spapr_pci: spapr_iommu: Make DMA window a subregion
[qemu/ar7.git] / qga / commands-posix.c
blob34ddba053144231dd0543f47a94090a8087d7da6
1 /*
2 * QEMU Guest Agent POSIX-specific command implementations
4 * Copyright IBM Corp. 2011
6 * Authors:
7 * Michael Roth <mdroth@linux.vnet.ibm.com>
8 * Michal Privoznik <mprivozn@redhat.com>
10 * This work is licensed under the terms of the GNU GPL, version 2 or later.
11 * See the COPYING file in the top-level directory.
14 #include <glib.h>
15 #include <sys/types.h>
16 #include <sys/ioctl.h>
17 #include <sys/wait.h>
18 #include <unistd.h>
19 #include <errno.h>
20 #include <fcntl.h>
21 #include <stdio.h>
22 #include <string.h>
23 #include <sys/stat.h>
24 #include <inttypes.h>
25 #include "qga/guest-agent-core.h"
26 #include "qga-qmp-commands.h"
27 #include "qapi/qmp/qerror.h"
28 #include "qemu/queue.h"
29 #include "qemu/host-utils.h"
31 #ifndef CONFIG_HAS_ENVIRON
32 #ifdef __APPLE__
33 #include <crt_externs.h>
34 #define environ (*_NSGetEnviron())
35 #else
36 extern char **environ;
37 #endif
38 #endif
40 #if defined(__linux__)
41 #include <mntent.h>
42 #include <linux/fs.h>
43 #include <ifaddrs.h>
44 #include <arpa/inet.h>
45 #include <sys/socket.h>
46 #include <net/if.h>
48 #ifdef FIFREEZE
49 #define CONFIG_FSFREEZE
50 #endif
51 #ifdef FITRIM
52 #define CONFIG_FSTRIM
53 #endif
54 #endif
56 static void ga_wait_child(pid_t pid, int *status, Error **errp)
58 pid_t rpid;
60 *status = 0;
62 do {
63 rpid = waitpid(pid, status, 0);
64 } while (rpid == -1 && errno == EINTR);
66 if (rpid == -1) {
67 error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
68 pid);
69 return;
72 g_assert(rpid == pid);
75 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
77 const char *shutdown_flag;
78 Error *local_err = NULL;
79 pid_t pid;
80 int status;
82 slog("guest-shutdown called, mode: %s", mode);
83 if (!has_mode || strcmp(mode, "powerdown") == 0) {
84 shutdown_flag = "-P";
85 } else if (strcmp(mode, "halt") == 0) {
86 shutdown_flag = "-H";
87 } else if (strcmp(mode, "reboot") == 0) {
88 shutdown_flag = "-r";
89 } else {
90 error_setg(errp,
91 "mode is invalid (valid values are: halt|powerdown|reboot");
92 return;
95 pid = fork();
96 if (pid == 0) {
97 /* child, start the shutdown */
98 setsid();
99 reopen_fd_to_null(0);
100 reopen_fd_to_null(1);
101 reopen_fd_to_null(2);
103 execle("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
104 "hypervisor initiated shutdown", (char*)NULL, environ);
105 _exit(EXIT_FAILURE);
106 } else if (pid < 0) {
107 error_setg_errno(errp, errno, "failed to create child process");
108 return;
111 ga_wait_child(pid, &status, &local_err);
112 if (local_err) {
113 error_propagate(errp, local_err);
114 return;
117 if (!WIFEXITED(status)) {
118 error_setg(errp, "child process has terminated abnormally");
119 return;
122 if (WEXITSTATUS(status)) {
123 error_setg(errp, "child process has failed to shutdown");
124 return;
127 /* succeeded */
130 int64_t qmp_guest_get_time(Error **errp)
132 int ret;
133 qemu_timeval tq;
134 int64_t time_ns;
136 ret = qemu_gettimeofday(&tq);
137 if (ret < 0) {
138 error_setg_errno(errp, errno, "Failed to get time");
139 return -1;
142 time_ns = tq.tv_sec * 1000000000LL + tq.tv_usec * 1000;
143 return time_ns;
146 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
148 int ret;
149 int status;
150 pid_t pid;
151 Error *local_err = NULL;
152 struct timeval tv;
154 /* If user has passed a time, validate and set it. */
155 if (has_time) {
156 /* year-2038 will overflow in case time_t is 32bit */
157 if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
158 error_setg(errp, "Time %" PRId64 " is too large", time_ns);
159 return;
162 tv.tv_sec = time_ns / 1000000000;
163 tv.tv_usec = (time_ns % 1000000000) / 1000;
165 ret = settimeofday(&tv, NULL);
166 if (ret < 0) {
167 error_setg_errno(errp, errno, "Failed to set time to guest");
168 return;
172 /* Now, if user has passed a time to set and the system time is set, we
173 * just need to synchronize the hardware clock. However, if no time was
174 * passed, user is requesting the opposite: set the system time from the
175 * hardware clock (RTC). */
176 pid = fork();
177 if (pid == 0) {
178 setsid();
179 reopen_fd_to_null(0);
180 reopen_fd_to_null(1);
181 reopen_fd_to_null(2);
183 /* Use '/sbin/hwclock -w' to set RTC from the system time,
184 * or '/sbin/hwclock -s' to set the system time from RTC. */
185 execle("/sbin/hwclock", "hwclock", has_time ? "-w" : "-s",
186 NULL, environ);
187 _exit(EXIT_FAILURE);
188 } else if (pid < 0) {
189 error_setg_errno(errp, errno, "failed to create child process");
190 return;
193 ga_wait_child(pid, &status, &local_err);
194 if (local_err) {
195 error_propagate(errp, local_err);
196 return;
199 if (!WIFEXITED(status)) {
200 error_setg(errp, "child process has terminated abnormally");
201 return;
204 if (WEXITSTATUS(status)) {
205 error_setg(errp, "hwclock failed to set hardware clock to system time");
206 return;
210 typedef struct GuestFileHandle {
211 uint64_t id;
212 FILE *fh;
213 QTAILQ_ENTRY(GuestFileHandle) next;
214 } GuestFileHandle;
216 static struct {
217 QTAILQ_HEAD(, GuestFileHandle) filehandles;
218 } guest_file_state;
220 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
222 GuestFileHandle *gfh;
223 int64_t handle;
225 handle = ga_get_fd_handle(ga_state, errp);
226 if (handle < 0) {
227 return -1;
230 gfh = g_malloc0(sizeof(GuestFileHandle));
231 gfh->id = handle;
232 gfh->fh = fh;
233 QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
235 return handle;
238 static GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
240 GuestFileHandle *gfh;
242 QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
244 if (gfh->id == id) {
245 return gfh;
249 error_setg(errp, "handle '%" PRId64 "' has not been found", id);
250 return NULL;
253 typedef const char * const ccpc;
255 #ifndef O_BINARY
256 #define O_BINARY 0
257 #endif
259 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
260 static const struct {
261 ccpc *forms;
262 int oflag_base;
263 } guest_file_open_modes[] = {
264 { (ccpc[]){ "r", NULL }, O_RDONLY },
265 { (ccpc[]){ "rb", NULL }, O_RDONLY | O_BINARY },
266 { (ccpc[]){ "w", NULL }, O_WRONLY | O_CREAT | O_TRUNC },
267 { (ccpc[]){ "wb", NULL }, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY },
268 { (ccpc[]){ "a", NULL }, O_WRONLY | O_CREAT | O_APPEND },
269 { (ccpc[]){ "ab", NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
270 { (ccpc[]){ "r+", NULL }, O_RDWR },
271 { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR | O_BINARY },
272 { (ccpc[]){ "w+", NULL }, O_RDWR | O_CREAT | O_TRUNC },
273 { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR | O_CREAT | O_TRUNC | O_BINARY },
274 { (ccpc[]){ "a+", NULL }, O_RDWR | O_CREAT | O_APPEND },
275 { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR | O_CREAT | O_APPEND | O_BINARY }
278 static int
279 find_open_flag(const char *mode_str, Error **errp)
281 unsigned mode;
283 for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
284 ccpc *form;
286 form = guest_file_open_modes[mode].forms;
287 while (*form != NULL && strcmp(*form, mode_str) != 0) {
288 ++form;
290 if (*form != NULL) {
291 break;
295 if (mode == ARRAY_SIZE(guest_file_open_modes)) {
296 error_setg(errp, "invalid file open mode '%s'", mode_str);
297 return -1;
299 return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
302 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
303 S_IRGRP | S_IWGRP | \
304 S_IROTH | S_IWOTH)
306 static FILE *
307 safe_open_or_create(const char *path, const char *mode, Error **errp)
309 Error *local_err = NULL;
310 int oflag;
312 oflag = find_open_flag(mode, &local_err);
313 if (local_err == NULL) {
314 int fd;
316 /* If the caller wants / allows creation of a new file, we implement it
317 * with a two step process: open() + (open() / fchmod()).
319 * First we insist on creating the file exclusively as a new file. If
320 * that succeeds, we're free to set any file-mode bits on it. (The
321 * motivation is that we want to set those file-mode bits independently
322 * of the current umask.)
324 * If the exclusive creation fails because the file already exists
325 * (EEXIST is not possible for any other reason), we just attempt to
326 * open the file, but in this case we won't be allowed to change the
327 * file-mode bits on the preexistent file.
329 * The pathname should never disappear between the two open()s in
330 * practice. If it happens, then someone very likely tried to race us.
331 * In this case just go ahead and report the ENOENT from the second
332 * open() to the caller.
334 * If the caller wants to open a preexistent file, then the first
335 * open() is decisive and its third argument is ignored, and the second
336 * open() and the fchmod() are never called.
338 fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
339 if (fd == -1 && errno == EEXIST) {
340 oflag &= ~(unsigned)O_CREAT;
341 fd = open(path, oflag);
344 if (fd == -1) {
345 error_setg_errno(&local_err, errno, "failed to open file '%s' "
346 "(mode: '%s')", path, mode);
347 } else {
348 qemu_set_cloexec(fd);
350 if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
351 error_setg_errno(&local_err, errno, "failed to set permission "
352 "0%03o on new file '%s' (mode: '%s')",
353 (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
354 } else {
355 FILE *f;
357 f = fdopen(fd, mode);
358 if (f == NULL) {
359 error_setg_errno(&local_err, errno, "failed to associate "
360 "stdio stream with file descriptor %d, "
361 "file '%s' (mode: '%s')", fd, path, mode);
362 } else {
363 return f;
367 close(fd);
368 if (oflag & O_CREAT) {
369 unlink(path);
374 error_propagate(errp, local_err);
375 return NULL;
378 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
379 Error **errp)
381 FILE *fh;
382 Error *local_err = NULL;
383 int fd;
384 int64_t ret = -1, handle;
386 if (!has_mode) {
387 mode = "r";
389 slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
390 fh = safe_open_or_create(path, mode, &local_err);
391 if (local_err != NULL) {
392 error_propagate(errp, local_err);
393 return -1;
396 /* set fd non-blocking to avoid common use cases (like reading from a
397 * named pipe) from hanging the agent
399 fd = fileno(fh);
400 ret = fcntl(fd, F_GETFL);
401 ret = fcntl(fd, F_SETFL, ret | O_NONBLOCK);
402 if (ret == -1) {
403 error_setg_errno(errp, errno, "failed to make file '%s' non-blocking",
404 path);
405 fclose(fh);
406 return -1;
409 handle = guest_file_handle_add(fh, errp);
410 if (handle < 0) {
411 fclose(fh);
412 return -1;
415 slog("guest-file-open, handle: %" PRId64, handle);
416 return handle;
419 void qmp_guest_file_close(int64_t handle, Error **errp)
421 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
422 int ret;
424 slog("guest-file-close called, handle: %" PRId64, handle);
425 if (!gfh) {
426 return;
429 ret = fclose(gfh->fh);
430 if (ret == EOF) {
431 error_setg_errno(errp, errno, "failed to close handle");
432 return;
435 QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
436 g_free(gfh);
439 struct GuestFileRead *qmp_guest_file_read(int64_t handle, bool has_count,
440 int64_t count, Error **errp)
442 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
443 GuestFileRead *read_data = NULL;
444 guchar *buf;
445 FILE *fh;
446 size_t read_count;
448 if (!gfh) {
449 return NULL;
452 if (!has_count) {
453 count = QGA_READ_COUNT_DEFAULT;
454 } else if (count < 0) {
455 error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
456 count);
457 return NULL;
460 fh = gfh->fh;
461 buf = g_malloc0(count+1);
462 read_count = fread(buf, 1, count, fh);
463 if (ferror(fh)) {
464 error_setg_errno(errp, errno, "failed to read file");
465 slog("guest-file-read failed, handle: %" PRId64, handle);
466 } else {
467 buf[read_count] = 0;
468 read_data = g_malloc0(sizeof(GuestFileRead));
469 read_data->count = read_count;
470 read_data->eof = feof(fh);
471 if (read_count) {
472 read_data->buf_b64 = g_base64_encode(buf, read_count);
475 g_free(buf);
476 clearerr(fh);
478 return read_data;
481 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
482 bool has_count, int64_t count,
483 Error **errp)
485 GuestFileWrite *write_data = NULL;
486 guchar *buf;
487 gsize buf_len;
488 int write_count;
489 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
490 FILE *fh;
492 if (!gfh) {
493 return NULL;
496 fh = gfh->fh;
497 buf = g_base64_decode(buf_b64, &buf_len);
499 if (!has_count) {
500 count = buf_len;
501 } else if (count < 0 || count > buf_len) {
502 error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
503 count);
504 g_free(buf);
505 return NULL;
508 write_count = fwrite(buf, 1, count, fh);
509 if (ferror(fh)) {
510 error_setg_errno(errp, errno, "failed to write to file");
511 slog("guest-file-write failed, handle: %" PRId64, handle);
512 } else {
513 write_data = g_malloc0(sizeof(GuestFileWrite));
514 write_data->count = write_count;
515 write_data->eof = feof(fh);
517 g_free(buf);
518 clearerr(fh);
520 return write_data;
523 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
524 int64_t whence, Error **errp)
526 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
527 GuestFileSeek *seek_data = NULL;
528 FILE *fh;
529 int ret;
531 if (!gfh) {
532 return NULL;
535 fh = gfh->fh;
536 ret = fseek(fh, offset, whence);
537 if (ret == -1) {
538 error_setg_errno(errp, errno, "failed to seek file");
539 } else {
540 seek_data = g_new0(GuestFileSeek, 1);
541 seek_data->position = ftell(fh);
542 seek_data->eof = feof(fh);
544 clearerr(fh);
546 return seek_data;
549 void qmp_guest_file_flush(int64_t handle, Error **errp)
551 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
552 FILE *fh;
553 int ret;
555 if (!gfh) {
556 return;
559 fh = gfh->fh;
560 ret = fflush(fh);
561 if (ret == EOF) {
562 error_setg_errno(errp, errno, "failed to flush file");
566 static void guest_file_init(void)
568 QTAILQ_INIT(&guest_file_state.filehandles);
571 /* linux-specific implementations. avoid this if at all possible. */
572 #if defined(__linux__)
574 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
575 typedef struct FsMount {
576 char *dirname;
577 char *devtype;
578 QTAILQ_ENTRY(FsMount) next;
579 } FsMount;
581 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
583 static void free_fs_mount_list(FsMountList *mounts)
585 FsMount *mount, *temp;
587 if (!mounts) {
588 return;
591 QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
592 QTAILQ_REMOVE(mounts, mount, next);
593 g_free(mount->dirname);
594 g_free(mount->devtype);
595 g_free(mount);
600 * Walk the mount table and build a list of local file systems
602 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
604 struct mntent *ment;
605 FsMount *mount;
606 char const *mtab = "/proc/self/mounts";
607 FILE *fp;
609 fp = setmntent(mtab, "r");
610 if (!fp) {
611 error_setg(errp, "failed to open mtab file: '%s'", mtab);
612 return;
615 while ((ment = getmntent(fp))) {
617 * An entry which device name doesn't start with a '/' is
618 * either a dummy file system or a network file system.
619 * Add special handling for smbfs and cifs as is done by
620 * coreutils as well.
622 if ((ment->mnt_fsname[0] != '/') ||
623 (strcmp(ment->mnt_type, "smbfs") == 0) ||
624 (strcmp(ment->mnt_type, "cifs") == 0)) {
625 continue;
628 mount = g_malloc0(sizeof(FsMount));
629 mount->dirname = g_strdup(ment->mnt_dir);
630 mount->devtype = g_strdup(ment->mnt_type);
632 QTAILQ_INSERT_TAIL(mounts, mount, next);
635 endmntent(fp);
637 #endif
639 #if defined(CONFIG_FSFREEZE)
641 typedef enum {
642 FSFREEZE_HOOK_THAW = 0,
643 FSFREEZE_HOOK_FREEZE,
644 } FsfreezeHookArg;
646 const char *fsfreeze_hook_arg_string[] = {
647 "thaw",
648 "freeze",
651 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
653 int status;
654 pid_t pid;
655 const char *hook;
656 const char *arg_str = fsfreeze_hook_arg_string[arg];
657 Error *local_err = NULL;
659 hook = ga_fsfreeze_hook(ga_state);
660 if (!hook) {
661 return;
663 if (access(hook, X_OK) != 0) {
664 error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
665 return;
668 slog("executing fsfreeze hook with arg '%s'", arg_str);
669 pid = fork();
670 if (pid == 0) {
671 setsid();
672 reopen_fd_to_null(0);
673 reopen_fd_to_null(1);
674 reopen_fd_to_null(2);
676 execle(hook, hook, arg_str, NULL, environ);
677 _exit(EXIT_FAILURE);
678 } else if (pid < 0) {
679 error_setg_errno(errp, errno, "failed to create child process");
680 return;
683 ga_wait_child(pid, &status, &local_err);
684 if (local_err) {
685 error_propagate(errp, local_err);
686 return;
689 if (!WIFEXITED(status)) {
690 error_setg(errp, "fsfreeze hook has terminated abnormally");
691 return;
694 status = WEXITSTATUS(status);
695 if (status) {
696 error_setg(errp, "fsfreeze hook has failed with status %d", status);
697 return;
702 * Return status of freeze/thaw
704 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
706 if (ga_is_frozen(ga_state)) {
707 return GUEST_FSFREEZE_STATUS_FROZEN;
710 return GUEST_FSFREEZE_STATUS_THAWED;
714 * Walk list of mounted file systems in the guest, and freeze the ones which
715 * are real local file systems.
717 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
719 int ret = 0, i = 0;
720 FsMountList mounts;
721 struct FsMount *mount;
722 Error *local_err = NULL;
723 int fd;
725 slog("guest-fsfreeze called");
727 execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
728 if (local_err) {
729 error_propagate(errp, local_err);
730 return -1;
733 QTAILQ_INIT(&mounts);
734 build_fs_mount_list(&mounts, &local_err);
735 if (local_err) {
736 error_propagate(errp, local_err);
737 return -1;
740 /* cannot risk guest agent blocking itself on a write in this state */
741 ga_set_frozen(ga_state);
743 QTAILQ_FOREACH_REVERSE(mount, &mounts, FsMountList, next) {
744 fd = qemu_open(mount->dirname, O_RDONLY);
745 if (fd == -1) {
746 error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
747 goto error;
750 /* we try to cull filesytems we know won't work in advance, but other
751 * filesytems may not implement fsfreeze for less obvious reasons.
752 * these will report EOPNOTSUPP. we simply ignore these when tallying
753 * the number of frozen filesystems.
755 * any other error means a failure to freeze a filesystem we
756 * expect to be freezable, so return an error in those cases
757 * and return system to thawed state.
759 ret = ioctl(fd, FIFREEZE);
760 if (ret == -1) {
761 if (errno != EOPNOTSUPP) {
762 error_setg_errno(errp, errno, "failed to freeze %s",
763 mount->dirname);
764 close(fd);
765 goto error;
767 } else {
768 i++;
770 close(fd);
773 free_fs_mount_list(&mounts);
774 return i;
776 error:
777 free_fs_mount_list(&mounts);
778 qmp_guest_fsfreeze_thaw(NULL);
779 return 0;
783 * Walk list of frozen file systems in the guest, and thaw them.
785 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
787 int ret;
788 FsMountList mounts;
789 FsMount *mount;
790 int fd, i = 0, logged;
791 Error *local_err = NULL;
793 QTAILQ_INIT(&mounts);
794 build_fs_mount_list(&mounts, &local_err);
795 if (local_err) {
796 error_propagate(errp, local_err);
797 return 0;
800 QTAILQ_FOREACH(mount, &mounts, next) {
801 logged = false;
802 fd = qemu_open(mount->dirname, O_RDONLY);
803 if (fd == -1) {
804 continue;
806 /* we have no way of knowing whether a filesystem was actually unfrozen
807 * as a result of a successful call to FITHAW, only that if an error
808 * was returned the filesystem was *not* unfrozen by that particular
809 * call.
811 * since multiple preceding FIFREEZEs require multiple calls to FITHAW
812 * to unfreeze, continuing issuing FITHAW until an error is returned,
813 * in which case either the filesystem is in an unfreezable state, or,
814 * more likely, it was thawed previously (and remains so afterward).
816 * also, since the most recent successful call is the one that did
817 * the actual unfreeze, we can use this to provide an accurate count
818 * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
819 * may * be useful for determining whether a filesystem was unfrozen
820 * during the freeze/thaw phase by a process other than qemu-ga.
822 do {
823 ret = ioctl(fd, FITHAW);
824 if (ret == 0 && !logged) {
825 i++;
826 logged = true;
828 } while (ret == 0);
829 close(fd);
832 ga_unset_frozen(ga_state);
833 free_fs_mount_list(&mounts);
835 execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
837 return i;
840 static void guest_fsfreeze_cleanup(void)
842 Error *err = NULL;
844 if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
845 qmp_guest_fsfreeze_thaw(&err);
846 if (err) {
847 slog("failed to clean up frozen filesystems: %s",
848 error_get_pretty(err));
849 error_free(err);
853 #endif /* CONFIG_FSFREEZE */
855 #if defined(CONFIG_FSTRIM)
857 * Walk list of mounted file systems in the guest, and trim them.
859 void qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
861 int ret = 0;
862 FsMountList mounts;
863 struct FsMount *mount;
864 int fd;
865 Error *local_err = NULL;
866 struct fstrim_range r = {
867 .start = 0,
868 .len = -1,
869 .minlen = has_minimum ? minimum : 0,
872 slog("guest-fstrim called");
874 QTAILQ_INIT(&mounts);
875 build_fs_mount_list(&mounts, &local_err);
876 if (local_err) {
877 error_propagate(errp, local_err);
878 return;
881 QTAILQ_FOREACH(mount, &mounts, next) {
882 fd = qemu_open(mount->dirname, O_RDONLY);
883 if (fd == -1) {
884 error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
885 goto error;
888 /* We try to cull filesytems we know won't work in advance, but other
889 * filesytems may not implement fstrim for less obvious reasons. These
890 * will report EOPNOTSUPP; we simply ignore these errors. Any other
891 * error means an unexpected error, so return it in those cases. In
892 * some other cases ENOTTY will be reported (e.g. CD-ROMs).
894 ret = ioctl(fd, FITRIM, &r);
895 if (ret == -1) {
896 if (errno != ENOTTY && errno != EOPNOTSUPP) {
897 error_setg_errno(errp, errno, "failed to trim %s",
898 mount->dirname);
899 close(fd);
900 goto error;
903 close(fd);
906 error:
907 free_fs_mount_list(&mounts);
909 #endif /* CONFIG_FSTRIM */
912 #define LINUX_SYS_STATE_FILE "/sys/power/state"
913 #define SUSPEND_SUPPORTED 0
914 #define SUSPEND_NOT_SUPPORTED 1
916 static void bios_supports_mode(const char *pmutils_bin, const char *pmutils_arg,
917 const char *sysfile_str, Error **errp)
919 Error *local_err = NULL;
920 char *pmutils_path;
921 pid_t pid;
922 int status;
924 pmutils_path = g_find_program_in_path(pmutils_bin);
926 pid = fork();
927 if (!pid) {
928 char buf[32]; /* hopefully big enough */
929 ssize_t ret;
930 int fd;
932 setsid();
933 reopen_fd_to_null(0);
934 reopen_fd_to_null(1);
935 reopen_fd_to_null(2);
937 if (pmutils_path) {
938 execle(pmutils_path, pmutils_bin, pmutils_arg, NULL, environ);
942 * If we get here either pm-utils is not installed or execle() has
943 * failed. Let's try the manual method if the caller wants it.
946 if (!sysfile_str) {
947 _exit(SUSPEND_NOT_SUPPORTED);
950 fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
951 if (fd < 0) {
952 _exit(SUSPEND_NOT_SUPPORTED);
955 ret = read(fd, buf, sizeof(buf)-1);
956 if (ret <= 0) {
957 _exit(SUSPEND_NOT_SUPPORTED);
959 buf[ret] = '\0';
961 if (strstr(buf, sysfile_str)) {
962 _exit(SUSPEND_SUPPORTED);
965 _exit(SUSPEND_NOT_SUPPORTED);
966 } else if (pid < 0) {
967 error_setg_errno(errp, errno, "failed to create child process");
968 goto out;
971 ga_wait_child(pid, &status, &local_err);
972 if (local_err) {
973 error_propagate(errp, local_err);
974 goto out;
977 if (!WIFEXITED(status)) {
978 error_setg(errp, "child process has terminated abnormally");
979 goto out;
982 switch (WEXITSTATUS(status)) {
983 case SUSPEND_SUPPORTED:
984 goto out;
985 case SUSPEND_NOT_SUPPORTED:
986 error_setg(errp,
987 "the requested suspend mode is not supported by the guest");
988 goto out;
989 default:
990 error_setg(errp,
991 "the helper program '%s' returned an unexpected exit status"
992 " code (%d)", pmutils_path, WEXITSTATUS(status));
993 goto out;
996 out:
997 g_free(pmutils_path);
1000 static void guest_suspend(const char *pmutils_bin, const char *sysfile_str,
1001 Error **errp)
1003 Error *local_err = NULL;
1004 char *pmutils_path;
1005 pid_t pid;
1006 int status;
1008 pmutils_path = g_find_program_in_path(pmutils_bin);
1010 pid = fork();
1011 if (pid == 0) {
1012 /* child */
1013 int fd;
1015 setsid();
1016 reopen_fd_to_null(0);
1017 reopen_fd_to_null(1);
1018 reopen_fd_to_null(2);
1020 if (pmutils_path) {
1021 execle(pmutils_path, pmutils_bin, NULL, environ);
1025 * If we get here either pm-utils is not installed or execle() has
1026 * failed. Let's try the manual method if the caller wants it.
1029 if (!sysfile_str) {
1030 _exit(EXIT_FAILURE);
1033 fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1034 if (fd < 0) {
1035 _exit(EXIT_FAILURE);
1038 if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1039 _exit(EXIT_FAILURE);
1042 _exit(EXIT_SUCCESS);
1043 } else if (pid < 0) {
1044 error_setg_errno(errp, errno, "failed to create child process");
1045 goto out;
1048 ga_wait_child(pid, &status, &local_err);
1049 if (local_err) {
1050 error_propagate(errp, local_err);
1051 goto out;
1054 if (!WIFEXITED(status)) {
1055 error_setg(errp, "child process has terminated abnormally");
1056 goto out;
1059 if (WEXITSTATUS(status)) {
1060 error_setg(errp, "child process has failed to suspend");
1061 goto out;
1064 out:
1065 g_free(pmutils_path);
1068 void qmp_guest_suspend_disk(Error **errp)
1070 Error *local_err = NULL;
1072 bios_supports_mode("pm-is-supported", "--hibernate", "disk", &local_err);
1073 if (local_err) {
1074 error_propagate(errp, local_err);
1075 return;
1078 guest_suspend("pm-hibernate", "disk", errp);
1081 void qmp_guest_suspend_ram(Error **errp)
1083 Error *local_err = NULL;
1085 bios_supports_mode("pm-is-supported", "--suspend", "mem", &local_err);
1086 if (local_err) {
1087 error_propagate(errp, local_err);
1088 return;
1091 guest_suspend("pm-suspend", "mem", errp);
1094 void qmp_guest_suspend_hybrid(Error **errp)
1096 Error *local_err = NULL;
1098 bios_supports_mode("pm-is-supported", "--suspend-hybrid", NULL,
1099 &local_err);
1100 if (local_err) {
1101 error_propagate(errp, local_err);
1102 return;
1105 guest_suspend("pm-suspend-hybrid", NULL, errp);
1108 static GuestNetworkInterfaceList *
1109 guest_find_interface(GuestNetworkInterfaceList *head,
1110 const char *name)
1112 for (; head; head = head->next) {
1113 if (strcmp(head->value->name, name) == 0) {
1114 break;
1118 return head;
1122 * Build information about guest interfaces
1124 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1126 GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
1127 struct ifaddrs *ifap, *ifa;
1129 if (getifaddrs(&ifap) < 0) {
1130 error_setg_errno(errp, errno, "getifaddrs failed");
1131 goto error;
1134 for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
1135 GuestNetworkInterfaceList *info;
1136 GuestIpAddressList **address_list = NULL, *address_item = NULL;
1137 char addr4[INET_ADDRSTRLEN];
1138 char addr6[INET6_ADDRSTRLEN];
1139 int sock;
1140 struct ifreq ifr;
1141 unsigned char *mac_addr;
1142 void *p;
1144 g_debug("Processing %s interface", ifa->ifa_name);
1146 info = guest_find_interface(head, ifa->ifa_name);
1148 if (!info) {
1149 info = g_malloc0(sizeof(*info));
1150 info->value = g_malloc0(sizeof(*info->value));
1151 info->value->name = g_strdup(ifa->ifa_name);
1153 if (!cur_item) {
1154 head = cur_item = info;
1155 } else {
1156 cur_item->next = info;
1157 cur_item = info;
1161 if (!info->value->has_hardware_address &&
1162 ifa->ifa_flags & SIOCGIFHWADDR) {
1163 /* we haven't obtained HW address yet */
1164 sock = socket(PF_INET, SOCK_STREAM, 0);
1165 if (sock == -1) {
1166 error_setg_errno(errp, errno, "failed to create socket");
1167 goto error;
1170 memset(&ifr, 0, sizeof(ifr));
1171 pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
1172 if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
1173 error_setg_errno(errp, errno,
1174 "failed to get MAC address of %s",
1175 ifa->ifa_name);
1176 close(sock);
1177 goto error;
1180 close(sock);
1181 mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
1183 info->value->hardware_address =
1184 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
1185 (int) mac_addr[0], (int) mac_addr[1],
1186 (int) mac_addr[2], (int) mac_addr[3],
1187 (int) mac_addr[4], (int) mac_addr[5]);
1189 info->value->has_hardware_address = true;
1192 if (ifa->ifa_addr &&
1193 ifa->ifa_addr->sa_family == AF_INET) {
1194 /* interface with IPv4 address */
1195 p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1196 if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
1197 error_setg_errno(errp, errno, "inet_ntop failed");
1198 goto error;
1201 address_item = g_malloc0(sizeof(*address_item));
1202 address_item->value = g_malloc0(sizeof(*address_item->value));
1203 address_item->value->ip_address = g_strdup(addr4);
1204 address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
1206 if (ifa->ifa_netmask) {
1207 /* Count the number of set bits in netmask.
1208 * This is safe as '1' and '0' cannot be shuffled in netmask. */
1209 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
1210 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
1212 } else if (ifa->ifa_addr &&
1213 ifa->ifa_addr->sa_family == AF_INET6) {
1214 /* interface with IPv6 address */
1215 p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
1216 if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
1217 error_setg_errno(errp, errno, "inet_ntop failed");
1218 goto error;
1221 address_item = g_malloc0(sizeof(*address_item));
1222 address_item->value = g_malloc0(sizeof(*address_item->value));
1223 address_item->value->ip_address = g_strdup(addr6);
1224 address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
1226 if (ifa->ifa_netmask) {
1227 /* Count the number of set bits in netmask.
1228 * This is safe as '1' and '0' cannot be shuffled in netmask. */
1229 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
1230 address_item->value->prefix =
1231 ctpop32(((uint32_t *) p)[0]) +
1232 ctpop32(((uint32_t *) p)[1]) +
1233 ctpop32(((uint32_t *) p)[2]) +
1234 ctpop32(((uint32_t *) p)[3]);
1238 if (!address_item) {
1239 continue;
1242 address_list = &info->value->ip_addresses;
1244 while (*address_list && (*address_list)->next) {
1245 address_list = &(*address_list)->next;
1248 if (!*address_list) {
1249 *address_list = address_item;
1250 } else {
1251 (*address_list)->next = address_item;
1254 info->value->has_ip_addresses = true;
1259 freeifaddrs(ifap);
1260 return head;
1262 error:
1263 freeifaddrs(ifap);
1264 qapi_free_GuestNetworkInterfaceList(head);
1265 return NULL;
1268 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
1270 static long sysconf_exact(int name, const char *name_str, Error **errp)
1272 long ret;
1274 errno = 0;
1275 ret = sysconf(name);
1276 if (ret == -1) {
1277 if (errno == 0) {
1278 error_setg(errp, "sysconf(%s): value indefinite", name_str);
1279 } else {
1280 error_setg_errno(errp, errno, "sysconf(%s)", name_str);
1283 return ret;
1286 /* Transfer online/offline status between @vcpu and the guest system.
1288 * On input either @errp or *@errp must be NULL.
1290 * In system-to-@vcpu direction, the following @vcpu fields are accessed:
1291 * - R: vcpu->logical_id
1292 * - W: vcpu->online
1293 * - W: vcpu->can_offline
1295 * In @vcpu-to-system direction, the following @vcpu fields are accessed:
1296 * - R: vcpu->logical_id
1297 * - R: vcpu->online
1299 * Written members remain unmodified on error.
1301 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
1302 Error **errp)
1304 char *dirpath;
1305 int dirfd;
1307 dirpath = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
1308 vcpu->logical_id);
1309 dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
1310 if (dirfd == -1) {
1311 error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
1312 } else {
1313 static const char fn[] = "online";
1314 int fd;
1315 int res;
1317 fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
1318 if (fd == -1) {
1319 if (errno != ENOENT) {
1320 error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
1321 } else if (sys2vcpu) {
1322 vcpu->online = true;
1323 vcpu->can_offline = false;
1324 } else if (!vcpu->online) {
1325 error_setg(errp, "logical processor #%" PRId64 " can't be "
1326 "offlined", vcpu->logical_id);
1327 } /* otherwise pretend successful re-onlining */
1328 } else {
1329 unsigned char status;
1331 res = pread(fd, &status, 1, 0);
1332 if (res == -1) {
1333 error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
1334 } else if (res == 0) {
1335 error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
1336 fn);
1337 } else if (sys2vcpu) {
1338 vcpu->online = (status != '0');
1339 vcpu->can_offline = true;
1340 } else if (vcpu->online != (status != '0')) {
1341 status = '0' + vcpu->online;
1342 if (pwrite(fd, &status, 1, 0) == -1) {
1343 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
1344 fn);
1346 } /* otherwise pretend successful re-(on|off)-lining */
1348 res = close(fd);
1349 g_assert(res == 0);
1352 res = close(dirfd);
1353 g_assert(res == 0);
1356 g_free(dirpath);
1359 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
1361 int64_t current;
1362 GuestLogicalProcessorList *head, **link;
1363 long sc_max;
1364 Error *local_err = NULL;
1366 current = 0;
1367 head = NULL;
1368 link = &head;
1369 sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
1371 while (local_err == NULL && current < sc_max) {
1372 GuestLogicalProcessor *vcpu;
1373 GuestLogicalProcessorList *entry;
1375 vcpu = g_malloc0(sizeof *vcpu);
1376 vcpu->logical_id = current++;
1377 vcpu->has_can_offline = true; /* lolspeak ftw */
1378 transfer_vcpu(vcpu, true, &local_err);
1380 entry = g_malloc0(sizeof *entry);
1381 entry->value = vcpu;
1383 *link = entry;
1384 link = &entry->next;
1387 if (local_err == NULL) {
1388 /* there's no guest with zero VCPUs */
1389 g_assert(head != NULL);
1390 return head;
1393 qapi_free_GuestLogicalProcessorList(head);
1394 error_propagate(errp, local_err);
1395 return NULL;
1398 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
1400 int64_t processed;
1401 Error *local_err = NULL;
1403 processed = 0;
1404 while (vcpus != NULL) {
1405 transfer_vcpu(vcpus->value, false, &local_err);
1406 if (local_err != NULL) {
1407 break;
1409 ++processed;
1410 vcpus = vcpus->next;
1413 if (local_err != NULL) {
1414 if (processed == 0) {
1415 error_propagate(errp, local_err);
1416 } else {
1417 error_free(local_err);
1421 return processed;
1424 #else /* defined(__linux__) */
1426 void qmp_guest_suspend_disk(Error **errp)
1428 error_set(errp, QERR_UNSUPPORTED);
1431 void qmp_guest_suspend_ram(Error **errp)
1433 error_set(errp, QERR_UNSUPPORTED);
1436 void qmp_guest_suspend_hybrid(Error **errp)
1438 error_set(errp, QERR_UNSUPPORTED);
1441 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1443 error_set(errp, QERR_UNSUPPORTED);
1444 return NULL;
1447 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
1449 error_set(errp, QERR_UNSUPPORTED);
1450 return NULL;
1453 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
1455 error_set(errp, QERR_UNSUPPORTED);
1456 return -1;
1459 #endif
1461 #if !defined(CONFIG_FSFREEZE)
1463 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1465 error_set(errp, QERR_UNSUPPORTED);
1467 return 0;
1470 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1472 error_set(errp, QERR_UNSUPPORTED);
1474 return 0;
1477 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1479 error_set(errp, QERR_UNSUPPORTED);
1481 return 0;
1483 #endif /* CONFIG_FSFREEZE */
1485 #if !defined(CONFIG_FSTRIM)
1486 void qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1488 error_set(errp, QERR_UNSUPPORTED);
1490 #endif
1492 /* register init/cleanup routines for stateful command groups */
1493 void ga_command_state_init(GAState *s, GACommandState *cs)
1495 #if defined(CONFIG_FSFREEZE)
1496 ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
1497 #endif
1498 ga_command_state_add(cs, guest_file_init, NULL);