virtio: update MemoryRegionCaches when guest set bad features
[qemu/ar7.git] / qga / commands-posix.c
blob3bffee99d4c97307eb1b930380b32eae2980214d
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 "qemu/osdep.h"
15 #include <sys/ioctl.h>
16 #include <sys/utsname.h>
17 #include <sys/wait.h>
18 #include <dirent.h>
19 #include "qemu-common.h"
20 #include "guest-agent-core.h"
21 #include "qga-qapi-commands.h"
22 #include "qapi/error.h"
23 #include "qapi/qmp/qerror.h"
24 #include "qemu/queue.h"
25 #include "qemu/host-utils.h"
26 #include "qemu/sockets.h"
27 #include "qemu/base64.h"
28 #include "qemu/cutils.h"
29 #include "commands-common.h"
31 #ifdef HAVE_UTMPX
32 #include <utmpx.h>
33 #endif
35 #ifndef CONFIG_HAS_ENVIRON
36 #ifdef __APPLE__
37 #include <crt_externs.h>
38 #define environ (*_NSGetEnviron())
39 #else
40 extern char **environ;
41 #endif
42 #endif
44 #if defined(__linux__)
45 #include <mntent.h>
46 #include <linux/fs.h>
47 #include <ifaddrs.h>
48 #include <arpa/inet.h>
49 #include <sys/socket.h>
50 #include <net/if.h>
51 #include <sys/statvfs.h>
53 #ifdef CONFIG_LIBUDEV
54 #include <libudev.h>
55 #endif
57 #ifdef FIFREEZE
58 #define CONFIG_FSFREEZE
59 #endif
60 #ifdef FITRIM
61 #define CONFIG_FSTRIM
62 #endif
63 #endif
65 static void ga_wait_child(pid_t pid, int *status, Error **errp)
67 pid_t rpid;
69 *status = 0;
71 do {
72 rpid = waitpid(pid, status, 0);
73 } while (rpid == -1 && errno == EINTR);
75 if (rpid == -1) {
76 error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
77 pid);
78 return;
81 g_assert(rpid == pid);
84 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
86 const char *shutdown_flag;
87 Error *local_err = NULL;
88 pid_t pid;
89 int status;
91 slog("guest-shutdown called, mode: %s", mode);
92 if (!has_mode || strcmp(mode, "powerdown") == 0) {
93 shutdown_flag = "-P";
94 } else if (strcmp(mode, "halt") == 0) {
95 shutdown_flag = "-H";
96 } else if (strcmp(mode, "reboot") == 0) {
97 shutdown_flag = "-r";
98 } else {
99 error_setg(errp,
100 "mode is invalid (valid values are: halt|powerdown|reboot");
101 return;
104 pid = fork();
105 if (pid == 0) {
106 /* child, start the shutdown */
107 setsid();
108 reopen_fd_to_null(0);
109 reopen_fd_to_null(1);
110 reopen_fd_to_null(2);
112 execle("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
113 "hypervisor initiated shutdown", (char*)NULL, environ);
114 _exit(EXIT_FAILURE);
115 } else if (pid < 0) {
116 error_setg_errno(errp, errno, "failed to create child process");
117 return;
120 ga_wait_child(pid, &status, &local_err);
121 if (local_err) {
122 error_propagate(errp, local_err);
123 return;
126 if (!WIFEXITED(status)) {
127 error_setg(errp, "child process has terminated abnormally");
128 return;
131 if (WEXITSTATUS(status)) {
132 error_setg(errp, "child process has failed to shutdown");
133 return;
136 /* succeeded */
139 int64_t qmp_guest_get_time(Error **errp)
141 int ret;
142 qemu_timeval tq;
144 ret = qemu_gettimeofday(&tq);
145 if (ret < 0) {
146 error_setg_errno(errp, errno, "Failed to get time");
147 return -1;
150 return tq.tv_sec * 1000000000LL + tq.tv_usec * 1000;
153 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
155 int ret;
156 int status;
157 pid_t pid;
158 Error *local_err = NULL;
159 struct timeval tv;
160 static const char hwclock_path[] = "/sbin/hwclock";
161 static int hwclock_available = -1;
163 if (hwclock_available < 0) {
164 hwclock_available = (access(hwclock_path, X_OK) == 0);
167 if (!hwclock_available) {
168 error_setg(errp, QERR_UNSUPPORTED);
169 return;
172 /* If user has passed a time, validate and set it. */
173 if (has_time) {
174 GDate date = { 0, };
176 /* year-2038 will overflow in case time_t is 32bit */
177 if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
178 error_setg(errp, "Time %" PRId64 " is too large", time_ns);
179 return;
182 tv.tv_sec = time_ns / 1000000000;
183 tv.tv_usec = (time_ns % 1000000000) / 1000;
184 g_date_set_time_t(&date, tv.tv_sec);
185 if (date.year < 1970 || date.year >= 2070) {
186 error_setg_errno(errp, errno, "Invalid time");
187 return;
190 ret = settimeofday(&tv, NULL);
191 if (ret < 0) {
192 error_setg_errno(errp, errno, "Failed to set time to guest");
193 return;
197 /* Now, if user has passed a time to set and the system time is set, we
198 * just need to synchronize the hardware clock. However, if no time was
199 * passed, user is requesting the opposite: set the system time from the
200 * hardware clock (RTC). */
201 pid = fork();
202 if (pid == 0) {
203 setsid();
204 reopen_fd_to_null(0);
205 reopen_fd_to_null(1);
206 reopen_fd_to_null(2);
208 /* Use '/sbin/hwclock -w' to set RTC from the system time,
209 * or '/sbin/hwclock -s' to set the system time from RTC. */
210 execle(hwclock_path, "hwclock", has_time ? "-w" : "-s",
211 NULL, environ);
212 _exit(EXIT_FAILURE);
213 } else if (pid < 0) {
214 error_setg_errno(errp, errno, "failed to create child process");
215 return;
218 ga_wait_child(pid, &status, &local_err);
219 if (local_err) {
220 error_propagate(errp, local_err);
221 return;
224 if (!WIFEXITED(status)) {
225 error_setg(errp, "child process has terminated abnormally");
226 return;
229 if (WEXITSTATUS(status)) {
230 error_setg(errp, "hwclock failed to set hardware clock to system time");
231 return;
235 typedef enum {
236 RW_STATE_NEW,
237 RW_STATE_READING,
238 RW_STATE_WRITING,
239 } RwState;
241 struct GuestFileHandle {
242 uint64_t id;
243 FILE *fh;
244 RwState state;
245 QTAILQ_ENTRY(GuestFileHandle) next;
248 static struct {
249 QTAILQ_HEAD(, GuestFileHandle) filehandles;
250 } guest_file_state = {
251 .filehandles = QTAILQ_HEAD_INITIALIZER(guest_file_state.filehandles),
254 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
256 GuestFileHandle *gfh;
257 int64_t handle;
259 handle = ga_get_fd_handle(ga_state, errp);
260 if (handle < 0) {
261 return -1;
264 gfh = g_new0(GuestFileHandle, 1);
265 gfh->id = handle;
266 gfh->fh = fh;
267 QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
269 return handle;
272 GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
274 GuestFileHandle *gfh;
276 QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
278 if (gfh->id == id) {
279 return gfh;
283 error_setg(errp, "handle '%" PRId64 "' has not been found", id);
284 return NULL;
287 typedef const char * const ccpc;
289 #ifndef O_BINARY
290 #define O_BINARY 0
291 #endif
293 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
294 static const struct {
295 ccpc *forms;
296 int oflag_base;
297 } guest_file_open_modes[] = {
298 { (ccpc[]){ "r", NULL }, O_RDONLY },
299 { (ccpc[]){ "rb", NULL }, O_RDONLY | O_BINARY },
300 { (ccpc[]){ "w", NULL }, O_WRONLY | O_CREAT | O_TRUNC },
301 { (ccpc[]){ "wb", NULL }, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY },
302 { (ccpc[]){ "a", NULL }, O_WRONLY | O_CREAT | O_APPEND },
303 { (ccpc[]){ "ab", NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
304 { (ccpc[]){ "r+", NULL }, O_RDWR },
305 { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR | O_BINARY },
306 { (ccpc[]){ "w+", NULL }, O_RDWR | O_CREAT | O_TRUNC },
307 { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR | O_CREAT | O_TRUNC | O_BINARY },
308 { (ccpc[]){ "a+", NULL }, O_RDWR | O_CREAT | O_APPEND },
309 { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR | O_CREAT | O_APPEND | O_BINARY }
312 static int
313 find_open_flag(const char *mode_str, Error **errp)
315 unsigned mode;
317 for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
318 ccpc *form;
320 form = guest_file_open_modes[mode].forms;
321 while (*form != NULL && strcmp(*form, mode_str) != 0) {
322 ++form;
324 if (*form != NULL) {
325 break;
329 if (mode == ARRAY_SIZE(guest_file_open_modes)) {
330 error_setg(errp, "invalid file open mode '%s'", mode_str);
331 return -1;
333 return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
336 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
337 S_IRGRP | S_IWGRP | \
338 S_IROTH | S_IWOTH)
340 static FILE *
341 safe_open_or_create(const char *path, const char *mode, Error **errp)
343 Error *local_err = NULL;
344 int oflag;
346 oflag = find_open_flag(mode, &local_err);
347 if (local_err == NULL) {
348 int fd;
350 /* If the caller wants / allows creation of a new file, we implement it
351 * with a two step process: open() + (open() / fchmod()).
353 * First we insist on creating the file exclusively as a new file. If
354 * that succeeds, we're free to set any file-mode bits on it. (The
355 * motivation is that we want to set those file-mode bits independently
356 * of the current umask.)
358 * If the exclusive creation fails because the file already exists
359 * (EEXIST is not possible for any other reason), we just attempt to
360 * open the file, but in this case we won't be allowed to change the
361 * file-mode bits on the preexistent file.
363 * The pathname should never disappear between the two open()s in
364 * practice. If it happens, then someone very likely tried to race us.
365 * In this case just go ahead and report the ENOENT from the second
366 * open() to the caller.
368 * If the caller wants to open a preexistent file, then the first
369 * open() is decisive and its third argument is ignored, and the second
370 * open() and the fchmod() are never called.
372 fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
373 if (fd == -1 && errno == EEXIST) {
374 oflag &= ~(unsigned)O_CREAT;
375 fd = open(path, oflag);
378 if (fd == -1) {
379 error_setg_errno(&local_err, errno, "failed to open file '%s' "
380 "(mode: '%s')", path, mode);
381 } else {
382 qemu_set_cloexec(fd);
384 if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
385 error_setg_errno(&local_err, errno, "failed to set permission "
386 "0%03o on new file '%s' (mode: '%s')",
387 (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
388 } else {
389 FILE *f;
391 f = fdopen(fd, mode);
392 if (f == NULL) {
393 error_setg_errno(&local_err, errno, "failed to associate "
394 "stdio stream with file descriptor %d, "
395 "file '%s' (mode: '%s')", fd, path, mode);
396 } else {
397 return f;
401 close(fd);
402 if (oflag & O_CREAT) {
403 unlink(path);
408 error_propagate(errp, local_err);
409 return NULL;
412 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
413 Error **errp)
415 FILE *fh;
416 Error *local_err = NULL;
417 int64_t handle;
419 if (!has_mode) {
420 mode = "r";
422 slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
423 fh = safe_open_or_create(path, mode, &local_err);
424 if (local_err != NULL) {
425 error_propagate(errp, local_err);
426 return -1;
429 /* set fd non-blocking to avoid common use cases (like reading from a
430 * named pipe) from hanging the agent
432 qemu_set_nonblock(fileno(fh));
434 handle = guest_file_handle_add(fh, errp);
435 if (handle < 0) {
436 fclose(fh);
437 return -1;
440 slog("guest-file-open, handle: %" PRId64, handle);
441 return handle;
444 void qmp_guest_file_close(int64_t handle, Error **errp)
446 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
447 int ret;
449 slog("guest-file-close called, handle: %" PRId64, handle);
450 if (!gfh) {
451 return;
454 ret = fclose(gfh->fh);
455 if (ret == EOF) {
456 error_setg_errno(errp, errno, "failed to close handle");
457 return;
460 QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
461 g_free(gfh);
464 GuestFileRead *guest_file_read_unsafe(GuestFileHandle *gfh,
465 int64_t count, Error **errp)
467 GuestFileRead *read_data = NULL;
468 guchar *buf;
469 FILE *fh = gfh->fh;
470 size_t read_count;
472 /* explicitly flush when switching from writing to reading */
473 if (gfh->state == RW_STATE_WRITING) {
474 int ret = fflush(fh);
475 if (ret == EOF) {
476 error_setg_errno(errp, errno, "failed to flush file");
477 return NULL;
479 gfh->state = RW_STATE_NEW;
482 buf = g_malloc0(count+1);
483 read_count = fread(buf, 1, count, fh);
484 if (ferror(fh)) {
485 error_setg_errno(errp, errno, "failed to read file");
486 } else {
487 buf[read_count] = 0;
488 read_data = g_new0(GuestFileRead, 1);
489 read_data->count = read_count;
490 read_data->eof = feof(fh);
491 if (read_count) {
492 read_data->buf_b64 = g_base64_encode(buf, read_count);
494 gfh->state = RW_STATE_READING;
496 g_free(buf);
497 clearerr(fh);
499 return read_data;
502 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
503 bool has_count, int64_t count,
504 Error **errp)
506 GuestFileWrite *write_data = NULL;
507 guchar *buf;
508 gsize buf_len;
509 int write_count;
510 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
511 FILE *fh;
513 if (!gfh) {
514 return NULL;
517 fh = gfh->fh;
519 if (gfh->state == RW_STATE_READING) {
520 int ret = fseek(fh, 0, SEEK_CUR);
521 if (ret == -1) {
522 error_setg_errno(errp, errno, "failed to seek file");
523 return NULL;
525 gfh->state = RW_STATE_NEW;
528 buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
529 if (!buf) {
530 return NULL;
533 if (!has_count) {
534 count = buf_len;
535 } else if (count < 0 || count > buf_len) {
536 error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
537 count);
538 g_free(buf);
539 return NULL;
542 write_count = fwrite(buf, 1, count, fh);
543 if (ferror(fh)) {
544 error_setg_errno(errp, errno, "failed to write to file");
545 slog("guest-file-write failed, handle: %" PRId64, handle);
546 } else {
547 write_data = g_new0(GuestFileWrite, 1);
548 write_data->count = write_count;
549 write_data->eof = feof(fh);
550 gfh->state = RW_STATE_WRITING;
552 g_free(buf);
553 clearerr(fh);
555 return write_data;
558 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
559 GuestFileWhence *whence_code,
560 Error **errp)
562 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
563 GuestFileSeek *seek_data = NULL;
564 FILE *fh;
565 int ret;
566 int whence;
567 Error *err = NULL;
569 if (!gfh) {
570 return NULL;
573 /* We stupidly exposed 'whence':'int' in our qapi */
574 whence = ga_parse_whence(whence_code, &err);
575 if (err) {
576 error_propagate(errp, err);
577 return NULL;
580 fh = gfh->fh;
581 ret = fseek(fh, offset, whence);
582 if (ret == -1) {
583 error_setg_errno(errp, errno, "failed to seek file");
584 if (errno == ESPIPE) {
585 /* file is non-seekable, stdio shouldn't be buffering anyways */
586 gfh->state = RW_STATE_NEW;
588 } else {
589 seek_data = g_new0(GuestFileSeek, 1);
590 seek_data->position = ftell(fh);
591 seek_data->eof = feof(fh);
592 gfh->state = RW_STATE_NEW;
594 clearerr(fh);
596 return seek_data;
599 void qmp_guest_file_flush(int64_t handle, Error **errp)
601 GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
602 FILE *fh;
603 int ret;
605 if (!gfh) {
606 return;
609 fh = gfh->fh;
610 ret = fflush(fh);
611 if (ret == EOF) {
612 error_setg_errno(errp, errno, "failed to flush file");
613 } else {
614 gfh->state = RW_STATE_NEW;
618 /* linux-specific implementations. avoid this if at all possible. */
619 #if defined(__linux__)
621 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
622 typedef struct FsMount {
623 char *dirname;
624 char *devtype;
625 unsigned int devmajor, devminor;
626 QTAILQ_ENTRY(FsMount) next;
627 } FsMount;
629 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
631 static void free_fs_mount_list(FsMountList *mounts)
633 FsMount *mount, *temp;
635 if (!mounts) {
636 return;
639 QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
640 QTAILQ_REMOVE(mounts, mount, next);
641 g_free(mount->dirname);
642 g_free(mount->devtype);
643 g_free(mount);
647 static int dev_major_minor(const char *devpath,
648 unsigned int *devmajor, unsigned int *devminor)
650 struct stat st;
652 *devmajor = 0;
653 *devminor = 0;
655 if (stat(devpath, &st) < 0) {
656 slog("failed to stat device file '%s': %s", devpath, strerror(errno));
657 return -1;
659 if (S_ISDIR(st.st_mode)) {
660 /* It is bind mount */
661 return -2;
663 if (S_ISBLK(st.st_mode)) {
664 *devmajor = major(st.st_rdev);
665 *devminor = minor(st.st_rdev);
666 return 0;
668 return -1;
672 * Walk the mount table and build a list of local file systems
674 static void build_fs_mount_list_from_mtab(FsMountList *mounts, Error **errp)
676 struct mntent *ment;
677 FsMount *mount;
678 char const *mtab = "/proc/self/mounts";
679 FILE *fp;
680 unsigned int devmajor, devminor;
682 fp = setmntent(mtab, "r");
683 if (!fp) {
684 error_setg(errp, "failed to open mtab file: '%s'", mtab);
685 return;
688 while ((ment = getmntent(fp))) {
690 * An entry which device name doesn't start with a '/' is
691 * either a dummy file system or a network file system.
692 * Add special handling for smbfs and cifs as is done by
693 * coreutils as well.
695 if ((ment->mnt_fsname[0] != '/') ||
696 (strcmp(ment->mnt_type, "smbfs") == 0) ||
697 (strcmp(ment->mnt_type, "cifs") == 0)) {
698 continue;
700 if (dev_major_minor(ment->mnt_fsname, &devmajor, &devminor) == -2) {
701 /* Skip bind mounts */
702 continue;
705 mount = g_new0(FsMount, 1);
706 mount->dirname = g_strdup(ment->mnt_dir);
707 mount->devtype = g_strdup(ment->mnt_type);
708 mount->devmajor = devmajor;
709 mount->devminor = devminor;
711 QTAILQ_INSERT_TAIL(mounts, mount, next);
714 endmntent(fp);
717 static void decode_mntname(char *name, int len)
719 int i, j = 0;
720 for (i = 0; i <= len; i++) {
721 if (name[i] != '\\') {
722 name[j++] = name[i];
723 } else if (name[i + 1] == '\\') {
724 name[j++] = '\\';
725 i++;
726 } else if (name[i + 1] >= '0' && name[i + 1] <= '3' &&
727 name[i + 2] >= '0' && name[i + 2] <= '7' &&
728 name[i + 3] >= '0' && name[i + 3] <= '7') {
729 name[j++] = (name[i + 1] - '0') * 64 +
730 (name[i + 2] - '0') * 8 +
731 (name[i + 3] - '0');
732 i += 3;
733 } else {
734 name[j++] = name[i];
739 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
741 FsMount *mount;
742 char const *mountinfo = "/proc/self/mountinfo";
743 FILE *fp;
744 char *line = NULL, *dash;
745 size_t n;
746 char check;
747 unsigned int devmajor, devminor;
748 int ret, dir_s, dir_e, type_s, type_e, dev_s, dev_e;
750 fp = fopen(mountinfo, "r");
751 if (!fp) {
752 build_fs_mount_list_from_mtab(mounts, errp);
753 return;
756 while (getline(&line, &n, fp) != -1) {
757 ret = sscanf(line, "%*u %*u %u:%u %*s %n%*s%n%c",
758 &devmajor, &devminor, &dir_s, &dir_e, &check);
759 if (ret < 3) {
760 continue;
762 dash = strstr(line + dir_e, " - ");
763 if (!dash) {
764 continue;
766 ret = sscanf(dash, " - %n%*s%n %n%*s%n%c",
767 &type_s, &type_e, &dev_s, &dev_e, &check);
768 if (ret < 1) {
769 continue;
771 line[dir_e] = 0;
772 dash[type_e] = 0;
773 dash[dev_e] = 0;
774 decode_mntname(line + dir_s, dir_e - dir_s);
775 decode_mntname(dash + dev_s, dev_e - dev_s);
776 if (devmajor == 0) {
777 /* btrfs reports major number = 0 */
778 if (strcmp("btrfs", dash + type_s) != 0 ||
779 dev_major_minor(dash + dev_s, &devmajor, &devminor) < 0) {
780 continue;
784 mount = g_new0(FsMount, 1);
785 mount->dirname = g_strdup(line + dir_s);
786 mount->devtype = g_strdup(dash + type_s);
787 mount->devmajor = devmajor;
788 mount->devminor = devminor;
790 QTAILQ_INSERT_TAIL(mounts, mount, next);
792 free(line);
794 fclose(fp);
796 #endif
798 #if defined(CONFIG_FSFREEZE)
800 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
802 char *path;
803 char *dpath;
804 char *driver = NULL;
805 char buf[PATH_MAX];
806 ssize_t len;
808 path = g_strndup(syspath, pathlen);
809 dpath = g_strdup_printf("%s/driver", path);
810 len = readlink(dpath, buf, sizeof(buf) - 1);
811 if (len != -1) {
812 buf[len] = 0;
813 driver = g_path_get_basename(buf);
815 g_free(dpath);
816 g_free(path);
817 return driver;
820 static int compare_uint(const void *_a, const void *_b)
822 unsigned int a = *(unsigned int *)_a;
823 unsigned int b = *(unsigned int *)_b;
825 return a < b ? -1 : a > b ? 1 : 0;
828 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
829 static int build_hosts(char const *syspath, char const *host, bool ata,
830 unsigned int *hosts, int hosts_max, Error **errp)
832 char *path;
833 DIR *dir;
834 struct dirent *entry;
835 int i = 0;
837 path = g_strndup(syspath, host - syspath);
838 dir = opendir(path);
839 if (!dir) {
840 error_setg_errno(errp, errno, "opendir(\"%s\")", path);
841 g_free(path);
842 return -1;
845 while (i < hosts_max) {
846 entry = readdir(dir);
847 if (!entry) {
848 break;
850 if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
851 ++i;
852 } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
853 ++i;
857 qsort(hosts, i, sizeof(hosts[0]), compare_uint);
859 g_free(path);
860 closedir(dir);
861 return i;
865 * Store disk device info for devices on the PCI bus.
866 * Returns true if information has been stored, or false for failure.
868 static bool build_guest_fsinfo_for_pci_dev(char const *syspath,
869 GuestDiskAddress *disk,
870 Error **errp)
872 unsigned int pci[4], host, hosts[8], tgt[3];
873 int i, nhosts = 0, pcilen;
874 GuestPCIAddress *pciaddr = disk->pci_controller;
875 bool has_ata = false, has_host = false, has_tgt = false;
876 char *p, *q, *driver = NULL;
877 bool ret = false;
879 p = strstr(syspath, "/devices/pci");
880 if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
881 pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
882 g_debug("only pci device is supported: sysfs path '%s'", syspath);
883 return false;
886 p += 12 + pcilen;
887 while (true) {
888 driver = get_pci_driver(syspath, p - syspath, errp);
889 if (driver && (g_str_equal(driver, "ata_piix") ||
890 g_str_equal(driver, "sym53c8xx") ||
891 g_str_equal(driver, "virtio-pci") ||
892 g_str_equal(driver, "ahci"))) {
893 break;
896 g_free(driver);
897 if (sscanf(p, "/%x:%x:%x.%x%n",
898 pci, pci + 1, pci + 2, pci + 3, &pcilen) == 4) {
899 p += pcilen;
900 continue;
903 g_debug("unsupported driver or sysfs path '%s'", syspath);
904 return false;
907 p = strstr(syspath, "/target");
908 if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
909 tgt, tgt + 1, tgt + 2) == 3) {
910 has_tgt = true;
913 p = strstr(syspath, "/ata");
914 if (p) {
915 q = p + 4;
916 has_ata = true;
917 } else {
918 p = strstr(syspath, "/host");
919 q = p + 5;
921 if (p && sscanf(q, "%u", &host) == 1) {
922 has_host = true;
923 nhosts = build_hosts(syspath, p, has_ata, hosts,
924 ARRAY_SIZE(hosts), errp);
925 if (nhosts < 0) {
926 goto cleanup;
930 pciaddr->domain = pci[0];
931 pciaddr->bus = pci[1];
932 pciaddr->slot = pci[2];
933 pciaddr->function = pci[3];
935 if (strcmp(driver, "ata_piix") == 0) {
936 /* a host per ide bus, target*:0:<unit>:0 */
937 if (!has_host || !has_tgt) {
938 g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
939 goto cleanup;
941 for (i = 0; i < nhosts; i++) {
942 if (host == hosts[i]) {
943 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
944 disk->bus = i;
945 disk->unit = tgt[1];
946 break;
949 if (i >= nhosts) {
950 g_debug("no host for '%s' (driver '%s')", syspath, driver);
951 goto cleanup;
953 } else if (strcmp(driver, "sym53c8xx") == 0) {
954 /* scsi(LSI Logic): target*:0:<unit>:0 */
955 if (!has_tgt) {
956 g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
957 goto cleanup;
959 disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
960 disk->unit = tgt[1];
961 } else if (strcmp(driver, "virtio-pci") == 0) {
962 if (has_tgt) {
963 /* virtio-scsi: target*:0:0:<unit> */
964 disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
965 disk->unit = tgt[2];
966 } else {
967 /* virtio-blk: 1 disk per 1 device */
968 disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
970 } else if (strcmp(driver, "ahci") == 0) {
971 /* ahci: 1 host per 1 unit */
972 if (!has_host || !has_tgt) {
973 g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
974 goto cleanup;
976 for (i = 0; i < nhosts; i++) {
977 if (host == hosts[i]) {
978 disk->unit = i;
979 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
980 break;
983 if (i >= nhosts) {
984 g_debug("no host for '%s' (driver '%s')", syspath, driver);
985 goto cleanup;
987 } else {
988 g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
989 goto cleanup;
992 ret = true;
994 cleanup:
995 g_free(driver);
996 return ret;
1000 * Store disk device info for non-PCI virtio devices (for example s390x
1001 * channel I/O devices). Returns true if information has been stored, or
1002 * false for failure.
1004 static bool build_guest_fsinfo_for_nonpci_virtio(char const *syspath,
1005 GuestDiskAddress *disk,
1006 Error **errp)
1008 unsigned int tgt[3];
1009 char *p;
1011 if (!strstr(syspath, "/virtio") || !strstr(syspath, "/block")) {
1012 g_debug("Unsupported virtio device '%s'", syspath);
1013 return false;
1016 p = strstr(syspath, "/target");
1017 if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
1018 &tgt[0], &tgt[1], &tgt[2]) == 3) {
1019 /* virtio-scsi: target*:0:<target>:<unit> */
1020 disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
1021 disk->bus = tgt[0];
1022 disk->target = tgt[1];
1023 disk->unit = tgt[2];
1024 } else {
1025 /* virtio-blk: 1 disk per 1 device */
1026 disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
1029 return true;
1032 /* Store disk device info specified by @sysfs into @fs */
1033 static void build_guest_fsinfo_for_real_device(char const *syspath,
1034 GuestFilesystemInfo *fs,
1035 Error **errp)
1037 GuestDiskAddress *disk;
1038 GuestPCIAddress *pciaddr;
1039 GuestDiskAddressList *list = NULL;
1040 bool has_hwinf;
1041 #ifdef CONFIG_LIBUDEV
1042 struct udev *udev = NULL;
1043 struct udev_device *udevice = NULL;
1044 #endif
1046 pciaddr = g_new0(GuestPCIAddress, 1);
1047 pciaddr->domain = -1; /* -1 means field is invalid */
1048 pciaddr->bus = -1;
1049 pciaddr->slot = -1;
1050 pciaddr->function = -1;
1052 disk = g_new0(GuestDiskAddress, 1);
1053 disk->pci_controller = pciaddr;
1054 disk->bus_type = GUEST_DISK_BUS_TYPE_UNKNOWN;
1056 list = g_new0(GuestDiskAddressList, 1);
1057 list->value = disk;
1059 #ifdef CONFIG_LIBUDEV
1060 udev = udev_new();
1061 udevice = udev_device_new_from_syspath(udev, syspath);
1062 if (udev == NULL || udevice == NULL) {
1063 g_debug("failed to query udev");
1064 } else {
1065 const char *devnode, *serial;
1066 devnode = udev_device_get_devnode(udevice);
1067 if (devnode != NULL) {
1068 disk->dev = g_strdup(devnode);
1069 disk->has_dev = true;
1071 serial = udev_device_get_property_value(udevice, "ID_SERIAL");
1072 if (serial != NULL && *serial != 0) {
1073 disk->serial = g_strdup(serial);
1074 disk->has_serial = true;
1078 udev_unref(udev);
1079 udev_device_unref(udevice);
1080 #endif
1082 if (strstr(syspath, "/devices/pci")) {
1083 has_hwinf = build_guest_fsinfo_for_pci_dev(syspath, disk, errp);
1084 } else if (strstr(syspath, "/virtio")) {
1085 has_hwinf = build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1086 } else {
1087 g_debug("Unsupported device type for '%s'", syspath);
1088 has_hwinf = false;
1091 if (has_hwinf || disk->has_dev || disk->has_serial) {
1092 list->next = fs->disk;
1093 fs->disk = list;
1094 } else {
1095 qapi_free_GuestDiskAddressList(list);
1099 static void build_guest_fsinfo_for_device(char const *devpath,
1100 GuestFilesystemInfo *fs,
1101 Error **errp);
1103 /* Store a list of slave devices of virtual volume specified by @syspath into
1104 * @fs */
1105 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
1106 GuestFilesystemInfo *fs,
1107 Error **errp)
1109 Error *err = NULL;
1110 DIR *dir;
1111 char *dirpath;
1112 struct dirent *entry;
1114 dirpath = g_strdup_printf("%s/slaves", syspath);
1115 dir = opendir(dirpath);
1116 if (!dir) {
1117 if (errno != ENOENT) {
1118 error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1120 g_free(dirpath);
1121 return;
1124 for (;;) {
1125 errno = 0;
1126 entry = readdir(dir);
1127 if (entry == NULL) {
1128 if (errno) {
1129 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1131 break;
1134 if (entry->d_type == DT_LNK) {
1135 char *path;
1137 g_debug(" slave device '%s'", entry->d_name);
1138 path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1139 build_guest_fsinfo_for_device(path, fs, &err);
1140 g_free(path);
1142 if (err) {
1143 error_propagate(errp, err);
1144 break;
1149 g_free(dirpath);
1150 closedir(dir);
1153 /* Dispatch to functions for virtual/real device */
1154 static void build_guest_fsinfo_for_device(char const *devpath,
1155 GuestFilesystemInfo *fs,
1156 Error **errp)
1158 char *syspath = realpath(devpath, NULL);
1160 if (!syspath) {
1161 error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1162 return;
1165 if (!fs->name) {
1166 fs->name = g_path_get_basename(syspath);
1169 g_debug(" parse sysfs path '%s'", syspath);
1171 if (strstr(syspath, "/devices/virtual/block/")) {
1172 build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1173 } else {
1174 build_guest_fsinfo_for_real_device(syspath, fs, errp);
1177 free(syspath);
1180 /* Return a list of the disk device(s)' info which @mount lies on */
1181 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1182 Error **errp)
1184 GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1185 struct statvfs buf;
1186 unsigned long used, nonroot_total, fr_size;
1187 char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1188 mount->devmajor, mount->devminor);
1190 fs->mountpoint = g_strdup(mount->dirname);
1191 fs->type = g_strdup(mount->devtype);
1192 build_guest_fsinfo_for_device(devpath, fs, errp);
1194 if (statvfs(fs->mountpoint, &buf) == 0) {
1195 fr_size = buf.f_frsize;
1196 used = buf.f_blocks - buf.f_bfree;
1197 nonroot_total = used + buf.f_bavail;
1198 fs->used_bytes = used * fr_size;
1199 fs->total_bytes = nonroot_total * fr_size;
1201 fs->has_total_bytes = true;
1202 fs->has_used_bytes = true;
1205 g_free(devpath);
1207 return fs;
1210 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1212 FsMountList mounts;
1213 struct FsMount *mount;
1214 GuestFilesystemInfoList *new, *ret = NULL;
1215 Error *local_err = NULL;
1217 QTAILQ_INIT(&mounts);
1218 build_fs_mount_list(&mounts, &local_err);
1219 if (local_err) {
1220 error_propagate(errp, local_err);
1221 return NULL;
1224 QTAILQ_FOREACH(mount, &mounts, next) {
1225 g_debug("Building guest fsinfo for '%s'", mount->dirname);
1227 new = g_malloc0(sizeof(*ret));
1228 new->value = build_guest_fsinfo(mount, &local_err);
1229 new->next = ret;
1230 ret = new;
1231 if (local_err) {
1232 error_propagate(errp, local_err);
1233 qapi_free_GuestFilesystemInfoList(ret);
1234 ret = NULL;
1235 break;
1239 free_fs_mount_list(&mounts);
1240 return ret;
1244 typedef enum {
1245 FSFREEZE_HOOK_THAW = 0,
1246 FSFREEZE_HOOK_FREEZE,
1247 } FsfreezeHookArg;
1249 static const char *fsfreeze_hook_arg_string[] = {
1250 "thaw",
1251 "freeze",
1254 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1256 int status;
1257 pid_t pid;
1258 const char *hook;
1259 const char *arg_str = fsfreeze_hook_arg_string[arg];
1260 Error *local_err = NULL;
1262 hook = ga_fsfreeze_hook(ga_state);
1263 if (!hook) {
1264 return;
1266 if (access(hook, X_OK) != 0) {
1267 error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1268 return;
1271 slog("executing fsfreeze hook with arg '%s'", arg_str);
1272 pid = fork();
1273 if (pid == 0) {
1274 setsid();
1275 reopen_fd_to_null(0);
1276 reopen_fd_to_null(1);
1277 reopen_fd_to_null(2);
1279 execle(hook, hook, arg_str, NULL, environ);
1280 _exit(EXIT_FAILURE);
1281 } else if (pid < 0) {
1282 error_setg_errno(errp, errno, "failed to create child process");
1283 return;
1286 ga_wait_child(pid, &status, &local_err);
1287 if (local_err) {
1288 error_propagate(errp, local_err);
1289 return;
1292 if (!WIFEXITED(status)) {
1293 error_setg(errp, "fsfreeze hook has terminated abnormally");
1294 return;
1297 status = WEXITSTATUS(status);
1298 if (status) {
1299 error_setg(errp, "fsfreeze hook has failed with status %d", status);
1300 return;
1305 * Return status of freeze/thaw
1307 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1309 if (ga_is_frozen(ga_state)) {
1310 return GUEST_FSFREEZE_STATUS_FROZEN;
1313 return GUEST_FSFREEZE_STATUS_THAWED;
1316 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1318 return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1322 * Walk list of mounted file systems in the guest, and freeze the ones which
1323 * are real local file systems.
1325 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1326 strList *mountpoints,
1327 Error **errp)
1329 int ret = 0, i = 0;
1330 strList *list;
1331 FsMountList mounts;
1332 struct FsMount *mount;
1333 Error *local_err = NULL;
1334 int fd;
1336 slog("guest-fsfreeze called");
1338 execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1339 if (local_err) {
1340 error_propagate(errp, local_err);
1341 return -1;
1344 QTAILQ_INIT(&mounts);
1345 build_fs_mount_list(&mounts, &local_err);
1346 if (local_err) {
1347 error_propagate(errp, local_err);
1348 return -1;
1351 /* cannot risk guest agent blocking itself on a write in this state */
1352 ga_set_frozen(ga_state);
1354 QTAILQ_FOREACH_REVERSE(mount, &mounts, next) {
1355 /* To issue fsfreeze in the reverse order of mounts, check if the
1356 * mount is listed in the list here */
1357 if (has_mountpoints) {
1358 for (list = mountpoints; list; list = list->next) {
1359 if (strcmp(list->value, mount->dirname) == 0) {
1360 break;
1363 if (!list) {
1364 continue;
1368 fd = qemu_open_old(mount->dirname, O_RDONLY);
1369 if (fd == -1) {
1370 error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1371 goto error;
1374 /* we try to cull filesystems we know won't work in advance, but other
1375 * filesystems may not implement fsfreeze for less obvious reasons.
1376 * these will report EOPNOTSUPP. we simply ignore these when tallying
1377 * the number of frozen filesystems.
1378 * if a filesystem is mounted more than once (aka bind mount) a
1379 * consecutive attempt to freeze an already frozen filesystem will
1380 * return EBUSY.
1382 * any other error means a failure to freeze a filesystem we
1383 * expect to be freezable, so return an error in those cases
1384 * and return system to thawed state.
1386 ret = ioctl(fd, FIFREEZE);
1387 if (ret == -1) {
1388 if (errno != EOPNOTSUPP && errno != EBUSY) {
1389 error_setg_errno(errp, errno, "failed to freeze %s",
1390 mount->dirname);
1391 close(fd);
1392 goto error;
1394 } else {
1395 i++;
1397 close(fd);
1400 free_fs_mount_list(&mounts);
1401 /* We may not issue any FIFREEZE here.
1402 * Just unset ga_state here and ready for the next call.
1404 if (i == 0) {
1405 ga_unset_frozen(ga_state);
1407 return i;
1409 error:
1410 free_fs_mount_list(&mounts);
1411 qmp_guest_fsfreeze_thaw(NULL);
1412 return 0;
1416 * Walk list of frozen file systems in the guest, and thaw them.
1418 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1420 int ret;
1421 FsMountList mounts;
1422 FsMount *mount;
1423 int fd, i = 0, logged;
1424 Error *local_err = NULL;
1426 QTAILQ_INIT(&mounts);
1427 build_fs_mount_list(&mounts, &local_err);
1428 if (local_err) {
1429 error_propagate(errp, local_err);
1430 return 0;
1433 QTAILQ_FOREACH(mount, &mounts, next) {
1434 logged = false;
1435 fd = qemu_open_old(mount->dirname, O_RDONLY);
1436 if (fd == -1) {
1437 continue;
1439 /* we have no way of knowing whether a filesystem was actually unfrozen
1440 * as a result of a successful call to FITHAW, only that if an error
1441 * was returned the filesystem was *not* unfrozen by that particular
1442 * call.
1444 * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1445 * to unfreeze, continuing issuing FITHAW until an error is returned,
1446 * in which case either the filesystem is in an unfreezable state, or,
1447 * more likely, it was thawed previously (and remains so afterward).
1449 * also, since the most recent successful call is the one that did
1450 * the actual unfreeze, we can use this to provide an accurate count
1451 * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1452 * may * be useful for determining whether a filesystem was unfrozen
1453 * during the freeze/thaw phase by a process other than qemu-ga.
1455 do {
1456 ret = ioctl(fd, FITHAW);
1457 if (ret == 0 && !logged) {
1458 i++;
1459 logged = true;
1461 } while (ret == 0);
1462 close(fd);
1465 ga_unset_frozen(ga_state);
1466 free_fs_mount_list(&mounts);
1468 execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1470 return i;
1473 static void guest_fsfreeze_cleanup(void)
1475 Error *err = NULL;
1477 if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1478 qmp_guest_fsfreeze_thaw(&err);
1479 if (err) {
1480 slog("failed to clean up frozen filesystems: %s",
1481 error_get_pretty(err));
1482 error_free(err);
1486 #endif /* CONFIG_FSFREEZE */
1488 #if defined(CONFIG_FSTRIM)
1490 * Walk list of mounted file systems in the guest, and trim them.
1492 GuestFilesystemTrimResponse *
1493 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1495 GuestFilesystemTrimResponse *response;
1496 GuestFilesystemTrimResultList *list;
1497 GuestFilesystemTrimResult *result;
1498 int ret = 0;
1499 FsMountList mounts;
1500 struct FsMount *mount;
1501 int fd;
1502 Error *local_err = NULL;
1503 struct fstrim_range r;
1505 slog("guest-fstrim called");
1507 QTAILQ_INIT(&mounts);
1508 build_fs_mount_list(&mounts, &local_err);
1509 if (local_err) {
1510 error_propagate(errp, local_err);
1511 return NULL;
1514 response = g_malloc0(sizeof(*response));
1516 QTAILQ_FOREACH(mount, &mounts, next) {
1517 result = g_malloc0(sizeof(*result));
1518 result->path = g_strdup(mount->dirname);
1520 list = g_malloc0(sizeof(*list));
1521 list->value = result;
1522 list->next = response->paths;
1523 response->paths = list;
1525 fd = qemu_open_old(mount->dirname, O_RDONLY);
1526 if (fd == -1) {
1527 result->error = g_strdup_printf("failed to open: %s",
1528 strerror(errno));
1529 result->has_error = true;
1530 continue;
1533 /* We try to cull filesystems we know won't work in advance, but other
1534 * filesystems may not implement fstrim for less obvious reasons.
1535 * These will report EOPNOTSUPP; while in some other cases ENOTTY
1536 * will be reported (e.g. CD-ROMs).
1537 * Any other error means an unexpected error.
1539 r.start = 0;
1540 r.len = -1;
1541 r.minlen = has_minimum ? minimum : 0;
1542 ret = ioctl(fd, FITRIM, &r);
1543 if (ret == -1) {
1544 result->has_error = true;
1545 if (errno == ENOTTY || errno == EOPNOTSUPP) {
1546 result->error = g_strdup("trim not supported");
1547 } else {
1548 result->error = g_strdup_printf("failed to trim: %s",
1549 strerror(errno));
1551 close(fd);
1552 continue;
1555 result->has_minimum = true;
1556 result->minimum = r.minlen;
1557 result->has_trimmed = true;
1558 result->trimmed = r.len;
1559 close(fd);
1562 free_fs_mount_list(&mounts);
1563 return response;
1565 #endif /* CONFIG_FSTRIM */
1568 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1569 #define SUSPEND_SUPPORTED 0
1570 #define SUSPEND_NOT_SUPPORTED 1
1572 typedef enum {
1573 SUSPEND_MODE_DISK = 0,
1574 SUSPEND_MODE_RAM = 1,
1575 SUSPEND_MODE_HYBRID = 2,
1576 } SuspendMode;
1579 * Executes a command in a child process using g_spawn_sync,
1580 * returning an int >= 0 representing the exit status of the
1581 * process.
1583 * If the program wasn't found in path, returns -1.
1585 * If a problem happened when creating the child process,
1586 * returns -1 and errp is set.
1588 static int run_process_child(const char *command[], Error **errp)
1590 int exit_status, spawn_flag;
1591 GError *g_err = NULL;
1592 bool success;
1594 spawn_flag = G_SPAWN_SEARCH_PATH | G_SPAWN_STDOUT_TO_DEV_NULL |
1595 G_SPAWN_STDERR_TO_DEV_NULL;
1597 success = g_spawn_sync(NULL, (char **)command, environ, spawn_flag,
1598 NULL, NULL, NULL, NULL,
1599 &exit_status, &g_err);
1601 if (success) {
1602 return WEXITSTATUS(exit_status);
1605 if (g_err && (g_err->code != G_SPAWN_ERROR_NOENT)) {
1606 error_setg(errp, "failed to create child process, error '%s'",
1607 g_err->message);
1610 g_error_free(g_err);
1611 return -1;
1614 static bool systemd_supports_mode(SuspendMode mode, Error **errp)
1616 const char *systemctl_args[3] = {"systemd-hibernate", "systemd-suspend",
1617 "systemd-hybrid-sleep"};
1618 const char *cmd[4] = {"systemctl", "status", systemctl_args[mode], NULL};
1619 int status;
1621 status = run_process_child(cmd, errp);
1624 * systemctl status uses LSB return codes so we can expect
1625 * status > 0 and be ok. To assert if the guest has support
1626 * for the selected suspend mode, status should be < 4. 4 is
1627 * the code for unknown service status, the return value when
1628 * the service does not exist. A common value is status = 3
1629 * (program is not running).
1631 if (status > 0 && status < 4) {
1632 return true;
1635 return false;
1638 static void systemd_suspend(SuspendMode mode, Error **errp)
1640 Error *local_err = NULL;
1641 const char *systemctl_args[3] = {"hibernate", "suspend", "hybrid-sleep"};
1642 const char *cmd[3] = {"systemctl", systemctl_args[mode], NULL};
1643 int status;
1645 status = run_process_child(cmd, &local_err);
1647 if (status == 0) {
1648 return;
1651 if ((status == -1) && !local_err) {
1652 error_setg(errp, "the helper program 'systemctl %s' was not found",
1653 systemctl_args[mode]);
1654 return;
1657 if (local_err) {
1658 error_propagate(errp, local_err);
1659 } else {
1660 error_setg(errp, "the helper program 'systemctl %s' returned an "
1661 "unexpected exit status code (%d)",
1662 systemctl_args[mode], status);
1666 static bool pmutils_supports_mode(SuspendMode mode, Error **errp)
1668 Error *local_err = NULL;
1669 const char *pmutils_args[3] = {"--hibernate", "--suspend",
1670 "--suspend-hybrid"};
1671 const char *cmd[3] = {"pm-is-supported", pmutils_args[mode], NULL};
1672 int status;
1674 status = run_process_child(cmd, &local_err);
1676 if (status == SUSPEND_SUPPORTED) {
1677 return true;
1680 if ((status == -1) && !local_err) {
1681 return false;
1684 if (local_err) {
1685 error_propagate(errp, local_err);
1686 } else {
1687 error_setg(errp,
1688 "the helper program '%s' returned an unexpected exit"
1689 " status code (%d)", "pm-is-supported", status);
1692 return false;
1695 static void pmutils_suspend(SuspendMode mode, Error **errp)
1697 Error *local_err = NULL;
1698 const char *pmutils_binaries[3] = {"pm-hibernate", "pm-suspend",
1699 "pm-suspend-hybrid"};
1700 const char *cmd[2] = {pmutils_binaries[mode], NULL};
1701 int status;
1703 status = run_process_child(cmd, &local_err);
1705 if (status == 0) {
1706 return;
1709 if ((status == -1) && !local_err) {
1710 error_setg(errp, "the helper program '%s' was not found",
1711 pmutils_binaries[mode]);
1712 return;
1715 if (local_err) {
1716 error_propagate(errp, local_err);
1717 } else {
1718 error_setg(errp,
1719 "the helper program '%s' returned an unexpected exit"
1720 " status code (%d)", pmutils_binaries[mode], status);
1724 static bool linux_sys_state_supports_mode(SuspendMode mode, Error **errp)
1726 const char *sysfile_strs[3] = {"disk", "mem", NULL};
1727 const char *sysfile_str = sysfile_strs[mode];
1728 char buf[32]; /* hopefully big enough */
1729 int fd;
1730 ssize_t ret;
1732 if (!sysfile_str) {
1733 error_setg(errp, "unknown guest suspend mode");
1734 return false;
1737 fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
1738 if (fd < 0) {
1739 return false;
1742 ret = read(fd, buf, sizeof(buf) - 1);
1743 close(fd);
1744 if (ret <= 0) {
1745 return false;
1747 buf[ret] = '\0';
1749 if (strstr(buf, sysfile_str)) {
1750 return true;
1752 return false;
1755 static void linux_sys_state_suspend(SuspendMode mode, Error **errp)
1757 Error *local_err = NULL;
1758 const char *sysfile_strs[3] = {"disk", "mem", NULL};
1759 const char *sysfile_str = sysfile_strs[mode];
1760 pid_t pid;
1761 int status;
1763 if (!sysfile_str) {
1764 error_setg(errp, "unknown guest suspend mode");
1765 return;
1768 pid = fork();
1769 if (!pid) {
1770 /* child */
1771 int fd;
1773 setsid();
1774 reopen_fd_to_null(0);
1775 reopen_fd_to_null(1);
1776 reopen_fd_to_null(2);
1778 fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1779 if (fd < 0) {
1780 _exit(EXIT_FAILURE);
1783 if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1784 _exit(EXIT_FAILURE);
1787 _exit(EXIT_SUCCESS);
1788 } else if (pid < 0) {
1789 error_setg_errno(errp, errno, "failed to create child process");
1790 return;
1793 ga_wait_child(pid, &status, &local_err);
1794 if (local_err) {
1795 error_propagate(errp, local_err);
1796 return;
1799 if (WEXITSTATUS(status)) {
1800 error_setg(errp, "child process has failed to suspend");
1805 static void guest_suspend(SuspendMode mode, Error **errp)
1807 Error *local_err = NULL;
1808 bool mode_supported = false;
1810 if (systemd_supports_mode(mode, &local_err)) {
1811 mode_supported = true;
1812 systemd_suspend(mode, &local_err);
1815 if (!local_err) {
1816 return;
1819 error_free(local_err);
1820 local_err = NULL;
1822 if (pmutils_supports_mode(mode, &local_err)) {
1823 mode_supported = true;
1824 pmutils_suspend(mode, &local_err);
1827 if (!local_err) {
1828 return;
1831 error_free(local_err);
1832 local_err = NULL;
1834 if (linux_sys_state_supports_mode(mode, &local_err)) {
1835 mode_supported = true;
1836 linux_sys_state_suspend(mode, &local_err);
1839 if (!mode_supported) {
1840 error_free(local_err);
1841 error_setg(errp,
1842 "the requested suspend mode is not supported by the guest");
1843 } else {
1844 error_propagate(errp, local_err);
1848 void qmp_guest_suspend_disk(Error **errp)
1850 guest_suspend(SUSPEND_MODE_DISK, errp);
1853 void qmp_guest_suspend_ram(Error **errp)
1855 guest_suspend(SUSPEND_MODE_RAM, errp);
1858 void qmp_guest_suspend_hybrid(Error **errp)
1860 guest_suspend(SUSPEND_MODE_HYBRID, errp);
1863 static GuestNetworkInterfaceList *
1864 guest_find_interface(GuestNetworkInterfaceList *head,
1865 const char *name)
1867 for (; head; head = head->next) {
1868 if (strcmp(head->value->name, name) == 0) {
1869 break;
1873 return head;
1876 static int guest_get_network_stats(const char *name,
1877 GuestNetworkInterfaceStat *stats)
1879 int name_len;
1880 char const *devinfo = "/proc/net/dev";
1881 FILE *fp;
1882 char *line = NULL, *colon;
1883 size_t n = 0;
1884 fp = fopen(devinfo, "r");
1885 if (!fp) {
1886 return -1;
1888 name_len = strlen(name);
1889 while (getline(&line, &n, fp) != -1) {
1890 long long dummy;
1891 long long rx_bytes;
1892 long long rx_packets;
1893 long long rx_errs;
1894 long long rx_dropped;
1895 long long tx_bytes;
1896 long long tx_packets;
1897 long long tx_errs;
1898 long long tx_dropped;
1899 char *trim_line;
1900 trim_line = g_strchug(line);
1901 if (trim_line[0] == '\0') {
1902 continue;
1904 colon = strchr(trim_line, ':');
1905 if (!colon) {
1906 continue;
1908 if (colon - name_len == trim_line &&
1909 strncmp(trim_line, name, name_len) == 0) {
1910 if (sscanf(colon + 1,
1911 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
1912 &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
1913 &dummy, &dummy, &dummy, &dummy,
1914 &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
1915 &dummy, &dummy, &dummy, &dummy) != 16) {
1916 continue;
1918 stats->rx_bytes = rx_bytes;
1919 stats->rx_packets = rx_packets;
1920 stats->rx_errs = rx_errs;
1921 stats->rx_dropped = rx_dropped;
1922 stats->tx_bytes = tx_bytes;
1923 stats->tx_packets = tx_packets;
1924 stats->tx_errs = tx_errs;
1925 stats->tx_dropped = tx_dropped;
1926 fclose(fp);
1927 g_free(line);
1928 return 0;
1931 fclose(fp);
1932 g_free(line);
1933 g_debug("/proc/net/dev: Interface '%s' not found", name);
1934 return -1;
1938 * Build information about guest interfaces
1940 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1942 GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
1943 struct ifaddrs *ifap, *ifa;
1945 if (getifaddrs(&ifap) < 0) {
1946 error_setg_errno(errp, errno, "getifaddrs failed");
1947 goto error;
1950 for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
1951 GuestNetworkInterfaceList *info;
1952 GuestIpAddressList **address_list = NULL, *address_item = NULL;
1953 GuestNetworkInterfaceStat *interface_stat = NULL;
1954 char addr4[INET_ADDRSTRLEN];
1955 char addr6[INET6_ADDRSTRLEN];
1956 int sock;
1957 struct ifreq ifr;
1958 unsigned char *mac_addr;
1959 void *p;
1961 g_debug("Processing %s interface", ifa->ifa_name);
1963 info = guest_find_interface(head, ifa->ifa_name);
1965 if (!info) {
1966 info = g_malloc0(sizeof(*info));
1967 info->value = g_malloc0(sizeof(*info->value));
1968 info->value->name = g_strdup(ifa->ifa_name);
1970 if (!cur_item) {
1971 head = cur_item = info;
1972 } else {
1973 cur_item->next = info;
1974 cur_item = info;
1978 if (!info->value->has_hardware_address &&
1979 ifa->ifa_flags & SIOCGIFHWADDR) {
1980 /* we haven't obtained HW address yet */
1981 sock = socket(PF_INET, SOCK_STREAM, 0);
1982 if (sock == -1) {
1983 error_setg_errno(errp, errno, "failed to create socket");
1984 goto error;
1987 memset(&ifr, 0, sizeof(ifr));
1988 pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
1989 if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
1990 error_setg_errno(errp, errno,
1991 "failed to get MAC address of %s",
1992 ifa->ifa_name);
1993 close(sock);
1994 goto error;
1997 close(sock);
1998 mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
2000 info->value->hardware_address =
2001 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
2002 (int) mac_addr[0], (int) mac_addr[1],
2003 (int) mac_addr[2], (int) mac_addr[3],
2004 (int) mac_addr[4], (int) mac_addr[5]);
2006 info->value->has_hardware_address = true;
2009 if (ifa->ifa_addr &&
2010 ifa->ifa_addr->sa_family == AF_INET) {
2011 /* interface with IPv4 address */
2012 p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
2013 if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
2014 error_setg_errno(errp, errno, "inet_ntop failed");
2015 goto error;
2018 address_item = g_malloc0(sizeof(*address_item));
2019 address_item->value = g_malloc0(sizeof(*address_item->value));
2020 address_item->value->ip_address = g_strdup(addr4);
2021 address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
2023 if (ifa->ifa_netmask) {
2024 /* Count the number of set bits in netmask.
2025 * This is safe as '1' and '0' cannot be shuffled in netmask. */
2026 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
2027 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
2029 } else if (ifa->ifa_addr &&
2030 ifa->ifa_addr->sa_family == AF_INET6) {
2031 /* interface with IPv6 address */
2032 p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
2033 if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
2034 error_setg_errno(errp, errno, "inet_ntop failed");
2035 goto error;
2038 address_item = g_malloc0(sizeof(*address_item));
2039 address_item->value = g_malloc0(sizeof(*address_item->value));
2040 address_item->value->ip_address = g_strdup(addr6);
2041 address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
2043 if (ifa->ifa_netmask) {
2044 /* Count the number of set bits in netmask.
2045 * This is safe as '1' and '0' cannot be shuffled in netmask. */
2046 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
2047 address_item->value->prefix =
2048 ctpop32(((uint32_t *) p)[0]) +
2049 ctpop32(((uint32_t *) p)[1]) +
2050 ctpop32(((uint32_t *) p)[2]) +
2051 ctpop32(((uint32_t *) p)[3]);
2055 if (!address_item) {
2056 continue;
2059 address_list = &info->value->ip_addresses;
2061 while (*address_list && (*address_list)->next) {
2062 address_list = &(*address_list)->next;
2065 if (!*address_list) {
2066 *address_list = address_item;
2067 } else {
2068 (*address_list)->next = address_item;
2071 info->value->has_ip_addresses = true;
2073 if (!info->value->has_statistics) {
2074 interface_stat = g_malloc0(sizeof(*interface_stat));
2075 if (guest_get_network_stats(info->value->name,
2076 interface_stat) == -1) {
2077 info->value->has_statistics = false;
2078 g_free(interface_stat);
2079 } else {
2080 info->value->statistics = interface_stat;
2081 info->value->has_statistics = true;
2086 freeifaddrs(ifap);
2087 return head;
2089 error:
2090 freeifaddrs(ifap);
2091 qapi_free_GuestNetworkInterfaceList(head);
2092 return NULL;
2095 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
2097 static long sysconf_exact(int name, const char *name_str, Error **errp)
2099 long ret;
2101 errno = 0;
2102 ret = sysconf(name);
2103 if (ret == -1) {
2104 if (errno == 0) {
2105 error_setg(errp, "sysconf(%s): value indefinite", name_str);
2106 } else {
2107 error_setg_errno(errp, errno, "sysconf(%s)", name_str);
2110 return ret;
2113 /* Transfer online/offline status between @vcpu and the guest system.
2115 * On input either @errp or *@errp must be NULL.
2117 * In system-to-@vcpu direction, the following @vcpu fields are accessed:
2118 * - R: vcpu->logical_id
2119 * - W: vcpu->online
2120 * - W: vcpu->can_offline
2122 * In @vcpu-to-system direction, the following @vcpu fields are accessed:
2123 * - R: vcpu->logical_id
2124 * - R: vcpu->online
2126 * Written members remain unmodified on error.
2128 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
2129 char *dirpath, Error **errp)
2131 int fd;
2132 int res;
2133 int dirfd;
2134 static const char fn[] = "online";
2136 dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2137 if (dirfd == -1) {
2138 error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2139 return;
2142 fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
2143 if (fd == -1) {
2144 if (errno != ENOENT) {
2145 error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
2146 } else if (sys2vcpu) {
2147 vcpu->online = true;
2148 vcpu->can_offline = false;
2149 } else if (!vcpu->online) {
2150 error_setg(errp, "logical processor #%" PRId64 " can't be "
2151 "offlined", vcpu->logical_id);
2152 } /* otherwise pretend successful re-onlining */
2153 } else {
2154 unsigned char status;
2156 res = pread(fd, &status, 1, 0);
2157 if (res == -1) {
2158 error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
2159 } else if (res == 0) {
2160 error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
2161 fn);
2162 } else if (sys2vcpu) {
2163 vcpu->online = (status != '0');
2164 vcpu->can_offline = true;
2165 } else if (vcpu->online != (status != '0')) {
2166 status = '0' + vcpu->online;
2167 if (pwrite(fd, &status, 1, 0) == -1) {
2168 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
2169 fn);
2171 } /* otherwise pretend successful re-(on|off)-lining */
2173 res = close(fd);
2174 g_assert(res == 0);
2177 res = close(dirfd);
2178 g_assert(res == 0);
2181 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2183 int64_t current;
2184 GuestLogicalProcessorList *head, **link;
2185 long sc_max;
2186 Error *local_err = NULL;
2188 current = 0;
2189 head = NULL;
2190 link = &head;
2191 sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
2193 while (local_err == NULL && current < sc_max) {
2194 GuestLogicalProcessor *vcpu;
2195 GuestLogicalProcessorList *entry;
2196 int64_t id = current++;
2197 char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2198 id);
2200 if (g_file_test(path, G_FILE_TEST_EXISTS)) {
2201 vcpu = g_malloc0(sizeof *vcpu);
2202 vcpu->logical_id = id;
2203 vcpu->has_can_offline = true; /* lolspeak ftw */
2204 transfer_vcpu(vcpu, true, path, &local_err);
2205 entry = g_malloc0(sizeof *entry);
2206 entry->value = vcpu;
2207 *link = entry;
2208 link = &entry->next;
2210 g_free(path);
2213 if (local_err == NULL) {
2214 /* there's no guest with zero VCPUs */
2215 g_assert(head != NULL);
2216 return head;
2219 qapi_free_GuestLogicalProcessorList(head);
2220 error_propagate(errp, local_err);
2221 return NULL;
2224 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2226 int64_t processed;
2227 Error *local_err = NULL;
2229 processed = 0;
2230 while (vcpus != NULL) {
2231 char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2232 vcpus->value->logical_id);
2234 transfer_vcpu(vcpus->value, false, path, &local_err);
2235 g_free(path);
2236 if (local_err != NULL) {
2237 break;
2239 ++processed;
2240 vcpus = vcpus->next;
2243 if (local_err != NULL) {
2244 if (processed == 0) {
2245 error_propagate(errp, local_err);
2246 } else {
2247 error_free(local_err);
2251 return processed;
2254 void qmp_guest_set_user_password(const char *username,
2255 const char *password,
2256 bool crypted,
2257 Error **errp)
2259 Error *local_err = NULL;
2260 char *passwd_path = NULL;
2261 pid_t pid;
2262 int status;
2263 int datafd[2] = { -1, -1 };
2264 char *rawpasswddata = NULL;
2265 size_t rawpasswdlen;
2266 char *chpasswddata = NULL;
2267 size_t chpasswdlen;
2269 rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2270 if (!rawpasswddata) {
2271 return;
2273 rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2274 rawpasswddata[rawpasswdlen] = '\0';
2276 if (strchr(rawpasswddata, '\n')) {
2277 error_setg(errp, "forbidden characters in raw password");
2278 goto out;
2281 if (strchr(username, '\n') ||
2282 strchr(username, ':')) {
2283 error_setg(errp, "forbidden characters in username");
2284 goto out;
2287 chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2288 chpasswdlen = strlen(chpasswddata);
2290 passwd_path = g_find_program_in_path("chpasswd");
2292 if (!passwd_path) {
2293 error_setg(errp, "cannot find 'passwd' program in PATH");
2294 goto out;
2297 if (pipe(datafd) < 0) {
2298 error_setg(errp, "cannot create pipe FDs");
2299 goto out;
2302 pid = fork();
2303 if (pid == 0) {
2304 close(datafd[1]);
2305 /* child */
2306 setsid();
2307 dup2(datafd[0], 0);
2308 reopen_fd_to_null(1);
2309 reopen_fd_to_null(2);
2311 if (crypted) {
2312 execle(passwd_path, "chpasswd", "-e", NULL, environ);
2313 } else {
2314 execle(passwd_path, "chpasswd", NULL, environ);
2316 _exit(EXIT_FAILURE);
2317 } else if (pid < 0) {
2318 error_setg_errno(errp, errno, "failed to create child process");
2319 goto out;
2321 close(datafd[0]);
2322 datafd[0] = -1;
2324 if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2325 error_setg_errno(errp, errno, "cannot write new account password");
2326 goto out;
2328 close(datafd[1]);
2329 datafd[1] = -1;
2331 ga_wait_child(pid, &status, &local_err);
2332 if (local_err) {
2333 error_propagate(errp, local_err);
2334 goto out;
2337 if (!WIFEXITED(status)) {
2338 error_setg(errp, "child process has terminated abnormally");
2339 goto out;
2342 if (WEXITSTATUS(status)) {
2343 error_setg(errp, "child process has failed to set user password");
2344 goto out;
2347 out:
2348 g_free(chpasswddata);
2349 g_free(rawpasswddata);
2350 g_free(passwd_path);
2351 if (datafd[0] != -1) {
2352 close(datafd[0]);
2354 if (datafd[1] != -1) {
2355 close(datafd[1]);
2359 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2360 int size, Error **errp)
2362 int fd;
2363 int res;
2365 errno = 0;
2366 fd = openat(dirfd, pathname, O_RDONLY);
2367 if (fd == -1) {
2368 error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2369 return;
2372 res = pread(fd, buf, size, 0);
2373 if (res == -1) {
2374 error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2375 } else if (res == 0) {
2376 error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2378 close(fd);
2381 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2382 const char *buf, int size, Error **errp)
2384 int fd;
2386 errno = 0;
2387 fd = openat(dirfd, pathname, O_WRONLY);
2388 if (fd == -1) {
2389 error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2390 return;
2393 if (pwrite(fd, buf, size, 0) == -1) {
2394 error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2397 close(fd);
2400 /* Transfer online/offline status between @mem_blk and the guest system.
2402 * On input either @errp or *@errp must be NULL.
2404 * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2405 * - R: mem_blk->phys_index
2406 * - W: mem_blk->online
2407 * - W: mem_blk->can_offline
2409 * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2410 * - R: mem_blk->phys_index
2411 * - R: mem_blk->online
2412 *- R: mem_blk->can_offline
2413 * Written members remain unmodified on error.
2415 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2416 GuestMemoryBlockResponse *result,
2417 Error **errp)
2419 char *dirpath;
2420 int dirfd;
2421 char *status;
2422 Error *local_err = NULL;
2424 if (!sys2memblk) {
2425 DIR *dp;
2427 if (!result) {
2428 error_setg(errp, "Internal error, 'result' should not be NULL");
2429 return;
2431 errno = 0;
2432 dp = opendir("/sys/devices/system/memory/");
2433 /* if there is no 'memory' directory in sysfs,
2434 * we think this VM does not support online/offline memory block,
2435 * any other solution?
2437 if (!dp) {
2438 if (errno == ENOENT) {
2439 result->response =
2440 GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2442 goto out1;
2444 closedir(dp);
2447 dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2448 mem_blk->phys_index);
2449 dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2450 if (dirfd == -1) {
2451 if (sys2memblk) {
2452 error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2453 } else {
2454 if (errno == ENOENT) {
2455 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2456 } else {
2457 result->response =
2458 GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2461 g_free(dirpath);
2462 goto out1;
2464 g_free(dirpath);
2466 status = g_malloc0(10);
2467 ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2468 if (local_err) {
2469 /* treat with sysfs file that not exist in old kernel */
2470 if (errno == ENOENT) {
2471 error_free(local_err);
2472 if (sys2memblk) {
2473 mem_blk->online = true;
2474 mem_blk->can_offline = false;
2475 } else if (!mem_blk->online) {
2476 result->response =
2477 GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2479 } else {
2480 if (sys2memblk) {
2481 error_propagate(errp, local_err);
2482 } else {
2483 error_free(local_err);
2484 result->response =
2485 GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2488 goto out2;
2491 if (sys2memblk) {
2492 char removable = '0';
2494 mem_blk->online = (strncmp(status, "online", 6) == 0);
2496 ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2497 if (local_err) {
2498 /* if no 'removable' file, it doesn't support offline mem blk */
2499 if (errno == ENOENT) {
2500 error_free(local_err);
2501 mem_blk->can_offline = false;
2502 } else {
2503 error_propagate(errp, local_err);
2505 } else {
2506 mem_blk->can_offline = (removable != '0');
2508 } else {
2509 if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2510 const char *new_state = mem_blk->online ? "online" : "offline";
2512 ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2513 &local_err);
2514 if (local_err) {
2515 error_free(local_err);
2516 result->response =
2517 GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2518 goto out2;
2521 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2522 result->has_error_code = false;
2523 } /* otherwise pretend successful re-(on|off)-lining */
2525 g_free(status);
2526 close(dirfd);
2527 return;
2529 out2:
2530 g_free(status);
2531 close(dirfd);
2532 out1:
2533 if (!sys2memblk) {
2534 result->has_error_code = true;
2535 result->error_code = errno;
2539 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2541 GuestMemoryBlockList *head, **link;
2542 Error *local_err = NULL;
2543 struct dirent *de;
2544 DIR *dp;
2546 head = NULL;
2547 link = &head;
2549 dp = opendir("/sys/devices/system/memory/");
2550 if (!dp) {
2551 /* it's ok if this happens to be a system that doesn't expose
2552 * memory blocks via sysfs, but otherwise we should report
2553 * an error
2555 if (errno != ENOENT) {
2556 error_setg_errno(errp, errno, "Can't open directory"
2557 "\"/sys/devices/system/memory/\"");
2559 return NULL;
2562 /* Note: the phys_index of memory block may be discontinuous,
2563 * this is because a memblk is the unit of the Sparse Memory design, which
2564 * allows discontinuous memory ranges (ex. NUMA), so here we should
2565 * traverse the memory block directory.
2567 while ((de = readdir(dp)) != NULL) {
2568 GuestMemoryBlock *mem_blk;
2569 GuestMemoryBlockList *entry;
2571 if ((strncmp(de->d_name, "memory", 6) != 0) ||
2572 !(de->d_type & DT_DIR)) {
2573 continue;
2576 mem_blk = g_malloc0(sizeof *mem_blk);
2577 /* The d_name is "memoryXXX", phys_index is block id, same as XXX */
2578 mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2579 mem_blk->has_can_offline = true; /* lolspeak ftw */
2580 transfer_memory_block(mem_blk, true, NULL, &local_err);
2581 if (local_err) {
2582 break;
2585 entry = g_malloc0(sizeof *entry);
2586 entry->value = mem_blk;
2588 *link = entry;
2589 link = &entry->next;
2592 closedir(dp);
2593 if (local_err == NULL) {
2594 /* there's no guest with zero memory blocks */
2595 if (head == NULL) {
2596 error_setg(errp, "guest reported zero memory blocks!");
2598 return head;
2601 qapi_free_GuestMemoryBlockList(head);
2602 error_propagate(errp, local_err);
2603 return NULL;
2606 GuestMemoryBlockResponseList *
2607 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2609 GuestMemoryBlockResponseList *head, **link;
2610 Error *local_err = NULL;
2612 head = NULL;
2613 link = &head;
2615 while (mem_blks != NULL) {
2616 GuestMemoryBlockResponse *result;
2617 GuestMemoryBlockResponseList *entry;
2618 GuestMemoryBlock *current_mem_blk = mem_blks->value;
2620 result = g_malloc0(sizeof(*result));
2621 result->phys_index = current_mem_blk->phys_index;
2622 transfer_memory_block(current_mem_blk, false, result, &local_err);
2623 if (local_err) { /* should never happen */
2624 goto err;
2626 entry = g_malloc0(sizeof *entry);
2627 entry->value = result;
2629 *link = entry;
2630 link = &entry->next;
2631 mem_blks = mem_blks->next;
2634 return head;
2635 err:
2636 qapi_free_GuestMemoryBlockResponseList(head);
2637 error_propagate(errp, local_err);
2638 return NULL;
2641 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2643 Error *local_err = NULL;
2644 char *dirpath;
2645 int dirfd;
2646 char *buf;
2647 GuestMemoryBlockInfo *info;
2649 dirpath = g_strdup_printf("/sys/devices/system/memory/");
2650 dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2651 if (dirfd == -1) {
2652 error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2653 g_free(dirpath);
2654 return NULL;
2656 g_free(dirpath);
2658 buf = g_malloc0(20);
2659 ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2660 close(dirfd);
2661 if (local_err) {
2662 g_free(buf);
2663 error_propagate(errp, local_err);
2664 return NULL;
2667 info = g_new0(GuestMemoryBlockInfo, 1);
2668 info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2670 g_free(buf);
2672 return info;
2675 #else /* defined(__linux__) */
2677 void qmp_guest_suspend_disk(Error **errp)
2679 error_setg(errp, QERR_UNSUPPORTED);
2682 void qmp_guest_suspend_ram(Error **errp)
2684 error_setg(errp, QERR_UNSUPPORTED);
2687 void qmp_guest_suspend_hybrid(Error **errp)
2689 error_setg(errp, QERR_UNSUPPORTED);
2692 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2694 error_setg(errp, QERR_UNSUPPORTED);
2695 return NULL;
2698 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2700 error_setg(errp, QERR_UNSUPPORTED);
2701 return NULL;
2704 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2706 error_setg(errp, QERR_UNSUPPORTED);
2707 return -1;
2710 void qmp_guest_set_user_password(const char *username,
2711 const char *password,
2712 bool crypted,
2713 Error **errp)
2715 error_setg(errp, QERR_UNSUPPORTED);
2718 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2720 error_setg(errp, QERR_UNSUPPORTED);
2721 return NULL;
2724 GuestMemoryBlockResponseList *
2725 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2727 error_setg(errp, QERR_UNSUPPORTED);
2728 return NULL;
2731 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2733 error_setg(errp, QERR_UNSUPPORTED);
2734 return NULL;
2737 #endif
2739 #if !defined(CONFIG_FSFREEZE)
2741 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2743 error_setg(errp, QERR_UNSUPPORTED);
2744 return NULL;
2747 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2749 error_setg(errp, QERR_UNSUPPORTED);
2751 return 0;
2754 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
2756 error_setg(errp, QERR_UNSUPPORTED);
2758 return 0;
2761 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
2762 strList *mountpoints,
2763 Error **errp)
2765 error_setg(errp, QERR_UNSUPPORTED);
2767 return 0;
2770 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
2772 error_setg(errp, QERR_UNSUPPORTED);
2774 return 0;
2776 #endif /* CONFIG_FSFREEZE */
2778 #if !defined(CONFIG_FSTRIM)
2779 GuestFilesystemTrimResponse *
2780 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
2782 error_setg(errp, QERR_UNSUPPORTED);
2783 return NULL;
2785 #endif
2787 /* add unsupported commands to the blacklist */
2788 GList *ga_command_blacklist_init(GList *blacklist)
2790 #if !defined(__linux__)
2792 const char *list[] = {
2793 "guest-suspend-disk", "guest-suspend-ram",
2794 "guest-suspend-hybrid", "guest-network-get-interfaces",
2795 "guest-get-vcpus", "guest-set-vcpus",
2796 "guest-get-memory-blocks", "guest-set-memory-blocks",
2797 "guest-get-memory-block-size", "guest-get-memory-block-info",
2798 NULL};
2799 char **p = (char **)list;
2801 while (*p) {
2802 blacklist = g_list_append(blacklist, g_strdup(*p++));
2805 #endif
2807 #if !defined(CONFIG_FSFREEZE)
2809 const char *list[] = {
2810 "guest-get-fsinfo", "guest-fsfreeze-status",
2811 "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
2812 "guest-fsfreeze-thaw", "guest-get-fsinfo", NULL};
2813 char **p = (char **)list;
2815 while (*p) {
2816 blacklist = g_list_append(blacklist, g_strdup(*p++));
2819 #endif
2821 #if !defined(CONFIG_FSTRIM)
2822 blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
2823 #endif
2825 blacklist = g_list_append(blacklist, g_strdup("guest-get-devices"));
2827 return blacklist;
2830 /* register init/cleanup routines for stateful command groups */
2831 void ga_command_state_init(GAState *s, GACommandState *cs)
2833 #if defined(CONFIG_FSFREEZE)
2834 ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
2835 #endif
2838 #ifdef HAVE_UTMPX
2840 #define QGA_MICRO_SECOND_TO_SECOND 1000000
2842 static double ga_get_login_time(struct utmpx *user_info)
2844 double seconds = (double)user_info->ut_tv.tv_sec;
2845 double useconds = (double)user_info->ut_tv.tv_usec;
2846 useconds /= QGA_MICRO_SECOND_TO_SECOND;
2847 return seconds + useconds;
2850 GuestUserList *qmp_guest_get_users(Error **errp)
2852 GHashTable *cache = NULL;
2853 GuestUserList *head = NULL, *cur_item = NULL;
2854 struct utmpx *user_info = NULL;
2855 gpointer value = NULL;
2856 GuestUser *user = NULL;
2857 GuestUserList *item = NULL;
2858 double login_time = 0;
2860 cache = g_hash_table_new(g_str_hash, g_str_equal);
2861 setutxent();
2863 for (;;) {
2864 user_info = getutxent();
2865 if (user_info == NULL) {
2866 break;
2867 } else if (user_info->ut_type != USER_PROCESS) {
2868 continue;
2869 } else if (g_hash_table_contains(cache, user_info->ut_user)) {
2870 value = g_hash_table_lookup(cache, user_info->ut_user);
2871 user = (GuestUser *)value;
2872 login_time = ga_get_login_time(user_info);
2873 /* We're ensuring the earliest login time to be sent */
2874 if (login_time < user->login_time) {
2875 user->login_time = login_time;
2877 continue;
2880 item = g_new0(GuestUserList, 1);
2881 item->value = g_new0(GuestUser, 1);
2882 item->value->user = g_strdup(user_info->ut_user);
2883 item->value->login_time = ga_get_login_time(user_info);
2885 g_hash_table_insert(cache, item->value->user, item->value);
2887 if (!cur_item) {
2888 head = cur_item = item;
2889 } else {
2890 cur_item->next = item;
2891 cur_item = item;
2894 endutxent();
2895 g_hash_table_destroy(cache);
2896 return head;
2899 #else
2901 GuestUserList *qmp_guest_get_users(Error **errp)
2903 error_setg(errp, QERR_UNSUPPORTED);
2904 return NULL;
2907 #endif
2909 /* Replace escaped special characters with theire real values. The replacement
2910 * is done in place -- returned value is in the original string.
2912 static void ga_osrelease_replace_special(gchar *value)
2914 gchar *p, *p2, quote;
2916 /* Trim the string at first space or semicolon if it is not enclosed in
2917 * single or double quotes. */
2918 if ((value[0] != '"') || (value[0] == '\'')) {
2919 p = strchr(value, ' ');
2920 if (p != NULL) {
2921 *p = 0;
2923 p = strchr(value, ';');
2924 if (p != NULL) {
2925 *p = 0;
2927 return;
2930 quote = value[0];
2931 p2 = value;
2932 p = value + 1;
2933 while (*p != 0) {
2934 if (*p == '\\') {
2935 p++;
2936 switch (*p) {
2937 case '$':
2938 case '\'':
2939 case '"':
2940 case '\\':
2941 case '`':
2942 break;
2943 default:
2944 /* Keep literal backslash followed by whatever is there */
2945 p--;
2946 break;
2948 } else if (*p == quote) {
2949 *p2 = 0;
2950 break;
2952 *(p2++) = *(p++);
2956 static GKeyFile *ga_parse_osrelease(const char *fname)
2958 gchar *content = NULL;
2959 gchar *content2 = NULL;
2960 GError *err = NULL;
2961 GKeyFile *keys = g_key_file_new();
2962 const char *group = "[os-release]\n";
2964 if (!g_file_get_contents(fname, &content, NULL, &err)) {
2965 slog("failed to read '%s', error: %s", fname, err->message);
2966 goto fail;
2969 if (!g_utf8_validate(content, -1, NULL)) {
2970 slog("file is not utf-8 encoded: %s", fname);
2971 goto fail;
2973 content2 = g_strdup_printf("%s%s", group, content);
2975 if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
2976 &err)) {
2977 slog("failed to parse file '%s', error: %s", fname, err->message);
2978 goto fail;
2981 g_free(content);
2982 g_free(content2);
2983 return keys;
2985 fail:
2986 g_error_free(err);
2987 g_free(content);
2988 g_free(content2);
2989 g_key_file_free(keys);
2990 return NULL;
2993 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
2995 GuestOSInfo *info = NULL;
2996 struct utsname kinfo;
2997 GKeyFile *osrelease = NULL;
2998 const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
3000 info = g_new0(GuestOSInfo, 1);
3002 if (uname(&kinfo) != 0) {
3003 error_setg_errno(errp, errno, "uname failed");
3004 } else {
3005 info->has_kernel_version = true;
3006 info->kernel_version = g_strdup(kinfo.version);
3007 info->has_kernel_release = true;
3008 info->kernel_release = g_strdup(kinfo.release);
3009 info->has_machine = true;
3010 info->machine = g_strdup(kinfo.machine);
3013 if (qga_os_release != NULL) {
3014 osrelease = ga_parse_osrelease(qga_os_release);
3015 } else {
3016 osrelease = ga_parse_osrelease("/etc/os-release");
3017 if (osrelease == NULL) {
3018 osrelease = ga_parse_osrelease("/usr/lib/os-release");
3022 if (osrelease != NULL) {
3023 char *value;
3025 #define GET_FIELD(field, osfield) do { \
3026 value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
3027 if (value != NULL) { \
3028 ga_osrelease_replace_special(value); \
3029 info->has_ ## field = true; \
3030 info->field = value; \
3032 } while (0)
3033 GET_FIELD(id, "ID");
3034 GET_FIELD(name, "NAME");
3035 GET_FIELD(pretty_name, "PRETTY_NAME");
3036 GET_FIELD(version, "VERSION");
3037 GET_FIELD(version_id, "VERSION_ID");
3038 GET_FIELD(variant, "VARIANT");
3039 GET_FIELD(variant_id, "VARIANT_ID");
3040 #undef GET_FIELD
3042 g_key_file_free(osrelease);
3045 return info;
3048 GuestDeviceInfoList *qmp_guest_get_devices(Error **errp)
3050 error_setg(errp, QERR_UNSUPPORTED);
3052 return NULL;