xref: /qemu/qga/commands-posix.c (revision 1fbf2665)
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 #include "commands-common.h"
29 
30 #ifdef HAVE_UTMPX
31 #include <utmpx.h>
32 #endif
33 
34 #if defined(__linux__)
35 #include <mntent.h>
36 #include <linux/fs.h>
37 #include <ifaddrs.h>
38 #include <arpa/inet.h>
39 #include <sys/socket.h>
40 #include <net/if.h>
41 #include <sys/statvfs.h>
42 
43 #ifdef CONFIG_LIBUDEV
44 #include <libudev.h>
45 #endif
46 
47 #ifdef FIFREEZE
48 #define CONFIG_FSFREEZE
49 #endif
50 #ifdef FITRIM
51 #define CONFIG_FSTRIM
52 #endif
53 #endif
54 
55 static void ga_wait_child(pid_t pid, int *status, Error **errp)
56 {
57     pid_t rpid;
58 
59     *status = 0;
60 
61     do {
62         rpid = waitpid(pid, status, 0);
63     } while (rpid == -1 && errno == EINTR);
64 
65     if (rpid == -1) {
66         error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
67                          pid);
68         return;
69     }
70 
71     g_assert(rpid == pid);
72 }
73 
74 void qmp_guest_shutdown(bool has_mode, const char *mode, Error **errp)
75 {
76     const char *shutdown_flag;
77     Error *local_err = NULL;
78     pid_t pid;
79     int status;
80 
81     slog("guest-shutdown called, mode: %s", mode);
82     if (!has_mode || strcmp(mode, "powerdown") == 0) {
83         shutdown_flag = "-P";
84     } else if (strcmp(mode, "halt") == 0) {
85         shutdown_flag = "-H";
86     } else if (strcmp(mode, "reboot") == 0) {
87         shutdown_flag = "-r";
88     } else {
89         error_setg(errp,
90                    "mode is invalid (valid values are: halt|powerdown|reboot");
91         return;
92     }
93 
94     pid = fork();
95     if (pid == 0) {
96         /* child, start the shutdown */
97         setsid();
98         reopen_fd_to_null(0);
99         reopen_fd_to_null(1);
100         reopen_fd_to_null(2);
101 
102         execl("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
103                "hypervisor initiated shutdown", (char *)NULL);
104         _exit(EXIT_FAILURE);
105     } else if (pid < 0) {
106         error_setg_errno(errp, errno, "failed to create child process");
107         return;
108     }
109 
110     ga_wait_child(pid, &status, &local_err);
111     if (local_err) {
112         error_propagate(errp, local_err);
113         return;
114     }
115 
116     if (!WIFEXITED(status)) {
117         error_setg(errp, "child process has terminated abnormally");
118         return;
119     }
120 
121     if (WEXITSTATUS(status)) {
122         error_setg(errp, "child process has failed to shutdown");
123         return;
124     }
125 
126     /* succeeded */
127 }
128 
129 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
130 {
131     int ret;
132     int status;
133     pid_t pid;
134     Error *local_err = NULL;
135     struct timeval tv;
136     static const char hwclock_path[] = "/sbin/hwclock";
137     static int hwclock_available = -1;
138 
139     if (hwclock_available < 0) {
140         hwclock_available = (access(hwclock_path, X_OK) == 0);
141     }
142 
143     if (!hwclock_available) {
144         error_setg(errp, QERR_UNSUPPORTED);
145         return;
146     }
147 
148     /* If user has passed a time, validate and set it. */
149     if (has_time) {
150         GDate date = { 0, };
151 
152         /* year-2038 will overflow in case time_t is 32bit */
153         if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
154             error_setg(errp, "Time %" PRId64 " is too large", time_ns);
155             return;
156         }
157 
158         tv.tv_sec = time_ns / 1000000000;
159         tv.tv_usec = (time_ns % 1000000000) / 1000;
160         g_date_set_time_t(&date, tv.tv_sec);
161         if (date.year < 1970 || date.year >= 2070) {
162             error_setg_errno(errp, errno, "Invalid time");
163             return;
164         }
165 
166         ret = settimeofday(&tv, NULL);
167         if (ret < 0) {
168             error_setg_errno(errp, errno, "Failed to set time to guest");
169             return;
170         }
171     }
172 
173     /* Now, if user has passed a time to set and the system time is set, we
174      * just need to synchronize the hardware clock. However, if no time was
175      * passed, user is requesting the opposite: set the system time from the
176      * hardware clock (RTC). */
177     pid = fork();
178     if (pid == 0) {
179         setsid();
180         reopen_fd_to_null(0);
181         reopen_fd_to_null(1);
182         reopen_fd_to_null(2);
183 
184         /* Use '/sbin/hwclock -w' to set RTC from the system time,
185          * or '/sbin/hwclock -s' to set the system time from RTC. */
186         execl(hwclock_path, "hwclock", has_time ? "-w" : "-s", NULL);
187         _exit(EXIT_FAILURE);
188     } else if (pid < 0) {
189         error_setg_errno(errp, errno, "failed to create child process");
190         return;
191     }
192 
193     ga_wait_child(pid, &status, &local_err);
194     if (local_err) {
195         error_propagate(errp, local_err);
196         return;
197     }
198 
199     if (!WIFEXITED(status)) {
200         error_setg(errp, "child process has terminated abnormally");
201         return;
202     }
203 
204     if (WEXITSTATUS(status)) {
205         error_setg(errp, "hwclock failed to set hardware clock to system time");
206         return;
207     }
208 }
209 
210 typedef enum {
211     RW_STATE_NEW,
212     RW_STATE_READING,
213     RW_STATE_WRITING,
214 } RwState;
215 
216 struct GuestFileHandle {
217     uint64_t id;
218     FILE *fh;
219     RwState state;
220     QTAILQ_ENTRY(GuestFileHandle) next;
221 };
222 
223 static struct {
224     QTAILQ_HEAD(, GuestFileHandle) filehandles;
225 } guest_file_state = {
226     .filehandles = QTAILQ_HEAD_INITIALIZER(guest_file_state.filehandles),
227 };
228 
229 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
230 {
231     GuestFileHandle *gfh;
232     int64_t handle;
233 
234     handle = ga_get_fd_handle(ga_state, errp);
235     if (handle < 0) {
236         return -1;
237     }
238 
239     gfh = g_new0(GuestFileHandle, 1);
240     gfh->id = handle;
241     gfh->fh = fh;
242     QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
243 
244     return handle;
245 }
246 
247 GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
248 {
249     GuestFileHandle *gfh;
250 
251     QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
252     {
253         if (gfh->id == id) {
254             return gfh;
255         }
256     }
257 
258     error_setg(errp, "handle '%" PRId64 "' has not been found", id);
259     return NULL;
260 }
261 
262 typedef const char * const ccpc;
263 
264 #ifndef O_BINARY
265 #define O_BINARY 0
266 #endif
267 
268 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
269 static const struct {
270     ccpc *forms;
271     int oflag_base;
272 } guest_file_open_modes[] = {
273     { (ccpc[]){ "r",          NULL }, O_RDONLY                                 },
274     { (ccpc[]){ "rb",         NULL }, O_RDONLY                      | O_BINARY },
275     { (ccpc[]){ "w",          NULL }, O_WRONLY | O_CREAT | O_TRUNC             },
276     { (ccpc[]){ "wb",         NULL }, O_WRONLY | O_CREAT | O_TRUNC  | O_BINARY },
277     { (ccpc[]){ "a",          NULL }, O_WRONLY | O_CREAT | O_APPEND            },
278     { (ccpc[]){ "ab",         NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
279     { (ccpc[]){ "r+",         NULL }, O_RDWR                                   },
280     { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR                        | O_BINARY },
281     { (ccpc[]){ "w+",         NULL }, O_RDWR   | O_CREAT | O_TRUNC             },
282     { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR   | O_CREAT | O_TRUNC  | O_BINARY },
283     { (ccpc[]){ "a+",         NULL }, O_RDWR   | O_CREAT | O_APPEND            },
284     { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR   | O_CREAT | O_APPEND | O_BINARY }
285 };
286 
287 static int
288 find_open_flag(const char *mode_str, Error **errp)
289 {
290     unsigned mode;
291 
292     for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
293         ccpc *form;
294 
295         form = guest_file_open_modes[mode].forms;
296         while (*form != NULL && strcmp(*form, mode_str) != 0) {
297             ++form;
298         }
299         if (*form != NULL) {
300             break;
301         }
302     }
303 
304     if (mode == ARRAY_SIZE(guest_file_open_modes)) {
305         error_setg(errp, "invalid file open mode '%s'", mode_str);
306         return -1;
307     }
308     return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
309 }
310 
311 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
312                                S_IRGRP | S_IWGRP | \
313                                S_IROTH | S_IWOTH)
314 
315 static FILE *
316 safe_open_or_create(const char *path, const char *mode, Error **errp)
317 {
318     Error *local_err = NULL;
319     int oflag;
320 
321     oflag = find_open_flag(mode, &local_err);
322     if (local_err == NULL) {
323         int fd;
324 
325         /* If the caller wants / allows creation of a new file, we implement it
326          * with a two step process: open() + (open() / fchmod()).
327          *
328          * First we insist on creating the file exclusively as a new file. If
329          * that succeeds, we're free to set any file-mode bits on it. (The
330          * motivation is that we want to set those file-mode bits independently
331          * of the current umask.)
332          *
333          * If the exclusive creation fails because the file already exists
334          * (EEXIST is not possible for any other reason), we just attempt to
335          * open the file, but in this case we won't be allowed to change the
336          * file-mode bits on the preexistent file.
337          *
338          * The pathname should never disappear between the two open()s in
339          * practice. If it happens, then someone very likely tried to race us.
340          * In this case just go ahead and report the ENOENT from the second
341          * open() to the caller.
342          *
343          * If the caller wants to open a preexistent file, then the first
344          * open() is decisive and its third argument is ignored, and the second
345          * open() and the fchmod() are never called.
346          */
347         fd = open(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
348         if (fd == -1 && errno == EEXIST) {
349             oflag &= ~(unsigned)O_CREAT;
350             fd = open(path, oflag);
351         }
352 
353         if (fd == -1) {
354             error_setg_errno(&local_err, errno, "failed to open file '%s' "
355                              "(mode: '%s')", path, mode);
356         } else {
357             qemu_set_cloexec(fd);
358 
359             if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
360                 error_setg_errno(&local_err, errno, "failed to set permission "
361                                  "0%03o on new file '%s' (mode: '%s')",
362                                  (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
363             } else {
364                 FILE *f;
365 
366                 f = fdopen(fd, mode);
367                 if (f == NULL) {
368                     error_setg_errno(&local_err, errno, "failed to associate "
369                                      "stdio stream with file descriptor %d, "
370                                      "file '%s' (mode: '%s')", fd, path, mode);
371                 } else {
372                     return f;
373                 }
374             }
375 
376             close(fd);
377             if (oflag & O_CREAT) {
378                 unlink(path);
379             }
380         }
381     }
382 
383     error_propagate(errp, local_err);
384     return NULL;
385 }
386 
387 int64_t qmp_guest_file_open(const char *path, bool has_mode, const char *mode,
388                             Error **errp)
389 {
390     FILE *fh;
391     Error *local_err = NULL;
392     int64_t handle;
393 
394     if (!has_mode) {
395         mode = "r";
396     }
397     slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
398     fh = safe_open_or_create(path, mode, &local_err);
399     if (local_err != NULL) {
400         error_propagate(errp, local_err);
401         return -1;
402     }
403 
404     /* set fd non-blocking to avoid common use cases (like reading from a
405      * named pipe) from hanging the agent
406      */
407     qemu_set_nonblock(fileno(fh));
408 
409     handle = guest_file_handle_add(fh, errp);
410     if (handle < 0) {
411         fclose(fh);
412         return -1;
413     }
414 
415     slog("guest-file-open, handle: %" PRId64, handle);
416     return handle;
417 }
418 
419 void qmp_guest_file_close(int64_t handle, Error **errp)
420 {
421     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
422     int ret;
423 
424     slog("guest-file-close called, handle: %" PRId64, handle);
425     if (!gfh) {
426         return;
427     }
428 
429     ret = fclose(gfh->fh);
430     if (ret == EOF) {
431         error_setg_errno(errp, errno, "failed to close handle");
432         return;
433     }
434 
435     QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
436     g_free(gfh);
437 }
438 
439 GuestFileRead *guest_file_read_unsafe(GuestFileHandle *gfh,
440                                       int64_t count, Error **errp)
441 {
442     GuestFileRead *read_data = NULL;
443     guchar *buf;
444     FILE *fh = gfh->fh;
445     size_t read_count;
446 
447     /* explicitly flush when switching from writing to reading */
448     if (gfh->state == RW_STATE_WRITING) {
449         int ret = fflush(fh);
450         if (ret == EOF) {
451             error_setg_errno(errp, errno, "failed to flush file");
452             return NULL;
453         }
454         gfh->state = RW_STATE_NEW;
455     }
456 
457     buf = g_malloc0(count + 1);
458     read_count = fread(buf, 1, count, fh);
459     if (ferror(fh)) {
460         error_setg_errno(errp, errno, "failed to read file");
461     } else {
462         buf[read_count] = 0;
463         read_data = g_new0(GuestFileRead, 1);
464         read_data->count = read_count;
465         read_data->eof = feof(fh);
466         if (read_count) {
467             read_data->buf_b64 = g_base64_encode(buf, read_count);
468         }
469         gfh->state = RW_STATE_READING;
470     }
471     g_free(buf);
472     clearerr(fh);
473 
474     return read_data;
475 }
476 
477 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
478                                      bool has_count, int64_t count,
479                                      Error **errp)
480 {
481     GuestFileWrite *write_data = NULL;
482     guchar *buf;
483     gsize buf_len;
484     int write_count;
485     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
486     FILE *fh;
487 
488     if (!gfh) {
489         return NULL;
490     }
491 
492     fh = gfh->fh;
493 
494     if (gfh->state == RW_STATE_READING) {
495         int ret = fseek(fh, 0, SEEK_CUR);
496         if (ret == -1) {
497             error_setg_errno(errp, errno, "failed to seek file");
498             return NULL;
499         }
500         gfh->state = RW_STATE_NEW;
501     }
502 
503     buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
504     if (!buf) {
505         return NULL;
506     }
507 
508     if (!has_count) {
509         count = buf_len;
510     } else if (count < 0 || count > buf_len) {
511         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
512                    count);
513         g_free(buf);
514         return NULL;
515     }
516 
517     write_count = fwrite(buf, 1, count, fh);
518     if (ferror(fh)) {
519         error_setg_errno(errp, errno, "failed to write to file");
520         slog("guest-file-write failed, handle: %" PRId64, handle);
521     } else {
522         write_data = g_new0(GuestFileWrite, 1);
523         write_data->count = write_count;
524         write_data->eof = feof(fh);
525         gfh->state = RW_STATE_WRITING;
526     }
527     g_free(buf);
528     clearerr(fh);
529 
530     return write_data;
531 }
532 
533 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
534                                           GuestFileWhence *whence_code,
535                                           Error **errp)
536 {
537     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
538     GuestFileSeek *seek_data = NULL;
539     FILE *fh;
540     int ret;
541     int whence;
542     Error *err = NULL;
543 
544     if (!gfh) {
545         return NULL;
546     }
547 
548     /* We stupidly exposed 'whence':'int' in our qapi */
549     whence = ga_parse_whence(whence_code, &err);
550     if (err) {
551         error_propagate(errp, err);
552         return NULL;
553     }
554 
555     fh = gfh->fh;
556     ret = fseek(fh, offset, whence);
557     if (ret == -1) {
558         error_setg_errno(errp, errno, "failed to seek file");
559         if (errno == ESPIPE) {
560             /* file is non-seekable, stdio shouldn't be buffering anyways */
561             gfh->state = RW_STATE_NEW;
562         }
563     } else {
564         seek_data = g_new0(GuestFileSeek, 1);
565         seek_data->position = ftell(fh);
566         seek_data->eof = feof(fh);
567         gfh->state = RW_STATE_NEW;
568     }
569     clearerr(fh);
570 
571     return seek_data;
572 }
573 
574 void qmp_guest_file_flush(int64_t handle, Error **errp)
575 {
576     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
577     FILE *fh;
578     int ret;
579 
580     if (!gfh) {
581         return;
582     }
583 
584     fh = gfh->fh;
585     ret = fflush(fh);
586     if (ret == EOF) {
587         error_setg_errno(errp, errno, "failed to flush file");
588     } else {
589         gfh->state = RW_STATE_NEW;
590     }
591 }
592 
593 /* linux-specific implementations. avoid this if at all possible. */
594 #if defined(__linux__)
595 
596 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
597 typedef struct FsMount {
598     char *dirname;
599     char *devtype;
600     unsigned int devmajor, devminor;
601     QTAILQ_ENTRY(FsMount) next;
602 } FsMount;
603 
604 typedef QTAILQ_HEAD(FsMountList, FsMount) FsMountList;
605 
606 static void free_fs_mount_list(FsMountList *mounts)
607 {
608      FsMount *mount, *temp;
609 
610      if (!mounts) {
611          return;
612      }
613 
614      QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
615          QTAILQ_REMOVE(mounts, mount, next);
616          g_free(mount->dirname);
617          g_free(mount->devtype);
618          g_free(mount);
619      }
620 }
621 
622 static int dev_major_minor(const char *devpath,
623                            unsigned int *devmajor, unsigned int *devminor)
624 {
625     struct stat st;
626 
627     *devmajor = 0;
628     *devminor = 0;
629 
630     if (stat(devpath, &st) < 0) {
631         slog("failed to stat device file '%s': %s", devpath, strerror(errno));
632         return -1;
633     }
634     if (S_ISDIR(st.st_mode)) {
635         /* It is bind mount */
636         return -2;
637     }
638     if (S_ISBLK(st.st_mode)) {
639         *devmajor = major(st.st_rdev);
640         *devminor = minor(st.st_rdev);
641         return 0;
642     }
643     return -1;
644 }
645 
646 /*
647  * Walk the mount table and build a list of local file systems
648  */
649 static void build_fs_mount_list_from_mtab(FsMountList *mounts, Error **errp)
650 {
651     struct mntent *ment;
652     FsMount *mount;
653     char const *mtab = "/proc/self/mounts";
654     FILE *fp;
655     unsigned int devmajor, devminor;
656 
657     fp = setmntent(mtab, "r");
658     if (!fp) {
659         error_setg(errp, "failed to open mtab file: '%s'", mtab);
660         return;
661     }
662 
663     while ((ment = getmntent(fp))) {
664         /*
665          * An entry which device name doesn't start with a '/' is
666          * either a dummy file system or a network file system.
667          * Add special handling for smbfs and cifs as is done by
668          * coreutils as well.
669          */
670         if ((ment->mnt_fsname[0] != '/') ||
671             (strcmp(ment->mnt_type, "smbfs") == 0) ||
672             (strcmp(ment->mnt_type, "cifs") == 0)) {
673             continue;
674         }
675         if (dev_major_minor(ment->mnt_fsname, &devmajor, &devminor) == -2) {
676             /* Skip bind mounts */
677             continue;
678         }
679 
680         mount = g_new0(FsMount, 1);
681         mount->dirname = g_strdup(ment->mnt_dir);
682         mount->devtype = g_strdup(ment->mnt_type);
683         mount->devmajor = devmajor;
684         mount->devminor = devminor;
685 
686         QTAILQ_INSERT_TAIL(mounts, mount, next);
687     }
688 
689     endmntent(fp);
690 }
691 
692 static void decode_mntname(char *name, int len)
693 {
694     int i, j = 0;
695     for (i = 0; i <= len; i++) {
696         if (name[i] != '\\') {
697             name[j++] = name[i];
698         } else if (name[i + 1] == '\\') {
699             name[j++] = '\\';
700             i++;
701         } else if (name[i + 1] >= '0' && name[i + 1] <= '3' &&
702                    name[i + 2] >= '0' && name[i + 2] <= '7' &&
703                    name[i + 3] >= '0' && name[i + 3] <= '7') {
704             name[j++] = (name[i + 1] - '0') * 64 +
705                         (name[i + 2] - '0') * 8 +
706                         (name[i + 3] - '0');
707             i += 3;
708         } else {
709             name[j++] = name[i];
710         }
711     }
712 }
713 
714 static void build_fs_mount_list(FsMountList *mounts, Error **errp)
715 {
716     FsMount *mount;
717     char const *mountinfo = "/proc/self/mountinfo";
718     FILE *fp;
719     char *line = NULL, *dash;
720     size_t n;
721     char check;
722     unsigned int devmajor, devminor;
723     int ret, dir_s, dir_e, type_s, type_e, dev_s, dev_e;
724 
725     fp = fopen(mountinfo, "r");
726     if (!fp) {
727         build_fs_mount_list_from_mtab(mounts, errp);
728         return;
729     }
730 
731     while (getline(&line, &n, fp) != -1) {
732         ret = sscanf(line, "%*u %*u %u:%u %*s %n%*s%n%c",
733                      &devmajor, &devminor, &dir_s, &dir_e, &check);
734         if (ret < 3) {
735             continue;
736         }
737         dash = strstr(line + dir_e, " - ");
738         if (!dash) {
739             continue;
740         }
741         ret = sscanf(dash, " - %n%*s%n %n%*s%n%c",
742                      &type_s, &type_e, &dev_s, &dev_e, &check);
743         if (ret < 1) {
744             continue;
745         }
746         line[dir_e] = 0;
747         dash[type_e] = 0;
748         dash[dev_e] = 0;
749         decode_mntname(line + dir_s, dir_e - dir_s);
750         decode_mntname(dash + dev_s, dev_e - dev_s);
751         if (devmajor == 0) {
752             /* btrfs reports major number = 0 */
753             if (strcmp("btrfs", dash + type_s) != 0 ||
754                 dev_major_minor(dash + dev_s, &devmajor, &devminor) < 0) {
755                 continue;
756             }
757         }
758 
759         mount = g_new0(FsMount, 1);
760         mount->dirname = g_strdup(line + dir_s);
761         mount->devtype = g_strdup(dash + type_s);
762         mount->devmajor = devmajor;
763         mount->devminor = devminor;
764 
765         QTAILQ_INSERT_TAIL(mounts, mount, next);
766     }
767     free(line);
768 
769     fclose(fp);
770 }
771 #endif
772 
773 #if defined(CONFIG_FSFREEZE)
774 
775 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
776 {
777     char *path;
778     char *dpath;
779     char *driver = NULL;
780     char buf[PATH_MAX];
781     ssize_t len;
782 
783     path = g_strndup(syspath, pathlen);
784     dpath = g_strdup_printf("%s/driver", path);
785     len = readlink(dpath, buf, sizeof(buf) - 1);
786     if (len != -1) {
787         buf[len] = 0;
788         driver = g_path_get_basename(buf);
789     }
790     g_free(dpath);
791     g_free(path);
792     return driver;
793 }
794 
795 static int compare_uint(const void *_a, const void *_b)
796 {
797     unsigned int a = *(unsigned int *)_a;
798     unsigned int b = *(unsigned int *)_b;
799 
800     return a < b ? -1 : a > b ? 1 : 0;
801 }
802 
803 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
804 static int build_hosts(char const *syspath, char const *host, bool ata,
805                        unsigned int *hosts, int hosts_max, Error **errp)
806 {
807     char *path;
808     DIR *dir;
809     struct dirent *entry;
810     int i = 0;
811 
812     path = g_strndup(syspath, host - syspath);
813     dir = opendir(path);
814     if (!dir) {
815         error_setg_errno(errp, errno, "opendir(\"%s\")", path);
816         g_free(path);
817         return -1;
818     }
819 
820     while (i < hosts_max) {
821         entry = readdir(dir);
822         if (!entry) {
823             break;
824         }
825         if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
826             ++i;
827         } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
828             ++i;
829         }
830     }
831 
832     qsort(hosts, i, sizeof(hosts[0]), compare_uint);
833 
834     g_free(path);
835     closedir(dir);
836     return i;
837 }
838 
839 /*
840  * Store disk device info for devices on the PCI bus.
841  * Returns true if information has been stored, or false for failure.
842  */
843 static bool build_guest_fsinfo_for_pci_dev(char const *syspath,
844                                            GuestDiskAddress *disk,
845                                            Error **errp)
846 {
847     unsigned int pci[4], host, hosts[8], tgt[3];
848     int i, nhosts = 0, pcilen;
849     GuestPCIAddress *pciaddr = disk->pci_controller;
850     bool has_ata = false, has_host = false, has_tgt = false;
851     char *p, *q, *driver = NULL;
852     bool ret = false;
853 
854     p = strstr(syspath, "/devices/pci");
855     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
856                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
857         g_debug("only pci device is supported: sysfs path '%s'", syspath);
858         return false;
859     }
860 
861     p += 12 + pcilen;
862     while (true) {
863         driver = get_pci_driver(syspath, p - syspath, errp);
864         if (driver && (g_str_equal(driver, "ata_piix") ||
865                        g_str_equal(driver, "sym53c8xx") ||
866                        g_str_equal(driver, "virtio-pci") ||
867                        g_str_equal(driver, "ahci"))) {
868             break;
869         }
870 
871         g_free(driver);
872         if (sscanf(p, "/%x:%x:%x.%x%n",
873                           pci, pci + 1, pci + 2, pci + 3, &pcilen) == 4) {
874             p += pcilen;
875             continue;
876         }
877 
878         g_debug("unsupported driver or sysfs path '%s'", syspath);
879         return false;
880     }
881 
882     p = strstr(syspath, "/target");
883     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
884                     tgt, tgt + 1, tgt + 2) == 3) {
885         has_tgt = true;
886     }
887 
888     p = strstr(syspath, "/ata");
889     if (p) {
890         q = p + 4;
891         has_ata = true;
892     } else {
893         p = strstr(syspath, "/host");
894         q = p + 5;
895     }
896     if (p && sscanf(q, "%u", &host) == 1) {
897         has_host = true;
898         nhosts = build_hosts(syspath, p, has_ata, hosts,
899                              ARRAY_SIZE(hosts), errp);
900         if (nhosts < 0) {
901             goto cleanup;
902         }
903     }
904 
905     pciaddr->domain = pci[0];
906     pciaddr->bus = pci[1];
907     pciaddr->slot = pci[2];
908     pciaddr->function = pci[3];
909 
910     if (strcmp(driver, "ata_piix") == 0) {
911         /* a host per ide bus, target*:0:<unit>:0 */
912         if (!has_host || !has_tgt) {
913             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
914             goto cleanup;
915         }
916         for (i = 0; i < nhosts; i++) {
917             if (host == hosts[i]) {
918                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
919                 disk->bus = i;
920                 disk->unit = tgt[1];
921                 break;
922             }
923         }
924         if (i >= nhosts) {
925             g_debug("no host for '%s' (driver '%s')", syspath, driver);
926             goto cleanup;
927         }
928     } else if (strcmp(driver, "sym53c8xx") == 0) {
929         /* scsi(LSI Logic): target*:0:<unit>:0 */
930         if (!has_tgt) {
931             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
932             goto cleanup;
933         }
934         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
935         disk->unit = tgt[1];
936     } else if (strcmp(driver, "virtio-pci") == 0) {
937         if (has_tgt) {
938             /* virtio-scsi: target*:0:0:<unit> */
939             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
940             disk->unit = tgt[2];
941         } else {
942             /* virtio-blk: 1 disk per 1 device */
943             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
944         }
945     } else if (strcmp(driver, "ahci") == 0) {
946         /* ahci: 1 host per 1 unit */
947         if (!has_host || !has_tgt) {
948             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
949             goto cleanup;
950         }
951         for (i = 0; i < nhosts; i++) {
952             if (host == hosts[i]) {
953                 disk->unit = i;
954                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
955                 break;
956             }
957         }
958         if (i >= nhosts) {
959             g_debug("no host for '%s' (driver '%s')", syspath, driver);
960             goto cleanup;
961         }
962     } else {
963         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
964         goto cleanup;
965     }
966 
967     ret = true;
968 
969 cleanup:
970     g_free(driver);
971     return ret;
972 }
973 
974 /*
975  * Store disk device info for non-PCI virtio devices (for example s390x
976  * channel I/O devices). Returns true if information has been stored, or
977  * false for failure.
978  */
979 static bool build_guest_fsinfo_for_nonpci_virtio(char const *syspath,
980                                                  GuestDiskAddress *disk,
981                                                  Error **errp)
982 {
983     unsigned int tgt[3];
984     char *p;
985 
986     if (!strstr(syspath, "/virtio") || !strstr(syspath, "/block")) {
987         g_debug("Unsupported virtio device '%s'", syspath);
988         return false;
989     }
990 
991     p = strstr(syspath, "/target");
992     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
993                     &tgt[0], &tgt[1], &tgt[2]) == 3) {
994         /* virtio-scsi: target*:0:<target>:<unit> */
995         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
996         disk->bus = tgt[0];
997         disk->target = tgt[1];
998         disk->unit = tgt[2];
999     } else {
1000         /* virtio-blk: 1 disk per 1 device */
1001         disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
1002     }
1003 
1004     return true;
1005 }
1006 
1007 /*
1008  * Store disk device info for CCW devices (s390x channel I/O devices).
1009  * Returns true if information has been stored, or false for failure.
1010  */
1011 static bool build_guest_fsinfo_for_ccw_dev(char const *syspath,
1012                                            GuestDiskAddress *disk,
1013                                            Error **errp)
1014 {
1015     unsigned int cssid, ssid, subchno, devno;
1016     char *p;
1017 
1018     p = strstr(syspath, "/devices/css");
1019     if (!p || sscanf(p + 12, "%*x/%x.%x.%x/%*x.%*x.%x/",
1020                      &cssid, &ssid, &subchno, &devno) < 4) {
1021         g_debug("could not parse ccw device sysfs path: %s", syspath);
1022         return false;
1023     }
1024 
1025     disk->has_ccw_address = true;
1026     disk->ccw_address = g_new0(GuestCCWAddress, 1);
1027     disk->ccw_address->cssid = cssid;
1028     disk->ccw_address->ssid = ssid;
1029     disk->ccw_address->subchno = subchno;
1030     disk->ccw_address->devno = devno;
1031 
1032     if (strstr(p, "/virtio")) {
1033         build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1034     }
1035 
1036     return true;
1037 }
1038 
1039 /* Store disk device info specified by @sysfs into @fs */
1040 static void build_guest_fsinfo_for_real_device(char const *syspath,
1041                                                GuestFilesystemInfo *fs,
1042                                                Error **errp)
1043 {
1044     GuestDiskAddress *disk;
1045     GuestPCIAddress *pciaddr;
1046     bool has_hwinf;
1047 #ifdef CONFIG_LIBUDEV
1048     struct udev *udev = NULL;
1049     struct udev_device *udevice = NULL;
1050 #endif
1051 
1052     pciaddr = g_new0(GuestPCIAddress, 1);
1053     pciaddr->domain = -1;                       /* -1 means field is invalid */
1054     pciaddr->bus = -1;
1055     pciaddr->slot = -1;
1056     pciaddr->function = -1;
1057 
1058     disk = g_new0(GuestDiskAddress, 1);
1059     disk->pci_controller = pciaddr;
1060     disk->bus_type = GUEST_DISK_BUS_TYPE_UNKNOWN;
1061 
1062 #ifdef CONFIG_LIBUDEV
1063     udev = udev_new();
1064     udevice = udev_device_new_from_syspath(udev, syspath);
1065     if (udev == NULL || udevice == NULL) {
1066         g_debug("failed to query udev");
1067     } else {
1068         const char *devnode, *serial;
1069         devnode = udev_device_get_devnode(udevice);
1070         if (devnode != NULL) {
1071             disk->dev = g_strdup(devnode);
1072             disk->has_dev = true;
1073         }
1074         serial = udev_device_get_property_value(udevice, "ID_SERIAL");
1075         if (serial != NULL && *serial != 0) {
1076             disk->serial = g_strdup(serial);
1077             disk->has_serial = true;
1078         }
1079     }
1080 
1081     udev_unref(udev);
1082     udev_device_unref(udevice);
1083 #endif
1084 
1085     if (strstr(syspath, "/devices/pci")) {
1086         has_hwinf = build_guest_fsinfo_for_pci_dev(syspath, disk, errp);
1087     } else if (strstr(syspath, "/devices/css")) {
1088         has_hwinf = build_guest_fsinfo_for_ccw_dev(syspath, disk, errp);
1089     } else if (strstr(syspath, "/virtio")) {
1090         has_hwinf = build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1091     } else {
1092         g_debug("Unsupported device type for '%s'", syspath);
1093         has_hwinf = false;
1094     }
1095 
1096     if (has_hwinf || disk->has_dev || disk->has_serial) {
1097         QAPI_LIST_PREPEND(fs->disk, disk);
1098     } else {
1099         qapi_free_GuestDiskAddress(disk);
1100     }
1101 }
1102 
1103 static void build_guest_fsinfo_for_device(char const *devpath,
1104                                           GuestFilesystemInfo *fs,
1105                                           Error **errp);
1106 
1107 /* Store a list of slave devices of virtual volume specified by @syspath into
1108  * @fs */
1109 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
1110                                                   GuestFilesystemInfo *fs,
1111                                                   Error **errp)
1112 {
1113     Error *err = NULL;
1114     DIR *dir;
1115     char *dirpath;
1116     struct dirent *entry;
1117 
1118     dirpath = g_strdup_printf("%s/slaves", syspath);
1119     dir = opendir(dirpath);
1120     if (!dir) {
1121         if (errno != ENOENT) {
1122             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1123         }
1124         g_free(dirpath);
1125         return;
1126     }
1127 
1128     for (;;) {
1129         errno = 0;
1130         entry = readdir(dir);
1131         if (entry == NULL) {
1132             if (errno) {
1133                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1134             }
1135             break;
1136         }
1137 
1138         if (entry->d_type == DT_LNK) {
1139             char *path;
1140 
1141             g_debug(" slave device '%s'", entry->d_name);
1142             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1143             build_guest_fsinfo_for_device(path, fs, &err);
1144             g_free(path);
1145 
1146             if (err) {
1147                 error_propagate(errp, err);
1148                 break;
1149             }
1150         }
1151     }
1152 
1153     g_free(dirpath);
1154     closedir(dir);
1155 }
1156 
1157 static bool is_disk_virtual(const char *devpath, Error **errp)
1158 {
1159     g_autofree char *syspath = realpath(devpath, NULL);
1160 
1161     if (!syspath) {
1162         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1163         return false;
1164     }
1165     return strstr(syspath, "/devices/virtual/block/") != NULL;
1166 }
1167 
1168 /* Dispatch to functions for virtual/real device */
1169 static void build_guest_fsinfo_for_device(char const *devpath,
1170                                           GuestFilesystemInfo *fs,
1171                                           Error **errp)
1172 {
1173     ERRP_GUARD();
1174     g_autofree char *syspath = NULL;
1175     bool is_virtual = false;
1176 
1177     syspath = realpath(devpath, NULL);
1178     if (!syspath) {
1179         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1180         return;
1181     }
1182 
1183     if (!fs->name) {
1184         fs->name = g_path_get_basename(syspath);
1185     }
1186 
1187     g_debug("  parse sysfs path '%s'", syspath);
1188     is_virtual = is_disk_virtual(syspath, errp);
1189     if (*errp != NULL) {
1190         return;
1191     }
1192     if (is_virtual) {
1193         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1194     } else {
1195         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1196     }
1197 }
1198 
1199 #ifdef CONFIG_LIBUDEV
1200 
1201 /*
1202  * Wrapper around build_guest_fsinfo_for_device() for getting just
1203  * the disk address.
1204  */
1205 static GuestDiskAddress *get_disk_address(const char *syspath, Error **errp)
1206 {
1207     g_autoptr(GuestFilesystemInfo) fs = NULL;
1208 
1209     fs = g_new0(GuestFilesystemInfo, 1);
1210     build_guest_fsinfo_for_device(syspath, fs, errp);
1211     if (fs->disk != NULL) {
1212         return g_steal_pointer(&fs->disk->value);
1213     }
1214     return NULL;
1215 }
1216 
1217 static char *get_alias_for_syspath(const char *syspath)
1218 {
1219     struct udev *udev = NULL;
1220     struct udev_device *udevice = NULL;
1221     char *ret = NULL;
1222 
1223     udev = udev_new();
1224     if (udev == NULL) {
1225         g_debug("failed to query udev");
1226         goto out;
1227     }
1228     udevice = udev_device_new_from_syspath(udev, syspath);
1229     if (udevice == NULL) {
1230         g_debug("failed to query udev for path: %s", syspath);
1231         goto out;
1232     } else {
1233         const char *alias = udev_device_get_property_value(
1234             udevice, "DM_NAME");
1235         /*
1236          * NULL means there was an error and empty string means there is no
1237          * alias. In case of no alias we return NULL instead of empty string.
1238          */
1239         if (alias == NULL) {
1240             g_debug("failed to query udev for device alias for: %s",
1241                 syspath);
1242         } else if (*alias != 0) {
1243             ret = g_strdup(alias);
1244         }
1245     }
1246 
1247 out:
1248     udev_unref(udev);
1249     udev_device_unref(udevice);
1250     return ret;
1251 }
1252 
1253 static char *get_device_for_syspath(const char *syspath)
1254 {
1255     struct udev *udev = NULL;
1256     struct udev_device *udevice = NULL;
1257     char *ret = NULL;
1258 
1259     udev = udev_new();
1260     if (udev == NULL) {
1261         g_debug("failed to query udev");
1262         goto out;
1263     }
1264     udevice = udev_device_new_from_syspath(udev, syspath);
1265     if (udevice == NULL) {
1266         g_debug("failed to query udev for path: %s", syspath);
1267         goto out;
1268     } else {
1269         ret = g_strdup(udev_device_get_devnode(udevice));
1270     }
1271 
1272 out:
1273     udev_unref(udev);
1274     udev_device_unref(udevice);
1275     return ret;
1276 }
1277 
1278 static void get_disk_deps(const char *disk_dir, GuestDiskInfo *disk)
1279 {
1280     g_autofree char *deps_dir = NULL;
1281     const gchar *dep;
1282     GDir *dp_deps = NULL;
1283 
1284     /* List dependent disks */
1285     deps_dir = g_strdup_printf("%s/slaves", disk_dir);
1286     g_debug("  listing entries in: %s", deps_dir);
1287     dp_deps = g_dir_open(deps_dir, 0, NULL);
1288     if (dp_deps == NULL) {
1289         g_debug("failed to list entries in %s", deps_dir);
1290         return;
1291     }
1292     disk->has_dependencies = true;
1293     while ((dep = g_dir_read_name(dp_deps)) != NULL) {
1294         g_autofree char *dep_dir = NULL;
1295         char *dev_name;
1296 
1297         /* Add dependent disks */
1298         dep_dir = g_strdup_printf("%s/%s", deps_dir, dep);
1299         dev_name = get_device_for_syspath(dep_dir);
1300         if (dev_name != NULL) {
1301             g_debug("  adding dependent device: %s", dev_name);
1302             QAPI_LIST_PREPEND(disk->dependencies, dev_name);
1303         }
1304     }
1305     g_dir_close(dp_deps);
1306 }
1307 
1308 /*
1309  * Detect partitions subdirectory, name is "<disk_name><number>" or
1310  * "<disk_name>p<number>"
1311  *
1312  * @disk_name -- last component of /sys path (e.g. sda)
1313  * @disk_dir -- sys path of the disk (e.g. /sys/block/sda)
1314  * @disk_dev -- device node of the disk (e.g. /dev/sda)
1315  */
1316 static GuestDiskInfoList *get_disk_partitions(
1317     GuestDiskInfoList *list,
1318     const char *disk_name, const char *disk_dir,
1319     const char *disk_dev)
1320 {
1321     GuestDiskInfoList *ret = list;
1322     struct dirent *de_disk;
1323     DIR *dp_disk = NULL;
1324     size_t len = strlen(disk_name);
1325 
1326     dp_disk = opendir(disk_dir);
1327     while ((de_disk = readdir(dp_disk)) != NULL) {
1328         g_autofree char *partition_dir = NULL;
1329         char *dev_name;
1330         GuestDiskInfo *partition;
1331 
1332         if (!(de_disk->d_type & DT_DIR)) {
1333             continue;
1334         }
1335 
1336         if (!(strncmp(disk_name, de_disk->d_name, len) == 0 &&
1337             ((*(de_disk->d_name + len) == 'p' &&
1338             isdigit(*(de_disk->d_name + len + 1))) ||
1339                 isdigit(*(de_disk->d_name + len))))) {
1340             continue;
1341         }
1342 
1343         partition_dir = g_strdup_printf("%s/%s",
1344             disk_dir, de_disk->d_name);
1345         dev_name = get_device_for_syspath(partition_dir);
1346         if (dev_name == NULL) {
1347             g_debug("Failed to get device name for syspath: %s",
1348                 disk_dir);
1349             continue;
1350         }
1351         partition = g_new0(GuestDiskInfo, 1);
1352         partition->name = dev_name;
1353         partition->partition = true;
1354         partition->has_dependencies = true;
1355         /* Add parent disk as dependent for easier tracking of hierarchy */
1356         QAPI_LIST_PREPEND(partition->dependencies, g_strdup(disk_dev));
1357 
1358         QAPI_LIST_PREPEND(ret, partition);
1359     }
1360     closedir(dp_disk);
1361 
1362     return ret;
1363 }
1364 
1365 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1366 {
1367     GuestDiskInfoList *ret = NULL;
1368     GuestDiskInfo *disk;
1369     DIR *dp = NULL;
1370     struct dirent *de = NULL;
1371 
1372     g_debug("listing /sys/block directory");
1373     dp = opendir("/sys/block");
1374     if (dp == NULL) {
1375         error_setg_errno(errp, errno, "Can't open directory \"/sys/block\"");
1376         return NULL;
1377     }
1378     while ((de = readdir(dp)) != NULL) {
1379         g_autofree char *disk_dir = NULL, *line = NULL,
1380             *size_path = NULL;
1381         char *dev_name;
1382         Error *local_err = NULL;
1383         if (de->d_type != DT_LNK) {
1384             g_debug("  skipping entry: %s", de->d_name);
1385             continue;
1386         }
1387 
1388         /* Check size and skip zero-sized disks */
1389         g_debug("  checking disk size");
1390         size_path = g_strdup_printf("/sys/block/%s/size", de->d_name);
1391         if (!g_file_get_contents(size_path, &line, NULL, NULL)) {
1392             g_debug("  failed to read disk size");
1393             continue;
1394         }
1395         if (g_strcmp0(line, "0\n") == 0) {
1396             g_debug("  skipping zero-sized disk");
1397             continue;
1398         }
1399 
1400         g_debug("  adding %s", de->d_name);
1401         disk_dir = g_strdup_printf("/sys/block/%s", de->d_name);
1402         dev_name = get_device_for_syspath(disk_dir);
1403         if (dev_name == NULL) {
1404             g_debug("Failed to get device name for syspath: %s",
1405                 disk_dir);
1406             continue;
1407         }
1408         disk = g_new0(GuestDiskInfo, 1);
1409         disk->name = dev_name;
1410         disk->partition = false;
1411         disk->alias = get_alias_for_syspath(disk_dir);
1412         disk->has_alias = (disk->alias != NULL);
1413         QAPI_LIST_PREPEND(ret, disk);
1414 
1415         /* Get address for non-virtual devices */
1416         bool is_virtual = is_disk_virtual(disk_dir, &local_err);
1417         if (local_err != NULL) {
1418             g_debug("  failed to check disk path, ignoring error: %s",
1419                 error_get_pretty(local_err));
1420             error_free(local_err);
1421             local_err = NULL;
1422             /* Don't try to get the address */
1423             is_virtual = true;
1424         }
1425         if (!is_virtual) {
1426             disk->address = get_disk_address(disk_dir, &local_err);
1427             if (local_err != NULL) {
1428                 g_debug("  failed to get device info, ignoring error: %s",
1429                     error_get_pretty(local_err));
1430                 error_free(local_err);
1431                 local_err = NULL;
1432             } else if (disk->address != NULL) {
1433                 disk->has_address = true;
1434             }
1435         }
1436 
1437         get_disk_deps(disk_dir, disk);
1438         ret = get_disk_partitions(ret, de->d_name, disk_dir, dev_name);
1439     }
1440 
1441     closedir(dp);
1442 
1443     return ret;
1444 }
1445 
1446 #else
1447 
1448 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1449 {
1450     error_setg(errp, QERR_UNSUPPORTED);
1451     return NULL;
1452 }
1453 
1454 #endif
1455 
1456 /* Return a list of the disk device(s)' info which @mount lies on */
1457 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1458                                                Error **errp)
1459 {
1460     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1461     struct statvfs buf;
1462     unsigned long used, nonroot_total, fr_size;
1463     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1464                                     mount->devmajor, mount->devminor);
1465 
1466     fs->mountpoint = g_strdup(mount->dirname);
1467     fs->type = g_strdup(mount->devtype);
1468     build_guest_fsinfo_for_device(devpath, fs, errp);
1469 
1470     if (statvfs(fs->mountpoint, &buf) == 0) {
1471         fr_size = buf.f_frsize;
1472         used = buf.f_blocks - buf.f_bfree;
1473         nonroot_total = used + buf.f_bavail;
1474         fs->used_bytes = used * fr_size;
1475         fs->total_bytes = nonroot_total * fr_size;
1476 
1477         fs->has_total_bytes = true;
1478         fs->has_used_bytes = true;
1479     }
1480 
1481     g_free(devpath);
1482 
1483     return fs;
1484 }
1485 
1486 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1487 {
1488     FsMountList mounts;
1489     struct FsMount *mount;
1490     GuestFilesystemInfoList *ret = NULL;
1491     Error *local_err = NULL;
1492 
1493     QTAILQ_INIT(&mounts);
1494     build_fs_mount_list(&mounts, &local_err);
1495     if (local_err) {
1496         error_propagate(errp, local_err);
1497         return NULL;
1498     }
1499 
1500     QTAILQ_FOREACH(mount, &mounts, next) {
1501         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1502 
1503         QAPI_LIST_PREPEND(ret, build_guest_fsinfo(mount, &local_err));
1504         if (local_err) {
1505             error_propagate(errp, local_err);
1506             qapi_free_GuestFilesystemInfoList(ret);
1507             ret = NULL;
1508             break;
1509         }
1510     }
1511 
1512     free_fs_mount_list(&mounts);
1513     return ret;
1514 }
1515 
1516 
1517 typedef enum {
1518     FSFREEZE_HOOK_THAW = 0,
1519     FSFREEZE_HOOK_FREEZE,
1520 } FsfreezeHookArg;
1521 
1522 static const char *fsfreeze_hook_arg_string[] = {
1523     "thaw",
1524     "freeze",
1525 };
1526 
1527 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
1528 {
1529     int status;
1530     pid_t pid;
1531     const char *hook;
1532     const char *arg_str = fsfreeze_hook_arg_string[arg];
1533     Error *local_err = NULL;
1534 
1535     hook = ga_fsfreeze_hook(ga_state);
1536     if (!hook) {
1537         return;
1538     }
1539     if (access(hook, X_OK) != 0) {
1540         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
1541         return;
1542     }
1543 
1544     slog("executing fsfreeze hook with arg '%s'", arg_str);
1545     pid = fork();
1546     if (pid == 0) {
1547         setsid();
1548         reopen_fd_to_null(0);
1549         reopen_fd_to_null(1);
1550         reopen_fd_to_null(2);
1551 
1552         execl(hook, hook, arg_str, NULL);
1553         _exit(EXIT_FAILURE);
1554     } else if (pid < 0) {
1555         error_setg_errno(errp, errno, "failed to create child process");
1556         return;
1557     }
1558 
1559     ga_wait_child(pid, &status, &local_err);
1560     if (local_err) {
1561         error_propagate(errp, local_err);
1562         return;
1563     }
1564 
1565     if (!WIFEXITED(status)) {
1566         error_setg(errp, "fsfreeze hook has terminated abnormally");
1567         return;
1568     }
1569 
1570     status = WEXITSTATUS(status);
1571     if (status) {
1572         error_setg(errp, "fsfreeze hook has failed with status %d", status);
1573         return;
1574     }
1575 }
1576 
1577 /*
1578  * Return status of freeze/thaw
1579  */
1580 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
1581 {
1582     if (ga_is_frozen(ga_state)) {
1583         return GUEST_FSFREEZE_STATUS_FROZEN;
1584     }
1585 
1586     return GUEST_FSFREEZE_STATUS_THAWED;
1587 }
1588 
1589 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
1590 {
1591     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
1592 }
1593 
1594 /*
1595  * Walk list of mounted file systems in the guest, and freeze the ones which
1596  * are real local file systems.
1597  */
1598 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
1599                                        strList *mountpoints,
1600                                        Error **errp)
1601 {
1602     int ret = 0, i = 0;
1603     strList *list;
1604     FsMountList mounts;
1605     struct FsMount *mount;
1606     Error *local_err = NULL;
1607     int fd;
1608 
1609     slog("guest-fsfreeze called");
1610 
1611     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
1612     if (local_err) {
1613         error_propagate(errp, local_err);
1614         return -1;
1615     }
1616 
1617     QTAILQ_INIT(&mounts);
1618     build_fs_mount_list(&mounts, &local_err);
1619     if (local_err) {
1620         error_propagate(errp, local_err);
1621         return -1;
1622     }
1623 
1624     /* cannot risk guest agent blocking itself on a write in this state */
1625     ga_set_frozen(ga_state);
1626 
1627     QTAILQ_FOREACH_REVERSE(mount, &mounts, next) {
1628         /* To issue fsfreeze in the reverse order of mounts, check if the
1629          * mount is listed in the list here */
1630         if (has_mountpoints) {
1631             for (list = mountpoints; list; list = list->next) {
1632                 if (strcmp(list->value, mount->dirname) == 0) {
1633                     break;
1634                 }
1635             }
1636             if (!list) {
1637                 continue;
1638             }
1639         }
1640 
1641         fd = qemu_open_old(mount->dirname, O_RDONLY);
1642         if (fd == -1) {
1643             error_setg_errno(errp, errno, "failed to open %s", mount->dirname);
1644             goto error;
1645         }
1646 
1647         /* we try to cull filesystems we know won't work in advance, but other
1648          * filesystems may not implement fsfreeze for less obvious reasons.
1649          * these will report EOPNOTSUPP. we simply ignore these when tallying
1650          * the number of frozen filesystems.
1651          * if a filesystem is mounted more than once (aka bind mount) a
1652          * consecutive attempt to freeze an already frozen filesystem will
1653          * return EBUSY.
1654          *
1655          * any other error means a failure to freeze a filesystem we
1656          * expect to be freezable, so return an error in those cases
1657          * and return system to thawed state.
1658          */
1659         ret = ioctl(fd, FIFREEZE);
1660         if (ret == -1) {
1661             if (errno != EOPNOTSUPP && errno != EBUSY) {
1662                 error_setg_errno(errp, errno, "failed to freeze %s",
1663                                  mount->dirname);
1664                 close(fd);
1665                 goto error;
1666             }
1667         } else {
1668             i++;
1669         }
1670         close(fd);
1671     }
1672 
1673     free_fs_mount_list(&mounts);
1674     /* We may not issue any FIFREEZE here.
1675      * Just unset ga_state here and ready for the next call.
1676      */
1677     if (i == 0) {
1678         ga_unset_frozen(ga_state);
1679     }
1680     return i;
1681 
1682 error:
1683     free_fs_mount_list(&mounts);
1684     qmp_guest_fsfreeze_thaw(NULL);
1685     return 0;
1686 }
1687 
1688 /*
1689  * Walk list of frozen file systems in the guest, and thaw them.
1690  */
1691 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
1692 {
1693     int ret;
1694     FsMountList mounts;
1695     FsMount *mount;
1696     int fd, i = 0, logged;
1697     Error *local_err = NULL;
1698 
1699     QTAILQ_INIT(&mounts);
1700     build_fs_mount_list(&mounts, &local_err);
1701     if (local_err) {
1702         error_propagate(errp, local_err);
1703         return 0;
1704     }
1705 
1706     QTAILQ_FOREACH(mount, &mounts, next) {
1707         logged = false;
1708         fd = qemu_open_old(mount->dirname, O_RDONLY);
1709         if (fd == -1) {
1710             continue;
1711         }
1712         /* we have no way of knowing whether a filesystem was actually unfrozen
1713          * as a result of a successful call to FITHAW, only that if an error
1714          * was returned the filesystem was *not* unfrozen by that particular
1715          * call.
1716          *
1717          * since multiple preceding FIFREEZEs require multiple calls to FITHAW
1718          * to unfreeze, continuing issuing FITHAW until an error is returned,
1719          * in which case either the filesystem is in an unfreezable state, or,
1720          * more likely, it was thawed previously (and remains so afterward).
1721          *
1722          * also, since the most recent successful call is the one that did
1723          * the actual unfreeze, we can use this to provide an accurate count
1724          * of the number of filesystems unfrozen by guest-fsfreeze-thaw, which
1725          * may * be useful for determining whether a filesystem was unfrozen
1726          * during the freeze/thaw phase by a process other than qemu-ga.
1727          */
1728         do {
1729             ret = ioctl(fd, FITHAW);
1730             if (ret == 0 && !logged) {
1731                 i++;
1732                 logged = true;
1733             }
1734         } while (ret == 0);
1735         close(fd);
1736     }
1737 
1738     ga_unset_frozen(ga_state);
1739     free_fs_mount_list(&mounts);
1740 
1741     execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
1742 
1743     return i;
1744 }
1745 
1746 static void guest_fsfreeze_cleanup(void)
1747 {
1748     Error *err = NULL;
1749 
1750     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
1751         qmp_guest_fsfreeze_thaw(&err);
1752         if (err) {
1753             slog("failed to clean up frozen filesystems: %s",
1754                  error_get_pretty(err));
1755             error_free(err);
1756         }
1757     }
1758 }
1759 #endif /* CONFIG_FSFREEZE */
1760 
1761 #if defined(CONFIG_FSTRIM)
1762 /*
1763  * Walk list of mounted file systems in the guest, and trim them.
1764  */
1765 GuestFilesystemTrimResponse *
1766 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1767 {
1768     GuestFilesystemTrimResponse *response;
1769     GuestFilesystemTrimResult *result;
1770     int ret = 0;
1771     FsMountList mounts;
1772     struct FsMount *mount;
1773     int fd;
1774     Error *local_err = NULL;
1775     struct fstrim_range r;
1776 
1777     slog("guest-fstrim called");
1778 
1779     QTAILQ_INIT(&mounts);
1780     build_fs_mount_list(&mounts, &local_err);
1781     if (local_err) {
1782         error_propagate(errp, local_err);
1783         return NULL;
1784     }
1785 
1786     response = g_malloc0(sizeof(*response));
1787 
1788     QTAILQ_FOREACH(mount, &mounts, next) {
1789         result = g_malloc0(sizeof(*result));
1790         result->path = g_strdup(mount->dirname);
1791 
1792         QAPI_LIST_PREPEND(response->paths, result);
1793 
1794         fd = qemu_open_old(mount->dirname, O_RDONLY);
1795         if (fd == -1) {
1796             result->error = g_strdup_printf("failed to open: %s",
1797                                             strerror(errno));
1798             result->has_error = true;
1799             continue;
1800         }
1801 
1802         /* We try to cull filesystems we know won't work in advance, but other
1803          * filesystems may not implement fstrim for less obvious reasons.
1804          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1805          * will be reported (e.g. CD-ROMs).
1806          * Any other error means an unexpected error.
1807          */
1808         r.start = 0;
1809         r.len = -1;
1810         r.minlen = has_minimum ? minimum : 0;
1811         ret = ioctl(fd, FITRIM, &r);
1812         if (ret == -1) {
1813             result->has_error = true;
1814             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1815                 result->error = g_strdup("trim not supported");
1816             } else {
1817                 result->error = g_strdup_printf("failed to trim: %s",
1818                                                 strerror(errno));
1819             }
1820             close(fd);
1821             continue;
1822         }
1823 
1824         result->has_minimum = true;
1825         result->minimum = r.minlen;
1826         result->has_trimmed = true;
1827         result->trimmed = r.len;
1828         close(fd);
1829     }
1830 
1831     free_fs_mount_list(&mounts);
1832     return response;
1833 }
1834 #endif /* CONFIG_FSTRIM */
1835 
1836 
1837 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1838 #define SUSPEND_SUPPORTED 0
1839 #define SUSPEND_NOT_SUPPORTED 1
1840 
1841 typedef enum {
1842     SUSPEND_MODE_DISK = 0,
1843     SUSPEND_MODE_RAM = 1,
1844     SUSPEND_MODE_HYBRID = 2,
1845 } SuspendMode;
1846 
1847 /*
1848  * Executes a command in a child process using g_spawn_sync,
1849  * returning an int >= 0 representing the exit status of the
1850  * process.
1851  *
1852  * If the program wasn't found in path, returns -1.
1853  *
1854  * If a problem happened when creating the child process,
1855  * returns -1 and errp is set.
1856  */
1857 static int run_process_child(const char *command[], Error **errp)
1858 {
1859     int exit_status, spawn_flag;
1860     GError *g_err = NULL;
1861     bool success;
1862 
1863     spawn_flag = G_SPAWN_SEARCH_PATH | G_SPAWN_STDOUT_TO_DEV_NULL |
1864                  G_SPAWN_STDERR_TO_DEV_NULL;
1865 
1866     success =  g_spawn_sync(NULL, (char **)command, NULL, spawn_flag,
1867                             NULL, NULL, NULL, NULL,
1868                             &exit_status, &g_err);
1869 
1870     if (success) {
1871         return WEXITSTATUS(exit_status);
1872     }
1873 
1874     if (g_err && (g_err->code != G_SPAWN_ERROR_NOENT)) {
1875         error_setg(errp, "failed to create child process, error '%s'",
1876                    g_err->message);
1877     }
1878 
1879     g_error_free(g_err);
1880     return -1;
1881 }
1882 
1883 static bool systemd_supports_mode(SuspendMode mode, Error **errp)
1884 {
1885     const char *systemctl_args[3] = {"systemd-hibernate", "systemd-suspend",
1886                                      "systemd-hybrid-sleep"};
1887     const char *cmd[4] = {"systemctl", "status", systemctl_args[mode], NULL};
1888     int status;
1889 
1890     status = run_process_child(cmd, errp);
1891 
1892     /*
1893      * systemctl status uses LSB return codes so we can expect
1894      * status > 0 and be ok. To assert if the guest has support
1895      * for the selected suspend mode, status should be < 4. 4 is
1896      * the code for unknown service status, the return value when
1897      * the service does not exist. A common value is status = 3
1898      * (program is not running).
1899      */
1900     if (status > 0 && status < 4) {
1901         return true;
1902     }
1903 
1904     return false;
1905 }
1906 
1907 static void systemd_suspend(SuspendMode mode, Error **errp)
1908 {
1909     Error *local_err = NULL;
1910     const char *systemctl_args[3] = {"hibernate", "suspend", "hybrid-sleep"};
1911     const char *cmd[3] = {"systemctl", systemctl_args[mode], NULL};
1912     int status;
1913 
1914     status = run_process_child(cmd, &local_err);
1915 
1916     if (status == 0) {
1917         return;
1918     }
1919 
1920     if ((status == -1) && !local_err) {
1921         error_setg(errp, "the helper program 'systemctl %s' was not found",
1922                    systemctl_args[mode]);
1923         return;
1924     }
1925 
1926     if (local_err) {
1927         error_propagate(errp, local_err);
1928     } else {
1929         error_setg(errp, "the helper program 'systemctl %s' returned an "
1930                    "unexpected exit status code (%d)",
1931                    systemctl_args[mode], status);
1932     }
1933 }
1934 
1935 static bool pmutils_supports_mode(SuspendMode mode, Error **errp)
1936 {
1937     Error *local_err = NULL;
1938     const char *pmutils_args[3] = {"--hibernate", "--suspend",
1939                                    "--suspend-hybrid"};
1940     const char *cmd[3] = {"pm-is-supported", pmutils_args[mode], NULL};
1941     int status;
1942 
1943     status = run_process_child(cmd, &local_err);
1944 
1945     if (status == SUSPEND_SUPPORTED) {
1946         return true;
1947     }
1948 
1949     if ((status == -1) && !local_err) {
1950         return false;
1951     }
1952 
1953     if (local_err) {
1954         error_propagate(errp, local_err);
1955     } else {
1956         error_setg(errp,
1957                    "the helper program '%s' returned an unexpected exit"
1958                    " status code (%d)", "pm-is-supported", status);
1959     }
1960 
1961     return false;
1962 }
1963 
1964 static void pmutils_suspend(SuspendMode mode, Error **errp)
1965 {
1966     Error *local_err = NULL;
1967     const char *pmutils_binaries[3] = {"pm-hibernate", "pm-suspend",
1968                                        "pm-suspend-hybrid"};
1969     const char *cmd[2] = {pmutils_binaries[mode], NULL};
1970     int status;
1971 
1972     status = run_process_child(cmd, &local_err);
1973 
1974     if (status == 0) {
1975         return;
1976     }
1977 
1978     if ((status == -1) && !local_err) {
1979         error_setg(errp, "the helper program '%s' was not found",
1980                    pmutils_binaries[mode]);
1981         return;
1982     }
1983 
1984     if (local_err) {
1985         error_propagate(errp, local_err);
1986     } else {
1987         error_setg(errp,
1988                    "the helper program '%s' returned an unexpected exit"
1989                    " status code (%d)", pmutils_binaries[mode], status);
1990     }
1991 }
1992 
1993 static bool linux_sys_state_supports_mode(SuspendMode mode, Error **errp)
1994 {
1995     const char *sysfile_strs[3] = {"disk", "mem", NULL};
1996     const char *sysfile_str = sysfile_strs[mode];
1997     char buf[32]; /* hopefully big enough */
1998     int fd;
1999     ssize_t ret;
2000 
2001     if (!sysfile_str) {
2002         error_setg(errp, "unknown guest suspend mode");
2003         return false;
2004     }
2005 
2006     fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
2007     if (fd < 0) {
2008         return false;
2009     }
2010 
2011     ret = read(fd, buf, sizeof(buf) - 1);
2012     close(fd);
2013     if (ret <= 0) {
2014         return false;
2015     }
2016     buf[ret] = '\0';
2017 
2018     if (strstr(buf, sysfile_str)) {
2019         return true;
2020     }
2021     return false;
2022 }
2023 
2024 static void linux_sys_state_suspend(SuspendMode mode, Error **errp)
2025 {
2026     Error *local_err = NULL;
2027     const char *sysfile_strs[3] = {"disk", "mem", NULL};
2028     const char *sysfile_str = sysfile_strs[mode];
2029     pid_t pid;
2030     int status;
2031 
2032     if (!sysfile_str) {
2033         error_setg(errp, "unknown guest suspend mode");
2034         return;
2035     }
2036 
2037     pid = fork();
2038     if (!pid) {
2039         /* child */
2040         int fd;
2041 
2042         setsid();
2043         reopen_fd_to_null(0);
2044         reopen_fd_to_null(1);
2045         reopen_fd_to_null(2);
2046 
2047         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
2048         if (fd < 0) {
2049             _exit(EXIT_FAILURE);
2050         }
2051 
2052         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
2053             _exit(EXIT_FAILURE);
2054         }
2055 
2056         _exit(EXIT_SUCCESS);
2057     } else if (pid < 0) {
2058         error_setg_errno(errp, errno, "failed to create child process");
2059         return;
2060     }
2061 
2062     ga_wait_child(pid, &status, &local_err);
2063     if (local_err) {
2064         error_propagate(errp, local_err);
2065         return;
2066     }
2067 
2068     if (WEXITSTATUS(status)) {
2069         error_setg(errp, "child process has failed to suspend");
2070     }
2071 
2072 }
2073 
2074 static void guest_suspend(SuspendMode mode, Error **errp)
2075 {
2076     Error *local_err = NULL;
2077     bool mode_supported = false;
2078 
2079     if (systemd_supports_mode(mode, &local_err)) {
2080         mode_supported = true;
2081         systemd_suspend(mode, &local_err);
2082     }
2083 
2084     if (!local_err) {
2085         return;
2086     }
2087 
2088     error_free(local_err);
2089     local_err = NULL;
2090 
2091     if (pmutils_supports_mode(mode, &local_err)) {
2092         mode_supported = true;
2093         pmutils_suspend(mode, &local_err);
2094     }
2095 
2096     if (!local_err) {
2097         return;
2098     }
2099 
2100     error_free(local_err);
2101     local_err = NULL;
2102 
2103     if (linux_sys_state_supports_mode(mode, &local_err)) {
2104         mode_supported = true;
2105         linux_sys_state_suspend(mode, &local_err);
2106     }
2107 
2108     if (!mode_supported) {
2109         error_free(local_err);
2110         error_setg(errp,
2111                    "the requested suspend mode is not supported by the guest");
2112     } else {
2113         error_propagate(errp, local_err);
2114     }
2115 }
2116 
2117 void qmp_guest_suspend_disk(Error **errp)
2118 {
2119     guest_suspend(SUSPEND_MODE_DISK, errp);
2120 }
2121 
2122 void qmp_guest_suspend_ram(Error **errp)
2123 {
2124     guest_suspend(SUSPEND_MODE_RAM, errp);
2125 }
2126 
2127 void qmp_guest_suspend_hybrid(Error **errp)
2128 {
2129     guest_suspend(SUSPEND_MODE_HYBRID, errp);
2130 }
2131 
2132 static GuestNetworkInterface *
2133 guest_find_interface(GuestNetworkInterfaceList *head,
2134                      const char *name)
2135 {
2136     for (; head; head = head->next) {
2137         if (strcmp(head->value->name, name) == 0) {
2138             return head->value;
2139         }
2140     }
2141 
2142     return NULL;
2143 }
2144 
2145 static int guest_get_network_stats(const char *name,
2146                        GuestNetworkInterfaceStat *stats)
2147 {
2148     int name_len;
2149     char const *devinfo = "/proc/net/dev";
2150     FILE *fp;
2151     char *line = NULL, *colon;
2152     size_t n = 0;
2153     fp = fopen(devinfo, "r");
2154     if (!fp) {
2155         return -1;
2156     }
2157     name_len = strlen(name);
2158     while (getline(&line, &n, fp) != -1) {
2159         long long dummy;
2160         long long rx_bytes;
2161         long long rx_packets;
2162         long long rx_errs;
2163         long long rx_dropped;
2164         long long tx_bytes;
2165         long long tx_packets;
2166         long long tx_errs;
2167         long long tx_dropped;
2168         char *trim_line;
2169         trim_line = g_strchug(line);
2170         if (trim_line[0] == '\0') {
2171             continue;
2172         }
2173         colon = strchr(trim_line, ':');
2174         if (!colon) {
2175             continue;
2176         }
2177         if (colon - name_len  == trim_line &&
2178            strncmp(trim_line, name, name_len) == 0) {
2179             if (sscanf(colon + 1,
2180                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
2181                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
2182                   &dummy, &dummy, &dummy, &dummy,
2183                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
2184                   &dummy, &dummy, &dummy, &dummy) != 16) {
2185                 continue;
2186             }
2187             stats->rx_bytes = rx_bytes;
2188             stats->rx_packets = rx_packets;
2189             stats->rx_errs = rx_errs;
2190             stats->rx_dropped = rx_dropped;
2191             stats->tx_bytes = tx_bytes;
2192             stats->tx_packets = tx_packets;
2193             stats->tx_errs = tx_errs;
2194             stats->tx_dropped = tx_dropped;
2195             fclose(fp);
2196             g_free(line);
2197             return 0;
2198         }
2199     }
2200     fclose(fp);
2201     g_free(line);
2202     g_debug("/proc/net/dev: Interface '%s' not found", name);
2203     return -1;
2204 }
2205 
2206 /*
2207  * Build information about guest interfaces
2208  */
2209 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2210 {
2211     GuestNetworkInterfaceList *head = NULL, **tail = &head;
2212     struct ifaddrs *ifap, *ifa;
2213 
2214     if (getifaddrs(&ifap) < 0) {
2215         error_setg_errno(errp, errno, "getifaddrs failed");
2216         goto error;
2217     }
2218 
2219     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
2220         GuestNetworkInterface *info;
2221         GuestIpAddressList **address_tail;
2222         GuestIpAddress *address_item = NULL;
2223         GuestNetworkInterfaceStat *interface_stat = NULL;
2224         char addr4[INET_ADDRSTRLEN];
2225         char addr6[INET6_ADDRSTRLEN];
2226         int sock;
2227         struct ifreq ifr;
2228         unsigned char *mac_addr;
2229         void *p;
2230 
2231         g_debug("Processing %s interface", ifa->ifa_name);
2232 
2233         info = guest_find_interface(head, ifa->ifa_name);
2234 
2235         if (!info) {
2236             info = g_malloc0(sizeof(*info));
2237             info->name = g_strdup(ifa->ifa_name);
2238 
2239             QAPI_LIST_APPEND(tail, info);
2240         }
2241 
2242         if (!info->has_hardware_address && ifa->ifa_flags & SIOCGIFHWADDR) {
2243             /* we haven't obtained HW address yet */
2244             sock = socket(PF_INET, SOCK_STREAM, 0);
2245             if (sock == -1) {
2246                 error_setg_errno(errp, errno, "failed to create socket");
2247                 goto error;
2248             }
2249 
2250             memset(&ifr, 0, sizeof(ifr));
2251             pstrcpy(ifr.ifr_name, IF_NAMESIZE, info->name);
2252             if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
2253                 error_setg_errno(errp, errno,
2254                                  "failed to get MAC address of %s",
2255                                  ifa->ifa_name);
2256                 close(sock);
2257                 goto error;
2258             }
2259 
2260             close(sock);
2261             mac_addr = (unsigned char *) &ifr.ifr_hwaddr.sa_data;
2262 
2263             info->hardware_address =
2264                 g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
2265                                 (int) mac_addr[0], (int) mac_addr[1],
2266                                 (int) mac_addr[2], (int) mac_addr[3],
2267                                 (int) mac_addr[4], (int) mac_addr[5]);
2268 
2269             info->has_hardware_address = true;
2270         }
2271 
2272         if (ifa->ifa_addr &&
2273             ifa->ifa_addr->sa_family == AF_INET) {
2274             /* interface with IPv4 address */
2275             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
2276             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
2277                 error_setg_errno(errp, errno, "inet_ntop failed");
2278                 goto error;
2279             }
2280 
2281             address_item = g_malloc0(sizeof(*address_item));
2282             address_item->ip_address = g_strdup(addr4);
2283             address_item->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
2284 
2285             if (ifa->ifa_netmask) {
2286                 /* Count the number of set bits in netmask.
2287                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2288                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
2289                 address_item->prefix = ctpop32(((uint32_t *) p)[0]);
2290             }
2291         } else if (ifa->ifa_addr &&
2292                    ifa->ifa_addr->sa_family == AF_INET6) {
2293             /* interface with IPv6 address */
2294             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
2295             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
2296                 error_setg_errno(errp, errno, "inet_ntop failed");
2297                 goto error;
2298             }
2299 
2300             address_item = g_malloc0(sizeof(*address_item));
2301             address_item->ip_address = g_strdup(addr6);
2302             address_item->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
2303 
2304             if (ifa->ifa_netmask) {
2305                 /* Count the number of set bits in netmask.
2306                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2307                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
2308                 address_item->prefix =
2309                     ctpop32(((uint32_t *) p)[0]) +
2310                     ctpop32(((uint32_t *) p)[1]) +
2311                     ctpop32(((uint32_t *) p)[2]) +
2312                     ctpop32(((uint32_t *) p)[3]);
2313             }
2314         }
2315 
2316         if (!address_item) {
2317             continue;
2318         }
2319 
2320         address_tail = &info->ip_addresses;
2321         while (*address_tail) {
2322             address_tail = &(*address_tail)->next;
2323         }
2324         QAPI_LIST_APPEND(address_tail, address_item);
2325 
2326         info->has_ip_addresses = true;
2327 
2328         if (!info->has_statistics) {
2329             interface_stat = g_malloc0(sizeof(*interface_stat));
2330             if (guest_get_network_stats(info->name, interface_stat) == -1) {
2331                 info->has_statistics = false;
2332                 g_free(interface_stat);
2333             } else {
2334                 info->statistics = interface_stat;
2335                 info->has_statistics = true;
2336             }
2337         }
2338     }
2339 
2340     freeifaddrs(ifap);
2341     return head;
2342 
2343 error:
2344     freeifaddrs(ifap);
2345     qapi_free_GuestNetworkInterfaceList(head);
2346     return NULL;
2347 }
2348 
2349 /* Transfer online/offline status between @vcpu and the guest system.
2350  *
2351  * On input either @errp or *@errp must be NULL.
2352  *
2353  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
2354  * - R: vcpu->logical_id
2355  * - W: vcpu->online
2356  * - W: vcpu->can_offline
2357  *
2358  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
2359  * - R: vcpu->logical_id
2360  * - R: vcpu->online
2361  *
2362  * Written members remain unmodified on error.
2363  */
2364 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
2365                           char *dirpath, Error **errp)
2366 {
2367     int fd;
2368     int res;
2369     int dirfd;
2370     static const char fn[] = "online";
2371 
2372     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2373     if (dirfd == -1) {
2374         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2375         return;
2376     }
2377 
2378     fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
2379     if (fd == -1) {
2380         if (errno != ENOENT) {
2381             error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
2382         } else if (sys2vcpu) {
2383             vcpu->online = true;
2384             vcpu->can_offline = false;
2385         } else if (!vcpu->online) {
2386             error_setg(errp, "logical processor #%" PRId64 " can't be "
2387                        "offlined", vcpu->logical_id);
2388         } /* otherwise pretend successful re-onlining */
2389     } else {
2390         unsigned char status;
2391 
2392         res = pread(fd, &status, 1, 0);
2393         if (res == -1) {
2394             error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
2395         } else if (res == 0) {
2396             error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
2397                        fn);
2398         } else if (sys2vcpu) {
2399             vcpu->online = (status != '0');
2400             vcpu->can_offline = true;
2401         } else if (vcpu->online != (status != '0')) {
2402             status = '0' + vcpu->online;
2403             if (pwrite(fd, &status, 1, 0) == -1) {
2404                 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
2405                                  fn);
2406             }
2407         } /* otherwise pretend successful re-(on|off)-lining */
2408 
2409         res = close(fd);
2410         g_assert(res == 0);
2411     }
2412 
2413     res = close(dirfd);
2414     g_assert(res == 0);
2415 }
2416 
2417 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2418 {
2419     GuestLogicalProcessorList *head, **tail;
2420     const char *cpu_dir = "/sys/devices/system/cpu";
2421     const gchar *line;
2422     g_autoptr(GDir) cpu_gdir = NULL;
2423     Error *local_err = NULL;
2424 
2425     head = NULL;
2426     tail = &head;
2427     cpu_gdir = g_dir_open(cpu_dir, 0, NULL);
2428 
2429     if (cpu_gdir == NULL) {
2430         error_setg_errno(errp, errno, "failed to list entries: %s", cpu_dir);
2431         return NULL;
2432     }
2433 
2434     while (local_err == NULL && (line = g_dir_read_name(cpu_gdir)) != NULL) {
2435         GuestLogicalProcessor *vcpu;
2436         int64_t id;
2437         if (sscanf(line, "cpu%" PRId64, &id)) {
2438             g_autofree char *path = g_strdup_printf("/sys/devices/system/cpu/"
2439                                                     "cpu%" PRId64 "/", id);
2440             vcpu = g_malloc0(sizeof *vcpu);
2441             vcpu->logical_id = id;
2442             vcpu->has_can_offline = true; /* lolspeak ftw */
2443             transfer_vcpu(vcpu, true, path, &local_err);
2444             QAPI_LIST_APPEND(tail, vcpu);
2445         }
2446     }
2447 
2448     if (local_err == NULL) {
2449         /* there's no guest with zero VCPUs */
2450         g_assert(head != NULL);
2451         return head;
2452     }
2453 
2454     qapi_free_GuestLogicalProcessorList(head);
2455     error_propagate(errp, local_err);
2456     return NULL;
2457 }
2458 
2459 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2460 {
2461     int64_t processed;
2462     Error *local_err = NULL;
2463 
2464     processed = 0;
2465     while (vcpus != NULL) {
2466         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2467                                      vcpus->value->logical_id);
2468 
2469         transfer_vcpu(vcpus->value, false, path, &local_err);
2470         g_free(path);
2471         if (local_err != NULL) {
2472             break;
2473         }
2474         ++processed;
2475         vcpus = vcpus->next;
2476     }
2477 
2478     if (local_err != NULL) {
2479         if (processed == 0) {
2480             error_propagate(errp, local_err);
2481         } else {
2482             error_free(local_err);
2483         }
2484     }
2485 
2486     return processed;
2487 }
2488 
2489 void qmp_guest_set_user_password(const char *username,
2490                                  const char *password,
2491                                  bool crypted,
2492                                  Error **errp)
2493 {
2494     Error *local_err = NULL;
2495     char *passwd_path = NULL;
2496     pid_t pid;
2497     int status;
2498     int datafd[2] = { -1, -1 };
2499     char *rawpasswddata = NULL;
2500     size_t rawpasswdlen;
2501     char *chpasswddata = NULL;
2502     size_t chpasswdlen;
2503 
2504     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2505     if (!rawpasswddata) {
2506         return;
2507     }
2508     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2509     rawpasswddata[rawpasswdlen] = '\0';
2510 
2511     if (strchr(rawpasswddata, '\n')) {
2512         error_setg(errp, "forbidden characters in raw password");
2513         goto out;
2514     }
2515 
2516     if (strchr(username, '\n') ||
2517         strchr(username, ':')) {
2518         error_setg(errp, "forbidden characters in username");
2519         goto out;
2520     }
2521 
2522     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2523     chpasswdlen = strlen(chpasswddata);
2524 
2525     passwd_path = g_find_program_in_path("chpasswd");
2526 
2527     if (!passwd_path) {
2528         error_setg(errp, "cannot find 'passwd' program in PATH");
2529         goto out;
2530     }
2531 
2532     if (pipe(datafd) < 0) {
2533         error_setg(errp, "cannot create pipe FDs");
2534         goto out;
2535     }
2536 
2537     pid = fork();
2538     if (pid == 0) {
2539         close(datafd[1]);
2540         /* child */
2541         setsid();
2542         dup2(datafd[0], 0);
2543         reopen_fd_to_null(1);
2544         reopen_fd_to_null(2);
2545 
2546         if (crypted) {
2547             execl(passwd_path, "chpasswd", "-e", NULL);
2548         } else {
2549             execl(passwd_path, "chpasswd", NULL);
2550         }
2551         _exit(EXIT_FAILURE);
2552     } else if (pid < 0) {
2553         error_setg_errno(errp, errno, "failed to create child process");
2554         goto out;
2555     }
2556     close(datafd[0]);
2557     datafd[0] = -1;
2558 
2559     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2560         error_setg_errno(errp, errno, "cannot write new account password");
2561         goto out;
2562     }
2563     close(datafd[1]);
2564     datafd[1] = -1;
2565 
2566     ga_wait_child(pid, &status, &local_err);
2567     if (local_err) {
2568         error_propagate(errp, local_err);
2569         goto out;
2570     }
2571 
2572     if (!WIFEXITED(status)) {
2573         error_setg(errp, "child process has terminated abnormally");
2574         goto out;
2575     }
2576 
2577     if (WEXITSTATUS(status)) {
2578         error_setg(errp, "child process has failed to set user password");
2579         goto out;
2580     }
2581 
2582 out:
2583     g_free(chpasswddata);
2584     g_free(rawpasswddata);
2585     g_free(passwd_path);
2586     if (datafd[0] != -1) {
2587         close(datafd[0]);
2588     }
2589     if (datafd[1] != -1) {
2590         close(datafd[1]);
2591     }
2592 }
2593 
2594 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2595                                int size, Error **errp)
2596 {
2597     int fd;
2598     int res;
2599 
2600     errno = 0;
2601     fd = openat(dirfd, pathname, O_RDONLY);
2602     if (fd == -1) {
2603         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2604         return;
2605     }
2606 
2607     res = pread(fd, buf, size, 0);
2608     if (res == -1) {
2609         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2610     } else if (res == 0) {
2611         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2612     }
2613     close(fd);
2614 }
2615 
2616 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2617                                 const char *buf, int size, Error **errp)
2618 {
2619     int fd;
2620 
2621     errno = 0;
2622     fd = openat(dirfd, pathname, O_WRONLY);
2623     if (fd == -1) {
2624         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2625         return;
2626     }
2627 
2628     if (pwrite(fd, buf, size, 0) == -1) {
2629         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2630     }
2631 
2632     close(fd);
2633 }
2634 
2635 /* Transfer online/offline status between @mem_blk and the guest system.
2636  *
2637  * On input either @errp or *@errp must be NULL.
2638  *
2639  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2640  * - R: mem_blk->phys_index
2641  * - W: mem_blk->online
2642  * - W: mem_blk->can_offline
2643  *
2644  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2645  * - R: mem_blk->phys_index
2646  * - R: mem_blk->online
2647  *-  R: mem_blk->can_offline
2648  * Written members remain unmodified on error.
2649  */
2650 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2651                                   GuestMemoryBlockResponse *result,
2652                                   Error **errp)
2653 {
2654     char *dirpath;
2655     int dirfd;
2656     char *status;
2657     Error *local_err = NULL;
2658 
2659     if (!sys2memblk) {
2660         DIR *dp;
2661 
2662         if (!result) {
2663             error_setg(errp, "Internal error, 'result' should not be NULL");
2664             return;
2665         }
2666         errno = 0;
2667         dp = opendir("/sys/devices/system/memory/");
2668          /* if there is no 'memory' directory in sysfs,
2669          * we think this VM does not support online/offline memory block,
2670          * any other solution?
2671          */
2672         if (!dp) {
2673             if (errno == ENOENT) {
2674                 result->response =
2675                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2676             }
2677             goto out1;
2678         }
2679         closedir(dp);
2680     }
2681 
2682     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2683                               mem_blk->phys_index);
2684     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2685     if (dirfd == -1) {
2686         if (sys2memblk) {
2687             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2688         } else {
2689             if (errno == ENOENT) {
2690                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2691             } else {
2692                 result->response =
2693                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2694             }
2695         }
2696         g_free(dirpath);
2697         goto out1;
2698     }
2699     g_free(dirpath);
2700 
2701     status = g_malloc0(10);
2702     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2703     if (local_err) {
2704         /* treat with sysfs file that not exist in old kernel */
2705         if (errno == ENOENT) {
2706             error_free(local_err);
2707             if (sys2memblk) {
2708                 mem_blk->online = true;
2709                 mem_blk->can_offline = false;
2710             } else if (!mem_blk->online) {
2711                 result->response =
2712                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2713             }
2714         } else {
2715             if (sys2memblk) {
2716                 error_propagate(errp, local_err);
2717             } else {
2718                 error_free(local_err);
2719                 result->response =
2720                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2721             }
2722         }
2723         goto out2;
2724     }
2725 
2726     if (sys2memblk) {
2727         char removable = '0';
2728 
2729         mem_blk->online = (strncmp(status, "online", 6) == 0);
2730 
2731         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2732         if (local_err) {
2733             /* if no 'removable' file, it doesn't support offline mem blk */
2734             if (errno == ENOENT) {
2735                 error_free(local_err);
2736                 mem_blk->can_offline = false;
2737             } else {
2738                 error_propagate(errp, local_err);
2739             }
2740         } else {
2741             mem_blk->can_offline = (removable != '0');
2742         }
2743     } else {
2744         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2745             const char *new_state = mem_blk->online ? "online" : "offline";
2746 
2747             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2748                                 &local_err);
2749             if (local_err) {
2750                 error_free(local_err);
2751                 result->response =
2752                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2753                 goto out2;
2754             }
2755 
2756             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2757             result->has_error_code = false;
2758         } /* otherwise pretend successful re-(on|off)-lining */
2759     }
2760     g_free(status);
2761     close(dirfd);
2762     return;
2763 
2764 out2:
2765     g_free(status);
2766     close(dirfd);
2767 out1:
2768     if (!sys2memblk) {
2769         result->has_error_code = true;
2770         result->error_code = errno;
2771     }
2772 }
2773 
2774 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2775 {
2776     GuestMemoryBlockList *head, **tail;
2777     Error *local_err = NULL;
2778     struct dirent *de;
2779     DIR *dp;
2780 
2781     head = NULL;
2782     tail = &head;
2783 
2784     dp = opendir("/sys/devices/system/memory/");
2785     if (!dp) {
2786         /* it's ok if this happens to be a system that doesn't expose
2787          * memory blocks via sysfs, but otherwise we should report
2788          * an error
2789          */
2790         if (errno != ENOENT) {
2791             error_setg_errno(errp, errno, "Can't open directory"
2792                              "\"/sys/devices/system/memory/\"");
2793         }
2794         return NULL;
2795     }
2796 
2797     /* Note: the phys_index of memory block may be discontinuous,
2798      * this is because a memblk is the unit of the Sparse Memory design, which
2799      * allows discontinuous memory ranges (ex. NUMA), so here we should
2800      * traverse the memory block directory.
2801      */
2802     while ((de = readdir(dp)) != NULL) {
2803         GuestMemoryBlock *mem_blk;
2804 
2805         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2806             !(de->d_type & DT_DIR)) {
2807             continue;
2808         }
2809 
2810         mem_blk = g_malloc0(sizeof *mem_blk);
2811         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2812         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2813         mem_blk->has_can_offline = true; /* lolspeak ftw */
2814         transfer_memory_block(mem_blk, true, NULL, &local_err);
2815         if (local_err) {
2816             break;
2817         }
2818 
2819         QAPI_LIST_APPEND(tail, mem_blk);
2820     }
2821 
2822     closedir(dp);
2823     if (local_err == NULL) {
2824         /* there's no guest with zero memory blocks */
2825         if (head == NULL) {
2826             error_setg(errp, "guest reported zero memory blocks!");
2827         }
2828         return head;
2829     }
2830 
2831     qapi_free_GuestMemoryBlockList(head);
2832     error_propagate(errp, local_err);
2833     return NULL;
2834 }
2835 
2836 GuestMemoryBlockResponseList *
2837 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2838 {
2839     GuestMemoryBlockResponseList *head, **tail;
2840     Error *local_err = NULL;
2841 
2842     head = NULL;
2843     tail = &head;
2844 
2845     while (mem_blks != NULL) {
2846         GuestMemoryBlockResponse *result;
2847         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2848 
2849         result = g_malloc0(sizeof(*result));
2850         result->phys_index = current_mem_blk->phys_index;
2851         transfer_memory_block(current_mem_blk, false, result, &local_err);
2852         if (local_err) { /* should never happen */
2853             goto err;
2854         }
2855 
2856         QAPI_LIST_APPEND(tail, result);
2857         mem_blks = mem_blks->next;
2858     }
2859 
2860     return head;
2861 err:
2862     qapi_free_GuestMemoryBlockResponseList(head);
2863     error_propagate(errp, local_err);
2864     return NULL;
2865 }
2866 
2867 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2868 {
2869     Error *local_err = NULL;
2870     char *dirpath;
2871     int dirfd;
2872     char *buf;
2873     GuestMemoryBlockInfo *info;
2874 
2875     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2876     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2877     if (dirfd == -1) {
2878         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2879         g_free(dirpath);
2880         return NULL;
2881     }
2882     g_free(dirpath);
2883 
2884     buf = g_malloc0(20);
2885     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2886     close(dirfd);
2887     if (local_err) {
2888         g_free(buf);
2889         error_propagate(errp, local_err);
2890         return NULL;
2891     }
2892 
2893     info = g_new0(GuestMemoryBlockInfo, 1);
2894     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2895 
2896     g_free(buf);
2897 
2898     return info;
2899 }
2900 
2901 #else /* defined(__linux__) */
2902 
2903 void qmp_guest_suspend_disk(Error **errp)
2904 {
2905     error_setg(errp, QERR_UNSUPPORTED);
2906 }
2907 
2908 void qmp_guest_suspend_ram(Error **errp)
2909 {
2910     error_setg(errp, QERR_UNSUPPORTED);
2911 }
2912 
2913 void qmp_guest_suspend_hybrid(Error **errp)
2914 {
2915     error_setg(errp, QERR_UNSUPPORTED);
2916 }
2917 
2918 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2919 {
2920     error_setg(errp, QERR_UNSUPPORTED);
2921     return NULL;
2922 }
2923 
2924 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2925 {
2926     error_setg(errp, QERR_UNSUPPORTED);
2927     return NULL;
2928 }
2929 
2930 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2931 {
2932     error_setg(errp, QERR_UNSUPPORTED);
2933     return -1;
2934 }
2935 
2936 void qmp_guest_set_user_password(const char *username,
2937                                  const char *password,
2938                                  bool crypted,
2939                                  Error **errp)
2940 {
2941     error_setg(errp, QERR_UNSUPPORTED);
2942 }
2943 
2944 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2945 {
2946     error_setg(errp, QERR_UNSUPPORTED);
2947     return NULL;
2948 }
2949 
2950 GuestMemoryBlockResponseList *
2951 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2952 {
2953     error_setg(errp, QERR_UNSUPPORTED);
2954     return NULL;
2955 }
2956 
2957 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2958 {
2959     error_setg(errp, QERR_UNSUPPORTED);
2960     return NULL;
2961 }
2962 
2963 #endif
2964 
2965 #if !defined(CONFIG_FSFREEZE)
2966 
2967 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
2968 {
2969     error_setg(errp, QERR_UNSUPPORTED);
2970     return NULL;
2971 }
2972 
2973 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
2974 {
2975     error_setg(errp, QERR_UNSUPPORTED);
2976 
2977     return 0;
2978 }
2979 
2980 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
2981 {
2982     error_setg(errp, QERR_UNSUPPORTED);
2983 
2984     return 0;
2985 }
2986 
2987 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
2988                                        strList *mountpoints,
2989                                        Error **errp)
2990 {
2991     error_setg(errp, QERR_UNSUPPORTED);
2992 
2993     return 0;
2994 }
2995 
2996 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
2997 {
2998     error_setg(errp, QERR_UNSUPPORTED);
2999 
3000     return 0;
3001 }
3002 
3003 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
3004 {
3005     error_setg(errp, QERR_UNSUPPORTED);
3006     return NULL;
3007 }
3008 
3009 #endif /* CONFIG_FSFREEZE */
3010 
3011 #if !defined(CONFIG_FSTRIM)
3012 GuestFilesystemTrimResponse *
3013 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
3014 {
3015     error_setg(errp, QERR_UNSUPPORTED);
3016     return NULL;
3017 }
3018 #endif
3019 
3020 /* add unsupported commands to the blacklist */
3021 GList *ga_command_blacklist_init(GList *blacklist)
3022 {
3023 #if !defined(__linux__)
3024     {
3025         const char *list[] = {
3026             "guest-suspend-disk", "guest-suspend-ram",
3027             "guest-suspend-hybrid", "guest-network-get-interfaces",
3028             "guest-get-vcpus", "guest-set-vcpus",
3029             "guest-get-memory-blocks", "guest-set-memory-blocks",
3030             "guest-get-memory-block-size", "guest-get-memory-block-info",
3031             NULL};
3032         char **p = (char **)list;
3033 
3034         while (*p) {
3035             blacklist = g_list_append(blacklist, g_strdup(*p++));
3036         }
3037     }
3038 #endif
3039 
3040 #if !defined(CONFIG_FSFREEZE)
3041     {
3042         const char *list[] = {
3043             "guest-get-fsinfo", "guest-fsfreeze-status",
3044             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
3045             "guest-fsfreeze-thaw", "guest-get-fsinfo",
3046             "guest-get-disks", NULL};
3047         char **p = (char **)list;
3048 
3049         while (*p) {
3050             blacklist = g_list_append(blacklist, g_strdup(*p++));
3051         }
3052     }
3053 #endif
3054 
3055 #if !defined(CONFIG_FSTRIM)
3056     blacklist = g_list_append(blacklist, g_strdup("guest-fstrim"));
3057 #endif
3058 
3059     blacklist = g_list_append(blacklist, g_strdup("guest-get-devices"));
3060 
3061     return blacklist;
3062 }
3063 
3064 /* register init/cleanup routines for stateful command groups */
3065 void ga_command_state_init(GAState *s, GACommandState *cs)
3066 {
3067 #if defined(CONFIG_FSFREEZE)
3068     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
3069 #endif
3070 }
3071 
3072 #ifdef HAVE_UTMPX
3073 
3074 #define QGA_MICRO_SECOND_TO_SECOND 1000000
3075 
3076 static double ga_get_login_time(struct utmpx *user_info)
3077 {
3078     double seconds = (double)user_info->ut_tv.tv_sec;
3079     double useconds = (double)user_info->ut_tv.tv_usec;
3080     useconds /= QGA_MICRO_SECOND_TO_SECOND;
3081     return seconds + useconds;
3082 }
3083 
3084 GuestUserList *qmp_guest_get_users(Error **errp)
3085 {
3086     GHashTable *cache = NULL;
3087     GuestUserList *head = NULL, **tail = &head;
3088     struct utmpx *user_info = NULL;
3089     gpointer value = NULL;
3090     GuestUser *user = NULL;
3091     double login_time = 0;
3092 
3093     cache = g_hash_table_new(g_str_hash, g_str_equal);
3094     setutxent();
3095 
3096     for (;;) {
3097         user_info = getutxent();
3098         if (user_info == NULL) {
3099             break;
3100         } else if (user_info->ut_type != USER_PROCESS) {
3101             continue;
3102         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
3103             value = g_hash_table_lookup(cache, user_info->ut_user);
3104             user = (GuestUser *)value;
3105             login_time = ga_get_login_time(user_info);
3106             /* We're ensuring the earliest login time to be sent */
3107             if (login_time < user->login_time) {
3108                 user->login_time = login_time;
3109             }
3110             continue;
3111         }
3112 
3113         user = g_new0(GuestUser, 1);
3114         user->user = g_strdup(user_info->ut_user);
3115         user->login_time = ga_get_login_time(user_info);
3116 
3117         g_hash_table_insert(cache, user->user, user);
3118 
3119         QAPI_LIST_APPEND(tail, user);
3120     }
3121     endutxent();
3122     g_hash_table_destroy(cache);
3123     return head;
3124 }
3125 
3126 #else
3127 
3128 GuestUserList *qmp_guest_get_users(Error **errp)
3129 {
3130     error_setg(errp, QERR_UNSUPPORTED);
3131     return NULL;
3132 }
3133 
3134 #endif
3135 
3136 /* Replace escaped special characters with theire real values. The replacement
3137  * is done in place -- returned value is in the original string.
3138  */
3139 static void ga_osrelease_replace_special(gchar *value)
3140 {
3141     gchar *p, *p2, quote;
3142 
3143     /* Trim the string at first space or semicolon if it is not enclosed in
3144      * single or double quotes. */
3145     if ((value[0] != '"') || (value[0] == '\'')) {
3146         p = strchr(value, ' ');
3147         if (p != NULL) {
3148             *p = 0;
3149         }
3150         p = strchr(value, ';');
3151         if (p != NULL) {
3152             *p = 0;
3153         }
3154         return;
3155     }
3156 
3157     quote = value[0];
3158     p2 = value;
3159     p = value + 1;
3160     while (*p != 0) {
3161         if (*p == '\\') {
3162             p++;
3163             switch (*p) {
3164             case '$':
3165             case '\'':
3166             case '"':
3167             case '\\':
3168             case '`':
3169                 break;
3170             default:
3171                 /* Keep literal backslash followed by whatever is there */
3172                 p--;
3173                 break;
3174             }
3175         } else if (*p == quote) {
3176             *p2 = 0;
3177             break;
3178         }
3179         *(p2++) = *(p++);
3180     }
3181 }
3182 
3183 static GKeyFile *ga_parse_osrelease(const char *fname)
3184 {
3185     gchar *content = NULL;
3186     gchar *content2 = NULL;
3187     GError *err = NULL;
3188     GKeyFile *keys = g_key_file_new();
3189     const char *group = "[os-release]\n";
3190 
3191     if (!g_file_get_contents(fname, &content, NULL, &err)) {
3192         slog("failed to read '%s', error: %s", fname, err->message);
3193         goto fail;
3194     }
3195 
3196     if (!g_utf8_validate(content, -1, NULL)) {
3197         slog("file is not utf-8 encoded: %s", fname);
3198         goto fail;
3199     }
3200     content2 = g_strdup_printf("%s%s", group, content);
3201 
3202     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
3203                                    &err)) {
3204         slog("failed to parse file '%s', error: %s", fname, err->message);
3205         goto fail;
3206     }
3207 
3208     g_free(content);
3209     g_free(content2);
3210     return keys;
3211 
3212 fail:
3213     g_error_free(err);
3214     g_free(content);
3215     g_free(content2);
3216     g_key_file_free(keys);
3217     return NULL;
3218 }
3219 
3220 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
3221 {
3222     GuestOSInfo *info = NULL;
3223     struct utsname kinfo;
3224     GKeyFile *osrelease = NULL;
3225     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
3226 
3227     info = g_new0(GuestOSInfo, 1);
3228 
3229     if (uname(&kinfo) != 0) {
3230         error_setg_errno(errp, errno, "uname failed");
3231     } else {
3232         info->has_kernel_version = true;
3233         info->kernel_version = g_strdup(kinfo.version);
3234         info->has_kernel_release = true;
3235         info->kernel_release = g_strdup(kinfo.release);
3236         info->has_machine = true;
3237         info->machine = g_strdup(kinfo.machine);
3238     }
3239 
3240     if (qga_os_release != NULL) {
3241         osrelease = ga_parse_osrelease(qga_os_release);
3242     } else {
3243         osrelease = ga_parse_osrelease("/etc/os-release");
3244         if (osrelease == NULL) {
3245             osrelease = ga_parse_osrelease("/usr/lib/os-release");
3246         }
3247     }
3248 
3249     if (osrelease != NULL) {
3250         char *value;
3251 
3252 #define GET_FIELD(field, osfield) do { \
3253     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
3254     if (value != NULL) { \
3255         ga_osrelease_replace_special(value); \
3256         info->has_ ## field = true; \
3257         info->field = value; \
3258     } \
3259 } while (0)
3260         GET_FIELD(id, "ID");
3261         GET_FIELD(name, "NAME");
3262         GET_FIELD(pretty_name, "PRETTY_NAME");
3263         GET_FIELD(version, "VERSION");
3264         GET_FIELD(version_id, "VERSION_ID");
3265         GET_FIELD(variant, "VARIANT");
3266         GET_FIELD(variant_id, "VARIANT_ID");
3267 #undef GET_FIELD
3268 
3269         g_key_file_free(osrelease);
3270     }
3271 
3272     return info;
3273 }
3274 
3275 GuestDeviceInfoList *qmp_guest_get_devices(Error **errp)
3276 {
3277     error_setg(errp, QERR_UNSUPPORTED);
3278 
3279     return NULL;
3280 }
3281 
3282 #ifndef HOST_NAME_MAX
3283 # ifdef _POSIX_HOST_NAME_MAX
3284 #  define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
3285 # else
3286 #  define HOST_NAME_MAX 255
3287 # endif
3288 #endif
3289 
3290 char *qga_get_host_name(Error **errp)
3291 {
3292     long len = -1;
3293     g_autofree char *hostname = NULL;
3294 
3295 #ifdef _SC_HOST_NAME_MAX
3296     len = sysconf(_SC_HOST_NAME_MAX);
3297 #endif /* _SC_HOST_NAME_MAX */
3298 
3299     if (len < 0) {
3300         len = HOST_NAME_MAX;
3301     }
3302 
3303     /* Unfortunately, gethostname() below does not guarantee a
3304      * NULL terminated string. Therefore, allocate one byte more
3305      * to be sure. */
3306     hostname = g_new0(char, len + 1);
3307 
3308     if (gethostname(hostname, len) < 0) {
3309         error_setg_errno(errp, errno,
3310                          "cannot get hostname");
3311         return NULL;
3312     }
3313 
3314     return g_steal_pointer(&hostname);
3315 }
3316