xref: /qemu/qga/commands-posix.c (revision 25b5ff1a)
1 /*
2  * QEMU Guest Agent POSIX-specific command implementations
3  *
4  * Copyright IBM Corp. 2011
5  *
6  * Authors:
7  *  Michael Roth      <mdroth@linux.vnet.ibm.com>
8  *  Michal Privoznik  <mprivozn@redhat.com>
9  *
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.
12  */
13 
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 "guest-agent-core.h"
20 #include "qga-qapi-commands.h"
21 #include "qapi/error.h"
22 #include "qapi/qmp/qerror.h"
23 #include "qemu/queue.h"
24 #include "qemu/host-utils.h"
25 #include "qemu/sockets.h"
26 #include "qemu/base64.h"
27 #include "qemu/cutils.h"
28 
29 #ifdef HAVE_UTMPX
30 #include <utmpx.h>
31 #endif
32 
33 #ifndef CONFIG_HAS_ENVIRON
34 #ifdef __APPLE__
35 #include <crt_externs.h>
36 #define environ (*_NSGetEnviron())
37 #else
38 extern char **environ;
39 #endif
40 #endif
41 
42 #if defined(__linux__)
43 #include <mntent.h>
44 #include <linux/fs.h>
45 #include <ifaddrs.h>
46 #include <arpa/inet.h>
47 #include <sys/socket.h>
48 #include <net/if.h>
49 #include <sys/statvfs.h>
50 
51 #ifdef FIFREEZE
52 #define CONFIG_FSFREEZE
53 #endif
54 #ifdef FITRIM
55 #define CONFIG_FSTRIM
56 #endif
57 #endif
58 
59 static void ga_wait_child(pid_t pid, int *status, Error **errp)
60 {
61     pid_t rpid;
62 
63     *status = 0;
64 
65     do {
66         rpid = waitpid(pid, status, 0);
67     } while (rpid == -1 && errno == EINTR);
68 
69     if (rpid == -1) {
70         error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
71                          pid);
72         return;
73     }
74 
75     g_assert(rpid == pid);
76 }
77 
78 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
79 {
80     const char *shutdown_flag;
81     Error *local_err = NULL;
82     pid_t pid;
83     int status;
84 
85     slog("guest-shutdown called, mode: %s", mode);
86     if (!has_mode || strcmp(mode, "powerdown") == 0) {
87         shutdown_flag = "-P";
88     } else if (strcmp(mode, "halt") == 0) {
89         shutdown_flag = "-H";
90     } else if (strcmp(mode, "reboot") == 0) {
91         shutdown_flag = "-r";
92     } else {
93         error_setg(errp,
94                    "mode is invalid (valid values are: halt|powerdown|reboot");
95         return;
96     }
97 
98     pid = fork();
99     if (pid == 0) {
100         /* child, start the shutdown */
101         setsid();
102         reopen_fd_to_null(0);
103         reopen_fd_to_null(1);
104         reopen_fd_to_null(2);
105 
106         execle("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
107                "hypervisor initiated shutdown", (char*)NULL, environ);
108         _exit(EXIT_FAILURE);
109     } else if (pid < 0) {
110         error_setg_errno(errp, errno, "failed to create child process");
111         return;
112     }
113 
114     ga_wait_child(pid, &status, &local_err);
115     if (local_err) {
116         error_propagate(errp, local_err);
117         return;
118     }
119 
120     if (!WIFEXITED(status)) {
121         error_setg(errp, "child process has terminated abnormally");
122         return;
123     }
124 
125     if (WEXITSTATUS(status)) {
126         error_setg(errp, "child process has failed to shutdown");
127         return;
128     }
129 
130     /* succeeded */
131 }
132 
133 int64_t qmp_guest_get_time(Error **errp)
134 {
135    int ret;
136    qemu_timeval tq;
137 
138    ret = qemu_gettimeofday(&tq);
139    if (ret < 0) {
140        error_setg_errno(errp, errno, "Failed to get time");
141        return -1;
142    }
143 
144    return tq.tv_sec * 1000000000LL + tq.tv_usec * 1000;
145 }
146 
147 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
148 {
149     int ret;
150     int status;
151     pid_t pid;
152     Error *local_err = NULL;
153     struct timeval tv;
154 
155     /* If user has passed a time, validate and set it. */
156     if (has_time) {
157         GDate date = { 0, };
158 
159         /* year-2038 will overflow in case time_t is 32bit */
160         if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
161             error_setg(errp, "Time %" PRId64 " is too large", time_ns);
162             return;
163         }
164 
165         tv.tv_sec = time_ns / 1000000000;
166         tv.tv_usec = (time_ns % 1000000000) / 1000;
167         g_date_set_time_t(&date, tv.tv_sec);
168         if (date.year < 1970 || date.year >= 2070) {
169             error_setg_errno(errp, errno, "Invalid time");
170             return;
171         }
172 
173         ret = settimeofday(&tv, NULL);
174         if (ret < 0) {
175             error_setg_errno(errp, errno, "Failed to set time to guest");
176             return;
177         }
178     }
179 
180     /* Now, if user has passed a time to set and the system time is set, we
181      * just need to synchronize the hardware clock. However, if no time was
182      * passed, user is requesting the opposite: set the system time from the
183      * hardware clock (RTC). */
184     pid = fork();
185     if (pid == 0) {
186         setsid();
187         reopen_fd_to_null(0);
188         reopen_fd_to_null(1);
189         reopen_fd_to_null(2);
190 
191         /* Use '/sbin/hwclock -w' to set RTC from the system time,
192          * or '/sbin/hwclock -s' to set the system time from RTC. */
193         execle("/sbin/hwclock", "hwclock", has_time ? "-w" : "-s",
194                NULL, environ);
195         _exit(EXIT_FAILURE);
196     } else if (pid < 0) {
197         error_setg_errno(errp, errno, "failed to create child process");
198         return;
199     }
200 
201     ga_wait_child(pid, &status, &local_err);
202     if (local_err) {
203         error_propagate(errp, local_err);
204         return;
205     }
206 
207     if (!WIFEXITED(status)) {
208         error_setg(errp, "child process has terminated abnormally");
209         return;
210     }
211 
212     if (WEXITSTATUS(status)) {
213         error_setg(errp, "hwclock failed to set hardware clock to system time");
214         return;
215     }
216 }
217 
218 typedef enum {
219     RW_STATE_NEW,
220     RW_STATE_READING,
221     RW_STATE_WRITING,
222 } RwState;
223 
224 typedef struct GuestFileHandle {
225     uint64_t id;
226     FILE *fh;
227     RwState state;
228     QTAILQ_ENTRY(GuestFileHandle) next;
229 } GuestFileHandle;
230 
231 static struct {
232     QTAILQ_HEAD(, GuestFileHandle) filehandles;
233 } guest_file_state = {
234     .filehandles = QTAILQ_HEAD_INITIALIZER(guest_file_state.filehandles),
235 };
236 
237 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
238 {
239     GuestFileHandle *gfh;
240     int64_t handle;
241 
242     handle = ga_get_fd_handle(ga_state, errp);
243     if (handle < 0) {
244         return -1;
245     }
246 
247     gfh = g_new0(GuestFileHandle, 1);
248     gfh->id = handle;
249     gfh->fh = fh;
250     QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
251 
252     return handle;
253 }
254 
255 static GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
256 {
257     GuestFileHandle *gfh;
258 
259     QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
260     {
261         if (gfh->id == id) {
262             return gfh;
263         }
264     }
265 
266     error_setg(errp, "handle '%" PRId64 "' has not been found", id);
267     return NULL;
268 }
269 
270 typedef const char * const ccpc;
271 
272 #ifndef O_BINARY
273 #define O_BINARY 0
274 #endif
275 
276 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
277 static const struct {
278     ccpc *forms;
279     int oflag_base;
280 } guest_file_open_modes[] = {
281     { (ccpc[]){ "r",          NULL }, O_RDONLY                                 },
282     { (ccpc[]){ "rb",         NULL }, O_RDONLY                      | O_BINARY },
283     { (ccpc[]){ "w",          NULL }, O_WRONLY | O_CREAT | O_TRUNC             },
284     { (ccpc[]){ "wb",         NULL }, O_WRONLY | O_CREAT | O_TRUNC  | O_BINARY },
285     { (ccpc[]){ "a",          NULL }, O_WRONLY | O_CREAT | O_APPEND            },
286     { (ccpc[]){ "ab",         NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
287     { (ccpc[]){ "r+",         NULL }, O_RDWR                                   },
288     { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR                        | O_BINARY },
289     { (ccpc[]){ "w+",         NULL }, O_RDWR   | O_CREAT | O_TRUNC             },
290     { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR   | O_CREAT | O_TRUNC  | O_BINARY },
291     { (ccpc[]){ "a+",         NULL }, O_RDWR   | O_CREAT | O_APPEND            },
292     { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR   | O_CREAT | O_APPEND | O_BINARY }
293 };
294 
295 static int
296 find_open_flag(const char *mode_str, Error **errp)
297 {
298     unsigned mode;
299 
300     for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
301         ccpc *form;
302 
303         form = guest_file_open_modes[mode].forms;
304         while (*form != NULL && strcmp(*form, mode_str) != 0) {
305             ++form;
306         }
307         if (*form != NULL) {
308             break;
309         }
310     }
311 
312     if (mode == ARRAY_SIZE(guest_file_open_modes)) {
313         error_setg(errp, "invalid file open mode '%s'", mode_str);
314         return -1;
315     }
316     return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
317 }
318 
319 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
320                                S_IRGRP | S_IWGRP | \
321                                S_IROTH | S_IWOTH)
322 
323 static FILE *
324 safe_open_or_create(const char *path, const char *mode, Error **errp)
325 {
326     Error *local_err = NULL;
327     int oflag;
328 
329     oflag = find_open_flag(mode, &local_err);
330     if (local_err == NULL) {
331         int fd;
332 
333         /* If the caller wants / allows creation of a new file, we implement it
334          * with a two step process: open() + (open() / fchmod()).
335          *
336          * First we insist on creating the file exclusively as a new file. If
337          * that succeeds, we're free to set any file-mode bits on it. (The
338          * motivation is that we want to set those file-mode bits independently
339          * of the current umask.)
340          *
341          * If the exclusive creation fails because the file already exists
342          * (EEXIST is not possible for any other reason), we just attempt to
343          * open the file, but in this case we won't be allowed to change the
344          * file-mode bits on the preexistent file.
345          *
346          * The pathname should never disappear between the two open()s in
347          * practice. If it happens, then someone very likely tried to race us.
348          * In this case just go ahead and report the ENOENT from the second
349          * open() to the caller.
350          *
351          * If the caller wants to open a preexistent file, then the first
352          * open() is decisive and its third argument is ignored, and the second
353          * open() and the fchmod() are never called.
354          */
355         fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
356         if (fd == -1 && errno == EEXIST) {
357             oflag &= ~(unsigned)O_CREAT;
358             fd = open(path, oflag);
359         }
360 
361         if (fd == -1) {
362             error_setg_errno(&local_err, errno, "failed to open file '%s' "
363                              "(mode: '%s')", path, mode);
364         } else {
365             qemu_set_cloexec(fd);
366 
367             if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
368                 error_setg_errno(&local_err, errno, "failed to set permission "
369                                  "0%03o on new file '%s' (mode: '%s')",
370                                  (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
371             } else {
372                 FILE *f;
373 
374                 f = fdopen(fd, mode);
375                 if (f == NULL) {
376                     error_setg_errno(&local_err, errno, "failed to associate "
377                                      "stdio stream with file descriptor %d, "
378                                      "file '%s' (mode: '%s')", fd, path, mode);
379                 } else {
380                     return f;
381                 }
382             }
383 
384             close(fd);
385             if (oflag & O_CREAT) {
386                 unlink(path);
387             }
388         }
389     }
390 
391     error_propagate(errp, local_err);
392     return NULL;
393 }
394 
395 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
396                             Error **errp)
397 {
398     FILE *fh;
399     Error *local_err = NULL;
400     int64_t handle;
401 
402     if (!has_mode) {
403         mode = "r";
404     }
405     slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
406     fh = safe_open_or_create(path, mode, &local_err);
407     if (local_err != NULL) {
408         error_propagate(errp, local_err);
409         return -1;
410     }
411 
412     /* set fd non-blocking to avoid common use cases (like reading from a
413      * named pipe) from hanging the agent
414      */
415     qemu_set_nonblock(fileno(fh));
416 
417     handle = guest_file_handle_add(fh, errp);
418     if (handle < 0) {
419         fclose(fh);
420         return -1;
421     }
422 
423     slog("guest-file-open, handle: %" PRId64, handle);
424     return handle;
425 }
426 
427 void qmp_guest_file_close(int64_t handle, Error **errp)
428 {
429     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
430     int ret;
431 
432     slog("guest-file-close called, handle: %" PRId64, handle);
433     if (!gfh) {
434         return;
435     }
436 
437     ret = fclose(gfh->fh);
438     if (ret == EOF) {
439         error_setg_errno(errp, errno, "failed to close handle");
440         return;
441     }
442 
443     QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
444     g_free(gfh);
445 }
446 
447 struct GuestFileRead *qmp_guest_file_read(int64_t handle, bool has_count,
448                                           int64_t count, Error **errp)
449 {
450     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
451     GuestFileRead *read_data = NULL;
452     guchar *buf;
453     FILE *fh;
454     size_t read_count;
455 
456     if (!gfh) {
457         return NULL;
458     }
459 
460     if (!has_count) {
461         count = QGA_READ_COUNT_DEFAULT;
462     } else if (count < 0 || count >= UINT32_MAX) {
463         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
464                    count);
465         return NULL;
466     }
467 
468     fh = gfh->fh;
469 
470     /* explicitly flush when switching from writing to reading */
471     if (gfh->state == RW_STATE_WRITING) {
472         int ret = fflush(fh);
473         if (ret == EOF) {
474             error_setg_errno(errp, errno, "failed to flush file");
475             return NULL;
476         }
477         gfh->state = RW_STATE_NEW;
478     }
479 
480     buf = g_malloc0(count+1);
481     read_count = fread(buf, 1, count, fh);
482     if (ferror(fh)) {
483         error_setg_errno(errp, errno, "failed to read file");
484         slog("guest-file-read failed, handle: %" PRId64, handle);
485     } else {
486         buf[read_count] = 0;
487         read_data = g_new0(GuestFileRead, 1);
488         read_data->count = read_count;
489         read_data->eof = feof(fh);
490         if (read_count) {
491             read_data->buf_b64 = g_base64_encode(buf, read_count);
492         }
493         gfh->state = RW_STATE_READING;
494     }
495     g_free(buf);
496     clearerr(fh);
497 
498     return read_data;
499 }
500 
501 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
502                                      bool has_count, int64_t count,
503                                      Error **errp)
504 {
505     GuestFileWrite *write_data = NULL;
506     guchar *buf;
507     gsize buf_len;
508     int write_count;
509     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
510     FILE *fh;
511 
512     if (!gfh) {
513         return NULL;
514     }
515 
516     fh = gfh->fh;
517 
518     if (gfh->state == RW_STATE_READING) {
519         int ret = fseek(fh, 0, SEEK_CUR);
520         if (ret == -1) {
521             error_setg_errno(errp, errno, "failed to seek file");
522             return NULL;
523         }
524         gfh->state = RW_STATE_NEW;
525     }
526 
527     buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
528     if (!buf) {
529         return NULL;
530     }
531 
532     if (!has_count) {
533         count = buf_len;
534     } else if (count < 0 || count > buf_len) {
535         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
536                    count);
537         g_free(buf);
538         return NULL;
539     }
540 
541     write_count = fwrite(buf, 1, count, fh);
542     if (ferror(fh)) {
543         error_setg_errno(errp, errno, "failed to write to file");
544         slog("guest-file-write failed, handle: %" PRId64, handle);
545     } else {
546         write_data = g_new0(GuestFileWrite, 1);
547         write_data->count = write_count;
548         write_data->eof = feof(fh);
549         gfh->state = RW_STATE_WRITING;
550     }
551     g_free(buf);
552     clearerr(fh);
553 
554     return write_data;
555 }
556 
557 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
558                                           GuestFileWhence *whence_code,
559                                           Error **errp)
560 {
561     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
562     GuestFileSeek *seek_data = NULL;
563     FILE *fh;
564     int ret;
565     int whence;
566     Error *err = NULL;
567 
568     if (!gfh) {
569         return NULL;
570     }
571 
572     /* We stupidly exposed 'whence':'int' in our qapi */
573     whence = ga_parse_whence(whence_code, &err);
574     if (err) {
575         error_propagate(errp, err);
576         return NULL;
577     }
578 
579     fh = gfh->fh;
580     ret = fseek(fh, offset, whence);
581     if (ret == -1) {
582         error_setg_errno(errp, errno, "failed to seek file");
583         if (errno == ESPIPE) {
584             /* file is non-seekable, stdio shouldn't be buffering anyways */
585             gfh->state = RW_STATE_NEW;
586         }
587     } else {
588         seek_data = g_new0(GuestFileSeek, 1);
589         seek_data->position = ftell(fh);
590         seek_data->eof = feof(fh);
591         gfh->state = RW_STATE_NEW;
592     }
593     clearerr(fh);
594 
595     return seek_data;
596 }
597 
598 void qmp_guest_file_flush(int64_t handle, Error **errp)
599 {
600     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
601     FILE *fh;
602     int ret;
603 
604     if (!gfh) {
605         return;
606     }
607 
608     fh = gfh->fh;
609     ret = fflush(fh);
610     if (ret == EOF) {
611         error_setg_errno(errp, errno, "failed to flush file");
612     } else {
613         gfh->state = RW_STATE_NEW;
614     }
615 }
616 
617 /* linux-specific implementations. avoid this if at all possible. */
618 #if defined(__linux__)
619 
620 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
621 typedef struct FsMount {
622     char *dirname;
623     char *devtype;
624     unsigned int devmajor, devminor;
625     QTAILQ_ENTRY(FsMount) next;
626 } FsMount;
627 
628 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
629 
630 static void free_fs_mount_list(FsMountList *mounts)
631 {
632      FsMount *mount, *temp;
633 
634      if (!mounts) {
635          return;
636      }
637 
638      QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
639          QTAILQ_REMOVE(mounts, mount, next);
640          g_free(mount->dirname);
641          g_free(mount->devtype);
642          g_free(mount);
643      }
644 }
645 
646 static int dev_major_minor(const char *devpath,
647                            unsigned int *devmajor, unsigned int *devminor)
648 {
649     struct stat st;
650 
651     *devmajor = 0;
652     *devminor = 0;
653 
654     if (stat(devpath, &st) < 0) {
655         slog("failed to stat device file '%s': %s", devpath, strerror(errno));
656         return -1;
657     }
658     if (S_ISDIR(st.st_mode)) {
659         /* It is bind mount */
660         return -2;
661     }
662     if (S_ISBLK(st.st_mode)) {
663         *devmajor = major(st.st_rdev);
664         *devminor = minor(st.st_rdev);
665         return 0;
666     }
667     return -1;
668 }
669 
670 /*
671  * Walk the mount table and build a list of local file systems
672  */
673 static void build_fs_mount_list_from_mtab(FsMountList *mounts, Error **errp)
674 {
675     struct mntent *ment;
676     FsMount *mount;
677     char const *mtab = "/proc/self/mounts";
678     FILE *fp;
679     unsigned int devmajor, devminor;
680 
681     fp = setmntent(mtab, "r");
682     if (!fp) {
683         error_setg(errp, "failed to open mtab file: '%s'", mtab);
684         return;
685     }
686 
687     while ((ment = getmntent(fp))) {
688         /*
689          * An entry which device name doesn't start with a '/' is
690          * either a dummy file system or a network file system.
691          * Add special handling for smbfs and cifs as is done by
692          * coreutils as well.
693          */
694         if ((ment->mnt_fsname[0] != '/') ||
695             (strcmp(ment->mnt_type, "smbfs") == 0) ||
696             (strcmp(ment->mnt_type, "cifs") == 0)) {
697             continue;
698         }
699         if (dev_major_minor(ment->mnt_fsname, &devmajor, &devminor) == -2) {
700             /* Skip bind mounts */
701             continue;
702         }
703 
704         mount = g_new0(FsMount, 1);
705         mount->dirname = g_strdup(ment->mnt_dir);
706         mount->devtype = g_strdup(ment->mnt_type);
707         mount->devmajor = devmajor;
708         mount->devminor = devminor;
709 
710         QTAILQ_INSERT_TAIL(mounts, mount, next);
711     }
712 
713     endmntent(fp);
714 }
715 
716 static void decode_mntname(char *name, int len)
717 {
718     int i, j = 0;
719     for (i = 0; i <= len; i++) {
720         if (name[i] != '\\') {
721             name[j++] = name[i];
722         } else if (name[i + 1] == '\\') {
723             name[j++] = '\\';
724             i++;
725         } else if (name[i + 1] >= '0' && name[i + 1] <= '3' &&
726                    name[i + 2] >= '0' && name[i + 2] <= '7' &&
727                    name[i + 3] >= '0' && name[i + 3] <= '7') {
728             name[j++] = (name[i + 1] - '0') * 64 +
729                         (name[i + 2] - '0') * 8 +
730                         (name[i + 3] - '0');
731             i += 3;
732         } else {
733             name[j++] = name[i];
734         }
735     }
736 }
737 
738 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
739 {
740     FsMount *mount;
741     char const *mountinfo = "/proc/self/mountinfo";
742     FILE *fp;
743     char *line = NULL, *dash;
744     size_t n;
745     char check;
746     unsigned int devmajor, devminor;
747     int ret, dir_s, dir_e, type_s, type_e, dev_s, dev_e;
748 
749     fp = fopen(mountinfo, "r");
750     if (!fp) {
751         build_fs_mount_list_from_mtab(mounts, errp);
752         return;
753     }
754 
755     while (getline(&line, &n, fp) != -1) {
756         ret = sscanf(line, "%*u %*u %u:%u %*s %n%*s%n%c",
757                      &devmajor, &devminor, &dir_s, &dir_e, &check);
758         if (ret < 3) {
759             continue;
760         }
761         dash = strstr(line + dir_e, " - ");
762         if (!dash) {
763             continue;
764         }
765         ret = sscanf(dash, " - %n%*s%n %n%*s%n%c",
766                      &type_s, &type_e, &dev_s, &dev_e, &check);
767         if (ret < 1) {
768             continue;
769         }
770         line[dir_e] = 0;
771         dash[type_e] = 0;
772         dash[dev_e] = 0;
773         decode_mntname(line + dir_s, dir_e - dir_s);
774         decode_mntname(dash + dev_s, dev_e - dev_s);
775         if (devmajor == 0) {
776             /* btrfs reports major number = 0 */
777             if (strcmp("btrfs", dash + type_s) != 0 ||
778                 dev_major_minor(dash + dev_s, &devmajor, &devminor) < 0) {
779                 continue;
780             }
781         }
782 
783         mount = g_new0(FsMount, 1);
784         mount->dirname = g_strdup(line + dir_s);
785         mount->devtype = g_strdup(dash + type_s);
786         mount->devmajor = devmajor;
787         mount->devminor = devminor;
788 
789         QTAILQ_INSERT_TAIL(mounts, mount, next);
790     }
791     free(line);
792 
793     fclose(fp);
794 }
795 #endif
796 
797 #if defined(CONFIG_FSFREEZE)
798 
799 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
800 {
801     char *path;
802     char *dpath;
803     char *driver = NULL;
804     char buf[PATH_MAX];
805     ssize_t len;
806 
807     path = g_strndup(syspath, pathlen);
808     dpath = g_strdup_printf("%s/driver", path);
809     len = readlink(dpath, buf, sizeof(buf) - 1);
810     if (len != -1) {
811         buf[len] = 0;
812         driver = g_path_get_basename(buf);
813     }
814     g_free(dpath);
815     g_free(path);
816     return driver;
817 }
818 
819 static int compare_uint(const void *_a, const void *_b)
820 {
821     unsigned int a = *(unsigned int *)_a;
822     unsigned int b = *(unsigned int *)_b;
823 
824     return a < b ? -1 : a > b ? 1 : 0;
825 }
826 
827 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
828 static int build_hosts(char const *syspath, char const *host, bool ata,
829                        unsigned int *hosts, int hosts_max, Error **errp)
830 {
831     char *path;
832     DIR *dir;
833     struct dirent *entry;
834     int i = 0;
835 
836     path = g_strndup(syspath, host - syspath);
837     dir = opendir(path);
838     if (!dir) {
839         error_setg_errno(errp, errno, "opendir(\"%s\")", path);
840         g_free(path);
841         return -1;
842     }
843 
844     while (i < hosts_max) {
845         entry = readdir(dir);
846         if (!entry) {
847             break;
848         }
849         if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
850             ++i;
851         } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
852             ++i;
853         }
854     }
855 
856     qsort(hosts, i, sizeof(hosts[0]), compare_uint);
857 
858     g_free(path);
859     closedir(dir);
860     return i;
861 }
862 
863 /* Store disk device info specified by @sysfs into @fs */
864 static void build_guest_fsinfo_for_real_device(char const *syspath,
865                                                GuestFilesystemInfo *fs,
866                                                Error **errp)
867 {
868     unsigned int pci[4], host, hosts[8], tgt[3];
869     int i, nhosts = 0, pcilen;
870     GuestDiskAddress *disk;
871     GuestPCIAddress *pciaddr;
872     GuestDiskAddressList *list = NULL;
873     bool has_ata = false, has_host = false, has_tgt = false;
874     char *p, *q, *driver = NULL;
875 
876     p = strstr(syspath, "/devices/pci");
877     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
878                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
879         g_debug("only pci device is supported: sysfs path \"%s\"", syspath);
880         return;
881     }
882 
883     driver = get_pci_driver(syspath, (p + 12 + pcilen) - syspath, errp);
884     if (!driver) {
885         goto cleanup;
886     }
887 
888     p = strstr(syspath, "/target");
889     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
890                     tgt, tgt + 1, tgt + 2) == 3) {
891         has_tgt = true;
892     }
893 
894     p = strstr(syspath, "/ata");
895     if (p) {
896         q = p + 4;
897         has_ata = true;
898     } else {
899         p = strstr(syspath, "/host");
900         q = p + 5;
901     }
902     if (p && sscanf(q, "%u", &host) == 1) {
903         has_host = true;
904         nhosts = build_hosts(syspath, p, has_ata, hosts,
905                              ARRAY_SIZE(hosts), errp);
906         if (nhosts < 0) {
907             goto cleanup;
908         }
909     }
910 
911     pciaddr = g_malloc0(sizeof(*pciaddr));
912     pciaddr->domain = pci[0];
913     pciaddr->bus = pci[1];
914     pciaddr->slot = pci[2];
915     pciaddr->function = pci[3];
916 
917     disk = g_malloc0(sizeof(*disk));
918     disk->pci_controller = pciaddr;
919 
920     list = g_malloc0(sizeof(*list));
921     list->value = disk;
922 
923     if (strcmp(driver, "ata_piix") == 0) {
924         /* a host per ide bus, target*:0:<unit>:0 */
925         if (!has_host || !has_tgt) {
926             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
927             goto cleanup;
928         }
929         for (i = 0; i < nhosts; i++) {
930             if (host == hosts[i]) {
931                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
932                 disk->bus = i;
933                 disk->unit = tgt[1];
934                 break;
935             }
936         }
937         if (i >= nhosts) {
938             g_debug("no host for '%s' (driver '%s')", syspath, driver);
939             goto cleanup;
940         }
941     } else if (strcmp(driver, "sym53c8xx") == 0) {
942         /* scsi(LSI Logic): target*:0:<unit>:0 */
943         if (!has_tgt) {
944             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
945             goto cleanup;
946         }
947         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
948         disk->unit = tgt[1];
949     } else if (strcmp(driver, "virtio-pci") == 0) {
950         if (has_tgt) {
951             /* virtio-scsi: target*:0:0:<unit> */
952             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
953             disk->unit = tgt[2];
954         } else {
955             /* virtio-blk: 1 disk per 1 device */
956             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
957         }
958     } else if (strcmp(driver, "ahci") == 0) {
959         /* ahci: 1 host per 1 unit */
960         if (!has_host || !has_tgt) {
961             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
962             goto cleanup;
963         }
964         for (i = 0; i < nhosts; i++) {
965             if (host == hosts[i]) {
966                 disk->unit = i;
967                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
968                 break;
969             }
970         }
971         if (i >= nhosts) {
972             g_debug("no host for '%s' (driver '%s')", syspath, driver);
973             goto cleanup;
974         }
975     } else {
976         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
977         goto cleanup;
978     }
979 
980     list->next = fs->disk;
981     fs->disk = list;
982     g_free(driver);
983     return;
984 
985 cleanup:
986     if (list) {
987         qapi_free_GuestDiskAddressList(list);
988     }
989     g_free(driver);
990 }
991 
992 static void build_guest_fsinfo_for_device(char const *devpath,
993                                           GuestFilesystemInfo *fs,
994                                           Error **errp);
995 
996 /* Store a list of slave devices of virtual volume specified by @syspath into
997  * @fs */
998 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
999                                                   GuestFilesystemInfo *fs,
1000                                                   Error **errp)
1001 {
1002     DIR *dir;
1003     char *dirpath;
1004     struct dirent *entry;
1005 
1006     dirpath = g_strdup_printf("%s/slaves", syspath);
1007     dir = opendir(dirpath);
1008     if (!dir) {
1009         if (errno != ENOENT) {
1010             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1011         }
1012         g_free(dirpath);
1013         return;
1014     }
1015 
1016     for (;;) {
1017         errno = 0;
1018         entry = readdir(dir);
1019         if (entry == NULL) {
1020             if (errno) {
1021                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1022             }
1023             break;
1024         }
1025 
1026         if (entry->d_type == DT_LNK) {
1027             char *path;
1028 
1029             g_debug(" slave device '%s'", entry->d_name);
1030             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1031             build_guest_fsinfo_for_device(path, fs, errp);
1032             g_free(path);
1033 
1034             if (*errp) {
1035                 break;
1036             }
1037         }
1038     }
1039 
1040     g_free(dirpath);
1041     closedir(dir);
1042 }
1043 
1044 /* Dispatch to functions for virtual/real device */
1045 static void build_guest_fsinfo_for_device(char const *devpath,
1046                                           GuestFilesystemInfo *fs,
1047                                           Error **errp)
1048 {
1049     char *syspath = realpath(devpath, NULL);
1050 
1051     if (!syspath) {
1052         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1053         return;
1054     }
1055 
1056     if (!fs->name) {
1057         fs->name = g_path_get_basename(syspath);
1058     }
1059 
1060     g_debug("  parse sysfs path '%s'", syspath);
1061 
1062     if (strstr(syspath, "/devices/virtual/block/")) {
1063         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1064     } else {
1065         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1066     }
1067 
1068     free(syspath);
1069 }
1070 
1071 /* Return a list of the disk device(s)' info which @mount lies on */
1072 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1073                                                Error **errp)
1074 {
1075     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1076     struct statvfs buf;
1077     unsigned long used, nonroot_total, fr_size;
1078     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1079                                     mount->devmajor, mount->devminor);
1080 
1081     fs->mountpoint = g_strdup(mount->dirname);
1082     fs->type = g_strdup(mount->devtype);
1083     build_guest_fsinfo_for_device(devpath, fs, errp);
1084 
1085     if (statvfs(fs->mountpoint, &buf) == 0) {
1086         fr_size = buf.f_frsize;
1087         used = buf.f_blocks - buf.f_bfree;
1088         nonroot_total = used + buf.f_bavail;
1089         fs->used_bytes = used * fr_size;
1090         fs->total_bytes = nonroot_total * fr_size;
1091 
1092         fs->has_total_bytes = true;
1093         fs->has_used_bytes = true;
1094     }
1095 
1096     g_free(devpath);
1097 
1098     return fs;
1099 }
1100 
1101 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1102 {
1103     FsMountList mounts;
1104     struct FsMount *mount;
1105     GuestFilesystemInfoList *new, *ret = NULL;
1106     Error *local_err = NULL;
1107 
1108     QTAILQ_INIT(&mounts);
1109     build_fs_mount_list(&mounts, &local_err);
1110     if (local_err) {
1111         error_propagate(errp, local_err);
1112         return NULL;
1113     }
1114 
1115     QTAILQ_FOREACH(mount, &mounts, next) {
1116         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1117 
1118         new = g_malloc0(sizeof(*ret));
1119         new->value = build_guest_fsinfo(mount, &local_err);
1120         new->next = ret;
1121         ret = new;
1122         if (local_err) {
1123             error_propagate(errp, local_err);
1124             qapi_free_GuestFilesystemInfoList(ret);
1125             ret = NULL;
1126             break;
1127         }
1128     }
1129 
1130     free_fs_mount_list(&mounts);
1131     return ret;
1132 }
1133 
1134 
1135 typedef enum {
1136     FSFREEZE_HOOK_THAW = 0,
1137     FSFREEZE_HOOK_FREEZE,
1138 } FsfreezeHookArg;
1139 
1140 static const char *fsfreeze_hook_arg_string[] = {
1141     "thaw",
1142     "freeze",
1143 };
1144 
1145 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1146 {
1147     int status;
1148     pid_t pid;
1149     const char *hook;
1150     const char *arg_str = fsfreeze_hook_arg_string[arg];
1151     Error *local_err = NULL;
1152 
1153     hook = ga_fsfreeze_hook(ga_state);
1154     if (!hook) {
1155         return;
1156     }
1157     if (access(hook, X_OK) != 0) {
1158         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1159         return;
1160     }
1161 
1162     slog("executing fsfreeze hook with arg '%s'", arg_str);
1163     pid = fork();
1164     if (pid == 0) {
1165         setsid();
1166         reopen_fd_to_null(0);
1167         reopen_fd_to_null(1);
1168         reopen_fd_to_null(2);
1169 
1170         execle(hook, hook, arg_str, NULL, environ);
1171         _exit(EXIT_FAILURE);
1172     } else if (pid < 0) {
1173         error_setg_errno(errp, errno, "failed to create child process");
1174         return;
1175     }
1176 
1177     ga_wait_child(pid, &status, &local_err);
1178     if (local_err) {
1179         error_propagate(errp, local_err);
1180         return;
1181     }
1182 
1183     if (!WIFEXITED(status)) {
1184         error_setg(errp, "fsfreeze hook has terminated abnormally");
1185         return;
1186     }
1187 
1188     status = WEXITSTATUS(status);
1189     if (status) {
1190         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1191         return;
1192     }
1193 }
1194 
1195 /*
1196  * Return status of freeze/thaw
1197  */
1198 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1199 {
1200     if (ga_is_frozen(ga_state)) {
1201         return GUEST_FSFREEZE_STATUS_FROZEN;
1202     }
1203 
1204     return GUEST_FSFREEZE_STATUS_THAWED;
1205 }
1206 
1207 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1208 {
1209     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1210 }
1211 
1212 /*
1213  * Walk list of mounted file systems in the guest, and freeze the ones which
1214  * are real local file systems.
1215  */
1216 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1217                                        strList *mountpoints,
1218                                        Error **errp)
1219 {
1220     int ret = 0, i = 0;
1221     strList *list;
1222     FsMountList mounts;
1223     struct FsMount *mount;
1224     Error *local_err = NULL;
1225     int fd;
1226 
1227     slog("guest-fsfreeze called");
1228 
1229     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1230     if (local_err) {
1231         error_propagate(errp, local_err);
1232         return -1;
1233     }
1234 
1235     QTAILQ_INIT(&mounts);
1236     build_fs_mount_list(&mounts, &local_err);
1237     if (local_err) {
1238         error_propagate(errp, local_err);
1239         return -1;
1240     }
1241 
1242     /* cannot risk guest agent blocking itself on a write in this state */
1243     ga_set_frozen(ga_state);
1244 
1245     QTAILQ_FOREACH_REVERSE(mount, &mounts, FsMountList, next) {
1246         /* To issue fsfreeze in the reverse order of mounts, check if the
1247          * mount is listed in the list here */
1248         if (has_mountpoints) {
1249             for (list = mountpoints; list; list = list->next) {
1250                 if (strcmp(list->value, mount->dirname) == 0) {
1251                     break;
1252                 }
1253             }
1254             if (!list) {
1255                 continue;
1256             }
1257         }
1258 
1259         fd = qemu_open(mount->dirname, O_RDONLY);
1260         if (fd == -1) {
1261             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1262             goto error;
1263         }
1264 
1265         /* we try to cull filesystems we know won't work in advance, but other
1266          * filesystems may not implement fsfreeze for less obvious reasons.
1267          * these will report EOPNOTSUPP. we simply ignore these when tallying
1268          * the number of frozen filesystems.
1269          * if a filesystem is mounted more than once (aka bind mount) a
1270          * consecutive attempt to freeze an already frozen filesystem will
1271          * return EBUSY.
1272          *
1273          * any other error means a failure to freeze a filesystem we
1274          * expect to be freezable, so return an error in those cases
1275          * and return system to thawed state.
1276          */
1277         ret = ioctl(fd, FIFREEZE);
1278         if (ret == -1) {
1279             if (errno != EOPNOTSUPP && errno != EBUSY) {
1280                 error_setg_errno(errp, errno, "failed to freeze %s",
1281                                  mount->dirname);
1282                 close(fd);
1283                 goto error;
1284             }
1285         } else {
1286             i++;
1287         }
1288         close(fd);
1289     }
1290 
1291     free_fs_mount_list(&mounts);
1292     /* We may not issue any FIFREEZE here.
1293      * Just unset ga_state here and ready for the next call.
1294      */
1295     if (i == 0) {
1296         ga_unset_frozen(ga_state);
1297     }
1298     return i;
1299 
1300 error:
1301     free_fs_mount_list(&mounts);
1302     qmp_guest_fsfreeze_thaw(NULL);
1303     return 0;
1304 }
1305 
1306 /*
1307  * Walk list of frozen file systems in the guest, and thaw them.
1308  */
1309 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1310 {
1311     int ret;
1312     FsMountList mounts;
1313     FsMount *mount;
1314     int fd, i = 0, logged;
1315     Error *local_err = NULL;
1316 
1317     QTAILQ_INIT(&mounts);
1318     build_fs_mount_list(&mounts, &local_err);
1319     if (local_err) {
1320         error_propagate(errp, local_err);
1321         return 0;
1322     }
1323 
1324     QTAILQ_FOREACH(mount, &mounts, next) {
1325         logged = false;
1326         fd = qemu_open(mount->dirname, O_RDONLY);
1327         if (fd == -1) {
1328             continue;
1329         }
1330         /* we have no way of knowing whether a filesystem was actually unfrozen
1331          * as a result of a successful call to FITHAW, only that if an error
1332          * was returned the filesystem was *not* unfrozen by that particular
1333          * call.
1334          *
1335          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1336          * to unfreeze, continuing issuing FITHAW until an error is returned,
1337          * in which case either the filesystem is in an unfreezable state, or,
1338          * more likely, it was thawed previously (and remains so afterward).
1339          *
1340          * also, since the most recent successful call is the one that did
1341          * the actual unfreeze, we can use this to provide an accurate count
1342          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1343          * may * be useful for determining whether a filesystem was unfrozen
1344          * during the freeze/thaw phase by a process other than qemu-ga.
1345          */
1346         do {
1347             ret = ioctl(fd, FITHAW);
1348             if (ret == 0 && !logged) {
1349                 i++;
1350                 logged = true;
1351             }
1352         } while (ret == 0);
1353         close(fd);
1354     }
1355 
1356     ga_unset_frozen(ga_state);
1357     free_fs_mount_list(&mounts);
1358 
1359     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1360 
1361     return i;
1362 }
1363 
1364 static void guest_fsfreeze_cleanup(void)
1365 {
1366     Error *err = NULL;
1367 
1368     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1369         qmp_guest_fsfreeze_thaw(&err);
1370         if (err) {
1371             slog("failed to clean up frozen filesystems: %s",
1372                  error_get_pretty(err));
1373             error_free(err);
1374         }
1375     }
1376 }
1377 #endif /* CONFIG_FSFREEZE */
1378 
1379 #if defined(CONFIG_FSTRIM)
1380 /*
1381  * Walk list of mounted file systems in the guest, and trim them.
1382  */
1383 GuestFilesystemTrimResponse *
1384 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1385 {
1386     GuestFilesystemTrimResponse *response;
1387     GuestFilesystemTrimResultList *list;
1388     GuestFilesystemTrimResult *result;
1389     int ret = 0;
1390     FsMountList mounts;
1391     struct FsMount *mount;
1392     int fd;
1393     Error *local_err = NULL;
1394     struct fstrim_range r;
1395 
1396     slog("guest-fstrim called");
1397 
1398     QTAILQ_INIT(&mounts);
1399     build_fs_mount_list(&mounts, &local_err);
1400     if (local_err) {
1401         error_propagate(errp, local_err);
1402         return NULL;
1403     }
1404 
1405     response = g_malloc0(sizeof(*response));
1406 
1407     QTAILQ_FOREACH(mount, &mounts, next) {
1408         result = g_malloc0(sizeof(*result));
1409         result->path = g_strdup(mount->dirname);
1410 
1411         list = g_malloc0(sizeof(*list));
1412         list->value = result;
1413         list->next = response->paths;
1414         response->paths = list;
1415 
1416         fd = qemu_open(mount->dirname, O_RDONLY);
1417         if (fd == -1) {
1418             result->error = g_strdup_printf("failed to open: %s",
1419                                             strerror(errno));
1420             result->has_error = true;
1421             continue;
1422         }
1423 
1424         /* We try to cull filesystems we know won't work in advance, but other
1425          * filesystems may not implement fstrim for less obvious reasons.
1426          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1427          * will be reported (e.g. CD-ROMs).
1428          * Any other error means an unexpected error.
1429          */
1430         r.start = 0;
1431         r.len = -1;
1432         r.minlen = has_minimum ? minimum : 0;
1433         ret = ioctl(fd, FITRIM, &r);
1434         if (ret == -1) {
1435             result->has_error = true;
1436             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1437                 result->error = g_strdup("trim not supported");
1438             } else {
1439                 result->error = g_strdup_printf("failed to trim: %s",
1440                                                 strerror(errno));
1441             }
1442             close(fd);
1443             continue;
1444         }
1445 
1446         result->has_minimum = true;
1447         result->minimum = r.minlen;
1448         result->has_trimmed = true;
1449         result->trimmed = r.len;
1450         close(fd);
1451     }
1452 
1453     free_fs_mount_list(&mounts);
1454     return response;
1455 }
1456 #endif /* CONFIG_FSTRIM */
1457 
1458 
1459 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1460 #define SUSPEND_SUPPORTED 0
1461 #define SUSPEND_NOT_SUPPORTED 1
1462 
1463 static void bios_supports_mode(const char *pmutils_bin, const char *pmutils_arg,
1464                                const char *sysfile_str, Error **errp)
1465 {
1466     Error *local_err = NULL;
1467     char *pmutils_path;
1468     pid_t pid;
1469     int status;
1470 
1471     pmutils_path = g_find_program_in_path(pmutils_bin);
1472 
1473     pid = fork();
1474     if (!pid) {
1475         char buf[32]; /* hopefully big enough */
1476         ssize_t ret;
1477         int fd;
1478 
1479         setsid();
1480         reopen_fd_to_null(0);
1481         reopen_fd_to_null(1);
1482         reopen_fd_to_null(2);
1483 
1484         if (pmutils_path) {
1485             execle(pmutils_path, pmutils_bin, pmutils_arg, NULL, environ);
1486         }
1487 
1488         /*
1489          * If we get here either pm-utils is not installed or execle() has
1490          * failed. Let's try the manual method if the caller wants it.
1491          */
1492 
1493         if (!sysfile_str) {
1494             _exit(SUSPEND_NOT_SUPPORTED);
1495         }
1496 
1497         fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
1498         if (fd < 0) {
1499             _exit(SUSPEND_NOT_SUPPORTED);
1500         }
1501 
1502         ret = read(fd, buf, sizeof(buf)-1);
1503         if (ret <= 0) {
1504             _exit(SUSPEND_NOT_SUPPORTED);
1505         }
1506         buf[ret] = '\0';
1507 
1508         if (strstr(buf, sysfile_str)) {
1509             _exit(SUSPEND_SUPPORTED);
1510         }
1511 
1512         _exit(SUSPEND_NOT_SUPPORTED);
1513     } else if (pid < 0) {
1514         error_setg_errno(errp, errno, "failed to create child process");
1515         goto out;
1516     }
1517 
1518     ga_wait_child(pid, &status, &local_err);
1519     if (local_err) {
1520         error_propagate(errp, local_err);
1521         goto out;
1522     }
1523 
1524     if (!WIFEXITED(status)) {
1525         error_setg(errp, "child process has terminated abnormally");
1526         goto out;
1527     }
1528 
1529     switch (WEXITSTATUS(status)) {
1530     case SUSPEND_SUPPORTED:
1531         goto out;
1532     case SUSPEND_NOT_SUPPORTED:
1533         error_setg(errp,
1534                    "the requested suspend mode is not supported by the guest");
1535         goto out;
1536     default:
1537         error_setg(errp,
1538                    "the helper program '%s' returned an unexpected exit status"
1539                    " code (%d)", pmutils_path, WEXITSTATUS(status));
1540         goto out;
1541     }
1542 
1543 out:
1544     g_free(pmutils_path);
1545 }
1546 
1547 static void guest_suspend(const char *pmutils_bin, const char *sysfile_str,
1548                           Error **errp)
1549 {
1550     Error *local_err = NULL;
1551     char *pmutils_path;
1552     pid_t pid;
1553     int status;
1554 
1555     pmutils_path = g_find_program_in_path(pmutils_bin);
1556 
1557     pid = fork();
1558     if (pid == 0) {
1559         /* child */
1560         int fd;
1561 
1562         setsid();
1563         reopen_fd_to_null(0);
1564         reopen_fd_to_null(1);
1565         reopen_fd_to_null(2);
1566 
1567         if (pmutils_path) {
1568             execle(pmutils_path, pmutils_bin, NULL, environ);
1569         }
1570 
1571         /*
1572          * If we get here either pm-utils is not installed or execle() has
1573          * failed. Let's try the manual method if the caller wants it.
1574          */
1575 
1576         if (!sysfile_str) {
1577             _exit(EXIT_FAILURE);
1578         }
1579 
1580         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1581         if (fd < 0) {
1582             _exit(EXIT_FAILURE);
1583         }
1584 
1585         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1586             _exit(EXIT_FAILURE);
1587         }
1588 
1589         _exit(EXIT_SUCCESS);
1590     } else if (pid < 0) {
1591         error_setg_errno(errp, errno, "failed to create child process");
1592         goto out;
1593     }
1594 
1595     ga_wait_child(pid, &status, &local_err);
1596     if (local_err) {
1597         error_propagate(errp, local_err);
1598         goto out;
1599     }
1600 
1601     if (!WIFEXITED(status)) {
1602         error_setg(errp, "child process has terminated abnormally");
1603         goto out;
1604     }
1605 
1606     if (WEXITSTATUS(status)) {
1607         error_setg(errp, "child process has failed to suspend");
1608         goto out;
1609     }
1610 
1611 out:
1612     g_free(pmutils_path);
1613 }
1614 
1615 void qmp_guest_suspend_disk(Error **errp)
1616 {
1617     Error *local_err = NULL;
1618 
1619     bios_supports_mode("pm-is-supported", "--hibernate", "disk", &local_err);
1620     if (local_err) {
1621         error_propagate(errp, local_err);
1622         return;
1623     }
1624 
1625     guest_suspend("pm-hibernate", "disk", errp);
1626 }
1627 
1628 void qmp_guest_suspend_ram(Error **errp)
1629 {
1630     Error *local_err = NULL;
1631 
1632     bios_supports_mode("pm-is-supported", "--suspend", "mem", &local_err);
1633     if (local_err) {
1634         error_propagate(errp, local_err);
1635         return;
1636     }
1637 
1638     guest_suspend("pm-suspend", "mem", errp);
1639 }
1640 
1641 void qmp_guest_suspend_hybrid(Error **errp)
1642 {
1643     Error *local_err = NULL;
1644 
1645     bios_supports_mode("pm-is-supported", "--suspend-hybrid", NULL,
1646                        &local_err);
1647     if (local_err) {
1648         error_propagate(errp, local_err);
1649         return;
1650     }
1651 
1652     guest_suspend("pm-suspend-hybrid", NULL, errp);
1653 }
1654 
1655 static GuestNetworkInterfaceList *
1656 guest_find_interface(GuestNetworkInterfaceList *head,
1657                      const char *name)
1658 {
1659     for (; head; head = head->next) {
1660         if (strcmp(head->value->name, name) == 0) {
1661             break;
1662         }
1663     }
1664 
1665     return head;
1666 }
1667 
1668 static int guest_get_network_stats(const char *name,
1669                        GuestNetworkInterfaceStat *stats)
1670 {
1671     int name_len;
1672     char const *devinfo = "/proc/net/dev";
1673     FILE *fp;
1674     char *line = NULL, *colon;
1675     size_t n = 0;
1676     fp = fopen(devinfo, "r");
1677     if (!fp) {
1678         return -1;
1679     }
1680     name_len = strlen(name);
1681     while (getline(&line, &n, fp) != -1) {
1682         long long dummy;
1683         long long rx_bytes;
1684         long long rx_packets;
1685         long long rx_errs;
1686         long long rx_dropped;
1687         long long tx_bytes;
1688         long long tx_packets;
1689         long long tx_errs;
1690         long long tx_dropped;
1691         char *trim_line;
1692         trim_line = g_strchug(line);
1693         if (trim_line[0] == '\0') {
1694             continue;
1695         }
1696         colon = strchr(trim_line, ':');
1697         if (!colon) {
1698             continue;
1699         }
1700         if (colon - name_len  == trim_line &&
1701            strncmp(trim_line, name, name_len) == 0) {
1702             if (sscanf(colon + 1,
1703                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
1704                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
1705                   &dummy, &dummy, &dummy, &dummy,
1706                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
1707                   &dummy, &dummy, &dummy, &dummy) != 16) {
1708                 continue;
1709             }
1710             stats->rx_bytes = rx_bytes;
1711             stats->rx_packets = rx_packets;
1712             stats->rx_errs = rx_errs;
1713             stats->rx_dropped = rx_dropped;
1714             stats->tx_bytes = tx_bytes;
1715             stats->tx_packets = tx_packets;
1716             stats->tx_errs = tx_errs;
1717             stats->tx_dropped = tx_dropped;
1718             fclose(fp);
1719             g_free(line);
1720             return 0;
1721         }
1722     }
1723     fclose(fp);
1724     g_free(line);
1725     g_debug("/proc/net/dev: Interface '%s' not found", name);
1726     return -1;
1727 }
1728 
1729 /*
1730  * Build information about guest interfaces
1731  */
1732 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
1733 {
1734     GuestNetworkInterfaceList *head = NULL, *cur_item = NULL;
1735     struct ifaddrs *ifap, *ifa;
1736 
1737     if (getifaddrs(&ifap) < 0) {
1738         error_setg_errno(errp, errno, "getifaddrs failed");
1739         goto error;
1740     }
1741 
1742     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
1743         GuestNetworkInterfaceList *info;
1744         GuestIpAddressList **address_list = NULL, *address_item = NULL;
1745         GuestNetworkInterfaceStat  *interface_stat = NULL;
1746         char addr4[INET_ADDRSTRLEN];
1747         char addr6[INET6_ADDRSTRLEN];
1748         int sock;
1749         struct ifreq ifr;
1750         unsigned char *mac_addr;
1751         void *p;
1752 
1753         g_debug("Processing %s interface", ifa->ifa_name);
1754 
1755         info = guest_find_interface(head, ifa->ifa_name);
1756 
1757         if (!info) {
1758             info = g_malloc0(sizeof(*info));
1759             info->value = g_malloc0(sizeof(*info->value));
1760             info->value->name = g_strdup(ifa->ifa_name);
1761 
1762             if (!cur_item) {
1763                 head = cur_item = info;
1764             } else {
1765                 cur_item->next = info;
1766                 cur_item = info;
1767             }
1768         }
1769 
1770         if (!info->value->has_hardware_address &&
1771             ifa->ifa_flags & SIOCGIFHWADDR) {
1772             /* we haven't obtained HW address yet */
1773             sock = socket(PF_INET, SOCK_STREAM, 0);
1774             if (sock == -1) {
1775                 error_setg_errno(errp, errno, "failed to create socket");
1776                 goto error;
1777             }
1778 
1779             memset(&ifr, 0, sizeof(ifr));
1780             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->value->name);
1781             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
1782                 error_setg_errno(errp, errno,
1783                                  "failed to get MAC address of %s",
1784                                  ifa->ifa_name);
1785                 close(sock);
1786                 goto error;
1787             }
1788 
1789             close(sock);
1790             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
1791 
1792             info->value->hardware_address =
1793                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
1794                                 (int) mac_addr[0], (int) mac_addr[1],
1795                                 (int) mac_addr[2], (int) mac_addr[3],
1796                                 (int) mac_addr[4], (int) mac_addr[5]);
1797 
1798             info->value->has_hardware_address = true;
1799         }
1800 
1801         if (ifa->ifa_addr &&
1802             ifa->ifa_addr->sa_family == AF_INET) {
1803             /* interface with IPv4 address */
1804             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1805             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
1806                 error_setg_errno(errp, errno, "inet_ntop failed");
1807                 goto error;
1808             }
1809 
1810             address_item = g_malloc0(sizeof(*address_item));
1811             address_item->value = g_malloc0(sizeof(*address_item->value));
1812             address_item->value->ip_address = g_strdup(addr4);
1813             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
1814 
1815             if (ifa->ifa_netmask) {
1816                 /* Count the number of set bits in netmask.
1817                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1818                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
1819                 address_item->value->prefix = ctpop32(((uint32_t *) p)[0]);
1820             }
1821         } else if (ifa->ifa_addr &&
1822                    ifa->ifa_addr->sa_family == AF_INET6) {
1823             /* interface with IPv6 address */
1824             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
1825             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
1826                 error_setg_errno(errp, errno, "inet_ntop failed");
1827                 goto error;
1828             }
1829 
1830             address_item = g_malloc0(sizeof(*address_item));
1831             address_item->value = g_malloc0(sizeof(*address_item->value));
1832             address_item->value->ip_address = g_strdup(addr6);
1833             address_item->value->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
1834 
1835             if (ifa->ifa_netmask) {
1836                 /* Count the number of set bits in netmask.
1837                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
1838                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
1839                 address_item->value->prefix =
1840                     ctpop32(((uint32_t *) p)[0]) +
1841                     ctpop32(((uint32_t *) p)[1]) +
1842                     ctpop32(((uint32_t *) p)[2]) +
1843                     ctpop32(((uint32_t *) p)[3]);
1844             }
1845         }
1846 
1847         if (!address_item) {
1848             continue;
1849         }
1850 
1851         address_list = &info->value->ip_addresses;
1852 
1853         while (*address_list && (*address_list)->next) {
1854             address_list = &(*address_list)->next;
1855         }
1856 
1857         if (!*address_list) {
1858             *address_list = address_item;
1859         } else {
1860             (*address_list)->next = address_item;
1861         }
1862 
1863         info->value->has_ip_addresses = true;
1864 
1865         if (!info->value->has_statistics) {
1866             interface_stat = g_malloc0(sizeof(*interface_stat));
1867             if (guest_get_network_stats(info->value->name,
1868                 interface_stat) == -1) {
1869                 info->value->has_statistics = false;
1870                 g_free(interface_stat);
1871             } else {
1872                 info->value->statistics = interface_stat;
1873                 info->value->has_statistics = true;
1874             }
1875         }
1876     }
1877 
1878     freeifaddrs(ifap);
1879     return head;
1880 
1881 error:
1882     freeifaddrs(ifap);
1883     qapi_free_GuestNetworkInterfaceList(head);
1884     return NULL;
1885 }
1886 
1887 #define SYSCONF_EXACT(name, errp) sysconf_exact((name), #name, (errp))
1888 
1889 static long sysconf_exact(int name, const char *name_str, Error **errp)
1890 {
1891     long ret;
1892 
1893     errno = 0;
1894     ret = sysconf(name);
1895     if (ret == -1) {
1896         if (errno == 0) {
1897             error_setg(errp, "sysconf(%s): value indefinite", name_str);
1898         } else {
1899             error_setg_errno(errp, errno, "sysconf(%s)", name_str);
1900         }
1901     }
1902     return ret;
1903 }
1904 
1905 /* Transfer online/offline status between @vcpu and the guest system.
1906  *
1907  * On input either @errp or *@errp must be NULL.
1908  *
1909  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
1910  * - R: vcpu->logical_id
1911  * - W: vcpu->online
1912  * - W: vcpu->can_offline
1913  *
1914  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
1915  * - R: vcpu->logical_id
1916  * - R: vcpu->online
1917  *
1918  * Written members remain unmodified on error.
1919  */
1920 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
1921                           Error **errp)
1922 {
1923     char *dirpath;
1924     int dirfd;
1925 
1926     dirpath = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
1927                               vcpu->logical_id);
1928     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
1929     if (dirfd == -1) {
1930         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
1931     } else {
1932         static const char fn[] = "online";
1933         int fd;
1934         int res;
1935 
1936         fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
1937         if (fd == -1) {
1938             if (errno != ENOENT) {
1939                 error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
1940             } else if (sys2vcpu) {
1941                 vcpu->online = true;
1942                 vcpu->can_offline = false;
1943             } else if (!vcpu->online) {
1944                 error_setg(errp, "logical processor #%" PRId64 " can't be "
1945                            "offlined", vcpu->logical_id);
1946             } /* otherwise pretend successful re-onlining */
1947         } else {
1948             unsigned char status;
1949 
1950             res = pread(fd, &status, 1, 0);
1951             if (res == -1) {
1952                 error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
1953             } else if (res == 0) {
1954                 error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
1955                            fn);
1956             } else if (sys2vcpu) {
1957                 vcpu->online = (status != '0');
1958                 vcpu->can_offline = true;
1959             } else if (vcpu->online != (status != '0')) {
1960                 status = '0' + vcpu->online;
1961                 if (pwrite(fd, &status, 1, 0) == -1) {
1962                     error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
1963                                      fn);
1964                 }
1965             } /* otherwise pretend successful re-(on|off)-lining */
1966 
1967             res = close(fd);
1968             g_assert(res == 0);
1969         }
1970 
1971         res = close(dirfd);
1972         g_assert(res == 0);
1973     }
1974 
1975     g_free(dirpath);
1976 }
1977 
1978 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
1979 {
1980     int64_t current;
1981     GuestLogicalProcessorList *head, **link;
1982     long sc_max;
1983     Error *local_err = NULL;
1984 
1985     current = 0;
1986     head = NULL;
1987     link = &head;
1988     sc_max = SYSCONF_EXACT(_SC_NPROCESSORS_CONF, &local_err);
1989 
1990     while (local_err == NULL && current < sc_max) {
1991         GuestLogicalProcessor *vcpu;
1992         GuestLogicalProcessorList *entry;
1993 
1994         vcpu = g_malloc0(sizeof *vcpu);
1995         vcpu->logical_id = current++;
1996         vcpu->has_can_offline = true; /* lolspeak ftw */
1997         transfer_vcpu(vcpu, true, &local_err);
1998 
1999         entry = g_malloc0(sizeof *entry);
2000         entry->value = vcpu;
2001 
2002         *link = entry;
2003         link = &entry->next;
2004     }
2005 
2006     if (local_err == NULL) {
2007         /* there's no guest with zero VCPUs */
2008         g_assert(head != NULL);
2009         return head;
2010     }
2011 
2012     qapi_free_GuestLogicalProcessorList(head);
2013     error_propagate(errp, local_err);
2014     return NULL;
2015 }
2016 
2017 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2018 {
2019     int64_t processed;
2020     Error *local_err = NULL;
2021 
2022     processed = 0;
2023     while (vcpus != NULL) {
2024         transfer_vcpu(vcpus->value, false, &local_err);
2025         if (local_err != NULL) {
2026             break;
2027         }
2028         ++processed;
2029         vcpus = vcpus->next;
2030     }
2031 
2032     if (local_err != NULL) {
2033         if (processed == 0) {
2034             error_propagate(errp, local_err);
2035         } else {
2036             error_free(local_err);
2037         }
2038     }
2039 
2040     return processed;
2041 }
2042 
2043 void qmp_guest_set_user_password(const char *username,
2044                                  const char *password,
2045                                  bool crypted,
2046                                  Error **errp)
2047 {
2048     Error *local_err = NULL;
2049     char *passwd_path = NULL;
2050     pid_t pid;
2051     int status;
2052     int datafd[2] = { -1, -1 };
2053     char *rawpasswddata = NULL;
2054     size_t rawpasswdlen;
2055     char *chpasswddata = NULL;
2056     size_t chpasswdlen;
2057 
2058     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2059     if (!rawpasswddata) {
2060         return;
2061     }
2062     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2063     rawpasswddata[rawpasswdlen] = '\0';
2064 
2065     if (strchr(rawpasswddata, '\n')) {
2066         error_setg(errp, "forbidden characters in raw password");
2067         goto out;
2068     }
2069 
2070     if (strchr(username, '\n') ||
2071         strchr(username, ':')) {
2072         error_setg(errp, "forbidden characters in username");
2073         goto out;
2074     }
2075 
2076     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2077     chpasswdlen = strlen(chpasswddata);
2078 
2079     passwd_path = g_find_program_in_path("chpasswd");
2080 
2081     if (!passwd_path) {
2082         error_setg(errp, "cannot find 'passwd' program in PATH");
2083         goto out;
2084     }
2085 
2086     if (pipe(datafd) < 0) {
2087         error_setg(errp, "cannot create pipe FDs");
2088         goto out;
2089     }
2090 
2091     pid = fork();
2092     if (pid == 0) {
2093         close(datafd[1]);
2094         /* child */
2095         setsid();
2096         dup2(datafd[0], 0);
2097         reopen_fd_to_null(1);
2098         reopen_fd_to_null(2);
2099 
2100         if (crypted) {
2101             execle(passwd_path, "chpasswd", "-e", NULL, environ);
2102         } else {
2103             execle(passwd_path, "chpasswd", NULL, environ);
2104         }
2105         _exit(EXIT_FAILURE);
2106     } else if (pid < 0) {
2107         error_setg_errno(errp, errno, "failed to create child process");
2108         goto out;
2109     }
2110     close(datafd[0]);
2111     datafd[0] = -1;
2112 
2113     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2114         error_setg_errno(errp, errno, "cannot write new account password");
2115         goto out;
2116     }
2117     close(datafd[1]);
2118     datafd[1] = -1;
2119 
2120     ga_wait_child(pid, &status, &local_err);
2121     if (local_err) {
2122         error_propagate(errp, local_err);
2123         goto out;
2124     }
2125 
2126     if (!WIFEXITED(status)) {
2127         error_setg(errp, "child process has terminated abnormally");
2128         goto out;
2129     }
2130 
2131     if (WEXITSTATUS(status)) {
2132         error_setg(errp, "child process has failed to set user password");
2133         goto out;
2134     }
2135 
2136 out:
2137     g_free(chpasswddata);
2138     g_free(rawpasswddata);
2139     g_free(passwd_path);
2140     if (datafd[0] != -1) {
2141         close(datafd[0]);
2142     }
2143     if (datafd[1] != -1) {
2144         close(datafd[1]);
2145     }
2146 }
2147 
2148 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2149                                int size, Error **errp)
2150 {
2151     int fd;
2152     int res;
2153 
2154     errno = 0;
2155     fd = openat(dirfd, pathname, O_RDONLY);
2156     if (fd == -1) {
2157         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2158         return;
2159     }
2160 
2161     res = pread(fd, buf, size, 0);
2162     if (res == -1) {
2163         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2164     } else if (res == 0) {
2165         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2166     }
2167     close(fd);
2168 }
2169 
2170 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2171                                 const char *buf, int size, Error **errp)
2172 {
2173     int fd;
2174 
2175     errno = 0;
2176     fd = openat(dirfd, pathname, O_WRONLY);
2177     if (fd == -1) {
2178         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2179         return;
2180     }
2181 
2182     if (pwrite(fd, buf, size, 0) == -1) {
2183         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2184     }
2185 
2186     close(fd);
2187 }
2188 
2189 /* Transfer online/offline status between @mem_blk and the guest system.
2190  *
2191  * On input either @errp or *@errp must be NULL.
2192  *
2193  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2194  * - R: mem_blk->phys_index
2195  * - W: mem_blk->online
2196  * - W: mem_blk->can_offline
2197  *
2198  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2199  * - R: mem_blk->phys_index
2200  * - R: mem_blk->online
2201  *-  R: mem_blk->can_offline
2202  * Written members remain unmodified on error.
2203  */
2204 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2205                                   GuestMemoryBlockResponse *result,
2206                                   Error **errp)
2207 {
2208     char *dirpath;
2209     int dirfd;
2210     char *status;
2211     Error *local_err = NULL;
2212 
2213     if (!sys2memblk) {
2214         DIR *dp;
2215 
2216         if (!result) {
2217             error_setg(errp, "Internal error, 'result' should not be NULL");
2218             return;
2219         }
2220         errno = 0;
2221         dp = opendir("/sys/devices/system/memory/");
2222          /* if there is no 'memory' directory in sysfs,
2223          * we think this VM does not support online/offline memory block,
2224          * any other solution?
2225          */
2226         if (!dp) {
2227             if (errno == ENOENT) {
2228                 result->response =
2229                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2230             }
2231             goto out1;
2232         }
2233         closedir(dp);
2234     }
2235 
2236     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2237                               mem_blk->phys_index);
2238     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2239     if (dirfd == -1) {
2240         if (sys2memblk) {
2241             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2242         } else {
2243             if (errno == ENOENT) {
2244                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2245             } else {
2246                 result->response =
2247                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2248             }
2249         }
2250         g_free(dirpath);
2251         goto out1;
2252     }
2253     g_free(dirpath);
2254 
2255     status = g_malloc0(10);
2256     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2257     if (local_err) {
2258         /* treat with sysfs file that not exist in old kernel */
2259         if (errno == ENOENT) {
2260             error_free(local_err);
2261             if (sys2memblk) {
2262                 mem_blk->online = true;
2263                 mem_blk->can_offline = false;
2264             } else if (!mem_blk->online) {
2265                 result->response =
2266                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2267             }
2268         } else {
2269             if (sys2memblk) {
2270                 error_propagate(errp, local_err);
2271             } else {
2272                 result->response =
2273                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2274             }
2275         }
2276         goto out2;
2277     }
2278 
2279     if (sys2memblk) {
2280         char removable = '0';
2281 
2282         mem_blk->online = (strncmp(status, "online", 6) == 0);
2283 
2284         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2285         if (local_err) {
2286             /* if no 'removable' file, it doesn't support offline mem blk */
2287             if (errno == ENOENT) {
2288                 error_free(local_err);
2289                 mem_blk->can_offline = false;
2290             } else {
2291                 error_propagate(errp, local_err);
2292             }
2293         } else {
2294             mem_blk->can_offline = (removable != '0');
2295         }
2296     } else {
2297         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2298             const char *new_state = mem_blk->online ? "online" : "offline";
2299 
2300             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2301                                 &local_err);
2302             if (local_err) {
2303                 error_free(local_err);
2304                 result->response =
2305                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2306                 goto out2;
2307             }
2308 
2309             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2310             result->has_error_code = false;
2311         } /* otherwise pretend successful re-(on|off)-lining */
2312     }
2313     g_free(status);
2314     close(dirfd);
2315     return;
2316 
2317 out2:
2318     g_free(status);
2319     close(dirfd);
2320 out1:
2321     if (!sys2memblk) {
2322         result->has_error_code = true;
2323         result->error_code = errno;
2324     }
2325 }
2326 
2327 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2328 {
2329     GuestMemoryBlockList *head, **link;
2330     Error *local_err = NULL;
2331     struct dirent *de;
2332     DIR *dp;
2333 
2334     head = NULL;
2335     link = &head;
2336 
2337     dp = opendir("/sys/devices/system/memory/");
2338     if (!dp) {
2339         /* it's ok if this happens to be a system that doesn't expose
2340          * memory blocks via sysfs, but otherwise we should report
2341          * an error
2342          */
2343         if (errno != ENOENT) {
2344             error_setg_errno(errp, errno, "Can't open directory"
2345                              "\"/sys/devices/system/memory/\"");
2346         }
2347         return NULL;
2348     }
2349 
2350     /* Note: the phys_index of memory block may be discontinuous,
2351      * this is because a memblk is the unit of the Sparse Memory design, which
2352      * allows discontinuous memory ranges (ex. NUMA), so here we should
2353      * traverse the memory block directory.
2354      */
2355     while ((de = readdir(dp)) != NULL) {
2356         GuestMemoryBlock *mem_blk;
2357         GuestMemoryBlockList *entry;
2358 
2359         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2360             !(de->d_type & DT_DIR)) {
2361             continue;
2362         }
2363 
2364         mem_blk = g_malloc0(sizeof *mem_blk);
2365         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2366         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2367         mem_blk->has_can_offline = true; /* lolspeak ftw */
2368         transfer_memory_block(mem_blk, true, NULL, &local_err);
2369 
2370         entry = g_malloc0(sizeof *entry);
2371         entry->value = mem_blk;
2372 
2373         *link = entry;
2374         link = &entry->next;
2375     }
2376 
2377     closedir(dp);
2378     if (local_err == NULL) {
2379         /* there's no guest with zero memory blocks */
2380         if (head == NULL) {
2381             error_setg(errp, "guest reported zero memory blocks!");
2382         }
2383         return head;
2384     }
2385 
2386     qapi_free_GuestMemoryBlockList(head);
2387     error_propagate(errp, local_err);
2388     return NULL;
2389 }
2390 
2391 GuestMemoryBlockResponseList *
2392 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2393 {
2394     GuestMemoryBlockResponseList *head, **link;
2395     Error *local_err = NULL;
2396 
2397     head = NULL;
2398     link = &head;
2399 
2400     while (mem_blks != NULL) {
2401         GuestMemoryBlockResponse *result;
2402         GuestMemoryBlockResponseList *entry;
2403         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2404 
2405         result = g_malloc0(sizeof(*result));
2406         result->phys_index = current_mem_blk->phys_index;
2407         transfer_memory_block(current_mem_blk, false, result, &local_err);
2408         if (local_err) { /* should never happen */
2409             goto err;
2410         }
2411         entry = g_malloc0(sizeof *entry);
2412         entry->value = result;
2413 
2414         *link = entry;
2415         link = &entry->next;
2416         mem_blks = mem_blks->next;
2417     }
2418 
2419     return head;
2420 err:
2421     qapi_free_GuestMemoryBlockResponseList(head);
2422     error_propagate(errp, local_err);
2423     return NULL;
2424 }
2425 
2426 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2427 {
2428     Error *local_err = NULL;
2429     char *dirpath;
2430     int dirfd;
2431     char *buf;
2432     GuestMemoryBlockInfo *info;
2433 
2434     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2435     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2436     if (dirfd == -1) {
2437         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2438         g_free(dirpath);
2439         return NULL;
2440     }
2441     g_free(dirpath);
2442 
2443     buf = g_malloc0(20);
2444     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2445     close(dirfd);
2446     if (local_err) {
2447         g_free(buf);
2448         error_propagate(errp, local_err);
2449         return NULL;
2450     }
2451 
2452     info = g_new0(GuestMemoryBlockInfo, 1);
2453     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2454 
2455     g_free(buf);
2456 
2457     return info;
2458 }
2459 
2460 #else /* defined(__linux__) */
2461 
2462 void qmp_guest_suspend_disk(Error **errp)
2463 {
2464     error_setg(errp, QERR_UNSUPPORTED);
2465 }
2466 
2467 void qmp_guest_suspend_ram(Error **errp)
2468 {
2469     error_setg(errp, QERR_UNSUPPORTED);
2470 }
2471 
2472 void qmp_guest_suspend_hybrid(Error **errp)
2473 {
2474     error_setg(errp, QERR_UNSUPPORTED);
2475 }
2476 
2477 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2478 {
2479     error_setg(errp, QERR_UNSUPPORTED);
2480     return NULL;
2481 }
2482 
2483 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2484 {
2485     error_setg(errp, QERR_UNSUPPORTED);
2486     return NULL;
2487 }
2488 
2489 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2490 {
2491     error_setg(errp, QERR_UNSUPPORTED);
2492     return -1;
2493 }
2494 
2495 void qmp_guest_set_user_password(const char *username,
2496                                  const char *password,
2497                                  bool crypted,
2498                                  Error **errp)
2499 {
2500     error_setg(errp, QERR_UNSUPPORTED);
2501 }
2502 
2503 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2504 {
2505     error_setg(errp, QERR_UNSUPPORTED);
2506     return NULL;
2507 }
2508 
2509 GuestMemoryBlockResponseList *
2510 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2511 {
2512     error_setg(errp, QERR_UNSUPPORTED);
2513     return NULL;
2514 }
2515 
2516 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2517 {
2518     error_setg(errp, QERR_UNSUPPORTED);
2519     return NULL;
2520 }
2521 
2522 #endif
2523 
2524 #if !defined(CONFIG_FSFREEZE)
2525 
2526 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2527 {
2528     error_setg(errp, QERR_UNSUPPORTED);
2529     return NULL;
2530 }
2531 
2532 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2533 {
2534     error_setg(errp, QERR_UNSUPPORTED);
2535 
2536     return 0;
2537 }
2538 
2539 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
2540 {
2541     error_setg(errp, QERR_UNSUPPORTED);
2542 
2543     return 0;
2544 }
2545 
2546 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
2547                                        strList *mountpoints,
2548                                        Error **errp)
2549 {
2550     error_setg(errp, QERR_UNSUPPORTED);
2551 
2552     return 0;
2553 }
2554 
2555 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
2556 {
2557     error_setg(errp, QERR_UNSUPPORTED);
2558 
2559     return 0;
2560 }
2561 #endif /* CONFIG_FSFREEZE */
2562 
2563 #if !defined(CONFIG_FSTRIM)
2564 GuestFilesystemTrimResponse *
2565 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
2566 {
2567     error_setg(errp, QERR_UNSUPPORTED);
2568     return NULL;
2569 }
2570 #endif
2571 
2572 /* add unsupported commands to the blacklist */
2573 GList *ga_command_blacklist_init(GList *blacklist)
2574 {
2575 #if !defined(__linux__)
2576     {
2577         const char *list[] = {
2578             "guest-suspend-disk", "guest-suspend-ram",
2579             "guest-suspend-hybrid", "guest-network-get-interfaces",
2580             "guest-get-vcpus", "guest-set-vcpus",
2581             "guest-get-memory-blocks", "guest-set-memory-blocks",
2582             "guest-get-memory-block-size", NULL};
2583         char **p = (char **)list;
2584 
2585         while (*p) {
2586             blacklist = g_list_append(blacklist, g_strdup(*p++));
2587         }
2588     }
2589 #endif
2590 
2591 #if !defined(CONFIG_FSFREEZE)
2592     {
2593         const char *list[] = {
2594             "guest-get-fsinfo", "guest-fsfreeze-status",
2595             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
2596             "guest-fsfreeze-thaw", "guest-get-fsinfo", NULL};
2597         char **p = (char **)list;
2598 
2599         while (*p) {
2600             blacklist = g_list_append(blacklist, g_strdup(*p++));
2601         }
2602     }
2603 #endif
2604 
2605 #if !defined(CONFIG_FSTRIM)
2606     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
2607 #endif
2608 
2609     return blacklist;
2610 }
2611 
2612 /* register init/cleanup routines for stateful command groups */
2613 void ga_command_state_init(GAState *s, GACommandState *cs)
2614 {
2615 #if defined(CONFIG_FSFREEZE)
2616     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
2617 #endif
2618 }
2619 
2620 #ifdef HAVE_UTMPX
2621 
2622 #define QGA_MICRO_SECOND_TO_SECOND 1000000
2623 
2624 static double ga_get_login_time(struct utmpx *user_info)
2625 {
2626     double seconds = (double)user_info->ut_tv.tv_sec;
2627     double useconds = (double)user_info->ut_tv.tv_usec;
2628     useconds /= QGA_MICRO_SECOND_TO_SECOND;
2629     return seconds + useconds;
2630 }
2631 
2632 GuestUserList *qmp_guest_get_users(Error **err)
2633 {
2634     GHashTable *cache = NULL;
2635     GuestUserList *head = NULL, *cur_item = NULL;
2636     struct utmpx *user_info = NULL;
2637     gpointer value = NULL;
2638     GuestUser *user = NULL;
2639     GuestUserList *item = NULL;
2640     double login_time = 0;
2641 
2642     cache = g_hash_table_new(g_str_hash, g_str_equal);
2643     setutxent();
2644 
2645     for (;;) {
2646         user_info = getutxent();
2647         if (user_info == NULL) {
2648             break;
2649         } else if (user_info->ut_type != USER_PROCESS) {
2650             continue;
2651         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
2652             value = g_hash_table_lookup(cache, user_info->ut_user);
2653             user = (GuestUser *)value;
2654             login_time = ga_get_login_time(user_info);
2655             /* We're ensuring the earliest login time to be sent */
2656             if (login_time < user->login_time) {
2657                 user->login_time = login_time;
2658             }
2659             continue;
2660         }
2661 
2662         item = g_new0(GuestUserList, 1);
2663         item->value = g_new0(GuestUser, 1);
2664         item->value->user = g_strdup(user_info->ut_user);
2665         item->value->login_time = ga_get_login_time(user_info);
2666 
2667         g_hash_table_insert(cache, item->value->user, item->value);
2668 
2669         if (!cur_item) {
2670             head = cur_item = item;
2671         } else {
2672             cur_item->next = item;
2673             cur_item = item;
2674         }
2675     }
2676     endutxent();
2677     g_hash_table_destroy(cache);
2678     return head;
2679 }
2680 
2681 #else
2682 
2683 GuestUserList *qmp_guest_get_users(Error **errp)
2684 {
2685     error_setg(errp, QERR_UNSUPPORTED);
2686     return NULL;
2687 }
2688 
2689 #endif
2690 
2691 /* Replace escaped special characters with theire real values. The replacement
2692  * is done in place -- returned value is in the original string.
2693  */
2694 static void ga_osrelease_replace_special(gchar *value)
2695 {
2696     gchar *p, *p2, quote;
2697 
2698     /* Trim the string at first space or semicolon if it is not enclosed in
2699      * single or double quotes. */
2700     if ((value[0] != '"') || (value[0] == '\'')) {
2701         p = strchr(value, ' ');
2702         if (p != NULL) {
2703             *p = 0;
2704         }
2705         p = strchr(value, ';');
2706         if (p != NULL) {
2707             *p = 0;
2708         }
2709         return;
2710     }
2711 
2712     quote = value[0];
2713     p2 = value;
2714     p = value + 1;
2715     while (*p != 0) {
2716         if (*p == '\\') {
2717             p++;
2718             switch (*p) {
2719             case '$':
2720             case '\'':
2721             case '"':
2722             case '\\':
2723             case '`':
2724                 break;
2725             default:
2726                 /* Keep literal backslash followed by whatever is there */
2727                 p--;
2728                 break;
2729             }
2730         } else if (*p == quote) {
2731             *p2 = 0;
2732             break;
2733         }
2734         *(p2++) = *(p++);
2735     }
2736 }
2737 
2738 static GKeyFile *ga_parse_osrelease(const char *fname)
2739 {
2740     gchar *content = NULL;
2741     gchar *content2 = NULL;
2742     GError *err = NULL;
2743     GKeyFile *keys = g_key_file_new();
2744     const char *group = "[os-release]\n";
2745 
2746     if (!g_file_get_contents(fname, &content, NULL, &err)) {
2747         slog("failed to read '%s', error: %s", fname, err->message);
2748         goto fail;
2749     }
2750 
2751     if (!g_utf8_validate(content, -1, NULL)) {
2752         slog("file is not utf-8 encoded: %s", fname);
2753         goto fail;
2754     }
2755     content2 = g_strdup_printf("%s%s", group, content);
2756 
2757     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
2758                                    &err)) {
2759         slog("failed to parse file '%s', error: %s", fname, err->message);
2760         goto fail;
2761     }
2762 
2763     g_free(content);
2764     g_free(content2);
2765     return keys;
2766 
2767 fail:
2768     g_error_free(err);
2769     g_free(content);
2770     g_free(content2);
2771     g_key_file_free(keys);
2772     return NULL;
2773 }
2774 
2775 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
2776 {
2777     GuestOSInfo *info = NULL;
2778     struct utsname kinfo;
2779     GKeyFile *osrelease = NULL;
2780     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
2781 
2782     info = g_new0(GuestOSInfo, 1);
2783 
2784     if (uname(&kinfo) != 0) {
2785         error_setg_errno(errp, errno, "uname failed");
2786     } else {
2787         info->has_kernel_version = true;
2788         info->kernel_version = g_strdup(kinfo.version);
2789         info->has_kernel_release = true;
2790         info->kernel_release = g_strdup(kinfo.release);
2791         info->has_machine = true;
2792         info->machine = g_strdup(kinfo.machine);
2793     }
2794 
2795     if (qga_os_release != NULL) {
2796         osrelease = ga_parse_osrelease(qga_os_release);
2797     } else {
2798         osrelease = ga_parse_osrelease("/etc/os-release");
2799         if (osrelease == NULL) {
2800             osrelease = ga_parse_osrelease("/usr/lib/os-release");
2801         }
2802     }
2803 
2804     if (osrelease != NULL) {
2805         char *value;
2806 
2807 #define GET_FIELD(field, osfield) do { \
2808     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
2809     if (value != NULL) { \
2810         ga_osrelease_replace_special(value); \
2811         info->has_ ## field = true; \
2812         info->field = value; \
2813     } \
2814 } while (0)
2815         GET_FIELD(id, "ID");
2816         GET_FIELD(name, "NAME");
2817         GET_FIELD(pretty_name, "PRETTY_NAME");
2818         GET_FIELD(version, "VERSION");
2819         GET_FIELD(version_id, "VERSION_ID");
2820         GET_FIELD(variant, "VARIANT");
2821         GET_FIELD(variant_id, "VARIANT_ID");
2822 #undef GET_FIELD
2823 
2824         g_key_file_free(osrelease);
2825     }
2826 
2827     return info;
2828 }
2829