_td_hack = 0;
1609 } else {
1610 fprintf(stderr, "qemu: invalid option value '%s'\n", value);
1611 exit(1);
1612 }
1613 }
1614 #endif
1615 }
1616
1617 #ifdef _WIN32
1618 static void socket_cleanup(void)
1619 {
1620 WSACleanup();
1621 }
1622
1623 static int socket_init(void)
1624 {
1625 WSADATA Data;
1626 int ret, err;
1627
1628 ret = WSAStartup(MAKEWORD(2,2), &Data);
1629 if (ret != 0) {
1630 err = WSAGetLastError();
1631 fprintf(stderr, "WSAStartup: %d\n", err);
1632 return -1;
1633 }
1634 atexit(socket_cleanup);
1635 return 0;
1636 }
1637 #endif
1638
1639 /***********************************************************/
1640 /* Bluetooth support */
1641 static int nb_hcis;
1642 static int cur_hci;
1643 static struct HCIInfo *hci_table[MAX_NICS];
1644
1645 static struct bt_vlan_s {
1646 struct bt_scatternet_s net;
1647 int id;
1648 struct bt_vlan_s *next;
1649 } *first_bt_vlan;
1650
1651 /* find or alloc a new bluetooth "VLAN" */
1652 static struct bt_scatternet_s *qemu_find_bt_vlan(int id)
1653 {
1654 struct bt_vlan_s **pvlan, *vlan;
1655 for (vlan = first_bt_vlan; vlan != NULL; vlan = vlan->next) {
1656 if (vlan->id == id)
1657 return &vlan->net;
1658 }
1659 vlan = qemu_mallocz(sizeof(struct bt_vlan_s));
1660 vlan->id = id;
1661 pvlan = &first_bt_vlan;
1662 while (*pvlan != NULL)
1663 pvlan = &(*pvlan)->next;
1664 *pvlan = vlan;
1665 return &vlan->net;
1666 }
1667
1668 static void null_hci_send(struct HCIInfo *hci, const uint8_t *data, int len)
1669 {
1670 }
1671
1672 static int null_hci_addr_set(struct HCIInfo *hci, const uint8_t *bd_addr)
1673 {
1674 return -ENOTSUP;
1675 }
1676
1677 static struct HCIInfo null_hci = {
1678 .cmd_send = null_hci_send,
1679 .sco_send = null_hci_send,
1680 .acl_send = null_hci_send,
1681 .bdaddr_set = null_hci_addr_set,
1682 };
1683
1684 struct HCIInfo *qemu_next_hci(void)
1685 {
1686 if (cur_hci == nb_hcis)
1687 return &null_hci;
1688
1689 return hci_table[cur_hci++];
1690 }
1691
1692 static struct HCIInfo *hci_init(const char *str)
1693 {
1694 char *endp;
1695 struct bt_scatternet_s *vlan = 0;
1696
1697 if (!strcmp(str, "null"))
1698 /* null */
1699 return &null_hci;
1700 else if (!strncmp(str, "host", 4) && (str[4] == '\0' || str[4] == ':'))
1701 /* host[:hciN] */
1702 return bt_host_hci(str[4] ? str + 5 : "hci0");
1703 else if (!strncmp(str, "hci", 3)) {
1704 /* hci[,vlan=n] */
1705 if (str[3]) {
1706 if (!strncmp(str + 3, ",vlan=", 6)) {
1707 vlan = qemu_find_bt_vlan(strtol(str + 9, &endp, 0));
1708 if (*endp)
1709 vlan = 0;
1710 }
1711 } else
1712 vlan = qemu_find_bt_vlan(0);
1713 if (vlan)
1714 return bt_new_hci(vlan);
1715 }
1716
1717 fprintf(stderr, "qemu: Unknown bluetooth HCI `%s'.\n", str);
1718
1719 return 0;
1720 }
1721
1722 static int bt_hci_parse(const char *str)
1723 {
1724 struct HCIInfo *hci;
1725 bdaddr_t bdaddr;
1726
1727 if (nb_hcis >= MAX_NICS) {
1728 fprintf(stderr, "qemu: Too many bluetooth HCIs (max %i).\n", MAX_NICS);
1729 return -1;
1730 }
1731
1732 hci = hci_init(str);
1733 if (!hci)
1734 return -1;
1735
1736 bdaddr.b[0] = 0x52;
1737 bdaddr.b[1] = 0x54;
1738 bdaddr.b[2] = 0x00;
1739 bdaddr.b[3] = 0x12;
1740 bdaddr.b[4] = 0x34;
1741 bdaddr.b[5] = 0x56 + nb_hcis;
1742 hci->bdaddr_set(hci, bdaddr.b);
1743
1744 hci_table[nb_hcis++] = hci;
1745
1746 return 0;
1747 }
1748
1749 static void bt_vhci_add(int vlan_id)
1750 {
1751 struct bt_scatternet_s *vlan = qemu_find_bt_vlan(vlan_id);
1752
1753 if (!vlan->slave)
1754 fprintf(stderr, "qemu: warning: adding a VHCI to "
1755 "an empty scatternet %i\n", vlan_id);
1756
1757 bt_vhci_init(bt_new_hci(vlan));
1758 }
1759
1760 static struct bt_device_s *bt_device_add(const char *opt)
1761 {
1762 struct bt_scatternet_s *vlan;
1763 int vlan_id = 0;
1764 char *endp = strstr(opt, ",vlan=");
1765 int len = (endp ? endp - opt : strlen(opt)) + 1;
1766 char devname[10];
1767
1768 pstrcpy(devname, MIN(sizeof(devname), len), opt);
1769
1770 if (endp) {
1771 vlan_id = strtol(endp + 6, &endp, 0);
1772 if (*endp) {
1773 fprintf(stderr, "qemu: unrecognised bluetooth vlan Id\n");
1774 return 0;
1775 }
1776 }
1777
1778 vlan = qemu_find_bt_vlan(vlan_id);
1779
1780 if (!vlan->slave)
1781 fprintf(stderr, "qemu: warning: adding a slave device to "
1782 "an empty scatternet %i\n", vlan_id);
1783
1784 if (!strcmp(devname, "keyboard"))
1785 return bt_keyboard_init(vlan);
1786
1787 fprintf(stderr, "qemu: unsupported bluetooth device `%s'\n", devname);
1788 return 0;
1789 }
1790
1791 static int bt_parse(const char *opt)
1792 {
1793 const char *endp, *p;
1794 int vlan;
1795
1796 if (strstart(opt, "hci", &endp)) {
1797 if (!*endp || *endp == ',') {
1798 if (*endp)
1799 if (!strstart(endp, ",vlan=", 0))
1800 opt = endp + 1;
1801
1802 return bt_hci_parse(opt);
1803 }
1804 } else if (strstart(opt, "vhci", &endp)) {
1805 if (!*endp || *endp == ',') {
1806 if (*endp) {
1807 if (strstart(endp, ",vlan=", &p)) {
1808 vlan = strtol(p, (char **) &endp, 0);
1809 if (*endp) {
1810 fprintf(stderr, "qemu: bad scatternet '%s'\n", p);
1811 return 1;
1812 }
1813 } else {
1814 fprintf(stderr, "qemu: bad parameter '%s'\n", endp + 1);
1815 return 1;
1816 }
1817 } else
1818 vlan = 0;
1819
1820 bt_vhci_add(vlan);
1821 return 0;
1822 }
1823 } else if (strstart(opt, "device:", &endp))
1824 return !bt_device_add(endp);
1825
1826 fprintf(stderr, "qemu: bad bluetooth parameter '%s'\n", opt);
1827 return 1;
1828 }
1829
1830 /***********************************************************/
1831 /* QEMU Block devices */
1832
1833 #define HD_ALIAS "index=%d,media=disk"
1834 #define CDROM_ALIAS "index=2,media=cdrom"
1835 #define FD_ALIAS "index=%d,if=floppy"
1836 #define PFLASH_ALIAS "if=pflash"
1837 #define MTD_ALIAS "if=mtd"
1838 #define SD_ALIAS "index=0,if=sd"
1839
1840 QemuOpts *drive_add(const char *file, const char *fmt, ...)
1841 {
1842 va_list ap;
1843 char optstr[1024];
1844 QemuOpts *opts;
1845
1846 va_start(ap, fmt);
1847 vsnprintf(optstr, sizeof(optstr), fmt, ap);
1848 va_end(ap);
1849
1850 opts = qemu_opts_parse(&qemu_drive_opts, optstr, 0);
1851 if (!opts) {
1852 fprintf(stderr, "%s: huh? duplicate? (%s)\n",
1853 __FUNCTION__, optstr);
1854 return NULL;
1855 }
1856 if (file)
1857 qemu_opt_set(opts, "file", file);
1858 return opts;
1859 }
1860
1861 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
1862 {
1863 DriveInfo *dinfo;
1864
1865 /* seek interface, bus and unit */
1866
1867 QTAILQ_FOREACH(dinfo, &drives, next) {
1868 if (dinfo->type == type &&
1869 dinfo->bus == bus &&
1870 dinfo->unit == unit)
1871 return dinfo;
1872 }
1873
1874 return NULL;
1875 }
1876
1877 DriveInfo *drive_get_by_id(const char *id)
1878 {
1879 DriveInfo *dinfo;
1880
1881 QTAILQ_FOREACH(dinfo, &drives, next) {
1882 if (strcmp(id, dinfo->id))
1883 continue;
1884 return dinfo;
1885 }
1886 return NULL;
1887 }
1888
1889 int drive_get_max_bus(BlockInterfaceType type)
1890 {
1891 int max_bus;
1892 DriveInfo *dinfo;
1893
1894 max_bus = -1;
1895 QTAILQ_FOREACH(dinfo, &drives, next) {
1896 if(dinfo->type == type &&
1897 dinfo->bus > max_bus)
1898 max_bus = dinfo->bus;
1899 }
1900 return max_bus;
1901 }
1902
1903 const char *drive_get_serial(BlockDriverState *bdrv)
1904 {
1905 DriveInfo *dinfo;
1906
1907 QTAILQ_FOREACH(dinfo, &drives, next) {
1908 if (dinfo->bdrv == bdrv)
1909 return dinfo->serial;
1910 }
1911
1912 return "\0";
1913 }
1914
1915 BlockInterfaceErrorAction drive_get_on_error(
1916 BlockDriverState *bdrv, int is_read)
1917 {
1918 DriveInfo *dinfo;
1919
1920 QTAILQ_FOREACH(dinfo, &drives, next) {
1921 if (dinfo->bdrv == bdrv)
1922 return is_read ? dinfo->on_read_error : dinfo->on_write_error;
1923 }
1924
1925 return is_read ? BLOCK_ERR_REPORT : BLOCK_ERR_STOP_ENOSPC;
1926 }
1927
1928 static void bdrv_format_print(void *opaque, const char *name)
1929 {
1930 fprintf(stderr, " %s", name);
1931 }
1932
1933 void drive_uninit(DriveInfo *dinfo)
1934 {
1935 qemu_opts_del(dinfo->opts);
1936 bdrv_delete(dinfo->bdrv);
1937 QTAILQ_REMOVE(&drives, dinfo, next);
1938 qemu_free(dinfo);
1939 }
1940
1941 static int parse_block_error_action(const char *buf, int is_read)
1942 {
1943 if (!strcmp(buf, "ignore")) {
1944 return BLOCK_ERR_IGNORE;
1945 } else if (!is_read && !strcmp(buf, "enospc")) {
1946 return BLOCK_ERR_STOP_ENOSPC;
1947 } else if (!strcmp(buf, "stop")) {
1948 return BLOCK_ERR_STOP_ANY;
1949 } else if (!strcmp(buf, "report")) {
1950 return BLOCK_ERR_REPORT;
1951 } else {
1952 fprintf(stderr, "qemu: '%s' invalid %s error action\n",
1953 buf, is_read ? "read" : "write");
1954 return -1;
1955 }
1956 }
1957
1958 DriveInfo *drive_init(QemuOpts *opts, void *opaque,
1959 int *fatal_error)
1960 {
1961 const char *buf;
1962 const char *file = NULL;
1963 char devname[128];
1964 const char *serial;
1965 const char *mediastr = "";
1966 BlockInterfaceType type;
1967 enum { MEDIA_DISK, MEDIA_CDROM } media;
1968 int bus_id, unit_id;
1969 int cyls, heads, secs, translation;
1970 BlockDriver *drv = NULL;
1971 QEMUMachine *machine = opaque;
1972 int max_devs;
1973 int index;
1974 int cache;
1975 int aio = 0;
1976 int ro = 0;
1977 int bdrv_flags;
1978 int on_read_error, on_write_error;
1979 const char *devaddr;
1980 DriveInfo *dinfo;
1981 int snapshot = 0;
1982
1983 *fatal_error = 1;
1984
1985 translation = BIOS_ATA_TRANSLATION_AUTO;
1986 cache = 1;
1987
1988 if (machine && machine->use_scsi) {
1989 type = IF_SCSI;
1990 max_devs = MAX_SCSI_DEVS;
1991 pstrcpy(devname, sizeof(devname), "scsi");
1992 } else {
1993 type = IF_IDE;
1994 max_devs = MAX_IDE_DEVS;
1995 pstrcpy(devname, sizeof(devname), "ide");
1996 }
1997 media = MEDIA_DISK;
1998
1999 /* extract parameters */
2000 bus_id = qemu_opt_get_number(opts, "bus", 0);
2001 unit_id = qemu_opt_get_number(opts, "unit", -1);
2002 index = qemu_opt_get_number(opts, "index", -1);
2003
2004 cyls = qemu_opt_get_number(opts, "cyls", 0);
2005 heads = qemu_opt_get_number(opts, "heads", 0);
2006 secs = qemu_opt_get_number(opts, "secs", 0);
2007
2008 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
2009 ro = qemu_opt_get_bool(opts, "readonly", 0);
2010
2011 file = qemu_opt_get(opts, "file");
2012 serial = qemu_opt_get(opts, "serial");
2013
2014 if ((buf = qemu_opt_get(opts, "if")) != NULL) {
2015 pstrcpy(devname, sizeof(devname), buf);
2016 if (!strcmp(buf, "ide")) {
2017 type = IF_IDE;
2018 max_devs = MAX_IDE_DEVS;
2019 } else if (!strcmp(buf, "scsi")) {
2020 type = IF_SCSI;
2021 max_devs = MAX_SCSI_DEVS;
2022 } else if (!strcmp(buf, "floppy")) {
2023 type = IF_FLOPPY;
2024 max_devs = 0;
2025 } else if (!strcmp(buf, "pflash")) {
2026 type = IF_PFLASH;
2027 max_devs = 0;
2028 } else if (!strcmp(buf, "mtd")) {
2029 type = IF_MTD;
2030 max_devs = 0;
2031 } else if (!strcmp(buf, "sd")) {
2032 type = IF_SD;
2033 max_devs = 0;
2034 } else if (!strcmp(buf, "virtio")) {
2035 type = IF_VIRTIO;
2036 max_devs = 0;
2037 } else if (!strcmp(buf, "xen")) {
2038 type = IF_XEN;
2039 max_devs = 0;
2040 } else if (!strcmp(buf, "none")) {
2041 type = IF_NONE;
2042 max_devs = 0;
2043 } else {
2044 fprintf(stderr, "qemu: unsupported bus type '%s'\n", buf);
2045 return NULL;
2046 }
2047 }
2048
2049 if (cyls || heads || secs) {
2050 if (cyls < 1 || (type == IF_IDE && cyls > 16383)) {
2051 fprintf(stderr, "qemu: '%s' invalid physical cyls number\n", buf);
2052 return NULL;
2053 }
2054 if (heads < 1 || (type == IF_IDE && heads > 16)) {
2055 fprintf(stderr, "qemu: '%s' invalid physical heads number\n", buf);
2056 return NULL;
2057 }
2058 if (secs < 1 || (type == IF_IDE && secs > 63)) {
2059 fprintf(stderr, "qemu: '%s' invalid physical secs number\n", buf);
2060 return NULL;
2061 }
2062 }
2063
2064 if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
2065 if (!cyls) {
2066 fprintf(stderr,
2067 "qemu: '%s' trans must be used with cyls,heads and secs\n",
2068 buf);
2069 return NULL;
2070 }
2071 if (!strcmp(buf, "none"))
2072 translation = BIOS_ATA_TRANSLATION_NONE;
2073 else if (!strcmp(buf, "lba"))
2074 translation = BIOS_ATA_TRANSLATION_LBA;
2075 else if (!strcmp(buf, "auto"))
2076 translation = BIOS_ATA_TRANSLATION_AUTO;
2077 else {
2078 fprintf(stderr, "qemu: '%s' invalid translation type\n", buf);
2079 return NULL;
2080 }
2081 }
2082
2083 if ((buf = qemu_opt_get(opts, "media")) != NULL) {
2084 if (!strcmp(buf, "disk")) {
2085 media = MEDIA_DISK;
2086 } else if (!strcmp(buf, "cdrom")) {
2087 if (cyls || secs || heads) {
2088 fprintf(stderr,
2089 "qemu: '%s' invalid physical CHS format\n", buf);
2090 return NULL;
2091 }
2092 media = MEDIA_CDROM;
2093 } else {
2094 fprintf(stderr, "qemu: '%s' invalid media\n", buf);
2095 return NULL;
2096 }
2097 }
2098
2099 if ((buf = qemu_opt_get(opts, "cache")) != NULL) {
2100 if (!strcmp(buf, "off") || !strcmp(buf, "none"))
2101 cache = 0;
2102 else if (!strcmp(buf, "writethrough"))
2103 cache = 1;
2104 else if (!strcmp(buf, "writeback"))
2105 cache = 2;
2106 else {
2107 fprintf(stderr, "qemu: invalid cache option\n");
2108 return NULL;
2109 }
2110 }
2111
2112 #ifdef CONFIG_LINUX_AIO
2113 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
2114 if (!strcmp(buf, "threads"))
2115 aio = 0;
2116 else if (!strcmp(buf, "native"))
2117 aio = 1;
2118 else {
2119 fprintf(stderr, "qemu: invalid aio option\n");
2120 return NULL;
2121 }
2122 }
2123 #endif
2124
2125 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
2126 if (strcmp(buf, "?") == 0) {
2127 fprintf(stderr, "qemu: Supported formats:");
2128 bdrv_iterate_format(bdrv_format_print, NULL);
2129 fprintf(stderr, "\n");
2130 return NULL;
2131 }
2132 drv = bdrv_find_whitelisted_format(buf);
2133 if (!drv) {
2134 fprintf(stderr, "qemu: '%s' invalid format\n", buf);
2135 return NULL;
2136 }
2137 }
2138
2139 on_write_error = BLOCK_ERR_STOP_ENOSPC;
2140 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
2141 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO) {
2142 fprintf(stderr, "werror is no supported by this format\n");
2143 return NULL;
2144 }
2145
2146 on_write_error = parse_block_error_action(buf, 0);
2147 if (on_write_error < 0) {
2148 return NULL;
2149 }
2150 }
2151
2152 on_read_error = BLOCK_ERR_REPORT;
2153 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
2154 if (type != IF_IDE && type != IF_VIRTIO) {
2155 fprintf(stderr, "rerror is no supported by this format\n");
2156 return NULL;
2157 }
2158
2159 on_read_error = parse_block_error_action(buf, 1);
2160 if (on_read_error < 0) {
2161 return NULL;
2162 }
2163 }
2164
2165 if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
2166 if (type != IF_VIRTIO) {
2167 fprintf(stderr, "addr is not supported\n");
2168 return NULL;
2169 }
2170 }
2171
2172 /* compute bus and unit according index */
2173
2174 if (index != -1) {
2175 if (bus_id != 0 || unit_id != -1) {
2176 fprintf(stderr,
2177 "qemu: index cannot be used with bus and unit\n");
2178 return NULL;
2179 }
2180 if (max_devs == 0)
2181 {
2182 unit_id = index;
2183 bus_id = 0;
2184 } else {
2185 unit_id = index % max_devs;
2186 bus_id = index / max_devs;
2187 }
2188 }
2189
2190 /* if user doesn't specify a unit_id,
2191 * try to find the first free
2192 */
2193
2194 if (unit_id == -1) {
2195 unit_id = 0;
2196 while (drive_get(type, bus_id, unit_id) != NULL) {
2197 unit_id++;
2198 if (max_devs && unit_id >= max_devs) {
2199 unit_id -= max_devs;
2200 bus_id++;
2201 }
2202 }
2203 }
2204
2205 /* check unit id */
2206
2207 if (max_devs && unit_id >= max_devs) {
2208 fprintf(stderr, "qemu: unit %d too big (max is %d)\n",
2209 unit_id, max_devs - 1);
2210 return NULL;
2211 }
2212
2213 /*
2214 * ignore multiple definitions
2215 */
2216
2217 if (drive_get(type, bus_id, unit_id) != NULL) {
2218 *fatal_error = 0;
2219 return NULL;
2220 }
2221
2222 /* init */
2223
2224 dinfo = qemu_mallocz(sizeof(*dinfo));
2225 if ((buf = qemu_opts_id(opts)) != NULL) {
2226 dinfo->id = qemu_strdup(buf);
2227 } else {
2228 /* no id supplied -> create one */
2229 dinfo->id = qemu_mallocz(32);
2230 if (type == IF_IDE || type == IF_SCSI)
2231 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
2232 if (max_devs)
2233 snprintf(dinfo->id, 32, "%s%i%s%i",
2234 devname, bus_id, mediastr, unit_id);
2235 else
2236 snprintf(dinfo->id, 32, "%s%s%i",
2237 devname, mediastr, unit_id);
2238 }
2239 dinfo->bdrv = bdrv_new(dinfo->id);
2240 dinfo->devaddr = devaddr;
2241 dinfo->type = type;
2242 dinfo->bus = bus_id;
2243 dinfo->unit = unit_id;
2244 dinfo->on_read_error = on_read_error;
2245 dinfo->on_write_error = on_write_error;
2246 dinfo->opts = opts;
2247 if (serial)
2248 strncpy(dinfo->serial, serial, sizeof(serial));
2249 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
2250
2251 switch(type) {
2252 case IF_IDE:
2253 case IF_SCSI:
2254 case IF_XEN:
2255 case IF_NONE:
2256 switch(media) {
2257 case MEDIA_DISK:
2258 if (cyls != 0) {
2259 bdrv_set_geometry_hint(dinfo->bdrv, cyls, heads, secs);
2260 bdrv_set_translation_hint(dinfo->bdrv, translation);
2261 }
2262 break;
2263 case MEDIA_CDROM:
2264 bdrv_set_type_hint(dinfo->bdrv, BDRV_TYPE_CDROM);
2265 break;
2266 }
2267 break;
2268 case IF_SD:
2269 /* FIXME: This isn't really a floppy, but it's a reasonable
2270 approximation. */
2271 case IF_FLOPPY:
2272 bdrv_set_type_hint(dinfo->bdrv, BDRV_TYPE_FLOPPY);
2273 break;
2274 case IF_PFLASH:
2275 case IF_MTD:
2276 break;
2277 case IF_VIRTIO:
2278 /* add virtio block device */
2279 opts = qemu_opts_create(&qemu_device_opts, NULL, 0);
2280 qemu_opt_set(opts, "driver", "virtio-blk-pci");
2281 qemu_opt_set(opts, "drive", dinfo->id);
2282 if (devaddr)
2283 qemu_opt_set(opts, "addr", devaddr);
2284 break;
2285 case IF_COUNT:
2286 abort();
2287 }
2288 if (!file) {
2289 *fatal_error = 0;
2290 return NULL;
2291 }
2292 bdrv_flags = 0;
2293 if (snapshot) {
2294 bdrv_flags |= BDRV_O_SNAPSHOT;
2295 cache = 2; /* always use write-back with snapshot */
2296 }
2297 if (cache == 0) /* no caching */
2298 bdrv_flags |= BDRV_O_NOCACHE;
2299 else if (cache == 2) /* write-back */
2300 bdrv_flags |= BDRV_O_CACHE_WB;
2301
2302 if (aio == 1) {
2303 bdrv_flags |= BDRV_O_NATIVE_AIO;
2304 } else {
2305 bdrv_flags &= ~BDRV_O_NATIVE_AIO;
2306 }
2307
2308 if (ro == 1) {
2309 if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY) {
2310 fprintf(stderr, "qemu: readonly flag not supported for drive with this interface\n");
2311 return NULL;
2312 }
2313 }
2314 /*
2315 * cdrom is read-only. Set it now, after above interface checking
2316 * since readonly attribute not explicitly required, so no error.
2317 */
2318 if (media == MEDIA_CDROM) {
2319 ro = 1;
2320 }
2321 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
2322
2323 if (bdrv_open2(dinfo->bdrv, file, bdrv_flags, drv) < 0) {
2324 fprintf(stderr, "qemu: could not open disk image %s: %s\n",
2325 file, strerror(errno));
2326 return NULL;
2327 }
2328
2329 if (bdrv_key_required(dinfo->bdrv))
2330 autostart = 0;
2331 *fatal_error = 0;
2332 return dinfo;
2333 }
2334
2335 static int drive_init_func(QemuOpts *opts, void *opaque)
2336 {
2337 QEMUMachine *machine = opaque;
2338 int fatal_error = 0;
2339
2340 if (drive_init(opts, machine, &fatal_error) == NULL) {
2341 if (fatal_error)
2342 return 1;
2343 }
2344 return 0;
2345 }
2346
2347 static int drive_enable_snapshot(QemuOpts *opts, void *opaque)
2348 {
2349 if (NULL == qemu_opt_get(opts, "snapshot")) {
2350 qemu_opt_set(opts, "snapshot", "on");
2351 }
2352 return 0;
2353 }
2354
2355 void qemu_register_boot_set(QEMUBootSetHandler *func, void *opaque)
2356 {
2357 boot_set_handler = func;
2358 boot_set_opaque = opaque;
2359 }
2360
2361 int qemu_boot_set(const char *boot_devices)
2362 {
2363 if (!boot_set_handler) {
2364 return -EINVAL;
2365 }
2366 return boot_set_handler(boot_set_opaque, boot_devices);
2367 }
2368
2369 static int parse_bootdevices(char *devices)
2370 {
2371 /* We just do some generic consistency checks */
2372 const char *p;
2373 int bitmap = 0;
2374
2375 for (p = devices; *p != '\0'; p++) {
2376 /* Allowed boot devices are:
2377 * a-b: floppy disk drives
2378 * c-f: IDE disk drives
2379 * g-m: machine implementation dependant drives
2380 * n-p: network devices
2381 * It's up to each machine implementation to check if the given boot
2382 * devices match the actual hardware implementation and firmware
2383 * features.
2384 */
2385 if (*p < 'a' || *p > 'p') {
2386 fprintf(stderr, "Invalid boot device '%c'\n", *p);
2387 exit(1);
2388 }
2389 if (bitmap & (1 << (*p - 'a'))) {
2390 fprintf(stderr, "Boot device '%c' was given twice\n", *p);
2391 exit(1);
2392 }
2393 bitmap |= 1 << (*p - 'a');
2394 }
2395 return bitmap;
2396 }
2397
2398 static void restore_boot_devices(void *opaque)
2399 {
2400 char *standard_boot_devices = opaque;
2401
2402 qemu_boot_set(standard_boot_devices);
2403
2404 qemu_unregister_reset(restore_boot_devices, standard_boot_devices);
2405 qemu_free(standard_boot_devices);
2406 }
2407
2408 static void numa_add(const char *optarg)
2409 {
2410 char option[128];
2411 char *endptr;
2412 unsigned long long value, endvalue;
2413 int nodenr;
2414
2415 optarg = get_opt_name(option, 128, optarg, ',') + 1;
2416 if (!strcmp(option, "node")) {
2417 if (get_param_value(option, 128, "nodeid", optarg) == 0) {
2418 nodenr = nb_numa_nodes;
2419 } else {
2420 nodenr = strtoull(option, NULL, 10);
2421 }
2422
2423 if (get_param_value(option, 128, "mem", optarg) == 0) {
2424 node_mem[nodenr] = 0;
2425 } else {
2426 value = strtoull(option, &endptr, 0);
2427 switch (*endptr) {
2428 case 0: case 'M': case 'm':
2429 value <<= 20;
2430 break;
2431 case 'G': case 'g':
2432 value <<= 30;
2433 break;
2434 }
2435 node_mem[nodenr] = value;
2436 }
2437 if (get_param_value(option, 128, "cpus", optarg) == 0) {
2438 node_cpumask[nodenr] = 0;
2439 } else {
2440 value = strtoull(option, &endptr, 10);
2441 if (value >= 64) {
2442 value = 63;
2443 fprintf(stderr, "only 64 CPUs in NUMA mode supported.\n");
2444 } else {
2445 if (*endptr == '-') {
2446 endvalue = strtoull(endptr+1, &endptr, 10);
2447 if (endvalue >= 63) {
2448 endvalue = 62;
2449 fprintf(stderr,
2450 "only 63 CPUs in NUMA mode supported.\n");
2451 }
2452 value = (2ULL << endvalue) - (1ULL << value);
2453 } else {
2454 value = 1ULL << value;
2455 }
2456 }
2457 node_cpumask[nodenr] = value;
2458 }
2459 nb_numa_nodes++;
2460 }
2461 return;
2462 }
2463
2464 static void smp_parse(const char *optarg)
2465 {
2466 int smp, sockets = 0, threads = 0, cores = 0;
2467 char *endptr;
2468 char option[128];
2469
2470 smp = strtoul(optarg, &endptr, 10);
2471 if (endptr != optarg) {
2472 if (*endptr == ',') {
2473 endptr++;
2474 }
2475 }
2476 if (get_param_value(option, 128, "sockets", endptr) != 0)
2477 sockets = strtoull(option, NULL, 10);
2478 if (get_param_value(option, 128, "cores", endptr) != 0)
2479 cores = strtoull(option, NULL, 10);
2480 if (get_param_value(option, 128, "threads", endptr) != 0)
2481 threads = strtoull(option, NULL, 10);
2482 if (get_param_value(option, 128, "maxcpus", endptr) != 0)
2483 max_cpus = strtoull(option, NULL, 10);
2484
2485 /* compute missing values, prefer sockets over cores over threads */
2486 if (smp == 0 || sockets == 0) {
2487 sockets = sockets > 0 ? sockets : 1;
2488 cores = cores > 0 ? cores : 1;
2489 threads = threads > 0 ? threads : 1;
2490 if (smp == 0) {
2491 smp = cores * threads * sockets;
2492 }
2493 } else {
2494 if (cores == 0) {
2495 threads = threads > 0 ? threads : 1;
2496 cores = smp / (sockets * threads);
2497 } else {
2498 if (sockets) {
2499 threads = smp / (cores * sockets);
2500 }
2501 }
2502 }
2503 smp_cpus = smp;
2504 smp_cores = cores > 0 ? cores : 1;
2505 smp_threads = threads > 0 ? threads : 1;
2506 if (max_cpus == 0)
2507 max_cpus = smp_cpus;
2508 }
2509
2510 /***********************************************************/
2511 /* USB devices */
2512
2513 static int usb_device_add(const char *devname, int is_hotplug)
2514 {
2515 const char *p;
2516 USBDevice *dev = NULL;
2517
2518 if (!usb_enabled)
2519 return -1;
2520
2521 /* drivers with .usbdevice_name entry in USBDeviceInfo */
2522 dev = usbdevice_create(devname);
2523 if (dev)
2524 goto done;
2525
2526 /* the other ones */
2527 if (strstart(devname, "host:", &p)) {
2528 dev = usb_host_device_open(p);
2529 } else if (!strcmp(devname, "bt") || strstart(devname, "bt:", &p)) {
2530 dev = usb_bt_init(devname[2] ? hci_init(p) :
2531 bt_new_hci(qemu_find_bt_vlan(0)));
2532 } else {
2533 return -1;
2534 }
2535 if (!dev)
2536 return -1;
2537
2538 done:
2539 return 0;
2540 }
2541
2542 static int usb_device_del(const char *devname)
2543 {
2544 int bus_num, addr;
2545 const char *p;
2546
2547 if (strstart(devname, "host:", &p))
2548 return usb_host_device_close(p);
2549
2550 if (!usb_enabled)
2551 return -1;
2552
2553 p = strchr(devname, '.');
2554 if (!p)
2555 return -1;
2556 bus_num = strtoul(devname, NULL, 0);
2557 addr = strtoul(p + 1, NULL, 0);
2558
2559 return usb_device_delete_addr(bus_num, addr);
2560 }
2561
2562 static int usb_parse(const char *cmdline)
2563 {
2564 int r;
2565 r = usb_device_add(cmdline, 0);
2566 if (r < 0) {
2567 fprintf(stderr, "qemu: could not add USB device '%s'\n", cmdline);
2568 }
2569 return r;
2570 }
2571
2572 void do_usb_add(Monitor *mon, const QDict *qdict)
2573 {
2574 const char *devname = qdict_get_str(qdict, "devname");
2575 if (usb_device_add(devname, 1) < 0) {
2576 error_report("could not add USB device '%s'", devname);
2577 }
2578 }
2579
2580 void do_usb_del(Monitor *mon, const QDict *qdict)
2581 {
2582 const char *devname = qdict_get_str(qdict, "devname");
2583 if (usb_device_del(devname) < 0) {
2584 error_report("could not delete USB device '%s'", devname);
2585 }
2586 }
2587
2588 /***********************************************************/
2589 /* PCMCIA/Cardbus */
2590
2591 static struct pcmcia_socket_entry_s {
2592 PCMCIASocket *socket;
2593 struct pcmcia_socket_entry_s *next;
2594 } *pcmcia_sockets = 0;
2595
2596 void pcmcia_socket_register(PCMCIASocket *socket)
2597 {
2598 struct pcmcia_socket_entry_s *entry;
2599
2600 entry = qemu_malloc(sizeof(struct pcmcia_socket_entry_s));
2601 entry->socket = socket;
2602 entry->next = pcmcia_sockets;
2603 pcmcia_sockets = entry;
2604 }
2605
2606 void pcmcia_socket_unregister(PCMCIASocket *socket)
2607 {
2608 struct pcmcia_socket_entry_s *entry, **ptr;
2609
2610 ptr = &pcmcia_sockets;
2611 for (entry = *ptr; entry; ptr = &entry->next, entry = *ptr)
2612 if (entry->socket == socket) {
2613 *ptr = entry->next;
2614 qemu_free(entry);
2615 }
2616 }
2617
2618 void pcmcia_info(Monitor *mon)
2619 {
2620 struct pcmcia_socket_entry_s *iter;
2621
2622 if (!pcmcia_sockets)
2623 monitor_printf(mon, "No PCMCIA sockets\n");
2624
2625 for (iter = pcmcia_sockets; iter; iter = iter->next)
2626 monitor_printf(mon, "%s: %s\n", iter->socket->slot_string,
2627 iter->socket->attached ? iter->socket->card_string :
2628 "Empty");
2629 }
2630
2631 /***********************************************************/
2632 /* I/O handling */
2633
2634 typedef struct IOHandlerRecord {
2635 int fd;
2636 IOCanRWHandler *fd_read_poll;
2637 IOHandler *fd_read;
2638 IOHandler *fd_write;
2639 int deleted;
2640 void *opaque;
2641 /* temporary data */
2642 struct pollfd *ufd;
2643 struct IOHandlerRecord *next;
2644 } IOHandlerRecord;
2645
2646 static IOHandlerRecord *first_io_handler;
2647
2648 /* XXX: fd_read_poll should be suppressed, but an API change is
2649 necessary in the character devices to suppress fd_can_read(). */
2650 int qemu_set_fd_handler2(int fd,
2651 IOCanRWHandler *fd_read_poll,
2652 IOHandler *fd_read,
2653 IOHandler *fd_write,
2654 void *opaque)
2655 {
2656 IOHandlerRecord **pioh, *ioh;
2657
2658 if (!fd_read && !fd_write) {
2659 pioh = &first_io_handler;
2660 for(;;) {
2661 ioh = *pioh;
2662 if (ioh == NULL)
2663 break;
2664 if (ioh->fd == fd) {
2665 ioh->deleted = 1;
2666 break;
2667 }
2668 pioh = &ioh->next;
2669 }
2670 } else {
2671 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
2672 if (ioh->fd == fd)
2673 goto found;
2674 }
2675 ioh = qemu_mallocz(sizeof(IOHandlerRecord));
2676 ioh->next = first_io_handler;
2677 first_io_handler = ioh;
2678 found:
2679 ioh->fd = fd;
2680 ioh->fd_read_poll = fd_read_poll;
2681 ioh->fd_read = fd_read;
2682 ioh->fd_write = fd_write;
2683 ioh->opaque = opaque;
2684 ioh->deleted = 0;
2685 }
2686 return 0;
2687 }
2688
2689 int qemu_set_fd_handler(int fd,
2690 IOHandler *fd_read,
2691 IOHandler *fd_write,
2692 void *opaque)
2693 {
2694 return qemu_set_fd_handler2(fd, NULL, fd_read, fd_write, opaque);
2695 }
2696
2697 #ifdef _WIN32
2698 /***********************************************************/
2699 /* Polling handling */
2700
2701 typedef struct PollingEntry {
2702 PollingFunc *func;
2703 void *opaque;
2704 struct PollingEntry *next;
2705 } PollingEntry;
2706
2707 static PollingEntry *first_polling_entry;
2708
2709 int qemu_add_polling_cb(PollingFunc *func, void *opaque)
2710 {
2711 PollingEntry **ppe, *pe;
2712 pe = qemu_mallocz(sizeof(PollingEntry));
2713 pe->func = func;
2714 pe->opaque = opaque;
2715 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next);
2716 *ppe = pe;
2717 return 0;
2718 }
2719
2720 void qemu_del_polling_cb(PollingFunc *func, void *opaque)
2721 {
2722 PollingEntry **ppe, *pe;
2723 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next) {
2724 pe = *ppe;
2725 if (pe->func == func && pe->opaque == opaque) {
2726 *ppe = pe->next;
2727 qemu_free(pe);
2728 break;
2729 }
2730 }
2731 }
2732
2733 /***********************************************************/
2734 /* Wait objects support */
2735 typedef struct WaitObjects {
2736 int num;
2737 HANDLE events[MAXIMUM_WAIT_OBJECTS + 1];
2738 WaitObjectFunc *func[MAXIMUM_WAIT_OBJECTS + 1];
2739 void *opaque[MAXIMUM_WAIT_OBJECTS + 1];
2740 } WaitObjects;
2741
2742 static WaitObjects wait_objects = {0};
2743
2744 int qemu_add_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
2745 {
2746 WaitObjects *w = &wait_objects;
2747
2748 if (w->num >= MAXIMUM_WAIT_OBJECTS)
2749 return -1;
2750 w->events[w->num] = handle;
2751 w->func[w->num] = func;
2752 w->opaque[w->num] = opaque;
2753 w->num++;
2754 return 0;
2755 }
2756
2757 void qemu_del_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
2758 {
2759 int i, found;
2760 WaitObjects *w = &wait_objects;
2761
2762 found = 0;
2763 for (i = 0; i < w->num; i++) {
2764 if (w->events[i] == handle)
2765 found = 1;
2766 if (found) {
2767 w->events[i] = w->events[i + 1];
2768 w->func[i] = w->func[i + 1];
2769 w->opaque[i] = w->opaque[i + 1];
2770 }
2771 }
2772 if (found)
2773 w->num--;
2774 }
2775 #endif
2776
2777 /***********************************************************/
2778 /* ram save/restore */
2779
2780 #define RAM_SAVE_FLAG_FULL 0x01 /* Obsolete, not used anymore */
2781 #define RAM_SAVE_FLAG_COMPRESS 0x02
2782 #define RAM_SAVE_FLAG_MEM_SIZE 0x04
2783 #define RAM_SAVE_FLAG_PAGE 0x08
2784 #define RAM_SAVE_FLAG_EOS 0x10
2785
2786 static int is_dup_page(uint8_t *page, uint8_t ch)
2787 {
2788 uint32_t val = ch << 24 | ch << 16 | ch << 8 | ch;
2789 uint32_t *array = (uint32_t *)page;
2790 int i;
2791
2792 for (i = 0; i < (TARGET_PAGE_SIZE / 4); i++) {
2793 if (array[i] != val)
2794 return 0;
2795 }
2796
2797 return 1;
2798 }
2799
2800 static int ram_save_block(QEMUFile *f)
2801 {
2802 static ram_addr_t current_addr = 0;
2803 ram_addr_t saved_addr = current_addr;
2804 ram_addr_t addr = 0;
2805 int found = 0;
2806
2807 while (addr < last_ram_offset) {
2808 if (cpu_physical_memory_get_dirty(current_addr, MIGRATION_DIRTY_FLAG)) {
2809 uint8_t *p;
2810
2811 cpu_physical_memory_reset_dirty(current_addr,
2812 current_addr + TARGET_PAGE_SIZE,
2813 MIGRATION_DIRTY_FLAG);
2814
2815 p = qemu_get_ram_ptr(current_addr);
2816
2817 if (is_dup_page(p, *p)) {
2818 qemu_put_be64(f, current_addr | RAM_SAVE_FLAG_COMPRESS);
2819 qemu_put_byte(f, *p);
2820 } else {
2821 qemu_put_be64(f, current_addr | RAM_SAVE_FLAG_PAGE);
2822 qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
2823 }
2824
2825 found = 1;
2826 break;
2827 }
2828 addr += TARGET_PAGE_SIZE;
2829 current_addr = (saved_addr + addr) % last_ram_offset;
2830 }
2831
2832 return found;
2833 }
2834
2835 static uint64_t bytes_transferred;
2836
2837 static ram_addr_t ram_save_remaining(void)
2838 {
2839 ram_addr_t addr;
2840 ram_addr_t count = 0;
2841
2842 for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
2843 if (cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
2844 count++;
2845 }
2846
2847 return count;
2848 }
2849
2850 uint64_t ram_bytes_remaining(void)
2851 {
2852 return ram_save_remaining() * TARGET_PAGE_SIZE;
2853 }
2854
2855 uint64_t ram_bytes_transferred(void)
2856 {
2857 return bytes_transferred;
2858 }
2859
2860 uint64_t ram_bytes_total(void)
2861 {
2862 return last_ram_offset;
2863 }
2864
2865 static int ram_save_live(Monitor *mon, QEMUFile *f, int stage, void *opaque)
2866 {
2867 ram_addr_t addr;
2868 uint64_t bytes_transferred_last;
2869 double bwidth = 0;
2870 uint64_t expected_time = 0;
2871
2872 if (stage < 0) {
2873 cpu_physical_memory_set_dirty_tracking(0);
2874 return 0;
2875 }
2876
2877 if (cpu_physical_sync_dirty_bitmap(0, TARGET_PHYS_ADDR_MAX) != 0) {
2878 qemu_file_set_error(f);
2879 return 0;
2880 }
2881
2882 if (stage == 1) {
2883 bytes_transferred = 0;
2884
2885 /* Make sure all dirty bits are set */
2886 for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
2887 if (!cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
2888 cpu_physical_memory_set_dirty(addr);
2889 }
2890
2891 /* Enable dirty memory tracking */
2892 cpu_physical_memory_set_dirty_tracking(1);
2893
2894 qemu_put_be64(f, last_ram_offset | RAM_SAVE_FLAG_MEM_SIZE);
2895 }
2896
2897 bytes_transferred_last = bytes_transferred;
2898 bwidth = qemu_get_clock_ns(rt_clock);
2899
2900 while (!qemu_file_rate_limit(f)) {
2901 int ret;
2902
2903 ret = ram_save_block(f);
2904 bytes_transferred += ret * TARGET_PAGE_SIZE;
2905 if (ret == 0) /* no more blocks */
2906 break;
2907 }
2908
2909 bwidth = qemu_get_clock_ns(rt_clock) - bwidth;
2910 bwidth = (bytes_transferred - bytes_transferred_last) / bwidth;
2911
2912 /* if we haven't transferred anything this round, force expected_time to a
2913 * a very high value, but without crashing */
2914 if (bwidth == 0)
2915 bwidth = 0.000001;
2916
2917 /* try transferring iterative blocks of memory */
2918 if (stage == 3) {
2919 /* flush all remaining blocks regardless of rate limiting */
2920 while (ram_save_block(f) != 0) {
2921 bytes_transferred += TARGET_PAGE_SIZE;
2922 }
2923 cpu_physical_memory_set_dirty_tracking(0);
2924 }
2925
2926 qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
2927
2928 expected_time = ram_save_remaining() * TARGET_PAGE_SIZE / bwidth;
2929
2930 return (stage == 2) && (expected_time <= migrate_max_downtime());
2931 }
2932
2933 static int ram_load(QEMUFile *f, void *opaque, int version_id)
2934 {
2935 ram_addr_t addr;
2936 int flags;
2937
2938 if (version_id != 3)
2939 return -EINVAL;
2940
2941 do {
2942 addr = qemu_get_be64(f);
2943
2944 flags = addr & ~TARGET_PAGE_MASK;
2945 addr &= TARGET_PAGE_MASK;
2946
2947 if (flags & RAM_SAVE_FLAG_MEM_SIZE) {
2948 if (addr != last_ram_offset)
2949 return -EINVAL;
2950 }
2951
2952 if (flags & RAM_SAVE_FLAG_COMPRESS) {
2953 uint8_t ch = qemu_get_byte(f);
2954 memset(qemu_get_ram_ptr(addr), ch, TARGET_PAGE_SIZE);
2955 #ifndef _WIN32
2956 if (ch == 0 &&
2957 (!kvm_enabled() || kvm_has_sync_mmu())) {
2958 madvise(qemu_get_ram_ptr(addr), TARGET_PAGE_SIZE, MADV_DONTNEED);
2959 }
2960 #endif
2961 } else if (flags & RAM_SAVE_FLAG_PAGE) {
2962 qemu_get_buffer(f, qemu_get_ram_ptr(addr), TARGET_PAGE_SIZE);
2963 }
2964 if (qemu_file_has_error(f)) {
2965 return -EIO;
2966 }
2967 } while (!(flags & RAM_SAVE_FLAG_EOS));
2968
2969 return 0;
2970 }
2971
2972 void qemu_service_io(void)
2973 {
2974 qemu_notify_event();
2975 }
2976
2977 /***********************************************************/
2978 /* machine registration */
2979
2980 static QEMUMachine *first_machine = NULL;
2981 QEMUMachine *current_machine = NULL;
2982
2983 int qemu_register_machine(QEMUMachine *m)
2984 {
2985 QEMUMachine **pm;
2986 pm = &first_machine;
2987 while (*pm != NULL)
2988 pm = &(*pm)->next;
2989 m->next = NULL;
2990 *pm = m;
2991 return 0;
2992 }
2993
2994 static QEMUMachine *find_machine(const char *name)
2995 {
2996 QEMUMachine *m;
2997
2998 for(m = first_machine; m != NULL; m = m->next) {
2999 if (!strcmp(m->name, name))
3000 return m;
3001 if (m->alias && !strcmp(m->alias, name))
3002 return m;
3003 }
3004 return NULL;
3005 }
3006
3007 static QEMUMachine *find_default_machine(void)
3008 {
3009 QEMUMachine *m;
3010
3011 for(m = first_machine; m != NULL; m = m->next) {
3012 if (m->is_default) {
3013 return m;
3014 }
3015 }
3016 return NULL;
3017 }
3018
3019 /***********************************************************/
3020 /* main execution loop */
3021
3022 static void gui_update(void *opaque)
3023 {
3024 uint64_t interval = GUI_REFRESH_INTERVAL;
3025 DisplayState *ds = opaque;
3026 DisplayChangeListener *dcl = ds->listeners;
3027
3028 qemu_flush_coalesced_mmio_buffer();
3029 dpy_refresh(ds);
3030
3031 while (dcl != NULL) {
3032 if (dcl->gui_timer_interval &&
3033 dcl->gui_timer_interval < interval)
3034 interval = dcl->gui_timer_interval;
3035 dcl = dcl->next;
3036 }
3037 qemu_mod_timer(ds->gui_timer, interval + qemu_get_clock(rt_clock));
3038 }
3039
3040 static void nographic_update(void *opaque)
3041 {
3042 uint64_t interval = GUI_REFRESH_INTERVAL;
3043
3044 qemu_flush_coalesced_mmio_buffer();
3045 qemu_mod_timer(nographic_timer, interval + qemu_get_clock(rt_clock));
3046 }
3047
3048 void cpu_synchronize_all_states(void)
3049 {
3050 CPUState *cpu;
3051
3052 for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) {
3053 cpu_synchronize_state(cpu);
3054 }
3055 }
3056
3057 void cpu_synchronize_all_post_reset(void)
3058 {
3059 CPUState *cpu;
3060
3061 for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) {
3062 cpu_synchronize_post_reset(cpu);
3063 }
3064 }
3065
3066 void cpu_synchronize_all_post_init(void)
3067 {
3068 CPUState *cpu;
3069
3070 for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) {
3071 cpu_synchronize_post_init(cpu);
3072 }
3073 }
3074
3075 struct vm_change_state_entry {
3076 VMChangeStateHandler *cb;
3077 void *opaque;
3078 QLIST_ENTRY (vm_change_state_entry) entries;
3079 };
3080
3081 static QLIST_HEAD(vm_change_state_head, vm_change_state_entry) vm_change_state_head;
3082
3083 VMChangeStateEntry *qemu_add_vm_change_state_handler(VMChangeStateHandler *cb,
3084 void *opaque)
3085 {
3086 VMChangeStateEntry *e;
3087
3088 e = qemu_mallocz(sizeof (*e));
3089
3090 e->cb = cb;
3091 e->opaque = opaque;
3092 QLIST_INSERT_HEAD(&vm_change_state_head, e, entries);
3093 return e;
3094 }
3095
3096 void qemu_del_vm_change_state_handler(VMChangeStateEntry *e)
3097 {
3098 QLIST_REMOVE (e, entries);
3099 qemu_free (e);
3100 }
3101
3102 static void vm_state_notify(int running, int reason)
3103 {
3104 VMChangeStateEntry *e;
3105
3106 for (e = vm_change_state_head.lh_first; e; e = e->entries.le_next) {
3107 e->cb(e->opaque, running, reason);
3108 }
3109 }
3110
3111 static void resume_all_vcpus(void);
3112 static void pause_all_vcpus(void);
3113
3114 void vm_start(void)
3115 {
3116 if (!vm_running) {
3117 cpu_enable_ticks();
3118 vm_running = 1;
3119 vm_state_notify(1, 0);
3120 resume_all_vcpus();
3121 }
3122 }
3123
3124 /* reset/shutdown handler */
3125
3126 typedef struct QEMUResetEntry {
3127 QTAILQ_ENTRY(QEMUResetEntry) entry;
3128 QEMUResetHandler *func;
3129 void *opaque;
3130 } QEMUResetEntry;
3131
3132 static QTAILQ_HEAD(reset_handlers, QEMUResetEntry) reset_handlers =
3133 QTAILQ_HEAD_INITIALIZER(reset_handlers);
3134 static int reset_requested;
3135 static int shutdown_requested;
3136 static int powerdown_requested;
3137 static int debug_requested;
3138 static int vmstop_requested;
3139
3140 int qemu_shutdown_requested(void)
3141 {
3142 int r = shutdown_requested;
3143 shutdown_requested = 0;
3144 return r;
3145 }
3146
3147 int qemu_reset_requested(void)
3148 {
3149 int r = reset_requested;
3150 reset_requested = 0;
3151 return r;
3152 }
3153
3154 int qemu_powerdown_requested(void)
3155 {
3156 int r = powerdown_requested;
3157 powerdown_requested = 0;
3158 return r;
3159 }
3160
3161 static int qemu_debug_requested(void)
3162 {
3163 int r = debug_requested;
3164 debug_requested = 0;
3165 return r;
3166 }
3167
3168 static int qemu_vmstop_requested(void)
3169 {
3170 int r = vmstop_requested;
3171 vmstop_requested = 0;
3172 return r;
3173 }
3174
3175 static void do_vm_stop(int reason)
3176 {
3177 if (vm_running) {
3178 cpu_disable_ticks();
3179 vm_running = 0;
3180 pause_all_vcpus();
3181 vm_state_notify(0, reason);
3182 monitor_protocol_event(QEVENT_STOP, NULL);
3183 }
3184 }
3185
3186 void qemu_register_reset(QEMUResetHandler *func, void *opaque)
3187 {
3188 QEMUResetEntry *re = qemu_mallocz(sizeof(QEMUResetEntry));
3189
3190 re->func = func;
3191 re->opaque = opaque;
3192 QTAILQ_INSERT_TAIL(&reset_handlers, re, entry);
3193 }
3194
3195 void qemu_unregister_reset(QEMUResetHandler *func, void *opaque)
3196 {
3197 QEMUResetEntry *re;
3198
3199 QTAILQ_FOREACH(re, &reset_handlers, entry) {
3200 if (re->func == func && re->opaque == opaque) {
3201 QTAILQ_REMOVE(&reset_handlers, re, entry);
3202 qemu_free(re);
3203 return;
3204 }
3205 }
3206 }
3207
3208 void qemu_system_reset(void)
3209 {
3210 QEMUResetEntry *re, *nre;
3211
3212 /* reset all devices */
3213 QTAILQ_FOREACH_SAFE(re, &reset_handlers, entry, nre) {
3214 re->func(re->opaque);
3215 }
3216 monitor_protocol_event(QEVENT_RESET, NULL);
3217 cpu_synchronize_all_post_reset();
3218 }
3219
3220 void qemu_system_reset_request(void)
3221 {
3222 if (no_reboot) {
3223 shutdown_requested = 1;
3224 } else {
3225 reset_requested = 1;
3226 }
3227 qemu_notify_event();
3228 }
3229
3230 void qemu_system_shutdown_request(void)
3231 {
3232 shutdown_requested = 1;
3233 qemu_notify_event();
3234 }
3235
3236 void qemu_system_powerdown_request(void)
3237 {
3238 powerdown_requested = 1;
3239 qemu_notify_event();
3240 }
3241
3242 #ifdef CONFIG_IOTHREAD
3243 static void qemu_system_vmstop_request(int reason)
3244 {
3245 vmstop_requested = reason;
3246 qemu_notify_event();
3247 }
3248 #endif
3249
3250 #ifndef _WIN32
3251 static int io_thread_fd = -1;
3252
3253 static void qemu_event_increment(void)
3254 {
3255 /* Write 8 bytes to be compatible with eventfd. */
3256 static uint64_t val = 1;
3257 ssize_t ret;
3258
3259 if (io_thread_fd == -1)
3260 return;
3261
3262 do {
3263 ret = write(io_thread_fd, &val, sizeof(val));
3264 } while (ret < 0 && errno == EINTR);
3265
3266 /* EAGAIN is fine, a read must be pending. */
3267 if (ret < 0 && errno != EAGAIN) {
3268 fprintf(stderr, "qemu_event_increment: write() filed: %s\n",
3269 strerror(errno));
3270 exit (1);
3271 }
3272 }
3273
3274 static void qemu_event_read(void *opaque)
3275 {
3276 int fd = (unsigned long)opaque;
3277 ssize_t len;
3278 char buffer[512];
3279
3280 /* Drain the notify pipe. For eventfd, only 8 bytes will be read. */
3281 do {
3282 len = read(fd, buffer, sizeof(buffer));
3283 } while ((len == -1 && errno == EINTR) || len == sizeof(buffer));
3284 }
3285
3286 static int qemu_event_init(void)
3287 {
3288 int err;
3289 int fds[2];
3290
3291 err = qemu_eventfd(fds);
3292 if (err == -1)
3293 return -errno;
3294
3295 err = fcntl_setfl(fds[0], O_NONBLOCK);
3296 if (err < 0)
3297 goto fail;
3298
3299 err = fcntl_setfl(fds[1], O_NONBLOCK);
3300 if (err < 0)
3301 goto fail;
3302
3303 qemu_set_fd_handler2(fds[0], NULL, qemu_event_read, NULL,
3304 (void *)(unsigned long)fds[0]);
3305
3306 io_thread_fd = fds[1];
3307 return 0;
3308
3309 fail:
3310 close(fds[0]);
3311 close(fds[1]);
3312 return err;
3313 }
3314 #else
3315 HANDLE qemu_event_handle;
3316
3317 static void dummy_event_handler(void *opaque)
3318 {
3319 }
3320
3321 static int qemu_event_init(void)
3322 {
3323 qemu_event_handle = CreateEvent(NULL, FALSE, FALSE, NULL);
3324 if (!qemu_event_handle) {
3325 fprintf(stderr, "Failed CreateEvent: %ld\n", GetLastError());
3326 return -1;
3327 }
3328 qemu_add_wait_object(qemu_event_handle, dummy_event_handler, NULL);
3329 return 0;
3330 }
3331
3332 static void qemu_event_increment(void)
3333 {
3334 if (!SetEvent(qemu_event_handle)) {
3335 fprintf(stderr, "qemu_event_increment: SetEvent failed: %ld\n",
3336 GetLastError());
3337 exit (1);
3338 }
3339 }
3340 #endif
3341
3342 static int cpu_can_run(CPUState *env)
3343 {
3344 if (env->stop)
3345 return 0;
3346 if (env->stopped)
3347 return 0;
3348 if (!vm_running)
3349 return 0;
3350 return 1;
3351 }
3352
3353 #ifndef CONFIG_IOTHREAD
3354 static int qemu_init_main_loop(void)
3355 {
3356 return qemu_event_init();
3357 }
3358
3359 void qemu_init_vcpu(void *_env)
3360 {
3361 CPUState *env = _env;
3362
3363 env->nr_cores = smp_cores;
3364 env->nr_threads = smp_threads;
3365 if (kvm_enabled())
3366 kvm_init_vcpu(env);
3367 return;
3368 }
3369
3370 int qemu_cpu_self(void *env)
3371 {
3372 return 1;
3373 }
3374
3375 static void resume_all_vcpus(void)
3376 {
3377 }
3378
3379 static void pause_all_vcpus(void)
3380 {
3381 }
3382
3383 void qemu_cpu_kick(void *env)
3384 {
3385 return;
3386 }
3387
3388 void qemu_notify_event(void)
3389 {
3390 CPUState *env = cpu_single_env;
3391
3392 qemu_event_increment ();
3393 if (env) {
3394 cpu_exit(env);
3395 }
3396 if (next_cpu && env != next_cpu) {
3397 cpu_exit(next_cpu);
3398 }
3399 }
3400
3401 void qemu_mutex_lock_iothread(void) {}
3402 void qemu_mutex_unlock_iothread(void) {}
3403
3404 void vm_stop(int reason)
3405 {
3406 do_vm_stop(reason);
3407 }
3408
3409 #else /* CONFIG_IOTHREAD */
3410
3411 #include "qemu-thread.h"
3412
3413 QemuMutex qemu_global_mutex;
3414 static QemuMutex qemu_fair_mutex;
3415
3416 static QemuThread io_thread;
3417
3418 static QemuThread *tcg_cpu_thread;
3419 static QemuCond *tcg_halt_cond;
3420
3421 static int qemu_system_ready;
3422 /* cpu creation */
3423 static QemuCond qemu_cpu_cond;
3424 /* system init */
3425 static QemuCond qemu_system_cond;
3426 static QemuCond qemu_pause_cond;
3427
3428 static void tcg_block_io_signals(void);
3429 static void kvm_block_io_signals(CPUState *env);
3430 static void unblock_io_signals(void);
3431 static int tcg_has_work(void);
3432 static int cpu_has_work(CPUState *env);
3433
3434 static int qemu_init_main_loop(void)
3435 {
3436 int ret;
3437
3438 ret = qemu_event_init();
3439 if (ret)
3440 return ret;
3441
3442 qemu_cond_init(&qemu_pause_cond);
3443 qemu_mutex_init(&qemu_fair_mutex);
3444 qemu_mutex_init(&qemu_global_mutex);
3445 qemu_mutex_lock(&qemu_global_mutex);
3446
3447 unblock_io_signals();
3448 qemu_thread_self(&io_thread);
3449
3450 return 0;
3451 }
3452
3453 static void qemu_wait_io_event_common(CPUState *env)
3454 {
3455 if (env->stop) {
3456 env->stop = 0;
3457 env->stopped = 1;
3458 qemu_cond_signal(&qemu_pause_cond);
3459 }
3460 }
3461
3462 static void qemu_wait_io_event(CPUState *env)
3463 {
3464 while (!tcg_has_work())
3465 qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000);
3466
3467 qemu_mutex_unlock(&qemu_global_mutex);
3468
3469 /*
3470 * Users of qemu_global_mutex can be starved, having no chance
3471 * to acquire it since this path will get to it first.
3472 * So use another lock to provide fairness.
3473 */
3474 qemu_mutex_lock(&qemu_fair_mutex);
3475 qemu_mutex_unlock(&qemu_fair_mutex);
3476
3477 qemu_mutex_lock(&qemu_global_mutex);
3478 qemu_wait_io_event_common(env);
3479 }
3480
3481 static void qemu_kvm_eat_signal(CPUState *env, int timeout)
3482 {
3483 struct timespec ts;
3484 int r, e;
3485 siginfo_t siginfo;
3486 sigset_t waitset;
3487
3488 ts.tv_sec = timeout / 1000;
3489 ts.tv_nsec = (timeout % 1000) * 1000000;
3490
3491 sigemptyset(&waitset);
3492 sigaddset(&waitset, SIG_IPI);
3493
3494 qemu_mutex_unlock(&qemu_global_mutex);
3495 r = sigtimedwait(&waitset, &siginfo, &ts);
3496 e = errno;
3497 qemu_mutex_lock(&qemu_global_mutex);
3498
3499 if (r == -1 && !(e == EAGAIN || e == EINTR)) {
3500 fprintf(stderr, "sigtimedwait: %s\n", strerror(e));
3501 exit(1);
3502 }
3503 }
3504
3505 static void qemu_kvm_wait_io_event(CPUState *env)
3506 {
3507 while (!cpu_has_work(env))
3508 qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000);
3509
3510 qemu_kvm_eat_signal(env, 0);
3511 qemu_wait_io_event_common(env);
3512 }
3513
3514 static int qemu_cpu_exec(CPUState *env);
3515
3516 static void *kvm_cpu_thread_fn(void *arg)
3517 {
3518 CPUState *env = arg;
3519
3520 qemu_thread_self(env->thread);
3521 if (kvm_enabled())
3522 kvm_init_vcpu(env);
3523
3524 kvm_block_io_signals(env);
3525
3526 /* signal CPU creation */
3527 qemu_mutex_lock(&qemu_global_mutex);
3528 env->created = 1;
3529 qemu_cond_signal(&qemu_cpu_cond);
3530
3531 /* and wait for machine initialization */
3532 while (!qemu_system_ready)
3533 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
3534
3535 while (1) {
3536 if (cpu_can_run(env))
3537 qemu_cpu_exec(env);
3538 qemu_kvm_wait_io_event(env);
3539 }
3540
3541 return NULL;
3542 }
3543
3544 static void tcg_cpu_exec(void);
3545
3546 static void *tcg_cpu_thread_fn(void *arg)
3547 {
3548 CPUState *env = arg;
3549
3550 tcg_block_io_signals();
3551 qemu_thread_self(env->thread);
3552
3553 /* signal CPU creation */
3554 qemu_mutex_lock(&qemu_global_mutex);
3555 for (env = first_cpu; env != NULL; env = env->next_cpu)
3556 env->created = 1;
3557 qemu_cond_signal(&qemu_cpu_cond);
3558
3559 /* and wait for machine initialization */
3560 while (!qemu_system_ready)
3561 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
3562
3563 while (1) {
3564 tcg_cpu_exec();
3565 qemu_wait_io_event(cur_cpu);
3566 }
3567
3568 return NULL;
3569 }
3570
3571 void qemu_cpu_kick(void *_env)
3572 {
3573 CPUState *env = _env;
3574 qemu_cond_broadcast(env->halt_cond);
3575 if (kvm_enabled())
3576 qemu_thread_signal(env->thread, SIG_IPI);
3577 }
3578
3579 int qemu_cpu_self(void *_env)
3580 {
3581 CPUState *env = _env;
3582 QemuThread this;
3583
3584 qemu_thread_self(&this);
3585
3586 return qemu_thread_equal(&this, env->thread);
3587 }
3588
3589 static void cpu_signal(int sig)
3590 {
3591 if (cpu_single_env)
3592 cpu_exit(cpu_single_env);
3593 }
3594
3595 static void tcg_block_io_signals(void)
3596 {
3597 sigset_t set;
3598 struct sigaction sigact;
3599
3600 sigemptyset(&set);
3601 sigaddset(&set, SIGUSR2);
3602 sigaddset(&set, SIGIO);
3603 sigaddset(&set, SIGALRM);
3604 sigaddset(&set, SIGCHLD);
3605 pthread_sigmask(SIG_BLOCK, &set, NULL);
3606
3607 sigemptyset(&set);
3608 sigaddset(&set, SIG_IPI);
3609 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
3610
3611 memset(&sigact, 0, sizeof(sigact));
3612 sigact.sa_handler = cpu_signal;
3613 sigaction(SIG_IPI, &sigact, NULL);
3614 }
3615
3616 static void dummy_signal(int sig)
3617 {
3618 }
3619
3620 static void kvm_block_io_signals(CPUState *env)
3621 {
3622 int r;
3623 sigset_t set;
3624 struct sigaction sigact;
3625
3626 sigemptyset(&set);
3627 sigaddset(&set, SIGUSR2);
3628 sigaddset(&set, SIGIO);
3629 sigaddset(&set, SIGALRM);
3630 sigaddset(&set, SIGCHLD);
3631 sigaddset(&set, SIG_IPI);
3632 pthread_sigmask(SIG_BLOCK, &set, NULL);
3633
3634 pthread_sigmask(SIG_BLOCK, NULL, &set);
3635 sigdelset(&set, SIG_IPI);
3636
3637 memset(&sigact, 0, sizeof(sigact));
3638 sigact.sa_handler = dummy_signal;
3639 sigaction(SIG_IPI, &sigact, NULL);
3640
3641 r = kvm_set_signal_mask(env, &set);
3642 if (r) {
3643 fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(r));
3644 exit(1);
3645 }
3646 }
3647
3648 static void unblock_io_signals(void)
3649 {
3650 sigset_t set;
3651
3652 sigemptyset(&set);
3653 sigaddset(&set, SIGUSR2);
3654 sigaddset(&set, SIGIO);
3655 sigaddset(&set, SIGALRM);
3656 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
3657
3658 sigemptyset(&set);
3659 sigaddset(&set, SIG_IPI);
3660 pthread_sigmask(SIG_BLOCK, &set, NULL);
3661 }
3662
3663 static void qemu_signal_lock(unsigned int msecs)
3664 {
3665 qemu_mutex_lock(&qemu_fair_mutex);
3666
3667 while (qemu_mutex_trylock(&qemu_global_mutex)) {
3668 qemu_thread_signal(tcg_cpu_thread, SIG_IPI);
3669 if (!qemu_mutex_timedlock(&qemu_global_mutex, msecs))
3670 break;
3671 }
3672 qemu_mutex_unlock(&qemu_fair_mutex);
3673 }
3674
3675 void qemu_mutex_lock_iothread(void)
3676 {
3677 if (kvm_enabled()) {
3678 qemu_mutex_lock(&qemu_fair_mutex);
3679 qemu_mutex_lock(&qemu_global_mutex);
3680 qemu_mutex_unlock(&qemu_fair_mutex);
3681 } else
3682 qemu_signal_lock(100);
3683 }
3684
3685 void qemu_mutex_unlock_iothread(void)
3686 {
3687 qemu_mutex_unlock(&qemu_global_mutex);
3688 }
3689
3690 static int all_vcpus_paused(void)
3691 {
3692 CPUState *penv = first_cpu;
3693
3694 while (penv) {
3695 if (!penv->stopped)
3696 return 0;
3697 penv = (CPUState *)penv->next_cpu;
3698 }
3699
3700 return 1;
3701 }
3702
3703 static void pause_all_vcpus(void)
3704 {
3705 CPUState *penv = first_cpu;
3706
3707 while (penv) {
3708 penv->stop = 1;
3709 qemu_thread_signal(penv->thread, SIG_IPI);
3710 qemu_cpu_kick(penv);
3711 penv = (CPUState *)penv->next_cpu;
3712 }
3713
3714 while (!all_vcpus_paused()) {
3715 qemu_cond_timedwait(&qemu_pause_cond, &qemu_global_mutex, 100);
3716 penv = first_cpu;
3717 while (penv) {
3718 qemu_thread_signal(penv->thread, SIG_IPI);
3719 penv = (CPUState *)penv->next_cpu;
3720 }
3721 }
3722 }
3723
3724 static void resume_all_vcpus(void)
3725 {
3726 CPUState *penv = first_cpu;
3727
3728 while (penv) {
3729 penv->stop = 0;
3730 penv->stopped = 0;
3731 qemu_thread_signal(penv->thread, SIG_IPI);
3732 qemu_cpu_kick(penv);
3733 penv = (CPUState *)penv->next_cpu;
3734 }
3735 }
3736
3737 static void tcg_init_vcpu(void *_env)
3738 {
3739 CPUState *env = _env;
3740 /* share a single thread for all cpus with TCG */
3741 if (!tcg_cpu_thread) {
3742 env->thread = qemu_mallocz(sizeof(QemuThread));
3743 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
3744 qemu_cond_init(env->halt_cond);
3745 qemu_thread_create(env->thread, tcg_cpu_thread_fn, env);
3746 while (env->created == 0)
3747 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
3748 tcg_cpu_thread = env->thread;
3749 tcg_halt_cond = env->halt_cond;
3750 } else {
3751 env->thread = tcg_cpu_thread;
3752 env->halt_cond = tcg_halt_cond;
3753 }
3754 }
3755
3756 static void kvm_start_vcpu(CPUState *env)
3757 {
3758 env->thread = qemu_mallocz(sizeof(QemuThread));
3759 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
3760 qemu_cond_init(env->halt_cond);
3761 qemu_thread_create(env->thread, kvm_cpu_thread_fn, env);
3762 while (env->created == 0)
3763 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
3764 }
3765
3766 void qemu_init_vcpu(void *_env)
3767 {
3768 CPUState *env = _env;
3769
3770 env->nr_cores = smp_cores;
3771 env->nr_threads = smp_threads;
3772 if (kvm_enabled())
3773 kvm_start_vcpu(env);
3774 else
3775 tcg_init_vcpu(env);
3776 }
3777
3778 void qemu_notify_event(void)
3779 {
3780 qemu_event_increment();
3781 }
3782
3783 void vm_stop(int reason)
3784 {
3785 QemuThread me;
3786 qemu_thread_self(&me);
3787
3788 if (!qemu_thread_equal(&me, &io_thread)) {
3789 qemu_system_vmstop_request(reason);
3790 /*
3791 * FIXME: should not return to device code in case
3792 * vm_stop() has been requested.
3793 */
3794 if (cpu_single_env) {
3795 cpu_exit(cpu_single_env);
3796 cpu_single_env->stop = 1;
3797 }
3798 return;
3799 }
3800 do_vm_stop(reason);
3801 }
3802
3803 #endif
3804
3805
3806 #ifdef _WIN32
3807 static void host_main_loop_wait(int *timeout)
3808 {
3809 int ret, ret2, i;
3810 PollingEntry *pe;
3811
3812
3813 /* XXX: need to suppress polling by better using win32 events */
3814 ret = 0;
3815 for(pe = first_polling_entry; pe != NULL; pe = pe->next) {
3816 ret |= pe->func(pe->opaque);
3817 }
3818 if (ret == 0) {
3819 int err;
3820 WaitObjects *w = &wait_objects;
3821
3822 ret = WaitForMultipleObjects(w->num, w->events, FALSE, *timeout);
3823 if (WAIT_OBJECT_0 + 0 <= ret && ret <= WAIT_OBJECT_0 + w->num - 1) {
3824 if (w->func[ret - WAIT_OBJECT_0])
3825 w->func[ret - WAIT_OBJECT_0](w->opaque[ret - WAIT_OBJECT_0]);
3826
3827 /* Check for additional signaled events */
3828 for(i = (ret - WAIT_OBJECT_0 + 1); i < w->num; i++) {
3829
3830 /* Check if event is signaled */
3831 ret2 = WaitForSingleObject(w->events[i], 0);
3832 if(ret2 == WAIT_OBJECT_0) {
3833 if (w->func[i])
3834 w->func[i](w->opaque[i]);
3835 } else if (ret2 == WAIT_TIMEOUT) {
3836 } else {
3837 err = GetLastError();
3838 fprintf(stderr, "WaitForSingleObject error %d %d\n", i, err);
3839 }
3840 }
3841 } else if (ret == WAIT_TIMEOUT) {
3842 } else {
3843 err = GetLastError();
3844 fprintf(stderr, "WaitForMultipleObjects error %d %d\n", ret, err);
3845 }
3846 }
3847
3848 *timeout = 0;
3849 }
3850 #else
3851 static void host_main_loop_wait(int *timeout)
3852 {
3853 }
3854 #endif
3855
3856 void main_loop_wait(int timeout)
3857 {
3858 IOHandlerRecord *ioh;
3859 fd_set rfds, wfds, xfds;
3860 int ret, nfds;
3861 struct timeval tv;
3862
3863 qemu_bh_update_timeout(&timeout);
3864
3865 host_main_loop_wait(&timeout);
3866
3867 /* poll any events */
3868 /* XXX: separate device handlers from system ones */
3869 nfds = -1;
3870 FD_ZERO(&rfds);
3871 FD_ZERO(&wfds);
3872 FD_ZERO(&xfds);
3873 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
3874 if (ioh->deleted)
3875 continue;
3876 if (ioh->fd_read &&
3877 (!ioh->fd_read_poll ||
3878 ioh->fd_read_poll(ioh->opaque) != 0)) {
3879 FD_SET(ioh->fd, &rfds);
3880 if (ioh->fd > nfds)
3881 nfds = ioh->fd;
3882 }
3883 if (ioh->fd_write) {
3884 FD_SET(ioh->fd, &wfds);
3885 if (ioh->fd > nfds)
3886 nfds = ioh->fd;
3887 }
3888 }
3889
3890 tv.tv_sec = timeout / 1000;
3891 tv.tv_usec = (timeout % 1000) * 1000;
3892
3893 slirp_select_fill(&nfds, &rfds, &wfds, &xfds);
3894
3895 qemu_mutex_unlock_iothread();
3896 ret = select(nfds + 1, &rfds, &wfds, &xfds, &tv);
3897 qemu_mutex_lock_iothread();
3898 if (ret > 0) {
3899 IOHandlerRecord **pioh;
3900
3901 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
3902 if (!ioh->deleted && ioh->fd_read && FD_ISSET(ioh->fd, &rfds)) {
3903 ioh->fd_read(ioh->opaque);
3904 }
3905 if (!ioh->deleted && ioh->fd_write && FD_ISSET(ioh->fd, &wfds)) {
3906 ioh->fd_write(ioh->opaque);
3907 }
3908 }
3909
3910 /* remove deleted IO handlers */
3911 pioh = &first_io_handler;
3912 while (*pioh) {
3913 ioh = *pioh;
3914 if (ioh->deleted) {
3915 *pioh = ioh->next;
3916 qemu_free(ioh);
3917 } else
3918 pioh = &ioh->next;
3919 }
3920 }
3921
3922 slirp_select_poll(&rfds, &wfds, &xfds, (ret < 0));
3923
3924 qemu_run_all_timers();
3925
3926 /* Check bottom-halves last in case any of the earlier events triggered
3927 them. */
3928 qemu_bh_poll();
3929
3930 }
3931
3932 static int qemu_cpu_exec(CPUState *env)
3933 {
3934 int ret;
3935 #ifdef CONFIG_PROFILER
3936 int64_t ti;
3937 #endif
3938
3939 #ifdef CONFIG_PROFILER
3940 ti = profile_getclock();
3941 #endif
3942 if (use_icount) {
3943 int64_t count;
3944 int decr;
3945 qemu_icount -= (env->icount_decr.u16.low + env->icount_extra);
3946 env->icount_decr.u16.low = 0;
3947 env->icount_extra = 0;
3948 count = qemu_next_deadline();
3949 count = (count + (1 << icount_time_shift) - 1)
3950 >> icount_time_shift;
3951 qemu_icount += count;
3952 decr = (count > 0xffff) ? 0xffff : count;
3953 count -= decr;
3954 env->icount_decr.u16.low = decr;
3955 env->icount_extra = count;
3956 }
3957 ret = cpu_exec(env);
3958 #ifdef CONFIG_PROFILER
3959 qemu_time += profile_getclock() - ti;
3960 #endif
3961 if (use_icount) {
3962 /* Fold pending instructions back into the
3963 instruction counter, and clear the interrupt flag. */
3964 qemu_icount -= (env->icount_decr.u16.low
3965 + env->icount_extra);
3966 env->icount_decr.u32 = 0;
3967 env->icount_extra = 0;
3968 }
3969 return ret;
3970 }
3971
3972 static void tcg_cpu_exec(void)
3973 {
3974 int ret = 0;
3975
3976 if (next_cpu == NULL)
3977 next_cpu = first_cpu;
3978 for (; next_cpu != NULL; next_cpu = next_cpu->next_cpu) {
3979 CPUState *env = cur_cpu = next_cpu;
3980
3981 qemu_clock_enable(vm_clock,
3982 (cur_cpu->singlestep_enabled & SSTEP_NOTIMER) == 0);
3983
3984 if (alarm_timer->pending)
3985 break;
3986 if (cpu_can_run(env))
3987 ret = qemu_cpu_exec(env);
3988 else if (env->stop)
3989 break;
3990
3991 if (ret == EXCP_DEBUG) {
3992 gdb_set_stop_cpu(env);
3993 debug_requested = 1;
3994 break;
3995 }
3996 }
3997 }
3998
3999 static int cpu_has_work(CPUState *env)
4000 {
4001 if (env->stop)
4002 return 1;
4003 if (env->stopped)
4004 return 0;
4005 if (!env->halted)
4006 return 1;
4007 if (qemu_cpu_has_work(env))
4008 return 1;
4009 return 0;
4010 }
4011
4012 static int tcg_has_work(void)
4013 {
4014 CPUState *env;
4015
4016 for (env = first_cpu; env != NULL; env = env->next_cpu)
4017 if (cpu_has_work(env))
4018 return 1;
4019 return 0;
4020 }
4021
4022 static int qemu_calculate_timeout(void)
4023 {
4024 #ifndef CONFIG_IOTHREAD
4025 int timeout;
4026
4027 if (!vm_running)
4028 timeout = 5000;
4029 else if (tcg_has_work())
4030 timeout = 0;
4031 else if (!use_icount)
4032 timeout = 5000;
4033 else {
4034 /* XXX: use timeout computed from timers */
4035 int64_t add;
4036 int64_t delta;
4037 /* Advance virtual time to the next event. */
4038 if (use_icount == 1) {
4039 /* When not using an adaptive execution frequency
4040 we tend to get badly out of sync with real time,
4041 so just delay for a reasonable amount of time. */
4042 delta = 0;
4043 } else {
4044 delta = cpu_get_icount() - cpu_get_clock();
4045 }
4046 if (delta > 0) {
4047 /* If virtual time is ahead of real time then just
4048 wait for IO. */
4049 timeout = (delta / 1000000) + 1;
4050 } else {
4051 /* Wait for either IO to occur or the next
4052 timer event. */
4053 add = qemu_next_deadline();
4054 /* We advance the timer before checking for IO.
4055 Limit the amount we advance so that early IO
4056 activity won't get the guest too far ahead. */
4057 if (add > 10000000)
4058 add = 10000000;
4059 delta += add;
4060 add = (add + (1 << icount_time_shift) - 1)
4061 >> icount_time_shift;
4062 qemu_icount += add;
4063 timeout = delta / 1000000;
4064 if (timeout < 0)
4065 timeout = 0;
4066 }
4067 }
4068
4069 return timeout;
4070 #else /* CONFIG_IOTHREAD */
4071 return 1000;
4072 #endif
4073 }
4074
4075 static int vm_can_run(void)
4076 {
4077 if (powerdown_requested)
4078 return 0;
4079 if (reset_requested)
4080 return 0;
4081 if (shutdown_requested)
4082 return 0;
4083 if (debug_requested)
4084 return 0;
4085 return 1;
4086 }
4087
4088 qemu_irq qemu_system_powerdown;
4089
4090 static void main_loop(void)
4091 {
4092 int r;
4093
4094 #ifdef CONFIG_IOTHREAD
4095 qemu_system_ready = 1;
4096 qemu_cond_broadcast(&qemu_system_cond);
4097 #endif
4098
4099 for (;;) {
4100 do {
4101 #ifdef CONFIG_PROFILER
4102 int64_t ti;
4103 #endif
4104 #ifndef CONFIG_IOTHREAD
4105 tcg_cpu_exec();
4106 #endif
4107 #ifdef CONFIG_PROFILER
4108 ti = profile_getclock();
4109 #endif
4110 main_loop_wait(qemu_calculate_timeout());
4111 #ifdef CONFIG_PROFILER
4112 dev_time += profile_getclock() - ti;
4113 #endif
4114 } while (vm_can_run());
4115
4116 if (qemu_debug_requested()) {
4117 vm_stop(EXCP_DEBUG);
4118 }
4119 if (qemu_shutdown_requested()) {
4120 monitor_protocol_event(QEVENT_SHUTDOWN, NULL);
4121 if (no_shutdown) {
4122 vm_stop(0);
4123 no_shutdown = 0;
4124 } else
4125 break;
4126 }
4127 if (qemu_reset_requested()) {
4128 pause_all_vcpus();
4129 qemu_system_reset();
4130 resume_all_vcpus();
4131 }
4132 if (qemu_powerdown_requested()) {
4133 monitor_protocol_event(QEVENT_POWERDOWN, NULL);
4134 qemu_irq_raise(qemu_system_powerdown);
4135 }
4136 if ((r = qemu_vmstop_requested())) {
4137 vm_stop(r);
4138 }
4139 }
4140 pause_all_vcpus();
4141 }
4142
4143 static void version(void)
4144 {
4145 printf("QEMU PC emulator version " QEMU_VERSION QEMU_PKGVERSION ", Copyright (c) 2003-2008 Fabrice Bellard\n");
4146 }
4147
4148 static void help(int exitcode)
4149 {
4150 const char *options_help =
4151 #define DEF(option, opt_arg, opt_enum, opt_help) \
4152 opt_help
4153 #define DEFHEADING(text) stringify(text) "\n"
4154 #include "qemu-options.h"
4155 #undef DEF
4156 #undef DEFHEADING
4157 #undef GEN_DOCS
4158 ;
4159 version();
4160 printf("usage: %s [options] [disk_image]\n"
4161 "\n"
4162 "'disk_image' is a raw hard image image for IDE hard disk 0\n"
4163 "\n"
4164 "%s\n"
4165 "During emulation, the following keys are useful:\n"
4166 "ctrl-alt-f toggle full screen\n"
4167 "ctrl-alt-n switch to virtual console 'n'\n"
4168 "ctrl-alt toggle mouse and keyboard grab\n"
4169 "\n"
4170 "When using -nographic, press 'ctrl-a h' to get some help.\n",
4171 "qemu",
4172 options_help);
4173 exit(exitcode);
4174 }
4175
4176 #define HAS_ARG 0x0001
4177
4178 enum {
4179 #define DEF(option, opt_arg, opt_enum, opt_help) \
4180 opt_enum,
4181 #define DEFHEADING(text)
4182 #include "qemu-options.h"
4183 #undef DEF
4184 #undef DEFHEADING
4185 #undef GEN_DOCS
4186 };
4187
4188 typedef struct QEMUOption {
4189 const char *name;
4190 int flags;
4191 int index;
4192 } QEMUOption;
4193
4194 static const QEMUOption qemu_options[] = {
4195 { "h", 0, QEMU_OPTION_h },
4196 #define DEF(option, opt_arg, opt_enum, opt_help) \
4197 { option, opt_arg, opt_enum },
4198 #define DEFHEADING(text)
4199 #include "qemu-options.h"
4200 #undef DEF
4201 #undef DEFHEADING
4202 #undef GEN_DOCS
4203 { NULL },
4204 };
4205
4206 #ifdef HAS_AUDIO
4207 struct soundhw soundhw[] = {
4208 #ifdef HAS_AUDIO_CHOICE
4209 #if defined(TARGET_I386) || defined(TARGET_MIPS)
4210 {
4211 "pcspk",
4212 "PC speaker",
4213 0,
4214 1,
4215 { .init_isa = pcspk_audio_init }
4216 },
4217 #endif
4218
4219 #ifdef CONFIG_SB16
4220 {
4221 "sb16",
4222 "Creative Sound Blaster 16",
4223 0,
4224 1,
4225 { .init_isa = SB16_init }
4226 },
4227 #endif
4228
4229 #ifdef CONFIG_CS4231A
4230 {
4231 "cs4231a",
4232 "CS4231A",
4233 0,
4234 1,
4235 { .init_isa = cs4231a_init }
4236 },
4237 #endif
4238
4239 #ifdef CONFIG_ADLIB
4240 {
4241 "adlib",
4242 #ifdef HAS_YMF262
4243 "Yamaha YMF262 (OPL3)",
4244 #else
4245 "Yamaha YM3812 (OPL2)",
4246 #endif
4247 0,
4248 1,
4249 { .init_isa = Adlib_init }
4250 },
4251 #endif
4252
4253 #ifdef CONFIG_GUS
4254 {
4255 "gus",
4256 "Gravis Ultrasound GF1",
4257 0,
4258 1,
4259 { .init_isa = GUS_init }
4260 },
4261 #endif
4262
4263 #ifdef CONFIG_AC97
4264 {
4265 "ac97",
4266 "Intel 82801AA AC97 Audio",
4267 0,
4268 0,
4269 { .init_pci = ac97_init }
4270 },
4271 #endif
4272
4273 #ifdef CONFIG_ES1370
4274 {
4275 "es1370",
4276 "ENSONIQ AudioPCI ES1370",
4277 0,
4278 0,
4279 { .init_pci = es1370_init }
4280 },
4281 #endif
4282
4283 #endif /* HAS_AUDIO_CHOICE */
4284
4285 { NULL, NULL, 0, 0, { NULL } }
4286 };
4287
4288 static void select_soundhw (const char *optarg)
4289 {
4290 struct soundhw *c;
4291
4292 if (*optarg == '?') {
4293 show_valid_cards:
4294
4295 printf ("Valid sound card names (comma separated):\n");
4296 for (c = soundhw; c->name; ++c) {
4297 printf ("%-11s %s\n", c->name, c->descr);
4298 }
4299 printf ("\n-soundhw all will enable all of the above\n");
4300 exit (*optarg != '?');
4301 }
4302 else {
4303 size_t l;
4304 const char *p;
4305 char *e;
4306 int bad_card = 0;
4307
4308 if (!strcmp (optarg, "all")) {
4309 for (c = soundhw; c->name; ++c) {
4310 c->enabled = 1;
4311 }
4312 return;
4313 }
4314
4315 p = optarg;
4316 while (*p) {
4317 e = strchr (p, ',');
4318 l = !e ? strlen (p) : (size_t) (e - p);
4319
4320 for (c = soundhw; c->name; ++c) {
4321 if (!strncmp (c->name, p, l) && !c->name[l]) {
4322 c->enabled = 1;
4323 break;
4324 }
4325 }
4326
4327 if (!c->name) {
4328 if (l > 80) {
4329 fprintf (stderr,
4330 "Unknown sound card name (too big to show)\n");
4331 }
4332 else {
4333 fprintf (stderr, "Unknown sound card name `%.*s'\n",
4334 (int) l, p);
4335 }
4336 bad_card = 1;
4337 }
4338 p += l + (e != NULL);
4339 }
4340
4341 if (bad_card)
4342 goto show_valid_cards;
4343 }
4344 }
4345 #endif
4346
4347 static void select_vgahw (const char *p)
4348 {
4349 const char *opts;
4350
4351 default_vga = 0;
4352 vga_interface_type = VGA_NONE;
4353 if (strstart(p, "std", &opts)) {
4354 vga_interface_type = VGA_STD;
4355 } else if (strstart(p, "cirrus", &opts)) {
4356 vga_interface_type = VGA_CIRRUS;
4357 } else if (strstart(p, "vmware", &opts)) {
4358 vga_interface_type = VGA_VMWARE;
4359 } else if (strstart(p, "xenfb", &opts)) {
4360 vga_interface_type = VGA_XENFB;
4361 } else if (!strstart(p, "none", &opts)) {
4362 invalid_vga:
4363 fprintf(stderr, "Unknown vga type: %s\n", p);
4364 exit(1);
4365 }
4366 while (*opts) {
4367 const char *nextopt;
4368
4369 if (strstart(opts, ",retrace=", &nextopt)) {
4370 opts = nextopt;
4371 if (strstart(opts, "dumb", &nextopt))
4372 vga_retrace_method = VGA_RETRACE_DUMB;
4373 else if (strstart(opts, "precise", &nextopt))
4374 vga_retrace_method = VGA_RETRACE_PRECISE;
4375 else goto invalid_vga;
4376 } else goto invalid_vga;
4377 opts = nextopt;
4378 }
4379 }
4380
4381 #ifdef TARGET_I386
4382 static int balloon_parse(const char *arg)
4383 {
4384 QemuOpts *opts;
4385
4386 if (strcmp(arg, "none") == 0) {
4387 return 0;
4388 }
4389
4390 if (!strncmp(arg, "virtio", 6)) {
4391 if (arg[6] == ',') {
4392 /* have params -> parse them */
4393 opts = qemu_opts_parse(&qemu_device_opts, arg+7, 0);
4394 if (!opts)
4395 return -1;
4396 } else {
4397 /* create empty opts */
4398 opts = qemu_opts_create(&qemu_device_opts, NULL, 0);
4399 }
4400 qemu_opt_set(opts, "driver", "virtio-balloon-pci");
4401 return 0;
4402 }
4403
4404 return -1;
4405 }
4406 #endif
4407
4408 #ifdef _WIN32
4409 static BOOL WINAPI qemu_ctrl_handler(DWORD type)
4410 {
4411 exit(STATUS_CONTROL_C_EXIT);
4412 return TRUE;
4413 }
4414 #endif
4415
4416 int qemu_uuid_parse(const char *str, uint8_t *uuid)
4417 {
4418 int ret;
4419
4420 if(strlen(str) != 36)
4421 return -1;
4422
4423 ret = sscanf(str, UUID_FMT, &uuid[0], &uuid[1], &uuid[2], &uuid[3],
4424 &uuid[4], &uuid[5], &uuid[6], &uuid[7], &uuid[8], &uuid[9],
4425 &uuid[10], &uuid[11], &uuid[12], &uuid[13], &uuid[14], &uuid[15]);
4426
4427 if(ret != 16)
4428 return -1;
4429
4430 #ifdef TARGET_I386
4431 smbios_add_field(1, offsetof(struct smbios_type_1, uuid), 16, uuid);
4432 #endif
4433
4434 return 0;
4435 }
4436
4437 #ifndef _WIN32
4438
4439 static void termsig_handler(int signal)
4440 {
4441 qemu_system_shutdown_request();
4442 }
4443
4444 static void sigchld_handler(int signal)
4445 {
4446 waitpid(-1, NULL, WNOHANG);
4447 }
4448
4449 static void sighandler_setup(void)
4450 {
4451 struct sigaction act;
4452
4453 memset(&act, 0, sizeof(act));
4454 act.sa_handler = termsig_handler;
4455 sigaction(SIGINT, &act, NULL);
4456 sigaction(SIGHUP, &act, NULL);
4457 sigaction(SIGTERM, &act, NULL);
4458
4459 act.sa_handler = sigchld_handler;
4460 act.sa_flags = SA_NOCLDSTOP;
4461 sigaction(SIGCHLD, &act, NULL);
4462 }
4463
4464 #endif
4465
4466 #ifdef _WIN32
4467 /* Look for support files in the same directory as the executable. */
4468 static char *find_datadir(const char *argv0)
4469 {
4470 char *p;
4471 char buf[MAX_PATH];
4472 DWORD len;
4473
4474 len = GetModuleFileName(NULL, buf, sizeof(buf) - 1);
4475 if (len == 0) {
4476 return NULL;
4477 }
4478
4479 buf[len] = 0;
4480 p = buf + len - 1;
4481 while (p != buf && *p != '\\')
4482 p--;
4483 *p = 0;
4484 if (access(buf, R_OK) == 0) {
4485 return qemu_strdup(buf);
4486 }
4487 return NULL;
4488 }
4489 #else /* !_WIN32 */
4490
4491 /* Find a likely location for support files using the location of the binary.
4492 For installed binaries this will be "$bindir/../share/qemu". When
4493 running from the build tree this will be "$bindir/../pc-bios". */
4494 #define SHARE_SUFFIX "/share/qemu"
4495 #define BUILD_SUFFIX "/pc-bios"
4496 static char *find_datadir(const char *argv0)
4497 {
4498 char *dir;
4499 char *p = NULL;
4500 char *res;
4501 char buf[PATH_MAX];
4502 size_t max_len;
4503
4504 #if defined(__linux__)
4505 {
4506 int len;
4507 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
4508 if (len > 0) {
4509 buf[len] = 0;
4510 p = buf;
4511 }
4512 }
4513 #elif defined(__FreeBSD__)
4514 {
4515 int len;
4516 len = readlink("/proc/curproc/file", buf, sizeof(buf) - 1);
4517 if (len > 0) {
4518 buf[len] = 0;
4519 p = buf;
4520 }
4521 }
4522 #endif
4523 /* If we don't have any way of figuring out the actual executable
4524 location then try argv[0]. */
4525 if (!p) {
4526 p = realpath(argv0, buf);
4527 if (!p) {
4528 return NULL;
4529 }
4530 }
4531 dir = dirname(p);
4532 dir = dirname(dir);
4533
4534 max_len = strlen(dir) +
4535 MAX(strlen(SHARE_SUFFIX), strlen(BUILD_SUFFIX)) + 1;
4536 res = qemu_mallocz(max_len);
4537 snprintf(res, max_len, "%s%s", dir, SHARE_SUFFIX);
4538 if (access(res, R_OK)) {
4539 snprintf(res, max_len, "%s%s", dir, BUILD_SUFFIX);
4540 if (access(res, R_OK)) {
4541 qemu_free(res);
4542 res = NULL;
4543 }
4544 }
4545
4546 return res;
4547 }
4548 #undef SHARE_SUFFIX
4549 #undef BUILD_SUFFIX
4550 #endif
4551
4552 char *qemu_find_file(int type, const char *name)
4553 {
4554 int len;
4555 const char *subdir;
4556 char *buf;
4557
4558 /* If name contains path separators then try it as a straight path. */
4559 if ((strchr(name, '/') || strchr(name, '\\'))
4560 && access(name, R_OK) == 0) {
4561 return qemu_strdup(name);
4562 }
4563 switch (type) {
4564 case QEMU_FILE_TYPE_BIOS:
4565 subdir = "";
4566 break;
4567 case QEMU_FILE_TYPE_KEYMAP:
4568 subdir = "keymaps/";
4569 break;
4570 default:
4571 abort();
4572 }
4573 len = strlen(data_dir) + strlen(name) + strlen(subdir) + 2;
4574 buf = qemu_mallocz(len);
4575 snprintf(buf, len, "%s/%s%s", data_dir, subdir, name);
4576 if (access(buf, R_OK)) {
4577 qemu_free(buf);
4578 return NULL;
4579 }
4580 return buf;
4581 }
4582
4583 static int device_help_func(QemuOpts *opts, void *opaque)
4584 {
4585 return qdev_device_help(opts);
4586 }
4587
4588 static int device_init_func(QemuOpts *opts, void *opaque)
4589 {
4590 DeviceState *dev;
4591
4592 dev = qdev_device_add(opts);
4593 if (!dev)
4594 return -1;
4595 return 0;
4596 }
4597
4598 static int chardev_init_func(QemuOpts *opts, void *opaque)
4599 {
4600 CharDriverState *chr;
4601
4602 chr = qemu_chr_open_opts(opts, NULL);
4603 if (!chr)
4604 return -1;
4605 return 0;
4606 }
4607
4608 static int mon_init_func(QemuOpts *opts, void *opaque)
4609 {
4610 CharDriverState *chr;
4611 const char *chardev;
4612 const char *mode;
4613 int flags;
4614
4615 mode = qemu_opt_get(opts, "mode");
4616 if (mode == NULL) {
4617 mode = "readline";
4618 }
4619 if (strcmp(mode, "readline") == 0) {
4620 flags = MONITOR_USE_READLINE;
4621 } else if (strcmp(mode, "control") == 0) {
4622 flags = MONITOR_USE_CONTROL;
4623 } else {
4624 fprintf(stderr, "unknown monitor mode \"%s\"\n", mode);
4625 exit(1);
4626 }
4627
4628 if (qemu_opt_get_bool(opts, "default", 0))
4629 flags |= MONITOR_IS_DEFAULT;
4630
4631 chardev = qemu_opt_get(opts, "chardev");
4632 chr = qemu_chr_find(chardev);
4633 if (chr == NULL) {
4634 fprintf(stderr, "chardev \"%s\" not found\n", chardev);
4635 exit(1);
4636 }
4637
4638 monitor_init(chr, flags);
4639 return 0;
4640 }
4641
4642 static void monitor_parse(const char *optarg, const char *mode)
4643 {
4644 static int monitor_device_index = 0;
4645 QemuOpts *opts;
4646 const char *p;
4647 char label[32];
4648 int def = 0;
4649
4650 if (strstart(optarg, "chardev:", &p)) {
4651 snprintf(label, sizeof(label), "%s", p);
4652 } else {
4653 if (monitor_device_index) {
4654 snprintf(label, sizeof(label), "monitor%d",
4655 monitor_device_index);
4656 } else {
4657 snprintf(label, sizeof(label), "monitor");
4658 def = 1;
4659 }
4660 opts = qemu_chr_parse_compat(label, optarg);
4661 if (!opts) {
4662 fprintf(stderr, "parse error: %s\n", optarg);
4663 exit(1);
4664 }
4665 }
4666
4667 opts = qemu_opts_create(&qemu_mon_opts, label, 1);
4668 if (!opts) {
4669 fprintf(stderr, "duplicate chardev: %s\n", label);
4670 exit(1);
4671 }
4672 qemu_opt_set(opts, "mode", mode);
4673 qemu_opt_set(opts, "chardev", label);
4674 if (def)
4675 qemu_opt_set(opts, "default", "on");
4676 monitor_device_index++;
4677 }
4678
4679 struct device_config {
4680 enum {
4681 DEV_USB, /* -usbdevice */
4682 DEV_BT, /* -bt */
4683 DEV_SERIAL, /* -serial */
4684 DEV_PARALLEL, /* -parallel */
4685 DEV_VIRTCON, /* -virtioconsole */
4686 DEV_DEBUGCON, /* -debugcon */
4687 } type;
4688 const char *cmdline;
4689 QTAILQ_ENTRY(device_config) next;
4690 };
4691 QTAILQ_HEAD(, device_config) device_configs = QTAILQ_HEAD_INITIALIZER(device_configs);
4692
4693 static void add_device_config(int type, const char *cmdline)
4694 {
4695 struct device_config *conf;
4696
4697 conf = qemu_mallocz(sizeof(*conf));
4698 conf->type = type;
4699 conf->cmdline = cmdline;
4700 QTAILQ_INSERT_TAIL(&device_configs, conf, next);
4701 }
4702
4703 static int foreach_device_config(int type, int (*func)(const char *cmdline))
4704 {
4705 struct device_config *conf;
4706 int rc;
4707
4708 QTAILQ_FOREACH(conf, &device_configs, next) {
4709 if (conf->type != type)
4710 continue;
4711 rc = func(conf->cmdline);
4712 if (0 != rc)
4713 return rc;
4714 }
4715 return 0;
4716 }
4717
4718 static int serial_parse(const char *devname)
4719 {
4720 static int index = 0;
4721 char label[32];
4722
4723 if (strcmp(devname, "none") == 0)
4724 return 0;
4725 if (index == MAX_SERIAL_PORTS) {
4726 fprintf(stderr, "qemu: too many serial ports\n");
4727 exit(1);
4728 }
4729 snprintf(label, sizeof(label), "serial%d", index);
4730 serial_hds[index] = qemu_chr_open(label, devname, NULL);
4731 if (!serial_hds[index]) {
4732 fprintf(stderr, "qemu: could not open serial device '%s': %s\n",
4733 devname, strerror(errno));
4734 return -1;
4735 }
4736 index++;
4737 return 0;
4738 }
4739
4740 static int parallel_parse(const char *devname)
4741 {
4742 static int index = 0;
4743 char label[32];
4744
4745 if (strcmp(devname, "none") == 0)
4746 return 0;
4747 if (index == MAX_PARALLEL_PORTS) {
4748 fprintf(stderr, "qemu: too many parallel ports\n");
4749 exit(1);
4750 }
4751 snprintf(label, sizeof(label), "parallel%d", index);
4752 parallel_hds[index] = qemu_chr_open(label, devname, NULL);
4753 if (!parallel_hds[index]) {
4754 fprintf(stderr, "qemu: could not open parallel device '%s': %s\n",
4755 devname, strerror(errno));
4756 return -1;
4757 }
4758 index++;
4759 return 0;
4760 }
4761
4762 static int virtcon_parse(const char *devname)
4763 {
4764 static int index = 0;
4765 char label[32];
4766 QemuOpts *bus_opts, *dev_opts;
4767
4768 if (strcmp(devname, "none") == 0)
4769 return 0;
4770 if (index == MAX_VIRTIO_CONSOLES) {
4771 fprintf(stderr, "qemu: too many virtio consoles\n");
4772 exit(1);
4773 }
4774
4775 bus_opts = qemu_opts_create(&qemu_device_opts, NULL, 0);
4776 qemu_opt_set(bus_opts, "driver", "virtio-serial");
4777
4778 dev_opts = qemu_opts_create(&qemu_device_opts, NULL, 0);
4779 qemu_opt_set(dev_opts, "driver", "virtconsole");
4780
4781 snprintf(label, sizeof(label), "virtcon%d", index);
4782 virtcon_hds[index] = qemu_chr_open(label, devname, NULL);
4783 if (!virtcon_hds[index]) {
4784 fprintf(stderr, "qemu: could not open virtio console '%s': %s\n",
4785 devname, strerror(errno));
4786 return -1;
4787 }
4788 qemu_opt_set(dev_opts, "chardev", label);
4789
4790 index++;
4791 return 0;
4792 }
4793
4794 static int debugcon_parse(const char *devname)
4795 {
4796 QemuOpts *opts;
4797
4798 if (!qemu_chr_open("debugcon", devname, NULL)) {
4799 exit(1);
4800 }
4801 opts = qemu_opts_create(&qemu_device_opts, "debugcon", 1);
4802 if (!opts) {
4803 fprintf(stderr, "qemu: already have a debugcon device\n");
4804 exit(1);
4805 }
4806 qemu_opt_set(opts, "driver", "isa-debugcon");
4807 qemu_opt_set(opts, "chardev", "debugcon");
4808 return 0;
4809 }
4810
4811 static const QEMUOption *lookup_opt(int argc, char **argv,
4812 const char **poptarg, int *poptind)
4813 {
4814 const QEMUOption *popt;
4815 int optind = *poptind;
4816 char *r = argv[optind];
4817 const char *optarg;
4818
4819 loc_set_cmdline(argv, optind, 1);
4820 optind++;
4821 /* Treat --foo the same as -foo. */
4822 if (r[1] == '-')
4823 r++;
4824 popt = qemu_options;
4825 for(;;) {
4826 if (!popt->name) {
4827 error_report("invalid option");
4828 exit(1);
4829 }
4830 if (!strcmp(popt->name, r + 1))
4831 break;
4832 popt++;
4833 }
4834 if (popt->flags & HAS_ARG) {
4835 if (optind >= argc) {
4836 error_report("requires an argument");
4837 exit(1);
4838 }
4839 optarg = argv[optind++];
4840 loc_set_cmdline(argv, optind - 2, 2);
4841 } else {
4842 optarg = NULL;
4843 }
4844
4845 *poptarg = optarg;
4846 *poptind = optind;
4847
4848 return popt;
4849 }
4850
4851 int main(int argc, char **argv, char **envp)
4852 {
4853 const char *gdbstub_dev = NULL;
4854 uint32_t boot_devices_bitmap = 0;
4855 int i;
4856 int snapshot, linux_boot, net_boot;
4857 const char *initrd_filename;
4858 const char *kernel_filename, *kernel_cmdline;
4859 char boot_devices[33] = "cad"; /* default to HD->floppy->CD-ROM */
4860 DisplayState *ds;
4861 DisplayChangeListener *dcl;
4862 int cyls, heads, secs, translation;
4863 QemuOpts *hda_opts = NULL, *opts;
4864 int optind;
4865 const char *optarg;
4866 const char *loadvm = NULL;
4867 QEMUMachine *machine;
4868 const char *cpu_model;
4869 #ifndef _WIN32
4870 int fds[2];
4871 #endif
4872 int tb_size;
4873 const char *pid_file = NULL;
4874 const char *incoming = NULL;
4875 #ifndef _WIN32
4876 int fd = 0;
4877 struct passwd *pwd = NULL;
4878 const char *chroot_dir = NULL;
4879 const char *run_as = NULL;
4880 #endif
4881 CPUState *env;
4882 int show_vnc_port = 0;
4883 int defconfig = 1;
4884
4885 error_set_progname(argv[0]);
4886
4887 init_clocks();
4888
4889 qemu_cache_utils_init(envp);
4890
4891 QLIST_INIT (&vm_change_state_head);
4892 #ifndef _WIN32
4893 {
4894 struct sigaction act;
4895 sigfillset(&act.sa_mask);
4896 act.sa_flags = 0;
4897 act.sa_handler = SIG_IGN;
4898 sigaction(SIGPIPE, &act, NULL);
4899 }
4900 #else
4901 SetConsoleCtrlHandler(qemu_ctrl_handler, TRUE);
4902 /* Note: cpu_interrupt() is currently not SMP safe, so we force
4903 QEMU to run on a single CPU */
4904 {
4905 HANDLE h;
4906 DWORD mask, smask;
4907 int i;
4908 h = GetCurrentProcess();
4909 if (GetProcessAffinityMask(h, &mask, &smask)) {
4910 for(i = 0; i < 32; i++) {
4911 if (mask & (1 << i))
4912 break;
4913 }
4914 if (i != 32) {
4915 mask = 1 << i;
4916 SetProcessAffinityMask(h, mask);
4917 }
4918 }
4919 }
4920 #endif
4921
4922 module_call_init(MODULE_INIT_MACHINE);
4923 machine = find_default_machine();
4924 cpu_model = NULL;
4925 initrd_filename = NULL;
4926 ram_size = 0;
4927 snapshot = 0;
4928 kernel_filename = NULL;
4929 kernel_cmdline = "";
4930 cyls = heads = secs = 0;
4931 translation = BIOS_ATA_TRANSLATION_AUTO;
4932
4933 for (i = 0; i < MAX_NODES; i++) {
4934 node_mem[i] = 0;
4935 node_cpumask[i] = 0;
4936 }
4937
4938 nb_numa_nodes = 0;
4939 nb_nics = 0;
4940
4941 tb_size = 0;
4942 autostart= 1;
4943
4944 /* first pass of option parsing */
4945 optind = 1;
4946 while (optind < argc) {
4947 if (argv[optind][0] != '-') {
4948 /* disk image */
4949 optind++;
4950 continue;
4951 } else {
4952 const QEMUOption *popt;
4953
4954 popt = lookup_opt(argc, argv, &optarg, &optind);
4955 switch (popt->index) {
4956 case QEMU_OPTION_nodefconfig:
4957 defconfig=0;
4958 break;
4959 }
4960 }
4961 }
4962
4963 if (defconfig) {
4964 const char *fname;
4965 FILE *fp;
4966
4967 fname = CONFIG_QEMU_CONFDIR "/qemu.conf";
4968 fp = fopen(fname, "r");
4969 if (fp) {
4970 if (qemu_config_parse(fp, fname) != 0) {
4971 exit(1);
4972 }
4973 fclose(fp);
4974 }
4975
4976 fname = CONFIG_QEMU_CONFDIR "/target-" TARGET_ARCH ".conf";
4977 fp = fopen(fname, "r");
4978 if (fp) {
4979 if (qemu_config_parse(fp, fname) != 0) {
4980 exit(1);
4981 }
4982 fclose(fp);
4983 }
4984 }
4985 #if defined(cpudef_setup)
4986 cpudef_setup(); /* parse cpu definitions in target config file */
4987 #endif
4988
4989 /* second pass of option parsing */
4990 optind = 1;
4991 for(;;) {
4992 if (optind >= argc)
4993 break;
4994 if (argv[optind][0] != '-') {
4995 hda_opts = drive_add(argv[optind++], HD_ALIAS, 0);
4996 } else {
4997 const QEMUOption *popt;
4998
4999 popt = lookup_opt(argc, argv, &optarg, &optind);
5000 switch(popt->index) {
5001 case QEMU_OPTION_M:
5002 machine = find_machine(optarg);
5003 if (!machine) {
5004 QEMUMachine *m;
5005 printf("Supported machines are:\n");
5006 for(m = first_machine; m != NULL; m = m->next) {
5007 if (m->alias)
5008 printf("%-10s %s (alias of %s)\n",
5009 m->alias, m->desc, m->name);
5010 printf("%-10s %s%s\n",
5011 m->name, m->desc,
5012 m->is_default ? " (default)" : "");
5013 }
5014 exit(*optarg != '?');
5015 }
5016 break;
5017 case QEMU_OPTION_cpu:
5018 /* hw initialization will check this */
5019 if (*optarg == '?') {
5020 /* XXX: implement xxx_cpu_list for targets that still miss it */
5021 #if defined(cpu_list_id)
5022 cpu_list_id(stdout, &fprintf, optarg);
5023 #elif defined(cpu_list)
5024 cpu_list(stdout, &fprintf); /* deprecated */
5025 #endif
5026 exit(0);
5027 } else {
5028 cpu_model = optarg;
5029 }
5030 break;
5031 case QEMU_OPTION_initrd:
5032 initrd_filename = optarg;
5033 break;
5034 case QEMU_OPTION_hda:
5035 if (cyls == 0)
5036 hda_opts = drive_add(optarg, HD_ALIAS, 0);
5037 else
5038 hda_opts = drive_add(optarg, HD_ALIAS
5039 ",cyls=%d,heads=%d,secs=%d%s",
5040 0, cyls, heads, secs,
5041 translation == BIOS_ATA_TRANSLATION_LBA ?
5042 ",trans=lba" :
5043 translation == BIOS_ATA_TRANSLATION_NONE ?
5044 ",trans=none" : "");
5045 break;
5046 case QEMU_OPTION_hdb:
5047 case QEMU_OPTION_hdc:
5048 case QEMU_OPTION_hdd:
5049 drive_add(optarg, HD_ALIAS, popt->index - QEMU_OPTION_hda);
5050 break;
5051 case QEMU_OPTION_drive:
5052 drive_add(NULL, "%s", optarg);
5053 break;
5054 case QEMU_OPTION_set:
5055 if (qemu_set_option(optarg) != 0)
5056 exit(1);
5057 break;
5058 case QEMU_OPTION_global:
5059 if (qemu_global_option(optarg) != 0)
5060 exit(1);
5061 break;
5062 case QEMU_OPTION_mtdblock:
5063 drive_add(optarg, MTD_ALIAS);
5064 break;
5065 case QEMU_OPTION_sd:
5066 drive_add(optarg, SD_ALIAS);
5067 break;
5068 case QEMU_OPTION_pflash:
5069 drive_add(optarg, PFLASH_ALIAS);
5070 break;
5071 case QEMU_OPTION_snapshot:
5072 snapshot = 1;
5073 break;
5074 case QEMU_OPTION_hdachs:
5075 {
5076 const char *p;
5077 p = optarg;
5078 cyls = strtol(p, (char **)&p, 0);
5079 if (cyls < 1 || cyls > 16383)
5080 goto chs_fail;
5081 if (*p != ',')
5082 goto chs_fail;
5083 p++;
5084 heads = strtol(p, (char **)&p, 0);
5085 if (heads < 1 || heads > 16)
5086 goto chs_fail;
5087 if (*p != ',')
5088 goto chs_fail;
5089 p++;
5090 secs = strtol(p, (char **)&p, 0);
5091 if (secs < 1 || secs > 63)
5092 goto chs_fail;
5093 if (*p == ',') {
5094 p++;
5095 if (!strcmp(p, "none"))
5096 translation = BIOS_ATA_TRANSLATION_NONE;
5097 else if (!strcmp(p, "lba"))
5098 translation = BIOS_ATA_TRANSLATION_LBA;
5099 else if (!strcmp(p, "auto"))
5100 translation = BIOS_ATA_TRANSLATION_AUTO;
5101 else
5102 goto chs_fail;
5103 } else if (*p != '\0') {
5104 chs_fail:
5105 fprintf(stderr, "qemu: invalid physical CHS format\n");
5106 exit(1);
5107 }
5108 if (hda_opts != NULL) {
5109 char num[16];
5110 snprintf(num, sizeof(num), "%d", cyls);
5111 qemu_opt_set(hda_opts, "cyls", num);
5112 snprintf(num, sizeof(num), "%d", heads);
5113 qemu_opt_set(hda_opts, "heads", num);
5114 snprintf(num, sizeof(num), "%d", secs);
5115 qemu_opt_set(hda_opts, "secs", num);
5116 if (translation == BIOS_ATA_TRANSLATION_LBA)
5117 qemu_opt_set(hda_opts, "trans", "lba");
5118 if (translation == BIOS_ATA_TRANSLATION_NONE)
5119 qemu_opt_set(hda_opts, "trans", "none");
5120 }
5121 }
5122 break;
5123 case QEMU_OPTION_numa:
5124 if (nb_numa_nodes >= MAX_NODES) {
5125 fprintf(stderr, "qemu: too many NUMA nodes\n");
5126 exit(1);
5127 }
5128 numa_add(optarg);
5129 break;
5130 case QEMU_OPTION_nographic:
5131 display_type = DT_NOGRAPHIC;
5132 break;
5133 #ifdef CONFIG_CURSES
5134 case QEMU_OPTION_curses:
5135 display_type = DT_CURSES;
5136 break;
5137 #endif
5138 case QEMU_OPTION_portrait:
5139 graphic_rotate = 1;
5140 break;
5141 case QEMU_OPTION_kernel:
5142 kernel_filename = optarg;
5143 break;
5144 case QEMU_OPTION_append:
5145 kernel_cmdline = optarg;
5146 break;
5147 case QEMU_OPTION_cdrom:
5148 drive_add(optarg, CDROM_ALIAS);
5149 break;
5150 case QEMU_OPTION_boot:
5151 {
5152 static const char * const params[] = {
5153 "order", "once", "menu", NULL
5154 };
5155 char buf[sizeof(boot_devices)];
5156 char *standard_boot_devices;
5157 int legacy = 0;
5158
5159 if (!strchr(optarg, '=')) {
5160 legacy = 1;
5161 pstrcpy(buf, sizeof(buf), optarg);
5162 } else if (check_params(buf, sizeof(buf), params, optarg) < 0) {
5163 fprintf(stderr,
5164 "qemu: unknown boot parameter '%s' in '%s'\n",
5165 buf, optarg);
5166 exit(1);
5167 }
5168
5169 if (legacy ||
5170 get_param_value(buf, sizeof(buf), "order", optarg)) {
5171 boot_devices_bitmap = parse_bootdevices(buf);
5172 pstrcpy(boot_devices, sizeof(boot_devices), buf);
5173 }
5174 if (!legacy) {
5175 if (get_param_value(buf, sizeof(buf),
5176 "once", optarg)) {
5177 boot_devices_bitmap |= parse_bootdevices(buf);
5178 standard_boot_devices = qemu_strdup(boot_devices);
5179 pstrcpy(boot_devices, sizeof(boot_devices), buf);
5180 qemu_register_reset(restore_boot_devices,
5181 standard_boot_devices);
5182 }
5183 if (get_param_value(buf, sizeof(buf),
5184 "menu", optarg)) {
5185 if (!strcmp(buf, "on")) {
5186 boot_menu = 1;
5187 } else if (!strcmp(buf, "off")) {
5188 boot_menu = 0;
5189 } else {
5190 fprintf(stderr,
5191 "qemu: invalid option value '%s'\n",
5192 buf);
5193 exit(1);
5194 }
5195 }
5196 }
5197 }
5198 break;
5199 case QEMU_OPTION_fda:
5200 case QEMU_OPTION_fdb:
5201 drive_add(optarg, FD_ALIAS, popt->index - QEMU_OPTION_fda);
5202 break;
5203 #ifdef TARGET_I386
5204 case QEMU_OPTION_no_fd_bootchk:
5205 fd_bootchk = 0;
5206 break;
5207 #endif
5208 case QEMU_OPTION_netdev:
5209 if (net_client_parse(&qemu_netdev_opts, optarg) == -1) {
5210 exit(1);
5211 }
5212 break;
5213 case QEMU_OPTION_net:
5214 if (net_client_parse(&qemu_net_opts, optarg) == -1) {
5215 exit(1);
5216 }
5217 break;
5218 #ifdef CONFIG_SLIRP
5219 case QEMU_OPTION_tftp:
5220 legacy_tftp_prefix = optarg;
5221 break;
5222 case QEMU_OPTION_bootp:
5223 legacy_bootp_filename = optarg;
5224 break;
5225 #ifndef _WIN32
5226 case QEMU_OPTION_smb:
5227 if (net_slirp_smb(optarg) < 0)
5228 exit(1);
5229 break;
5230 #endif
5231 case QEMU_OPTION_redir:
5232 if (net_slirp_redir(optarg) < 0)
5233 exit(1);
5234 break;
5235 #endif
5236 case QEMU_OPTION_bt:
5237 add_device_config(DEV_BT, optarg);
5238 break;
5239 #ifdef HAS_AUDIO
5240 case QEMU_OPTION_audio_help:
5241 AUD_help ();
5242 exit (0);
5243 break;
5244 case QEMU_OPTION_soundhw:
5245 select_soundhw (optarg);
5246 break;
5247 #endif
5248 case QEMU_OPTION_h:
5249 help(0);
5250 break;
5251 case QEMU_OPTION_version:
5252 version();
5253 exit(0);
5254 break;
5255 case QEMU_OPTION_m: {
5256 uint64_t value;
5257 char *ptr;
5258
5259 value = strtoul(optarg, &ptr, 10);
5260 switch (*ptr) {
5261 case 0: case 'M': case 'm':
5262 value <<= 20;
5263 break;
5264 case 'G': case 'g':
5265 value <<= 30;
5266 break;
5267 default:
5268 fprintf(stderr, "qemu: invalid ram size: %s\n", optarg);
5269 exit(1);
5270 }
5271
5272 /* On 32-bit hosts, QEMU is limited by virtual address space */
5273 if (value > (2047 << 20) && HOST_LONG_BITS == 32) {
5274 fprintf(stderr, "qemu: at most 2047 MB RAM can be simulated\n");
5275 exit(1);
5276 }
5277 if (value != (uint64_t)(ram_addr_t)value) {
5278 fprintf(stderr, "qemu: ram size too large\n");
5279 exit(1);
5280 }
5281 ram_size = value;
5282 break;
5283 }
5284 case QEMU_OPTION_mempath:
5285 mem_path = optarg;
5286 break;
5287 #ifdef MAP_POPULATE
5288 case QEMU_OPTION_mem_prealloc:
5289 mem_prealloc = 1;
5290 break;
5291 #endif
5292 case QEMU_OPTION_d:
5293 {
5294 int mask;
5295 const CPULogItem *item;
5296
5297 mask = cpu_str_to_log_mask(optarg);
5298 if (!mask) {
5299 printf("Log items (comma separated):\n");
5300 for(item = cpu_log_items; item->mask != 0; item++) {
5301 printf("%-10s %s\n", item->name, item->help);
5302 }
5303 exit(1);
5304 }
5305 cpu_set_log(mask);
5306 }
5307 break;
5308 case QEMU_OPTION_s:
5309 gdbstub_dev = "tcp::" DEFAULT_GDBSTUB_PORT;
5310 break;
5311 case QEMU_OPTION_gdb:
5312 gdbstub_dev = optarg;
5313 break;
5314 case QEMU_OPTION_L:
5315 data_dir = optarg;
5316 break;
5317 case QEMU_OPTION_bios:
5318 bios_name = optarg;
5319 break;
5320 case QEMU_OPTION_singlestep:
5321 singlestep = 1;
5322 break;
5323 case QEMU_OPTION_S:
5324 autostart = 0;
5325 break;
5326 case QEMU_OPTION_k:
5327 keyboard_layout = optarg;
5328 break;
5329 case QEMU_OPTION_localtime:
5330 rtc_utc = 0;
5331 break;
5332 case QEMU_OPTION_vga:
5333 select_vgahw (optarg);
5334 break;
5335 #if defined(TARGET_PPC) || defined(TARGET_SPARC)
5336 case QEMU_OPTION_g:
5337 {
5338 const char *p;
5339 int w, h, depth;
5340 p = optarg;
5341 w = strtol(p, (char **)&p, 10);
5342 if (w <= 0) {
5343 graphic_error:
5344 fprintf(stderr, "qemu: invalid resolution or depth\n");
5345 exit(1);
5346 }
5347 if (*p != 'x')
5348 goto graphic_error;
5349 p++;
5350 h = strtol(p, (char **)&p, 10);
5351 if (h <= 0)
5352 goto graphic_error;
5353 if (*p == 'x') {
5354 p++;
5355 depth = strtol(p, (char **)&p, 10);
5356 if (depth != 8 && depth != 15 && depth != 16 &&
5357 depth != 24 && depth != 32)
5358 goto graphic_error;
5359 } else if (*p == '\0') {
5360 depth = graphic_depth;
5361 } else {
5362 goto graphic_error;
5363 }
5364
5365 graphic_width = w;
5366 graphic_height = h;
5367 graphic_depth = depth;
5368 }
5369 break;
5370 #endif
5371 case QEMU_OPTION_echr:
5372 {
5373 char *r;
5374 term_escape_char = strtol(optarg, &r, 0);
5375 if (r == optarg)
5376 printf("Bad argument to echr\n");
5377 break;
5378 }
5379 case QEMU_OPTION_monitor:
5380 monitor_parse(optarg, "readline");
5381 default_monitor = 0;
5382 break;
5383 case QEMU_OPTION_qmp:
5384 monitor_parse(optarg, "control");
5385 default_monitor = 0;
5386 break;
5387 case QEMU_OPTION_mon:
5388 opts = qemu_opts_parse(&qemu_mon_opts, optarg, 1);
5389 if (!opts) {
5390 fprintf(stderr, "parse error: %s\n", optarg);
5391 exit(1);
5392 }
5393 default_monitor = 0;
5394 break;
5395 case QEMU_OPTION_chardev:
5396 opts = qemu_opts_parse(&qemu_chardev_opts, optarg, 1);
5397 if (!opts) {
5398 fprintf(stderr, "parse error: %s\n", optarg);
5399 exit(1);
5400 }
5401 break;
5402 case QEMU_OPTION_serial:
5403 add_device_config(DEV_SERIAL, optarg);
5404 default_serial = 0;
5405 if (strncmp(optarg, "mon:", 4) == 0) {
5406 default_monitor = 0;
5407 }
5408 break;
5409 case QEMU_OPTION_watchdog:
5410 if (watchdog) {
5411 fprintf(stderr,
5412 "qemu: only one watchdog option may be given\n");
5413 return 1;
5414 }
5415 watchdog = optarg;
5416 break;
5417 case QEMU_OPTION_watchdog_action:
5418 if (select_watchdog_action(optarg) == -1) {
5419 fprintf(stderr, "Unknown -watchdog-action parameter\n");
5420 exit(1);
5421 }
5422 break;
5423 case QEMU_OPTION_virtiocon:
5424 add_device_config(DEV_VIRTCON, optarg);
5425 default_virtcon = 0;
5426 if (strncmp(optarg, "mon:", 4) == 0) {
5427 default_monitor = 0;
5428 }
5429 break;
5430 case QEMU_OPTION_parallel:
5431 add_device_config(DEV_PARALLEL, optarg);
5432 default_parallel = 0;
5433 if (strncmp(optarg, "mon:", 4) == 0) {
5434 default_monitor = 0;
5435 }
5436 break;
5437 case QEMU_OPTION_debugcon:
5438 add_device_config(DEV_DEBUGCON, optarg);
5439 break;
5440 case QEMU_OPTION_loadvm:
5441 loadvm = optarg;
5442 break;
5443 case QEMU_OPTION_full_screen:
5444 full_screen = 1;
5445 break;
5446 #ifdef CONFIG_SDL
5447 case QEMU_OPTION_no_frame:
5448 no_frame = 1;
5449 break;
5450 case QEMU_OPTION_alt_grab:
5451 alt_grab = 1;
5452 break;
5453 case QEMU_OPTION_ctrl_grab:
5454 ctrl_grab = 1;
5455 break;
5456 case QEMU_OPTION_no_quit:
5457 no_quit = 1;
5458 break;
5459 case QEMU_OPTION_sdl:
5460 display_type = DT_SDL;
5461 break;
5462 #endif
5463 case QEMU_OPTION_pidfile:
5464 pid_file = optarg;
5465 break;
5466 #ifdef TARGET_I386
5467 case QEMU_OPTION_win2k_hack:
5468 win2k_install_hack = 1;
5469 break;
5470 case QEMU_OPTION_rtc_td_hack:
5471 rtc_td_hack = 1;
5472 break;
5473 case QEMU_OPTION_acpitable:
5474 if(acpi_table_add(optarg) < 0) {
5475 fprintf(stderr, "Wrong acpi table provided\n");
5476 exit(1);
5477 }
5478 break;
5479 case QEMU_OPTION_smbios:
5480 if(smbios_entry_add(optarg) < 0) {
5481 fprintf(stderr, "Wrong smbios provided\n");
5482 exit(1);
5483 }
5484 break;
5485 #endif
5486 #ifdef CONFIG_KVM
5487 case QEMU_OPTION_enable_kvm:
5488 kvm_allowed = 1;
5489 break;
5490 #endif
5491 case QEMU_OPTION_usb:
5492 usb_enabled = 1;
5493 break;
5494 case QEMU_OPTION_usbdevice:
5495 usb_enabled = 1;
5496 add_device_config(DEV_USB, optarg);
5497 break;
5498 case QEMU_OPTION_device:
5499 if (!qemu_opts_parse(&qemu_device_opts, optarg, 1)) {
5500 exit(1);
5501 }
5502 break;
5503 case QEMU_OPTION_smp:
5504 smp_parse(optarg);
5505 if (smp_cpus < 1) {
5506 fprintf(stderr, "Invalid number of CPUs\n");
5507 exit(1);
5508 }
5509 if (max_cpus < smp_cpus) {
5510 fprintf(stderr, "maxcpus must be equal to or greater than "
5511 "smp\n");
5512 exit(1);
5513 }
5514 if (max_cpus > 255) {
5515 fprintf(stderr, "Unsupported number of maxcpus\n");
5516 exit(1);
5517 }
5518 break;
5519 case QEMU_OPTION_vnc:
5520 display_type = DT_VNC;
5521 vnc_display = optarg;
5522 break;
5523 #ifdef TARGET_I386
5524 case QEMU_OPTION_no_acpi:
5525 acpi_enabled = 0;
5526 break;
5527 case QEMU_OPTION_no_hpet:
5528 no_hpet = 1;
5529 break;
5530 case QEMU_OPTION_balloon:
5531 if (balloon_parse(optarg) < 0) {
5532 fprintf(stderr, "Unknown -balloon argument %s\n", optarg);
5533 exit(1);
5534 }
5535 break;
5536 #endif
5537 case QEMU_OPTION_no_reboot:
5538 no_reboot = 1;
5539 break;
5540 case QEMU_OPTION_no_shutdown:
5541 no_shutdown = 1;
5542 break;
5543 case QEMU_OPTION_show_cursor:
5544 cursor_hide = 0;
5545 break;
5546 case QEMU_OPTION_uuid:
5547 if(qemu_uuid_parse(optarg, qemu_uuid) < 0) {
5548 fprintf(stderr, "Fail to parse UUID string."
5549 " Wrong format.\n");
5550 exit(1);
5551 }
5552 break;
5553 #ifndef _WIN32
5554 case QEMU_OPTION_daemonize:
5555 daemonize = 1;
5556 break;
5557 #endif
5558 case QEMU_OPTION_option_rom:
5559 if (nb_option_roms >= MAX_OPTION_ROMS) {
5560 fprintf(stderr, "Too many option ROMs\n");
5561 exit(1);
5562 }
5563 option_rom[nb_option_roms] = optarg;
5564 nb_option_roms++;
5565 break;
5566 #if defined(TARGET_ARM) || defined(TARGET_M68K)
5567 case QEMU_OPTION_semihosting:
5568 semihosting_enabled = 1;
5569 break;
5570 #endif
5571 case QEMU_OPTION_name:
5572 qemu_name = qemu_strdup(optarg);
5573 {
5574 char *p = strchr(qemu_name, ',');
5575 if (p != NULL) {
5576 *p++ = 0;
5577 if (strncmp(p, "process=", 8)) {
5578 fprintf(stderr, "Unknown subargument %s to -name", p);
5579 exit(1);
5580 }
5581 p += 8;
5582 set_proc_name(p);
5583 }
5584 }
5585 break;
5586 #if defined(TARGET_SPARC) || defined(TARGET_PPC)
5587 case QEMU_OPTION_prom_env:
5588 if (nb_prom_envs >= MAX_PROM_ENVS) {
5589 fprintf(stderr, "Too many prom variables\n");
5590 exit(1);
5591 }
5592 prom_envs[nb_prom_envs] = optarg;
5593 nb_prom_envs++;
5594 break;
5595 #endif
5596 #ifdef TARGET_ARM
5597 case QEMU_OPTION_old_param:
5598 old_param = 1;
5599 break;
5600 #endif
5601 case QEMU_OPTION_clock:
5602 configure_alarms(optarg);
5603 break;
5604 case QEMU_OPTION_startdate:
5605 configure_rtc_date_offset(optarg, 1);
5606 break;
5607 case QEMU_OPTION_rtc:
5608 opts = qemu_opts_parse(&qemu_rtc_opts, optarg, 0);
5609 if (!opts) {
5610 fprintf(stderr, "parse error: %s\n", optarg);
5611 exit(1);
5612 }
5613 configure_rtc(opts);
5614 break;
5615 case QEMU_OPTION_tb_size:
5616 tb_size = strtol(optarg, NULL, 0);
5617 if (tb_size < 0)
5618 tb_size = 0;
5619 break;
5620 case QEMU_OPTION_icount:
5621 use_icount = 1;
5622 if (strcmp(optarg, "auto") == 0) {
5623 icount_time_shift = -1;
5624 } else {
5625 icount_time_shift = strtol(optarg, NULL, 0);
5626 }
5627 break;
5628 case QEMU_OPTION_incoming:
5629 incoming = optarg;
5630 break;
5631 case QEMU_OPTION_nodefaults:
5632 default_serial = 0;
5633 default_parallel = 0;
5634 default_virtcon = 0;
5635 default_monitor = 0;
5636 default_vga = 0;
5637 default_net = 0;
5638 default_floppy = 0;
5639 default_cdrom = 0;
5640 default_sdcard = 0;
5641 break;
5642 #ifndef _WIN32
5643 case QEMU_OPTION_chroot:
5644 chroot_dir = optarg;
5645 break;
5646 case QEMU_OPTION_runas:
5647 run_as = optarg;
5648 break;
5649 #endif
5650 #ifdef CONFIG_XEN
5651 case QEMU_OPTION_xen_domid:
5652 xen_domid = atoi(optarg);
5653 break;
5654 case QEMU_OPTION_xen_create:
5655 xen_mode = XEN_CREATE;
5656 break;
5657 case QEMU_OPTION_xen_attach:
5658 xen_mode = XEN_ATTACH;
5659 break;
5660 #endif
5661 case QEMU_OPTION_readconfig:
5662 {
5663 FILE *fp;
5664 fp = fopen(optarg, "r");
5665 if (fp == NULL) {
5666 fprintf(stderr, "open %s: %s\n", optarg, strerror(errno));
5667 exit(1);
5668 }
5669 if (qemu_config_parse(fp, optarg) != 0) {
5670 exit(1);
5671 }
5672 fclose(fp);
5673 break;
5674 }
5675 case QEMU_OPTION_writeconfig:
5676 {
5677 FILE *fp;
5678 if (strcmp(optarg, "-") == 0) {
5679 fp = stdout;
5680 } else {
5681 fp = fopen(optarg, "w");
5682 if (fp == NULL) {
5683 fprintf(stderr, "open %s: %s\n", optarg, strerror(errno));
5684 exit(1);
5685 }
5686 }
5687 qemu_config_write(fp);
5688 fclose(fp);
5689 break;
5690 }
5691 }
5692 }
5693 }
5694 loc_set_none();
5695
5696 /* If no data_dir is specified then try to find it relative to the
5697 executable path. */
5698 if (!data_dir) {
5699 data_dir = find_datadir(argv[0]);
5700 }
5701 /* If all else fails use the install patch specified when building. */
5702 if (!data_dir) {
5703 data_dir = CONFIG_QEMU_SHAREDIR;
5704 }
5705
5706 /*
5707 * Default to max_cpus = smp_cpus, in case the user doesn't
5708 * specify a max_cpus value.
5709 */
5710 if (!max_cpus)
5711 max_cpus = smp_cpus;
5712
5713 machine->max_cpus = machine->max_cpus ?: 1; /* Default to UP */
5714 if (smp_cpus > machine->max_cpus) {
5715 fprintf(stderr, "Number of SMP cpus requested (%d), exceeds max cpus "
5716 "supported by machine `%s' (%d)\n", smp_cpus, machine->name,
5717 machine->max_cpus);
5718 exit(1);
5719 }
5720
5721 qemu_opts_foreach(&qemu_device_opts, default_driver_check, NULL, 0);
5722 qemu_opts_foreach(&qemu_global_opts, default_driver_check, NULL, 0);
5723
5724 if (machine->no_serial) {
5725 default_serial = 0;
5726 }
5727 if (machine->no_parallel) {
5728 default_parallel = 0;
5729 }
5730 if (!machine->use_virtcon) {
5731 default_virtcon = 0;
5732 }
5733 if (machine->no_vga) {
5734 default_vga = 0;
5735 }
5736 if (machine->no_floppy) {
5737 default_floppy = 0;
5738 }
5739 if (machine->no_cdrom) {
5740 default_cdrom = 0;
5741 }
5742 if (machine->no_sdcard) {
5743 default_sdcard = 0;
5744 }
5745
5746 if (display_type == DT_NOGRAPHIC) {
5747 if (default_parallel)
5748 add_device_config(DEV_PARALLEL, "null");
5749 if (default_serial && default_monitor) {
5750 add_device_config(DEV_SERIAL, "mon:stdio");
5751 } else if (default_virtcon && default_monitor) {
5752 add_device_config(DEV_VIRTCON, "mon:stdio");
5753 } else {
5754 if (default_serial)
5755 add_device_config(DEV_SERIAL, "stdio");
5756 if (default_virtcon)
5757 add_device_config(DEV_VIRTCON, "stdio");
5758 if (default_monitor)
5759 monitor_parse("stdio", "readline");
5760 }
5761 } else {
5762 if (default_serial)
5763 add_device_config(DEV_SERIAL, "vc:80Cx24C");
5764 if (default_parallel)
5765 add_device_config(DEV_PARALLEL, "vc:80Cx24C");
5766 if (default_monitor)
5767 monitor_parse("vc:80Cx24C", "readline");
5768 if (default_virtcon)
5769 add_device_config(DEV_VIRTCON, "vc:80Cx24C");
5770 }
5771 if (default_vga)
5772 vga_interface_type = VGA_CIRRUS;
5773
5774 if (qemu_opts_foreach(&qemu_chardev_opts, chardev_init_func, NULL, 1) != 0)
5775 exit(1);
5776
5777 #ifndef _WIN32
5778 if (daemonize) {
5779 pid_t pid;
5780
5781 if (pipe(fds) == -1)
5782 exit(1);
5783
5784 pid = fork();
5785 if (pid > 0) {
5786 uint8_t status;
5787 ssize_t len;
5788
5789 close(fds[1]);
5790
5791 again:
5792 len = read(fds[0], &status, 1);
5793 if (len == -1 && (errno == EINTR))
5794 goto again;
5795
5796 if (len != 1)
5797 exit(1);
5798 else if (status == 1) {
5799 fprintf(stderr, "Could not acquire pidfile: %s\n", strerror(errno));
5800 exit(1);
5801 } else
5802 exit(0);
5803 } else if (pid < 0)
5804 exit(1);
5805
5806 close(fds[0]);
5807 qemu_set_cloexec(fds[1]);
5808
5809 setsid();
5810
5811 pid = fork();
5812 if (pid > 0)
5813 exit(0);
5814 else if (pid < 0)
5815 exit(1);
5816
5817 umask(027);
5818
5819 signal(SIGTSTP, SIG_IGN);
5820 signal(SIGTTOU, SIG_IGN);
5821 signal(SIGTTIN, SIG_IGN);
5822 }
5823 #endif
5824
5825 if (pid_file && qemu_create_pidfile(pid_file) != 0) {
5826 #ifndef _WIN32
5827 if (daemonize) {
5828 uint8_t status = 1;
5829 if (write(fds[1], &status, 1) != 1) {
5830 perror("daemonize. Writing to pipe\n");
5831 }
5832 } else
5833 #endif
5834 fprintf(stderr, "Could not acquire pid file: %s\n", strerror(errno));
5835 exit(1);
5836 }
5837
5838 if (kvm_enabled()) {
5839 int ret;
5840
5841 ret = kvm_init(smp_cpus);
5842 if (ret < 0) {
5843 fprintf(stderr, "failed to initialize KVM\n");
5844 exit(1);
5845 }
5846 }
5847
5848 if (qemu_init_main_loop()) {
5849 fprintf(stderr, "qemu_init_main_loop failed\n");
5850 exit(1);
5851 }
5852 linux_boot = (kernel_filename != NULL);
5853
5854 if (!linux_boot && *kernel_cmdline != '\0') {
5855 fprintf(stderr, "-append only allowed with -kernel option\n");
5856 exit(1);
5857 }
5858
5859 if (!linux_boot && initrd_filename != NULL) {
5860 fprintf(stderr, "-initrd only allowed with -kernel option\n");
5861 exit(1);
5862 }
5863
5864 #ifndef _WIN32
5865 /* Win32 doesn't support line-buffering and requires size >= 2 */
5866 setvbuf(stdout, NULL, _IOLBF, 0);
5867 #endif
5868
5869 if (init_timer_alarm() < 0) {
5870 fprintf(stderr, "could not initialize alarm timer\n");
5871 exit(1);
5872 }
5873 if (use_icount && icount_time_shift < 0) {
5874 use_icount = 2;
5875 /* 125MIPS seems a reasonable initial guess at the guest speed.
5876 It will be corrected fairly quickly anyway. */
5877 icount_time_shift = 3;
5878 init_icount_adjust();
5879 }
5880
5881 #ifdef _WIN32
5882 socket_init();
5883 #endif
5884
5885 if (net_init_clients() < 0) {
5886 exit(1);
5887 }
5888
5889 net_boot = (boot_devices_bitmap >> ('n' - 'a')) & 0xF;
5890 net_set_boot_mask(net_boot);
5891
5892 /* init the bluetooth world */
5893 if (foreach_device_config(DEV_BT, bt_parse))
5894 exit(1);
5895
5896 /* init the memory */
5897 if (ram_size == 0)
5898 ram_size = DEFAULT_RAM_SIZE * 1024 * 1024;
5899
5900 /* init the dynamic translator */
5901 cpu_exec_init_all(tb_size * 1024 * 1024);
5902
5903 bdrv_init_with_whitelist();
5904
5905 blk_mig_init();
5906
5907 if (default_cdrom) {
5908 /* we always create the cdrom drive, even if no disk is there */
5909 drive_add(NULL, CDROM_ALIAS);
5910 }
5911
5912 if (default_floppy) {
5913 /* we always create at least one floppy */
5914 drive_add(NULL, FD_ALIAS, 0);
5915 }
5916
5917 if (default_sdcard) {
5918 /* we always create one sd slot, even if no card is in it */
5919 drive_add(NULL, SD_ALIAS);
5920 }
5921
5922 /* open the virtual block devices */
5923 if (snapshot)
5924 qemu_opts_foreach(&qemu_drive_opts, drive_enable_snapshot, NULL, 0);
5925 if (qemu_opts_foreach(&qemu_drive_opts, drive_init_func, machine, 1) != 0)
5926 exit(1);
5927
5928 vmstate_register(0, &vmstate_timers ,&timers_state);
5929 register_savevm_live("ram", 0, 3, NULL, ram_save_live, NULL,
5930 ram_load, NULL);
5931
5932 if (nb_numa_nodes > 0) {
5933 int i;
5934
5935 if (nb_numa_nodes > smp_cpus) {
5936 nb_numa_nodes = smp_cpus;
5937 }
5938
5939 /* If no memory size if given for any node, assume the default case
5940 * and distribute the available memory equally across all nodes
5941 */
5942 for (i = 0; i < nb_numa_nodes; i++) {
5943 if (node_mem[i] != 0)
5944 break;
5945 }
5946 if (i == nb_numa_nodes) {
5947 uint64_t usedmem = 0;
5948
5949 /* On Linux, the each node's border has to be 8MB aligned,
5950 * the final node gets the rest.
5951 */
5952 for (i = 0; i < nb_numa_nodes - 1; i++) {
5953 node_mem[i] = (ram_size / nb_numa_nodes) & ~((1 << 23UL) - 1);
5954 usedmem += node_mem[i];
5955 }
5956 node_mem[i] = ram_size - usedmem;
5957 }
5958
5959 for (i = 0; i < nb_numa_nodes; i++) {
5960 if (node_cpumask[i] != 0)
5961 break;
5962 }
5963 /* assigning the VCPUs round-robin is easier to implement, guest OSes
5964 * must cope with this anyway, because there are BIOSes out there in
5965 * real machines which also use this scheme.
5966 */
5967 if (i == nb_numa_nodes) {
5968 for (i = 0; i < smp_cpus; i++) {
5969 node_cpumask[i % nb_numa_nodes] |= 1 << i;
5970 }
5971 }
5972 }
5973
5974 if (foreach_device_config(DEV_SERIAL, serial_parse) < 0)
5975 exit(1);
5976 if (foreach_device_config(DEV_PARALLEL, parallel_parse) < 0)
5977 exit(1);
5978 if (foreach_device_config(DEV_VIRTCON, virtcon_parse) < 0)
5979 exit(1);
5980 if (foreach_device_config(DEV_DEBUGCON, debugcon_parse) < 0)
5981 exit(1);
5982
5983 module_call_init(MODULE_INIT_DEVICE);
5984
5985 if (qemu_opts_foreach(&qemu_device_opts, device_help_func, NULL, 0) != 0)
5986 exit(0);
5987
5988 if (watchdog) {
5989 i = select_watchdog(watchdog);
5990 if (i > 0)
5991 exit (i == 1 ? 1 : 0);
5992 }
5993
5994 if (machine->compat_props) {
5995 qdev_prop_register_global_list(machine->compat_props);
5996 }
5997 qemu_add_globals();
5998
5999 machine->init(ram_size, boot_devices,
6000 kernel_filename, kernel_cmdline, initrd_filename, cpu_model);
6001
6002 cpu_synchronize_all_post_init();
6003
6004 #ifndef _WIN32
6005 /* must be after terminal init, SDL library changes signal handlers */
6006 sighandler_setup();
6007 #endif
6008
6009 for (env = first_cpu; env != NULL; env = env->next_cpu) {
6010 for (i = 0; i < nb_numa_nodes; i++) {
6011 if (node_cpumask[i] & (1 << env->cpu_index)) {
6012 env->numa_node = i;
6013 }
6014 }
6015 }
6016
6017 current_machine = machine;
6018
6019 /* init USB devices */
6020 if (usb_enabled) {
6021 if (foreach_device_config(DEV_USB, usb_parse) < 0)
6022 exit(1);
6023 }
6024
6025 /* init generic devices */
6026 if (qemu_opts_foreach(&qemu_device_opts, device_init_func, NULL, 1) != 0)
6027 exit(1);
6028
6029 net_check_clients();
6030
6031 /* just use the first displaystate for the moment */
6032 ds = get_displaystate();
6033
6034 if (display_type == DT_DEFAULT) {
6035 #if defined(CONFIG_SDL) || defined(CONFIG_COCOA)
6036 display_type = DT_SDL;
6037 #else
6038 display_type = DT_VNC;
6039 vnc_display = "localhost:0,to=99";
6040 show_vnc_port = 1;
6041 #endif
6042 }
6043
6044
6045 switch (display_type) {
6046 case DT_NOGRAPHIC:
6047 break;
6048 #if defined(CONFIG_CURSES)
6049 case DT_CURSES:
6050 curses_display_init(ds, full_screen);
6051 break;
6052 #endif
6053 #if defined(CONFIG_SDL)
6054 case DT_SDL:
6055 sdl_display_init(ds, full_screen, no_frame);
6056 break;
6057 #elif defined(CONFIG_COCOA)
6058 case DT_SDL:
6059 cocoa_display_init(ds, full_screen);
6060 break;
6061 #endif
6062 case DT_VNC:
6063 vnc_display_init(ds);
6064 if (vnc_display_open(ds, vnc_display) < 0)
6065 exit(1);
6066
6067 if (show_vnc_port) {
6068 printf("VNC server running on `%s'\n", vnc_display_local_addr(ds));
6069 }
6070 break;
6071 default:
6072 break;
6073 }
6074 dpy_resize(ds);
6075
6076 dcl = ds->listeners;
6077 while (dcl != NULL) {
6078 if (dcl->dpy_refresh != NULL) {
6079 ds->gui_timer = qemu_new_timer(rt_clock, gui_update, ds);
6080 qemu_mod_timer(ds->gui_timer, qemu_get_clock(rt_clock));
6081 }
6082 dcl = dcl->next;
6083 }
6084
6085 if (display_type == DT_NOGRAPHIC || display_type == DT_VNC) {
6086 nographic_timer = qemu_new_timer(rt_clock, nographic_update, NULL);
6087 qemu_mod_timer(nographic_timer, qemu_get_clock(rt_clock));
6088 }
6089
6090 text_consoles_set_display(ds);
6091
6092 if (qemu_opts_foreach(&qemu_mon_opts, mon_init_func, NULL, 1) != 0)
6093 exit(1);
6094
6095 if (gdbstub_dev && gdbserver_start(gdbstub_dev) < 0) {
6096 fprintf(stderr, "qemu: could not open gdbserver on device '%s'\n",
6097 gdbstub_dev);
6098 exit(1);
6099 }
6100
6101 qdev_machine_creation_done();
6102
6103 if (rom_load_all() != 0) {
6104 fprintf(stderr, "rom loading failed\n");
6105 exit(1);
6106 }
6107
6108 qemu_system_reset();
6109 if (loadvm) {
6110 if (load_vmstate(loadvm) < 0) {
6111 autostart = 0;
6112 }
6113 }
6114
6115 if (incoming) {
6116 qemu_start_incoming_migration(incoming);
6117 } else if (autostart) {
6118 vm_start();
6119 }
6120
6121 #ifndef _WIN32
6122 if (daemonize) {
6123 uint8_t status = 0;
6124 ssize_t len;
6125
6126 again1:
6127 len = write(fds[1], &status, 1);
6128 if (len == -1 && (errno == EINTR))
6129 goto again1;
6130
6131 if (len != 1)
6132 exit(1);
6133
6134 if (chdir("/")) {
6135 perror("not able to chdir to /");
6136 exit(1);
6137 }
6138 TFR(fd = qemu_open("/dev/null", O_RDWR));
6139 if (fd == -1)
6140 exit(1);
6141 }
6142
6143 if (run_as) {
6144 pwd = getpwnam(run_as);
6145 if (!pwd) {
6146 fprintf(stderr, "User \"%s\" doesn't exist\n", run_as);
6147 exit(1);
6148 }
6149 }
6150
6151 if (chroot_dir) {
6152 if (chroot(chroot_dir) < 0) {
6153 fprintf(stderr, "chroot failed\n");
6154 exit(1);
6155 }
6156 if (chdir("/")) {
6157 perror("not able to chdir to /");
6158 exit(1);
6159 }
6160 }
6161
6162 if (run_as) {
6163 if (setgid(pwd->pw_gid) < 0) {
6164 fprintf(stderr, "Failed to setgid(%d)\n", pwd->pw_gid);
6165 exit(1);
6166 }
6167 if (setuid(pwd->pw_uid) < 0) {
6168 fprintf(stderr, "Failed to setuid(%d)\n", pwd->pw_uid);
6169 exit(1);
6170 }
6171 if (setuid(0) != -1) {
6172 fprintf(stderr, "Dropping privileges failed\n");
6173 exit(1);
6174 }
6175 }
6176
6177 if (daemonize) {
6178 dup2(fd, 0);
6179 dup2(fd, 1);
6180 dup2(fd, 2);
6181
6182 close(fd);
6183 }
6184 #endif
6185
6186 main_loop();
6187 quit_timers();
6188 net_cleanup();
6189
6190 return 0;
6191 }
AMD IOMMU emulation for KVM