Instead of writing a zero page, madvise it away
[qemu/aliguori-queue.git] / vl.c
blobca26450e305fab21c9bc1c3334a0a93ca183b3ad
1 /*
2 * QEMU System Emulator
4 * Copyright (c) 2003-2008 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
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>
32 /* Needed early for HOST_BSD etc. */
33 #include "config-host.h"
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 #if defined(__NetBSD__)
48 #include <net/if_tap.h>
49 #endif
50 #ifdef __linux__
51 #include <linux/if_tun.h>
52 #endif
53 #include <arpa/inet.h>
54 #include <dirent.h>
55 #include <netdb.h>
56 #include <sys/select.h>
57 #ifdef HOST_BSD
58 #include <sys/stat.h>
59 #if defined(__FreeBSD__) || defined(__DragonFly__)
60 #include <libutil.h>
61 #else
62 #include <util.h>
63 #endif
64 #elif defined (__GLIBC__) && defined (__FreeBSD_kernel__)
65 #include <freebsd/stdlib.h>
66 #else
67 #ifdef __linux__
68 #include <pty.h>
69 #include <malloc.h>
70 #include <linux/rtc.h>
72 /* For the benefit of older linux systems which don't supply it,
73 we use a local copy of hpet.h. */
74 /* #include <linux/hpet.h> */
75 #include "hpet.h"
77 #include <linux/ppdev.h>
78 #include <linux/parport.h>
79 #endif
80 #ifdef __sun__
81 #include <sys/stat.h>
82 #include <sys/ethernet.h>
83 #include <sys/sockio.h>
84 #include <netinet/arp.h>
85 #include <netinet/in.h>
86 #include <netinet/in_systm.h>
87 #include <netinet/ip.h>
88 #include <netinet/ip_icmp.h> // must come after ip.h
89 #include <netinet/udp.h>
90 #include <netinet/tcp.h>
91 #include <net/if.h>
92 #include <syslog.h>
93 #include <stropts.h>
94 #endif
95 #endif
96 #endif
98 #if defined(__OpenBSD__)
99 #include <util.h>
100 #endif
102 #if defined(CONFIG_VDE)
103 #include <libvdeplug.h>
104 #endif
106 #ifdef _WIN32
107 #include <windows.h>
108 #include <malloc.h>
109 #include <sys/timeb.h>
110 #include <mmsystem.h>
111 #define getopt_long_only getopt_long
112 #define memalign(align, size) malloc(size)
113 #endif
115 #ifdef CONFIG_SDL
116 #ifdef __APPLE__
117 #include <SDL.h>
118 int qemu_main(int argc, char **argv, char **envp);
119 int main(int argc, char **argv)
121 qemu_main(argc, argv, NULL);
123 #undef main
124 #define main qemu_main
125 #endif
126 #endif /* CONFIG_SDL */
128 #ifdef CONFIG_COCOA
129 #undef main
130 #define main qemu_main
131 #endif /* CONFIG_COCOA */
133 #include "hw/hw.h"
134 #include "hw/boards.h"
135 #include "hw/usb.h"
136 #include "hw/pcmcia.h"
137 #include "hw/pc.h"
138 #include "hw/audiodev.h"
139 #include "hw/isa.h"
140 #include "hw/baum.h"
141 #include "hw/bt.h"
142 #include "hw/watchdog.h"
143 #include "hw/smbios.h"
144 #include "hw/xen.h"
145 #include "bt-host.h"
146 #include "net.h"
147 #include "monitor.h"
148 #include "console.h"
149 #include "sysemu.h"
150 #include "gdbstub.h"
151 #include "qemu-timer.h"
152 #include "qemu-char.h"
153 #include "cache-utils.h"
154 #include "block.h"
155 #include "dma.h"
156 #include "audio/audio.h"
157 #include "migration.h"
158 #include "kvm.h"
159 #include "balloon.h"
160 #include "qemu-option.h"
162 #include "disas.h"
164 #include "exec-all.h"
166 #include "qemu_socket.h"
168 #if defined(CONFIG_SLIRP)
169 #include "libslirp.h"
170 #endif
172 //#define DEBUG_UNUSED_IOPORT
173 //#define DEBUG_IOPORT
174 //#define DEBUG_NET
175 //#define DEBUG_SLIRP
178 #ifdef DEBUG_IOPORT
179 # define LOG_IOPORT(...) qemu_log_mask(CPU_LOG_IOPORT, ## __VA_ARGS__)
180 #else
181 # define LOG_IOPORT(...) do { } while (0)
182 #endif
184 #define DEFAULT_RAM_SIZE 128
186 /* Max number of USB devices that can be specified on the commandline. */
187 #define MAX_USB_CMDLINE 8
189 /* Max number of bluetooth switches on the commandline. */
190 #define MAX_BT_CMDLINE 10
192 /* XXX: use a two level table to limit memory usage */
193 #define MAX_IOPORTS 65536
195 static const char *data_dir;
196 const char *bios_name = NULL;
197 static void *ioport_opaque[MAX_IOPORTS];
198 static IOPortReadFunc *ioport_read_table[3][MAX_IOPORTS];
199 static IOPortWriteFunc *ioport_write_table[3][MAX_IOPORTS];
200 /* Note: drives_table[MAX_DRIVES] is a dummy block driver if none available
201 to store the VM snapshots */
202 DriveInfo drives_table[MAX_DRIVES+1];
203 int nb_drives;
204 enum vga_retrace_method vga_retrace_method = VGA_RETRACE_DUMB;
205 static DisplayState *display_state;
206 DisplayType display_type = DT_DEFAULT;
207 const char* keyboard_layout = NULL;
208 int64_t ticks_per_sec;
209 ram_addr_t ram_size;
210 int nb_nics;
211 NICInfo nd_table[MAX_NICS];
212 int vm_running;
213 static int autostart;
214 static int rtc_utc = 1;
215 static int rtc_date_offset = -1; /* -1 means no change */
216 int cirrus_vga_enabled = 1;
217 int std_vga_enabled = 0;
218 int vmsvga_enabled = 0;
219 int xenfb_enabled = 0;
220 #ifdef TARGET_SPARC
221 int graphic_width = 1024;
222 int graphic_height = 768;
223 int graphic_depth = 8;
224 #else
225 int graphic_width = 800;
226 int graphic_height = 600;
227 int graphic_depth = 15;
228 #endif
229 static int full_screen = 0;
230 #ifdef CONFIG_SDL
231 static int no_frame = 0;
232 #endif
233 int no_quit = 0;
234 CharDriverState *serial_hds[MAX_SERIAL_PORTS];
235 CharDriverState *parallel_hds[MAX_PARALLEL_PORTS];
236 CharDriverState *virtcon_hds[MAX_VIRTIO_CONSOLES];
237 #ifdef TARGET_I386
238 int win2k_install_hack = 0;
239 int rtc_td_hack = 0;
240 #endif
241 int usb_enabled = 0;
242 int singlestep = 0;
243 int smp_cpus = 1;
244 const char *vnc_display;
245 int acpi_enabled = 1;
246 int no_hpet = 0;
247 int no_virtio_balloon = 0;
248 int fd_bootchk = 1;
249 int no_reboot = 0;
250 int no_shutdown = 0;
251 int cursor_hide = 1;
252 int graphic_rotate = 0;
253 #ifndef _WIN32
254 int daemonize = 0;
255 #endif
256 WatchdogTimerModel *watchdog = NULL;
257 int watchdog_action = WDT_RESET;
258 const char *option_rom[MAX_OPTION_ROMS];
259 int nb_option_roms;
260 int semihosting_enabled = 0;
261 #ifdef TARGET_ARM
262 int old_param = 0;
263 #endif
264 const char *qemu_name;
265 int alt_grab = 0;
266 #if defined(TARGET_SPARC) || defined(TARGET_PPC)
267 unsigned int nb_prom_envs = 0;
268 const char *prom_envs[MAX_PROM_ENVS];
269 #endif
270 int nb_drives_opt;
271 struct drive_opt drives_opt[MAX_DRIVES];
273 int nb_numa_nodes;
274 uint64_t node_mem[MAX_NODES];
275 uint64_t node_cpumask[MAX_NODES];
277 static CPUState *cur_cpu;
278 static CPUState *next_cpu;
279 static int timer_alarm_pending = 1;
280 /* Conversion factor from emulated instructions to virtual clock ticks. */
281 static int icount_time_shift;
282 /* Arbitrarily pick 1MIPS as the minimum allowable speed. */
283 #define MAX_ICOUNT_SHIFT 10
284 /* Compensate for varying guest execution speed. */
285 static int64_t qemu_icount_bias;
286 static QEMUTimer *icount_rt_timer;
287 static QEMUTimer *icount_vm_timer;
288 static QEMUTimer *nographic_timer;
290 uint8_t qemu_uuid[16];
292 /***********************************************************/
293 /* x86 ISA bus support */
295 target_phys_addr_t isa_mem_base = 0;
296 PicState2 *isa_pic;
298 static IOPortReadFunc default_ioport_readb, default_ioport_readw, default_ioport_readl;
299 static IOPortWriteFunc default_ioport_writeb, default_ioport_writew, default_ioport_writel;
301 static uint32_t ioport_read(int index, uint32_t address)
303 static IOPortReadFunc *default_func[3] = {
304 default_ioport_readb,
305 default_ioport_readw,
306 default_ioport_readl
308 IOPortReadFunc *func = ioport_read_table[index][address];
309 if (!func)
310 func = default_func[index];
311 return func(ioport_opaque[address], address);
314 static void ioport_write(int index, uint32_t address, uint32_t data)
316 static IOPortWriteFunc *default_func[3] = {
317 default_ioport_writeb,
318 default_ioport_writew,
319 default_ioport_writel
321 IOPortWriteFunc *func = ioport_write_table[index][address];
322 if (!func)
323 func = default_func[index];
324 func(ioport_opaque[address], address, data);
327 static uint32_t default_ioport_readb(void *opaque, uint32_t address)
329 #ifdef DEBUG_UNUSED_IOPORT
330 fprintf(stderr, "unused inb: port=0x%04x\n", address);
331 #endif
332 return 0xff;
335 static void default_ioport_writeb(void *opaque, uint32_t address, uint32_t data)
337 #ifdef DEBUG_UNUSED_IOPORT
338 fprintf(stderr, "unused outb: port=0x%04x data=0x%02x\n", address, data);
339 #endif
342 /* default is to make two byte accesses */
343 static uint32_t default_ioport_readw(void *opaque, uint32_t address)
345 uint32_t data;
346 data = ioport_read(0, address);
347 address = (address + 1) & (MAX_IOPORTS - 1);
348 data |= ioport_read(0, address) << 8;
349 return data;
352 static void default_ioport_writew(void *opaque, uint32_t address, uint32_t data)
354 ioport_write(0, address, data & 0xff);
355 address = (address + 1) & (MAX_IOPORTS - 1);
356 ioport_write(0, address, (data >> 8) & 0xff);
359 static uint32_t default_ioport_readl(void *opaque, uint32_t address)
361 #ifdef DEBUG_UNUSED_IOPORT
362 fprintf(stderr, "unused inl: port=0x%04x\n", address);
363 #endif
364 return 0xffffffff;
367 static void default_ioport_writel(void *opaque, uint32_t address, uint32_t data)
369 #ifdef DEBUG_UNUSED_IOPORT
370 fprintf(stderr, "unused outl: port=0x%04x data=0x%02x\n", address, data);
371 #endif
374 /* size is the word size in byte */
375 int register_ioport_read(int start, int length, int size,
376 IOPortReadFunc *func, void *opaque)
378 int i, bsize;
380 if (size == 1) {
381 bsize = 0;
382 } else if (size == 2) {
383 bsize = 1;
384 } else if (size == 4) {
385 bsize = 2;
386 } else {
387 hw_error("register_ioport_read: invalid size");
388 return -1;
390 for(i = start; i < start + length; i += size) {
391 ioport_read_table[bsize][i] = func;
392 if (ioport_opaque[i] != NULL && ioport_opaque[i] != opaque)
393 hw_error("register_ioport_read: invalid opaque");
394 ioport_opaque[i] = opaque;
396 return 0;
399 /* size is the word size in byte */
400 int register_ioport_write(int start, int length, int size,
401 IOPortWriteFunc *func, void *opaque)
403 int i, bsize;
405 if (size == 1) {
406 bsize = 0;
407 } else if (size == 2) {
408 bsize = 1;
409 } else if (size == 4) {
410 bsize = 2;
411 } else {
412 hw_error("register_ioport_write: invalid size");
413 return -1;
415 for(i = start; i < start + length; i += size) {
416 ioport_write_table[bsize][i] = func;
417 if (ioport_opaque[i] != NULL && ioport_opaque[i] != opaque)
418 hw_error("register_ioport_write: invalid opaque");
419 ioport_opaque[i] = opaque;
421 return 0;
424 void isa_unassign_ioport(int start, int length)
426 int i;
428 for(i = start; i < start + length; i++) {
429 ioport_read_table[0][i] = default_ioport_readb;
430 ioport_read_table[1][i] = default_ioport_readw;
431 ioport_read_table[2][i] = default_ioport_readl;
433 ioport_write_table[0][i] = default_ioport_writeb;
434 ioport_write_table[1][i] = default_ioport_writew;
435 ioport_write_table[2][i] = default_ioport_writel;
437 ioport_opaque[i] = NULL;
441 /***********************************************************/
443 void cpu_outb(CPUState *env, int addr, int val)
445 LOG_IOPORT("outb: %04x %02x\n", addr, val);
446 ioport_write(0, addr, val);
447 #ifdef CONFIG_KQEMU
448 if (env)
449 env->last_io_time = cpu_get_time_fast();
450 #endif
453 void cpu_outw(CPUState *env, int addr, int val)
455 LOG_IOPORT("outw: %04x %04x\n", addr, val);
456 ioport_write(1, addr, val);
457 #ifdef CONFIG_KQEMU
458 if (env)
459 env->last_io_time = cpu_get_time_fast();
460 #endif
463 void cpu_outl(CPUState *env, int addr, int val)
465 LOG_IOPORT("outl: %04x %08x\n", addr, val);
466 ioport_write(2, addr, val);
467 #ifdef CONFIG_KQEMU
468 if (env)
469 env->last_io_time = cpu_get_time_fast();
470 #endif
473 int cpu_inb(CPUState *env, int addr)
475 int val;
476 val = ioport_read(0, addr);
477 LOG_IOPORT("inb : %04x %02x\n", addr, val);
478 #ifdef CONFIG_KQEMU
479 if (env)
480 env->last_io_time = cpu_get_time_fast();
481 #endif
482 return val;
485 int cpu_inw(CPUState *env, int addr)
487 int val;
488 val = ioport_read(1, addr);
489 LOG_IOPORT("inw : %04x %04x\n", addr, val);
490 #ifdef CONFIG_KQEMU
491 if (env)
492 env->last_io_time = cpu_get_time_fast();
493 #endif
494 return val;
497 int cpu_inl(CPUState *env, int addr)
499 int val;
500 val = ioport_read(2, addr);
501 LOG_IOPORT("inl : %04x %08x\n", addr, val);
502 #ifdef CONFIG_KQEMU
503 if (env)
504 env->last_io_time = cpu_get_time_fast();
505 #endif
506 return val;
509 /***********************************************************/
510 void hw_error(const char *fmt, ...)
512 va_list ap;
513 CPUState *env;
515 va_start(ap, fmt);
516 fprintf(stderr, "qemu: hardware error: ");
517 vfprintf(stderr, fmt, ap);
518 fprintf(stderr, "\n");
519 for(env = first_cpu; env != NULL; env = env->next_cpu) {
520 fprintf(stderr, "CPU #%d:\n", env->cpu_index);
521 #ifdef TARGET_I386
522 cpu_dump_state(env, stderr, fprintf, X86_DUMP_FPU);
523 #else
524 cpu_dump_state(env, stderr, fprintf, 0);
525 #endif
527 va_end(ap);
528 abort();
531 /***************/
532 /* ballooning */
534 static QEMUBalloonEvent *qemu_balloon_event;
535 void *qemu_balloon_event_opaque;
537 void qemu_add_balloon_handler(QEMUBalloonEvent *func, void *opaque)
539 qemu_balloon_event = func;
540 qemu_balloon_event_opaque = opaque;
543 void qemu_balloon(ram_addr_t target)
545 if (qemu_balloon_event)
546 qemu_balloon_event(qemu_balloon_event_opaque, target);
549 ram_addr_t qemu_balloon_status(void)
551 if (qemu_balloon_event)
552 return qemu_balloon_event(qemu_balloon_event_opaque, 0);
553 return 0;
556 /***********************************************************/
557 /* keyboard/mouse */
559 static QEMUPutKBDEvent *qemu_put_kbd_event;
560 static void *qemu_put_kbd_event_opaque;
561 static QEMUPutMouseEntry *qemu_put_mouse_event_head;
562 static QEMUPutMouseEntry *qemu_put_mouse_event_current;
564 void qemu_add_kbd_event_handler(QEMUPutKBDEvent *func, void *opaque)
566 qemu_put_kbd_event_opaque = opaque;
567 qemu_put_kbd_event = func;
570 QEMUPutMouseEntry *qemu_add_mouse_event_handler(QEMUPutMouseEvent *func,
571 void *opaque, int absolute,
572 const char *name)
574 QEMUPutMouseEntry *s, *cursor;
576 s = qemu_mallocz(sizeof(QEMUPutMouseEntry));
578 s->qemu_put_mouse_event = func;
579 s->qemu_put_mouse_event_opaque = opaque;
580 s->qemu_put_mouse_event_absolute = absolute;
581 s->qemu_put_mouse_event_name = qemu_strdup(name);
582 s->next = NULL;
584 if (!qemu_put_mouse_event_head) {
585 qemu_put_mouse_event_head = qemu_put_mouse_event_current = s;
586 return s;
589 cursor = qemu_put_mouse_event_head;
590 while (cursor->next != NULL)
591 cursor = cursor->next;
593 cursor->next = s;
594 qemu_put_mouse_event_current = s;
596 return s;
599 void qemu_remove_mouse_event_handler(QEMUPutMouseEntry *entry)
601 QEMUPutMouseEntry *prev = NULL, *cursor;
603 if (!qemu_put_mouse_event_head || entry == NULL)
604 return;
606 cursor = qemu_put_mouse_event_head;
607 while (cursor != NULL && cursor != entry) {
608 prev = cursor;
609 cursor = cursor->next;
612 if (cursor == NULL) // does not exist or list empty
613 return;
614 else if (prev == NULL) { // entry is head
615 qemu_put_mouse_event_head = cursor->next;
616 if (qemu_put_mouse_event_current == entry)
617 qemu_put_mouse_event_current = cursor->next;
618 qemu_free(entry->qemu_put_mouse_event_name);
619 qemu_free(entry);
620 return;
623 prev->next = entry->next;
625 if (qemu_put_mouse_event_current == entry)
626 qemu_put_mouse_event_current = prev;
628 qemu_free(entry->qemu_put_mouse_event_name);
629 qemu_free(entry);
632 void kbd_put_keycode(int keycode)
634 if (qemu_put_kbd_event) {
635 qemu_put_kbd_event(qemu_put_kbd_event_opaque, keycode);
639 void kbd_mouse_event(int dx, int dy, int dz, int buttons_state)
641 QEMUPutMouseEvent *mouse_event;
642 void *mouse_event_opaque;
643 int width;
645 if (!qemu_put_mouse_event_current) {
646 return;
649 mouse_event =
650 qemu_put_mouse_event_current->qemu_put_mouse_event;
651 mouse_event_opaque =
652 qemu_put_mouse_event_current->qemu_put_mouse_event_opaque;
654 if (mouse_event) {
655 if (graphic_rotate) {
656 if (qemu_put_mouse_event_current->qemu_put_mouse_event_absolute)
657 width = 0x7fff;
658 else
659 width = graphic_width - 1;
660 mouse_event(mouse_event_opaque,
661 width - dy, dx, dz, buttons_state);
662 } else
663 mouse_event(mouse_event_opaque,
664 dx, dy, dz, buttons_state);
668 int kbd_mouse_is_absolute(void)
670 if (!qemu_put_mouse_event_current)
671 return 0;
673 return qemu_put_mouse_event_current->qemu_put_mouse_event_absolute;
676 void do_info_mice(Monitor *mon)
678 QEMUPutMouseEntry *cursor;
679 int index = 0;
681 if (!qemu_put_mouse_event_head) {
682 monitor_printf(mon, "No mouse devices connected\n");
683 return;
686 monitor_printf(mon, "Mouse devices available:\n");
687 cursor = qemu_put_mouse_event_head;
688 while (cursor != NULL) {
689 monitor_printf(mon, "%c Mouse #%d: %s\n",
690 (cursor == qemu_put_mouse_event_current ? '*' : ' '),
691 index, cursor->qemu_put_mouse_event_name);
692 index++;
693 cursor = cursor->next;
697 void do_mouse_set(Monitor *mon, int index)
699 QEMUPutMouseEntry *cursor;
700 int i = 0;
702 if (!qemu_put_mouse_event_head) {
703 monitor_printf(mon, "No mouse devices connected\n");
704 return;
707 cursor = qemu_put_mouse_event_head;
708 while (cursor != NULL && index != i) {
709 i++;
710 cursor = cursor->next;
713 if (cursor != NULL)
714 qemu_put_mouse_event_current = cursor;
715 else
716 monitor_printf(mon, "Mouse at given index not found\n");
719 /* compute with 96 bit intermediate result: (a*b)/c */
720 uint64_t muldiv64(uint64_t a, uint32_t b, uint32_t c)
722 union {
723 uint64_t ll;
724 struct {
725 #ifdef WORDS_BIGENDIAN
726 uint32_t high, low;
727 #else
728 uint32_t low, high;
729 #endif
730 } l;
731 } u, res;
732 uint64_t rl, rh;
734 u.ll = a;
735 rl = (uint64_t)u.l.low * (uint64_t)b;
736 rh = (uint64_t)u.l.high * (uint64_t)b;
737 rh += (rl >> 32);
738 res.l.high = rh / c;
739 res.l.low = (((rh % c) << 32) + (rl & 0xffffffff)) / c;
740 return res.ll;
743 /***********************************************************/
744 /* real time host monotonic timer */
746 #define QEMU_TIMER_BASE 1000000000LL
748 #ifdef WIN32
750 static int64_t clock_freq;
752 static void init_get_clock(void)
754 LARGE_INTEGER freq;
755 int ret;
756 ret = QueryPerformanceFrequency(&freq);
757 if (ret == 0) {
758 fprintf(stderr, "Could not calibrate ticks\n");
759 exit(1);
761 clock_freq = freq.QuadPart;
764 static int64_t get_clock(void)
766 LARGE_INTEGER ti;
767 QueryPerformanceCounter(&ti);
768 return muldiv64(ti.QuadPart, QEMU_TIMER_BASE, clock_freq);
771 #else
773 static int use_rt_clock;
775 static void init_get_clock(void)
777 use_rt_clock = 0;
778 #if defined(__linux__) || (defined(__FreeBSD__) && __FreeBSD_version >= 500000) \
779 || defined(__DragonFly__)
781 struct timespec ts;
782 if (clock_gettime(CLOCK_MONOTONIC, &ts) == 0) {
783 use_rt_clock = 1;
786 #endif
789 static int64_t get_clock(void)
791 #if defined(__linux__) || (defined(__FreeBSD__) && __FreeBSD_version >= 500000) \
792 || defined(__DragonFly__)
793 if (use_rt_clock) {
794 struct timespec ts;
795 clock_gettime(CLOCK_MONOTONIC, &ts);
796 return ts.tv_sec * 1000000000LL + ts.tv_nsec;
797 } else
798 #endif
800 /* XXX: using gettimeofday leads to problems if the date
801 changes, so it should be avoided. */
802 struct timeval tv;
803 gettimeofday(&tv, NULL);
804 return tv.tv_sec * 1000000000LL + (tv.tv_usec * 1000);
807 #endif
809 /* Return the virtual CPU time, based on the instruction counter. */
810 static int64_t cpu_get_icount(void)
812 int64_t icount;
813 CPUState *env = cpu_single_env;;
814 icount = qemu_icount;
815 if (env) {
816 if (!can_do_io(env))
817 fprintf(stderr, "Bad clock read\n");
818 icount -= (env->icount_decr.u16.low + env->icount_extra);
820 return qemu_icount_bias + (icount << icount_time_shift);
823 /***********************************************************/
824 /* guest cycle counter */
826 static int64_t cpu_ticks_prev;
827 static int64_t cpu_ticks_offset;
828 static int64_t cpu_clock_offset;
829 static int cpu_ticks_enabled;
831 /* return the host CPU cycle counter and handle stop/restart */
832 int64_t cpu_get_ticks(void)
834 if (use_icount) {
835 return cpu_get_icount();
837 if (!cpu_ticks_enabled) {
838 return cpu_ticks_offset;
839 } else {
840 int64_t ticks;
841 ticks = cpu_get_real_ticks();
842 if (cpu_ticks_prev > ticks) {
843 /* Note: non increasing ticks may happen if the host uses
844 software suspend */
845 cpu_ticks_offset += cpu_ticks_prev - ticks;
847 cpu_ticks_prev = ticks;
848 return ticks + cpu_ticks_offset;
852 /* return the host CPU monotonic timer and handle stop/restart */
853 static int64_t cpu_get_clock(void)
855 int64_t ti;
856 if (!cpu_ticks_enabled) {
857 return cpu_clock_offset;
858 } else {
859 ti = get_clock();
860 return ti + cpu_clock_offset;
864 /* enable cpu_get_ticks() */
865 void cpu_enable_ticks(void)
867 if (!cpu_ticks_enabled) {
868 cpu_ticks_offset -= cpu_get_real_ticks();
869 cpu_clock_offset -= get_clock();
870 cpu_ticks_enabled = 1;
874 /* disable cpu_get_ticks() : the clock is stopped. You must not call
875 cpu_get_ticks() after that. */
876 void cpu_disable_ticks(void)
878 if (cpu_ticks_enabled) {
879 cpu_ticks_offset = cpu_get_ticks();
880 cpu_clock_offset = cpu_get_clock();
881 cpu_ticks_enabled = 0;
885 /***********************************************************/
886 /* timers */
888 #define QEMU_TIMER_REALTIME 0
889 #define QEMU_TIMER_VIRTUAL 1
891 struct QEMUClock {
892 int type;
893 /* XXX: add frequency */
896 struct QEMUTimer {
897 QEMUClock *clock;
898 int64_t expire_time;
899 QEMUTimerCB *cb;
900 void *opaque;
901 struct QEMUTimer *next;
904 struct qemu_alarm_timer {
905 char const *name;
906 unsigned int flags;
908 int (*start)(struct qemu_alarm_timer *t);
909 void (*stop)(struct qemu_alarm_timer *t);
910 void (*rearm)(struct qemu_alarm_timer *t);
911 void *priv;
914 #define ALARM_FLAG_DYNTICKS 0x1
915 #define ALARM_FLAG_EXPIRED 0x2
917 static inline int alarm_has_dynticks(struct qemu_alarm_timer *t)
919 return t && (t->flags & ALARM_FLAG_DYNTICKS);
922 static void qemu_rearm_alarm_timer(struct qemu_alarm_timer *t)
924 if (!alarm_has_dynticks(t))
925 return;
927 t->rearm(t);
930 /* TODO: MIN_TIMER_REARM_US should be optimized */
931 #define MIN_TIMER_REARM_US 250
933 static struct qemu_alarm_timer *alarm_timer;
935 #ifdef _WIN32
937 struct qemu_alarm_win32 {
938 MMRESULT timerId;
939 unsigned int period;
940 } alarm_win32_data = {0, -1};
942 static int win32_start_timer(struct qemu_alarm_timer *t);
943 static void win32_stop_timer(struct qemu_alarm_timer *t);
944 static void win32_rearm_timer(struct qemu_alarm_timer *t);
946 #else
948 static int unix_start_timer(struct qemu_alarm_timer *t);
949 static void unix_stop_timer(struct qemu_alarm_timer *t);
951 #ifdef __linux__
953 static int dynticks_start_timer(struct qemu_alarm_timer *t);
954 static void dynticks_stop_timer(struct qemu_alarm_timer *t);
955 static void dynticks_rearm_timer(struct qemu_alarm_timer *t);
957 static int hpet_start_timer(struct qemu_alarm_timer *t);
958 static void hpet_stop_timer(struct qemu_alarm_timer *t);
960 static int rtc_start_timer(struct qemu_alarm_timer *t);
961 static void rtc_stop_timer(struct qemu_alarm_timer *t);
963 #endif /* __linux__ */
965 #endif /* _WIN32 */
967 /* Correlation between real and virtual time is always going to be
968 fairly approximate, so ignore small variation.
969 When the guest is idle real and virtual time will be aligned in
970 the IO wait loop. */
971 #define ICOUNT_WOBBLE (QEMU_TIMER_BASE / 10)
973 static void icount_adjust(void)
975 int64_t cur_time;
976 int64_t cur_icount;
977 int64_t delta;
978 static int64_t last_delta;
979 /* If the VM is not running, then do nothing. */
980 if (!vm_running)
981 return;
983 cur_time = cpu_get_clock();
984 cur_icount = qemu_get_clock(vm_clock);
985 delta = cur_icount - cur_time;
986 /* FIXME: This is a very crude algorithm, somewhat prone to oscillation. */
987 if (delta > 0
988 && last_delta + ICOUNT_WOBBLE < delta * 2
989 && icount_time_shift > 0) {
990 /* The guest is getting too far ahead. Slow time down. */
991 icount_time_shift--;
993 if (delta < 0
994 && last_delta - ICOUNT_WOBBLE > delta * 2
995 && icount_time_shift < MAX_ICOUNT_SHIFT) {
996 /* The guest is getting too far behind. Speed time up. */
997 icount_time_shift++;
999 last_delta = delta;
1000 qemu_icount_bias = cur_icount - (qemu_icount << icount_time_shift);
1003 static void icount_adjust_rt(void * opaque)
1005 qemu_mod_timer(icount_rt_timer,
1006 qemu_get_clock(rt_clock) + 1000);
1007 icount_adjust();
1010 static void icount_adjust_vm(void * opaque)
1012 qemu_mod_timer(icount_vm_timer,
1013 qemu_get_clock(vm_clock) + QEMU_TIMER_BASE / 10);
1014 icount_adjust();
1017 static void init_icount_adjust(void)
1019 /* Have both realtime and virtual time triggers for speed adjustment.
1020 The realtime trigger catches emulated time passing too slowly,
1021 the virtual time trigger catches emulated time passing too fast.
1022 Realtime triggers occur even when idle, so use them less frequently
1023 than VM triggers. */
1024 icount_rt_timer = qemu_new_timer(rt_clock, icount_adjust_rt, NULL);
1025 qemu_mod_timer(icount_rt_timer,
1026 qemu_get_clock(rt_clock) + 1000);
1027 icount_vm_timer = qemu_new_timer(vm_clock, icount_adjust_vm, NULL);
1028 qemu_mod_timer(icount_vm_timer,
1029 qemu_get_clock(vm_clock) + QEMU_TIMER_BASE / 10);
1032 static struct qemu_alarm_timer alarm_timers[] = {
1033 #ifndef _WIN32
1034 #ifdef __linux__
1035 {"dynticks", ALARM_FLAG_DYNTICKS, dynticks_start_timer,
1036 dynticks_stop_timer, dynticks_rearm_timer, NULL},
1037 /* HPET - if available - is preferred */
1038 {"hpet", 0, hpet_start_timer, hpet_stop_timer, NULL, NULL},
1039 /* ...otherwise try RTC */
1040 {"rtc", 0, rtc_start_timer, rtc_stop_timer, NULL, NULL},
1041 #endif
1042 {"unix", 0, unix_start_timer, unix_stop_timer, NULL, NULL},
1043 #else
1044 {"dynticks", ALARM_FLAG_DYNTICKS, win32_start_timer,
1045 win32_stop_timer, win32_rearm_timer, &alarm_win32_data},
1046 {"win32", 0, win32_start_timer,
1047 win32_stop_timer, NULL, &alarm_win32_data},
1048 #endif
1049 {NULL, }
1052 static void show_available_alarms(void)
1054 int i;
1056 printf("Available alarm timers, in order of precedence:\n");
1057 for (i = 0; alarm_timers[i].name; i++)
1058 printf("%s\n", alarm_timers[i].name);
1061 static void configure_alarms(char const *opt)
1063 int i;
1064 int cur = 0;
1065 int count = ARRAY_SIZE(alarm_timers) - 1;
1066 char *arg;
1067 char *name;
1068 struct qemu_alarm_timer tmp;
1070 if (!strcmp(opt, "?")) {
1071 show_available_alarms();
1072 exit(0);
1075 arg = strdup(opt);
1077 /* Reorder the array */
1078 name = strtok(arg, ",");
1079 while (name) {
1080 for (i = 0; i < count && alarm_timers[i].name; i++) {
1081 if (!strcmp(alarm_timers[i].name, name))
1082 break;
1085 if (i == count) {
1086 fprintf(stderr, "Unknown clock %s\n", name);
1087 goto next;
1090 if (i < cur)
1091 /* Ignore */
1092 goto next;
1094 /* Swap */
1095 tmp = alarm_timers[i];
1096 alarm_timers[i] = alarm_timers[cur];
1097 alarm_timers[cur] = tmp;
1099 cur++;
1100 next:
1101 name = strtok(NULL, ",");
1104 free(arg);
1106 if (cur) {
1107 /* Disable remaining timers */
1108 for (i = cur; i < count; i++)
1109 alarm_timers[i].name = NULL;
1110 } else {
1111 show_available_alarms();
1112 exit(1);
1116 QEMUClock *rt_clock;
1117 QEMUClock *vm_clock;
1119 static QEMUTimer *active_timers[2];
1121 static QEMUClock *qemu_new_clock(int type)
1123 QEMUClock *clock;
1124 clock = qemu_mallocz(sizeof(QEMUClock));
1125 clock->type = type;
1126 return clock;
1129 QEMUTimer *qemu_new_timer(QEMUClock *clock, QEMUTimerCB *cb, void *opaque)
1131 QEMUTimer *ts;
1133 ts = qemu_mallocz(sizeof(QEMUTimer));
1134 ts->clock = clock;
1135 ts->cb = cb;
1136 ts->opaque = opaque;
1137 return ts;
1140 void qemu_free_timer(QEMUTimer *ts)
1142 qemu_free(ts);
1145 /* stop a timer, but do not dealloc it */
1146 void qemu_del_timer(QEMUTimer *ts)
1148 QEMUTimer **pt, *t;
1150 /* NOTE: this code must be signal safe because
1151 qemu_timer_expired() can be called from a signal. */
1152 pt = &active_timers[ts->clock->type];
1153 for(;;) {
1154 t = *pt;
1155 if (!t)
1156 break;
1157 if (t == ts) {
1158 *pt = t->next;
1159 break;
1161 pt = &t->next;
1165 /* modify the current timer so that it will be fired when current_time
1166 >= expire_time. The corresponding callback will be called. */
1167 void qemu_mod_timer(QEMUTimer *ts, int64_t expire_time)
1169 QEMUTimer **pt, *t;
1171 qemu_del_timer(ts);
1173 /* add the timer in the sorted list */
1174 /* NOTE: this code must be signal safe because
1175 qemu_timer_expired() can be called from a signal. */
1176 pt = &active_timers[ts->clock->type];
1177 for(;;) {
1178 t = *pt;
1179 if (!t)
1180 break;
1181 if (t->expire_time > expire_time)
1182 break;
1183 pt = &t->next;
1185 ts->expire_time = expire_time;
1186 ts->next = *pt;
1187 *pt = ts;
1189 /* Rearm if necessary */
1190 if (pt == &active_timers[ts->clock->type]) {
1191 if ((alarm_timer->flags & ALARM_FLAG_EXPIRED) == 0) {
1192 qemu_rearm_alarm_timer(alarm_timer);
1194 /* Interrupt execution to force deadline recalculation. */
1195 if (use_icount)
1196 qemu_notify_event();
1200 int qemu_timer_pending(QEMUTimer *ts)
1202 QEMUTimer *t;
1203 for(t = active_timers[ts->clock->type]; t != NULL; t = t->next) {
1204 if (t == ts)
1205 return 1;
1207 return 0;
1210 static inline int qemu_timer_expired(QEMUTimer *timer_head, int64_t current_time)
1212 if (!timer_head)
1213 return 0;
1214 return (timer_head->expire_time <= current_time);
1217 static void qemu_run_timers(QEMUTimer **ptimer_head, int64_t current_time)
1219 QEMUTimer *ts;
1221 for(;;) {
1222 ts = *ptimer_head;
1223 if (!ts || ts->expire_time > current_time)
1224 break;
1225 /* remove timer from the list before calling the callback */
1226 *ptimer_head = ts->next;
1227 ts->next = NULL;
1229 /* run the callback (the timer list can be modified) */
1230 ts->cb(ts->opaque);
1234 int64_t qemu_get_clock(QEMUClock *clock)
1236 switch(clock->type) {
1237 case QEMU_TIMER_REALTIME:
1238 return get_clock() / 1000000;
1239 default:
1240 case QEMU_TIMER_VIRTUAL:
1241 if (use_icount) {
1242 return cpu_get_icount();
1243 } else {
1244 return cpu_get_clock();
1249 static void init_timers(void)
1251 init_get_clock();
1252 ticks_per_sec = QEMU_TIMER_BASE;
1253 rt_clock = qemu_new_clock(QEMU_TIMER_REALTIME);
1254 vm_clock = qemu_new_clock(QEMU_TIMER_VIRTUAL);
1257 /* save a timer */
1258 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
1260 uint64_t expire_time;
1262 if (qemu_timer_pending(ts)) {
1263 expire_time = ts->expire_time;
1264 } else {
1265 expire_time = -1;
1267 qemu_put_be64(f, expire_time);
1270 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
1272 uint64_t expire_time;
1274 expire_time = qemu_get_be64(f);
1275 if (expire_time != -1) {
1276 qemu_mod_timer(ts, expire_time);
1277 } else {
1278 qemu_del_timer(ts);
1282 static void timer_save(QEMUFile *f, void *opaque)
1284 if (cpu_ticks_enabled) {
1285 hw_error("cannot save state if virtual timers are running");
1287 qemu_put_be64(f, cpu_ticks_offset);
1288 qemu_put_be64(f, ticks_per_sec);
1289 qemu_put_be64(f, cpu_clock_offset);
1292 static int timer_load(QEMUFile *f, void *opaque, int version_id)
1294 if (version_id != 1 && version_id != 2)
1295 return -EINVAL;
1296 if (cpu_ticks_enabled) {
1297 return -EINVAL;
1299 cpu_ticks_offset=qemu_get_be64(f);
1300 ticks_per_sec=qemu_get_be64(f);
1301 if (version_id == 2) {
1302 cpu_clock_offset=qemu_get_be64(f);
1304 return 0;
1307 static void qemu_event_increment(void);
1309 #ifdef _WIN32
1310 static void CALLBACK host_alarm_handler(UINT uTimerID, UINT uMsg,
1311 DWORD_PTR dwUser, DWORD_PTR dw1,
1312 DWORD_PTR dw2)
1313 #else
1314 static void host_alarm_handler(int host_signum)
1315 #endif
1317 #if 0
1318 #define DISP_FREQ 1000
1320 static int64_t delta_min = INT64_MAX;
1321 static int64_t delta_max, delta_cum, last_clock, delta, ti;
1322 static int count;
1323 ti = qemu_get_clock(vm_clock);
1324 if (last_clock != 0) {
1325 delta = ti - last_clock;
1326 if (delta < delta_min)
1327 delta_min = delta;
1328 if (delta > delta_max)
1329 delta_max = delta;
1330 delta_cum += delta;
1331 if (++count == DISP_FREQ) {
1332 printf("timer: min=%" PRId64 " us max=%" PRId64 " us avg=%" PRId64 " us avg_freq=%0.3f Hz\n",
1333 muldiv64(delta_min, 1000000, ticks_per_sec),
1334 muldiv64(delta_max, 1000000, ticks_per_sec),
1335 muldiv64(delta_cum, 1000000 / DISP_FREQ, ticks_per_sec),
1336 (double)ticks_per_sec / ((double)delta_cum / DISP_FREQ));
1337 count = 0;
1338 delta_min = INT64_MAX;
1339 delta_max = 0;
1340 delta_cum = 0;
1343 last_clock = ti;
1345 #endif
1346 if (alarm_has_dynticks(alarm_timer) ||
1347 (!use_icount &&
1348 qemu_timer_expired(active_timers[QEMU_TIMER_VIRTUAL],
1349 qemu_get_clock(vm_clock))) ||
1350 qemu_timer_expired(active_timers[QEMU_TIMER_REALTIME],
1351 qemu_get_clock(rt_clock))) {
1352 qemu_event_increment();
1353 if (alarm_timer) alarm_timer->flags |= ALARM_FLAG_EXPIRED;
1355 #ifndef CONFIG_IOTHREAD
1356 if (next_cpu) {
1357 /* stop the currently executing cpu because a timer occured */
1358 cpu_exit(next_cpu);
1359 #ifdef CONFIG_KQEMU
1360 if (next_cpu->kqemu_enabled) {
1361 kqemu_cpu_interrupt(next_cpu);
1363 #endif
1365 #endif
1366 timer_alarm_pending = 1;
1367 qemu_notify_event();
1371 static int64_t qemu_next_deadline(void)
1373 int64_t delta;
1375 if (active_timers[QEMU_TIMER_VIRTUAL]) {
1376 delta = active_timers[QEMU_TIMER_VIRTUAL]->expire_time -
1377 qemu_get_clock(vm_clock);
1378 } else {
1379 /* To avoid problems with overflow limit this to 2^32. */
1380 delta = INT32_MAX;
1383 if (delta < 0)
1384 delta = 0;
1386 return delta;
1389 #if defined(__linux__) || defined(_WIN32)
1390 static uint64_t qemu_next_deadline_dyntick(void)
1392 int64_t delta;
1393 int64_t rtdelta;
1395 if (use_icount)
1396 delta = INT32_MAX;
1397 else
1398 delta = (qemu_next_deadline() + 999) / 1000;
1400 if (active_timers[QEMU_TIMER_REALTIME]) {
1401 rtdelta = (active_timers[QEMU_TIMER_REALTIME]->expire_time -
1402 qemu_get_clock(rt_clock))*1000;
1403 if (rtdelta < delta)
1404 delta = rtdelta;
1407 if (delta < MIN_TIMER_REARM_US)
1408 delta = MIN_TIMER_REARM_US;
1410 return delta;
1412 #endif
1414 #ifndef _WIN32
1416 /* Sets a specific flag */
1417 static int fcntl_setfl(int fd, int flag)
1419 int flags;
1421 flags = fcntl(fd, F_GETFL);
1422 if (flags == -1)
1423 return -errno;
1425 if (fcntl(fd, F_SETFL, flags | flag) == -1)
1426 return -errno;
1428 return 0;
1431 #if defined(__linux__)
1433 #define RTC_FREQ 1024
1435 static void enable_sigio_timer(int fd)
1437 struct sigaction act;
1439 /* timer signal */
1440 sigfillset(&act.sa_mask);
1441 act.sa_flags = 0;
1442 act.sa_handler = host_alarm_handler;
1444 sigaction(SIGIO, &act, NULL);
1445 fcntl_setfl(fd, O_ASYNC);
1446 fcntl(fd, F_SETOWN, getpid());
1449 static int hpet_start_timer(struct qemu_alarm_timer *t)
1451 struct hpet_info info;
1452 int r, fd;
1454 fd = open("/dev/hpet", O_RDONLY);
1455 if (fd < 0)
1456 return -1;
1458 /* Set frequency */
1459 r = ioctl(fd, HPET_IRQFREQ, RTC_FREQ);
1460 if (r < 0) {
1461 fprintf(stderr, "Could not configure '/dev/hpet' to have a 1024Hz timer. This is not a fatal\n"
1462 "error, but for better emulation accuracy type:\n"
1463 "'echo 1024 > /proc/sys/dev/hpet/max-user-freq' as root.\n");
1464 goto fail;
1467 /* Check capabilities */
1468 r = ioctl(fd, HPET_INFO, &info);
1469 if (r < 0)
1470 goto fail;
1472 /* Enable periodic mode */
1473 r = ioctl(fd, HPET_EPI, 0);
1474 if (info.hi_flags && (r < 0))
1475 goto fail;
1477 /* Enable interrupt */
1478 r = ioctl(fd, HPET_IE_ON, 0);
1479 if (r < 0)
1480 goto fail;
1482 enable_sigio_timer(fd);
1483 t->priv = (void *)(long)fd;
1485 return 0;
1486 fail:
1487 close(fd);
1488 return -1;
1491 static void hpet_stop_timer(struct qemu_alarm_timer *t)
1493 int fd = (long)t->priv;
1495 close(fd);
1498 static int rtc_start_timer(struct qemu_alarm_timer *t)
1500 int rtc_fd;
1501 unsigned long current_rtc_freq = 0;
1503 TFR(rtc_fd = open("/dev/rtc", O_RDONLY));
1504 if (rtc_fd < 0)
1505 return -1;
1506 ioctl(rtc_fd, RTC_IRQP_READ, &current_rtc_freq);
1507 if (current_rtc_freq != RTC_FREQ &&
1508 ioctl(rtc_fd, RTC_IRQP_SET, RTC_FREQ) < 0) {
1509 fprintf(stderr, "Could not configure '/dev/rtc' to have a 1024 Hz timer. This is not a fatal\n"
1510 "error, but for better emulation accuracy either use a 2.6 host Linux kernel or\n"
1511 "type 'echo 1024 > /proc/sys/dev/rtc/max-user-freq' as root.\n");
1512 goto fail;
1514 if (ioctl(rtc_fd, RTC_PIE_ON, 0) < 0) {
1515 fail:
1516 close(rtc_fd);
1517 return -1;
1520 enable_sigio_timer(rtc_fd);
1522 t->priv = (void *)(long)rtc_fd;
1524 return 0;
1527 static void rtc_stop_timer(struct qemu_alarm_timer *t)
1529 int rtc_fd = (long)t->priv;
1531 close(rtc_fd);
1534 static int dynticks_start_timer(struct qemu_alarm_timer *t)
1536 struct sigevent ev;
1537 timer_t host_timer;
1538 struct sigaction act;
1540 sigfillset(&act.sa_mask);
1541 act.sa_flags = 0;
1542 act.sa_handler = host_alarm_handler;
1544 sigaction(SIGALRM, &act, NULL);
1547 * Initialize ev struct to 0 to avoid valgrind complaining
1548 * about uninitialized data in timer_create call
1550 memset(&ev, 0, sizeof(ev));
1551 ev.sigev_value.sival_int = 0;
1552 ev.sigev_notify = SIGEV_SIGNAL;
1553 ev.sigev_signo = SIGALRM;
1555 if (timer_create(CLOCK_REALTIME, &ev, &host_timer)) {
1556 perror("timer_create");
1558 /* disable dynticks */
1559 fprintf(stderr, "Dynamic Ticks disabled\n");
1561 return -1;
1564 t->priv = (void *)(long)host_timer;
1566 return 0;
1569 static void dynticks_stop_timer(struct qemu_alarm_timer *t)
1571 timer_t host_timer = (timer_t)(long)t->priv;
1573 timer_delete(host_timer);
1576 static void dynticks_rearm_timer(struct qemu_alarm_timer *t)
1578 timer_t host_timer = (timer_t)(long)t->priv;
1579 struct itimerspec timeout;
1580 int64_t nearest_delta_us = INT64_MAX;
1581 int64_t current_us;
1583 if (!active_timers[QEMU_TIMER_REALTIME] &&
1584 !active_timers[QEMU_TIMER_VIRTUAL])
1585 return;
1587 nearest_delta_us = qemu_next_deadline_dyntick();
1589 /* check whether a timer is already running */
1590 if (timer_gettime(host_timer, &timeout)) {
1591 perror("gettime");
1592 fprintf(stderr, "Internal timer error: aborting\n");
1593 exit(1);
1595 current_us = timeout.it_value.tv_sec * 1000000 + timeout.it_value.tv_nsec/1000;
1596 if (current_us && current_us <= nearest_delta_us)
1597 return;
1599 timeout.it_interval.tv_sec = 0;
1600 timeout.it_interval.tv_nsec = 0; /* 0 for one-shot timer */
1601 timeout.it_value.tv_sec = nearest_delta_us / 1000000;
1602 timeout.it_value.tv_nsec = (nearest_delta_us % 1000000) * 1000;
1603 if (timer_settime(host_timer, 0 /* RELATIVE */, &timeout, NULL)) {
1604 perror("settime");
1605 fprintf(stderr, "Internal timer error: aborting\n");
1606 exit(1);
1610 #endif /* defined(__linux__) */
1612 static int unix_start_timer(struct qemu_alarm_timer *t)
1614 struct sigaction act;
1615 struct itimerval itv;
1616 int err;
1618 /* timer signal */
1619 sigfillset(&act.sa_mask);
1620 act.sa_flags = 0;
1621 act.sa_handler = host_alarm_handler;
1623 sigaction(SIGALRM, &act, NULL);
1625 itv.it_interval.tv_sec = 0;
1626 /* for i386 kernel 2.6 to get 1 ms */
1627 itv.it_interval.tv_usec = 999;
1628 itv.it_value.tv_sec = 0;
1629 itv.it_value.tv_usec = 10 * 1000;
1631 err = setitimer(ITIMER_REAL, &itv, NULL);
1632 if (err)
1633 return -1;
1635 return 0;
1638 static void unix_stop_timer(struct qemu_alarm_timer *t)
1640 struct itimerval itv;
1642 memset(&itv, 0, sizeof(itv));
1643 setitimer(ITIMER_REAL, &itv, NULL);
1646 #endif /* !defined(_WIN32) */
1649 #ifdef _WIN32
1651 static int win32_start_timer(struct qemu_alarm_timer *t)
1653 TIMECAPS tc;
1654 struct qemu_alarm_win32 *data = t->priv;
1655 UINT flags;
1657 memset(&tc, 0, sizeof(tc));
1658 timeGetDevCaps(&tc, sizeof(tc));
1660 if (data->period < tc.wPeriodMin)
1661 data->period = tc.wPeriodMin;
1663 timeBeginPeriod(data->period);
1665 flags = TIME_CALLBACK_FUNCTION;
1666 if (alarm_has_dynticks(t))
1667 flags |= TIME_ONESHOT;
1668 else
1669 flags |= TIME_PERIODIC;
1671 data->timerId = timeSetEvent(1, // interval (ms)
1672 data->period, // resolution
1673 host_alarm_handler, // function
1674 (DWORD)t, // parameter
1675 flags);
1677 if (!data->timerId) {
1678 perror("Failed to initialize win32 alarm timer");
1679 timeEndPeriod(data->period);
1680 return -1;
1683 return 0;
1686 static void win32_stop_timer(struct qemu_alarm_timer *t)
1688 struct qemu_alarm_win32 *data = t->priv;
1690 timeKillEvent(data->timerId);
1691 timeEndPeriod(data->period);
1694 static void win32_rearm_timer(struct qemu_alarm_timer *t)
1696 struct qemu_alarm_win32 *data = t->priv;
1697 uint64_t nearest_delta_us;
1699 if (!active_timers[QEMU_TIMER_REALTIME] &&
1700 !active_timers[QEMU_TIMER_VIRTUAL])
1701 return;
1703 nearest_delta_us = qemu_next_deadline_dyntick();
1704 nearest_delta_us /= 1000;
1706 timeKillEvent(data->timerId);
1708 data->timerId = timeSetEvent(1,
1709 data->period,
1710 host_alarm_handler,
1711 (DWORD)t,
1712 TIME_ONESHOT | TIME_PERIODIC);
1714 if (!data->timerId) {
1715 perror("Failed to re-arm win32 alarm timer");
1717 timeEndPeriod(data->period);
1718 exit(1);
1722 #endif /* _WIN32 */
1724 static int init_timer_alarm(void)
1726 struct qemu_alarm_timer *t = NULL;
1727 int i, err = -1;
1729 for (i = 0; alarm_timers[i].name; i++) {
1730 t = &alarm_timers[i];
1732 err = t->start(t);
1733 if (!err)
1734 break;
1737 if (err) {
1738 err = -ENOENT;
1739 goto fail;
1742 alarm_timer = t;
1744 return 0;
1746 fail:
1747 return err;
1750 static void quit_timers(void)
1752 alarm_timer->stop(alarm_timer);
1753 alarm_timer = NULL;
1756 /***********************************************************/
1757 /* host time/date access */
1758 void qemu_get_timedate(struct tm *tm, int offset)
1760 time_t ti;
1761 struct tm *ret;
1763 time(&ti);
1764 ti += offset;
1765 if (rtc_date_offset == -1) {
1766 if (rtc_utc)
1767 ret = gmtime(&ti);
1768 else
1769 ret = localtime(&ti);
1770 } else {
1771 ti -= rtc_date_offset;
1772 ret = gmtime(&ti);
1775 memcpy(tm, ret, sizeof(struct tm));
1778 int qemu_timedate_diff(struct tm *tm)
1780 time_t seconds;
1782 if (rtc_date_offset == -1)
1783 if (rtc_utc)
1784 seconds = mktimegm(tm);
1785 else
1786 seconds = mktime(tm);
1787 else
1788 seconds = mktimegm(tm) + rtc_date_offset;
1790 return seconds - time(NULL);
1793 #ifdef _WIN32
1794 static void socket_cleanup(void)
1796 WSACleanup();
1799 static int socket_init(void)
1801 WSADATA Data;
1802 int ret, err;
1804 ret = WSAStartup(MAKEWORD(2,2), &Data);
1805 if (ret != 0) {
1806 err = WSAGetLastError();
1807 fprintf(stderr, "WSAStartup: %d\n", err);
1808 return -1;
1810 atexit(socket_cleanup);
1811 return 0;
1813 #endif
1815 int get_param_value(char *buf, int buf_size,
1816 const char *tag, const char *str)
1818 const char *p;
1819 char option[128];
1821 p = str;
1822 for(;;) {
1823 p = get_opt_name(option, sizeof(option), p, '=');
1824 if (*p != '=')
1825 break;
1826 p++;
1827 if (!strcmp(tag, option)) {
1828 (void)get_opt_value(buf, buf_size, p);
1829 return strlen(buf);
1830 } else {
1831 p = get_opt_value(NULL, 0, p);
1833 if (*p != ',')
1834 break;
1835 p++;
1837 return 0;
1840 int check_params(char *buf, int buf_size,
1841 const char * const *params, const char *str)
1843 const char *p;
1844 int i;
1846 p = str;
1847 while (*p != '\0') {
1848 p = get_opt_name(buf, buf_size, p, '=');
1849 if (*p != '=') {
1850 return -1;
1852 p++;
1853 for (i = 0; params[i] != NULL; i++) {
1854 if (!strcmp(params[i], buf)) {
1855 break;
1858 if (params[i] == NULL) {
1859 return -1;
1861 p = get_opt_value(NULL, 0, p);
1862 if (*p != ',') {
1863 break;
1865 p++;
1867 return 0;
1870 /***********************************************************/
1871 /* Bluetooth support */
1872 static int nb_hcis;
1873 static int cur_hci;
1874 static struct HCIInfo *hci_table[MAX_NICS];
1876 static struct bt_vlan_s {
1877 struct bt_scatternet_s net;
1878 int id;
1879 struct bt_vlan_s *next;
1880 } *first_bt_vlan;
1882 /* find or alloc a new bluetooth "VLAN" */
1883 static struct bt_scatternet_s *qemu_find_bt_vlan(int id)
1885 struct bt_vlan_s **pvlan, *vlan;
1886 for (vlan = first_bt_vlan; vlan != NULL; vlan = vlan->next) {
1887 if (vlan->id == id)
1888 return &vlan->net;
1890 vlan = qemu_mallocz(sizeof(struct bt_vlan_s));
1891 vlan->id = id;
1892 pvlan = &first_bt_vlan;
1893 while (*pvlan != NULL)
1894 pvlan = &(*pvlan)->next;
1895 *pvlan = vlan;
1896 return &vlan->net;
1899 static void null_hci_send(struct HCIInfo *hci, const uint8_t *data, int len)
1903 static int null_hci_addr_set(struct HCIInfo *hci, const uint8_t *bd_addr)
1905 return -ENOTSUP;
1908 static struct HCIInfo null_hci = {
1909 .cmd_send = null_hci_send,
1910 .sco_send = null_hci_send,
1911 .acl_send = null_hci_send,
1912 .bdaddr_set = null_hci_addr_set,
1915 struct HCIInfo *qemu_next_hci(void)
1917 if (cur_hci == nb_hcis)
1918 return &null_hci;
1920 return hci_table[cur_hci++];
1923 static struct HCIInfo *hci_init(const char *str)
1925 char *endp;
1926 struct bt_scatternet_s *vlan = 0;
1928 if (!strcmp(str, "null"))
1929 /* null */
1930 return &null_hci;
1931 else if (!strncmp(str, "host", 4) && (str[4] == '\0' || str[4] == ':'))
1932 /* host[:hciN] */
1933 return bt_host_hci(str[4] ? str + 5 : "hci0");
1934 else if (!strncmp(str, "hci", 3)) {
1935 /* hci[,vlan=n] */
1936 if (str[3]) {
1937 if (!strncmp(str + 3, ",vlan=", 6)) {
1938 vlan = qemu_find_bt_vlan(strtol(str + 9, &endp, 0));
1939 if (*endp)
1940 vlan = 0;
1942 } else
1943 vlan = qemu_find_bt_vlan(0);
1944 if (vlan)
1945 return bt_new_hci(vlan);
1948 fprintf(stderr, "qemu: Unknown bluetooth HCI `%s'.\n", str);
1950 return 0;
1953 static int bt_hci_parse(const char *str)
1955 struct HCIInfo *hci;
1956 bdaddr_t bdaddr;
1958 if (nb_hcis >= MAX_NICS) {
1959 fprintf(stderr, "qemu: Too many bluetooth HCIs (max %i).\n", MAX_NICS);
1960 return -1;
1963 hci = hci_init(str);
1964 if (!hci)
1965 return -1;
1967 bdaddr.b[0] = 0x52;
1968 bdaddr.b[1] = 0x54;
1969 bdaddr.b[2] = 0x00;
1970 bdaddr.b[3] = 0x12;
1971 bdaddr.b[4] = 0x34;
1972 bdaddr.b[5] = 0x56 + nb_hcis;
1973 hci->bdaddr_set(hci, bdaddr.b);
1975 hci_table[nb_hcis++] = hci;
1977 return 0;
1980 static void bt_vhci_add(int vlan_id)
1982 struct bt_scatternet_s *vlan = qemu_find_bt_vlan(vlan_id);
1984 if (!vlan->slave)
1985 fprintf(stderr, "qemu: warning: adding a VHCI to "
1986 "an empty scatternet %i\n", vlan_id);
1988 bt_vhci_init(bt_new_hci(vlan));
1991 static struct bt_device_s *bt_device_add(const char *opt)
1993 struct bt_scatternet_s *vlan;
1994 int vlan_id = 0;
1995 char *endp = strstr(opt, ",vlan=");
1996 int len = (endp ? endp - opt : strlen(opt)) + 1;
1997 char devname[10];
1999 pstrcpy(devname, MIN(sizeof(devname), len), opt);
2001 if (endp) {
2002 vlan_id = strtol(endp + 6, &endp, 0);
2003 if (*endp) {
2004 fprintf(stderr, "qemu: unrecognised bluetooth vlan Id\n");
2005 return 0;
2009 vlan = qemu_find_bt_vlan(vlan_id);
2011 if (!vlan->slave)
2012 fprintf(stderr, "qemu: warning: adding a slave device to "
2013 "an empty scatternet %i\n", vlan_id);
2015 if (!strcmp(devname, "keyboard"))
2016 return bt_keyboard_init(vlan);
2018 fprintf(stderr, "qemu: unsupported bluetooth device `%s'\n", devname);
2019 return 0;
2022 static int bt_parse(const char *opt)
2024 const char *endp, *p;
2025 int vlan;
2027 if (strstart(opt, "hci", &endp)) {
2028 if (!*endp || *endp == ',') {
2029 if (*endp)
2030 if (!strstart(endp, ",vlan=", 0))
2031 opt = endp + 1;
2033 return bt_hci_parse(opt);
2035 } else if (strstart(opt, "vhci", &endp)) {
2036 if (!*endp || *endp == ',') {
2037 if (*endp) {
2038 if (strstart(endp, ",vlan=", &p)) {
2039 vlan = strtol(p, (char **) &endp, 0);
2040 if (*endp) {
2041 fprintf(stderr, "qemu: bad scatternet '%s'\n", p);
2042 return 1;
2044 } else {
2045 fprintf(stderr, "qemu: bad parameter '%s'\n", endp + 1);
2046 return 1;
2048 } else
2049 vlan = 0;
2051 bt_vhci_add(vlan);
2052 return 0;
2054 } else if (strstart(opt, "device:", &endp))
2055 return !bt_device_add(endp);
2057 fprintf(stderr, "qemu: bad bluetooth parameter '%s'\n", opt);
2058 return 1;
2061 /***********************************************************/
2062 /* QEMU Block devices */
2064 #define HD_ALIAS "index=%d,media=disk"
2065 #define CDROM_ALIAS "index=2,media=cdrom"
2066 #define FD_ALIAS "index=%d,if=floppy"
2067 #define PFLASH_ALIAS "if=pflash"
2068 #define MTD_ALIAS "if=mtd"
2069 #define SD_ALIAS "index=0,if=sd"
2071 static int drive_opt_get_free_idx(void)
2073 int index;
2075 for (index = 0; index < MAX_DRIVES; index++)
2076 if (!drives_opt[index].used) {
2077 drives_opt[index].used = 1;
2078 return index;
2081 return -1;
2084 static int drive_get_free_idx(void)
2086 int index;
2088 for (index = 0; index < MAX_DRIVES; index++)
2089 if (!drives_table[index].used) {
2090 drives_table[index].used = 1;
2091 return index;
2094 return -1;
2097 int drive_add(const char *file, const char *fmt, ...)
2099 va_list ap;
2100 int index = drive_opt_get_free_idx();
2102 if (nb_drives_opt >= MAX_DRIVES || index == -1) {
2103 fprintf(stderr, "qemu: too many drives\n");
2104 return -1;
2107 drives_opt[index].file = file;
2108 va_start(ap, fmt);
2109 vsnprintf(drives_opt[index].opt,
2110 sizeof(drives_opt[0].opt), fmt, ap);
2111 va_end(ap);
2113 nb_drives_opt++;
2114 return index;
2117 void drive_remove(int index)
2119 drives_opt[index].used = 0;
2120 nb_drives_opt--;
2123 int drive_get_index(BlockInterfaceType type, int bus, int unit)
2125 int index;
2127 /* seek interface, bus and unit */
2129 for (index = 0; index < MAX_DRIVES; index++)
2130 if (drives_table[index].type == type &&
2131 drives_table[index].bus == bus &&
2132 drives_table[index].unit == unit &&
2133 drives_table[index].used)
2134 return index;
2136 return -1;
2139 int drive_get_max_bus(BlockInterfaceType type)
2141 int max_bus;
2142 int index;
2144 max_bus = -1;
2145 for (index = 0; index < nb_drives; index++) {
2146 if(drives_table[index].type == type &&
2147 drives_table[index].bus > max_bus)
2148 max_bus = drives_table[index].bus;
2150 return max_bus;
2153 const char *drive_get_serial(BlockDriverState *bdrv)
2155 int index;
2157 for (index = 0; index < nb_drives; index++)
2158 if (drives_table[index].bdrv == bdrv)
2159 return drives_table[index].serial;
2161 return "\0";
2164 BlockInterfaceErrorAction drive_get_onerror(BlockDriverState *bdrv)
2166 int index;
2168 for (index = 0; index < nb_drives; index++)
2169 if (drives_table[index].bdrv == bdrv)
2170 return drives_table[index].onerror;
2172 return BLOCK_ERR_STOP_ENOSPC;
2175 static void bdrv_format_print(void *opaque, const char *name)
2177 fprintf(stderr, " %s", name);
2180 void drive_uninit(BlockDriverState *bdrv)
2182 int i;
2184 for (i = 0; i < MAX_DRIVES; i++)
2185 if (drives_table[i].bdrv == bdrv) {
2186 drives_table[i].bdrv = NULL;
2187 drives_table[i].used = 0;
2188 drive_remove(drives_table[i].drive_opt_idx);
2189 nb_drives--;
2190 break;
2194 int drive_init(struct drive_opt *arg, int snapshot, void *opaque)
2196 char buf[128];
2197 char file[1024];
2198 char devname[128];
2199 char serial[21];
2200 const char *mediastr = "";
2201 BlockInterfaceType type;
2202 enum { MEDIA_DISK, MEDIA_CDROM } media;
2203 int bus_id, unit_id;
2204 int cyls, heads, secs, translation;
2205 BlockDriverState *bdrv;
2206 BlockDriver *drv = NULL;
2207 QEMUMachine *machine = opaque;
2208 int max_devs;
2209 int index;
2210 int cache;
2211 int bdrv_flags, onerror;
2212 int drives_table_idx;
2213 char *str = arg->opt;
2214 static const char * const params[] = { "bus", "unit", "if", "index",
2215 "cyls", "heads", "secs", "trans",
2216 "media", "snapshot", "file",
2217 "cache", "format", "serial", "werror",
2218 NULL };
2220 if (check_params(buf, sizeof(buf), params, str) < 0) {
2221 fprintf(stderr, "qemu: unknown parameter '%s' in '%s'\n",
2222 buf, str);
2223 return -1;
2226 file[0] = 0;
2227 cyls = heads = secs = 0;
2228 bus_id = 0;
2229 unit_id = -1;
2230 translation = BIOS_ATA_TRANSLATION_AUTO;
2231 index = -1;
2232 cache = 3;
2234 if (machine->use_scsi) {
2235 type = IF_SCSI;
2236 max_devs = MAX_SCSI_DEVS;
2237 pstrcpy(devname, sizeof(devname), "scsi");
2238 } else {
2239 type = IF_IDE;
2240 max_devs = MAX_IDE_DEVS;
2241 pstrcpy(devname, sizeof(devname), "ide");
2243 media = MEDIA_DISK;
2245 /* extract parameters */
2247 if (get_param_value(buf, sizeof(buf), "bus", str)) {
2248 bus_id = strtol(buf, NULL, 0);
2249 if (bus_id < 0) {
2250 fprintf(stderr, "qemu: '%s' invalid bus id\n", str);
2251 return -1;
2255 if (get_param_value(buf, sizeof(buf), "unit", str)) {
2256 unit_id = strtol(buf, NULL, 0);
2257 if (unit_id < 0) {
2258 fprintf(stderr, "qemu: '%s' invalid unit id\n", str);
2259 return -1;
2263 if (get_param_value(buf, sizeof(buf), "if", str)) {
2264 pstrcpy(devname, sizeof(devname), buf);
2265 if (!strcmp(buf, "ide")) {
2266 type = IF_IDE;
2267 max_devs = MAX_IDE_DEVS;
2268 } else if (!strcmp(buf, "scsi")) {
2269 type = IF_SCSI;
2270 max_devs = MAX_SCSI_DEVS;
2271 } else if (!strcmp(buf, "floppy")) {
2272 type = IF_FLOPPY;
2273 max_devs = 0;
2274 } else if (!strcmp(buf, "pflash")) {
2275 type = IF_PFLASH;
2276 max_devs = 0;
2277 } else if (!strcmp(buf, "mtd")) {
2278 type = IF_MTD;
2279 max_devs = 0;
2280 } else if (!strcmp(buf, "sd")) {
2281 type = IF_SD;
2282 max_devs = 0;
2283 } else if (!strcmp(buf, "virtio")) {
2284 type = IF_VIRTIO;
2285 max_devs = 0;
2286 } else if (!strcmp(buf, "xen")) {
2287 type = IF_XEN;
2288 max_devs = 0;
2289 } else {
2290 fprintf(stderr, "qemu: '%s' unsupported bus type '%s'\n", str, buf);
2291 return -1;
2295 if (get_param_value(buf, sizeof(buf), "index", str)) {
2296 index = strtol(buf, NULL, 0);
2297 if (index < 0) {
2298 fprintf(stderr, "qemu: '%s' invalid index\n", str);
2299 return -1;
2303 if (get_param_value(buf, sizeof(buf), "cyls", str)) {
2304 cyls = strtol(buf, NULL, 0);
2307 if (get_param_value(buf, sizeof(buf), "heads", str)) {
2308 heads = strtol(buf, NULL, 0);
2311 if (get_param_value(buf, sizeof(buf), "secs", str)) {
2312 secs = strtol(buf, NULL, 0);
2315 if (cyls || heads || secs) {
2316 if (cyls < 1 || cyls > 16383) {
2317 fprintf(stderr, "qemu: '%s' invalid physical cyls number\n", str);
2318 return -1;
2320 if (heads < 1 || heads > 16) {
2321 fprintf(stderr, "qemu: '%s' invalid physical heads number\n", str);
2322 return -1;
2324 if (secs < 1 || secs > 63) {
2325 fprintf(stderr, "qemu: '%s' invalid physical secs number\n", str);
2326 return -1;
2330 if (get_param_value(buf, sizeof(buf), "trans", str)) {
2331 if (!cyls) {
2332 fprintf(stderr,
2333 "qemu: '%s' trans must be used with cyls,heads and secs\n",
2334 str);
2335 return -1;
2337 if (!strcmp(buf, "none"))
2338 translation = BIOS_ATA_TRANSLATION_NONE;
2339 else if (!strcmp(buf, "lba"))
2340 translation = BIOS_ATA_TRANSLATION_LBA;
2341 else if (!strcmp(buf, "auto"))
2342 translation = BIOS_ATA_TRANSLATION_AUTO;
2343 else {
2344 fprintf(stderr, "qemu: '%s' invalid translation type\n", str);
2345 return -1;
2349 if (get_param_value(buf, sizeof(buf), "media", str)) {
2350 if (!strcmp(buf, "disk")) {
2351 media = MEDIA_DISK;
2352 } else if (!strcmp(buf, "cdrom")) {
2353 if (cyls || secs || heads) {
2354 fprintf(stderr,
2355 "qemu: '%s' invalid physical CHS format\n", str);
2356 return -1;
2358 media = MEDIA_CDROM;
2359 } else {
2360 fprintf(stderr, "qemu: '%s' invalid media\n", str);
2361 return -1;
2365 if (get_param_value(buf, sizeof(buf), "snapshot", str)) {
2366 if (!strcmp(buf, "on"))
2367 snapshot = 1;
2368 else if (!strcmp(buf, "off"))
2369 snapshot = 0;
2370 else {
2371 fprintf(stderr, "qemu: '%s' invalid snapshot option\n", str);
2372 return -1;
2376 if (get_param_value(buf, sizeof(buf), "cache", str)) {
2377 if (!strcmp(buf, "off") || !strcmp(buf, "none"))
2378 cache = 0;
2379 else if (!strcmp(buf, "writethrough"))
2380 cache = 1;
2381 else if (!strcmp(buf, "writeback"))
2382 cache = 2;
2383 else {
2384 fprintf(stderr, "qemu: invalid cache option\n");
2385 return -1;
2389 if (get_param_value(buf, sizeof(buf), "format", str)) {
2390 if (strcmp(buf, "?") == 0) {
2391 fprintf(stderr, "qemu: Supported formats:");
2392 bdrv_iterate_format(bdrv_format_print, NULL);
2393 fprintf(stderr, "\n");
2394 return -1;
2396 drv = bdrv_find_format(buf);
2397 if (!drv) {
2398 fprintf(stderr, "qemu: '%s' invalid format\n", buf);
2399 return -1;
2403 if (arg->file == NULL)
2404 get_param_value(file, sizeof(file), "file", str);
2405 else
2406 pstrcpy(file, sizeof(file), arg->file);
2408 if (!get_param_value(serial, sizeof(serial), "serial", str))
2409 memset(serial, 0, sizeof(serial));
2411 onerror = BLOCK_ERR_STOP_ENOSPC;
2412 if (get_param_value(buf, sizeof(serial), "werror", str)) {
2413 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO) {
2414 fprintf(stderr, "werror is no supported by this format\n");
2415 return -1;
2417 if (!strcmp(buf, "ignore"))
2418 onerror = BLOCK_ERR_IGNORE;
2419 else if (!strcmp(buf, "enospc"))
2420 onerror = BLOCK_ERR_STOP_ENOSPC;
2421 else if (!strcmp(buf, "stop"))
2422 onerror = BLOCK_ERR_STOP_ANY;
2423 else if (!strcmp(buf, "report"))
2424 onerror = BLOCK_ERR_REPORT;
2425 else {
2426 fprintf(stderr, "qemu: '%s' invalid write error action\n", buf);
2427 return -1;
2431 /* compute bus and unit according index */
2433 if (index != -1) {
2434 if (bus_id != 0 || unit_id != -1) {
2435 fprintf(stderr,
2436 "qemu: '%s' index cannot be used with bus and unit\n", str);
2437 return -1;
2439 if (max_devs == 0)
2441 unit_id = index;
2442 bus_id = 0;
2443 } else {
2444 unit_id = index % max_devs;
2445 bus_id = index / max_devs;
2449 /* if user doesn't specify a unit_id,
2450 * try to find the first free
2453 if (unit_id == -1) {
2454 unit_id = 0;
2455 while (drive_get_index(type, bus_id, unit_id) != -1) {
2456 unit_id++;
2457 if (max_devs && unit_id >= max_devs) {
2458 unit_id -= max_devs;
2459 bus_id++;
2464 /* check unit id */
2466 if (max_devs && unit_id >= max_devs) {
2467 fprintf(stderr, "qemu: '%s' unit %d too big (max is %d)\n",
2468 str, unit_id, max_devs - 1);
2469 return -1;
2473 * ignore multiple definitions
2476 if (drive_get_index(type, bus_id, unit_id) != -1)
2477 return -2;
2479 /* init */
2481 if (type == IF_IDE || type == IF_SCSI)
2482 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
2483 if (max_devs)
2484 snprintf(buf, sizeof(buf), "%s%i%s%i",
2485 devname, bus_id, mediastr, unit_id);
2486 else
2487 snprintf(buf, sizeof(buf), "%s%s%i",
2488 devname, mediastr, unit_id);
2489 bdrv = bdrv_new(buf);
2490 drives_table_idx = drive_get_free_idx();
2491 drives_table[drives_table_idx].bdrv = bdrv;
2492 drives_table[drives_table_idx].type = type;
2493 drives_table[drives_table_idx].bus = bus_id;
2494 drives_table[drives_table_idx].unit = unit_id;
2495 drives_table[drives_table_idx].onerror = onerror;
2496 drives_table[drives_table_idx].drive_opt_idx = arg - drives_opt;
2497 strncpy(drives_table[drives_table_idx].serial, serial, sizeof(serial));
2498 nb_drives++;
2500 switch(type) {
2501 case IF_IDE:
2502 case IF_SCSI:
2503 case IF_XEN:
2504 switch(media) {
2505 case MEDIA_DISK:
2506 if (cyls != 0) {
2507 bdrv_set_geometry_hint(bdrv, cyls, heads, secs);
2508 bdrv_set_translation_hint(bdrv, translation);
2510 break;
2511 case MEDIA_CDROM:
2512 bdrv_set_type_hint(bdrv, BDRV_TYPE_CDROM);
2513 break;
2515 break;
2516 case IF_SD:
2517 /* FIXME: This isn't really a floppy, but it's a reasonable
2518 approximation. */
2519 case IF_FLOPPY:
2520 bdrv_set_type_hint(bdrv, BDRV_TYPE_FLOPPY);
2521 break;
2522 case IF_PFLASH:
2523 case IF_MTD:
2524 case IF_VIRTIO:
2525 break;
2526 case IF_COUNT:
2527 abort();
2529 if (!file[0])
2530 return -2;
2531 bdrv_flags = 0;
2532 if (snapshot) {
2533 bdrv_flags |= BDRV_O_SNAPSHOT;
2534 cache = 2; /* always use write-back with snapshot */
2536 if (cache == 0) /* no caching */
2537 bdrv_flags |= BDRV_O_NOCACHE;
2538 else if (cache == 2) /* write-back */
2539 bdrv_flags |= BDRV_O_CACHE_WB;
2540 else if (cache == 3) /* not specified */
2541 bdrv_flags |= BDRV_O_CACHE_DEF;
2542 if (bdrv_open2(bdrv, file, bdrv_flags, drv) < 0) {
2543 fprintf(stderr, "qemu: could not open disk image %s\n",
2544 file);
2545 return -1;
2547 if (bdrv_key_required(bdrv))
2548 autostart = 0;
2549 return drives_table_idx;
2552 static void numa_add(const char *optarg)
2554 char option[128];
2555 char *endptr;
2556 unsigned long long value, endvalue;
2557 int nodenr;
2559 optarg = get_opt_name(option, 128, optarg, ',') + 1;
2560 if (!strcmp(option, "node")) {
2561 if (get_param_value(option, 128, "nodeid", optarg) == 0) {
2562 nodenr = nb_numa_nodes;
2563 } else {
2564 nodenr = strtoull(option, NULL, 10);
2567 if (get_param_value(option, 128, "mem", optarg) == 0) {
2568 node_mem[nodenr] = 0;
2569 } else {
2570 value = strtoull(option, &endptr, 0);
2571 switch (*endptr) {
2572 case 0: case 'M': case 'm':
2573 value <<= 20;
2574 break;
2575 case 'G': case 'g':
2576 value <<= 30;
2577 break;
2579 node_mem[nodenr] = value;
2581 if (get_param_value(option, 128, "cpus", optarg) == 0) {
2582 node_cpumask[nodenr] = 0;
2583 } else {
2584 value = strtoull(option, &endptr, 10);
2585 if (value >= 64) {
2586 value = 63;
2587 fprintf(stderr, "only 64 CPUs in NUMA mode supported.\n");
2588 } else {
2589 if (*endptr == '-') {
2590 endvalue = strtoull(endptr+1, &endptr, 10);
2591 if (endvalue >= 63) {
2592 endvalue = 62;
2593 fprintf(stderr,
2594 "only 63 CPUs in NUMA mode supported.\n");
2596 value = (1 << (endvalue + 1)) - (1 << value);
2597 } else {
2598 value = 1 << value;
2601 node_cpumask[nodenr] = value;
2603 nb_numa_nodes++;
2605 return;
2608 /***********************************************************/
2609 /* USB devices */
2611 static USBPort *used_usb_ports;
2612 static USBPort *free_usb_ports;
2614 /* ??? Maybe change this to register a hub to keep track of the topology. */
2615 void qemu_register_usb_port(USBPort *port, void *opaque, int index,
2616 usb_attachfn attach)
2618 port->opaque = opaque;
2619 port->index = index;
2620 port->attach = attach;
2621 port->next = free_usb_ports;
2622 free_usb_ports = port;
2625 int usb_device_add_dev(USBDevice *dev)
2627 USBPort *port;
2629 /* Find a USB port to add the device to. */
2630 port = free_usb_ports;
2631 if (!port->next) {
2632 USBDevice *hub;
2634 /* Create a new hub and chain it on. */
2635 free_usb_ports = NULL;
2636 port->next = used_usb_ports;
2637 used_usb_ports = port;
2639 hub = usb_hub_init(VM_USB_HUB_SIZE);
2640 usb_attach(port, hub);
2641 port = free_usb_ports;
2644 free_usb_ports = port->next;
2645 port->next = used_usb_ports;
2646 used_usb_ports = port;
2647 usb_attach(port, dev);
2648 return 0;
2651 static void usb_msd_password_cb(void *opaque, int err)
2653 USBDevice *dev = opaque;
2655 if (!err)
2656 usb_device_add_dev(dev);
2657 else
2658 dev->handle_destroy(dev);
2661 static int usb_device_add(const char *devname, int is_hotplug)
2663 const char *p;
2664 USBDevice *dev;
2666 if (!free_usb_ports)
2667 return -1;
2669 if (strstart(devname, "host:", &p)) {
2670 dev = usb_host_device_open(p);
2671 } else if (!strcmp(devname, "mouse")) {
2672 dev = usb_mouse_init();
2673 } else if (!strcmp(devname, "tablet")) {
2674 dev = usb_tablet_init();
2675 } else if (!strcmp(devname, "keyboard")) {
2676 dev = usb_keyboard_init();
2677 } else if (strstart(devname, "disk:", &p)) {
2678 BlockDriverState *bs;
2680 dev = usb_msd_init(p);
2681 if (!dev)
2682 return -1;
2683 bs = usb_msd_get_bdrv(dev);
2684 if (bdrv_key_required(bs)) {
2685 autostart = 0;
2686 if (is_hotplug) {
2687 monitor_read_bdrv_key_start(cur_mon, bs, usb_msd_password_cb,
2688 dev);
2689 return 0;
2692 } else if (!strcmp(devname, "wacom-tablet")) {
2693 dev = usb_wacom_init();
2694 } else if (strstart(devname, "serial:", &p)) {
2695 dev = usb_serial_init(p);
2696 #ifdef CONFIG_BRLAPI
2697 } else if (!strcmp(devname, "braille")) {
2698 dev = usb_baum_init();
2699 #endif
2700 } else if (strstart(devname, "net:", &p)) {
2701 int nic = nb_nics;
2703 if (net_client_init(NULL, "nic", p) < 0)
2704 return -1;
2705 nd_table[nic].model = "usb";
2706 dev = usb_net_init(&nd_table[nic]);
2707 } else if (!strcmp(devname, "bt") || strstart(devname, "bt:", &p)) {
2708 dev = usb_bt_init(devname[2] ? hci_init(p) :
2709 bt_new_hci(qemu_find_bt_vlan(0)));
2710 } else {
2711 return -1;
2713 if (!dev)
2714 return -1;
2716 return usb_device_add_dev(dev);
2719 int usb_device_del_addr(int bus_num, int addr)
2721 USBPort *port;
2722 USBPort **lastp;
2723 USBDevice *dev;
2725 if (!used_usb_ports)
2726 return -1;
2728 if (bus_num != 0)
2729 return -1;
2731 lastp = &used_usb_ports;
2732 port = used_usb_ports;
2733 while (port && port->dev->addr != addr) {
2734 lastp = &port->next;
2735 port = port->next;
2738 if (!port)
2739 return -1;
2741 dev = port->dev;
2742 *lastp = port->next;
2743 usb_attach(port, NULL);
2744 dev->handle_destroy(dev);
2745 port->next = free_usb_ports;
2746 free_usb_ports = port;
2747 return 0;
2750 static int usb_device_del(const char *devname)
2752 int bus_num, addr;
2753 const char *p;
2755 if (strstart(devname, "host:", &p))
2756 return usb_host_device_close(p);
2758 if (!used_usb_ports)
2759 return -1;
2761 p = strchr(devname, '.');
2762 if (!p)
2763 return -1;
2764 bus_num = strtoul(devname, NULL, 0);
2765 addr = strtoul(p + 1, NULL, 0);
2767 return usb_device_del_addr(bus_num, addr);
2770 void do_usb_add(Monitor *mon, const char *devname)
2772 usb_device_add(devname, 1);
2775 void do_usb_del(Monitor *mon, const char *devname)
2777 usb_device_del(devname);
2780 void usb_info(Monitor *mon)
2782 USBDevice *dev;
2783 USBPort *port;
2784 const char *speed_str;
2786 if (!usb_enabled) {
2787 monitor_printf(mon, "USB support not enabled\n");
2788 return;
2791 for (port = used_usb_ports; port; port = port->next) {
2792 dev = port->dev;
2793 if (!dev)
2794 continue;
2795 switch(dev->speed) {
2796 case USB_SPEED_LOW:
2797 speed_str = "1.5";
2798 break;
2799 case USB_SPEED_FULL:
2800 speed_str = "12";
2801 break;
2802 case USB_SPEED_HIGH:
2803 speed_str = "480";
2804 break;
2805 default:
2806 speed_str = "?";
2807 break;
2809 monitor_printf(mon, " Device %d.%d, Speed %s Mb/s, Product %s\n",
2810 0, dev->addr, speed_str, dev->devname);
2814 /***********************************************************/
2815 /* PCMCIA/Cardbus */
2817 static struct pcmcia_socket_entry_s {
2818 PCMCIASocket *socket;
2819 struct pcmcia_socket_entry_s *next;
2820 } *pcmcia_sockets = 0;
2822 void pcmcia_socket_register(PCMCIASocket *socket)
2824 struct pcmcia_socket_entry_s *entry;
2826 entry = qemu_malloc(sizeof(struct pcmcia_socket_entry_s));
2827 entry->socket = socket;
2828 entry->next = pcmcia_sockets;
2829 pcmcia_sockets = entry;
2832 void pcmcia_socket_unregister(PCMCIASocket *socket)
2834 struct pcmcia_socket_entry_s *entry, **ptr;
2836 ptr = &pcmcia_sockets;
2837 for (entry = *ptr; entry; ptr = &entry->next, entry = *ptr)
2838 if (entry->socket == socket) {
2839 *ptr = entry->next;
2840 qemu_free(entry);
2844 void pcmcia_info(Monitor *mon)
2846 struct pcmcia_socket_entry_s *iter;
2848 if (!pcmcia_sockets)
2849 monitor_printf(mon, "No PCMCIA sockets\n");
2851 for (iter = pcmcia_sockets; iter; iter = iter->next)
2852 monitor_printf(mon, "%s: %s\n", iter->socket->slot_string,
2853 iter->socket->attached ? iter->socket->card_string :
2854 "Empty");
2857 /***********************************************************/
2858 /* register display */
2860 struct DisplayAllocator default_allocator = {
2861 defaultallocator_create_displaysurface,
2862 defaultallocator_resize_displaysurface,
2863 defaultallocator_free_displaysurface
2866 void register_displaystate(DisplayState *ds)
2868 DisplayState **s;
2869 s = &display_state;
2870 while (*s != NULL)
2871 s = &(*s)->next;
2872 ds->next = NULL;
2873 *s = ds;
2876 DisplayState *get_displaystate(void)
2878 return display_state;
2881 DisplayAllocator *register_displayallocator(DisplayState *ds, DisplayAllocator *da)
2883 if(ds->allocator == &default_allocator) ds->allocator = da;
2884 return ds->allocator;
2887 /* dumb display */
2889 static void dumb_display_init(void)
2891 DisplayState *ds = qemu_mallocz(sizeof(DisplayState));
2892 ds->allocator = &default_allocator;
2893 ds->surface = qemu_create_displaysurface(ds, 640, 480);
2894 register_displaystate(ds);
2897 /***********************************************************/
2898 /* I/O handling */
2900 typedef struct IOHandlerRecord {
2901 int fd;
2902 IOCanRWHandler *fd_read_poll;
2903 IOHandler *fd_read;
2904 IOHandler *fd_write;
2905 int deleted;
2906 void *opaque;
2907 /* temporary data */
2908 struct pollfd *ufd;
2909 struct IOHandlerRecord *next;
2910 } IOHandlerRecord;
2912 static IOHandlerRecord *first_io_handler;
2914 /* XXX: fd_read_poll should be suppressed, but an API change is
2915 necessary in the character devices to suppress fd_can_read(). */
2916 int qemu_set_fd_handler2(int fd,
2917 IOCanRWHandler *fd_read_poll,
2918 IOHandler *fd_read,
2919 IOHandler *fd_write,
2920 void *opaque)
2922 IOHandlerRecord **pioh, *ioh;
2924 if (!fd_read && !fd_write) {
2925 pioh = &first_io_handler;
2926 for(;;) {
2927 ioh = *pioh;
2928 if (ioh == NULL)
2929 break;
2930 if (ioh->fd == fd) {
2931 ioh->deleted = 1;
2932 break;
2934 pioh = &ioh->next;
2936 } else {
2937 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
2938 if (ioh->fd == fd)
2939 goto found;
2941 ioh = qemu_mallocz(sizeof(IOHandlerRecord));
2942 ioh->next = first_io_handler;
2943 first_io_handler = ioh;
2944 found:
2945 ioh->fd = fd;
2946 ioh->fd_read_poll = fd_read_poll;
2947 ioh->fd_read = fd_read;
2948 ioh->fd_write = fd_write;
2949 ioh->opaque = opaque;
2950 ioh->deleted = 0;
2952 return 0;
2955 int qemu_set_fd_handler(int fd,
2956 IOHandler *fd_read,
2957 IOHandler *fd_write,
2958 void *opaque)
2960 return qemu_set_fd_handler2(fd, NULL, fd_read, fd_write, opaque);
2963 #ifdef _WIN32
2964 /***********************************************************/
2965 /* Polling handling */
2967 typedef struct PollingEntry {
2968 PollingFunc *func;
2969 void *opaque;
2970 struct PollingEntry *next;
2971 } PollingEntry;
2973 static PollingEntry *first_polling_entry;
2975 int qemu_add_polling_cb(PollingFunc *func, void *opaque)
2977 PollingEntry **ppe, *pe;
2978 pe = qemu_mallocz(sizeof(PollingEntry));
2979 pe->func = func;
2980 pe->opaque = opaque;
2981 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next);
2982 *ppe = pe;
2983 return 0;
2986 void qemu_del_polling_cb(PollingFunc *func, void *opaque)
2988 PollingEntry **ppe, *pe;
2989 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next) {
2990 pe = *ppe;
2991 if (pe->func == func && pe->opaque == opaque) {
2992 *ppe = pe->next;
2993 qemu_free(pe);
2994 break;
2999 /***********************************************************/
3000 /* Wait objects support */
3001 typedef struct WaitObjects {
3002 int num;
3003 HANDLE events[MAXIMUM_WAIT_OBJECTS + 1];
3004 WaitObjectFunc *func[MAXIMUM_WAIT_OBJECTS + 1];
3005 void *opaque[MAXIMUM_WAIT_OBJECTS + 1];
3006 } WaitObjects;
3008 static WaitObjects wait_objects = {0};
3010 int qemu_add_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
3012 WaitObjects *w = &wait_objects;
3014 if (w->num >= MAXIMUM_WAIT_OBJECTS)
3015 return -1;
3016 w->events[w->num] = handle;
3017 w->func[w->num] = func;
3018 w->opaque[w->num] = opaque;
3019 w->num++;
3020 return 0;
3023 void qemu_del_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
3025 int i, found;
3026 WaitObjects *w = &wait_objects;
3028 found = 0;
3029 for (i = 0; i < w->num; i++) {
3030 if (w->events[i] == handle)
3031 found = 1;
3032 if (found) {
3033 w->events[i] = w->events[i + 1];
3034 w->func[i] = w->func[i + 1];
3035 w->opaque[i] = w->opaque[i + 1];
3038 if (found)
3039 w->num--;
3041 #endif
3043 /***********************************************************/
3044 /* ram save/restore */
3046 static int ram_get_page(QEMUFile *f, uint8_t *buf, int len)
3048 int v;
3050 v = qemu_get_byte(f);
3051 switch(v) {
3052 case 0:
3053 if (qemu_get_buffer(f, buf, len) != len)
3054 return -EIO;
3055 break;
3056 case 1:
3057 v = qemu_get_byte(f);
3058 memset(buf, v, len);
3059 break;
3060 default:
3061 return -EINVAL;
3064 if (qemu_file_has_error(f))
3065 return -EIO;
3067 return 0;
3070 static int ram_load_v1(QEMUFile *f, void *opaque)
3072 int ret;
3073 ram_addr_t i;
3075 if (qemu_get_be32(f) != last_ram_offset)
3076 return -EINVAL;
3077 for(i = 0; i < last_ram_offset; i+= TARGET_PAGE_SIZE) {
3078 ret = ram_get_page(f, qemu_get_ram_ptr(i), TARGET_PAGE_SIZE);
3079 if (ret)
3080 return ret;
3082 return 0;
3085 #define BDRV_HASH_BLOCK_SIZE 1024
3086 #define IOBUF_SIZE 4096
3087 #define RAM_CBLOCK_MAGIC 0xfabe
3089 typedef struct RamDecompressState {
3090 z_stream zstream;
3091 QEMUFile *f;
3092 uint8_t buf[IOBUF_SIZE];
3093 } RamDecompressState;
3095 static int ram_decompress_open(RamDecompressState *s, QEMUFile *f)
3097 int ret;
3098 memset(s, 0, sizeof(*s));
3099 s->f = f;
3100 ret = inflateInit(&s->zstream);
3101 if (ret != Z_OK)
3102 return -1;
3103 return 0;
3106 static int ram_decompress_buf(RamDecompressState *s, uint8_t *buf, int len)
3108 int ret, clen;
3110 s->zstream.avail_out = len;
3111 s->zstream.next_out = buf;
3112 while (s->zstream.avail_out > 0) {
3113 if (s->zstream.avail_in == 0) {
3114 if (qemu_get_be16(s->f) != RAM_CBLOCK_MAGIC)
3115 return -1;
3116 clen = qemu_get_be16(s->f);
3117 if (clen > IOBUF_SIZE)
3118 return -1;
3119 qemu_get_buffer(s->f, s->buf, clen);
3120 s->zstream.avail_in = clen;
3121 s->zstream.next_in = s->buf;
3123 ret = inflate(&s->zstream, Z_PARTIAL_FLUSH);
3124 if (ret != Z_OK && ret != Z_STREAM_END) {
3125 return -1;
3128 return 0;
3131 static void ram_decompress_close(RamDecompressState *s)
3133 inflateEnd(&s->zstream);
3136 #define RAM_SAVE_FLAG_FULL 0x01
3137 #define RAM_SAVE_FLAG_COMPRESS 0x02
3138 #define RAM_SAVE_FLAG_MEM_SIZE 0x04
3139 #define RAM_SAVE_FLAG_PAGE 0x08
3140 #define RAM_SAVE_FLAG_EOS 0x10
3142 static int is_dup_page(uint8_t *page, uint8_t ch)
3144 uint32_t val = ch << 24 | ch << 16 | ch << 8 | ch;
3145 uint32_t *array = (uint32_t *)page;
3146 int i;
3148 for (i = 0; i < (TARGET_PAGE_SIZE / 4); i++) {
3149 if (array[i] != val)
3150 return 0;
3153 return 1;
3156 static int ram_save_block(QEMUFile *f)
3158 static ram_addr_t current_addr = 0;
3159 ram_addr_t saved_addr = current_addr;
3160 ram_addr_t addr = 0;
3161 int found = 0;
3163 while (addr < last_ram_offset) {
3164 if (cpu_physical_memory_get_dirty(current_addr, MIGRATION_DIRTY_FLAG)) {
3165 uint8_t *p;
3167 cpu_physical_memory_reset_dirty(current_addr,
3168 current_addr + TARGET_PAGE_SIZE,
3169 MIGRATION_DIRTY_FLAG);
3171 p = qemu_get_ram_ptr(current_addr);
3173 if (is_dup_page(p, *p)) {
3174 qemu_put_be64(f, current_addr | RAM_SAVE_FLAG_COMPRESS);
3175 qemu_put_byte(f, *p);
3176 } else {
3177 qemu_put_be64(f, current_addr | RAM_SAVE_FLAG_PAGE);
3178 qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
3181 found = 1;
3182 break;
3184 addr += TARGET_PAGE_SIZE;
3185 current_addr = (saved_addr + addr) % last_ram_offset;
3188 return found;
3191 static uint64_t bytes_transferred = 0;
3193 static ram_addr_t ram_save_remaining(void)
3195 ram_addr_t addr;
3196 ram_addr_t count = 0;
3198 for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
3199 if (cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
3200 count++;
3203 return count;
3206 uint64_t ram_bytes_remaining(void)
3208 return ram_save_remaining() * TARGET_PAGE_SIZE;
3211 uint64_t ram_bytes_transferred(void)
3213 return bytes_transferred;
3216 uint64_t ram_bytes_total(void)
3218 return last_ram_offset;
3221 static int ram_save_live(QEMUFile *f, int stage, void *opaque)
3223 ram_addr_t addr;
3224 uint64_t bytes_transferred_last;
3225 double bwidth = 0;
3226 uint64_t expected_time = 0;
3228 if (cpu_physical_sync_dirty_bitmap(0, TARGET_PHYS_ADDR_MAX) != 0) {
3229 qemu_file_set_error(f);
3230 return 0;
3233 if (stage == 1) {
3234 /* Make sure all dirty bits are set */
3235 for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
3236 if (!cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
3237 cpu_physical_memory_set_dirty(addr);
3240 /* Enable dirty memory tracking */
3241 cpu_physical_memory_set_dirty_tracking(1);
3243 qemu_put_be64(f, last_ram_offset | RAM_SAVE_FLAG_MEM_SIZE);
3246 bytes_transferred_last = bytes_transferred;
3247 bwidth = get_clock();
3249 while (!qemu_file_rate_limit(f)) {
3250 int ret;
3252 ret = ram_save_block(f);
3253 bytes_transferred += ret * TARGET_PAGE_SIZE;
3254 if (ret == 0) /* no more blocks */
3255 break;
3258 bwidth = get_clock() - bwidth;
3259 bwidth = (bytes_transferred - bytes_transferred_last) / bwidth;
3261 /* if we haven't transferred anything this round, force expected_time to a
3262 * a very high value, but without crashing */
3263 if (bwidth == 0)
3264 bwidth = 0.000001;
3266 /* try transferring iterative blocks of memory */
3268 if (stage == 3) {
3270 /* flush all remaining blocks regardless of rate limiting */
3271 while (ram_save_block(f) != 0) {
3272 bytes_transferred += TARGET_PAGE_SIZE;
3274 cpu_physical_memory_set_dirty_tracking(0);
3277 qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
3279 expected_time = ram_save_remaining() * TARGET_PAGE_SIZE / bwidth;
3281 return (stage == 2) && (expected_time <= migrate_max_downtime());
3284 static int ram_load_dead(QEMUFile *f, void *opaque)
3286 RamDecompressState s1, *s = &s1;
3287 uint8_t buf[10];
3288 ram_addr_t i;
3290 if (ram_decompress_open(s, f) < 0)
3291 return -EINVAL;
3292 for(i = 0; i < last_ram_offset; i+= BDRV_HASH_BLOCK_SIZE) {
3293 if (ram_decompress_buf(s, buf, 1) < 0) {
3294 fprintf(stderr, "Error while reading ram block header\n");
3295 goto error;
3297 if (buf[0] == 0) {
3298 if (ram_decompress_buf(s, qemu_get_ram_ptr(i),
3299 BDRV_HASH_BLOCK_SIZE) < 0) {
3300 fprintf(stderr, "Error while reading ram block address=0x%08" PRIx64, (uint64_t)i);
3301 goto error;
3303 } else {
3304 error:
3305 printf("Error block header\n");
3306 return -EINVAL;
3309 ram_decompress_close(s);
3311 return 0;
3314 static int ram_load(QEMUFile *f, void *opaque, int version_id)
3316 ram_addr_t addr;
3317 int flags;
3319 if (version_id == 1)
3320 return ram_load_v1(f, opaque);
3322 if (version_id == 2) {
3323 if (qemu_get_be32(f) != last_ram_offset)
3324 return -EINVAL;
3325 return ram_load_dead(f, opaque);
3328 if (version_id != 3)
3329 return -EINVAL;
3331 do {
3332 addr = qemu_get_be64(f);
3334 flags = addr & ~TARGET_PAGE_MASK;
3335 addr &= TARGET_PAGE_MASK;
3337 if (flags & RAM_SAVE_FLAG_MEM_SIZE) {
3338 if (addr != last_ram_offset)
3339 return -EINVAL;
3342 if (flags & RAM_SAVE_FLAG_FULL) {
3343 if (ram_load_dead(f, opaque) < 0)
3344 return -EINVAL;
3347 if (flags & RAM_SAVE_FLAG_COMPRESS) {
3348 uint8_t ch = qemu_get_byte(f);
3349 #if defined(__linux__)
3350 if (ch == 0 &&
3351 (!kvm_enabled() || kvm_has_sync_mmu())) {
3352 madvise(qemu_get_ram_ptr(addr), TARGET_PAGE_SIZE, MADV_DONTNEED);
3353 } else
3354 #endif
3355 memset(qemu_get_ram_ptr(addr), ch, TARGET_PAGE_SIZE);
3356 } else if (flags & RAM_SAVE_FLAG_PAGE)
3357 qemu_get_buffer(f, qemu_get_ram_ptr(addr), TARGET_PAGE_SIZE);
3358 } while (!(flags & RAM_SAVE_FLAG_EOS));
3360 return 0;
3363 void qemu_service_io(void)
3365 qemu_notify_event();
3368 /***********************************************************/
3369 /* bottom halves (can be seen as timers which expire ASAP) */
3371 struct QEMUBH {
3372 QEMUBHFunc *cb;
3373 void *opaque;
3374 int scheduled;
3375 int idle;
3376 int deleted;
3377 QEMUBH *next;
3380 static QEMUBH *first_bh = NULL;
3382 QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
3384 QEMUBH *bh;
3385 bh = qemu_mallocz(sizeof(QEMUBH));
3386 bh->cb = cb;
3387 bh->opaque = opaque;
3388 bh->next = first_bh;
3389 first_bh = bh;
3390 return bh;
3393 int qemu_bh_poll(void)
3395 QEMUBH *bh, **bhp;
3396 int ret;
3398 ret = 0;
3399 for (bh = first_bh; bh; bh = bh->next) {
3400 if (!bh->deleted && bh->scheduled) {
3401 bh->scheduled = 0;
3402 if (!bh->idle)
3403 ret = 1;
3404 bh->idle = 0;
3405 bh->cb(bh->opaque);
3409 /* remove deleted bhs */
3410 bhp = &first_bh;
3411 while (*bhp) {
3412 bh = *bhp;
3413 if (bh->deleted) {
3414 *bhp = bh->next;
3415 qemu_free(bh);
3416 } else
3417 bhp = &bh->next;
3420 return ret;
3423 void qemu_bh_schedule_idle(QEMUBH *bh)
3425 if (bh->scheduled)
3426 return;
3427 bh->scheduled = 1;
3428 bh->idle = 1;
3431 void qemu_bh_schedule(QEMUBH *bh)
3433 if (bh->scheduled)
3434 return;
3435 bh->scheduled = 1;
3436 bh->idle = 0;
3437 /* stop the currently executing CPU to execute the BH ASAP */
3438 qemu_notify_event();
3441 void qemu_bh_cancel(QEMUBH *bh)
3443 bh->scheduled = 0;
3446 void qemu_bh_delete(QEMUBH *bh)
3448 bh->scheduled = 0;
3449 bh->deleted = 1;
3452 static void qemu_bh_update_timeout(int *timeout)
3454 QEMUBH *bh;
3456 for (bh = first_bh; bh; bh = bh->next) {
3457 if (!bh->deleted && bh->scheduled) {
3458 if (bh->idle) {
3459 /* idle bottom halves will be polled at least
3460 * every 10ms */
3461 *timeout = MIN(10, *timeout);
3462 } else {
3463 /* non-idle bottom halves will be executed
3464 * immediately */
3465 *timeout = 0;
3466 break;
3472 /***********************************************************/
3473 /* machine registration */
3475 static QEMUMachine *first_machine = NULL;
3476 QEMUMachine *current_machine = NULL;
3478 int qemu_register_machine(QEMUMachine *m)
3480 QEMUMachine **pm;
3481 pm = &first_machine;
3482 while (*pm != NULL)
3483 pm = &(*pm)->next;
3484 m->next = NULL;
3485 *pm = m;
3486 return 0;
3489 static QEMUMachine *find_machine(const char *name)
3491 QEMUMachine *m;
3493 for(m = first_machine; m != NULL; m = m->next) {
3494 if (!strcmp(m->name, name))
3495 return m;
3497 return NULL;
3500 static QEMUMachine *find_default_machine(void)
3502 QEMUMachine *m;
3504 for(m = first_machine; m != NULL; m = m->next) {
3505 if (m->is_default) {
3506 return m;
3509 return NULL;
3512 /***********************************************************/
3513 /* main execution loop */
3515 static void gui_update(void *opaque)
3517 uint64_t interval = GUI_REFRESH_INTERVAL;
3518 DisplayState *ds = opaque;
3519 DisplayChangeListener *dcl = ds->listeners;
3521 dpy_refresh(ds);
3523 while (dcl != NULL) {
3524 if (dcl->gui_timer_interval &&
3525 dcl->gui_timer_interval < interval)
3526 interval = dcl->gui_timer_interval;
3527 dcl = dcl->next;
3529 qemu_mod_timer(ds->gui_timer, interval + qemu_get_clock(rt_clock));
3532 static void nographic_update(void *opaque)
3534 uint64_t interval = GUI_REFRESH_INTERVAL;
3536 qemu_mod_timer(nographic_timer, interval + qemu_get_clock(rt_clock));
3539 struct vm_change_state_entry {
3540 VMChangeStateHandler *cb;
3541 void *opaque;
3542 LIST_ENTRY (vm_change_state_entry) entries;
3545 static LIST_HEAD(vm_change_state_head, vm_change_state_entry) vm_change_state_head;
3547 VMChangeStateEntry *qemu_add_vm_change_state_handler(VMChangeStateHandler *cb,
3548 void *opaque)
3550 VMChangeStateEntry *e;
3552 e = qemu_mallocz(sizeof (*e));
3554 e->cb = cb;
3555 e->opaque = opaque;
3556 LIST_INSERT_HEAD(&vm_change_state_head, e, entries);
3557 return e;
3560 void qemu_del_vm_change_state_handler(VMChangeStateEntry *e)
3562 LIST_REMOVE (e, entries);
3563 qemu_free (e);
3566 static void vm_state_notify(int running, int reason)
3568 VMChangeStateEntry *e;
3570 for (e = vm_change_state_head.lh_first; e; e = e->entries.le_next) {
3571 e->cb(e->opaque, running, reason);
3575 static void resume_all_vcpus(void);
3576 static void pause_all_vcpus(void);
3578 void vm_start(void)
3580 if (!vm_running) {
3581 cpu_enable_ticks();
3582 vm_running = 1;
3583 vm_state_notify(1, 0);
3584 qemu_rearm_alarm_timer(alarm_timer);
3585 resume_all_vcpus();
3589 /* reset/shutdown handler */
3591 typedef struct QEMUResetEntry {
3592 QEMUResetHandler *func;
3593 void *opaque;
3594 int order;
3595 struct QEMUResetEntry *next;
3596 } QEMUResetEntry;
3598 static QEMUResetEntry *first_reset_entry;
3599 static int reset_requested;
3600 static int shutdown_requested;
3601 static int powerdown_requested;
3602 static int debug_requested;
3603 static int vmstop_requested;
3605 int qemu_shutdown_requested(void)
3607 int r = shutdown_requested;
3608 shutdown_requested = 0;
3609 return r;
3612 int qemu_reset_requested(void)
3614 int r = reset_requested;
3615 reset_requested = 0;
3616 return r;
3619 int qemu_powerdown_requested(void)
3621 int r = powerdown_requested;
3622 powerdown_requested = 0;
3623 return r;
3626 static int qemu_debug_requested(void)
3628 int r = debug_requested;
3629 debug_requested = 0;
3630 return r;
3633 static int qemu_vmstop_requested(void)
3635 int r = vmstop_requested;
3636 vmstop_requested = 0;
3637 return r;
3640 static void do_vm_stop(int reason)
3642 if (vm_running) {
3643 cpu_disable_ticks();
3644 vm_running = 0;
3645 pause_all_vcpus();
3646 vm_state_notify(0, reason);
3650 void qemu_register_reset(QEMUResetHandler *func, int order, void *opaque)
3652 QEMUResetEntry **pre, *re;
3654 pre = &first_reset_entry;
3655 while (*pre != NULL && (*pre)->order >= order) {
3656 pre = &(*pre)->next;
3658 re = qemu_mallocz(sizeof(QEMUResetEntry));
3659 re->func = func;
3660 re->opaque = opaque;
3661 re->order = order;
3662 re->next = NULL;
3663 *pre = re;
3666 void qemu_system_reset(void)
3668 QEMUResetEntry *re;
3670 /* reset all devices */
3671 for(re = first_reset_entry; re != NULL; re = re->next) {
3672 re->func(re->opaque);
3676 void qemu_system_reset_request(void)
3678 if (no_reboot) {
3679 shutdown_requested = 1;
3680 } else {
3681 reset_requested = 1;
3683 qemu_notify_event();
3686 void qemu_system_shutdown_request(void)
3688 shutdown_requested = 1;
3689 qemu_notify_event();
3692 void qemu_system_powerdown_request(void)
3694 powerdown_requested = 1;
3695 qemu_notify_event();
3698 #ifdef CONFIG_IOTHREAD
3699 static void qemu_system_vmstop_request(int reason)
3701 vmstop_requested = reason;
3702 qemu_notify_event();
3704 #endif
3706 #ifndef _WIN32
3707 static int io_thread_fd = -1;
3709 static void qemu_event_increment(void)
3711 static const char byte = 0;
3713 if (io_thread_fd == -1)
3714 return;
3716 write(io_thread_fd, &byte, sizeof(byte));
3719 static void qemu_event_read(void *opaque)
3721 int fd = (unsigned long)opaque;
3722 ssize_t len;
3724 /* Drain the notify pipe */
3725 do {
3726 char buffer[512];
3727 len = read(fd, buffer, sizeof(buffer));
3728 } while ((len == -1 && errno == EINTR) || len > 0);
3731 static int qemu_event_init(void)
3733 int err;
3734 int fds[2];
3736 err = pipe(fds);
3737 if (err == -1)
3738 return -errno;
3740 err = fcntl_setfl(fds[0], O_NONBLOCK);
3741 if (err < 0)
3742 goto fail;
3744 err = fcntl_setfl(fds[1], O_NONBLOCK);
3745 if (err < 0)
3746 goto fail;
3748 qemu_set_fd_handler2(fds[0], NULL, qemu_event_read, NULL,
3749 (void *)(unsigned long)fds[0]);
3751 io_thread_fd = fds[1];
3752 return 0;
3754 fail:
3755 close(fds[0]);
3756 close(fds[1]);
3757 return err;
3759 #else
3760 HANDLE qemu_event_handle;
3762 static void dummy_event_handler(void *opaque)
3766 static int qemu_event_init(void)
3768 qemu_event_handle = CreateEvent(NULL, FALSE, FALSE, NULL);
3769 if (!qemu_event_handle) {
3770 perror("Failed CreateEvent");
3771 return -1;
3773 qemu_add_wait_object(qemu_event_handle, dummy_event_handler, NULL);
3774 return 0;
3777 static void qemu_event_increment(void)
3779 SetEvent(qemu_event_handle);
3781 #endif
3783 static int cpu_can_run(CPUState *env)
3785 if (env->stop)
3786 return 0;
3787 if (env->stopped)
3788 return 0;
3789 return 1;
3792 #ifndef CONFIG_IOTHREAD
3793 static int qemu_init_main_loop(void)
3795 return qemu_event_init();
3798 void qemu_init_vcpu(void *_env)
3800 CPUState *env = _env;
3802 if (kvm_enabled())
3803 kvm_init_vcpu(env);
3804 return;
3807 int qemu_cpu_self(void *env)
3809 return 1;
3812 static void resume_all_vcpus(void)
3816 static void pause_all_vcpus(void)
3820 void qemu_cpu_kick(void *env)
3822 return;
3825 void qemu_notify_event(void)
3827 CPUState *env = cpu_single_env;
3829 if (env) {
3830 cpu_exit(env);
3831 #ifdef USE_KQEMU
3832 if (env->kqemu_enabled)
3833 kqemu_cpu_interrupt(env);
3834 #endif
3838 #define qemu_mutex_lock_iothread() do { } while (0)
3839 #define qemu_mutex_unlock_iothread() do { } while (0)
3841 void vm_stop(int reason)
3843 do_vm_stop(reason);
3846 #else /* CONFIG_IOTHREAD */
3848 #include "qemu-thread.h"
3850 QemuMutex qemu_global_mutex;
3851 static QemuMutex qemu_fair_mutex;
3853 static QemuThread io_thread;
3855 static QemuThread *tcg_cpu_thread;
3856 static QemuCond *tcg_halt_cond;
3858 static int qemu_system_ready;
3859 /* cpu creation */
3860 static QemuCond qemu_cpu_cond;
3861 /* system init */
3862 static QemuCond qemu_system_cond;
3863 static QemuCond qemu_pause_cond;
3865 static void block_io_signals(void);
3866 static void unblock_io_signals(void);
3867 static int tcg_has_work(void);
3869 static int qemu_init_main_loop(void)
3871 int ret;
3873 ret = qemu_event_init();
3874 if (ret)
3875 return ret;
3877 qemu_cond_init(&qemu_pause_cond);
3878 qemu_mutex_init(&qemu_fair_mutex);
3879 qemu_mutex_init(&qemu_global_mutex);
3880 qemu_mutex_lock(&qemu_global_mutex);
3882 unblock_io_signals();
3883 qemu_thread_self(&io_thread);
3885 return 0;
3888 static void qemu_wait_io_event(CPUState *env)
3890 while (!tcg_has_work())
3891 qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000);
3893 qemu_mutex_unlock(&qemu_global_mutex);
3896 * Users of qemu_global_mutex can be starved, having no chance
3897 * to acquire it since this path will get to it first.
3898 * So use another lock to provide fairness.
3900 qemu_mutex_lock(&qemu_fair_mutex);
3901 qemu_mutex_unlock(&qemu_fair_mutex);
3903 qemu_mutex_lock(&qemu_global_mutex);
3904 if (env->stop) {
3905 env->stop = 0;
3906 env->stopped = 1;
3907 qemu_cond_signal(&qemu_pause_cond);
3911 static int qemu_cpu_exec(CPUState *env);
3913 static void *kvm_cpu_thread_fn(void *arg)
3915 CPUState *env = arg;
3917 block_io_signals();
3918 qemu_thread_self(env->thread);
3920 /* signal CPU creation */
3921 qemu_mutex_lock(&qemu_global_mutex);
3922 env->created = 1;
3923 qemu_cond_signal(&qemu_cpu_cond);
3925 /* and wait for machine initialization */
3926 while (!qemu_system_ready)
3927 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
3929 while (1) {
3930 if (cpu_can_run(env))
3931 qemu_cpu_exec(env);
3932 qemu_wait_io_event(env);
3935 return NULL;
3938 static void tcg_cpu_exec(void);
3940 static void *tcg_cpu_thread_fn(void *arg)
3942 CPUState *env = arg;
3944 block_io_signals();
3945 qemu_thread_self(env->thread);
3947 /* signal CPU creation */
3948 qemu_mutex_lock(&qemu_global_mutex);
3949 for (env = first_cpu; env != NULL; env = env->next_cpu)
3950 env->created = 1;
3951 qemu_cond_signal(&qemu_cpu_cond);
3953 /* and wait for machine initialization */
3954 while (!qemu_system_ready)
3955 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
3957 while (1) {
3958 tcg_cpu_exec();
3959 qemu_wait_io_event(cur_cpu);
3962 return NULL;
3965 void qemu_cpu_kick(void *_env)
3967 CPUState *env = _env;
3968 qemu_cond_broadcast(env->halt_cond);
3969 if (kvm_enabled())
3970 qemu_thread_signal(env->thread, SIGUSR1);
3973 int qemu_cpu_self(void *env)
3975 return (cpu_single_env != NULL);
3978 static void cpu_signal(int sig)
3980 if (cpu_single_env)
3981 cpu_exit(cpu_single_env);
3984 static void block_io_signals(void)
3986 sigset_t set;
3987 struct sigaction sigact;
3989 sigemptyset(&set);
3990 sigaddset(&set, SIGUSR2);
3991 sigaddset(&set, SIGIO);
3992 sigaddset(&set, SIGALRM);
3993 pthread_sigmask(SIG_BLOCK, &set, NULL);
3995 sigemptyset(&set);
3996 sigaddset(&set, SIGUSR1);
3997 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
3999 memset(&sigact, 0, sizeof(sigact));
4000 sigact.sa_handler = cpu_signal;
4001 sigaction(SIGUSR1, &sigact, NULL);
4004 static void unblock_io_signals(void)
4006 sigset_t set;
4008 sigemptyset(&set);
4009 sigaddset(&set, SIGUSR2);
4010 sigaddset(&set, SIGIO);
4011 sigaddset(&set, SIGALRM);
4012 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
4014 sigemptyset(&set);
4015 sigaddset(&set, SIGUSR1);
4016 pthread_sigmask(SIG_BLOCK, &set, NULL);
4019 static void qemu_signal_lock(unsigned int msecs)
4021 qemu_mutex_lock(&qemu_fair_mutex);
4023 while (qemu_mutex_trylock(&qemu_global_mutex)) {
4024 qemu_thread_signal(tcg_cpu_thread, SIGUSR1);
4025 if (!qemu_mutex_timedlock(&qemu_global_mutex, msecs))
4026 break;
4028 qemu_mutex_unlock(&qemu_fair_mutex);
4031 static void qemu_mutex_lock_iothread(void)
4033 if (kvm_enabled()) {
4034 qemu_mutex_lock(&qemu_fair_mutex);
4035 qemu_mutex_lock(&qemu_global_mutex);
4036 qemu_mutex_unlock(&qemu_fair_mutex);
4037 } else
4038 qemu_signal_lock(100);
4041 static void qemu_mutex_unlock_iothread(void)
4043 qemu_mutex_unlock(&qemu_global_mutex);
4046 static int all_vcpus_paused(void)
4048 CPUState *penv = first_cpu;
4050 while (penv) {
4051 if (!penv->stopped)
4052 return 0;
4053 penv = (CPUState *)penv->next_cpu;
4056 return 1;
4059 static void pause_all_vcpus(void)
4061 CPUState *penv = first_cpu;
4063 while (penv) {
4064 penv->stop = 1;
4065 qemu_thread_signal(penv->thread, SIGUSR1);
4066 qemu_cpu_kick(penv);
4067 penv = (CPUState *)penv->next_cpu;
4070 while (!all_vcpus_paused()) {
4071 qemu_cond_timedwait(&qemu_pause_cond, &qemu_global_mutex, 100);
4072 penv = first_cpu;
4073 while (penv) {
4074 qemu_thread_signal(penv->thread, SIGUSR1);
4075 penv = (CPUState *)penv->next_cpu;
4080 static void resume_all_vcpus(void)
4082 CPUState *penv = first_cpu;
4084 while (penv) {
4085 penv->stop = 0;
4086 penv->stopped = 0;
4087 qemu_thread_signal(penv->thread, SIGUSR1);
4088 qemu_cpu_kick(penv);
4089 penv = (CPUState *)penv->next_cpu;
4093 static void tcg_init_vcpu(void *_env)
4095 CPUState *env = _env;
4096 /* share a single thread for all cpus with TCG */
4097 if (!tcg_cpu_thread) {
4098 env->thread = qemu_mallocz(sizeof(QemuThread));
4099 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
4100 qemu_cond_init(env->halt_cond);
4101 qemu_thread_create(env->thread, tcg_cpu_thread_fn, env);
4102 while (env->created == 0)
4103 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
4104 tcg_cpu_thread = env->thread;
4105 tcg_halt_cond = env->halt_cond;
4106 } else {
4107 env->thread = tcg_cpu_thread;
4108 env->halt_cond = tcg_halt_cond;
4112 static void kvm_start_vcpu(CPUState *env)
4114 kvm_init_vcpu(env);
4115 env->thread = qemu_mallocz(sizeof(QemuThread));
4116 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
4117 qemu_cond_init(env->halt_cond);
4118 qemu_thread_create(env->thread, kvm_cpu_thread_fn, env);
4119 while (env->created == 0)
4120 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
4123 void qemu_init_vcpu(void *_env)
4125 CPUState *env = _env;
4127 if (kvm_enabled())
4128 kvm_start_vcpu(env);
4129 else
4130 tcg_init_vcpu(env);
4133 void qemu_notify_event(void)
4135 qemu_event_increment();
4138 void vm_stop(int reason)
4140 QemuThread me;
4141 qemu_thread_self(&me);
4143 if (!qemu_thread_equal(&me, &io_thread)) {
4144 qemu_system_vmstop_request(reason);
4146 * FIXME: should not return to device code in case
4147 * vm_stop() has been requested.
4149 if (cpu_single_env) {
4150 cpu_exit(cpu_single_env);
4151 cpu_single_env->stop = 1;
4153 return;
4155 do_vm_stop(reason);
4158 #endif
4161 #ifdef _WIN32
4162 static void host_main_loop_wait(int *timeout)
4164 int ret, ret2, i;
4165 PollingEntry *pe;
4168 /* XXX: need to suppress polling by better using win32 events */
4169 ret = 0;
4170 for(pe = first_polling_entry; pe != NULL; pe = pe->next) {
4171 ret |= pe->func(pe->opaque);
4173 if (ret == 0) {
4174 int err;
4175 WaitObjects *w = &wait_objects;
4177 ret = WaitForMultipleObjects(w->num, w->events, FALSE, *timeout);
4178 if (WAIT_OBJECT_0 + 0 <= ret && ret <= WAIT_OBJECT_0 + w->num - 1) {
4179 if (w->func[ret - WAIT_OBJECT_0])
4180 w->func[ret - WAIT_OBJECT_0](w->opaque[ret - WAIT_OBJECT_0]);
4182 /* Check for additional signaled events */
4183 for(i = (ret - WAIT_OBJECT_0 + 1); i < w->num; i++) {
4185 /* Check if event is signaled */
4186 ret2 = WaitForSingleObject(w->events[i], 0);
4187 if(ret2 == WAIT_OBJECT_0) {
4188 if (w->func[i])
4189 w->func[i](w->opaque[i]);
4190 } else if (ret2 == WAIT_TIMEOUT) {
4191 } else {
4192 err = GetLastError();
4193 fprintf(stderr, "WaitForSingleObject error %d %d\n", i, err);
4196 } else if (ret == WAIT_TIMEOUT) {
4197 } else {
4198 err = GetLastError();
4199 fprintf(stderr, "WaitForMultipleObjects error %d %d\n", ret, err);
4203 *timeout = 0;
4205 #else
4206 static void host_main_loop_wait(int *timeout)
4209 #endif
4211 void main_loop_wait(int timeout)
4213 IOHandlerRecord *ioh;
4214 fd_set rfds, wfds, xfds;
4215 int ret, nfds;
4216 struct timeval tv;
4218 qemu_bh_update_timeout(&timeout);
4220 host_main_loop_wait(&timeout);
4222 /* poll any events */
4223 /* XXX: separate device handlers from system ones */
4224 nfds = -1;
4225 FD_ZERO(&rfds);
4226 FD_ZERO(&wfds);
4227 FD_ZERO(&xfds);
4228 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
4229 if (ioh->deleted)
4230 continue;
4231 if (ioh->fd_read &&
4232 (!ioh->fd_read_poll ||
4233 ioh->fd_read_poll(ioh->opaque) != 0)) {
4234 FD_SET(ioh->fd, &rfds);
4235 if (ioh->fd > nfds)
4236 nfds = ioh->fd;
4238 if (ioh->fd_write) {
4239 FD_SET(ioh->fd, &wfds);
4240 if (ioh->fd > nfds)
4241 nfds = ioh->fd;
4245 tv.tv_sec = timeout / 1000;
4246 tv.tv_usec = (timeout % 1000) * 1000;
4248 #if defined(CONFIG_SLIRP)
4249 if (slirp_is_inited()) {
4250 slirp_select_fill(&nfds, &rfds, &wfds, &xfds);
4252 #endif
4253 qemu_mutex_unlock_iothread();
4254 ret = select(nfds + 1, &rfds, &wfds, &xfds, &tv);
4255 qemu_mutex_lock_iothread();
4256 if (ret > 0) {
4257 IOHandlerRecord **pioh;
4259 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
4260 if (!ioh->deleted && ioh->fd_read && FD_ISSET(ioh->fd, &rfds)) {
4261 ioh->fd_read(ioh->opaque);
4263 if (!ioh->deleted && ioh->fd_write && FD_ISSET(ioh->fd, &wfds)) {
4264 ioh->fd_write(ioh->opaque);
4268 /* remove deleted IO handlers */
4269 pioh = &first_io_handler;
4270 while (*pioh) {
4271 ioh = *pioh;
4272 if (ioh->deleted) {
4273 *pioh = ioh->next;
4274 qemu_free(ioh);
4275 } else
4276 pioh = &ioh->next;
4279 #if defined(CONFIG_SLIRP)
4280 if (slirp_is_inited()) {
4281 if (ret < 0) {
4282 FD_ZERO(&rfds);
4283 FD_ZERO(&wfds);
4284 FD_ZERO(&xfds);
4286 slirp_select_poll(&rfds, &wfds, &xfds);
4288 #endif
4290 /* rearm timer, if not periodic */
4291 if (alarm_timer->flags & ALARM_FLAG_EXPIRED) {
4292 alarm_timer->flags &= ~ALARM_FLAG_EXPIRED;
4293 qemu_rearm_alarm_timer(alarm_timer);
4296 /* vm time timers */
4297 if (vm_running) {
4298 if (!cur_cpu || likely(!(cur_cpu->singlestep_enabled & SSTEP_NOTIMER)))
4299 qemu_run_timers(&active_timers[QEMU_TIMER_VIRTUAL],
4300 qemu_get_clock(vm_clock));
4303 /* real time timers */
4304 qemu_run_timers(&active_timers[QEMU_TIMER_REALTIME],
4305 qemu_get_clock(rt_clock));
4307 /* Check bottom-halves last in case any of the earlier events triggered
4308 them. */
4309 qemu_bh_poll();
4313 static int qemu_cpu_exec(CPUState *env)
4315 int ret;
4316 #ifdef CONFIG_PROFILER
4317 int64_t ti;
4318 #endif
4320 #ifdef CONFIG_PROFILER
4321 ti = profile_getclock();
4322 #endif
4323 if (use_icount) {
4324 int64_t count;
4325 int decr;
4326 qemu_icount -= (env->icount_decr.u16.low + env->icount_extra);
4327 env->icount_decr.u16.low = 0;
4328 env->icount_extra = 0;
4329 count = qemu_next_deadline();
4330 count = (count + (1 << icount_time_shift) - 1)
4331 >> icount_time_shift;
4332 qemu_icount += count;
4333 decr = (count > 0xffff) ? 0xffff : count;
4334 count -= decr;
4335 env->icount_decr.u16.low = decr;
4336 env->icount_extra = count;
4338 ret = cpu_exec(env);
4339 #ifdef CONFIG_PROFILER
4340 qemu_time += profile_getclock() - ti;
4341 #endif
4342 if (use_icount) {
4343 /* Fold pending instructions back into the
4344 instruction counter, and clear the interrupt flag. */
4345 qemu_icount -= (env->icount_decr.u16.low
4346 + env->icount_extra);
4347 env->icount_decr.u32 = 0;
4348 env->icount_extra = 0;
4350 return ret;
4353 static void tcg_cpu_exec(void)
4355 int ret = 0;
4357 if (next_cpu == NULL)
4358 next_cpu = first_cpu;
4359 for (; next_cpu != NULL; next_cpu = next_cpu->next_cpu) {
4360 CPUState *env = cur_cpu = next_cpu;
4362 if (!vm_running)
4363 break;
4364 if (timer_alarm_pending) {
4365 timer_alarm_pending = 0;
4366 break;
4368 if (cpu_can_run(env))
4369 ret = qemu_cpu_exec(env);
4370 if (ret == EXCP_DEBUG) {
4371 gdb_set_stop_cpu(env);
4372 debug_requested = 1;
4373 break;
4378 static int cpu_has_work(CPUState *env)
4380 if (env->stop)
4381 return 1;
4382 if (env->stopped)
4383 return 0;
4384 if (!env->halted)
4385 return 1;
4386 if (qemu_cpu_has_work(env))
4387 return 1;
4388 return 0;
4391 static int tcg_has_work(void)
4393 CPUState *env;
4395 for (env = first_cpu; env != NULL; env = env->next_cpu)
4396 if (cpu_has_work(env))
4397 return 1;
4398 return 0;
4401 static int qemu_calculate_timeout(void)
4403 #ifndef CONFIG_IOTHREAD
4404 int timeout;
4406 if (!vm_running)
4407 timeout = 5000;
4408 else if (tcg_has_work())
4409 timeout = 0;
4410 else if (!use_icount)
4411 timeout = 5000;
4412 else {
4413 /* XXX: use timeout computed from timers */
4414 int64_t add;
4415 int64_t delta;
4416 /* Advance virtual time to the next event. */
4417 if (use_icount == 1) {
4418 /* When not using an adaptive execution frequency
4419 we tend to get badly out of sync with real time,
4420 so just delay for a reasonable amount of time. */
4421 delta = 0;
4422 } else {
4423 delta = cpu_get_icount() - cpu_get_clock();
4425 if (delta > 0) {
4426 /* If virtual time is ahead of real time then just
4427 wait for IO. */
4428 timeout = (delta / 1000000) + 1;
4429 } else {
4430 /* Wait for either IO to occur or the next
4431 timer event. */
4432 add = qemu_next_deadline();
4433 /* We advance the timer before checking for IO.
4434 Limit the amount we advance so that early IO
4435 activity won't get the guest too far ahead. */
4436 if (add > 10000000)
4437 add = 10000000;
4438 delta += add;
4439 add = (add + (1 << icount_time_shift) - 1)
4440 >> icount_time_shift;
4441 qemu_icount += add;
4442 timeout = delta / 1000000;
4443 if (timeout < 0)
4444 timeout = 0;
4448 return timeout;
4449 #else /* CONFIG_IOTHREAD */
4450 return 1000;
4451 #endif
4454 static int vm_can_run(void)
4456 if (powerdown_requested)
4457 return 0;
4458 if (reset_requested)
4459 return 0;
4460 if (shutdown_requested)
4461 return 0;
4462 if (debug_requested)
4463 return 0;
4464 return 1;
4467 static void main_loop(void)
4469 int r;
4471 #ifdef CONFIG_IOTHREAD
4472 qemu_system_ready = 1;
4473 qemu_cond_broadcast(&qemu_system_cond);
4474 #endif
4476 for (;;) {
4477 do {
4478 #ifdef CONFIG_PROFILER
4479 int64_t ti;
4480 #endif
4481 #ifndef CONFIG_IOTHREAD
4482 tcg_cpu_exec();
4483 #endif
4484 #ifdef CONFIG_PROFILER
4485 ti = profile_getclock();
4486 #endif
4487 main_loop_wait(qemu_calculate_timeout());
4488 #ifdef CONFIG_PROFILER
4489 dev_time += profile_getclock() - ti;
4490 #endif
4491 } while (vm_can_run());
4493 if (qemu_debug_requested())
4494 vm_stop(EXCP_DEBUG);
4495 if (qemu_shutdown_requested()) {
4496 if (no_shutdown) {
4497 vm_stop(0);
4498 no_shutdown = 0;
4499 } else
4500 break;
4502 if (qemu_reset_requested()) {
4503 pause_all_vcpus();
4504 qemu_system_reset();
4505 resume_all_vcpus();
4507 if (qemu_powerdown_requested())
4508 qemu_system_powerdown();
4509 if ((r = qemu_vmstop_requested()))
4510 vm_stop(r);
4512 pause_all_vcpus();
4515 static void version(void)
4517 printf("QEMU PC emulator version " QEMU_VERSION QEMU_PKGVERSION ", Copyright (c) 2003-2008 Fabrice Bellard\n");
4520 static void help(int exitcode)
4522 version();
4523 printf("usage: %s [options] [disk_image]\n"
4524 "\n"
4525 "'disk_image' is a raw hard image image for IDE hard disk 0\n"
4526 "\n"
4527 #define DEF(option, opt_arg, opt_enum, opt_help) \
4528 opt_help
4529 #define DEFHEADING(text) stringify(text) "\n"
4530 #include "qemu-options.h"
4531 #undef DEF
4532 #undef DEFHEADING
4533 #undef GEN_DOCS
4534 "\n"
4535 "During emulation, the following keys are useful:\n"
4536 "ctrl-alt-f toggle full screen\n"
4537 "ctrl-alt-n switch to virtual console 'n'\n"
4538 "ctrl-alt toggle mouse and keyboard grab\n"
4539 "\n"
4540 "When using -nographic, press 'ctrl-a h' to get some help.\n"
4542 "qemu",
4543 DEFAULT_RAM_SIZE,
4544 #ifndef _WIN32
4545 DEFAULT_NETWORK_SCRIPT,
4546 DEFAULT_NETWORK_DOWN_SCRIPT,
4547 #endif
4548 DEFAULT_GDBSTUB_PORT,
4549 "/tmp/qemu.log");
4550 exit(exitcode);
4553 #define HAS_ARG 0x0001
4555 enum {
4556 #define DEF(option, opt_arg, opt_enum, opt_help) \
4557 opt_enum,
4558 #define DEFHEADING(text)
4559 #include "qemu-options.h"
4560 #undef DEF
4561 #undef DEFHEADING
4562 #undef GEN_DOCS
4565 typedef struct QEMUOption {
4566 const char *name;
4567 int flags;
4568 int index;
4569 } QEMUOption;
4571 static const QEMUOption qemu_options[] = {
4572 { "h", 0, QEMU_OPTION_h },
4573 #define DEF(option, opt_arg, opt_enum, opt_help) \
4574 { option, opt_arg, opt_enum },
4575 #define DEFHEADING(text)
4576 #include "qemu-options.h"
4577 #undef DEF
4578 #undef DEFHEADING
4579 #undef GEN_DOCS
4580 { NULL },
4583 #ifdef HAS_AUDIO
4584 struct soundhw soundhw[] = {
4585 #ifdef HAS_AUDIO_CHOICE
4586 #if defined(TARGET_I386) || defined(TARGET_MIPS)
4588 "pcspk",
4589 "PC speaker",
4592 { .init_isa = pcspk_audio_init }
4594 #endif
4596 #ifdef CONFIG_SB16
4598 "sb16",
4599 "Creative Sound Blaster 16",
4602 { .init_isa = SB16_init }
4604 #endif
4606 #ifdef CONFIG_CS4231A
4608 "cs4231a",
4609 "CS4231A",
4612 { .init_isa = cs4231a_init }
4614 #endif
4616 #ifdef CONFIG_ADLIB
4618 "adlib",
4619 #ifdef HAS_YMF262
4620 "Yamaha YMF262 (OPL3)",
4621 #else
4622 "Yamaha YM3812 (OPL2)",
4623 #endif
4626 { .init_isa = Adlib_init }
4628 #endif
4630 #ifdef CONFIG_GUS
4632 "gus",
4633 "Gravis Ultrasound GF1",
4636 { .init_isa = GUS_init }
4638 #endif
4640 #ifdef CONFIG_AC97
4642 "ac97",
4643 "Intel 82801AA AC97 Audio",
4646 { .init_pci = ac97_init }
4648 #endif
4650 #ifdef CONFIG_ES1370
4652 "es1370",
4653 "ENSONIQ AudioPCI ES1370",
4656 { .init_pci = es1370_init }
4658 #endif
4660 #endif /* HAS_AUDIO_CHOICE */
4662 { NULL, NULL, 0, 0, { NULL } }
4665 static void select_soundhw (const char *optarg)
4667 struct soundhw *c;
4669 if (*optarg == '?') {
4670 show_valid_cards:
4672 printf ("Valid sound card names (comma separated):\n");
4673 for (c = soundhw; c->name; ++c) {
4674 printf ("%-11s %s\n", c->name, c->descr);
4676 printf ("\n-soundhw all will enable all of the above\n");
4677 exit (*optarg != '?');
4679 else {
4680 size_t l;
4681 const char *p;
4682 char *e;
4683 int bad_card = 0;
4685 if (!strcmp (optarg, "all")) {
4686 for (c = soundhw; c->name; ++c) {
4687 c->enabled = 1;
4689 return;
4692 p = optarg;
4693 while (*p) {
4694 e = strchr (p, ',');
4695 l = !e ? strlen (p) : (size_t) (e - p);
4697 for (c = soundhw; c->name; ++c) {
4698 if (!strncmp (c->name, p, l)) {
4699 c->enabled = 1;
4700 break;
4704 if (!c->name) {
4705 if (l > 80) {
4706 fprintf (stderr,
4707 "Unknown sound card name (too big to show)\n");
4709 else {
4710 fprintf (stderr, "Unknown sound card name `%.*s'\n",
4711 (int) l, p);
4713 bad_card = 1;
4715 p += l + (e != NULL);
4718 if (bad_card)
4719 goto show_valid_cards;
4722 #endif
4724 static void select_vgahw (const char *p)
4726 const char *opts;
4728 cirrus_vga_enabled = 0;
4729 std_vga_enabled = 0;
4730 vmsvga_enabled = 0;
4731 xenfb_enabled = 0;
4732 if (strstart(p, "std", &opts)) {
4733 std_vga_enabled = 1;
4734 } else if (strstart(p, "cirrus", &opts)) {
4735 cirrus_vga_enabled = 1;
4736 } else if (strstart(p, "vmware", &opts)) {
4737 vmsvga_enabled = 1;
4738 } else if (strstart(p, "xenfb", &opts)) {
4739 xenfb_enabled = 1;
4740 } else if (!strstart(p, "none", &opts)) {
4741 invalid_vga:
4742 fprintf(stderr, "Unknown vga type: %s\n", p);
4743 exit(1);
4745 while (*opts) {
4746 const char *nextopt;
4748 if (strstart(opts, ",retrace=", &nextopt)) {
4749 opts = nextopt;
4750 if (strstart(opts, "dumb", &nextopt))
4751 vga_retrace_method = VGA_RETRACE_DUMB;
4752 else if (strstart(opts, "precise", &nextopt))
4753 vga_retrace_method = VGA_RETRACE_PRECISE;
4754 else goto invalid_vga;
4755 } else goto invalid_vga;
4756 opts = nextopt;
4760 #ifdef _WIN32
4761 static BOOL WINAPI qemu_ctrl_handler(DWORD type)
4763 exit(STATUS_CONTROL_C_EXIT);
4764 return TRUE;
4766 #endif
4768 int qemu_uuid_parse(const char *str, uint8_t *uuid)
4770 int ret;
4772 if(strlen(str) != 36)
4773 return -1;
4775 ret = sscanf(str, UUID_FMT, &uuid[0], &uuid[1], &uuid[2], &uuid[3],
4776 &uuid[4], &uuid[5], &uuid[6], &uuid[7], &uuid[8], &uuid[9],
4777 &uuid[10], &uuid[11], &uuid[12], &uuid[13], &uuid[14], &uuid[15]);
4779 if(ret != 16)
4780 return -1;
4782 #ifdef TARGET_I386
4783 smbios_add_field(1, offsetof(struct smbios_type_1, uuid), 16, uuid);
4784 #endif
4786 return 0;
4789 #define MAX_NET_CLIENTS 32
4791 #ifndef _WIN32
4793 static void termsig_handler(int signal)
4795 qemu_system_shutdown_request();
4798 static void sigchld_handler(int signal)
4800 waitpid(-1, NULL, WNOHANG);
4803 static void sighandler_setup(void)
4805 struct sigaction act;
4807 memset(&act, 0, sizeof(act));
4808 act.sa_handler = termsig_handler;
4809 sigaction(SIGINT, &act, NULL);
4810 sigaction(SIGHUP, &act, NULL);
4811 sigaction(SIGTERM, &act, NULL);
4813 act.sa_handler = sigchld_handler;
4814 act.sa_flags = SA_NOCLDSTOP;
4815 sigaction(SIGCHLD, &act, NULL);
4818 #endif
4820 #ifdef _WIN32
4821 /* Look for support files in the same directory as the executable. */
4822 static char *find_datadir(const char *argv0)
4824 char *p;
4825 char buf[MAX_PATH];
4826 DWORD len;
4828 len = GetModuleFileName(NULL, buf, sizeof(buf) - 1);
4829 if (len == 0) {
4830 return NULL;
4833 buf[len] = 0;
4834 p = buf + len - 1;
4835 while (p != buf && *p != '\\')
4836 p--;
4837 *p = 0;
4838 if (access(buf, R_OK) == 0) {
4839 return qemu_strdup(buf);
4841 return NULL;
4843 #else /* !_WIN32 */
4845 /* Find a likely location for support files using the location of the binary.
4846 For installed binaries this will be "$bindir/../share/qemu". When
4847 running from the build tree this will be "$bindir/../pc-bios". */
4848 #define SHARE_SUFFIX "/share/qemu"
4849 #define BUILD_SUFFIX "/pc-bios"
4850 static char *find_datadir(const char *argv0)
4852 char *dir;
4853 char *p = NULL;
4854 char *res;
4855 #ifdef PATH_MAX
4856 char buf[PATH_MAX];
4857 #endif
4858 size_t max_len;
4860 #if defined(__linux__)
4862 int len;
4863 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
4864 if (len > 0) {
4865 buf[len] = 0;
4866 p = buf;
4869 #elif defined(__FreeBSD__)
4871 int len;
4872 len = readlink("/proc/curproc/file", buf, sizeof(buf) - 1);
4873 if (len > 0) {
4874 buf[len] = 0;
4875 p = buf;
4878 #endif
4879 /* If we don't have any way of figuring out the actual executable
4880 location then try argv[0]. */
4881 if (!p) {
4882 #ifdef PATH_MAX
4883 p = buf;
4884 #endif
4885 p = realpath(argv0, p);
4886 if (!p) {
4887 return NULL;
4890 dir = dirname(p);
4891 dir = dirname(dir);
4893 max_len = strlen(dir) +
4894 MAX(strlen(SHARE_SUFFIX), strlen(BUILD_SUFFIX)) + 1;
4895 res = qemu_mallocz(max_len);
4896 snprintf(res, max_len, "%s%s", dir, SHARE_SUFFIX);
4897 if (access(res, R_OK)) {
4898 snprintf(res, max_len, "%s%s", dir, BUILD_SUFFIX);
4899 if (access(res, R_OK)) {
4900 qemu_free(res);
4901 res = NULL;
4904 #ifndef PATH_MAX
4905 free(p);
4906 #endif
4907 return res;
4909 #undef SHARE_SUFFIX
4910 #undef BUILD_SUFFIX
4911 #endif
4913 char *qemu_find_file(int type, const char *name)
4915 int len;
4916 const char *subdir;
4917 char *buf;
4919 /* If name contains path separators then try it as a straight path. */
4920 if ((strchr(name, '/') || strchr(name, '\\'))
4921 && access(name, R_OK) == 0) {
4922 return strdup(name);
4924 switch (type) {
4925 case QEMU_FILE_TYPE_BIOS:
4926 subdir = "";
4927 break;
4928 case QEMU_FILE_TYPE_KEYMAP:
4929 subdir = "keymaps/";
4930 break;
4931 default:
4932 abort();
4934 len = strlen(data_dir) + strlen(name) + strlen(subdir) + 2;
4935 buf = qemu_mallocz(len);
4936 snprintf(buf, len, "%s/%s%s", data_dir, subdir, name);
4937 if (access(buf, R_OK)) {
4938 qemu_free(buf);
4939 return NULL;
4941 return buf;
4944 int main(int argc, char **argv, char **envp)
4946 const char *gdbstub_dev = NULL;
4947 uint32_t boot_devices_bitmap = 0;
4948 int i;
4949 int snapshot, linux_boot, net_boot;
4950 const char *initrd_filename;
4951 const char *kernel_filename, *kernel_cmdline;
4952 const char *boot_devices = "";
4953 DisplayState *ds;
4954 DisplayChangeListener *dcl;
4955 int cyls, heads, secs, translation;
4956 const char *net_clients[MAX_NET_CLIENTS];
4957 int nb_net_clients;
4958 const char *bt_opts[MAX_BT_CMDLINE];
4959 int nb_bt_opts;
4960 int hda_index;
4961 int optind;
4962 const char *r, *optarg;
4963 CharDriverState *monitor_hd = NULL;
4964 const char *monitor_device;
4965 const char *serial_devices[MAX_SERIAL_PORTS];
4966 int serial_device_index;
4967 const char *parallel_devices[MAX_PARALLEL_PORTS];
4968 int parallel_device_index;
4969 const char *virtio_consoles[MAX_VIRTIO_CONSOLES];
4970 int virtio_console_index;
4971 const char *loadvm = NULL;
4972 QEMUMachine *machine;
4973 const char *cpu_model;
4974 const char *usb_devices[MAX_USB_CMDLINE];
4975 int usb_devices_index;
4976 #ifndef _WIN32
4977 int fds[2];
4978 #endif
4979 int tb_size;
4980 const char *pid_file = NULL;
4981 const char *incoming = NULL;
4982 #ifndef _WIN32
4983 int fd = 0;
4984 struct passwd *pwd = NULL;
4985 const char *chroot_dir = NULL;
4986 const char *run_as = NULL;
4987 #endif
4988 CPUState *env;
4989 int show_vnc_port = 0;
4991 qemu_cache_utils_init(envp);
4993 LIST_INIT (&vm_change_state_head);
4994 #ifndef _WIN32
4996 struct sigaction act;
4997 sigfillset(&act.sa_mask);
4998 act.sa_flags = 0;
4999 act.sa_handler = SIG_IGN;
5000 sigaction(SIGPIPE, &act, NULL);
5002 #else
5003 SetConsoleCtrlHandler(qemu_ctrl_handler, TRUE);
5004 /* Note: cpu_interrupt() is currently not SMP safe, so we force
5005 QEMU to run on a single CPU */
5007 HANDLE h;
5008 DWORD mask, smask;
5009 int i;
5010 h = GetCurrentProcess();
5011 if (GetProcessAffinityMask(h, &mask, &smask)) {
5012 for(i = 0; i < 32; i++) {
5013 if (mask & (1 << i))
5014 break;
5016 if (i != 32) {
5017 mask = 1 << i;
5018 SetProcessAffinityMask(h, mask);
5022 #endif
5024 module_call_init(MODULE_INIT_MACHINE);
5025 machine = find_default_machine();
5026 cpu_model = NULL;
5027 initrd_filename = NULL;
5028 ram_size = 0;
5029 snapshot = 0;
5030 kernel_filename = NULL;
5031 kernel_cmdline = "";
5032 cyls = heads = secs = 0;
5033 translation = BIOS_ATA_TRANSLATION_AUTO;
5034 monitor_device = "vc:80Cx24C";
5036 serial_devices[0] = "vc:80Cx24C";
5037 for(i = 1; i < MAX_SERIAL_PORTS; i++)
5038 serial_devices[i] = NULL;
5039 serial_device_index = 0;
5041 parallel_devices[0] = "vc:80Cx24C";
5042 for(i = 1; i < MAX_PARALLEL_PORTS; i++)
5043 parallel_devices[i] = NULL;
5044 parallel_device_index = 0;
5046 for(i = 0; i < MAX_VIRTIO_CONSOLES; i++)
5047 virtio_consoles[i] = NULL;
5048 virtio_console_index = 0;
5050 for (i = 0; i < MAX_NODES; i++) {
5051 node_mem[i] = 0;
5052 node_cpumask[i] = 0;
5055 usb_devices_index = 0;
5057 nb_net_clients = 0;
5058 nb_bt_opts = 0;
5059 nb_drives = 0;
5060 nb_drives_opt = 0;
5061 nb_numa_nodes = 0;
5062 hda_index = -1;
5064 nb_nics = 0;
5066 tb_size = 0;
5067 autostart= 1;
5069 register_watchdogs();
5071 optind = 1;
5072 for(;;) {
5073 if (optind >= argc)
5074 break;
5075 r = argv[optind];
5076 if (r[0] != '-') {
5077 hda_index = drive_add(argv[optind++], HD_ALIAS, 0);
5078 } else {
5079 const QEMUOption *popt;
5081 optind++;
5082 /* Treat --foo the same as -foo. */
5083 if (r[1] == '-')
5084 r++;
5085 popt = qemu_options;
5086 for(;;) {
5087 if (!popt->name) {
5088 fprintf(stderr, "%s: invalid option -- '%s'\n",
5089 argv[0], r);
5090 exit(1);
5092 if (!strcmp(popt->name, r + 1))
5093 break;
5094 popt++;
5096 if (popt->flags & HAS_ARG) {
5097 if (optind >= argc) {
5098 fprintf(stderr, "%s: option '%s' requires an argument\n",
5099 argv[0], r);
5100 exit(1);
5102 optarg = argv[optind++];
5103 } else {
5104 optarg = NULL;
5107 switch(popt->index) {
5108 case QEMU_OPTION_M:
5109 machine = find_machine(optarg);
5110 if (!machine) {
5111 QEMUMachine *m;
5112 printf("Supported machines are:\n");
5113 for(m = first_machine; m != NULL; m = m->next) {
5114 printf("%-10s %s%s\n",
5115 m->name, m->desc,
5116 m->is_default ? " (default)" : "");
5118 exit(*optarg != '?');
5120 break;
5121 case QEMU_OPTION_cpu:
5122 /* hw initialization will check this */
5123 if (*optarg == '?') {
5124 /* XXX: implement xxx_cpu_list for targets that still miss it */
5125 #if defined(cpu_list)
5126 cpu_list(stdout, &fprintf);
5127 #endif
5128 exit(0);
5129 } else {
5130 cpu_model = optarg;
5132 break;
5133 case QEMU_OPTION_initrd:
5134 initrd_filename = optarg;
5135 break;
5136 case QEMU_OPTION_hda:
5137 if (cyls == 0)
5138 hda_index = drive_add(optarg, HD_ALIAS, 0);
5139 else
5140 hda_index = drive_add(optarg, HD_ALIAS
5141 ",cyls=%d,heads=%d,secs=%d%s",
5142 0, cyls, heads, secs,
5143 translation == BIOS_ATA_TRANSLATION_LBA ?
5144 ",trans=lba" :
5145 translation == BIOS_ATA_TRANSLATION_NONE ?
5146 ",trans=none" : "");
5147 break;
5148 case QEMU_OPTION_hdb:
5149 case QEMU_OPTION_hdc:
5150 case QEMU_OPTION_hdd:
5151 drive_add(optarg, HD_ALIAS, popt->index - QEMU_OPTION_hda);
5152 break;
5153 case QEMU_OPTION_drive:
5154 drive_add(NULL, "%s", optarg);
5155 break;
5156 case QEMU_OPTION_mtdblock:
5157 drive_add(optarg, MTD_ALIAS);
5158 break;
5159 case QEMU_OPTION_sd:
5160 drive_add(optarg, SD_ALIAS);
5161 break;
5162 case QEMU_OPTION_pflash:
5163 drive_add(optarg, PFLASH_ALIAS);
5164 break;
5165 case QEMU_OPTION_snapshot:
5166 snapshot = 1;
5167 break;
5168 case QEMU_OPTION_hdachs:
5170 const char *p;
5171 p = optarg;
5172 cyls = strtol(p, (char **)&p, 0);
5173 if (cyls < 1 || cyls > 16383)
5174 goto chs_fail;
5175 if (*p != ',')
5176 goto chs_fail;
5177 p++;
5178 heads = strtol(p, (char **)&p, 0);
5179 if (heads < 1 || heads > 16)
5180 goto chs_fail;
5181 if (*p != ',')
5182 goto chs_fail;
5183 p++;
5184 secs = strtol(p, (char **)&p, 0);
5185 if (secs < 1 || secs > 63)
5186 goto chs_fail;
5187 if (*p == ',') {
5188 p++;
5189 if (!strcmp(p, "none"))
5190 translation = BIOS_ATA_TRANSLATION_NONE;
5191 else if (!strcmp(p, "lba"))
5192 translation = BIOS_ATA_TRANSLATION_LBA;
5193 else if (!strcmp(p, "auto"))
5194 translation = BIOS_ATA_TRANSLATION_AUTO;
5195 else
5196 goto chs_fail;
5197 } else if (*p != '\0') {
5198 chs_fail:
5199 fprintf(stderr, "qemu: invalid physical CHS format\n");
5200 exit(1);
5202 if (hda_index != -1)
5203 snprintf(drives_opt[hda_index].opt,
5204 sizeof(drives_opt[hda_index].opt),
5205 HD_ALIAS ",cyls=%d,heads=%d,secs=%d%s",
5206 0, cyls, heads, secs,
5207 translation == BIOS_ATA_TRANSLATION_LBA ?
5208 ",trans=lba" :
5209 translation == BIOS_ATA_TRANSLATION_NONE ?
5210 ",trans=none" : "");
5212 break;
5213 case QEMU_OPTION_numa:
5214 if (nb_numa_nodes >= MAX_NODES) {
5215 fprintf(stderr, "qemu: too many NUMA nodes\n");
5216 exit(1);
5218 numa_add(optarg);
5219 break;
5220 case QEMU_OPTION_nographic:
5221 display_type = DT_NOGRAPHIC;
5222 break;
5223 #ifdef CONFIG_CURSES
5224 case QEMU_OPTION_curses:
5225 display_type = DT_CURSES;
5226 break;
5227 #endif
5228 case QEMU_OPTION_portrait:
5229 graphic_rotate = 1;
5230 break;
5231 case QEMU_OPTION_kernel:
5232 kernel_filename = optarg;
5233 break;
5234 case QEMU_OPTION_append:
5235 kernel_cmdline = optarg;
5236 break;
5237 case QEMU_OPTION_cdrom:
5238 drive_add(optarg, CDROM_ALIAS);
5239 break;
5240 case QEMU_OPTION_boot:
5241 boot_devices = optarg;
5242 /* We just do some generic consistency checks */
5244 /* Could easily be extended to 64 devices if needed */
5245 const char *p;
5247 boot_devices_bitmap = 0;
5248 for (p = boot_devices; *p != '\0'; p++) {
5249 /* Allowed boot devices are:
5250 * a b : floppy disk drives
5251 * c ... f : IDE disk drives
5252 * g ... m : machine implementation dependant drives
5253 * n ... p : network devices
5254 * It's up to each machine implementation to check
5255 * if the given boot devices match the actual hardware
5256 * implementation and firmware features.
5258 if (*p < 'a' || *p > 'q') {
5259 fprintf(stderr, "Invalid boot device '%c'\n", *p);
5260 exit(1);
5262 if (boot_devices_bitmap & (1 << (*p - 'a'))) {
5263 fprintf(stderr,
5264 "Boot device '%c' was given twice\n",*p);
5265 exit(1);
5267 boot_devices_bitmap |= 1 << (*p - 'a');
5270 break;
5271 case QEMU_OPTION_fda:
5272 case QEMU_OPTION_fdb:
5273 drive_add(optarg, FD_ALIAS, popt->index - QEMU_OPTION_fda);
5274 break;
5275 #ifdef TARGET_I386
5276 case QEMU_OPTION_no_fd_bootchk:
5277 fd_bootchk = 0;
5278 break;
5279 #endif
5280 case QEMU_OPTION_net:
5281 if (nb_net_clients >= MAX_NET_CLIENTS) {
5282 fprintf(stderr, "qemu: too many network clients\n");
5283 exit(1);
5285 net_clients[nb_net_clients] = optarg;
5286 nb_net_clients++;
5287 break;
5288 #ifdef CONFIG_SLIRP
5289 case QEMU_OPTION_tftp:
5290 tftp_prefix = optarg;
5291 break;
5292 case QEMU_OPTION_bootp:
5293 bootp_filename = optarg;
5294 break;
5295 #ifndef _WIN32
5296 case QEMU_OPTION_smb:
5297 net_slirp_smb(optarg);
5298 break;
5299 #endif
5300 case QEMU_OPTION_redir:
5301 net_slirp_redir(NULL, optarg, NULL);
5302 break;
5303 #endif
5304 case QEMU_OPTION_bt:
5305 if (nb_bt_opts >= MAX_BT_CMDLINE) {
5306 fprintf(stderr, "qemu: too many bluetooth options\n");
5307 exit(1);
5309 bt_opts[nb_bt_opts++] = optarg;
5310 break;
5311 #ifdef HAS_AUDIO
5312 case QEMU_OPTION_audio_help:
5313 AUD_help ();
5314 exit (0);
5315 break;
5316 case QEMU_OPTION_soundhw:
5317 select_soundhw (optarg);
5318 break;
5319 #endif
5320 case QEMU_OPTION_h:
5321 help(0);
5322 break;
5323 case QEMU_OPTION_version:
5324 version();
5325 exit(0);
5326 break;
5327 case QEMU_OPTION_m: {
5328 uint64_t value;
5329 char *ptr;
5331 value = strtoul(optarg, &ptr, 10);
5332 switch (*ptr) {
5333 case 0: case 'M': case 'm':
5334 value <<= 20;
5335 break;
5336 case 'G': case 'g':
5337 value <<= 30;
5338 break;
5339 default:
5340 fprintf(stderr, "qemu: invalid ram size: %s\n", optarg);
5341 exit(1);
5344 /* On 32-bit hosts, QEMU is limited by virtual address space */
5345 if (value > (2047 << 20)
5346 #ifndef CONFIG_KQEMU
5347 && HOST_LONG_BITS == 32
5348 #endif
5350 fprintf(stderr, "qemu: at most 2047 MB RAM can be simulated\n");
5351 exit(1);
5353 if (value != (uint64_t)(ram_addr_t)value) {
5354 fprintf(stderr, "qemu: ram size too large\n");
5355 exit(1);
5357 ram_size = value;
5358 break;
5360 case QEMU_OPTION_d:
5362 int mask;
5363 const CPULogItem *item;
5365 mask = cpu_str_to_log_mask(optarg);
5366 if (!mask) {
5367 printf("Log items (comma separated):\n");
5368 for(item = cpu_log_items; item->mask != 0; item++) {
5369 printf("%-10s %s\n", item->name, item->help);
5371 exit(1);
5373 cpu_set_log(mask);
5375 break;
5376 case QEMU_OPTION_s:
5377 gdbstub_dev = "tcp::" DEFAULT_GDBSTUB_PORT;
5378 break;
5379 case QEMU_OPTION_gdb:
5380 gdbstub_dev = optarg;
5381 break;
5382 case QEMU_OPTION_L:
5383 data_dir = optarg;
5384 break;
5385 case QEMU_OPTION_bios:
5386 bios_name = optarg;
5387 break;
5388 case QEMU_OPTION_singlestep:
5389 singlestep = 1;
5390 break;
5391 case QEMU_OPTION_S:
5392 autostart = 0;
5393 break;
5394 #ifndef _WIN32
5395 case QEMU_OPTION_k:
5396 keyboard_layout = optarg;
5397 break;
5398 #endif
5399 case QEMU_OPTION_localtime:
5400 rtc_utc = 0;
5401 break;
5402 case QEMU_OPTION_vga:
5403 select_vgahw (optarg);
5404 break;
5405 #if defined(TARGET_PPC) || defined(TARGET_SPARC)
5406 case QEMU_OPTION_g:
5408 const char *p;
5409 int w, h, depth;
5410 p = optarg;
5411 w = strtol(p, (char **)&p, 10);
5412 if (w <= 0) {
5413 graphic_error:
5414 fprintf(stderr, "qemu: invalid resolution or depth\n");
5415 exit(1);
5417 if (*p != 'x')
5418 goto graphic_error;
5419 p++;
5420 h = strtol(p, (char **)&p, 10);
5421 if (h <= 0)
5422 goto graphic_error;
5423 if (*p == 'x') {
5424 p++;
5425 depth = strtol(p, (char **)&p, 10);
5426 if (depth != 8 && depth != 15 && depth != 16 &&
5427 depth != 24 && depth != 32)
5428 goto graphic_error;
5429 } else if (*p == '\0') {
5430 depth = graphic_depth;
5431 } else {
5432 goto graphic_error;
5435 graphic_width = w;
5436 graphic_height = h;
5437 graphic_depth = depth;
5439 break;
5440 #endif
5441 case QEMU_OPTION_echr:
5443 char *r;
5444 term_escape_char = strtol(optarg, &r, 0);
5445 if (r == optarg)
5446 printf("Bad argument to echr\n");
5447 break;
5449 case QEMU_OPTION_monitor:
5450 monitor_device = optarg;
5451 break;
5452 case QEMU_OPTION_serial:
5453 if (serial_device_index >= MAX_SERIAL_PORTS) {
5454 fprintf(stderr, "qemu: too many serial ports\n");
5455 exit(1);
5457 serial_devices[serial_device_index] = optarg;
5458 serial_device_index++;
5459 break;
5460 case QEMU_OPTION_watchdog:
5461 i = select_watchdog(optarg);
5462 if (i > 0)
5463 exit (i == 1 ? 1 : 0);
5464 break;
5465 case QEMU_OPTION_watchdog_action:
5466 if (select_watchdog_action(optarg) == -1) {
5467 fprintf(stderr, "Unknown -watchdog-action parameter\n");
5468 exit(1);
5470 break;
5471 case QEMU_OPTION_virtiocon:
5472 if (virtio_console_index >= MAX_VIRTIO_CONSOLES) {
5473 fprintf(stderr, "qemu: too many virtio consoles\n");
5474 exit(1);
5476 virtio_consoles[virtio_console_index] = optarg;
5477 virtio_console_index++;
5478 break;
5479 case QEMU_OPTION_parallel:
5480 if (parallel_device_index >= MAX_PARALLEL_PORTS) {
5481 fprintf(stderr, "qemu: too many parallel ports\n");
5482 exit(1);
5484 parallel_devices[parallel_device_index] = optarg;
5485 parallel_device_index++;
5486 break;
5487 case QEMU_OPTION_loadvm:
5488 loadvm = optarg;
5489 break;
5490 case QEMU_OPTION_full_screen:
5491 full_screen = 1;
5492 break;
5493 #ifdef CONFIG_SDL
5494 case QEMU_OPTION_no_frame:
5495 no_frame = 1;
5496 break;
5497 case QEMU_OPTION_alt_grab:
5498 alt_grab = 1;
5499 break;
5500 case QEMU_OPTION_no_quit:
5501 no_quit = 1;
5502 break;
5503 case QEMU_OPTION_sdl:
5504 display_type = DT_SDL;
5505 break;
5506 #endif
5507 case QEMU_OPTION_pidfile:
5508 pid_file = optarg;
5509 break;
5510 #ifdef TARGET_I386
5511 case QEMU_OPTION_win2k_hack:
5512 win2k_install_hack = 1;
5513 break;
5514 case QEMU_OPTION_rtc_td_hack:
5515 rtc_td_hack = 1;
5516 break;
5517 case QEMU_OPTION_acpitable:
5518 if(acpi_table_add(optarg) < 0) {
5519 fprintf(stderr, "Wrong acpi table provided\n");
5520 exit(1);
5522 break;
5523 case QEMU_OPTION_smbios:
5524 if(smbios_entry_add(optarg) < 0) {
5525 fprintf(stderr, "Wrong smbios provided\n");
5526 exit(1);
5528 break;
5529 #endif
5530 #ifdef CONFIG_KQEMU
5531 case QEMU_OPTION_no_kqemu:
5532 kqemu_allowed = 0;
5533 break;
5534 case QEMU_OPTION_kernel_kqemu:
5535 kqemu_allowed = 2;
5536 break;
5537 #endif
5538 #ifdef CONFIG_KVM
5539 case QEMU_OPTION_enable_kvm:
5540 kvm_allowed = 1;
5541 #ifdef CONFIG_KQEMU
5542 kqemu_allowed = 0;
5543 #endif
5544 break;
5545 #endif
5546 case QEMU_OPTION_usb:
5547 usb_enabled = 1;
5548 break;
5549 case QEMU_OPTION_usbdevice:
5550 usb_enabled = 1;
5551 if (usb_devices_index >= MAX_USB_CMDLINE) {
5552 fprintf(stderr, "Too many USB devices\n");
5553 exit(1);
5555 usb_devices[usb_devices_index] = optarg;
5556 usb_devices_index++;
5557 break;
5558 case QEMU_OPTION_smp:
5559 smp_cpus = atoi(optarg);
5560 if (smp_cpus < 1) {
5561 fprintf(stderr, "Invalid number of CPUs\n");
5562 exit(1);
5564 break;
5565 case QEMU_OPTION_vnc:
5566 display_type = DT_VNC;
5567 vnc_display = optarg;
5568 break;
5569 #ifdef TARGET_I386
5570 case QEMU_OPTION_no_acpi:
5571 acpi_enabled = 0;
5572 break;
5573 case QEMU_OPTION_no_hpet:
5574 no_hpet = 1;
5575 break;
5576 case QEMU_OPTION_no_virtio_balloon:
5577 no_virtio_balloon = 1;
5578 break;
5579 #endif
5580 case QEMU_OPTION_no_reboot:
5581 no_reboot = 1;
5582 break;
5583 case QEMU_OPTION_no_shutdown:
5584 no_shutdown = 1;
5585 break;
5586 case QEMU_OPTION_show_cursor:
5587 cursor_hide = 0;
5588 break;
5589 case QEMU_OPTION_uuid:
5590 if(qemu_uuid_parse(optarg, qemu_uuid) < 0) {
5591 fprintf(stderr, "Fail to parse UUID string."
5592 " Wrong format.\n");
5593 exit(1);
5595 break;
5596 #ifndef _WIN32
5597 case QEMU_OPTION_daemonize:
5598 daemonize = 1;
5599 break;
5600 #endif
5601 case QEMU_OPTION_option_rom:
5602 if (nb_option_roms >= MAX_OPTION_ROMS) {
5603 fprintf(stderr, "Too many option ROMs\n");
5604 exit(1);
5606 option_rom[nb_option_roms] = optarg;
5607 nb_option_roms++;
5608 break;
5609 #if defined(TARGET_ARM) || defined(TARGET_M68K)
5610 case QEMU_OPTION_semihosting:
5611 semihosting_enabled = 1;
5612 break;
5613 #endif
5614 case QEMU_OPTION_name:
5615 qemu_name = optarg;
5616 break;
5617 #if defined(TARGET_SPARC) || defined(TARGET_PPC)
5618 case QEMU_OPTION_prom_env:
5619 if (nb_prom_envs >= MAX_PROM_ENVS) {
5620 fprintf(stderr, "Too many prom variables\n");
5621 exit(1);
5623 prom_envs[nb_prom_envs] = optarg;
5624 nb_prom_envs++;
5625 break;
5626 #endif
5627 #ifdef TARGET_ARM
5628 case QEMU_OPTION_old_param:
5629 old_param = 1;
5630 break;
5631 #endif
5632 case QEMU_OPTION_clock:
5633 configure_alarms(optarg);
5634 break;
5635 case QEMU_OPTION_startdate:
5637 struct tm tm;
5638 time_t rtc_start_date;
5639 if (!strcmp(optarg, "now")) {
5640 rtc_date_offset = -1;
5641 } else {
5642 if (sscanf(optarg, "%d-%d-%dT%d:%d:%d",
5643 &tm.tm_year,
5644 &tm.tm_mon,
5645 &tm.tm_mday,
5646 &tm.tm_hour,
5647 &tm.tm_min,
5648 &tm.tm_sec) == 6) {
5649 /* OK */
5650 } else if (sscanf(optarg, "%d-%d-%d",
5651 &tm.tm_year,
5652 &tm.tm_mon,
5653 &tm.tm_mday) == 3) {
5654 tm.tm_hour = 0;
5655 tm.tm_min = 0;
5656 tm.tm_sec = 0;
5657 } else {
5658 goto date_fail;
5660 tm.tm_year -= 1900;
5661 tm.tm_mon--;
5662 rtc_start_date = mktimegm(&tm);
5663 if (rtc_start_date == -1) {
5664 date_fail:
5665 fprintf(stderr, "Invalid date format. Valid format are:\n"
5666 "'now' or '2006-06-17T16:01:21' or '2006-06-17'\n");
5667 exit(1);
5669 rtc_date_offset = time(NULL) - rtc_start_date;
5672 break;
5673 case QEMU_OPTION_tb_size:
5674 tb_size = strtol(optarg, NULL, 0);
5675 if (tb_size < 0)
5676 tb_size = 0;
5677 break;
5678 case QEMU_OPTION_icount:
5679 use_icount = 1;
5680 if (strcmp(optarg, "auto") == 0) {
5681 icount_time_shift = -1;
5682 } else {
5683 icount_time_shift = strtol(optarg, NULL, 0);
5685 break;
5686 case QEMU_OPTION_incoming:
5687 incoming = optarg;
5688 break;
5689 #ifndef _WIN32
5690 case QEMU_OPTION_chroot:
5691 chroot_dir = optarg;
5692 break;
5693 case QEMU_OPTION_runas:
5694 run_as = optarg;
5695 break;
5696 #endif
5697 #ifdef CONFIG_XEN
5698 case QEMU_OPTION_xen_domid:
5699 xen_domid = atoi(optarg);
5700 break;
5701 case QEMU_OPTION_xen_create:
5702 xen_mode = XEN_CREATE;
5703 break;
5704 case QEMU_OPTION_xen_attach:
5705 xen_mode = XEN_ATTACH;
5706 break;
5707 #endif
5712 /* If no data_dir is specified then try to find it relative to the
5713 executable path. */
5714 if (!data_dir) {
5715 data_dir = find_datadir(argv[0]);
5717 /* If all else fails use the install patch specified when building. */
5718 if (!data_dir) {
5719 data_dir = CONFIG_QEMU_SHAREDIR;
5722 #if defined(CONFIG_KVM) && defined(CONFIG_KQEMU)
5723 if (kvm_allowed && kqemu_allowed) {
5724 fprintf(stderr,
5725 "You can not enable both KVM and kqemu at the same time\n");
5726 exit(1);
5728 #endif
5730 machine->max_cpus = machine->max_cpus ?: 1; /* Default to UP */
5731 if (smp_cpus > machine->max_cpus) {
5732 fprintf(stderr, "Number of SMP cpus requested (%d), exceeds max cpus "
5733 "supported by machine `%s' (%d)\n", smp_cpus, machine->name,
5734 machine->max_cpus);
5735 exit(1);
5738 if (display_type == DT_NOGRAPHIC) {
5739 if (serial_device_index == 0)
5740 serial_devices[0] = "stdio";
5741 if (parallel_device_index == 0)
5742 parallel_devices[0] = "null";
5743 if (strncmp(monitor_device, "vc", 2) == 0)
5744 monitor_device = "stdio";
5747 #ifndef _WIN32
5748 if (daemonize) {
5749 pid_t pid;
5751 if (pipe(fds) == -1)
5752 exit(1);
5754 pid = fork();
5755 if (pid > 0) {
5756 uint8_t status;
5757 ssize_t len;
5759 close(fds[1]);
5761 again:
5762 len = read(fds[0], &status, 1);
5763 if (len == -1 && (errno == EINTR))
5764 goto again;
5766 if (len != 1)
5767 exit(1);
5768 else if (status == 1) {
5769 fprintf(stderr, "Could not acquire pidfile\n");
5770 exit(1);
5771 } else
5772 exit(0);
5773 } else if (pid < 0)
5774 exit(1);
5776 setsid();
5778 pid = fork();
5779 if (pid > 0)
5780 exit(0);
5781 else if (pid < 0)
5782 exit(1);
5784 umask(027);
5786 signal(SIGTSTP, SIG_IGN);
5787 signal(SIGTTOU, SIG_IGN);
5788 signal(SIGTTIN, SIG_IGN);
5791 if (pid_file && qemu_create_pidfile(pid_file) != 0) {
5792 if (daemonize) {
5793 uint8_t status = 1;
5794 write(fds[1], &status, 1);
5795 } else
5796 fprintf(stderr, "Could not acquire pid file\n");
5797 exit(1);
5799 #endif
5801 #ifdef CONFIG_KQEMU
5802 if (smp_cpus > 1)
5803 kqemu_allowed = 0;
5804 #endif
5805 if (qemu_init_main_loop()) {
5806 fprintf(stderr, "qemu_init_main_loop failed\n");
5807 exit(1);
5809 linux_boot = (kernel_filename != NULL);
5811 if (!linux_boot && *kernel_cmdline != '\0') {
5812 fprintf(stderr, "-append only allowed with -kernel option\n");
5813 exit(1);
5816 if (!linux_boot && initrd_filename != NULL) {
5817 fprintf(stderr, "-initrd only allowed with -kernel option\n");
5818 exit(1);
5821 /* boot to floppy or the default cd if no hard disk defined yet */
5822 if (!boot_devices[0]) {
5823 boot_devices = "cad";
5825 setvbuf(stdout, NULL, _IOLBF, 0);
5827 init_timers();
5828 if (init_timer_alarm() < 0) {
5829 fprintf(stderr, "could not initialize alarm timer\n");
5830 exit(1);
5832 if (use_icount && icount_time_shift < 0) {
5833 use_icount = 2;
5834 /* 125MIPS seems a reasonable initial guess at the guest speed.
5835 It will be corrected fairly quickly anyway. */
5836 icount_time_shift = 3;
5837 init_icount_adjust();
5840 #ifdef _WIN32
5841 socket_init();
5842 #endif
5844 /* init network clients */
5845 if (nb_net_clients == 0) {
5846 /* if no clients, we use a default config */
5847 net_clients[nb_net_clients++] = "nic";
5848 #ifdef CONFIG_SLIRP
5849 net_clients[nb_net_clients++] = "user";
5850 #endif
5853 for(i = 0;i < nb_net_clients; i++) {
5854 if (net_client_parse(net_clients[i]) < 0)
5855 exit(1);
5858 net_boot = (boot_devices_bitmap >> ('n' - 'a')) & 0xF;
5859 net_set_boot_mask(net_boot);
5861 net_client_check();
5863 /* init the bluetooth world */
5864 for (i = 0; i < nb_bt_opts; i++)
5865 if (bt_parse(bt_opts[i]))
5866 exit(1);
5868 /* init the memory */
5869 if (ram_size == 0)
5870 ram_size = DEFAULT_RAM_SIZE * 1024 * 1024;
5872 #ifdef CONFIG_KQEMU
5873 /* FIXME: This is a nasty hack because kqemu can't cope with dynamic
5874 guest ram allocation. It needs to go away. */
5875 if (kqemu_allowed) {
5876 kqemu_phys_ram_size = ram_size + 8 * 1024 * 1024 + 4 * 1024 * 1024;
5877 kqemu_phys_ram_base = qemu_vmalloc(kqemu_phys_ram_size);
5878 if (!kqemu_phys_ram_base) {
5879 fprintf(stderr, "Could not allocate physical memory\n");
5880 exit(1);
5883 #endif
5885 /* init the dynamic translator */
5886 cpu_exec_init_all(tb_size * 1024 * 1024);
5888 bdrv_init();
5890 /* we always create the cdrom drive, even if no disk is there */
5892 if (nb_drives_opt < MAX_DRIVES)
5893 drive_add(NULL, CDROM_ALIAS);
5895 /* we always create at least one floppy */
5897 if (nb_drives_opt < MAX_DRIVES)
5898 drive_add(NULL, FD_ALIAS, 0);
5900 /* we always create one sd slot, even if no card is in it */
5902 if (nb_drives_opt < MAX_DRIVES)
5903 drive_add(NULL, SD_ALIAS);
5905 /* open the virtual block devices */
5907 for(i = 0; i < nb_drives_opt; i++)
5908 if (drive_init(&drives_opt[i], snapshot, machine) == -1)
5909 exit(1);
5911 register_savevm("timer", 0, 2, timer_save, timer_load, NULL);
5912 register_savevm_live("ram", 0, 3, ram_save_live, NULL, ram_load, NULL);
5914 #ifndef _WIN32
5915 /* must be after terminal init, SDL library changes signal handlers */
5916 sighandler_setup();
5917 #endif
5919 /* Maintain compatibility with multiple stdio monitors */
5920 if (!strcmp(monitor_device,"stdio")) {
5921 for (i = 0; i < MAX_SERIAL_PORTS; i++) {
5922 const char *devname = serial_devices[i];
5923 if (devname && !strcmp(devname,"mon:stdio")) {
5924 monitor_device = NULL;
5925 break;
5926 } else if (devname && !strcmp(devname,"stdio")) {
5927 monitor_device = NULL;
5928 serial_devices[i] = "mon:stdio";
5929 break;
5934 if (nb_numa_nodes > 0) {
5935 int i;
5937 if (nb_numa_nodes > smp_cpus) {
5938 nb_numa_nodes = smp_cpus;
5941 /* If no memory size if given for any node, assume the default case
5942 * and distribute the available memory equally across all nodes
5944 for (i = 0; i < nb_numa_nodes; i++) {
5945 if (node_mem[i] != 0)
5946 break;
5948 if (i == nb_numa_nodes) {
5949 uint64_t usedmem = 0;
5951 /* On Linux, the each node's border has to be 8MB aligned,
5952 * the final node gets the rest.
5954 for (i = 0; i < nb_numa_nodes - 1; i++) {
5955 node_mem[i] = (ram_size / nb_numa_nodes) & ~((1 << 23UL) - 1);
5956 usedmem += node_mem[i];
5958 node_mem[i] = ram_size - usedmem;
5961 for (i = 0; i < nb_numa_nodes; i++) {
5962 if (node_cpumask[i] != 0)
5963 break;
5965 /* assigning the VCPUs round-robin is easier to implement, guest OSes
5966 * must cope with this anyway, because there are BIOSes out there in
5967 * real machines which also use this scheme.
5969 if (i == nb_numa_nodes) {
5970 for (i = 0; i < smp_cpus; i++) {
5971 node_cpumask[i % nb_numa_nodes] |= 1 << i;
5976 if (kvm_enabled()) {
5977 int ret;
5979 ret = kvm_init(smp_cpus);
5980 if (ret < 0) {
5981 fprintf(stderr, "failed to initialize KVM\n");
5982 exit(1);
5986 if (monitor_device) {
5987 monitor_hd = qemu_chr_open("monitor", monitor_device, NULL);
5988 if (!monitor_hd) {
5989 fprintf(stderr, "qemu: could not open monitor device '%s'\n", monitor_device);
5990 exit(1);
5994 for(i = 0; i < MAX_SERIAL_PORTS; i++) {
5995 const char *devname = serial_devices[i];
5996 if (devname && strcmp(devname, "none")) {
5997 char label[32];
5998 snprintf(label, sizeof(label), "serial%d", i);
5999 serial_hds[i] = qemu_chr_open(label, devname, NULL);
6000 if (!serial_hds[i]) {
6001 fprintf(stderr, "qemu: could not open serial device '%s'\n",
6002 devname);
6003 exit(1);
6008 for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
6009 const char *devname = parallel_devices[i];
6010 if (devname && strcmp(devname, "none")) {
6011 char label[32];
6012 snprintf(label, sizeof(label), "parallel%d", i);
6013 parallel_hds[i] = qemu_chr_open(label, devname, NULL);
6014 if (!parallel_hds[i]) {
6015 fprintf(stderr, "qemu: could not open parallel device '%s'\n",
6016 devname);
6017 exit(1);
6022 for(i = 0; i < MAX_VIRTIO_CONSOLES; i++) {
6023 const char *devname = virtio_consoles[i];
6024 if (devname && strcmp(devname, "none")) {
6025 char label[32];
6026 snprintf(label, sizeof(label), "virtcon%d", i);
6027 virtcon_hds[i] = qemu_chr_open(label, devname, NULL);
6028 if (!virtcon_hds[i]) {
6029 fprintf(stderr, "qemu: could not open virtio console '%s'\n",
6030 devname);
6031 exit(1);
6036 module_call_init(MODULE_INIT_DEVICE);
6038 machine->init(ram_size, boot_devices,
6039 kernel_filename, kernel_cmdline, initrd_filename, cpu_model);
6042 for (env = first_cpu; env != NULL; env = env->next_cpu) {
6043 for (i = 0; i < nb_numa_nodes; i++) {
6044 if (node_cpumask[i] & (1 << env->cpu_index)) {
6045 env->numa_node = i;
6050 current_machine = machine;
6052 /* Set KVM's vcpu state to qemu's initial CPUState. */
6053 if (kvm_enabled()) {
6054 int ret;
6056 ret = kvm_sync_vcpus();
6057 if (ret < 0) {
6058 fprintf(stderr, "failed to initialize vcpus\n");
6059 exit(1);
6063 /* init USB devices */
6064 if (usb_enabled) {
6065 for(i = 0; i < usb_devices_index; i++) {
6066 if (usb_device_add(usb_devices[i], 0) < 0) {
6067 fprintf(stderr, "Warning: could not add USB device %s\n",
6068 usb_devices[i]);
6073 if (!display_state)
6074 dumb_display_init();
6075 /* just use the first displaystate for the moment */
6076 ds = display_state;
6078 if (display_type == DT_DEFAULT) {
6079 #if defined(CONFIG_SDL) || defined(CONFIG_COCOA)
6080 display_type = DT_SDL;
6081 #else
6082 display_type = DT_VNC;
6083 vnc_display = "localhost:0,to=99";
6084 show_vnc_port = 1;
6085 #endif
6089 switch (display_type) {
6090 case DT_NOGRAPHIC:
6091 break;
6092 #if defined(CONFIG_CURSES)
6093 case DT_CURSES:
6094 curses_display_init(ds, full_screen);
6095 break;
6096 #endif
6097 #if defined(CONFIG_SDL)
6098 case DT_SDL:
6099 sdl_display_init(ds, full_screen, no_frame);
6100 break;
6101 #elif defined(CONFIG_COCOA)
6102 case DT_SDL:
6103 cocoa_display_init(ds, full_screen);
6104 break;
6105 #endif
6106 case DT_VNC:
6107 vnc_display_init(ds);
6108 if (vnc_display_open(ds, vnc_display) < 0)
6109 exit(1);
6111 if (show_vnc_port) {
6112 printf("VNC server running on `%s'\n", vnc_display_local_addr(ds));
6114 break;
6115 default:
6116 break;
6118 dpy_resize(ds);
6120 dcl = ds->listeners;
6121 while (dcl != NULL) {
6122 if (dcl->dpy_refresh != NULL) {
6123 ds->gui_timer = qemu_new_timer(rt_clock, gui_update, ds);
6124 qemu_mod_timer(ds->gui_timer, qemu_get_clock(rt_clock));
6126 dcl = dcl->next;
6129 if (display_type == DT_NOGRAPHIC || display_type == DT_VNC) {
6130 nographic_timer = qemu_new_timer(rt_clock, nographic_update, NULL);
6131 qemu_mod_timer(nographic_timer, qemu_get_clock(rt_clock));
6134 text_consoles_set_display(display_state);
6135 qemu_chr_initial_reset();
6137 if (monitor_device && monitor_hd)
6138 monitor_init(monitor_hd, MONITOR_USE_READLINE | MONITOR_IS_DEFAULT);
6140 for(i = 0; i < MAX_SERIAL_PORTS; i++) {
6141 const char *devname = serial_devices[i];
6142 if (devname && strcmp(devname, "none")) {
6143 if (strstart(devname, "vc", 0))
6144 qemu_chr_printf(serial_hds[i], "serial%d console\r\n", i);
6148 for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
6149 const char *devname = parallel_devices[i];
6150 if (devname && strcmp(devname, "none")) {
6151 if (strstart(devname, "vc", 0))
6152 qemu_chr_printf(parallel_hds[i], "parallel%d console\r\n", i);
6156 for(i = 0; i < MAX_VIRTIO_CONSOLES; i++) {
6157 const char *devname = virtio_consoles[i];
6158 if (virtcon_hds[i] && devname) {
6159 if (strstart(devname, "vc", 0))
6160 qemu_chr_printf(virtcon_hds[i], "virtio console%d\r\n", i);
6164 if (gdbstub_dev && gdbserver_start(gdbstub_dev) < 0) {
6165 fprintf(stderr, "qemu: could not open gdbserver on device '%s'\n",
6166 gdbstub_dev);
6167 exit(1);
6170 if (loadvm)
6171 do_loadvm(cur_mon, loadvm);
6173 if (incoming) {
6174 autostart = 0; /* fixme how to deal with -daemonize */
6175 qemu_start_incoming_migration(incoming);
6178 if (autostart)
6179 vm_start();
6181 #ifndef _WIN32
6182 if (daemonize) {
6183 uint8_t status = 0;
6184 ssize_t len;
6186 again1:
6187 len = write(fds[1], &status, 1);
6188 if (len == -1 && (errno == EINTR))
6189 goto again1;
6191 if (len != 1)
6192 exit(1);
6194 chdir("/");
6195 TFR(fd = open("/dev/null", O_RDWR));
6196 if (fd == -1)
6197 exit(1);
6200 if (run_as) {
6201 pwd = getpwnam(run_as);
6202 if (!pwd) {
6203 fprintf(stderr, "User \"%s\" doesn't exist\n", run_as);
6204 exit(1);
6208 if (chroot_dir) {
6209 if (chroot(chroot_dir) < 0) {
6210 fprintf(stderr, "chroot failed\n");
6211 exit(1);
6213 chdir("/");
6216 if (run_as) {
6217 if (setgid(pwd->pw_gid) < 0) {
6218 fprintf(stderr, "Failed to setgid(%d)\n", pwd->pw_gid);
6219 exit(1);
6221 if (setuid(pwd->pw_uid) < 0) {
6222 fprintf(stderr, "Failed to setuid(%d)\n", pwd->pw_uid);
6223 exit(1);
6225 if (setuid(0) != -1) {
6226 fprintf(stderr, "Dropping privileges failed\n");
6227 exit(1);
6231 if (daemonize) {
6232 dup2(fd, 0);
6233 dup2(fd, 1);
6234 dup2(fd, 2);
6236 close(fd);
6238 #endif
6240 main_loop();
6241 quit_timers();
6242 net_cleanup();
6244 return 0;