qcow2: Make cache=writethrough default
[qemu-kvm/fedora.git] / vl.c
blob5189ad537e939332577d3248c5eb8a38acc3e992
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 #if defined(__APPLE__) || defined(main)
117 #include <SDL.h>
118 int qemu_main(int argc, char **argv, char **envp);
119 int main(int argc, char **argv)
121 return 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 #include "slirp/libslirp.h"
170 //#define DEBUG_UNUSED_IOPORT
171 //#define DEBUG_IOPORT
172 //#define DEBUG_NET
173 //#define DEBUG_SLIRP
176 #ifdef DEBUG_IOPORT
177 # define LOG_IOPORT(...) qemu_log_mask(CPU_LOG_IOPORT, ## __VA_ARGS__)
178 #else
179 # define LOG_IOPORT(...) do { } while (0)
180 #endif
182 #define DEFAULT_RAM_SIZE 128
184 /* Max number of USB devices that can be specified on the commandline. */
185 #define MAX_USB_CMDLINE 8
187 /* Max number of bluetooth switches on the commandline. */
188 #define MAX_BT_CMDLINE 10
190 /* XXX: use a two level table to limit memory usage */
191 #define MAX_IOPORTS 65536
193 static const char *data_dir;
194 const char *bios_name = NULL;
195 static void *ioport_opaque[MAX_IOPORTS];
196 static IOPortReadFunc *ioport_read_table[3][MAX_IOPORTS];
197 static IOPortWriteFunc *ioport_write_table[3][MAX_IOPORTS];
198 /* Note: drives_table[MAX_DRIVES] is a dummy block driver if none available
199 to store the VM snapshots */
200 DriveInfo drives_table[MAX_DRIVES+1];
201 int nb_drives;
202 enum vga_retrace_method vga_retrace_method = VGA_RETRACE_DUMB;
203 static DisplayState *display_state;
204 DisplayType display_type = DT_DEFAULT;
205 const char* keyboard_layout = NULL;
206 int64_t ticks_per_sec;
207 ram_addr_t ram_size;
208 int nb_nics;
209 NICInfo nd_table[MAX_NICS];
210 int vm_running;
211 static int autostart;
212 static int rtc_utc = 1;
213 static int rtc_date_offset = -1; /* -1 means no change */
214 int cirrus_vga_enabled = 1;
215 int std_vga_enabled = 0;
216 int vmsvga_enabled = 0;
217 int xenfb_enabled = 0;
218 #ifdef TARGET_SPARC
219 int graphic_width = 1024;
220 int graphic_height = 768;
221 int graphic_depth = 8;
222 #else
223 int graphic_width = 800;
224 int graphic_height = 600;
225 int graphic_depth = 15;
226 #endif
227 static int full_screen = 0;
228 #ifdef CONFIG_SDL
229 static int no_frame = 0;
230 #endif
231 int no_quit = 0;
232 CharDriverState *serial_hds[MAX_SERIAL_PORTS];
233 CharDriverState *parallel_hds[MAX_PARALLEL_PORTS];
234 CharDriverState *virtcon_hds[MAX_VIRTIO_CONSOLES];
235 #ifdef TARGET_I386
236 int win2k_install_hack = 0;
237 int rtc_td_hack = 0;
238 #endif
239 int usb_enabled = 0;
240 int singlestep = 0;
241 int smp_cpus = 1;
242 const char *vnc_display;
243 int acpi_enabled = 1;
244 int no_hpet = 0;
245 int virtio_balloon = 1;
246 const char *virtio_balloon_devaddr;
247 int fd_bootchk = 1;
248 int no_reboot = 0;
249 int no_shutdown = 0;
250 int cursor_hide = 1;
251 int graphic_rotate = 0;
252 #ifndef _WIN32
253 int daemonize = 0;
254 #endif
255 WatchdogTimerModel *watchdog = NULL;
256 int watchdog_action = WDT_RESET;
257 const char *option_rom[MAX_OPTION_ROMS];
258 int nb_option_roms;
259 int semihosting_enabled = 0;
260 #ifdef TARGET_ARM
261 int old_param = 0;
262 #endif
263 const char *qemu_name;
264 int alt_grab = 0;
265 #if defined(TARGET_SPARC) || defined(TARGET_PPC)
266 unsigned int nb_prom_envs = 0;
267 const char *prom_envs[MAX_PROM_ENVS];
268 #endif
269 int nb_drives_opt;
270 struct drive_opt drives_opt[MAX_DRIVES];
272 int nb_numa_nodes;
273 uint64_t node_mem[MAX_NODES];
274 uint64_t node_cpumask[MAX_NODES];
276 static CPUState *cur_cpu;
277 static CPUState *next_cpu;
278 static int timer_alarm_pending = 1;
279 /* Conversion factor from emulated instructions to virtual clock ticks. */
280 static int icount_time_shift;
281 /* Arbitrarily pick 1MIPS as the minimum allowable speed. */
282 #define MAX_ICOUNT_SHIFT 10
283 /* Compensate for varying guest execution speed. */
284 static int64_t qemu_icount_bias;
285 static QEMUTimer *icount_rt_timer;
286 static QEMUTimer *icount_vm_timer;
287 static QEMUTimer *nographic_timer;
289 uint8_t qemu_uuid[16];
291 /***********************************************************/
292 /* x86 ISA bus support */
294 target_phys_addr_t isa_mem_base = 0;
295 PicState2 *isa_pic;
297 static IOPortReadFunc default_ioport_readb, default_ioport_readw, default_ioport_readl;
298 static IOPortWriteFunc default_ioport_writeb, default_ioport_writew, default_ioport_writel;
300 static uint32_t ioport_read(int index, uint32_t address)
302 static IOPortReadFunc *default_func[3] = {
303 default_ioport_readb,
304 default_ioport_readw,
305 default_ioport_readl
307 IOPortReadFunc *func = ioport_read_table[index][address];
308 if (!func)
309 func = default_func[index];
310 return func(ioport_opaque[address], address);
313 static void ioport_write(int index, uint32_t address, uint32_t data)
315 static IOPortWriteFunc *default_func[3] = {
316 default_ioport_writeb,
317 default_ioport_writew,
318 default_ioport_writel
320 IOPortWriteFunc *func = ioport_write_table[index][address];
321 if (!func)
322 func = default_func[index];
323 func(ioport_opaque[address], address, data);
326 static uint32_t default_ioport_readb(void *opaque, uint32_t address)
328 #ifdef DEBUG_UNUSED_IOPORT
329 fprintf(stderr, "unused inb: port=0x%04x\n", address);
330 #endif
331 return 0xff;
334 static void default_ioport_writeb(void *opaque, uint32_t address, uint32_t data)
336 #ifdef DEBUG_UNUSED_IOPORT
337 fprintf(stderr, "unused outb: port=0x%04x data=0x%02x\n", address, data);
338 #endif
341 /* default is to make two byte accesses */
342 static uint32_t default_ioport_readw(void *opaque, uint32_t address)
344 uint32_t data;
345 data = ioport_read(0, address);
346 address = (address + 1) & (MAX_IOPORTS - 1);
347 data |= ioport_read(0, address) << 8;
348 return data;
351 static void default_ioport_writew(void *opaque, uint32_t address, uint32_t data)
353 ioport_write(0, address, data & 0xff);
354 address = (address + 1) & (MAX_IOPORTS - 1);
355 ioport_write(0, address, (data >> 8) & 0xff);
358 static uint32_t default_ioport_readl(void *opaque, uint32_t address)
360 #ifdef DEBUG_UNUSED_IOPORT
361 fprintf(stderr, "unused inl: port=0x%04x\n", address);
362 #endif
363 return 0xffffffff;
366 static void default_ioport_writel(void *opaque, uint32_t address, uint32_t data)
368 #ifdef DEBUG_UNUSED_IOPORT
369 fprintf(stderr, "unused outl: port=0x%04x data=0x%02x\n", address, data);
370 #endif
373 /* size is the word size in byte */
374 int register_ioport_read(int start, int length, int size,
375 IOPortReadFunc *func, void *opaque)
377 int i, bsize;
379 if (size == 1) {
380 bsize = 0;
381 } else if (size == 2) {
382 bsize = 1;
383 } else if (size == 4) {
384 bsize = 2;
385 } else {
386 hw_error("register_ioport_read: invalid size");
387 return -1;
389 for(i = start; i < start + length; i += size) {
390 ioport_read_table[bsize][i] = func;
391 if (ioport_opaque[i] != NULL && ioport_opaque[i] != opaque)
392 hw_error("register_ioport_read: invalid opaque");
393 ioport_opaque[i] = opaque;
395 return 0;
398 /* size is the word size in byte */
399 int register_ioport_write(int start, int length, int size,
400 IOPortWriteFunc *func, void *opaque)
402 int i, bsize;
404 if (size == 1) {
405 bsize = 0;
406 } else if (size == 2) {
407 bsize = 1;
408 } else if (size == 4) {
409 bsize = 2;
410 } else {
411 hw_error("register_ioport_write: invalid size");
412 return -1;
414 for(i = start; i < start + length; i += size) {
415 ioport_write_table[bsize][i] = func;
416 if (ioport_opaque[i] != NULL && ioport_opaque[i] != opaque)
417 hw_error("register_ioport_write: invalid opaque");
418 ioport_opaque[i] = opaque;
420 return 0;
423 void isa_unassign_ioport(int start, int length)
425 int i;
427 for(i = start; i < start + length; i++) {
428 ioport_read_table[0][i] = default_ioport_readb;
429 ioport_read_table[1][i] = default_ioport_readw;
430 ioport_read_table[2][i] = default_ioport_readl;
432 ioport_write_table[0][i] = default_ioport_writeb;
433 ioport_write_table[1][i] = default_ioport_writew;
434 ioport_write_table[2][i] = default_ioport_writel;
436 ioport_opaque[i] = NULL;
440 /***********************************************************/
442 void cpu_outb(CPUState *env, int addr, int val)
444 LOG_IOPORT("outb: %04x %02x\n", addr, val);
445 ioport_write(0, addr, val);
446 #ifdef CONFIG_KQEMU
447 if (env)
448 env->last_io_time = cpu_get_time_fast();
449 #endif
452 void cpu_outw(CPUState *env, int addr, int val)
454 LOG_IOPORT("outw: %04x %04x\n", addr, val);
455 ioport_write(1, addr, val);
456 #ifdef CONFIG_KQEMU
457 if (env)
458 env->last_io_time = cpu_get_time_fast();
459 #endif
462 void cpu_outl(CPUState *env, int addr, int val)
464 LOG_IOPORT("outl: %04x %08x\n", addr, val);
465 ioport_write(2, addr, val);
466 #ifdef CONFIG_KQEMU
467 if (env)
468 env->last_io_time = cpu_get_time_fast();
469 #endif
472 int cpu_inb(CPUState *env, int addr)
474 int val;
475 val = ioport_read(0, addr);
476 LOG_IOPORT("inb : %04x %02x\n", addr, val);
477 #ifdef CONFIG_KQEMU
478 if (env)
479 env->last_io_time = cpu_get_time_fast();
480 #endif
481 return val;
484 int cpu_inw(CPUState *env, int addr)
486 int val;
487 val = ioport_read(1, addr);
488 LOG_IOPORT("inw : %04x %04x\n", addr, val);
489 #ifdef CONFIG_KQEMU
490 if (env)
491 env->last_io_time = cpu_get_time_fast();
492 #endif
493 return val;
496 int cpu_inl(CPUState *env, int addr)
498 int val;
499 val = ioport_read(2, addr);
500 LOG_IOPORT("inl : %04x %08x\n", addr, val);
501 #ifdef CONFIG_KQEMU
502 if (env)
503 env->last_io_time = cpu_get_time_fast();
504 #endif
505 return val;
508 /***********************************************************/
509 void hw_error(const char *fmt, ...)
511 va_list ap;
512 CPUState *env;
514 va_start(ap, fmt);
515 fprintf(stderr, "qemu: hardware error: ");
516 vfprintf(stderr, fmt, ap);
517 fprintf(stderr, "\n");
518 for(env = first_cpu; env != NULL; env = env->next_cpu) {
519 fprintf(stderr, "CPU #%d:\n", env->cpu_index);
520 #ifdef TARGET_I386
521 cpu_dump_state(env, stderr, fprintf, X86_DUMP_FPU);
522 #else
523 cpu_dump_state(env, stderr, fprintf, 0);
524 #endif
526 va_end(ap);
527 abort();
530 /***************/
531 /* ballooning */
533 static QEMUBalloonEvent *qemu_balloon_event;
534 void *qemu_balloon_event_opaque;
536 void qemu_add_balloon_handler(QEMUBalloonEvent *func, void *opaque)
538 qemu_balloon_event = func;
539 qemu_balloon_event_opaque = opaque;
542 void qemu_balloon(ram_addr_t target)
544 if (qemu_balloon_event)
545 qemu_balloon_event(qemu_balloon_event_opaque, target);
548 ram_addr_t qemu_balloon_status(void)
550 if (qemu_balloon_event)
551 return qemu_balloon_event(qemu_balloon_event_opaque, 0);
552 return 0;
555 /***********************************************************/
556 /* keyboard/mouse */
558 static QEMUPutKBDEvent *qemu_put_kbd_event;
559 static void *qemu_put_kbd_event_opaque;
560 static QEMUPutMouseEntry *qemu_put_mouse_event_head;
561 static QEMUPutMouseEntry *qemu_put_mouse_event_current;
563 void qemu_add_kbd_event_handler(QEMUPutKBDEvent *func, void *opaque)
565 qemu_put_kbd_event_opaque = opaque;
566 qemu_put_kbd_event = func;
569 QEMUPutMouseEntry *qemu_add_mouse_event_handler(QEMUPutMouseEvent *func,
570 void *opaque, int absolute,
571 const char *name)
573 QEMUPutMouseEntry *s, *cursor;
575 s = qemu_mallocz(sizeof(QEMUPutMouseEntry));
577 s->qemu_put_mouse_event = func;
578 s->qemu_put_mouse_event_opaque = opaque;
579 s->qemu_put_mouse_event_absolute = absolute;
580 s->qemu_put_mouse_event_name = qemu_strdup(name);
581 s->next = NULL;
583 if (!qemu_put_mouse_event_head) {
584 qemu_put_mouse_event_head = qemu_put_mouse_event_current = s;
585 return s;
588 cursor = qemu_put_mouse_event_head;
589 while (cursor->next != NULL)
590 cursor = cursor->next;
592 cursor->next = s;
593 qemu_put_mouse_event_current = s;
595 return s;
598 void qemu_remove_mouse_event_handler(QEMUPutMouseEntry *entry)
600 QEMUPutMouseEntry *prev = NULL, *cursor;
602 if (!qemu_put_mouse_event_head || entry == NULL)
603 return;
605 cursor = qemu_put_mouse_event_head;
606 while (cursor != NULL && cursor != entry) {
607 prev = cursor;
608 cursor = cursor->next;
611 if (cursor == NULL) // does not exist or list empty
612 return;
613 else if (prev == NULL) { // entry is head
614 qemu_put_mouse_event_head = cursor->next;
615 if (qemu_put_mouse_event_current == entry)
616 qemu_put_mouse_event_current = cursor->next;
617 qemu_free(entry->qemu_put_mouse_event_name);
618 qemu_free(entry);
619 return;
622 prev->next = entry->next;
624 if (qemu_put_mouse_event_current == entry)
625 qemu_put_mouse_event_current = prev;
627 qemu_free(entry->qemu_put_mouse_event_name);
628 qemu_free(entry);
631 void kbd_put_keycode(int keycode)
633 if (qemu_put_kbd_event) {
634 qemu_put_kbd_event(qemu_put_kbd_event_opaque, keycode);
638 void kbd_mouse_event(int dx, int dy, int dz, int buttons_state)
640 QEMUPutMouseEvent *mouse_event;
641 void *mouse_event_opaque;
642 int width;
644 if (!qemu_put_mouse_event_current) {
645 return;
648 mouse_event =
649 qemu_put_mouse_event_current->qemu_put_mouse_event;
650 mouse_event_opaque =
651 qemu_put_mouse_event_current->qemu_put_mouse_event_opaque;
653 if (mouse_event) {
654 if (graphic_rotate) {
655 if (qemu_put_mouse_event_current->qemu_put_mouse_event_absolute)
656 width = 0x7fff;
657 else
658 width = graphic_width - 1;
659 mouse_event(mouse_event_opaque,
660 width - dy, dx, dz, buttons_state);
661 } else
662 mouse_event(mouse_event_opaque,
663 dx, dy, dz, buttons_state);
667 int kbd_mouse_is_absolute(void)
669 if (!qemu_put_mouse_event_current)
670 return 0;
672 return qemu_put_mouse_event_current->qemu_put_mouse_event_absolute;
675 void do_info_mice(Monitor *mon)
677 QEMUPutMouseEntry *cursor;
678 int index = 0;
680 if (!qemu_put_mouse_event_head) {
681 monitor_printf(mon, "No mouse devices connected\n");
682 return;
685 monitor_printf(mon, "Mouse devices available:\n");
686 cursor = qemu_put_mouse_event_head;
687 while (cursor != NULL) {
688 monitor_printf(mon, "%c Mouse #%d: %s\n",
689 (cursor == qemu_put_mouse_event_current ? '*' : ' '),
690 index, cursor->qemu_put_mouse_event_name);
691 index++;
692 cursor = cursor->next;
696 void do_mouse_set(Monitor *mon, int index)
698 QEMUPutMouseEntry *cursor;
699 int i = 0;
701 if (!qemu_put_mouse_event_head) {
702 monitor_printf(mon, "No mouse devices connected\n");
703 return;
706 cursor = qemu_put_mouse_event_head;
707 while (cursor != NULL && index != i) {
708 i++;
709 cursor = cursor->next;
712 if (cursor != NULL)
713 qemu_put_mouse_event_current = cursor;
714 else
715 monitor_printf(mon, "Mouse at given index not found\n");
718 /* compute with 96 bit intermediate result: (a*b)/c */
719 uint64_t muldiv64(uint64_t a, uint32_t b, uint32_t c)
721 union {
722 uint64_t ll;
723 struct {
724 #ifdef WORDS_BIGENDIAN
725 uint32_t high, low;
726 #else
727 uint32_t low, high;
728 #endif
729 } l;
730 } u, res;
731 uint64_t rl, rh;
733 u.ll = a;
734 rl = (uint64_t)u.l.low * (uint64_t)b;
735 rh = (uint64_t)u.l.high * (uint64_t)b;
736 rh += (rl >> 32);
737 res.l.high = rh / c;
738 res.l.low = (((rh % c) << 32) + (rl & 0xffffffff)) / c;
739 return res.ll;
742 /***********************************************************/
743 /* real time host monotonic timer */
745 #define QEMU_TIMER_BASE 1000000000LL
747 #ifdef WIN32
749 static int64_t clock_freq;
751 static void init_get_clock(void)
753 LARGE_INTEGER freq;
754 int ret;
755 ret = QueryPerformanceFrequency(&freq);
756 if (ret == 0) {
757 fprintf(stderr, "Could not calibrate ticks\n");
758 exit(1);
760 clock_freq = freq.QuadPart;
763 static int64_t get_clock(void)
765 LARGE_INTEGER ti;
766 QueryPerformanceCounter(&ti);
767 return muldiv64(ti.QuadPart, QEMU_TIMER_BASE, clock_freq);
770 #else
772 static int use_rt_clock;
774 static void init_get_clock(void)
776 use_rt_clock = 0;
777 #if defined(__linux__) || (defined(__FreeBSD__) && __FreeBSD_version >= 500000) \
778 || defined(__DragonFly__)
780 struct timespec ts;
781 if (clock_gettime(CLOCK_MONOTONIC, &ts) == 0) {
782 use_rt_clock = 1;
785 #endif
788 static int64_t get_clock(void)
790 #if defined(__linux__) || (defined(__FreeBSD__) && __FreeBSD_version >= 500000) \
791 || defined(__DragonFly__)
792 if (use_rt_clock) {
793 struct timespec ts;
794 clock_gettime(CLOCK_MONOTONIC, &ts);
795 return ts.tv_sec * 1000000000LL + ts.tv_nsec;
796 } else
797 #endif
799 /* XXX: using gettimeofday leads to problems if the date
800 changes, so it should be avoided. */
801 struct timeval tv;
802 gettimeofday(&tv, NULL);
803 return tv.tv_sec * 1000000000LL + (tv.tv_usec * 1000);
806 #endif
808 /* Return the virtual CPU time, based on the instruction counter. */
809 static int64_t cpu_get_icount(void)
811 int64_t icount;
812 CPUState *env = cpu_single_env;;
813 icount = qemu_icount;
814 if (env) {
815 if (!can_do_io(env))
816 fprintf(stderr, "Bad clock read\n");
817 icount -= (env->icount_decr.u16.low + env->icount_extra);
819 return qemu_icount_bias + (icount << icount_time_shift);
822 /***********************************************************/
823 /* guest cycle counter */
825 static int64_t cpu_ticks_prev;
826 static int64_t cpu_ticks_offset;
827 static int64_t cpu_clock_offset;
828 static int cpu_ticks_enabled;
830 /* return the host CPU cycle counter and handle stop/restart */
831 int64_t cpu_get_ticks(void)
833 if (use_icount) {
834 return cpu_get_icount();
836 if (!cpu_ticks_enabled) {
837 return cpu_ticks_offset;
838 } else {
839 int64_t ticks;
840 ticks = cpu_get_real_ticks();
841 if (cpu_ticks_prev > ticks) {
842 /* Note: non increasing ticks may happen if the host uses
843 software suspend */
844 cpu_ticks_offset += cpu_ticks_prev - ticks;
846 cpu_ticks_prev = ticks;
847 return ticks + cpu_ticks_offset;
851 /* return the host CPU monotonic timer and handle stop/restart */
852 static int64_t cpu_get_clock(void)
854 int64_t ti;
855 if (!cpu_ticks_enabled) {
856 return cpu_clock_offset;
857 } else {
858 ti = get_clock();
859 return ti + cpu_clock_offset;
863 /* enable cpu_get_ticks() */
864 void cpu_enable_ticks(void)
866 if (!cpu_ticks_enabled) {
867 cpu_ticks_offset -= cpu_get_real_ticks();
868 cpu_clock_offset -= get_clock();
869 cpu_ticks_enabled = 1;
873 /* disable cpu_get_ticks() : the clock is stopped. You must not call
874 cpu_get_ticks() after that. */
875 void cpu_disable_ticks(void)
877 if (cpu_ticks_enabled) {
878 cpu_ticks_offset = cpu_get_ticks();
879 cpu_clock_offset = cpu_get_clock();
880 cpu_ticks_enabled = 0;
884 /***********************************************************/
885 /* timers */
887 #define QEMU_TIMER_REALTIME 0
888 #define QEMU_TIMER_VIRTUAL 1
890 struct QEMUClock {
891 int type;
892 /* XXX: add frequency */
895 struct QEMUTimer {
896 QEMUClock *clock;
897 int64_t expire_time;
898 QEMUTimerCB *cb;
899 void *opaque;
900 struct QEMUTimer *next;
903 struct qemu_alarm_timer {
904 char const *name;
905 unsigned int flags;
907 int (*start)(struct qemu_alarm_timer *t);
908 void (*stop)(struct qemu_alarm_timer *t);
909 void (*rearm)(struct qemu_alarm_timer *t);
910 void *priv;
913 #define ALARM_FLAG_DYNTICKS 0x1
914 #define ALARM_FLAG_EXPIRED 0x2
916 static inline int alarm_has_dynticks(struct qemu_alarm_timer *t)
918 return t && (t->flags & ALARM_FLAG_DYNTICKS);
921 static void qemu_rearm_alarm_timer(struct qemu_alarm_timer *t)
923 if (!alarm_has_dynticks(t))
924 return;
926 t->rearm(t);
929 /* TODO: MIN_TIMER_REARM_US should be optimized */
930 #define MIN_TIMER_REARM_US 250
932 static struct qemu_alarm_timer *alarm_timer;
934 #ifdef _WIN32
936 struct qemu_alarm_win32 {
937 MMRESULT timerId;
938 unsigned int period;
939 } alarm_win32_data = {0, -1};
941 static int win32_start_timer(struct qemu_alarm_timer *t);
942 static void win32_stop_timer(struct qemu_alarm_timer *t);
943 static void win32_rearm_timer(struct qemu_alarm_timer *t);
945 #else
947 static int unix_start_timer(struct qemu_alarm_timer *t);
948 static void unix_stop_timer(struct qemu_alarm_timer *t);
950 #ifdef __linux__
952 static int dynticks_start_timer(struct qemu_alarm_timer *t);
953 static void dynticks_stop_timer(struct qemu_alarm_timer *t);
954 static void dynticks_rearm_timer(struct qemu_alarm_timer *t);
956 static int hpet_start_timer(struct qemu_alarm_timer *t);
957 static void hpet_stop_timer(struct qemu_alarm_timer *t);
959 static int rtc_start_timer(struct qemu_alarm_timer *t);
960 static void rtc_stop_timer(struct qemu_alarm_timer *t);
962 #endif /* __linux__ */
964 #endif /* _WIN32 */
966 /* Correlation between real and virtual time is always going to be
967 fairly approximate, so ignore small variation.
968 When the guest is idle real and virtual time will be aligned in
969 the IO wait loop. */
970 #define ICOUNT_WOBBLE (QEMU_TIMER_BASE / 10)
972 static void icount_adjust(void)
974 int64_t cur_time;
975 int64_t cur_icount;
976 int64_t delta;
977 static int64_t last_delta;
978 /* If the VM is not running, then do nothing. */
979 if (!vm_running)
980 return;
982 cur_time = cpu_get_clock();
983 cur_icount = qemu_get_clock(vm_clock);
984 delta = cur_icount - cur_time;
985 /* FIXME: This is a very crude algorithm, somewhat prone to oscillation. */
986 if (delta > 0
987 && last_delta + ICOUNT_WOBBLE < delta * 2
988 && icount_time_shift > 0) {
989 /* The guest is getting too far ahead. Slow time down. */
990 icount_time_shift--;
992 if (delta < 0
993 && last_delta - ICOUNT_WOBBLE > delta * 2
994 && icount_time_shift < MAX_ICOUNT_SHIFT) {
995 /* The guest is getting too far behind. Speed time up. */
996 icount_time_shift++;
998 last_delta = delta;
999 qemu_icount_bias = cur_icount - (qemu_icount << icount_time_shift);
1002 static void icount_adjust_rt(void * opaque)
1004 qemu_mod_timer(icount_rt_timer,
1005 qemu_get_clock(rt_clock) + 1000);
1006 icount_adjust();
1009 static void icount_adjust_vm(void * opaque)
1011 qemu_mod_timer(icount_vm_timer,
1012 qemu_get_clock(vm_clock) + QEMU_TIMER_BASE / 10);
1013 icount_adjust();
1016 static void init_icount_adjust(void)
1018 /* Have both realtime and virtual time triggers for speed adjustment.
1019 The realtime trigger catches emulated time passing too slowly,
1020 the virtual time trigger catches emulated time passing too fast.
1021 Realtime triggers occur even when idle, so use them less frequently
1022 than VM triggers. */
1023 icount_rt_timer = qemu_new_timer(rt_clock, icount_adjust_rt, NULL);
1024 qemu_mod_timer(icount_rt_timer,
1025 qemu_get_clock(rt_clock) + 1000);
1026 icount_vm_timer = qemu_new_timer(vm_clock, icount_adjust_vm, NULL);
1027 qemu_mod_timer(icount_vm_timer,
1028 qemu_get_clock(vm_clock) + QEMU_TIMER_BASE / 10);
1031 static struct qemu_alarm_timer alarm_timers[] = {
1032 #ifndef _WIN32
1033 #ifdef __linux__
1034 {"dynticks", ALARM_FLAG_DYNTICKS, dynticks_start_timer,
1035 dynticks_stop_timer, dynticks_rearm_timer, NULL},
1036 /* HPET - if available - is preferred */
1037 {"hpet", 0, hpet_start_timer, hpet_stop_timer, NULL, NULL},
1038 /* ...otherwise try RTC */
1039 {"rtc", 0, rtc_start_timer, rtc_stop_timer, NULL, NULL},
1040 #endif
1041 {"unix", 0, unix_start_timer, unix_stop_timer, NULL, NULL},
1042 #else
1043 {"dynticks", ALARM_FLAG_DYNTICKS, win32_start_timer,
1044 win32_stop_timer, win32_rearm_timer, &alarm_win32_data},
1045 {"win32", 0, win32_start_timer,
1046 win32_stop_timer, NULL, &alarm_win32_data},
1047 #endif
1048 {NULL, }
1051 static void show_available_alarms(void)
1053 int i;
1055 printf("Available alarm timers, in order of precedence:\n");
1056 for (i = 0; alarm_timers[i].name; i++)
1057 printf("%s\n", alarm_timers[i].name);
1060 static void configure_alarms(char const *opt)
1062 int i;
1063 int cur = 0;
1064 int count = ARRAY_SIZE(alarm_timers) - 1;
1065 char *arg;
1066 char *name;
1067 struct qemu_alarm_timer tmp;
1069 if (!strcmp(opt, "?")) {
1070 show_available_alarms();
1071 exit(0);
1074 arg = strdup(opt);
1076 /* Reorder the array */
1077 name = strtok(arg, ",");
1078 while (name) {
1079 for (i = 0; i < count && alarm_timers[i].name; i++) {
1080 if (!strcmp(alarm_timers[i].name, name))
1081 break;
1084 if (i == count) {
1085 fprintf(stderr, "Unknown clock %s\n", name);
1086 goto next;
1089 if (i < cur)
1090 /* Ignore */
1091 goto next;
1093 /* Swap */
1094 tmp = alarm_timers[i];
1095 alarm_timers[i] = alarm_timers[cur];
1096 alarm_timers[cur] = tmp;
1098 cur++;
1099 next:
1100 name = strtok(NULL, ",");
1103 free(arg);
1105 if (cur) {
1106 /* Disable remaining timers */
1107 for (i = cur; i < count; i++)
1108 alarm_timers[i].name = NULL;
1109 } else {
1110 show_available_alarms();
1111 exit(1);
1115 QEMUClock *rt_clock;
1116 QEMUClock *vm_clock;
1118 static QEMUTimer *active_timers[2];
1120 static QEMUClock *qemu_new_clock(int type)
1122 QEMUClock *clock;
1123 clock = qemu_mallocz(sizeof(QEMUClock));
1124 clock->type = type;
1125 return clock;
1128 QEMUTimer *qemu_new_timer(QEMUClock *clock, QEMUTimerCB *cb, void *opaque)
1130 QEMUTimer *ts;
1132 ts = qemu_mallocz(sizeof(QEMUTimer));
1133 ts->clock = clock;
1134 ts->cb = cb;
1135 ts->opaque = opaque;
1136 return ts;
1139 void qemu_free_timer(QEMUTimer *ts)
1141 qemu_free(ts);
1144 /* stop a timer, but do not dealloc it */
1145 void qemu_del_timer(QEMUTimer *ts)
1147 QEMUTimer **pt, *t;
1149 /* NOTE: this code must be signal safe because
1150 qemu_timer_expired() can be called from a signal. */
1151 pt = &active_timers[ts->clock->type];
1152 for(;;) {
1153 t = *pt;
1154 if (!t)
1155 break;
1156 if (t == ts) {
1157 *pt = t->next;
1158 break;
1160 pt = &t->next;
1164 /* modify the current timer so that it will be fired when current_time
1165 >= expire_time. The corresponding callback will be called. */
1166 void qemu_mod_timer(QEMUTimer *ts, int64_t expire_time)
1168 QEMUTimer **pt, *t;
1170 qemu_del_timer(ts);
1172 /* add the timer in the sorted list */
1173 /* NOTE: this code must be signal safe because
1174 qemu_timer_expired() can be called from a signal. */
1175 pt = &active_timers[ts->clock->type];
1176 for(;;) {
1177 t = *pt;
1178 if (!t)
1179 break;
1180 if (t->expire_time > expire_time)
1181 break;
1182 pt = &t->next;
1184 ts->expire_time = expire_time;
1185 ts->next = *pt;
1186 *pt = ts;
1188 /* Rearm if necessary */
1189 if (pt == &active_timers[ts->clock->type]) {
1190 if ((alarm_timer->flags & ALARM_FLAG_EXPIRED) == 0) {
1191 qemu_rearm_alarm_timer(alarm_timer);
1193 /* Interrupt execution to force deadline recalculation. */
1194 if (use_icount)
1195 qemu_notify_event();
1199 int qemu_timer_pending(QEMUTimer *ts)
1201 QEMUTimer *t;
1202 for(t = active_timers[ts->clock->type]; t != NULL; t = t->next) {
1203 if (t == ts)
1204 return 1;
1206 return 0;
1209 static inline int qemu_timer_expired(QEMUTimer *timer_head, int64_t current_time)
1211 if (!timer_head)
1212 return 0;
1213 return (timer_head->expire_time <= current_time);
1216 static void qemu_run_timers(QEMUTimer **ptimer_head, int64_t current_time)
1218 QEMUTimer *ts;
1220 for(;;) {
1221 ts = *ptimer_head;
1222 if (!ts || ts->expire_time > current_time)
1223 break;
1224 /* remove timer from the list before calling the callback */
1225 *ptimer_head = ts->next;
1226 ts->next = NULL;
1228 /* run the callback (the timer list can be modified) */
1229 ts->cb(ts->opaque);
1233 int64_t qemu_get_clock(QEMUClock *clock)
1235 switch(clock->type) {
1236 case QEMU_TIMER_REALTIME:
1237 return get_clock() / 1000000;
1238 default:
1239 case QEMU_TIMER_VIRTUAL:
1240 if (use_icount) {
1241 return cpu_get_icount();
1242 } else {
1243 return cpu_get_clock();
1248 static void init_timers(void)
1250 init_get_clock();
1251 ticks_per_sec = QEMU_TIMER_BASE;
1252 rt_clock = qemu_new_clock(QEMU_TIMER_REALTIME);
1253 vm_clock = qemu_new_clock(QEMU_TIMER_VIRTUAL);
1256 /* save a timer */
1257 void qemu_put_timer(QEMUFile *f, QEMUTimer *ts)
1259 uint64_t expire_time;
1261 if (qemu_timer_pending(ts)) {
1262 expire_time = ts->expire_time;
1263 } else {
1264 expire_time = -1;
1266 qemu_put_be64(f, expire_time);
1269 void qemu_get_timer(QEMUFile *f, QEMUTimer *ts)
1271 uint64_t expire_time;
1273 expire_time = qemu_get_be64(f);
1274 if (expire_time != -1) {
1275 qemu_mod_timer(ts, expire_time);
1276 } else {
1277 qemu_del_timer(ts);
1281 static void timer_save(QEMUFile *f, void *opaque)
1283 if (cpu_ticks_enabled) {
1284 hw_error("cannot save state if virtual timers are running");
1286 qemu_put_be64(f, cpu_ticks_offset);
1287 qemu_put_be64(f, ticks_per_sec);
1288 qemu_put_be64(f, cpu_clock_offset);
1291 static int timer_load(QEMUFile *f, void *opaque, int version_id)
1293 if (version_id != 1 && version_id != 2)
1294 return -EINVAL;
1295 if (cpu_ticks_enabled) {
1296 return -EINVAL;
1298 cpu_ticks_offset=qemu_get_be64(f);
1299 ticks_per_sec=qemu_get_be64(f);
1300 if (version_id == 2) {
1301 cpu_clock_offset=qemu_get_be64(f);
1303 return 0;
1306 static void qemu_event_increment(void);
1308 #ifdef _WIN32
1309 static void CALLBACK host_alarm_handler(UINT uTimerID, UINT uMsg,
1310 DWORD_PTR dwUser, DWORD_PTR dw1,
1311 DWORD_PTR dw2)
1312 #else
1313 static void host_alarm_handler(int host_signum)
1314 #endif
1316 #if 0
1317 #define DISP_FREQ 1000
1319 static int64_t delta_min = INT64_MAX;
1320 static int64_t delta_max, delta_cum, last_clock, delta, ti;
1321 static int count;
1322 ti = qemu_get_clock(vm_clock);
1323 if (last_clock != 0) {
1324 delta = ti - last_clock;
1325 if (delta < delta_min)
1326 delta_min = delta;
1327 if (delta > delta_max)
1328 delta_max = delta;
1329 delta_cum += delta;
1330 if (++count == DISP_FREQ) {
1331 printf("timer: min=%" PRId64 " us max=%" PRId64 " us avg=%" PRId64 " us avg_freq=%0.3f Hz\n",
1332 muldiv64(delta_min, 1000000, ticks_per_sec),
1333 muldiv64(delta_max, 1000000, ticks_per_sec),
1334 muldiv64(delta_cum, 1000000 / DISP_FREQ, ticks_per_sec),
1335 (double)ticks_per_sec / ((double)delta_cum / DISP_FREQ));
1336 count = 0;
1337 delta_min = INT64_MAX;
1338 delta_max = 0;
1339 delta_cum = 0;
1342 last_clock = ti;
1344 #endif
1345 if (alarm_has_dynticks(alarm_timer) ||
1346 (!use_icount &&
1347 qemu_timer_expired(active_timers[QEMU_TIMER_VIRTUAL],
1348 qemu_get_clock(vm_clock))) ||
1349 qemu_timer_expired(active_timers[QEMU_TIMER_REALTIME],
1350 qemu_get_clock(rt_clock))) {
1351 qemu_event_increment();
1352 if (alarm_timer) alarm_timer->flags |= ALARM_FLAG_EXPIRED;
1354 #ifndef CONFIG_IOTHREAD
1355 if (next_cpu) {
1356 /* stop the currently executing cpu because a timer occured */
1357 cpu_exit(next_cpu);
1358 #ifdef CONFIG_KQEMU
1359 if (next_cpu->kqemu_enabled) {
1360 kqemu_cpu_interrupt(next_cpu);
1362 #endif
1364 #endif
1365 timer_alarm_pending = 1;
1366 qemu_notify_event();
1370 static int64_t qemu_next_deadline(void)
1372 int64_t delta;
1374 if (active_timers[QEMU_TIMER_VIRTUAL]) {
1375 delta = active_timers[QEMU_TIMER_VIRTUAL]->expire_time -
1376 qemu_get_clock(vm_clock);
1377 } else {
1378 /* To avoid problems with overflow limit this to 2^32. */
1379 delta = INT32_MAX;
1382 if (delta < 0)
1383 delta = 0;
1385 return delta;
1388 #if defined(__linux__) || defined(_WIN32)
1389 static uint64_t qemu_next_deadline_dyntick(void)
1391 int64_t delta;
1392 int64_t rtdelta;
1394 if (use_icount)
1395 delta = INT32_MAX;
1396 else
1397 delta = (qemu_next_deadline() + 999) / 1000;
1399 if (active_timers[QEMU_TIMER_REALTIME]) {
1400 rtdelta = (active_timers[QEMU_TIMER_REALTIME]->expire_time -
1401 qemu_get_clock(rt_clock))*1000;
1402 if (rtdelta < delta)
1403 delta = rtdelta;
1406 if (delta < MIN_TIMER_REARM_US)
1407 delta = MIN_TIMER_REARM_US;
1409 return delta;
1411 #endif
1413 #ifndef _WIN32
1415 /* Sets a specific flag */
1416 static int fcntl_setfl(int fd, int flag)
1418 int flags;
1420 flags = fcntl(fd, F_GETFL);
1421 if (flags == -1)
1422 return -errno;
1424 if (fcntl(fd, F_SETFL, flags | flag) == -1)
1425 return -errno;
1427 return 0;
1430 #if defined(__linux__)
1432 #define RTC_FREQ 1024
1434 static void enable_sigio_timer(int fd)
1436 struct sigaction act;
1438 /* timer signal */
1439 sigfillset(&act.sa_mask);
1440 act.sa_flags = 0;
1441 act.sa_handler = host_alarm_handler;
1443 sigaction(SIGIO, &act, NULL);
1444 fcntl_setfl(fd, O_ASYNC);
1445 fcntl(fd, F_SETOWN, getpid());
1448 static int hpet_start_timer(struct qemu_alarm_timer *t)
1450 struct hpet_info info;
1451 int r, fd;
1453 fd = open("/dev/hpet", O_RDONLY);
1454 if (fd < 0)
1455 return -1;
1457 /* Set frequency */
1458 r = ioctl(fd, HPET_IRQFREQ, RTC_FREQ);
1459 if (r < 0) {
1460 fprintf(stderr, "Could not configure '/dev/hpet' to have a 1024Hz timer. This is not a fatal\n"
1461 "error, but for better emulation accuracy type:\n"
1462 "'echo 1024 > /proc/sys/dev/hpet/max-user-freq' as root.\n");
1463 goto fail;
1466 /* Check capabilities */
1467 r = ioctl(fd, HPET_INFO, &info);
1468 if (r < 0)
1469 goto fail;
1471 /* Enable periodic mode */
1472 r = ioctl(fd, HPET_EPI, 0);
1473 if (info.hi_flags && (r < 0))
1474 goto fail;
1476 /* Enable interrupt */
1477 r = ioctl(fd, HPET_IE_ON, 0);
1478 if (r < 0)
1479 goto fail;
1481 enable_sigio_timer(fd);
1482 t->priv = (void *)(long)fd;
1484 return 0;
1485 fail:
1486 close(fd);
1487 return -1;
1490 static void hpet_stop_timer(struct qemu_alarm_timer *t)
1492 int fd = (long)t->priv;
1494 close(fd);
1497 static int rtc_start_timer(struct qemu_alarm_timer *t)
1499 int rtc_fd;
1500 unsigned long current_rtc_freq = 0;
1502 TFR(rtc_fd = open("/dev/rtc", O_RDONLY));
1503 if (rtc_fd < 0)
1504 return -1;
1505 ioctl(rtc_fd, RTC_IRQP_READ, &current_rtc_freq);
1506 if (current_rtc_freq != RTC_FREQ &&
1507 ioctl(rtc_fd, RTC_IRQP_SET, RTC_FREQ) < 0) {
1508 fprintf(stderr, "Could not configure '/dev/rtc' to have a 1024 Hz timer. This is not a fatal\n"
1509 "error, but for better emulation accuracy either use a 2.6 host Linux kernel or\n"
1510 "type 'echo 1024 > /proc/sys/dev/rtc/max-user-freq' as root.\n");
1511 goto fail;
1513 if (ioctl(rtc_fd, RTC_PIE_ON, 0) < 0) {
1514 fail:
1515 close(rtc_fd);
1516 return -1;
1519 enable_sigio_timer(rtc_fd);
1521 t->priv = (void *)(long)rtc_fd;
1523 return 0;
1526 static void rtc_stop_timer(struct qemu_alarm_timer *t)
1528 int rtc_fd = (long)t->priv;
1530 close(rtc_fd);
1533 static int dynticks_start_timer(struct qemu_alarm_timer *t)
1535 struct sigevent ev;
1536 timer_t host_timer;
1537 struct sigaction act;
1539 sigfillset(&act.sa_mask);
1540 act.sa_flags = 0;
1541 act.sa_handler = host_alarm_handler;
1543 sigaction(SIGALRM, &act, NULL);
1546 * Initialize ev struct to 0 to avoid valgrind complaining
1547 * about uninitialized data in timer_create call
1549 memset(&ev, 0, sizeof(ev));
1550 ev.sigev_value.sival_int = 0;
1551 ev.sigev_notify = SIGEV_SIGNAL;
1552 ev.sigev_signo = SIGALRM;
1554 if (timer_create(CLOCK_REALTIME, &ev, &host_timer)) {
1555 perror("timer_create");
1557 /* disable dynticks */
1558 fprintf(stderr, "Dynamic Ticks disabled\n");
1560 return -1;
1563 t->priv = (void *)(long)host_timer;
1565 return 0;
1568 static void dynticks_stop_timer(struct qemu_alarm_timer *t)
1570 timer_t host_timer = (timer_t)(long)t->priv;
1572 timer_delete(host_timer);
1575 static void dynticks_rearm_timer(struct qemu_alarm_timer *t)
1577 timer_t host_timer = (timer_t)(long)t->priv;
1578 struct itimerspec timeout;
1579 int64_t nearest_delta_us = INT64_MAX;
1580 int64_t current_us;
1582 if (!active_timers[QEMU_TIMER_REALTIME] &&
1583 !active_timers[QEMU_TIMER_VIRTUAL])
1584 return;
1586 nearest_delta_us = qemu_next_deadline_dyntick();
1588 /* check whether a timer is already running */
1589 if (timer_gettime(host_timer, &timeout)) {
1590 perror("gettime");
1591 fprintf(stderr, "Internal timer error: aborting\n");
1592 exit(1);
1594 current_us = timeout.it_value.tv_sec * 1000000 + timeout.it_value.tv_nsec/1000;
1595 if (current_us && current_us <= nearest_delta_us)
1596 return;
1598 timeout.it_interval.tv_sec = 0;
1599 timeout.it_interval.tv_nsec = 0; /* 0 for one-shot timer */
1600 timeout.it_value.tv_sec = nearest_delta_us / 1000000;
1601 timeout.it_value.tv_nsec = (nearest_delta_us % 1000000) * 1000;
1602 if (timer_settime(host_timer, 0 /* RELATIVE */, &timeout, NULL)) {
1603 perror("settime");
1604 fprintf(stderr, "Internal timer error: aborting\n");
1605 exit(1);
1609 #endif /* defined(__linux__) */
1611 static int unix_start_timer(struct qemu_alarm_timer *t)
1613 struct sigaction act;
1614 struct itimerval itv;
1615 int err;
1617 /* timer signal */
1618 sigfillset(&act.sa_mask);
1619 act.sa_flags = 0;
1620 act.sa_handler = host_alarm_handler;
1622 sigaction(SIGALRM, &act, NULL);
1624 itv.it_interval.tv_sec = 0;
1625 /* for i386 kernel 2.6 to get 1 ms */
1626 itv.it_interval.tv_usec = 999;
1627 itv.it_value.tv_sec = 0;
1628 itv.it_value.tv_usec = 10 * 1000;
1630 err = setitimer(ITIMER_REAL, &itv, NULL);
1631 if (err)
1632 return -1;
1634 return 0;
1637 static void unix_stop_timer(struct qemu_alarm_timer *t)
1639 struct itimerval itv;
1641 memset(&itv, 0, sizeof(itv));
1642 setitimer(ITIMER_REAL, &itv, NULL);
1645 #endif /* !defined(_WIN32) */
1648 #ifdef _WIN32
1650 static int win32_start_timer(struct qemu_alarm_timer *t)
1652 TIMECAPS tc;
1653 struct qemu_alarm_win32 *data = t->priv;
1654 UINT flags;
1656 memset(&tc, 0, sizeof(tc));
1657 timeGetDevCaps(&tc, sizeof(tc));
1659 if (data->period < tc.wPeriodMin)
1660 data->period = tc.wPeriodMin;
1662 timeBeginPeriod(data->period);
1664 flags = TIME_CALLBACK_FUNCTION;
1665 if (alarm_has_dynticks(t))
1666 flags |= TIME_ONESHOT;
1667 else
1668 flags |= TIME_PERIODIC;
1670 data->timerId = timeSetEvent(1, // interval (ms)
1671 data->period, // resolution
1672 host_alarm_handler, // function
1673 (DWORD)t, // parameter
1674 flags);
1676 if (!data->timerId) {
1677 perror("Failed to initialize win32 alarm timer");
1678 timeEndPeriod(data->period);
1679 return -1;
1682 return 0;
1685 static void win32_stop_timer(struct qemu_alarm_timer *t)
1687 struct qemu_alarm_win32 *data = t->priv;
1689 timeKillEvent(data->timerId);
1690 timeEndPeriod(data->period);
1693 static void win32_rearm_timer(struct qemu_alarm_timer *t)
1695 struct qemu_alarm_win32 *data = t->priv;
1696 uint64_t nearest_delta_us;
1698 if (!active_timers[QEMU_TIMER_REALTIME] &&
1699 !active_timers[QEMU_TIMER_VIRTUAL])
1700 return;
1702 nearest_delta_us = qemu_next_deadline_dyntick();
1703 nearest_delta_us /= 1000;
1705 timeKillEvent(data->timerId);
1707 data->timerId = timeSetEvent(1,
1708 data->period,
1709 host_alarm_handler,
1710 (DWORD)t,
1711 TIME_ONESHOT | TIME_PERIODIC);
1713 if (!data->timerId) {
1714 perror("Failed to re-arm win32 alarm timer");
1716 timeEndPeriod(data->period);
1717 exit(1);
1721 #endif /* _WIN32 */
1723 static int init_timer_alarm(void)
1725 struct qemu_alarm_timer *t = NULL;
1726 int i, err = -1;
1728 for (i = 0; alarm_timers[i].name; i++) {
1729 t = &alarm_timers[i];
1731 err = t->start(t);
1732 if (!err)
1733 break;
1736 if (err) {
1737 err = -ENOENT;
1738 goto fail;
1741 alarm_timer = t;
1743 return 0;
1745 fail:
1746 return err;
1749 static void quit_timers(void)
1751 alarm_timer->stop(alarm_timer);
1752 alarm_timer = NULL;
1755 /***********************************************************/
1756 /* host time/date access */
1757 void qemu_get_timedate(struct tm *tm, int offset)
1759 time_t ti;
1760 struct tm *ret;
1762 time(&ti);
1763 ti += offset;
1764 if (rtc_date_offset == -1) {
1765 if (rtc_utc)
1766 ret = gmtime(&ti);
1767 else
1768 ret = localtime(&ti);
1769 } else {
1770 ti -= rtc_date_offset;
1771 ret = gmtime(&ti);
1774 memcpy(tm, ret, sizeof(struct tm));
1777 int qemu_timedate_diff(struct tm *tm)
1779 time_t seconds;
1781 if (rtc_date_offset == -1)
1782 if (rtc_utc)
1783 seconds = mktimegm(tm);
1784 else
1785 seconds = mktime(tm);
1786 else
1787 seconds = mktimegm(tm) + rtc_date_offset;
1789 return seconds - time(NULL);
1792 #ifdef _WIN32
1793 static void socket_cleanup(void)
1795 WSACleanup();
1798 static int socket_init(void)
1800 WSADATA Data;
1801 int ret, err;
1803 ret = WSAStartup(MAKEWORD(2,2), &Data);
1804 if (ret != 0) {
1805 err = WSAGetLastError();
1806 fprintf(stderr, "WSAStartup: %d\n", err);
1807 return -1;
1809 atexit(socket_cleanup);
1810 return 0;
1812 #endif
1814 int get_next_param_value(char *buf, int buf_size,
1815 const char *tag, const char **pstr)
1817 const char *p;
1818 char option[128];
1820 p = *pstr;
1821 for(;;) {
1822 p = get_opt_name(option, sizeof(option), p, '=');
1823 if (*p != '=')
1824 break;
1825 p++;
1826 if (!strcmp(tag, option)) {
1827 *pstr = get_opt_value(buf, buf_size, p);
1828 if (**pstr == ',') {
1829 (*pstr)++;
1831 return strlen(buf);
1832 } else {
1833 p = get_opt_value(NULL, 0, p);
1835 if (*p != ',')
1836 break;
1837 p++;
1839 return 0;
1842 int get_param_value(char *buf, int buf_size,
1843 const char *tag, const char *str)
1845 return get_next_param_value(buf, buf_size, tag, &str);
1848 int check_params(char *buf, int buf_size,
1849 const char * const *params, const char *str)
1851 const char *p;
1852 int i;
1854 p = str;
1855 while (*p != '\0') {
1856 p = get_opt_name(buf, buf_size, p, '=');
1857 if (*p != '=') {
1858 return -1;
1860 p++;
1861 for (i = 0; params[i] != NULL; i++) {
1862 if (!strcmp(params[i], buf)) {
1863 break;
1866 if (params[i] == NULL) {
1867 return -1;
1869 p = get_opt_value(NULL, 0, p);
1870 if (*p != ',') {
1871 break;
1873 p++;
1875 return 0;
1878 /***********************************************************/
1879 /* Bluetooth support */
1880 static int nb_hcis;
1881 static int cur_hci;
1882 static struct HCIInfo *hci_table[MAX_NICS];
1884 static struct bt_vlan_s {
1885 struct bt_scatternet_s net;
1886 int id;
1887 struct bt_vlan_s *next;
1888 } *first_bt_vlan;
1890 /* find or alloc a new bluetooth "VLAN" */
1891 static struct bt_scatternet_s *qemu_find_bt_vlan(int id)
1893 struct bt_vlan_s **pvlan, *vlan;
1894 for (vlan = first_bt_vlan; vlan != NULL; vlan = vlan->next) {
1895 if (vlan->id == id)
1896 return &vlan->net;
1898 vlan = qemu_mallocz(sizeof(struct bt_vlan_s));
1899 vlan->id = id;
1900 pvlan = &first_bt_vlan;
1901 while (*pvlan != NULL)
1902 pvlan = &(*pvlan)->next;
1903 *pvlan = vlan;
1904 return &vlan->net;
1907 static void null_hci_send(struct HCIInfo *hci, const uint8_t *data, int len)
1911 static int null_hci_addr_set(struct HCIInfo *hci, const uint8_t *bd_addr)
1913 return -ENOTSUP;
1916 static struct HCIInfo null_hci = {
1917 .cmd_send = null_hci_send,
1918 .sco_send = null_hci_send,
1919 .acl_send = null_hci_send,
1920 .bdaddr_set = null_hci_addr_set,
1923 struct HCIInfo *qemu_next_hci(void)
1925 if (cur_hci == nb_hcis)
1926 return &null_hci;
1928 return hci_table[cur_hci++];
1931 static struct HCIInfo *hci_init(const char *str)
1933 char *endp;
1934 struct bt_scatternet_s *vlan = 0;
1936 if (!strcmp(str, "null"))
1937 /* null */
1938 return &null_hci;
1939 else if (!strncmp(str, "host", 4) && (str[4] == '\0' || str[4] == ':'))
1940 /* host[:hciN] */
1941 return bt_host_hci(str[4] ? str + 5 : "hci0");
1942 else if (!strncmp(str, "hci", 3)) {
1943 /* hci[,vlan=n] */
1944 if (str[3]) {
1945 if (!strncmp(str + 3, ",vlan=", 6)) {
1946 vlan = qemu_find_bt_vlan(strtol(str + 9, &endp, 0));
1947 if (*endp)
1948 vlan = 0;
1950 } else
1951 vlan = qemu_find_bt_vlan(0);
1952 if (vlan)
1953 return bt_new_hci(vlan);
1956 fprintf(stderr, "qemu: Unknown bluetooth HCI `%s'.\n", str);
1958 return 0;
1961 static int bt_hci_parse(const char *str)
1963 struct HCIInfo *hci;
1964 bdaddr_t bdaddr;
1966 if (nb_hcis >= MAX_NICS) {
1967 fprintf(stderr, "qemu: Too many bluetooth HCIs (max %i).\n", MAX_NICS);
1968 return -1;
1971 hci = hci_init(str);
1972 if (!hci)
1973 return -1;
1975 bdaddr.b[0] = 0x52;
1976 bdaddr.b[1] = 0x54;
1977 bdaddr.b[2] = 0x00;
1978 bdaddr.b[3] = 0x12;
1979 bdaddr.b[4] = 0x34;
1980 bdaddr.b[5] = 0x56 + nb_hcis;
1981 hci->bdaddr_set(hci, bdaddr.b);
1983 hci_table[nb_hcis++] = hci;
1985 return 0;
1988 static void bt_vhci_add(int vlan_id)
1990 struct bt_scatternet_s *vlan = qemu_find_bt_vlan(vlan_id);
1992 if (!vlan->slave)
1993 fprintf(stderr, "qemu: warning: adding a VHCI to "
1994 "an empty scatternet %i\n", vlan_id);
1996 bt_vhci_init(bt_new_hci(vlan));
1999 static struct bt_device_s *bt_device_add(const char *opt)
2001 struct bt_scatternet_s *vlan;
2002 int vlan_id = 0;
2003 char *endp = strstr(opt, ",vlan=");
2004 int len = (endp ? endp - opt : strlen(opt)) + 1;
2005 char devname[10];
2007 pstrcpy(devname, MIN(sizeof(devname), len), opt);
2009 if (endp) {
2010 vlan_id = strtol(endp + 6, &endp, 0);
2011 if (*endp) {
2012 fprintf(stderr, "qemu: unrecognised bluetooth vlan Id\n");
2013 return 0;
2017 vlan = qemu_find_bt_vlan(vlan_id);
2019 if (!vlan->slave)
2020 fprintf(stderr, "qemu: warning: adding a slave device to "
2021 "an empty scatternet %i\n", vlan_id);
2023 if (!strcmp(devname, "keyboard"))
2024 return bt_keyboard_init(vlan);
2026 fprintf(stderr, "qemu: unsupported bluetooth device `%s'\n", devname);
2027 return 0;
2030 static int bt_parse(const char *opt)
2032 const char *endp, *p;
2033 int vlan;
2035 if (strstart(opt, "hci", &endp)) {
2036 if (!*endp || *endp == ',') {
2037 if (*endp)
2038 if (!strstart(endp, ",vlan=", 0))
2039 opt = endp + 1;
2041 return bt_hci_parse(opt);
2043 } else if (strstart(opt, "vhci", &endp)) {
2044 if (!*endp || *endp == ',') {
2045 if (*endp) {
2046 if (strstart(endp, ",vlan=", &p)) {
2047 vlan = strtol(p, (char **) &endp, 0);
2048 if (*endp) {
2049 fprintf(stderr, "qemu: bad scatternet '%s'\n", p);
2050 return 1;
2052 } else {
2053 fprintf(stderr, "qemu: bad parameter '%s'\n", endp + 1);
2054 return 1;
2056 } else
2057 vlan = 0;
2059 bt_vhci_add(vlan);
2060 return 0;
2062 } else if (strstart(opt, "device:", &endp))
2063 return !bt_device_add(endp);
2065 fprintf(stderr, "qemu: bad bluetooth parameter '%s'\n", opt);
2066 return 1;
2069 /***********************************************************/
2070 /* QEMU Block devices */
2072 #define HD_ALIAS "index=%d,media=disk"
2073 #define CDROM_ALIAS "index=2,media=cdrom"
2074 #define FD_ALIAS "index=%d,if=floppy"
2075 #define PFLASH_ALIAS "if=pflash"
2076 #define MTD_ALIAS "if=mtd"
2077 #define SD_ALIAS "index=0,if=sd"
2079 static int drive_opt_get_free_idx(void)
2081 int index;
2083 for (index = 0; index < MAX_DRIVES; index++)
2084 if (!drives_opt[index].used) {
2085 drives_opt[index].used = 1;
2086 return index;
2089 return -1;
2092 static int drive_get_free_idx(void)
2094 int index;
2096 for (index = 0; index < MAX_DRIVES; index++)
2097 if (!drives_table[index].used) {
2098 drives_table[index].used = 1;
2099 return index;
2102 return -1;
2105 int drive_add(const char *file, const char *fmt, ...)
2107 va_list ap;
2108 int index = drive_opt_get_free_idx();
2110 if (nb_drives_opt >= MAX_DRIVES || index == -1) {
2111 fprintf(stderr, "qemu: too many drives\n");
2112 return -1;
2115 drives_opt[index].file = file;
2116 va_start(ap, fmt);
2117 vsnprintf(drives_opt[index].opt,
2118 sizeof(drives_opt[0].opt), fmt, ap);
2119 va_end(ap);
2121 nb_drives_opt++;
2122 return index;
2125 void drive_remove(int index)
2127 drives_opt[index].used = 0;
2128 nb_drives_opt--;
2131 int drive_get_index(BlockInterfaceType type, int bus, int unit)
2133 int index;
2135 /* seek interface, bus and unit */
2137 for (index = 0; index < MAX_DRIVES; index++)
2138 if (drives_table[index].type == type &&
2139 drives_table[index].bus == bus &&
2140 drives_table[index].unit == unit &&
2141 drives_table[index].used)
2142 return index;
2144 return -1;
2147 int drive_get_max_bus(BlockInterfaceType type)
2149 int max_bus;
2150 int index;
2152 max_bus = -1;
2153 for (index = 0; index < nb_drives; index++) {
2154 if(drives_table[index].type == type &&
2155 drives_table[index].bus > max_bus)
2156 max_bus = drives_table[index].bus;
2158 return max_bus;
2161 const char *drive_get_serial(BlockDriverState *bdrv)
2163 int index;
2165 for (index = 0; index < nb_drives; index++)
2166 if (drives_table[index].bdrv == bdrv)
2167 return drives_table[index].serial;
2169 return "\0";
2172 BlockInterfaceErrorAction drive_get_onerror(BlockDriverState *bdrv)
2174 int index;
2176 for (index = 0; index < nb_drives; index++)
2177 if (drives_table[index].bdrv == bdrv)
2178 return drives_table[index].onerror;
2180 return BLOCK_ERR_STOP_ENOSPC;
2183 static void bdrv_format_print(void *opaque, const char *name)
2185 fprintf(stderr, " %s", name);
2188 void drive_uninit(BlockDriverState *bdrv)
2190 int i;
2192 for (i = 0; i < MAX_DRIVES; i++)
2193 if (drives_table[i].bdrv == bdrv) {
2194 drives_table[i].bdrv = NULL;
2195 drives_table[i].used = 0;
2196 drive_remove(drives_table[i].drive_opt_idx);
2197 nb_drives--;
2198 break;
2202 int drive_init(struct drive_opt *arg, int snapshot, void *opaque)
2204 char buf[128];
2205 char file[1024];
2206 char devname[128];
2207 char serial[21];
2208 const char *mediastr = "";
2209 BlockInterfaceType type;
2210 enum { MEDIA_DISK, MEDIA_CDROM } media;
2211 int bus_id, unit_id;
2212 int cyls, heads, secs, translation;
2213 BlockDriverState *bdrv;
2214 BlockDriver *drv = NULL;
2215 QEMUMachine *machine = opaque;
2216 int max_devs;
2217 int index;
2218 int cache;
2219 int bdrv_flags, onerror;
2220 const char *devaddr;
2221 int drives_table_idx;
2222 char *str = arg->opt;
2223 static const char * const params[] = { "bus", "unit", "if", "index",
2224 "cyls", "heads", "secs", "trans",
2225 "media", "snapshot", "file",
2226 "cache", "format", "serial",
2227 "werror", "addr",
2228 NULL };
2230 if (check_params(buf, sizeof(buf), params, str) < 0) {
2231 fprintf(stderr, "qemu: unknown parameter '%s' in '%s'\n",
2232 buf, str);
2233 return -1;
2236 file[0] = 0;
2237 cyls = heads = secs = 0;
2238 bus_id = 0;
2239 unit_id = -1;
2240 translation = BIOS_ATA_TRANSLATION_AUTO;
2241 index = -1;
2242 cache = 1;
2244 if (machine->use_scsi) {
2245 type = IF_SCSI;
2246 max_devs = MAX_SCSI_DEVS;
2247 pstrcpy(devname, sizeof(devname), "scsi");
2248 } else {
2249 type = IF_IDE;
2250 max_devs = MAX_IDE_DEVS;
2251 pstrcpy(devname, sizeof(devname), "ide");
2253 media = MEDIA_DISK;
2255 /* extract parameters */
2257 if (get_param_value(buf, sizeof(buf), "bus", str)) {
2258 bus_id = strtol(buf, NULL, 0);
2259 if (bus_id < 0) {
2260 fprintf(stderr, "qemu: '%s' invalid bus id\n", str);
2261 return -1;
2265 if (get_param_value(buf, sizeof(buf), "unit", str)) {
2266 unit_id = strtol(buf, NULL, 0);
2267 if (unit_id < 0) {
2268 fprintf(stderr, "qemu: '%s' invalid unit id\n", str);
2269 return -1;
2273 if (get_param_value(buf, sizeof(buf), "if", str)) {
2274 pstrcpy(devname, sizeof(devname), buf);
2275 if (!strcmp(buf, "ide")) {
2276 type = IF_IDE;
2277 max_devs = MAX_IDE_DEVS;
2278 } else if (!strcmp(buf, "scsi")) {
2279 type = IF_SCSI;
2280 max_devs = MAX_SCSI_DEVS;
2281 } else if (!strcmp(buf, "floppy")) {
2282 type = IF_FLOPPY;
2283 max_devs = 0;
2284 } else if (!strcmp(buf, "pflash")) {
2285 type = IF_PFLASH;
2286 max_devs = 0;
2287 } else if (!strcmp(buf, "mtd")) {
2288 type = IF_MTD;
2289 max_devs = 0;
2290 } else if (!strcmp(buf, "sd")) {
2291 type = IF_SD;
2292 max_devs = 0;
2293 } else if (!strcmp(buf, "virtio")) {
2294 type = IF_VIRTIO;
2295 max_devs = 0;
2296 } else if (!strcmp(buf, "xen")) {
2297 type = IF_XEN;
2298 max_devs = 0;
2299 } else {
2300 fprintf(stderr, "qemu: '%s' unsupported bus type '%s'\n", str, buf);
2301 return -1;
2305 if (get_param_value(buf, sizeof(buf), "index", str)) {
2306 index = strtol(buf, NULL, 0);
2307 if (index < 0) {
2308 fprintf(stderr, "qemu: '%s' invalid index\n", str);
2309 return -1;
2313 if (get_param_value(buf, sizeof(buf), "cyls", str)) {
2314 cyls = strtol(buf, NULL, 0);
2317 if (get_param_value(buf, sizeof(buf), "heads", str)) {
2318 heads = strtol(buf, NULL, 0);
2321 if (get_param_value(buf, sizeof(buf), "secs", str)) {
2322 secs = strtol(buf, NULL, 0);
2325 if (cyls || heads || secs) {
2326 if (cyls < 1 || cyls > 16383) {
2327 fprintf(stderr, "qemu: '%s' invalid physical cyls number\n", str);
2328 return -1;
2330 if (heads < 1 || heads > 16) {
2331 fprintf(stderr, "qemu: '%s' invalid physical heads number\n", str);
2332 return -1;
2334 if (secs < 1 || secs > 63) {
2335 fprintf(stderr, "qemu: '%s' invalid physical secs number\n", str);
2336 return -1;
2340 if (get_param_value(buf, sizeof(buf), "trans", str)) {
2341 if (!cyls) {
2342 fprintf(stderr,
2343 "qemu: '%s' trans must be used with cyls,heads and secs\n",
2344 str);
2345 return -1;
2347 if (!strcmp(buf, "none"))
2348 translation = BIOS_ATA_TRANSLATION_NONE;
2349 else if (!strcmp(buf, "lba"))
2350 translation = BIOS_ATA_TRANSLATION_LBA;
2351 else if (!strcmp(buf, "auto"))
2352 translation = BIOS_ATA_TRANSLATION_AUTO;
2353 else {
2354 fprintf(stderr, "qemu: '%s' invalid translation type\n", str);
2355 return -1;
2359 if (get_param_value(buf, sizeof(buf), "media", str)) {
2360 if (!strcmp(buf, "disk")) {
2361 media = MEDIA_DISK;
2362 } else if (!strcmp(buf, "cdrom")) {
2363 if (cyls || secs || heads) {
2364 fprintf(stderr,
2365 "qemu: '%s' invalid physical CHS format\n", str);
2366 return -1;
2368 media = MEDIA_CDROM;
2369 } else {
2370 fprintf(stderr, "qemu: '%s' invalid media\n", str);
2371 return -1;
2375 if (get_param_value(buf, sizeof(buf), "snapshot", str)) {
2376 if (!strcmp(buf, "on"))
2377 snapshot = 1;
2378 else if (!strcmp(buf, "off"))
2379 snapshot = 0;
2380 else {
2381 fprintf(stderr, "qemu: '%s' invalid snapshot option\n", str);
2382 return -1;
2386 if (get_param_value(buf, sizeof(buf), "cache", str)) {
2387 if (!strcmp(buf, "off") || !strcmp(buf, "none"))
2388 cache = 0;
2389 else if (!strcmp(buf, "writethrough"))
2390 cache = 1;
2391 else if (!strcmp(buf, "writeback"))
2392 cache = 2;
2393 else {
2394 fprintf(stderr, "qemu: invalid cache option\n");
2395 return -1;
2399 if (get_param_value(buf, sizeof(buf), "format", str)) {
2400 if (strcmp(buf, "?") == 0) {
2401 fprintf(stderr, "qemu: Supported formats:");
2402 bdrv_iterate_format(bdrv_format_print, NULL);
2403 fprintf(stderr, "\n");
2404 return -1;
2406 drv = bdrv_find_format(buf);
2407 if (!drv) {
2408 fprintf(stderr, "qemu: '%s' invalid format\n", buf);
2409 return -1;
2413 if (arg->file == NULL)
2414 get_param_value(file, sizeof(file), "file", str);
2415 else
2416 pstrcpy(file, sizeof(file), arg->file);
2418 if (!get_param_value(serial, sizeof(serial), "serial", str))
2419 memset(serial, 0, sizeof(serial));
2421 onerror = BLOCK_ERR_STOP_ENOSPC;
2422 if (get_param_value(buf, sizeof(serial), "werror", str)) {
2423 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO) {
2424 fprintf(stderr, "werror is no supported by this format\n");
2425 return -1;
2427 if (!strcmp(buf, "ignore"))
2428 onerror = BLOCK_ERR_IGNORE;
2429 else if (!strcmp(buf, "enospc"))
2430 onerror = BLOCK_ERR_STOP_ENOSPC;
2431 else if (!strcmp(buf, "stop"))
2432 onerror = BLOCK_ERR_STOP_ANY;
2433 else if (!strcmp(buf, "report"))
2434 onerror = BLOCK_ERR_REPORT;
2435 else {
2436 fprintf(stderr, "qemu: '%s' invalid write error action\n", buf);
2437 return -1;
2441 devaddr = NULL;
2442 if (get_param_value(buf, sizeof(buf), "addr", str)) {
2443 if (type != IF_VIRTIO) {
2444 fprintf(stderr, "addr is not supported by in '%s'\n", str);
2445 return -1;
2447 devaddr = strdup(buf);
2450 /* compute bus and unit according index */
2452 if (index != -1) {
2453 if (bus_id != 0 || unit_id != -1) {
2454 fprintf(stderr,
2455 "qemu: '%s' index cannot be used with bus and unit\n", str);
2456 return -1;
2458 if (max_devs == 0)
2460 unit_id = index;
2461 bus_id = 0;
2462 } else {
2463 unit_id = index % max_devs;
2464 bus_id = index / max_devs;
2468 /* if user doesn't specify a unit_id,
2469 * try to find the first free
2472 if (unit_id == -1) {
2473 unit_id = 0;
2474 while (drive_get_index(type, bus_id, unit_id) != -1) {
2475 unit_id++;
2476 if (max_devs && unit_id >= max_devs) {
2477 unit_id -= max_devs;
2478 bus_id++;
2483 /* check unit id */
2485 if (max_devs && unit_id >= max_devs) {
2486 fprintf(stderr, "qemu: '%s' unit %d too big (max is %d)\n",
2487 str, unit_id, max_devs - 1);
2488 return -1;
2492 * ignore multiple definitions
2495 if (drive_get_index(type, bus_id, unit_id) != -1)
2496 return -2;
2498 /* init */
2500 if (type == IF_IDE || type == IF_SCSI)
2501 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
2502 if (max_devs)
2503 snprintf(buf, sizeof(buf), "%s%i%s%i",
2504 devname, bus_id, mediastr, unit_id);
2505 else
2506 snprintf(buf, sizeof(buf), "%s%s%i",
2507 devname, mediastr, unit_id);
2508 bdrv = bdrv_new(buf);
2509 drives_table_idx = drive_get_free_idx();
2510 drives_table[drives_table_idx].bdrv = bdrv;
2511 drives_table[drives_table_idx].devaddr = devaddr;
2512 drives_table[drives_table_idx].type = type;
2513 drives_table[drives_table_idx].bus = bus_id;
2514 drives_table[drives_table_idx].unit = unit_id;
2515 drives_table[drives_table_idx].onerror = onerror;
2516 drives_table[drives_table_idx].drive_opt_idx = arg - drives_opt;
2517 strncpy(drives_table[drives_table_idx].serial, serial, sizeof(serial));
2518 nb_drives++;
2520 switch(type) {
2521 case IF_IDE:
2522 case IF_SCSI:
2523 case IF_XEN:
2524 switch(media) {
2525 case MEDIA_DISK:
2526 if (cyls != 0) {
2527 bdrv_set_geometry_hint(bdrv, cyls, heads, secs);
2528 bdrv_set_translation_hint(bdrv, translation);
2530 break;
2531 case MEDIA_CDROM:
2532 bdrv_set_type_hint(bdrv, BDRV_TYPE_CDROM);
2533 break;
2535 break;
2536 case IF_SD:
2537 /* FIXME: This isn't really a floppy, but it's a reasonable
2538 approximation. */
2539 case IF_FLOPPY:
2540 bdrv_set_type_hint(bdrv, BDRV_TYPE_FLOPPY);
2541 break;
2542 case IF_PFLASH:
2543 case IF_MTD:
2544 case IF_VIRTIO:
2545 break;
2546 case IF_COUNT:
2547 abort();
2549 if (!file[0])
2550 return -2;
2551 bdrv_flags = 0;
2552 if (snapshot) {
2553 bdrv_flags |= BDRV_O_SNAPSHOT;
2554 cache = 2; /* always use write-back with snapshot */
2556 if (cache == 0) /* no caching */
2557 bdrv_flags |= BDRV_O_NOCACHE;
2558 else if (cache == 2) /* write-back */
2559 bdrv_flags |= BDRV_O_CACHE_WB;
2560 if (bdrv_open2(bdrv, file, bdrv_flags, drv) < 0) {
2561 fprintf(stderr, "qemu: could not open disk image %s\n",
2562 file);
2563 return -1;
2565 if (bdrv_key_required(bdrv))
2566 autostart = 0;
2567 return drives_table_idx;
2570 static void numa_add(const char *optarg)
2572 char option[128];
2573 char *endptr;
2574 unsigned long long value, endvalue;
2575 int nodenr;
2577 optarg = get_opt_name(option, 128, optarg, ',') + 1;
2578 if (!strcmp(option, "node")) {
2579 if (get_param_value(option, 128, "nodeid", optarg) == 0) {
2580 nodenr = nb_numa_nodes;
2581 } else {
2582 nodenr = strtoull(option, NULL, 10);
2585 if (get_param_value(option, 128, "mem", optarg) == 0) {
2586 node_mem[nodenr] = 0;
2587 } else {
2588 value = strtoull(option, &endptr, 0);
2589 switch (*endptr) {
2590 case 0: case 'M': case 'm':
2591 value <<= 20;
2592 break;
2593 case 'G': case 'g':
2594 value <<= 30;
2595 break;
2597 node_mem[nodenr] = value;
2599 if (get_param_value(option, 128, "cpus", optarg) == 0) {
2600 node_cpumask[nodenr] = 0;
2601 } else {
2602 value = strtoull(option, &endptr, 10);
2603 if (value >= 64) {
2604 value = 63;
2605 fprintf(stderr, "only 64 CPUs in NUMA mode supported.\n");
2606 } else {
2607 if (*endptr == '-') {
2608 endvalue = strtoull(endptr+1, &endptr, 10);
2609 if (endvalue >= 63) {
2610 endvalue = 62;
2611 fprintf(stderr,
2612 "only 63 CPUs in NUMA mode supported.\n");
2614 value = (1 << (endvalue + 1)) - (1 << value);
2615 } else {
2616 value = 1 << value;
2619 node_cpumask[nodenr] = value;
2621 nb_numa_nodes++;
2623 return;
2626 /***********************************************************/
2627 /* USB devices */
2629 static USBPort *used_usb_ports;
2630 static USBPort *free_usb_ports;
2632 /* ??? Maybe change this to register a hub to keep track of the topology. */
2633 void qemu_register_usb_port(USBPort *port, void *opaque, int index,
2634 usb_attachfn attach)
2636 port->opaque = opaque;
2637 port->index = index;
2638 port->attach = attach;
2639 port->next = free_usb_ports;
2640 free_usb_ports = port;
2643 int usb_device_add_dev(USBDevice *dev)
2645 USBPort *port;
2647 /* Find a USB port to add the device to. */
2648 port = free_usb_ports;
2649 if (!port->next) {
2650 USBDevice *hub;
2652 /* Create a new hub and chain it on. */
2653 free_usb_ports = NULL;
2654 port->next = used_usb_ports;
2655 used_usb_ports = port;
2657 hub = usb_hub_init(VM_USB_HUB_SIZE);
2658 usb_attach(port, hub);
2659 port = free_usb_ports;
2662 free_usb_ports = port->next;
2663 port->next = used_usb_ports;
2664 used_usb_ports = port;
2665 usb_attach(port, dev);
2666 return 0;
2669 static void usb_msd_password_cb(void *opaque, int err)
2671 USBDevice *dev = opaque;
2673 if (!err)
2674 usb_device_add_dev(dev);
2675 else
2676 dev->handle_destroy(dev);
2679 static int usb_device_add(const char *devname, int is_hotplug)
2681 const char *p;
2682 USBDevice *dev;
2684 if (!free_usb_ports)
2685 return -1;
2687 if (strstart(devname, "host:", &p)) {
2688 dev = usb_host_device_open(p);
2689 } else if (!strcmp(devname, "mouse")) {
2690 dev = usb_mouse_init();
2691 } else if (!strcmp(devname, "tablet")) {
2692 dev = usb_tablet_init();
2693 } else if (!strcmp(devname, "keyboard")) {
2694 dev = usb_keyboard_init();
2695 } else if (strstart(devname, "disk:", &p)) {
2696 BlockDriverState *bs;
2698 dev = usb_msd_init(p);
2699 if (!dev)
2700 return -1;
2701 bs = usb_msd_get_bdrv(dev);
2702 if (bdrv_key_required(bs)) {
2703 autostart = 0;
2704 if (is_hotplug) {
2705 monitor_read_bdrv_key_start(cur_mon, bs, usb_msd_password_cb,
2706 dev);
2707 return 0;
2710 } else if (!strcmp(devname, "wacom-tablet")) {
2711 dev = usb_wacom_init();
2712 } else if (strstart(devname, "serial:", &p)) {
2713 dev = usb_serial_init(p);
2714 #ifdef CONFIG_BRLAPI
2715 } else if (!strcmp(devname, "braille")) {
2716 dev = usb_baum_init();
2717 #endif
2718 } else if (strstart(devname, "net:", &p)) {
2719 int nic = nb_nics;
2721 if (net_client_init(NULL, "nic", p) < 0)
2722 return -1;
2723 nd_table[nic].model = "usb";
2724 dev = usb_net_init(&nd_table[nic]);
2725 } else if (!strcmp(devname, "bt") || strstart(devname, "bt:", &p)) {
2726 dev = usb_bt_init(devname[2] ? hci_init(p) :
2727 bt_new_hci(qemu_find_bt_vlan(0)));
2728 } else {
2729 return -1;
2731 if (!dev)
2732 return -1;
2734 return usb_device_add_dev(dev);
2737 int usb_device_del_addr(int bus_num, int addr)
2739 USBPort *port;
2740 USBPort **lastp;
2741 USBDevice *dev;
2743 if (!used_usb_ports)
2744 return -1;
2746 if (bus_num != 0)
2747 return -1;
2749 lastp = &used_usb_ports;
2750 port = used_usb_ports;
2751 while (port && port->dev->addr != addr) {
2752 lastp = &port->next;
2753 port = port->next;
2756 if (!port)
2757 return -1;
2759 dev = port->dev;
2760 *lastp = port->next;
2761 usb_attach(port, NULL);
2762 dev->handle_destroy(dev);
2763 port->next = free_usb_ports;
2764 free_usb_ports = port;
2765 return 0;
2768 static int usb_device_del(const char *devname)
2770 int bus_num, addr;
2771 const char *p;
2773 if (strstart(devname, "host:", &p))
2774 return usb_host_device_close(p);
2776 if (!used_usb_ports)
2777 return -1;
2779 p = strchr(devname, '.');
2780 if (!p)
2781 return -1;
2782 bus_num = strtoul(devname, NULL, 0);
2783 addr = strtoul(p + 1, NULL, 0);
2785 return usb_device_del_addr(bus_num, addr);
2788 void do_usb_add(Monitor *mon, const char *devname)
2790 usb_device_add(devname, 1);
2793 void do_usb_del(Monitor *mon, const char *devname)
2795 usb_device_del(devname);
2798 void usb_info(Monitor *mon)
2800 USBDevice *dev;
2801 USBPort *port;
2802 const char *speed_str;
2804 if (!usb_enabled) {
2805 monitor_printf(mon, "USB support not enabled\n");
2806 return;
2809 for (port = used_usb_ports; port; port = port->next) {
2810 dev = port->dev;
2811 if (!dev)
2812 continue;
2813 switch(dev->speed) {
2814 case USB_SPEED_LOW:
2815 speed_str = "1.5";
2816 break;
2817 case USB_SPEED_FULL:
2818 speed_str = "12";
2819 break;
2820 case USB_SPEED_HIGH:
2821 speed_str = "480";
2822 break;
2823 default:
2824 speed_str = "?";
2825 break;
2827 monitor_printf(mon, " Device %d.%d, Speed %s Mb/s, Product %s\n",
2828 0, dev->addr, speed_str, dev->devname);
2832 /***********************************************************/
2833 /* PCMCIA/Cardbus */
2835 static struct pcmcia_socket_entry_s {
2836 PCMCIASocket *socket;
2837 struct pcmcia_socket_entry_s *next;
2838 } *pcmcia_sockets = 0;
2840 void pcmcia_socket_register(PCMCIASocket *socket)
2842 struct pcmcia_socket_entry_s *entry;
2844 entry = qemu_malloc(sizeof(struct pcmcia_socket_entry_s));
2845 entry->socket = socket;
2846 entry->next = pcmcia_sockets;
2847 pcmcia_sockets = entry;
2850 void pcmcia_socket_unregister(PCMCIASocket *socket)
2852 struct pcmcia_socket_entry_s *entry, **ptr;
2854 ptr = &pcmcia_sockets;
2855 for (entry = *ptr; entry; ptr = &entry->next, entry = *ptr)
2856 if (entry->socket == socket) {
2857 *ptr = entry->next;
2858 qemu_free(entry);
2862 void pcmcia_info(Monitor *mon)
2864 struct pcmcia_socket_entry_s *iter;
2866 if (!pcmcia_sockets)
2867 monitor_printf(mon, "No PCMCIA sockets\n");
2869 for (iter = pcmcia_sockets; iter; iter = iter->next)
2870 monitor_printf(mon, "%s: %s\n", iter->socket->slot_string,
2871 iter->socket->attached ? iter->socket->card_string :
2872 "Empty");
2875 /***********************************************************/
2876 /* register display */
2878 struct DisplayAllocator default_allocator = {
2879 defaultallocator_create_displaysurface,
2880 defaultallocator_resize_displaysurface,
2881 defaultallocator_free_displaysurface
2884 void register_displaystate(DisplayState *ds)
2886 DisplayState **s;
2887 s = &display_state;
2888 while (*s != NULL)
2889 s = &(*s)->next;
2890 ds->next = NULL;
2891 *s = ds;
2894 DisplayState *get_displaystate(void)
2896 return display_state;
2899 DisplayAllocator *register_displayallocator(DisplayState *ds, DisplayAllocator *da)
2901 if(ds->allocator == &default_allocator) ds->allocator = da;
2902 return ds->allocator;
2905 /* dumb display */
2907 static void dumb_display_init(void)
2909 DisplayState *ds = qemu_mallocz(sizeof(DisplayState));
2910 ds->allocator = &default_allocator;
2911 ds->surface = qemu_create_displaysurface(ds, 640, 480);
2912 register_displaystate(ds);
2915 /***********************************************************/
2916 /* I/O handling */
2918 typedef struct IOHandlerRecord {
2919 int fd;
2920 IOCanRWHandler *fd_read_poll;
2921 IOHandler *fd_read;
2922 IOHandler *fd_write;
2923 int deleted;
2924 void *opaque;
2925 /* temporary data */
2926 struct pollfd *ufd;
2927 struct IOHandlerRecord *next;
2928 } IOHandlerRecord;
2930 static IOHandlerRecord *first_io_handler;
2932 /* XXX: fd_read_poll should be suppressed, but an API change is
2933 necessary in the character devices to suppress fd_can_read(). */
2934 int qemu_set_fd_handler2(int fd,
2935 IOCanRWHandler *fd_read_poll,
2936 IOHandler *fd_read,
2937 IOHandler *fd_write,
2938 void *opaque)
2940 IOHandlerRecord **pioh, *ioh;
2942 if (!fd_read && !fd_write) {
2943 pioh = &first_io_handler;
2944 for(;;) {
2945 ioh = *pioh;
2946 if (ioh == NULL)
2947 break;
2948 if (ioh->fd == fd) {
2949 ioh->deleted = 1;
2950 break;
2952 pioh = &ioh->next;
2954 } else {
2955 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
2956 if (ioh->fd == fd)
2957 goto found;
2959 ioh = qemu_mallocz(sizeof(IOHandlerRecord));
2960 ioh->next = first_io_handler;
2961 first_io_handler = ioh;
2962 found:
2963 ioh->fd = fd;
2964 ioh->fd_read_poll = fd_read_poll;
2965 ioh->fd_read = fd_read;
2966 ioh->fd_write = fd_write;
2967 ioh->opaque = opaque;
2968 ioh->deleted = 0;
2970 return 0;
2973 int qemu_set_fd_handler(int fd,
2974 IOHandler *fd_read,
2975 IOHandler *fd_write,
2976 void *opaque)
2978 return qemu_set_fd_handler2(fd, NULL, fd_read, fd_write, opaque);
2981 #ifdef _WIN32
2982 /***********************************************************/
2983 /* Polling handling */
2985 typedef struct PollingEntry {
2986 PollingFunc *func;
2987 void *opaque;
2988 struct PollingEntry *next;
2989 } PollingEntry;
2991 static PollingEntry *first_polling_entry;
2993 int qemu_add_polling_cb(PollingFunc *func, void *opaque)
2995 PollingEntry **ppe, *pe;
2996 pe = qemu_mallocz(sizeof(PollingEntry));
2997 pe->func = func;
2998 pe->opaque = opaque;
2999 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next);
3000 *ppe = pe;
3001 return 0;
3004 void qemu_del_polling_cb(PollingFunc *func, void *opaque)
3006 PollingEntry **ppe, *pe;
3007 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next) {
3008 pe = *ppe;
3009 if (pe->func == func && pe->opaque == opaque) {
3010 *ppe = pe->next;
3011 qemu_free(pe);
3012 break;
3017 /***********************************************************/
3018 /* Wait objects support */
3019 typedef struct WaitObjects {
3020 int num;
3021 HANDLE events[MAXIMUM_WAIT_OBJECTS + 1];
3022 WaitObjectFunc *func[MAXIMUM_WAIT_OBJECTS + 1];
3023 void *opaque[MAXIMUM_WAIT_OBJECTS + 1];
3024 } WaitObjects;
3026 static WaitObjects wait_objects = {0};
3028 int qemu_add_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
3030 WaitObjects *w = &wait_objects;
3032 if (w->num >= MAXIMUM_WAIT_OBJECTS)
3033 return -1;
3034 w->events[w->num] = handle;
3035 w->func[w->num] = func;
3036 w->opaque[w->num] = opaque;
3037 w->num++;
3038 return 0;
3041 void qemu_del_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
3043 int i, found;
3044 WaitObjects *w = &wait_objects;
3046 found = 0;
3047 for (i = 0; i < w->num; i++) {
3048 if (w->events[i] == handle)
3049 found = 1;
3050 if (found) {
3051 w->events[i] = w->events[i + 1];
3052 w->func[i] = w->func[i + 1];
3053 w->opaque[i] = w->opaque[i + 1];
3056 if (found)
3057 w->num--;
3059 #endif
3061 /***********************************************************/
3062 /* ram save/restore */
3064 static int ram_get_page(QEMUFile *f, uint8_t *buf, int len)
3066 int v;
3068 v = qemu_get_byte(f);
3069 switch(v) {
3070 case 0:
3071 if (qemu_get_buffer(f, buf, len) != len)
3072 return -EIO;
3073 break;
3074 case 1:
3075 v = qemu_get_byte(f);
3076 memset(buf, v, len);
3077 break;
3078 default:
3079 return -EINVAL;
3082 if (qemu_file_has_error(f))
3083 return -EIO;
3085 return 0;
3088 static int ram_load_v1(QEMUFile *f, void *opaque)
3090 int ret;
3091 ram_addr_t i;
3093 if (qemu_get_be32(f) != last_ram_offset)
3094 return -EINVAL;
3095 for(i = 0; i < last_ram_offset; i+= TARGET_PAGE_SIZE) {
3096 ret = ram_get_page(f, qemu_get_ram_ptr(i), TARGET_PAGE_SIZE);
3097 if (ret)
3098 return ret;
3100 return 0;
3103 #define BDRV_HASH_BLOCK_SIZE 1024
3104 #define IOBUF_SIZE 4096
3105 #define RAM_CBLOCK_MAGIC 0xfabe
3107 typedef struct RamDecompressState {
3108 z_stream zstream;
3109 QEMUFile *f;
3110 uint8_t buf[IOBUF_SIZE];
3111 } RamDecompressState;
3113 static int ram_decompress_open(RamDecompressState *s, QEMUFile *f)
3115 int ret;
3116 memset(s, 0, sizeof(*s));
3117 s->f = f;
3118 ret = inflateInit(&s->zstream);
3119 if (ret != Z_OK)
3120 return -1;
3121 return 0;
3124 static int ram_decompress_buf(RamDecompressState *s, uint8_t *buf, int len)
3126 int ret, clen;
3128 s->zstream.avail_out = len;
3129 s->zstream.next_out = buf;
3130 while (s->zstream.avail_out > 0) {
3131 if (s->zstream.avail_in == 0) {
3132 if (qemu_get_be16(s->f) != RAM_CBLOCK_MAGIC)
3133 return -1;
3134 clen = qemu_get_be16(s->f);
3135 if (clen > IOBUF_SIZE)
3136 return -1;
3137 qemu_get_buffer(s->f, s->buf, clen);
3138 s->zstream.avail_in = clen;
3139 s->zstream.next_in = s->buf;
3141 ret = inflate(&s->zstream, Z_PARTIAL_FLUSH);
3142 if (ret != Z_OK && ret != Z_STREAM_END) {
3143 return -1;
3146 return 0;
3149 static void ram_decompress_close(RamDecompressState *s)
3151 inflateEnd(&s->zstream);
3154 #define RAM_SAVE_FLAG_FULL 0x01
3155 #define RAM_SAVE_FLAG_COMPRESS 0x02
3156 #define RAM_SAVE_FLAG_MEM_SIZE 0x04
3157 #define RAM_SAVE_FLAG_PAGE 0x08
3158 #define RAM_SAVE_FLAG_EOS 0x10
3160 static int is_dup_page(uint8_t *page, uint8_t ch)
3162 uint32_t val = ch << 24 | ch << 16 | ch << 8 | ch;
3163 uint32_t *array = (uint32_t *)page;
3164 int i;
3166 for (i = 0; i < (TARGET_PAGE_SIZE / 4); i++) {
3167 if (array[i] != val)
3168 return 0;
3171 return 1;
3174 static int ram_save_block(QEMUFile *f)
3176 static ram_addr_t current_addr = 0;
3177 ram_addr_t saved_addr = current_addr;
3178 ram_addr_t addr = 0;
3179 int found = 0;
3181 while (addr < last_ram_offset) {
3182 if (cpu_physical_memory_get_dirty(current_addr, MIGRATION_DIRTY_FLAG)) {
3183 uint8_t *p;
3185 cpu_physical_memory_reset_dirty(current_addr,
3186 current_addr + TARGET_PAGE_SIZE,
3187 MIGRATION_DIRTY_FLAG);
3189 p = qemu_get_ram_ptr(current_addr);
3191 if (is_dup_page(p, *p)) {
3192 qemu_put_be64(f, current_addr | RAM_SAVE_FLAG_COMPRESS);
3193 qemu_put_byte(f, *p);
3194 } else {
3195 qemu_put_be64(f, current_addr | RAM_SAVE_FLAG_PAGE);
3196 qemu_put_buffer(f, p, TARGET_PAGE_SIZE);
3199 found = 1;
3200 break;
3202 addr += TARGET_PAGE_SIZE;
3203 current_addr = (saved_addr + addr) % last_ram_offset;
3206 return found;
3209 static uint64_t bytes_transferred = 0;
3211 static ram_addr_t ram_save_remaining(void)
3213 ram_addr_t addr;
3214 ram_addr_t count = 0;
3216 for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
3217 if (cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
3218 count++;
3221 return count;
3224 uint64_t ram_bytes_remaining(void)
3226 return ram_save_remaining() * TARGET_PAGE_SIZE;
3229 uint64_t ram_bytes_transferred(void)
3231 return bytes_transferred;
3234 uint64_t ram_bytes_total(void)
3236 return last_ram_offset;
3239 static int ram_save_live(QEMUFile *f, int stage, void *opaque)
3241 ram_addr_t addr;
3242 uint64_t bytes_transferred_last;
3243 double bwidth = 0;
3244 uint64_t expected_time = 0;
3246 if (cpu_physical_sync_dirty_bitmap(0, TARGET_PHYS_ADDR_MAX) != 0) {
3247 qemu_file_set_error(f);
3248 return 0;
3251 if (stage == 1) {
3252 /* Make sure all dirty bits are set */
3253 for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
3254 if (!cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
3255 cpu_physical_memory_set_dirty(addr);
3258 /* Enable dirty memory tracking */
3259 cpu_physical_memory_set_dirty_tracking(1);
3261 qemu_put_be64(f, last_ram_offset | RAM_SAVE_FLAG_MEM_SIZE);
3264 bytes_transferred_last = bytes_transferred;
3265 bwidth = get_clock();
3267 while (!qemu_file_rate_limit(f)) {
3268 int ret;
3270 ret = ram_save_block(f);
3271 bytes_transferred += ret * TARGET_PAGE_SIZE;
3272 if (ret == 0) /* no more blocks */
3273 break;
3276 bwidth = get_clock() - bwidth;
3277 bwidth = (bytes_transferred - bytes_transferred_last) / bwidth;
3279 /* if we haven't transferred anything this round, force expected_time to a
3280 * a very high value, but without crashing */
3281 if (bwidth == 0)
3282 bwidth = 0.000001;
3284 /* try transferring iterative blocks of memory */
3286 if (stage == 3) {
3288 /* flush all remaining blocks regardless of rate limiting */
3289 while (ram_save_block(f) != 0) {
3290 bytes_transferred += TARGET_PAGE_SIZE;
3292 cpu_physical_memory_set_dirty_tracking(0);
3295 qemu_put_be64(f, RAM_SAVE_FLAG_EOS);
3297 expected_time = ram_save_remaining() * TARGET_PAGE_SIZE / bwidth;
3299 return (stage == 2) && (expected_time <= migrate_max_downtime());
3302 static int ram_load_dead(QEMUFile *f, void *opaque)
3304 RamDecompressState s1, *s = &s1;
3305 uint8_t buf[10];
3306 ram_addr_t i;
3308 if (ram_decompress_open(s, f) < 0)
3309 return -EINVAL;
3310 for(i = 0; i < last_ram_offset; i+= BDRV_HASH_BLOCK_SIZE) {
3311 if (ram_decompress_buf(s, buf, 1) < 0) {
3312 fprintf(stderr, "Error while reading ram block header\n");
3313 goto error;
3315 if (buf[0] == 0) {
3316 if (ram_decompress_buf(s, qemu_get_ram_ptr(i),
3317 BDRV_HASH_BLOCK_SIZE) < 0) {
3318 fprintf(stderr, "Error while reading ram block address=0x%08" PRIx64, (uint64_t)i);
3319 goto error;
3321 } else {
3322 error:
3323 printf("Error block header\n");
3324 return -EINVAL;
3327 ram_decompress_close(s);
3329 return 0;
3332 static int ram_load(QEMUFile *f, void *opaque, int version_id)
3334 ram_addr_t addr;
3335 int flags;
3337 if (version_id == 1)
3338 return ram_load_v1(f, opaque);
3340 if (version_id == 2) {
3341 if (qemu_get_be32(f) != last_ram_offset)
3342 return -EINVAL;
3343 return ram_load_dead(f, opaque);
3346 if (version_id != 3)
3347 return -EINVAL;
3349 do {
3350 addr = qemu_get_be64(f);
3352 flags = addr & ~TARGET_PAGE_MASK;
3353 addr &= TARGET_PAGE_MASK;
3355 if (flags & RAM_SAVE_FLAG_MEM_SIZE) {
3356 if (addr != last_ram_offset)
3357 return -EINVAL;
3360 if (flags & RAM_SAVE_FLAG_FULL) {
3361 if (ram_load_dead(f, opaque) < 0)
3362 return -EINVAL;
3365 if (flags & RAM_SAVE_FLAG_COMPRESS) {
3366 uint8_t ch = qemu_get_byte(f);
3367 memset(qemu_get_ram_ptr(addr), ch, TARGET_PAGE_SIZE);
3368 #ifndef _WIN32
3369 if (ch == 0 &&
3370 (!kvm_enabled() || kvm_has_sync_mmu())) {
3371 madvise(qemu_get_ram_ptr(addr), TARGET_PAGE_SIZE, MADV_DONTNEED);
3373 #endif
3374 } else if (flags & RAM_SAVE_FLAG_PAGE)
3375 qemu_get_buffer(f, qemu_get_ram_ptr(addr), TARGET_PAGE_SIZE);
3376 } while (!(flags & RAM_SAVE_FLAG_EOS));
3378 return 0;
3381 void qemu_service_io(void)
3383 qemu_notify_event();
3386 /***********************************************************/
3387 /* bottom halves (can be seen as timers which expire ASAP) */
3389 struct QEMUBH {
3390 QEMUBHFunc *cb;
3391 void *opaque;
3392 int scheduled;
3393 int idle;
3394 int deleted;
3395 QEMUBH *next;
3398 static QEMUBH *first_bh = NULL;
3400 QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
3402 QEMUBH *bh;
3403 bh = qemu_mallocz(sizeof(QEMUBH));
3404 bh->cb = cb;
3405 bh->opaque = opaque;
3406 bh->next = first_bh;
3407 first_bh = bh;
3408 return bh;
3411 int qemu_bh_poll(void)
3413 QEMUBH *bh, **bhp;
3414 int ret;
3416 ret = 0;
3417 for (bh = first_bh; bh; bh = bh->next) {
3418 if (!bh->deleted && bh->scheduled) {
3419 bh->scheduled = 0;
3420 if (!bh->idle)
3421 ret = 1;
3422 bh->idle = 0;
3423 bh->cb(bh->opaque);
3427 /* remove deleted bhs */
3428 bhp = &first_bh;
3429 while (*bhp) {
3430 bh = *bhp;
3431 if (bh->deleted) {
3432 *bhp = bh->next;
3433 qemu_free(bh);
3434 } else
3435 bhp = &bh->next;
3438 return ret;
3441 void qemu_bh_schedule_idle(QEMUBH *bh)
3443 if (bh->scheduled)
3444 return;
3445 bh->scheduled = 1;
3446 bh->idle = 1;
3449 void qemu_bh_schedule(QEMUBH *bh)
3451 if (bh->scheduled)
3452 return;
3453 bh->scheduled = 1;
3454 bh->idle = 0;
3455 /* stop the currently executing CPU to execute the BH ASAP */
3456 qemu_notify_event();
3459 void qemu_bh_cancel(QEMUBH *bh)
3461 bh->scheduled = 0;
3464 void qemu_bh_delete(QEMUBH *bh)
3466 bh->scheduled = 0;
3467 bh->deleted = 1;
3470 static void qemu_bh_update_timeout(int *timeout)
3472 QEMUBH *bh;
3474 for (bh = first_bh; bh; bh = bh->next) {
3475 if (!bh->deleted && bh->scheduled) {
3476 if (bh->idle) {
3477 /* idle bottom halves will be polled at least
3478 * every 10ms */
3479 *timeout = MIN(10, *timeout);
3480 } else {
3481 /* non-idle bottom halves will be executed
3482 * immediately */
3483 *timeout = 0;
3484 break;
3490 /***********************************************************/
3491 /* machine registration */
3493 static QEMUMachine *first_machine = NULL;
3494 QEMUMachine *current_machine = NULL;
3496 int qemu_register_machine(QEMUMachine *m)
3498 QEMUMachine **pm;
3499 pm = &first_machine;
3500 while (*pm != NULL)
3501 pm = &(*pm)->next;
3502 m->next = NULL;
3503 *pm = m;
3504 return 0;
3507 static QEMUMachine *find_machine(const char *name)
3509 QEMUMachine *m;
3511 for(m = first_machine; m != NULL; m = m->next) {
3512 if (!strcmp(m->name, name))
3513 return m;
3515 return NULL;
3518 static QEMUMachine *find_default_machine(void)
3520 QEMUMachine *m;
3522 for(m = first_machine; m != NULL; m = m->next) {
3523 if (m->is_default) {
3524 return m;
3527 return NULL;
3530 /***********************************************************/
3531 /* main execution loop */
3533 static void gui_update(void *opaque)
3535 uint64_t interval = GUI_REFRESH_INTERVAL;
3536 DisplayState *ds = opaque;
3537 DisplayChangeListener *dcl = ds->listeners;
3539 dpy_refresh(ds);
3541 while (dcl != NULL) {
3542 if (dcl->gui_timer_interval &&
3543 dcl->gui_timer_interval < interval)
3544 interval = dcl->gui_timer_interval;
3545 dcl = dcl->next;
3547 qemu_mod_timer(ds->gui_timer, interval + qemu_get_clock(rt_clock));
3550 static void nographic_update(void *opaque)
3552 uint64_t interval = GUI_REFRESH_INTERVAL;
3554 qemu_mod_timer(nographic_timer, interval + qemu_get_clock(rt_clock));
3557 struct vm_change_state_entry {
3558 VMChangeStateHandler *cb;
3559 void *opaque;
3560 LIST_ENTRY (vm_change_state_entry) entries;
3563 static LIST_HEAD(vm_change_state_head, vm_change_state_entry) vm_change_state_head;
3565 VMChangeStateEntry *qemu_add_vm_change_state_handler(VMChangeStateHandler *cb,
3566 void *opaque)
3568 VMChangeStateEntry *e;
3570 e = qemu_mallocz(sizeof (*e));
3572 e->cb = cb;
3573 e->opaque = opaque;
3574 LIST_INSERT_HEAD(&vm_change_state_head, e, entries);
3575 return e;
3578 void qemu_del_vm_change_state_handler(VMChangeStateEntry *e)
3580 LIST_REMOVE (e, entries);
3581 qemu_free (e);
3584 static void vm_state_notify(int running, int reason)
3586 VMChangeStateEntry *e;
3588 for (e = vm_change_state_head.lh_first; e; e = e->entries.le_next) {
3589 e->cb(e->opaque, running, reason);
3593 static void resume_all_vcpus(void);
3594 static void pause_all_vcpus(void);
3596 void vm_start(void)
3598 if (!vm_running) {
3599 cpu_enable_ticks();
3600 vm_running = 1;
3601 vm_state_notify(1, 0);
3602 qemu_rearm_alarm_timer(alarm_timer);
3603 resume_all_vcpus();
3607 /* reset/shutdown handler */
3609 typedef struct QEMUResetEntry {
3610 QEMUResetHandler *func;
3611 void *opaque;
3612 struct QEMUResetEntry *next;
3613 } QEMUResetEntry;
3615 static QEMUResetEntry *first_reset_entry;
3616 static int reset_requested;
3617 static int shutdown_requested;
3618 static int powerdown_requested;
3619 static int debug_requested;
3620 static int vmstop_requested;
3622 int qemu_shutdown_requested(void)
3624 int r = shutdown_requested;
3625 shutdown_requested = 0;
3626 return r;
3629 int qemu_reset_requested(void)
3631 int r = reset_requested;
3632 reset_requested = 0;
3633 return r;
3636 int qemu_powerdown_requested(void)
3638 int r = powerdown_requested;
3639 powerdown_requested = 0;
3640 return r;
3643 static int qemu_debug_requested(void)
3645 int r = debug_requested;
3646 debug_requested = 0;
3647 return r;
3650 static int qemu_vmstop_requested(void)
3652 int r = vmstop_requested;
3653 vmstop_requested = 0;
3654 return r;
3657 static void do_vm_stop(int reason)
3659 if (vm_running) {
3660 cpu_disable_ticks();
3661 vm_running = 0;
3662 pause_all_vcpus();
3663 vm_state_notify(0, reason);
3667 void qemu_register_reset(QEMUResetHandler *func, void *opaque)
3669 QEMUResetEntry **pre, *re;
3671 pre = &first_reset_entry;
3672 while (*pre != NULL)
3673 pre = &(*pre)->next;
3674 re = qemu_mallocz(sizeof(QEMUResetEntry));
3675 re->func = func;
3676 re->opaque = opaque;
3677 re->next = NULL;
3678 *pre = re;
3681 void qemu_system_reset(void)
3683 QEMUResetEntry *re;
3685 /* reset all devices */
3686 for(re = first_reset_entry; re != NULL; re = re->next) {
3687 re->func(re->opaque);
3691 void qemu_system_reset_request(void)
3693 if (no_reboot) {
3694 shutdown_requested = 1;
3695 } else {
3696 reset_requested = 1;
3698 qemu_notify_event();
3701 void qemu_system_shutdown_request(void)
3703 shutdown_requested = 1;
3704 qemu_notify_event();
3707 void qemu_system_powerdown_request(void)
3709 powerdown_requested = 1;
3710 qemu_notify_event();
3713 #ifdef CONFIG_IOTHREAD
3714 static void qemu_system_vmstop_request(int reason)
3716 vmstop_requested = reason;
3717 qemu_notify_event();
3719 #endif
3721 #ifndef _WIN32
3722 static int io_thread_fd = -1;
3724 static void qemu_event_increment(void)
3726 static const char byte = 0;
3728 if (io_thread_fd == -1)
3729 return;
3731 write(io_thread_fd, &byte, sizeof(byte));
3734 static void qemu_event_read(void *opaque)
3736 int fd = (unsigned long)opaque;
3737 ssize_t len;
3739 /* Drain the notify pipe */
3740 do {
3741 char buffer[512];
3742 len = read(fd, buffer, sizeof(buffer));
3743 } while ((len == -1 && errno == EINTR) || len > 0);
3746 static int qemu_event_init(void)
3748 int err;
3749 int fds[2];
3751 err = pipe(fds);
3752 if (err == -1)
3753 return -errno;
3755 err = fcntl_setfl(fds[0], O_NONBLOCK);
3756 if (err < 0)
3757 goto fail;
3759 err = fcntl_setfl(fds[1], O_NONBLOCK);
3760 if (err < 0)
3761 goto fail;
3763 qemu_set_fd_handler2(fds[0], NULL, qemu_event_read, NULL,
3764 (void *)(unsigned long)fds[0]);
3766 io_thread_fd = fds[1];
3767 return 0;
3769 fail:
3770 close(fds[0]);
3771 close(fds[1]);
3772 return err;
3774 #else
3775 HANDLE qemu_event_handle;
3777 static void dummy_event_handler(void *opaque)
3781 static int qemu_event_init(void)
3783 qemu_event_handle = CreateEvent(NULL, FALSE, FALSE, NULL);
3784 if (!qemu_event_handle) {
3785 perror("Failed CreateEvent");
3786 return -1;
3788 qemu_add_wait_object(qemu_event_handle, dummy_event_handler, NULL);
3789 return 0;
3792 static void qemu_event_increment(void)
3794 SetEvent(qemu_event_handle);
3796 #endif
3798 static int cpu_can_run(CPUState *env)
3800 if (env->stop)
3801 return 0;
3802 if (env->stopped)
3803 return 0;
3804 return 1;
3807 #ifndef CONFIG_IOTHREAD
3808 static int qemu_init_main_loop(void)
3810 return qemu_event_init();
3813 void qemu_init_vcpu(void *_env)
3815 CPUState *env = _env;
3817 if (kvm_enabled())
3818 kvm_init_vcpu(env);
3819 return;
3822 int qemu_cpu_self(void *env)
3824 return 1;
3827 static void resume_all_vcpus(void)
3831 static void pause_all_vcpus(void)
3835 void qemu_cpu_kick(void *env)
3837 return;
3840 void qemu_notify_event(void)
3842 CPUState *env = cpu_single_env;
3844 if (env) {
3845 cpu_exit(env);
3846 #ifdef USE_KQEMU
3847 if (env->kqemu_enabled)
3848 kqemu_cpu_interrupt(env);
3849 #endif
3853 #define qemu_mutex_lock_iothread() do { } while (0)
3854 #define qemu_mutex_unlock_iothread() do { } while (0)
3856 void vm_stop(int reason)
3858 do_vm_stop(reason);
3861 #else /* CONFIG_IOTHREAD */
3863 #include "qemu-thread.h"
3865 QemuMutex qemu_global_mutex;
3866 static QemuMutex qemu_fair_mutex;
3868 static QemuThread io_thread;
3870 static QemuThread *tcg_cpu_thread;
3871 static QemuCond *tcg_halt_cond;
3873 static int qemu_system_ready;
3874 /* cpu creation */
3875 static QemuCond qemu_cpu_cond;
3876 /* system init */
3877 static QemuCond qemu_system_cond;
3878 static QemuCond qemu_pause_cond;
3880 static void block_io_signals(void);
3881 static void unblock_io_signals(void);
3882 static int tcg_has_work(void);
3884 static int qemu_init_main_loop(void)
3886 int ret;
3888 ret = qemu_event_init();
3889 if (ret)
3890 return ret;
3892 qemu_cond_init(&qemu_pause_cond);
3893 qemu_mutex_init(&qemu_fair_mutex);
3894 qemu_mutex_init(&qemu_global_mutex);
3895 qemu_mutex_lock(&qemu_global_mutex);
3897 unblock_io_signals();
3898 qemu_thread_self(&io_thread);
3900 return 0;
3903 static void qemu_wait_io_event(CPUState *env)
3905 while (!tcg_has_work())
3906 qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000);
3908 qemu_mutex_unlock(&qemu_global_mutex);
3911 * Users of qemu_global_mutex can be starved, having no chance
3912 * to acquire it since this path will get to it first.
3913 * So use another lock to provide fairness.
3915 qemu_mutex_lock(&qemu_fair_mutex);
3916 qemu_mutex_unlock(&qemu_fair_mutex);
3918 qemu_mutex_lock(&qemu_global_mutex);
3919 if (env->stop) {
3920 env->stop = 0;
3921 env->stopped = 1;
3922 qemu_cond_signal(&qemu_pause_cond);
3926 static int qemu_cpu_exec(CPUState *env);
3928 static void *kvm_cpu_thread_fn(void *arg)
3930 CPUState *env = arg;
3932 block_io_signals();
3933 qemu_thread_self(env->thread);
3935 /* signal CPU creation */
3936 qemu_mutex_lock(&qemu_global_mutex);
3937 env->created = 1;
3938 qemu_cond_signal(&qemu_cpu_cond);
3940 /* and wait for machine initialization */
3941 while (!qemu_system_ready)
3942 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
3944 while (1) {
3945 if (cpu_can_run(env))
3946 qemu_cpu_exec(env);
3947 qemu_wait_io_event(env);
3950 return NULL;
3953 static void tcg_cpu_exec(void);
3955 static void *tcg_cpu_thread_fn(void *arg)
3957 CPUState *env = arg;
3959 block_io_signals();
3960 qemu_thread_self(env->thread);
3962 /* signal CPU creation */
3963 qemu_mutex_lock(&qemu_global_mutex);
3964 for (env = first_cpu; env != NULL; env = env->next_cpu)
3965 env->created = 1;
3966 qemu_cond_signal(&qemu_cpu_cond);
3968 /* and wait for machine initialization */
3969 while (!qemu_system_ready)
3970 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
3972 while (1) {
3973 tcg_cpu_exec();
3974 qemu_wait_io_event(cur_cpu);
3977 return NULL;
3980 void qemu_cpu_kick(void *_env)
3982 CPUState *env = _env;
3983 qemu_cond_broadcast(env->halt_cond);
3984 if (kvm_enabled())
3985 qemu_thread_signal(env->thread, SIGUSR1);
3988 int qemu_cpu_self(void *env)
3990 return (cpu_single_env != NULL);
3993 static void cpu_signal(int sig)
3995 if (cpu_single_env)
3996 cpu_exit(cpu_single_env);
3999 static void block_io_signals(void)
4001 sigset_t set;
4002 struct sigaction sigact;
4004 sigemptyset(&set);
4005 sigaddset(&set, SIGUSR2);
4006 sigaddset(&set, SIGIO);
4007 sigaddset(&set, SIGALRM);
4008 pthread_sigmask(SIG_BLOCK, &set, NULL);
4010 sigemptyset(&set);
4011 sigaddset(&set, SIGUSR1);
4012 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
4014 memset(&sigact, 0, sizeof(sigact));
4015 sigact.sa_handler = cpu_signal;
4016 sigaction(SIGUSR1, &sigact, NULL);
4019 static void unblock_io_signals(void)
4021 sigset_t set;
4023 sigemptyset(&set);
4024 sigaddset(&set, SIGUSR2);
4025 sigaddset(&set, SIGIO);
4026 sigaddset(&set, SIGALRM);
4027 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
4029 sigemptyset(&set);
4030 sigaddset(&set, SIGUSR1);
4031 pthread_sigmask(SIG_BLOCK, &set, NULL);
4034 static void qemu_signal_lock(unsigned int msecs)
4036 qemu_mutex_lock(&qemu_fair_mutex);
4038 while (qemu_mutex_trylock(&qemu_global_mutex)) {
4039 qemu_thread_signal(tcg_cpu_thread, SIGUSR1);
4040 if (!qemu_mutex_timedlock(&qemu_global_mutex, msecs))
4041 break;
4043 qemu_mutex_unlock(&qemu_fair_mutex);
4046 static void qemu_mutex_lock_iothread(void)
4048 if (kvm_enabled()) {
4049 qemu_mutex_lock(&qemu_fair_mutex);
4050 qemu_mutex_lock(&qemu_global_mutex);
4051 qemu_mutex_unlock(&qemu_fair_mutex);
4052 } else
4053 qemu_signal_lock(100);
4056 static void qemu_mutex_unlock_iothread(void)
4058 qemu_mutex_unlock(&qemu_global_mutex);
4061 static int all_vcpus_paused(void)
4063 CPUState *penv = first_cpu;
4065 while (penv) {
4066 if (!penv->stopped)
4067 return 0;
4068 penv = (CPUState *)penv->next_cpu;
4071 return 1;
4074 static void pause_all_vcpus(void)
4076 CPUState *penv = first_cpu;
4078 while (penv) {
4079 penv->stop = 1;
4080 qemu_thread_signal(penv->thread, SIGUSR1);
4081 qemu_cpu_kick(penv);
4082 penv = (CPUState *)penv->next_cpu;
4085 while (!all_vcpus_paused()) {
4086 qemu_cond_timedwait(&qemu_pause_cond, &qemu_global_mutex, 100);
4087 penv = first_cpu;
4088 while (penv) {
4089 qemu_thread_signal(penv->thread, SIGUSR1);
4090 penv = (CPUState *)penv->next_cpu;
4095 static void resume_all_vcpus(void)
4097 CPUState *penv = first_cpu;
4099 while (penv) {
4100 penv->stop = 0;
4101 penv->stopped = 0;
4102 qemu_thread_signal(penv->thread, SIGUSR1);
4103 qemu_cpu_kick(penv);
4104 penv = (CPUState *)penv->next_cpu;
4108 static void tcg_init_vcpu(void *_env)
4110 CPUState *env = _env;
4111 /* share a single thread for all cpus with TCG */
4112 if (!tcg_cpu_thread) {
4113 env->thread = qemu_mallocz(sizeof(QemuThread));
4114 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
4115 qemu_cond_init(env->halt_cond);
4116 qemu_thread_create(env->thread, tcg_cpu_thread_fn, env);
4117 while (env->created == 0)
4118 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
4119 tcg_cpu_thread = env->thread;
4120 tcg_halt_cond = env->halt_cond;
4121 } else {
4122 env->thread = tcg_cpu_thread;
4123 env->halt_cond = tcg_halt_cond;
4127 static void kvm_start_vcpu(CPUState *env)
4129 kvm_init_vcpu(env);
4130 env->thread = qemu_mallocz(sizeof(QemuThread));
4131 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
4132 qemu_cond_init(env->halt_cond);
4133 qemu_thread_create(env->thread, kvm_cpu_thread_fn, env);
4134 while (env->created == 0)
4135 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
4138 void qemu_init_vcpu(void *_env)
4140 CPUState *env = _env;
4142 if (kvm_enabled())
4143 kvm_start_vcpu(env);
4144 else
4145 tcg_init_vcpu(env);
4148 void qemu_notify_event(void)
4150 qemu_event_increment();
4153 void vm_stop(int reason)
4155 QemuThread me;
4156 qemu_thread_self(&me);
4158 if (!qemu_thread_equal(&me, &io_thread)) {
4159 qemu_system_vmstop_request(reason);
4161 * FIXME: should not return to device code in case
4162 * vm_stop() has been requested.
4164 if (cpu_single_env) {
4165 cpu_exit(cpu_single_env);
4166 cpu_single_env->stop = 1;
4168 return;
4170 do_vm_stop(reason);
4173 #endif
4176 #ifdef _WIN32
4177 static void host_main_loop_wait(int *timeout)
4179 int ret, ret2, i;
4180 PollingEntry *pe;
4183 /* XXX: need to suppress polling by better using win32 events */
4184 ret = 0;
4185 for(pe = first_polling_entry; pe != NULL; pe = pe->next) {
4186 ret |= pe->func(pe->opaque);
4188 if (ret == 0) {
4189 int err;
4190 WaitObjects *w = &wait_objects;
4192 ret = WaitForMultipleObjects(w->num, w->events, FALSE, *timeout);
4193 if (WAIT_OBJECT_0 + 0 <= ret && ret <= WAIT_OBJECT_0 + w->num - 1) {
4194 if (w->func[ret - WAIT_OBJECT_0])
4195 w->func[ret - WAIT_OBJECT_0](w->opaque[ret - WAIT_OBJECT_0]);
4197 /* Check for additional signaled events */
4198 for(i = (ret - WAIT_OBJECT_0 + 1); i < w->num; i++) {
4200 /* Check if event is signaled */
4201 ret2 = WaitForSingleObject(w->events[i], 0);
4202 if(ret2 == WAIT_OBJECT_0) {
4203 if (w->func[i])
4204 w->func[i](w->opaque[i]);
4205 } else if (ret2 == WAIT_TIMEOUT) {
4206 } else {
4207 err = GetLastError();
4208 fprintf(stderr, "WaitForSingleObject error %d %d\n", i, err);
4211 } else if (ret == WAIT_TIMEOUT) {
4212 } else {
4213 err = GetLastError();
4214 fprintf(stderr, "WaitForMultipleObjects error %d %d\n", ret, err);
4218 *timeout = 0;
4220 #else
4221 static void host_main_loop_wait(int *timeout)
4224 #endif
4226 void main_loop_wait(int timeout)
4228 IOHandlerRecord *ioh;
4229 fd_set rfds, wfds, xfds;
4230 int ret, nfds;
4231 struct timeval tv;
4233 qemu_bh_update_timeout(&timeout);
4235 host_main_loop_wait(&timeout);
4237 /* poll any events */
4238 /* XXX: separate device handlers from system ones */
4239 nfds = -1;
4240 FD_ZERO(&rfds);
4241 FD_ZERO(&wfds);
4242 FD_ZERO(&xfds);
4243 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
4244 if (ioh->deleted)
4245 continue;
4246 if (ioh->fd_read &&
4247 (!ioh->fd_read_poll ||
4248 ioh->fd_read_poll(ioh->opaque) != 0)) {
4249 FD_SET(ioh->fd, &rfds);
4250 if (ioh->fd > nfds)
4251 nfds = ioh->fd;
4253 if (ioh->fd_write) {
4254 FD_SET(ioh->fd, &wfds);
4255 if (ioh->fd > nfds)
4256 nfds = ioh->fd;
4260 tv.tv_sec = timeout / 1000;
4261 tv.tv_usec = (timeout % 1000) * 1000;
4263 slirp_select_fill(&nfds, &rfds, &wfds, &xfds);
4265 qemu_mutex_unlock_iothread();
4266 ret = select(nfds + 1, &rfds, &wfds, &xfds, &tv);
4267 qemu_mutex_lock_iothread();
4268 if (ret > 0) {
4269 IOHandlerRecord **pioh;
4271 for(ioh = first_io_handler; ioh != NULL; ioh = ioh->next) {
4272 if (!ioh->deleted && ioh->fd_read && FD_ISSET(ioh->fd, &rfds)) {
4273 ioh->fd_read(ioh->opaque);
4275 if (!ioh->deleted && ioh->fd_write && FD_ISSET(ioh->fd, &wfds)) {
4276 ioh->fd_write(ioh->opaque);
4280 /* remove deleted IO handlers */
4281 pioh = &first_io_handler;
4282 while (*pioh) {
4283 ioh = *pioh;
4284 if (ioh->deleted) {
4285 *pioh = ioh->next;
4286 qemu_free(ioh);
4287 } else
4288 pioh = &ioh->next;
4292 slirp_select_poll(&rfds, &wfds, &xfds, (ret < 0));
4294 /* rearm timer, if not periodic */
4295 if (alarm_timer->flags & ALARM_FLAG_EXPIRED) {
4296 alarm_timer->flags &= ~ALARM_FLAG_EXPIRED;
4297 qemu_rearm_alarm_timer(alarm_timer);
4300 /* vm time timers */
4301 if (vm_running) {
4302 if (!cur_cpu || likely(!(cur_cpu->singlestep_enabled & SSTEP_NOTIMER)))
4303 qemu_run_timers(&active_timers[QEMU_TIMER_VIRTUAL],
4304 qemu_get_clock(vm_clock));
4307 /* real time timers */
4308 qemu_run_timers(&active_timers[QEMU_TIMER_REALTIME],
4309 qemu_get_clock(rt_clock));
4311 /* Check bottom-halves last in case any of the earlier events triggered
4312 them. */
4313 qemu_bh_poll();
4317 static int qemu_cpu_exec(CPUState *env)
4319 int ret;
4320 #ifdef CONFIG_PROFILER
4321 int64_t ti;
4322 #endif
4324 #ifdef CONFIG_PROFILER
4325 ti = profile_getclock();
4326 #endif
4327 if (use_icount) {
4328 int64_t count;
4329 int decr;
4330 qemu_icount -= (env->icount_decr.u16.low + env->icount_extra);
4331 env->icount_decr.u16.low = 0;
4332 env->icount_extra = 0;
4333 count = qemu_next_deadline();
4334 count = (count + (1 << icount_time_shift) - 1)
4335 >> icount_time_shift;
4336 qemu_icount += count;
4337 decr = (count > 0xffff) ? 0xffff : count;
4338 count -= decr;
4339 env->icount_decr.u16.low = decr;
4340 env->icount_extra = count;
4342 ret = cpu_exec(env);
4343 #ifdef CONFIG_PROFILER
4344 qemu_time += profile_getclock() - ti;
4345 #endif
4346 if (use_icount) {
4347 /* Fold pending instructions back into the
4348 instruction counter, and clear the interrupt flag. */
4349 qemu_icount -= (env->icount_decr.u16.low
4350 + env->icount_extra);
4351 env->icount_decr.u32 = 0;
4352 env->icount_extra = 0;
4354 return ret;
4357 static void tcg_cpu_exec(void)
4359 int ret = 0;
4361 if (next_cpu == NULL)
4362 next_cpu = first_cpu;
4363 for (; next_cpu != NULL; next_cpu = next_cpu->next_cpu) {
4364 CPUState *env = cur_cpu = next_cpu;
4366 if (!vm_running)
4367 break;
4368 if (timer_alarm_pending) {
4369 timer_alarm_pending = 0;
4370 break;
4372 if (cpu_can_run(env))
4373 ret = qemu_cpu_exec(env);
4374 if (ret == EXCP_DEBUG) {
4375 gdb_set_stop_cpu(env);
4376 debug_requested = 1;
4377 break;
4382 static int cpu_has_work(CPUState *env)
4384 if (env->stop)
4385 return 1;
4386 if (env->stopped)
4387 return 0;
4388 if (!env->halted)
4389 return 1;
4390 if (qemu_cpu_has_work(env))
4391 return 1;
4392 return 0;
4395 static int tcg_has_work(void)
4397 CPUState *env;
4399 for (env = first_cpu; env != NULL; env = env->next_cpu)
4400 if (cpu_has_work(env))
4401 return 1;
4402 return 0;
4405 static int qemu_calculate_timeout(void)
4407 #ifndef CONFIG_IOTHREAD
4408 int timeout;
4410 if (!vm_running)
4411 timeout = 5000;
4412 else if (tcg_has_work())
4413 timeout = 0;
4414 else if (!use_icount)
4415 timeout = 5000;
4416 else {
4417 /* XXX: use timeout computed from timers */
4418 int64_t add;
4419 int64_t delta;
4420 /* Advance virtual time to the next event. */
4421 if (use_icount == 1) {
4422 /* When not using an adaptive execution frequency
4423 we tend to get badly out of sync with real time,
4424 so just delay for a reasonable amount of time. */
4425 delta = 0;
4426 } else {
4427 delta = cpu_get_icount() - cpu_get_clock();
4429 if (delta > 0) {
4430 /* If virtual time is ahead of real time then just
4431 wait for IO. */
4432 timeout = (delta / 1000000) + 1;
4433 } else {
4434 /* Wait for either IO to occur or the next
4435 timer event. */
4436 add = qemu_next_deadline();
4437 /* We advance the timer before checking for IO.
4438 Limit the amount we advance so that early IO
4439 activity won't get the guest too far ahead. */
4440 if (add > 10000000)
4441 add = 10000000;
4442 delta += add;
4443 add = (add + (1 << icount_time_shift) - 1)
4444 >> icount_time_shift;
4445 qemu_icount += add;
4446 timeout = delta / 1000000;
4447 if (timeout < 0)
4448 timeout = 0;
4452 return timeout;
4453 #else /* CONFIG_IOTHREAD */
4454 return 1000;
4455 #endif
4458 static int vm_can_run(void)
4460 if (powerdown_requested)
4461 return 0;
4462 if (reset_requested)
4463 return 0;
4464 if (shutdown_requested)
4465 return 0;
4466 if (debug_requested)
4467 return 0;
4468 return 1;
4471 static void main_loop(void)
4473 int r;
4475 #ifdef CONFIG_IOTHREAD
4476 qemu_system_ready = 1;
4477 qemu_cond_broadcast(&qemu_system_cond);
4478 #endif
4480 for (;;) {
4481 do {
4482 #ifdef CONFIG_PROFILER
4483 int64_t ti;
4484 #endif
4485 #ifndef CONFIG_IOTHREAD
4486 tcg_cpu_exec();
4487 #endif
4488 #ifdef CONFIG_PROFILER
4489 ti = profile_getclock();
4490 #endif
4491 main_loop_wait(qemu_calculate_timeout());
4492 #ifdef CONFIG_PROFILER
4493 dev_time += profile_getclock() - ti;
4494 #endif
4495 } while (vm_can_run());
4497 if (qemu_debug_requested())
4498 vm_stop(EXCP_DEBUG);
4499 if (qemu_shutdown_requested()) {
4500 if (no_shutdown) {
4501 vm_stop(0);
4502 no_shutdown = 0;
4503 } else
4504 break;
4506 if (qemu_reset_requested()) {
4507 pause_all_vcpus();
4508 qemu_system_reset();
4509 resume_all_vcpus();
4511 if (qemu_powerdown_requested())
4512 qemu_system_powerdown();
4513 if ((r = qemu_vmstop_requested()))
4514 vm_stop(r);
4516 pause_all_vcpus();
4519 static void version(void)
4521 printf("QEMU PC emulator version " QEMU_VERSION QEMU_PKGVERSION ", Copyright (c) 2003-2008 Fabrice Bellard\n");
4524 static void help(int exitcode)
4526 version();
4527 printf("usage: %s [options] [disk_image]\n"
4528 "\n"
4529 "'disk_image' is a raw hard image image for IDE hard disk 0\n"
4530 "\n"
4531 #define DEF(option, opt_arg, opt_enum, opt_help) \
4532 opt_help
4533 #define DEFHEADING(text) stringify(text) "\n"
4534 #include "qemu-options.h"
4535 #undef DEF
4536 #undef DEFHEADING
4537 #undef GEN_DOCS
4538 "\n"
4539 "During emulation, the following keys are useful:\n"
4540 "ctrl-alt-f toggle full screen\n"
4541 "ctrl-alt-n switch to virtual console 'n'\n"
4542 "ctrl-alt toggle mouse and keyboard grab\n"
4543 "\n"
4544 "When using -nographic, press 'ctrl-a h' to get some help.\n"
4546 "qemu",
4547 DEFAULT_RAM_SIZE,
4548 #ifndef _WIN32
4549 DEFAULT_NETWORK_SCRIPT,
4550 DEFAULT_NETWORK_DOWN_SCRIPT,
4551 #endif
4552 DEFAULT_GDBSTUB_PORT,
4553 "/tmp/qemu.log");
4554 exit(exitcode);
4557 #define HAS_ARG 0x0001
4559 enum {
4560 #define DEF(option, opt_arg, opt_enum, opt_help) \
4561 opt_enum,
4562 #define DEFHEADING(text)
4563 #include "qemu-options.h"
4564 #undef DEF
4565 #undef DEFHEADING
4566 #undef GEN_DOCS
4569 typedef struct QEMUOption {
4570 const char *name;
4571 int flags;
4572 int index;
4573 } QEMUOption;
4575 static const QEMUOption qemu_options[] = {
4576 { "h", 0, QEMU_OPTION_h },
4577 #define DEF(option, opt_arg, opt_enum, opt_help) \
4578 { option, opt_arg, opt_enum },
4579 #define DEFHEADING(text)
4580 #include "qemu-options.h"
4581 #undef DEF
4582 #undef DEFHEADING
4583 #undef GEN_DOCS
4584 { NULL },
4587 #ifdef HAS_AUDIO
4588 struct soundhw soundhw[] = {
4589 #ifdef HAS_AUDIO_CHOICE
4590 #if defined(TARGET_I386) || defined(TARGET_MIPS)
4592 "pcspk",
4593 "PC speaker",
4596 { .init_isa = pcspk_audio_init }
4598 #endif
4600 #ifdef CONFIG_SB16
4602 "sb16",
4603 "Creative Sound Blaster 16",
4606 { .init_isa = SB16_init }
4608 #endif
4610 #ifdef CONFIG_CS4231A
4612 "cs4231a",
4613 "CS4231A",
4616 { .init_isa = cs4231a_init }
4618 #endif
4620 #ifdef CONFIG_ADLIB
4622 "adlib",
4623 #ifdef HAS_YMF262
4624 "Yamaha YMF262 (OPL3)",
4625 #else
4626 "Yamaha YM3812 (OPL2)",
4627 #endif
4630 { .init_isa = Adlib_init }
4632 #endif
4634 #ifdef CONFIG_GUS
4636 "gus",
4637 "Gravis Ultrasound GF1",
4640 { .init_isa = GUS_init }
4642 #endif
4644 #ifdef CONFIG_AC97
4646 "ac97",
4647 "Intel 82801AA AC97 Audio",
4650 { .init_pci = ac97_init }
4652 #endif
4654 #ifdef CONFIG_ES1370
4656 "es1370",
4657 "ENSONIQ AudioPCI ES1370",
4660 { .init_pci = es1370_init }
4662 #endif
4664 #endif /* HAS_AUDIO_CHOICE */
4666 { NULL, NULL, 0, 0, { NULL } }
4669 static void select_soundhw (const char *optarg)
4671 struct soundhw *c;
4673 if (*optarg == '?') {
4674 show_valid_cards:
4676 printf ("Valid sound card names (comma separated):\n");
4677 for (c = soundhw; c->name; ++c) {
4678 printf ("%-11s %s\n", c->name, c->descr);
4680 printf ("\n-soundhw all will enable all of the above\n");
4681 exit (*optarg != '?');
4683 else {
4684 size_t l;
4685 const char *p;
4686 char *e;
4687 int bad_card = 0;
4689 if (!strcmp (optarg, "all")) {
4690 for (c = soundhw; c->name; ++c) {
4691 c->enabled = 1;
4693 return;
4696 p = optarg;
4697 while (*p) {
4698 e = strchr (p, ',');
4699 l = !e ? strlen (p) : (size_t) (e - p);
4701 for (c = soundhw; c->name; ++c) {
4702 if (!strncmp (c->name, p, l)) {
4703 c->enabled = 1;
4704 break;
4708 if (!c->name) {
4709 if (l > 80) {
4710 fprintf (stderr,
4711 "Unknown sound card name (too big to show)\n");
4713 else {
4714 fprintf (stderr, "Unknown sound card name `%.*s'\n",
4715 (int) l, p);
4717 bad_card = 1;
4719 p += l + (e != NULL);
4722 if (bad_card)
4723 goto show_valid_cards;
4726 #endif
4728 static void select_vgahw (const char *p)
4730 const char *opts;
4732 cirrus_vga_enabled = 0;
4733 std_vga_enabled = 0;
4734 vmsvga_enabled = 0;
4735 xenfb_enabled = 0;
4736 if (strstart(p, "std", &opts)) {
4737 std_vga_enabled = 1;
4738 } else if (strstart(p, "cirrus", &opts)) {
4739 cirrus_vga_enabled = 1;
4740 } else if (strstart(p, "vmware", &opts)) {
4741 vmsvga_enabled = 1;
4742 } else if (strstart(p, "xenfb", &opts)) {
4743 xenfb_enabled = 1;
4744 } else if (!strstart(p, "none", &opts)) {
4745 invalid_vga:
4746 fprintf(stderr, "Unknown vga type: %s\n", p);
4747 exit(1);
4749 while (*opts) {
4750 const char *nextopt;
4752 if (strstart(opts, ",retrace=", &nextopt)) {
4753 opts = nextopt;
4754 if (strstart(opts, "dumb", &nextopt))
4755 vga_retrace_method = VGA_RETRACE_DUMB;
4756 else if (strstart(opts, "precise", &nextopt))
4757 vga_retrace_method = VGA_RETRACE_PRECISE;
4758 else goto invalid_vga;
4759 } else goto invalid_vga;
4760 opts = nextopt;
4764 #ifdef TARGET_I386
4765 static int balloon_parse(const char *arg)
4767 char buf[128];
4768 const char *p;
4770 if (!strcmp(arg, "none")) {
4771 virtio_balloon = 0;
4772 } else if (!strncmp(arg, "virtio", 6)) {
4773 virtio_balloon = 1;
4774 if (arg[6] == ',') {
4775 p = arg + 7;
4776 if (get_param_value(buf, sizeof(buf), "addr", p)) {
4777 virtio_balloon_devaddr = strdup(buf);
4780 } else {
4781 return -1;
4783 return 0;
4785 #endif
4787 #ifdef _WIN32
4788 static BOOL WINAPI qemu_ctrl_handler(DWORD type)
4790 exit(STATUS_CONTROL_C_EXIT);
4791 return TRUE;
4793 #endif
4795 int qemu_uuid_parse(const char *str, uint8_t *uuid)
4797 int ret;
4799 if(strlen(str) != 36)
4800 return -1;
4802 ret = sscanf(str, UUID_FMT, &uuid[0], &uuid[1], &uuid[2], &uuid[3],
4803 &uuid[4], &uuid[5], &uuid[6], &uuid[7], &uuid[8], &uuid[9],
4804 &uuid[10], &uuid[11], &uuid[12], &uuid[13], &uuid[14], &uuid[15]);
4806 if(ret != 16)
4807 return -1;
4809 #ifdef TARGET_I386
4810 smbios_add_field(1, offsetof(struct smbios_type_1, uuid), 16, uuid);
4811 #endif
4813 return 0;
4816 #define MAX_NET_CLIENTS 32
4818 #ifndef _WIN32
4820 static void termsig_handler(int signal)
4822 qemu_system_shutdown_request();
4825 static void sigchld_handler(int signal)
4827 waitpid(-1, NULL, WNOHANG);
4830 static void sighandler_setup(void)
4832 struct sigaction act;
4834 memset(&act, 0, sizeof(act));
4835 act.sa_handler = termsig_handler;
4836 sigaction(SIGINT, &act, NULL);
4837 sigaction(SIGHUP, &act, NULL);
4838 sigaction(SIGTERM, &act, NULL);
4840 act.sa_handler = sigchld_handler;
4841 act.sa_flags = SA_NOCLDSTOP;
4842 sigaction(SIGCHLD, &act, NULL);
4845 #endif
4847 #ifdef _WIN32
4848 /* Look for support files in the same directory as the executable. */
4849 static char *find_datadir(const char *argv0)
4851 char *p;
4852 char buf[MAX_PATH];
4853 DWORD len;
4855 len = GetModuleFileName(NULL, buf, sizeof(buf) - 1);
4856 if (len == 0) {
4857 return NULL;
4860 buf[len] = 0;
4861 p = buf + len - 1;
4862 while (p != buf && *p != '\\')
4863 p--;
4864 *p = 0;
4865 if (access(buf, R_OK) == 0) {
4866 return qemu_strdup(buf);
4868 return NULL;
4870 #else /* !_WIN32 */
4872 /* Find a likely location for support files using the location of the binary.
4873 For installed binaries this will be "$bindir/../share/qemu". When
4874 running from the build tree this will be "$bindir/../pc-bios". */
4875 #define SHARE_SUFFIX "/share/qemu"
4876 #define BUILD_SUFFIX "/pc-bios"
4877 static char *find_datadir(const char *argv0)
4879 char *dir;
4880 char *p = NULL;
4881 char *res;
4882 #ifdef PATH_MAX
4883 char buf[PATH_MAX];
4884 #endif
4885 size_t max_len;
4887 #if defined(__linux__)
4889 int len;
4890 len = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
4891 if (len > 0) {
4892 buf[len] = 0;
4893 p = buf;
4896 #elif defined(__FreeBSD__)
4898 int len;
4899 len = readlink("/proc/curproc/file", buf, sizeof(buf) - 1);
4900 if (len > 0) {
4901 buf[len] = 0;
4902 p = buf;
4905 #endif
4906 /* If we don't have any way of figuring out the actual executable
4907 location then try argv[0]. */
4908 if (!p) {
4909 #ifdef PATH_MAX
4910 p = buf;
4911 #endif
4912 p = realpath(argv0, p);
4913 if (!p) {
4914 return NULL;
4917 dir = dirname(p);
4918 dir = dirname(dir);
4920 max_len = strlen(dir) +
4921 MAX(strlen(SHARE_SUFFIX), strlen(BUILD_SUFFIX)) + 1;
4922 res = qemu_mallocz(max_len);
4923 snprintf(res, max_len, "%s%s", dir, SHARE_SUFFIX);
4924 if (access(res, R_OK)) {
4925 snprintf(res, max_len, "%s%s", dir, BUILD_SUFFIX);
4926 if (access(res, R_OK)) {
4927 qemu_free(res);
4928 res = NULL;
4931 #ifndef PATH_MAX
4932 free(p);
4933 #endif
4934 return res;
4936 #undef SHARE_SUFFIX
4937 #undef BUILD_SUFFIX
4938 #endif
4940 char *qemu_find_file(int type, const char *name)
4942 int len;
4943 const char *subdir;
4944 char *buf;
4946 /* If name contains path separators then try it as a straight path. */
4947 if ((strchr(name, '/') || strchr(name, '\\'))
4948 && access(name, R_OK) == 0) {
4949 return strdup(name);
4951 switch (type) {
4952 case QEMU_FILE_TYPE_BIOS:
4953 subdir = "";
4954 break;
4955 case QEMU_FILE_TYPE_KEYMAP:
4956 subdir = "keymaps/";
4957 break;
4958 default:
4959 abort();
4961 len = strlen(data_dir) + strlen(name) + strlen(subdir) + 2;
4962 buf = qemu_mallocz(len);
4963 snprintf(buf, len, "%s/%s%s", data_dir, subdir, name);
4964 if (access(buf, R_OK)) {
4965 qemu_free(buf);
4966 return NULL;
4968 return buf;
4971 int main(int argc, char **argv, char **envp)
4973 const char *gdbstub_dev = NULL;
4974 uint32_t boot_devices_bitmap = 0;
4975 int i;
4976 int snapshot, linux_boot, net_boot;
4977 const char *initrd_filename;
4978 const char *kernel_filename, *kernel_cmdline;
4979 const char *boot_devices = "";
4980 DisplayState *ds;
4981 DisplayChangeListener *dcl;
4982 int cyls, heads, secs, translation;
4983 const char *net_clients[MAX_NET_CLIENTS];
4984 int nb_net_clients;
4985 const char *bt_opts[MAX_BT_CMDLINE];
4986 int nb_bt_opts;
4987 int hda_index;
4988 int optind;
4989 const char *r, *optarg;
4990 CharDriverState *monitor_hd = NULL;
4991 const char *monitor_device;
4992 const char *serial_devices[MAX_SERIAL_PORTS];
4993 int serial_device_index;
4994 const char *parallel_devices[MAX_PARALLEL_PORTS];
4995 int parallel_device_index;
4996 const char *virtio_consoles[MAX_VIRTIO_CONSOLES];
4997 int virtio_console_index;
4998 const char *loadvm = NULL;
4999 QEMUMachine *machine;
5000 const char *cpu_model;
5001 const char *usb_devices[MAX_USB_CMDLINE];
5002 int usb_devices_index;
5003 #ifndef _WIN32
5004 int fds[2];
5005 #endif
5006 int tb_size;
5007 const char *pid_file = NULL;
5008 const char *incoming = NULL;
5009 #ifndef _WIN32
5010 int fd = 0;
5011 struct passwd *pwd = NULL;
5012 const char *chroot_dir = NULL;
5013 const char *run_as = NULL;
5014 #endif
5015 CPUState *env;
5016 int show_vnc_port = 0;
5018 qemu_cache_utils_init(envp);
5020 LIST_INIT (&vm_change_state_head);
5021 #ifndef _WIN32
5023 struct sigaction act;
5024 sigfillset(&act.sa_mask);
5025 act.sa_flags = 0;
5026 act.sa_handler = SIG_IGN;
5027 sigaction(SIGPIPE, &act, NULL);
5029 #else
5030 SetConsoleCtrlHandler(qemu_ctrl_handler, TRUE);
5031 /* Note: cpu_interrupt() is currently not SMP safe, so we force
5032 QEMU to run on a single CPU */
5034 HANDLE h;
5035 DWORD mask, smask;
5036 int i;
5037 h = GetCurrentProcess();
5038 if (GetProcessAffinityMask(h, &mask, &smask)) {
5039 for(i = 0; i < 32; i++) {
5040 if (mask & (1 << i))
5041 break;
5043 if (i != 32) {
5044 mask = 1 << i;
5045 SetProcessAffinityMask(h, mask);
5049 #endif
5051 module_call_init(MODULE_INIT_MACHINE);
5052 machine = find_default_machine();
5053 cpu_model = NULL;
5054 initrd_filename = NULL;
5055 ram_size = 0;
5056 snapshot = 0;
5057 kernel_filename = NULL;
5058 kernel_cmdline = "";
5059 cyls = heads = secs = 0;
5060 translation = BIOS_ATA_TRANSLATION_AUTO;
5061 monitor_device = "vc:80Cx24C";
5063 serial_devices[0] = "vc:80Cx24C";
5064 for(i = 1; i < MAX_SERIAL_PORTS; i++)
5065 serial_devices[i] = NULL;
5066 serial_device_index = 0;
5068 parallel_devices[0] = "vc:80Cx24C";
5069 for(i = 1; i < MAX_PARALLEL_PORTS; i++)
5070 parallel_devices[i] = NULL;
5071 parallel_device_index = 0;
5073 for(i = 0; i < MAX_VIRTIO_CONSOLES; i++)
5074 virtio_consoles[i] = NULL;
5075 virtio_console_index = 0;
5077 for (i = 0; i < MAX_NODES; i++) {
5078 node_mem[i] = 0;
5079 node_cpumask[i] = 0;
5082 usb_devices_index = 0;
5084 nb_net_clients = 0;
5085 nb_bt_opts = 0;
5086 nb_drives = 0;
5087 nb_drives_opt = 0;
5088 nb_numa_nodes = 0;
5089 hda_index = -1;
5091 nb_nics = 0;
5093 tb_size = 0;
5094 autostart= 1;
5096 register_watchdogs();
5098 optind = 1;
5099 for(;;) {
5100 if (optind >= argc)
5101 break;
5102 r = argv[optind];
5103 if (r[0] != '-') {
5104 hda_index = drive_add(argv[optind++], HD_ALIAS, 0);
5105 } else {
5106 const QEMUOption *popt;
5108 optind++;
5109 /* Treat --foo the same as -foo. */
5110 if (r[1] == '-')
5111 r++;
5112 popt = qemu_options;
5113 for(;;) {
5114 if (!popt->name) {
5115 fprintf(stderr, "%s: invalid option -- '%s'\n",
5116 argv[0], r);
5117 exit(1);
5119 if (!strcmp(popt->name, r + 1))
5120 break;
5121 popt++;
5123 if (popt->flags & HAS_ARG) {
5124 if (optind >= argc) {
5125 fprintf(stderr, "%s: option '%s' requires an argument\n",
5126 argv[0], r);
5127 exit(1);
5129 optarg = argv[optind++];
5130 } else {
5131 optarg = NULL;
5134 switch(popt->index) {
5135 case QEMU_OPTION_M:
5136 machine = find_machine(optarg);
5137 if (!machine) {
5138 QEMUMachine *m;
5139 printf("Supported machines are:\n");
5140 for(m = first_machine; m != NULL; m = m->next) {
5141 printf("%-10s %s%s\n",
5142 m->name, m->desc,
5143 m->is_default ? " (default)" : "");
5145 exit(*optarg != '?');
5147 break;
5148 case QEMU_OPTION_cpu:
5149 /* hw initialization will check this */
5150 if (*optarg == '?') {
5151 /* XXX: implement xxx_cpu_list for targets that still miss it */
5152 #if defined(cpu_list)
5153 cpu_list(stdout, &fprintf);
5154 #endif
5155 exit(0);
5156 } else {
5157 cpu_model = optarg;
5159 break;
5160 case QEMU_OPTION_initrd:
5161 initrd_filename = optarg;
5162 break;
5163 case QEMU_OPTION_hda:
5164 if (cyls == 0)
5165 hda_index = drive_add(optarg, HD_ALIAS, 0);
5166 else
5167 hda_index = drive_add(optarg, HD_ALIAS
5168 ",cyls=%d,heads=%d,secs=%d%s",
5169 0, cyls, heads, secs,
5170 translation == BIOS_ATA_TRANSLATION_LBA ?
5171 ",trans=lba" :
5172 translation == BIOS_ATA_TRANSLATION_NONE ?
5173 ",trans=none" : "");
5174 break;
5175 case QEMU_OPTION_hdb:
5176 case QEMU_OPTION_hdc:
5177 case QEMU_OPTION_hdd:
5178 drive_add(optarg, HD_ALIAS, popt->index - QEMU_OPTION_hda);
5179 break;
5180 case QEMU_OPTION_drive:
5181 drive_add(NULL, "%s", optarg);
5182 break;
5183 case QEMU_OPTION_mtdblock:
5184 drive_add(optarg, MTD_ALIAS);
5185 break;
5186 case QEMU_OPTION_sd:
5187 drive_add(optarg, SD_ALIAS);
5188 break;
5189 case QEMU_OPTION_pflash:
5190 drive_add(optarg, PFLASH_ALIAS);
5191 break;
5192 case QEMU_OPTION_snapshot:
5193 snapshot = 1;
5194 break;
5195 case QEMU_OPTION_hdachs:
5197 const char *p;
5198 p = optarg;
5199 cyls = strtol(p, (char **)&p, 0);
5200 if (cyls < 1 || cyls > 16383)
5201 goto chs_fail;
5202 if (*p != ',')
5203 goto chs_fail;
5204 p++;
5205 heads = strtol(p, (char **)&p, 0);
5206 if (heads < 1 || heads > 16)
5207 goto chs_fail;
5208 if (*p != ',')
5209 goto chs_fail;
5210 p++;
5211 secs = strtol(p, (char **)&p, 0);
5212 if (secs < 1 || secs > 63)
5213 goto chs_fail;
5214 if (*p == ',') {
5215 p++;
5216 if (!strcmp(p, "none"))
5217 translation = BIOS_ATA_TRANSLATION_NONE;
5218 else if (!strcmp(p, "lba"))
5219 translation = BIOS_ATA_TRANSLATION_LBA;
5220 else if (!strcmp(p, "auto"))
5221 translation = BIOS_ATA_TRANSLATION_AUTO;
5222 else
5223 goto chs_fail;
5224 } else if (*p != '\0') {
5225 chs_fail:
5226 fprintf(stderr, "qemu: invalid physical CHS format\n");
5227 exit(1);
5229 if (hda_index != -1)
5230 snprintf(drives_opt[hda_index].opt,
5231 sizeof(drives_opt[hda_index].opt),
5232 HD_ALIAS ",cyls=%d,heads=%d,secs=%d%s",
5233 0, cyls, heads, secs,
5234 translation == BIOS_ATA_TRANSLATION_LBA ?
5235 ",trans=lba" :
5236 translation == BIOS_ATA_TRANSLATION_NONE ?
5237 ",trans=none" : "");
5239 break;
5240 case QEMU_OPTION_numa:
5241 if (nb_numa_nodes >= MAX_NODES) {
5242 fprintf(stderr, "qemu: too many NUMA nodes\n");
5243 exit(1);
5245 numa_add(optarg);
5246 break;
5247 case QEMU_OPTION_nographic:
5248 display_type = DT_NOGRAPHIC;
5249 break;
5250 #ifdef CONFIG_CURSES
5251 case QEMU_OPTION_curses:
5252 display_type = DT_CURSES;
5253 break;
5254 #endif
5255 case QEMU_OPTION_portrait:
5256 graphic_rotate = 1;
5257 break;
5258 case QEMU_OPTION_kernel:
5259 kernel_filename = optarg;
5260 break;
5261 case QEMU_OPTION_append:
5262 kernel_cmdline = optarg;
5263 break;
5264 case QEMU_OPTION_cdrom:
5265 drive_add(optarg, CDROM_ALIAS);
5266 break;
5267 case QEMU_OPTION_boot:
5268 boot_devices = optarg;
5269 /* We just do some generic consistency checks */
5271 /* Could easily be extended to 64 devices if needed */
5272 const char *p;
5274 boot_devices_bitmap = 0;
5275 for (p = boot_devices; *p != '\0'; p++) {
5276 /* Allowed boot devices are:
5277 * a b : floppy disk drives
5278 * c ... f : IDE disk drives
5279 * g ... m : machine implementation dependant drives
5280 * n ... p : network devices
5281 * It's up to each machine implementation to check
5282 * if the given boot devices match the actual hardware
5283 * implementation and firmware features.
5285 if (*p < 'a' || *p > 'q') {
5286 fprintf(stderr, "Invalid boot device '%c'\n", *p);
5287 exit(1);
5289 if (boot_devices_bitmap & (1 << (*p - 'a'))) {
5290 fprintf(stderr,
5291 "Boot device '%c' was given twice\n",*p);
5292 exit(1);
5294 boot_devices_bitmap |= 1 << (*p - 'a');
5297 break;
5298 case QEMU_OPTION_fda:
5299 case QEMU_OPTION_fdb:
5300 drive_add(optarg, FD_ALIAS, popt->index - QEMU_OPTION_fda);
5301 break;
5302 #ifdef TARGET_I386
5303 case QEMU_OPTION_no_fd_bootchk:
5304 fd_bootchk = 0;
5305 break;
5306 #endif
5307 case QEMU_OPTION_net:
5308 if (nb_net_clients >= MAX_NET_CLIENTS) {
5309 fprintf(stderr, "qemu: too many network clients\n");
5310 exit(1);
5312 net_clients[nb_net_clients] = optarg;
5313 nb_net_clients++;
5314 break;
5315 #ifdef CONFIG_SLIRP
5316 case QEMU_OPTION_tftp:
5317 legacy_tftp_prefix = optarg;
5318 break;
5319 case QEMU_OPTION_bootp:
5320 legacy_bootp_filename = optarg;
5321 break;
5322 #ifndef _WIN32
5323 case QEMU_OPTION_smb:
5324 net_slirp_smb(optarg);
5325 break;
5326 #endif
5327 case QEMU_OPTION_redir:
5328 net_slirp_redir(optarg);
5329 break;
5330 #endif
5331 case QEMU_OPTION_bt:
5332 if (nb_bt_opts >= MAX_BT_CMDLINE) {
5333 fprintf(stderr, "qemu: too many bluetooth options\n");
5334 exit(1);
5336 bt_opts[nb_bt_opts++] = optarg;
5337 break;
5338 #ifdef HAS_AUDIO
5339 case QEMU_OPTION_audio_help:
5340 AUD_help ();
5341 exit (0);
5342 break;
5343 case QEMU_OPTION_soundhw:
5344 select_soundhw (optarg);
5345 break;
5346 #endif
5347 case QEMU_OPTION_h:
5348 help(0);
5349 break;
5350 case QEMU_OPTION_version:
5351 version();
5352 exit(0);
5353 break;
5354 case QEMU_OPTION_m: {
5355 uint64_t value;
5356 char *ptr;
5358 value = strtoul(optarg, &ptr, 10);
5359 switch (*ptr) {
5360 case 0: case 'M': case 'm':
5361 value <<= 20;
5362 break;
5363 case 'G': case 'g':
5364 value <<= 30;
5365 break;
5366 default:
5367 fprintf(stderr, "qemu: invalid ram size: %s\n", optarg);
5368 exit(1);
5371 /* On 32-bit hosts, QEMU is limited by virtual address space */
5372 if (value > (2047 << 20)
5373 #ifndef CONFIG_KQEMU
5374 && HOST_LONG_BITS == 32
5375 #endif
5377 fprintf(stderr, "qemu: at most 2047 MB RAM can be simulated\n");
5378 exit(1);
5380 if (value != (uint64_t)(ram_addr_t)value) {
5381 fprintf(stderr, "qemu: ram size too large\n");
5382 exit(1);
5384 ram_size = value;
5385 break;
5387 case QEMU_OPTION_d:
5389 int mask;
5390 const CPULogItem *item;
5392 mask = cpu_str_to_log_mask(optarg);
5393 if (!mask) {
5394 printf("Log items (comma separated):\n");
5395 for(item = cpu_log_items; item->mask != 0; item++) {
5396 printf("%-10s %s\n", item->name, item->help);
5398 exit(1);
5400 cpu_set_log(mask);
5402 break;
5403 case QEMU_OPTION_s:
5404 gdbstub_dev = "tcp::" DEFAULT_GDBSTUB_PORT;
5405 break;
5406 case QEMU_OPTION_gdb:
5407 gdbstub_dev = optarg;
5408 break;
5409 case QEMU_OPTION_L:
5410 data_dir = optarg;
5411 break;
5412 case QEMU_OPTION_bios:
5413 bios_name = optarg;
5414 break;
5415 case QEMU_OPTION_singlestep:
5416 singlestep = 1;
5417 break;
5418 case QEMU_OPTION_S:
5419 autostart = 0;
5420 break;
5421 #ifndef _WIN32
5422 case QEMU_OPTION_k:
5423 keyboard_layout = optarg;
5424 break;
5425 #endif
5426 case QEMU_OPTION_localtime:
5427 rtc_utc = 0;
5428 break;
5429 case QEMU_OPTION_vga:
5430 select_vgahw (optarg);
5431 break;
5432 #if defined(TARGET_PPC) || defined(TARGET_SPARC)
5433 case QEMU_OPTION_g:
5435 const char *p;
5436 int w, h, depth;
5437 p = optarg;
5438 w = strtol(p, (char **)&p, 10);
5439 if (w <= 0) {
5440 graphic_error:
5441 fprintf(stderr, "qemu: invalid resolution or depth\n");
5442 exit(1);
5444 if (*p != 'x')
5445 goto graphic_error;
5446 p++;
5447 h = strtol(p, (char **)&p, 10);
5448 if (h <= 0)
5449 goto graphic_error;
5450 if (*p == 'x') {
5451 p++;
5452 depth = strtol(p, (char **)&p, 10);
5453 if (depth != 8 && depth != 15 && depth != 16 &&
5454 depth != 24 && depth != 32)
5455 goto graphic_error;
5456 } else if (*p == '\0') {
5457 depth = graphic_depth;
5458 } else {
5459 goto graphic_error;
5462 graphic_width = w;
5463 graphic_height = h;
5464 graphic_depth = depth;
5466 break;
5467 #endif
5468 case QEMU_OPTION_echr:
5470 char *r;
5471 term_escape_char = strtol(optarg, &r, 0);
5472 if (r == optarg)
5473 printf("Bad argument to echr\n");
5474 break;
5476 case QEMU_OPTION_monitor:
5477 monitor_device = optarg;
5478 break;
5479 case QEMU_OPTION_serial:
5480 if (serial_device_index >= MAX_SERIAL_PORTS) {
5481 fprintf(stderr, "qemu: too many serial ports\n");
5482 exit(1);
5484 serial_devices[serial_device_index] = optarg;
5485 serial_device_index++;
5486 break;
5487 case QEMU_OPTION_watchdog:
5488 i = select_watchdog(optarg);
5489 if (i > 0)
5490 exit (i == 1 ? 1 : 0);
5491 break;
5492 case QEMU_OPTION_watchdog_action:
5493 if (select_watchdog_action(optarg) == -1) {
5494 fprintf(stderr, "Unknown -watchdog-action parameter\n");
5495 exit(1);
5497 break;
5498 case QEMU_OPTION_virtiocon:
5499 if (virtio_console_index >= MAX_VIRTIO_CONSOLES) {
5500 fprintf(stderr, "qemu: too many virtio consoles\n");
5501 exit(1);
5503 virtio_consoles[virtio_console_index] = optarg;
5504 virtio_console_index++;
5505 break;
5506 case QEMU_OPTION_parallel:
5507 if (parallel_device_index >= MAX_PARALLEL_PORTS) {
5508 fprintf(stderr, "qemu: too many parallel ports\n");
5509 exit(1);
5511 parallel_devices[parallel_device_index] = optarg;
5512 parallel_device_index++;
5513 break;
5514 case QEMU_OPTION_loadvm:
5515 loadvm = optarg;
5516 break;
5517 case QEMU_OPTION_full_screen:
5518 full_screen = 1;
5519 break;
5520 #ifdef CONFIG_SDL
5521 case QEMU_OPTION_no_frame:
5522 no_frame = 1;
5523 break;
5524 case QEMU_OPTION_alt_grab:
5525 alt_grab = 1;
5526 break;
5527 case QEMU_OPTION_no_quit:
5528 no_quit = 1;
5529 break;
5530 case QEMU_OPTION_sdl:
5531 display_type = DT_SDL;
5532 break;
5533 #endif
5534 case QEMU_OPTION_pidfile:
5535 pid_file = optarg;
5536 break;
5537 #ifdef TARGET_I386
5538 case QEMU_OPTION_win2k_hack:
5539 win2k_install_hack = 1;
5540 break;
5541 case QEMU_OPTION_rtc_td_hack:
5542 rtc_td_hack = 1;
5543 break;
5544 case QEMU_OPTION_acpitable:
5545 if(acpi_table_add(optarg) < 0) {
5546 fprintf(stderr, "Wrong acpi table provided\n");
5547 exit(1);
5549 break;
5550 case QEMU_OPTION_smbios:
5551 if(smbios_entry_add(optarg) < 0) {
5552 fprintf(stderr, "Wrong smbios provided\n");
5553 exit(1);
5555 break;
5556 #endif
5557 #ifdef CONFIG_KQEMU
5558 case QEMU_OPTION_no_kqemu:
5559 kqemu_allowed = 0;
5560 break;
5561 case QEMU_OPTION_kernel_kqemu:
5562 kqemu_allowed = 2;
5563 break;
5564 #endif
5565 #ifdef CONFIG_KVM
5566 case QEMU_OPTION_enable_kvm:
5567 kvm_allowed = 1;
5568 #ifdef CONFIG_KQEMU
5569 kqemu_allowed = 0;
5570 #endif
5571 break;
5572 #endif
5573 case QEMU_OPTION_usb:
5574 usb_enabled = 1;
5575 break;
5576 case QEMU_OPTION_usbdevice:
5577 usb_enabled = 1;
5578 if (usb_devices_index >= MAX_USB_CMDLINE) {
5579 fprintf(stderr, "Too many USB devices\n");
5580 exit(1);
5582 usb_devices[usb_devices_index] = optarg;
5583 usb_devices_index++;
5584 break;
5585 case QEMU_OPTION_smp:
5586 smp_cpus = atoi(optarg);
5587 if (smp_cpus < 1) {
5588 fprintf(stderr, "Invalid number of CPUs\n");
5589 exit(1);
5591 break;
5592 case QEMU_OPTION_vnc:
5593 display_type = DT_VNC;
5594 vnc_display = optarg;
5595 break;
5596 #ifdef TARGET_I386
5597 case QEMU_OPTION_no_acpi:
5598 acpi_enabled = 0;
5599 break;
5600 case QEMU_OPTION_no_hpet:
5601 no_hpet = 1;
5602 break;
5603 case QEMU_OPTION_balloon:
5604 if (balloon_parse(optarg) < 0) {
5605 fprintf(stderr, "Unknown -balloon argument %s\n", optarg);
5606 exit(1);
5608 break;
5609 #endif
5610 case QEMU_OPTION_no_reboot:
5611 no_reboot = 1;
5612 break;
5613 case QEMU_OPTION_no_shutdown:
5614 no_shutdown = 1;
5615 break;
5616 case QEMU_OPTION_show_cursor:
5617 cursor_hide = 0;
5618 break;
5619 case QEMU_OPTION_uuid:
5620 if(qemu_uuid_parse(optarg, qemu_uuid) < 0) {
5621 fprintf(stderr, "Fail to parse UUID string."
5622 " Wrong format.\n");
5623 exit(1);
5625 break;
5626 #ifndef _WIN32
5627 case QEMU_OPTION_daemonize:
5628 daemonize = 1;
5629 break;
5630 #endif
5631 case QEMU_OPTION_option_rom:
5632 if (nb_option_roms >= MAX_OPTION_ROMS) {
5633 fprintf(stderr, "Too many option ROMs\n");
5634 exit(1);
5636 option_rom[nb_option_roms] = optarg;
5637 nb_option_roms++;
5638 break;
5639 #if defined(TARGET_ARM) || defined(TARGET_M68K)
5640 case QEMU_OPTION_semihosting:
5641 semihosting_enabled = 1;
5642 break;
5643 #endif
5644 case QEMU_OPTION_name:
5645 qemu_name = optarg;
5646 break;
5647 #if defined(TARGET_SPARC) || defined(TARGET_PPC)
5648 case QEMU_OPTION_prom_env:
5649 if (nb_prom_envs >= MAX_PROM_ENVS) {
5650 fprintf(stderr, "Too many prom variables\n");
5651 exit(1);
5653 prom_envs[nb_prom_envs] = optarg;
5654 nb_prom_envs++;
5655 break;
5656 #endif
5657 #ifdef TARGET_ARM
5658 case QEMU_OPTION_old_param:
5659 old_param = 1;
5660 break;
5661 #endif
5662 case QEMU_OPTION_clock:
5663 configure_alarms(optarg);
5664 break;
5665 case QEMU_OPTION_startdate:
5667 struct tm tm;
5668 time_t rtc_start_date;
5669 if (!strcmp(optarg, "now")) {
5670 rtc_date_offset = -1;
5671 } else {
5672 if (sscanf(optarg, "%d-%d-%dT%d:%d:%d",
5673 &tm.tm_year,
5674 &tm.tm_mon,
5675 &tm.tm_mday,
5676 &tm.tm_hour,
5677 &tm.tm_min,
5678 &tm.tm_sec) == 6) {
5679 /* OK */
5680 } else if (sscanf(optarg, "%d-%d-%d",
5681 &tm.tm_year,
5682 &tm.tm_mon,
5683 &tm.tm_mday) == 3) {
5684 tm.tm_hour = 0;
5685 tm.tm_min = 0;
5686 tm.tm_sec = 0;
5687 } else {
5688 goto date_fail;
5690 tm.tm_year -= 1900;
5691 tm.tm_mon--;
5692 rtc_start_date = mktimegm(&tm);
5693 if (rtc_start_date == -1) {
5694 date_fail:
5695 fprintf(stderr, "Invalid date format. Valid format are:\n"
5696 "'now' or '2006-06-17T16:01:21' or '2006-06-17'\n");
5697 exit(1);
5699 rtc_date_offset = time(NULL) - rtc_start_date;
5702 break;
5703 case QEMU_OPTION_tb_size:
5704 tb_size = strtol(optarg, NULL, 0);
5705 if (tb_size < 0)
5706 tb_size = 0;
5707 break;
5708 case QEMU_OPTION_icount:
5709 use_icount = 1;
5710 if (strcmp(optarg, "auto") == 0) {
5711 icount_time_shift = -1;
5712 } else {
5713 icount_time_shift = strtol(optarg, NULL, 0);
5715 break;
5716 case QEMU_OPTION_incoming:
5717 incoming = optarg;
5718 break;
5719 #ifndef _WIN32
5720 case QEMU_OPTION_chroot:
5721 chroot_dir = optarg;
5722 break;
5723 case QEMU_OPTION_runas:
5724 run_as = optarg;
5725 break;
5726 #endif
5727 #ifdef CONFIG_XEN
5728 case QEMU_OPTION_xen_domid:
5729 xen_domid = atoi(optarg);
5730 break;
5731 case QEMU_OPTION_xen_create:
5732 xen_mode = XEN_CREATE;
5733 break;
5734 case QEMU_OPTION_xen_attach:
5735 xen_mode = XEN_ATTACH;
5736 break;
5737 #endif
5742 /* If no data_dir is specified then try to find it relative to the
5743 executable path. */
5744 if (!data_dir) {
5745 data_dir = find_datadir(argv[0]);
5747 /* If all else fails use the install patch specified when building. */
5748 if (!data_dir) {
5749 data_dir = CONFIG_QEMU_SHAREDIR;
5752 #if defined(CONFIG_KVM) && defined(CONFIG_KQEMU)
5753 if (kvm_allowed && kqemu_allowed) {
5754 fprintf(stderr,
5755 "You can not enable both KVM and kqemu at the same time\n");
5756 exit(1);
5758 #endif
5760 machine->max_cpus = machine->max_cpus ?: 1; /* Default to UP */
5761 if (smp_cpus > machine->max_cpus) {
5762 fprintf(stderr, "Number of SMP cpus requested (%d), exceeds max cpus "
5763 "supported by machine `%s' (%d)\n", smp_cpus, machine->name,
5764 machine->max_cpus);
5765 exit(1);
5768 if (display_type == DT_NOGRAPHIC) {
5769 if (serial_device_index == 0)
5770 serial_devices[0] = "stdio";
5771 if (parallel_device_index == 0)
5772 parallel_devices[0] = "null";
5773 if (strncmp(monitor_device, "vc", 2) == 0)
5774 monitor_device = "stdio";
5777 #ifndef _WIN32
5778 if (daemonize) {
5779 pid_t pid;
5781 if (pipe(fds) == -1)
5782 exit(1);
5784 pid = fork();
5785 if (pid > 0) {
5786 uint8_t status;
5787 ssize_t len;
5789 close(fds[1]);
5791 again:
5792 len = read(fds[0], &status, 1);
5793 if (len == -1 && (errno == EINTR))
5794 goto again;
5796 if (len != 1)
5797 exit(1);
5798 else if (status == 1) {
5799 fprintf(stderr, "Could not acquire pidfile\n");
5800 exit(1);
5801 } else
5802 exit(0);
5803 } else if (pid < 0)
5804 exit(1);
5806 setsid();
5808 pid = fork();
5809 if (pid > 0)
5810 exit(0);
5811 else if (pid < 0)
5812 exit(1);
5814 umask(027);
5816 signal(SIGTSTP, SIG_IGN);
5817 signal(SIGTTOU, SIG_IGN);
5818 signal(SIGTTIN, SIG_IGN);
5821 if (pid_file && qemu_create_pidfile(pid_file) != 0) {
5822 if (daemonize) {
5823 uint8_t status = 1;
5824 write(fds[1], &status, 1);
5825 } else
5826 fprintf(stderr, "Could not acquire pid file\n");
5827 exit(1);
5829 #endif
5831 #ifdef CONFIG_KQEMU
5832 if (smp_cpus > 1)
5833 kqemu_allowed = 0;
5834 #endif
5835 if (qemu_init_main_loop()) {
5836 fprintf(stderr, "qemu_init_main_loop failed\n");
5837 exit(1);
5839 linux_boot = (kernel_filename != NULL);
5841 if (!linux_boot && *kernel_cmdline != '\0') {
5842 fprintf(stderr, "-append only allowed with -kernel option\n");
5843 exit(1);
5846 if (!linux_boot && initrd_filename != NULL) {
5847 fprintf(stderr, "-initrd only allowed with -kernel option\n");
5848 exit(1);
5851 /* boot to floppy or the default cd if no hard disk defined yet */
5852 if (!boot_devices[0]) {
5853 boot_devices = "cad";
5855 setvbuf(stdout, NULL, _IOLBF, 0);
5857 init_timers();
5858 if (init_timer_alarm() < 0) {
5859 fprintf(stderr, "could not initialize alarm timer\n");
5860 exit(1);
5862 if (use_icount && icount_time_shift < 0) {
5863 use_icount = 2;
5864 /* 125MIPS seems a reasonable initial guess at the guest speed.
5865 It will be corrected fairly quickly anyway. */
5866 icount_time_shift = 3;
5867 init_icount_adjust();
5870 #ifdef _WIN32
5871 socket_init();
5872 #endif
5874 /* init network clients */
5875 if (nb_net_clients == 0) {
5876 /* if no clients, we use a default config */
5877 net_clients[nb_net_clients++] = "nic";
5878 #ifdef CONFIG_SLIRP
5879 net_clients[nb_net_clients++] = "user";
5880 #endif
5883 for(i = 0;i < nb_net_clients; i++) {
5884 if (net_client_parse(net_clients[i]) < 0)
5885 exit(1);
5888 net_boot = (boot_devices_bitmap >> ('n' - 'a')) & 0xF;
5889 net_set_boot_mask(net_boot);
5891 net_client_check();
5893 /* init the bluetooth world */
5894 for (i = 0; i < nb_bt_opts; i++)
5895 if (bt_parse(bt_opts[i]))
5896 exit(1);
5898 /* init the memory */
5899 if (ram_size == 0)
5900 ram_size = DEFAULT_RAM_SIZE * 1024 * 1024;
5902 #ifdef CONFIG_KQEMU
5903 /* FIXME: This is a nasty hack because kqemu can't cope with dynamic
5904 guest ram allocation. It needs to go away. */
5905 if (kqemu_allowed) {
5906 kqemu_phys_ram_size = ram_size + 8 * 1024 * 1024 + 4 * 1024 * 1024;
5907 kqemu_phys_ram_base = qemu_vmalloc(kqemu_phys_ram_size);
5908 if (!kqemu_phys_ram_base) {
5909 fprintf(stderr, "Could not allocate physical memory\n");
5910 exit(1);
5913 #endif
5915 /* init the dynamic translator */
5916 cpu_exec_init_all(tb_size * 1024 * 1024);
5918 bdrv_init();
5920 /* we always create the cdrom drive, even if no disk is there */
5922 if (nb_drives_opt < MAX_DRIVES)
5923 drive_add(NULL, CDROM_ALIAS);
5925 /* we always create at least one floppy */
5927 if (nb_drives_opt < MAX_DRIVES)
5928 drive_add(NULL, FD_ALIAS, 0);
5930 /* we always create one sd slot, even if no card is in it */
5932 if (nb_drives_opt < MAX_DRIVES)
5933 drive_add(NULL, SD_ALIAS);
5935 /* open the virtual block devices */
5937 for(i = 0; i < nb_drives_opt; i++)
5938 if (drive_init(&drives_opt[i], snapshot, machine) == -1)
5939 exit(1);
5941 register_savevm("timer", 0, 2, timer_save, timer_load, NULL);
5942 register_savevm_live("ram", 0, 3, ram_save_live, NULL, ram_load, NULL);
5944 #ifndef _WIN32
5945 /* must be after terminal init, SDL library changes signal handlers */
5946 sighandler_setup();
5947 #endif
5949 /* Maintain compatibility with multiple stdio monitors */
5950 if (!strcmp(monitor_device,"stdio")) {
5951 for (i = 0; i < MAX_SERIAL_PORTS; i++) {
5952 const char *devname = serial_devices[i];
5953 if (devname && !strcmp(devname,"mon:stdio")) {
5954 monitor_device = NULL;
5955 break;
5956 } else if (devname && !strcmp(devname,"stdio")) {
5957 monitor_device = NULL;
5958 serial_devices[i] = "mon:stdio";
5959 break;
5964 if (nb_numa_nodes > 0) {
5965 int i;
5967 if (nb_numa_nodes > smp_cpus) {
5968 nb_numa_nodes = smp_cpus;
5971 /* If no memory size if given for any node, assume the default case
5972 * and distribute the available memory equally across all nodes
5974 for (i = 0; i < nb_numa_nodes; i++) {
5975 if (node_mem[i] != 0)
5976 break;
5978 if (i == nb_numa_nodes) {
5979 uint64_t usedmem = 0;
5981 /* On Linux, the each node's border has to be 8MB aligned,
5982 * the final node gets the rest.
5984 for (i = 0; i < nb_numa_nodes - 1; i++) {
5985 node_mem[i] = (ram_size / nb_numa_nodes) & ~((1 << 23UL) - 1);
5986 usedmem += node_mem[i];
5988 node_mem[i] = ram_size - usedmem;
5991 for (i = 0; i < nb_numa_nodes; i++) {
5992 if (node_cpumask[i] != 0)
5993 break;
5995 /* assigning the VCPUs round-robin is easier to implement, guest OSes
5996 * must cope with this anyway, because there are BIOSes out there in
5997 * real machines which also use this scheme.
5999 if (i == nb_numa_nodes) {
6000 for (i = 0; i < smp_cpus; i++) {
6001 node_cpumask[i % nb_numa_nodes] |= 1 << i;
6006 if (kvm_enabled()) {
6007 int ret;
6009 ret = kvm_init(smp_cpus);
6010 if (ret < 0) {
6011 fprintf(stderr, "failed to initialize KVM\n");
6012 exit(1);
6016 if (monitor_device) {
6017 monitor_hd = qemu_chr_open("monitor", monitor_device, NULL);
6018 if (!monitor_hd) {
6019 fprintf(stderr, "qemu: could not open monitor device '%s'\n", monitor_device);
6020 exit(1);
6024 for(i = 0; i < MAX_SERIAL_PORTS; i++) {
6025 const char *devname = serial_devices[i];
6026 if (devname && strcmp(devname, "none")) {
6027 char label[32];
6028 snprintf(label, sizeof(label), "serial%d", i);
6029 serial_hds[i] = qemu_chr_open(label, devname, NULL);
6030 if (!serial_hds[i]) {
6031 fprintf(stderr, "qemu: could not open serial device '%s'\n",
6032 devname);
6033 exit(1);
6038 for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
6039 const char *devname = parallel_devices[i];
6040 if (devname && strcmp(devname, "none")) {
6041 char label[32];
6042 snprintf(label, sizeof(label), "parallel%d", i);
6043 parallel_hds[i] = qemu_chr_open(label, devname, NULL);
6044 if (!parallel_hds[i]) {
6045 fprintf(stderr, "qemu: could not open parallel device '%s'\n",
6046 devname);
6047 exit(1);
6052 for(i = 0; i < MAX_VIRTIO_CONSOLES; i++) {
6053 const char *devname = virtio_consoles[i];
6054 if (devname && strcmp(devname, "none")) {
6055 char label[32];
6056 snprintf(label, sizeof(label), "virtcon%d", i);
6057 virtcon_hds[i] = qemu_chr_open(label, devname, NULL);
6058 if (!virtcon_hds[i]) {
6059 fprintf(stderr, "qemu: could not open virtio console '%s'\n",
6060 devname);
6061 exit(1);
6066 module_call_init(MODULE_INIT_DEVICE);
6068 machine->init(ram_size, boot_devices,
6069 kernel_filename, kernel_cmdline, initrd_filename, cpu_model);
6072 for (env = first_cpu; env != NULL; env = env->next_cpu) {
6073 for (i = 0; i < nb_numa_nodes; i++) {
6074 if (node_cpumask[i] & (1 << env->cpu_index)) {
6075 env->numa_node = i;
6080 current_machine = machine;
6082 /* init USB devices */
6083 if (usb_enabled) {
6084 for(i = 0; i < usb_devices_index; i++) {
6085 if (usb_device_add(usb_devices[i], 0) < 0) {
6086 fprintf(stderr, "Warning: could not add USB device %s\n",
6087 usb_devices[i]);
6092 if (!display_state)
6093 dumb_display_init();
6094 /* just use the first displaystate for the moment */
6095 ds = display_state;
6097 if (display_type == DT_DEFAULT) {
6098 #if defined(CONFIG_SDL) || defined(CONFIG_COCOA)
6099 display_type = DT_SDL;
6100 #else
6101 display_type = DT_VNC;
6102 vnc_display = "localhost:0,to=99";
6103 show_vnc_port = 1;
6104 #endif
6108 switch (display_type) {
6109 case DT_NOGRAPHIC:
6110 break;
6111 #if defined(CONFIG_CURSES)
6112 case DT_CURSES:
6113 curses_display_init(ds, full_screen);
6114 break;
6115 #endif
6116 #if defined(CONFIG_SDL)
6117 case DT_SDL:
6118 sdl_display_init(ds, full_screen, no_frame);
6119 break;
6120 #elif defined(CONFIG_COCOA)
6121 case DT_SDL:
6122 cocoa_display_init(ds, full_screen);
6123 break;
6124 #endif
6125 case DT_VNC:
6126 vnc_display_init(ds);
6127 if (vnc_display_open(ds, vnc_display) < 0)
6128 exit(1);
6130 if (show_vnc_port) {
6131 printf("VNC server running on `%s'\n", vnc_display_local_addr(ds));
6133 break;
6134 default:
6135 break;
6137 dpy_resize(ds);
6139 dcl = ds->listeners;
6140 while (dcl != NULL) {
6141 if (dcl->dpy_refresh != NULL) {
6142 ds->gui_timer = qemu_new_timer(rt_clock, gui_update, ds);
6143 qemu_mod_timer(ds->gui_timer, qemu_get_clock(rt_clock));
6145 dcl = dcl->next;
6148 if (display_type == DT_NOGRAPHIC || display_type == DT_VNC) {
6149 nographic_timer = qemu_new_timer(rt_clock, nographic_update, NULL);
6150 qemu_mod_timer(nographic_timer, qemu_get_clock(rt_clock));
6153 text_consoles_set_display(display_state);
6154 qemu_chr_initial_reset();
6156 if (monitor_device && monitor_hd)
6157 monitor_init(monitor_hd, MONITOR_USE_READLINE | MONITOR_IS_DEFAULT);
6159 for(i = 0; i < MAX_SERIAL_PORTS; i++) {
6160 const char *devname = serial_devices[i];
6161 if (devname && strcmp(devname, "none")) {
6162 if (strstart(devname, "vc", 0))
6163 qemu_chr_printf(serial_hds[i], "serial%d console\r\n", i);
6167 for(i = 0; i < MAX_PARALLEL_PORTS; i++) {
6168 const char *devname = parallel_devices[i];
6169 if (devname && strcmp(devname, "none")) {
6170 if (strstart(devname, "vc", 0))
6171 qemu_chr_printf(parallel_hds[i], "parallel%d console\r\n", i);
6175 for(i = 0; i < MAX_VIRTIO_CONSOLES; i++) {
6176 const char *devname = virtio_consoles[i];
6177 if (virtcon_hds[i] && devname) {
6178 if (strstart(devname, "vc", 0))
6179 qemu_chr_printf(virtcon_hds[i], "virtio console%d\r\n", i);
6183 if (gdbstub_dev && gdbserver_start(gdbstub_dev) < 0) {
6184 fprintf(stderr, "qemu: could not open gdbserver on device '%s'\n",
6185 gdbstub_dev);
6186 exit(1);
6189 if (loadvm)
6190 do_loadvm(cur_mon, loadvm);
6192 if (incoming) {
6193 autostart = 0; /* fixme how to deal with -daemonize */
6194 qemu_start_incoming_migration(incoming);
6197 if (autostart)
6198 vm_start();
6200 #ifndef _WIN32
6201 if (daemonize) {
6202 uint8_t status = 0;
6203 ssize_t len;
6205 again1:
6206 len = write(fds[1], &status, 1);
6207 if (len == -1 && (errno == EINTR))
6208 goto again1;
6210 if (len != 1)
6211 exit(1);
6213 chdir("/");
6214 TFR(fd = open("/dev/null", O_RDWR));
6215 if (fd == -1)
6216 exit(1);
6219 if (run_as) {
6220 pwd = getpwnam(run_as);
6221 if (!pwd) {
6222 fprintf(stderr, "User \"%s\" doesn't exist\n", run_as);
6223 exit(1);
6227 if (chroot_dir) {
6228 if (chroot(chroot_dir) < 0) {
6229 fprintf(stderr, "chroot failed\n");
6230 exit(1);
6232 chdir("/");
6235 if (run_as) {
6236 if (setgid(pwd->pw_gid) < 0) {
6237 fprintf(stderr, "Failed to setgid(%d)\n", pwd->pw_gid);
6238 exit(1);
6240 if (setuid(pwd->pw_uid) < 0) {
6241 fprintf(stderr, "Failed to setuid(%d)\n", pwd->pw_uid);
6242 exit(1);
6244 if (setuid(0) != -1) {
6245 fprintf(stderr, "Dropping privileges failed\n");
6246 exit(1);
6250 if (daemonize) {
6251 dup2(fd, 0);
6252 dup2(fd, 1);
6253 dup2(fd, 2);
6255 close(fd);
6257 #endif
6259 main_loop();
6260 quit_timers();
6261 net_cleanup();
6263 return 0;