xref: /openbmc/qemu/qga/commands-posix.c (revision f6e33708)
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 "qga-qapi-commands.h"
20 #include "qapi/error.h"
21 #include "qapi/qmp/qerror.h"
22 #include "qemu/host-utils.h"
23 #include "qemu/sockets.h"
24 #include "qemu/base64.h"
25 #include "qemu/cutils.h"
26 #include "commands-common.h"
27 #include "block/nvme.h"
28 #include "cutils.h"
29 
30 #ifdef HAVE_UTMPX
31 #include <utmpx.h>
32 #endif
33 
34 #if defined(__linux__)
35 #include <mntent.h>
36 #include <sys/statvfs.h>
37 #include <linux/nvme_ioctl.h>
38 
39 #ifdef CONFIG_LIBUDEV
40 #include <libudev.h>
41 #endif
42 #endif
43 
44 #ifdef HAVE_GETIFADDRS
45 #include <arpa/inet.h>
46 #include <sys/socket.h>
47 #include <net/if.h>
48 #if defined(__NetBSD__) || defined(__OpenBSD__) || defined(CONFIG_SOLARIS)
49 #include <net/if_arp.h>
50 #include <netinet/if_ether.h>
51 #if !defined(ETHER_ADDR_LEN) && defined(ETHERADDRL)
52 #define ETHER_ADDR_LEN ETHERADDRL
53 #endif
54 #else
55 #include <net/ethernet.h>
56 #endif
57 #ifdef CONFIG_SOLARIS
58 #include <sys/sockio.h>
59 #endif
60 #endif
61 
62 static void ga_wait_child(pid_t pid, int *status, Error **errp)
63 {
64     pid_t rpid;
65 
66     *status = 0;
67 
68     rpid = RETRY_ON_EINTR(waitpid(pid, status, 0));
69 
70     if (rpid == -1) {
71         error_setg_errno(errp, errno, "failed to wait for child (pid: %d)",
72                          pid);
73         return;
74     }
75 
76     g_assert(rpid == pid);
77 }
78 
79 void qmp_guest_shutdown(const char *mode, Error **errp)
80 {
81     const char *shutdown_flag;
82     Error *local_err = NULL;
83     pid_t pid;
84     int status;
85 
86 #ifdef CONFIG_SOLARIS
87     const char *powerdown_flag = "-i5";
88     const char *halt_flag = "-i0";
89     const char *reboot_flag = "-i6";
90 #elif defined(CONFIG_BSD)
91     const char *powerdown_flag = "-p";
92     const char *halt_flag = "-h";
93     const char *reboot_flag = "-r";
94 #else
95     const char *powerdown_flag = "-P";
96     const char *halt_flag = "-H";
97     const char *reboot_flag = "-r";
98 #endif
99 
100     slog("guest-shutdown called, mode: %s", mode);
101     if (!mode || strcmp(mode, "powerdown") == 0) {
102         shutdown_flag = powerdown_flag;
103     } else if (strcmp(mode, "halt") == 0) {
104         shutdown_flag = halt_flag;
105     } else if (strcmp(mode, "reboot") == 0) {
106         shutdown_flag = reboot_flag;
107     } else {
108         error_setg(errp,
109                    "mode is invalid (valid values are: halt|powerdown|reboot");
110         return;
111     }
112 
113     pid = fork();
114     if (pid == 0) {
115         /* child, start the shutdown */
116         setsid();
117         reopen_fd_to_null(0);
118         reopen_fd_to_null(1);
119         reopen_fd_to_null(2);
120 
121 #ifdef CONFIG_SOLARIS
122         execl("/sbin/shutdown", "shutdown", shutdown_flag, "-g0", "-y",
123               "hypervisor initiated shutdown", (char *)NULL);
124 #elif defined(CONFIG_BSD)
125         execl("/sbin/shutdown", "shutdown", shutdown_flag, "+0",
126                "hypervisor initiated shutdown", (char *)NULL);
127 #else
128         execl("/sbin/shutdown", "shutdown", "-h", shutdown_flag, "+0",
129                "hypervisor initiated shutdown", (char *)NULL);
130 #endif
131         _exit(EXIT_FAILURE);
132     } else if (pid < 0) {
133         error_setg_errno(errp, errno, "failed to create child process");
134         return;
135     }
136 
137     ga_wait_child(pid, &status, &local_err);
138     if (local_err) {
139         error_propagate(errp, local_err);
140         return;
141     }
142 
143     if (!WIFEXITED(status)) {
144         error_setg(errp, "child process has terminated abnormally");
145         return;
146     }
147 
148     if (WEXITSTATUS(status)) {
149         error_setg(errp, "child process has failed to shutdown");
150         return;
151     }
152 
153     /* succeeded */
154 }
155 
156 void qmp_guest_set_time(bool has_time, int64_t time_ns, Error **errp)
157 {
158     int ret;
159     int status;
160     pid_t pid;
161     Error *local_err = NULL;
162     struct timeval tv;
163     static const char hwclock_path[] = "/sbin/hwclock";
164     static int hwclock_available = -1;
165 
166     if (hwclock_available < 0) {
167         hwclock_available = (access(hwclock_path, X_OK) == 0);
168     }
169 
170     if (!hwclock_available) {
171         error_setg(errp, QERR_UNSUPPORTED);
172         return;
173     }
174 
175     /* If user has passed a time, validate and set it. */
176     if (has_time) {
177         GDate date = { 0, };
178 
179         /* year-2038 will overflow in case time_t is 32bit */
180         if (time_ns / 1000000000 != (time_t)(time_ns / 1000000000)) {
181             error_setg(errp, "Time %" PRId64 " is too large", time_ns);
182             return;
183         }
184 
185         tv.tv_sec = time_ns / 1000000000;
186         tv.tv_usec = (time_ns % 1000000000) / 1000;
187         g_date_set_time_t(&date, tv.tv_sec);
188         if (date.year < 1970 || date.year >= 2070) {
189             error_setg_errno(errp, errno, "Invalid time");
190             return;
191         }
192 
193         ret = settimeofday(&tv, NULL);
194         if (ret < 0) {
195             error_setg_errno(errp, errno, "Failed to set time to guest");
196             return;
197         }
198     }
199 
200     /* Now, if user has passed a time to set and the system time is set, we
201      * just need to synchronize the hardware clock. However, if no time was
202      * passed, user is requesting the opposite: set the system time from the
203      * hardware clock (RTC). */
204     pid = fork();
205     if (pid == 0) {
206         setsid();
207         reopen_fd_to_null(0);
208         reopen_fd_to_null(1);
209         reopen_fd_to_null(2);
210 
211         /* Use '/sbin/hwclock -w' to set RTC from the system time,
212          * or '/sbin/hwclock -s' to set the system time from RTC. */
213         execl(hwclock_path, "hwclock", has_time ? "-w" : "-s", NULL);
214         _exit(EXIT_FAILURE);
215     } else if (pid < 0) {
216         error_setg_errno(errp, errno, "failed to create child process");
217         return;
218     }
219 
220     ga_wait_child(pid, &status, &local_err);
221     if (local_err) {
222         error_propagate(errp, local_err);
223         return;
224     }
225 
226     if (!WIFEXITED(status)) {
227         error_setg(errp, "child process has terminated abnormally");
228         return;
229     }
230 
231     if (WEXITSTATUS(status)) {
232         error_setg(errp, "hwclock failed to set hardware clock to system time");
233         return;
234     }
235 }
236 
237 typedef enum {
238     RW_STATE_NEW,
239     RW_STATE_READING,
240     RW_STATE_WRITING,
241 } RwState;
242 
243 struct GuestFileHandle {
244     uint64_t id;
245     FILE *fh;
246     RwState state;
247     QTAILQ_ENTRY(GuestFileHandle) next;
248 };
249 
250 static struct {
251     QTAILQ_HEAD(, GuestFileHandle) filehandles;
252 } guest_file_state = {
253     .filehandles = QTAILQ_HEAD_INITIALIZER(guest_file_state.filehandles),
254 };
255 
256 static int64_t guest_file_handle_add(FILE *fh, Error **errp)
257 {
258     GuestFileHandle *gfh;
259     int64_t handle;
260 
261     handle = ga_get_fd_handle(ga_state, errp);
262     if (handle < 0) {
263         return -1;
264     }
265 
266     gfh = g_new0(GuestFileHandle, 1);
267     gfh->id = handle;
268     gfh->fh = fh;
269     QTAILQ_INSERT_TAIL(&guest_file_state.filehandles, gfh, next);
270 
271     return handle;
272 }
273 
274 GuestFileHandle *guest_file_handle_find(int64_t id, Error **errp)
275 {
276     GuestFileHandle *gfh;
277 
278     QTAILQ_FOREACH(gfh, &guest_file_state.filehandles, next)
279     {
280         if (gfh->id == id) {
281             return gfh;
282         }
283     }
284 
285     error_setg(errp, "handle '%" PRId64 "' has not been found", id);
286     return NULL;
287 }
288 
289 typedef const char * const ccpc;
290 
291 #ifndef O_BINARY
292 #define O_BINARY 0
293 #endif
294 
295 /* http://pubs.opengroup.org/onlinepubs/9699919799/functions/fopen.html */
296 static const struct {
297     ccpc *forms;
298     int oflag_base;
299 } guest_file_open_modes[] = {
300     { (ccpc[]){ "r",          NULL }, O_RDONLY                                 },
301     { (ccpc[]){ "rb",         NULL }, O_RDONLY                      | O_BINARY },
302     { (ccpc[]){ "w",          NULL }, O_WRONLY | O_CREAT | O_TRUNC             },
303     { (ccpc[]){ "wb",         NULL }, O_WRONLY | O_CREAT | O_TRUNC  | O_BINARY },
304     { (ccpc[]){ "a",          NULL }, O_WRONLY | O_CREAT | O_APPEND            },
305     { (ccpc[]){ "ab",         NULL }, O_WRONLY | O_CREAT | O_APPEND | O_BINARY },
306     { (ccpc[]){ "r+",         NULL }, O_RDWR                                   },
307     { (ccpc[]){ "rb+", "r+b", NULL }, O_RDWR                        | O_BINARY },
308     { (ccpc[]){ "w+",         NULL }, O_RDWR   | O_CREAT | O_TRUNC             },
309     { (ccpc[]){ "wb+", "w+b", NULL }, O_RDWR   | O_CREAT | O_TRUNC  | O_BINARY },
310     { (ccpc[]){ "a+",         NULL }, O_RDWR   | O_CREAT | O_APPEND            },
311     { (ccpc[]){ "ab+", "a+b", NULL }, O_RDWR   | O_CREAT | O_APPEND | O_BINARY }
312 };
313 
314 static int
315 find_open_flag(const char *mode_str, Error **errp)
316 {
317     unsigned mode;
318 
319     for (mode = 0; mode < ARRAY_SIZE(guest_file_open_modes); ++mode) {
320         ccpc *form;
321 
322         form = guest_file_open_modes[mode].forms;
323         while (*form != NULL && strcmp(*form, mode_str) != 0) {
324             ++form;
325         }
326         if (*form != NULL) {
327             break;
328         }
329     }
330 
331     if (mode == ARRAY_SIZE(guest_file_open_modes)) {
332         error_setg(errp, "invalid file open mode '%s'", mode_str);
333         return -1;
334     }
335     return guest_file_open_modes[mode].oflag_base | O_NOCTTY | O_NONBLOCK;
336 }
337 
338 #define DEFAULT_NEW_FILE_MODE (S_IRUSR | S_IWUSR | \
339                                S_IRGRP | S_IWGRP | \
340                                S_IROTH | S_IWOTH)
341 
342 static FILE *
343 safe_open_or_create(const char *path, const char *mode, Error **errp)
344 {
345     int oflag;
346     int fd = -1;
347     FILE *f = NULL;
348 
349     oflag = find_open_flag(mode, errp);
350     if (oflag < 0) {
351         goto end;
352     }
353 
354     /* If the caller wants / allows creation of a new file, we implement it
355      * with a two step process: open() + (open() / fchmod()).
356      *
357      * First we insist on creating the file exclusively as a new file. If
358      * that succeeds, we're free to set any file-mode bits on it. (The
359      * motivation is that we want to set those file-mode bits independently
360      * of the current umask.)
361      *
362      * If the exclusive creation fails because the file already exists
363      * (EEXIST is not possible for any other reason), we just attempt to
364      * open the file, but in this case we won't be allowed to change the
365      * file-mode bits on the preexistent file.
366      *
367      * The pathname should never disappear between the two open()s in
368      * practice. If it happens, then someone very likely tried to race us.
369      * In this case just go ahead and report the ENOENT from the second
370      * open() to the caller.
371      *
372      * If the caller wants to open a preexistent file, then the first
373      * open() is decisive and its third argument is ignored, and the second
374      * open() and the fchmod() are never called.
375      */
376     fd = qga_open_cloexec(path, oflag | ((oflag & O_CREAT) ? O_EXCL : 0), 0);
377     if (fd == -1 && errno == EEXIST) {
378         oflag &= ~(unsigned)O_CREAT;
379         fd = qga_open_cloexec(path, oflag, 0);
380     }
381     if (fd == -1) {
382         error_setg_errno(errp, errno,
383                          "failed to open file '%s' (mode: '%s')",
384                          path, mode);
385         goto end;
386     }
387 
388     if ((oflag & O_CREAT) && fchmod(fd, DEFAULT_NEW_FILE_MODE) == -1) {
389         error_setg_errno(errp, errno, "failed to set permission "
390                          "0%03o on new file '%s' (mode: '%s')",
391                          (unsigned)DEFAULT_NEW_FILE_MODE, path, mode);
392         goto end;
393     }
394 
395     f = fdopen(fd, mode);
396     if (f == NULL) {
397         error_setg_errno(errp, errno, "failed to associate stdio stream with "
398                          "file descriptor %d, file '%s' (mode: '%s')",
399                          fd, path, mode);
400     }
401 
402 end:
403     if (f == NULL && fd != -1) {
404         close(fd);
405         if (oflag & O_CREAT) {
406             unlink(path);
407         }
408     }
409     return f;
410 }
411 
412 int64_t qmp_guest_file_open(const char *path, const char *mode,
413                             Error **errp)
414 {
415     FILE *fh;
416     Error *local_err = NULL;
417     int64_t handle;
418 
419     if (!mode) {
420         mode = "r";
421     }
422     slog("guest-file-open called, filepath: %s, mode: %s", path, mode);
423     fh = safe_open_or_create(path, mode, &local_err);
424     if (local_err != NULL) {
425         error_propagate(errp, local_err);
426         return -1;
427     }
428 
429     /* set fd non-blocking to avoid common use cases (like reading from a
430      * named pipe) from hanging the agent
431      */
432     if (!g_unix_set_fd_nonblocking(fileno(fh), true, NULL)) {
433         fclose(fh);
434         error_setg_errno(errp, errno, "Failed to set FD nonblocking");
435         return -1;
436     }
437 
438     handle = guest_file_handle_add(fh, errp);
439     if (handle < 0) {
440         fclose(fh);
441         return -1;
442     }
443 
444     slog("guest-file-open, handle: %" PRId64, handle);
445     return handle;
446 }
447 
448 void qmp_guest_file_close(int64_t handle, Error **errp)
449 {
450     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
451     int ret;
452 
453     slog("guest-file-close called, handle: %" PRId64, handle);
454     if (!gfh) {
455         return;
456     }
457 
458     ret = fclose(gfh->fh);
459     if (ret == EOF) {
460         error_setg_errno(errp, errno, "failed to close handle");
461         return;
462     }
463 
464     QTAILQ_REMOVE(&guest_file_state.filehandles, gfh, next);
465     g_free(gfh);
466 }
467 
468 GuestFileRead *guest_file_read_unsafe(GuestFileHandle *gfh,
469                                       int64_t count, Error **errp)
470 {
471     GuestFileRead *read_data = NULL;
472     guchar *buf;
473     FILE *fh = gfh->fh;
474     size_t read_count;
475 
476     /* explicitly flush when switching from writing to reading */
477     if (gfh->state == RW_STATE_WRITING) {
478         int ret = fflush(fh);
479         if (ret == EOF) {
480             error_setg_errno(errp, errno, "failed to flush file");
481             return NULL;
482         }
483         gfh->state = RW_STATE_NEW;
484     }
485 
486     buf = g_malloc0(count + 1);
487     read_count = fread(buf, 1, count, fh);
488     if (ferror(fh)) {
489         error_setg_errno(errp, errno, "failed to read file");
490     } else {
491         buf[read_count] = 0;
492         read_data = g_new0(GuestFileRead, 1);
493         read_data->count = read_count;
494         read_data->eof = feof(fh);
495         if (read_count) {
496             read_data->buf_b64 = g_base64_encode(buf, read_count);
497         }
498         gfh->state = RW_STATE_READING;
499     }
500     g_free(buf);
501     clearerr(fh);
502 
503     return read_data;
504 }
505 
506 GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
507                                      bool has_count, int64_t count,
508                                      Error **errp)
509 {
510     GuestFileWrite *write_data = NULL;
511     guchar *buf;
512     gsize buf_len;
513     int write_count;
514     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
515     FILE *fh;
516 
517     if (!gfh) {
518         return NULL;
519     }
520 
521     fh = gfh->fh;
522 
523     if (gfh->state == RW_STATE_READING) {
524         int ret = fseek(fh, 0, SEEK_CUR);
525         if (ret == -1) {
526             error_setg_errno(errp, errno, "failed to seek file");
527             return NULL;
528         }
529         gfh->state = RW_STATE_NEW;
530     }
531 
532     buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
533     if (!buf) {
534         return NULL;
535     }
536 
537     if (!has_count) {
538         count = buf_len;
539     } else if (count < 0 || count > buf_len) {
540         error_setg(errp, "value '%" PRId64 "' is invalid for argument count",
541                    count);
542         g_free(buf);
543         return NULL;
544     }
545 
546     write_count = fwrite(buf, 1, count, fh);
547     if (ferror(fh)) {
548         error_setg_errno(errp, errno, "failed to write to file");
549         slog("guest-file-write failed, handle: %" PRId64, handle);
550     } else {
551         write_data = g_new0(GuestFileWrite, 1);
552         write_data->count = write_count;
553         write_data->eof = feof(fh);
554         gfh->state = RW_STATE_WRITING;
555     }
556     g_free(buf);
557     clearerr(fh);
558 
559     return write_data;
560 }
561 
562 struct GuestFileSeek *qmp_guest_file_seek(int64_t handle, int64_t offset,
563                                           GuestFileWhence *whence_code,
564                                           Error **errp)
565 {
566     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
567     GuestFileSeek *seek_data = NULL;
568     FILE *fh;
569     int ret;
570     int whence;
571     Error *err = NULL;
572 
573     if (!gfh) {
574         return NULL;
575     }
576 
577     /* We stupidly exposed 'whence':'int' in our qapi */
578     whence = ga_parse_whence(whence_code, &err);
579     if (err) {
580         error_propagate(errp, err);
581         return NULL;
582     }
583 
584     fh = gfh->fh;
585     ret = fseek(fh, offset, whence);
586     if (ret == -1) {
587         error_setg_errno(errp, errno, "failed to seek file");
588         if (errno == ESPIPE) {
589             /* file is non-seekable, stdio shouldn't be buffering anyways */
590             gfh->state = RW_STATE_NEW;
591         }
592     } else {
593         seek_data = g_new0(GuestFileSeek, 1);
594         seek_data->position = ftell(fh);
595         seek_data->eof = feof(fh);
596         gfh->state = RW_STATE_NEW;
597     }
598     clearerr(fh);
599 
600     return seek_data;
601 }
602 
603 void qmp_guest_file_flush(int64_t handle, Error **errp)
604 {
605     GuestFileHandle *gfh = guest_file_handle_find(handle, errp);
606     FILE *fh;
607     int ret;
608 
609     if (!gfh) {
610         return;
611     }
612 
613     fh = gfh->fh;
614     ret = fflush(fh);
615     if (ret == EOF) {
616         error_setg_errno(errp, errno, "failed to flush file");
617     } else {
618         gfh->state = RW_STATE_NEW;
619     }
620 }
621 
622 #if defined(CONFIG_FSFREEZE) || defined(CONFIG_FSTRIM)
623 void free_fs_mount_list(FsMountList *mounts)
624 {
625      FsMount *mount, *temp;
626 
627      if (!mounts) {
628          return;
629      }
630 
631      QTAILQ_FOREACH_SAFE(mount, mounts, next, temp) {
632          QTAILQ_REMOVE(mounts, mount, next);
633          g_free(mount->dirname);
634          g_free(mount->devtype);
635          g_free(mount);
636      }
637 }
638 #endif
639 
640 #if defined(CONFIG_FSFREEZE)
641 typedef enum {
642     FSFREEZE_HOOK_THAW = 0,
643     FSFREEZE_HOOK_FREEZE,
644 } FsfreezeHookArg;
645 
646 static const char *fsfreeze_hook_arg_string[] = {
647     "thaw",
648     "freeze",
649 };
650 
651 static void execute_fsfreeze_hook(FsfreezeHookArg arg, Error **errp)
652 {
653     int status;
654     pid_t pid;
655     const char *hook;
656     const char *arg_str = fsfreeze_hook_arg_string[arg];
657     Error *local_err = NULL;
658 
659     hook = ga_fsfreeze_hook(ga_state);
660     if (!hook) {
661         return;
662     }
663     if (access(hook, X_OK) != 0) {
664         error_setg_errno(errp, errno, "can't access fsfreeze hook '%s'", hook);
665         return;
666     }
667 
668     slog("executing fsfreeze hook with arg '%s'", arg_str);
669     pid = fork();
670     if (pid == 0) {
671         setsid();
672         reopen_fd_to_null(0);
673         reopen_fd_to_null(1);
674         reopen_fd_to_null(2);
675 
676         execl(hook, hook, arg_str, NULL);
677         _exit(EXIT_FAILURE);
678     } else if (pid < 0) {
679         error_setg_errno(errp, errno, "failed to create child process");
680         return;
681     }
682 
683     ga_wait_child(pid, &status, &local_err);
684     if (local_err) {
685         error_propagate(errp, local_err);
686         return;
687     }
688 
689     if (!WIFEXITED(status)) {
690         error_setg(errp, "fsfreeze hook has terminated abnormally");
691         return;
692     }
693 
694     status = WEXITSTATUS(status);
695     if (status) {
696         error_setg(errp, "fsfreeze hook has failed with status %d", status);
697         return;
698     }
699 }
700 
701 /*
702  * Return status of freeze/thaw
703  */
704 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
705 {
706     if (ga_is_frozen(ga_state)) {
707         return GUEST_FSFREEZE_STATUS_FROZEN;
708     }
709 
710     return GUEST_FSFREEZE_STATUS_THAWED;
711 }
712 
713 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
714 {
715     return qmp_guest_fsfreeze_freeze_list(false, NULL, errp);
716 }
717 
718 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
719                                        strList *mountpoints,
720                                        Error **errp)
721 {
722     int ret;
723     FsMountList mounts;
724     Error *local_err = NULL;
725 
726     slog("guest-fsfreeze called");
727 
728     execute_fsfreeze_hook(FSFREEZE_HOOK_FREEZE, &local_err);
729     if (local_err) {
730         error_propagate(errp, local_err);
731         return -1;
732     }
733 
734     QTAILQ_INIT(&mounts);
735     if (!build_fs_mount_list(&mounts, &local_err)) {
736         error_propagate(errp, local_err);
737         return -1;
738     }
739 
740     /* cannot risk guest agent blocking itself on a write in this state */
741     ga_set_frozen(ga_state);
742 
743     ret = qmp_guest_fsfreeze_do_freeze_list(has_mountpoints, mountpoints,
744                                             mounts, errp);
745 
746     free_fs_mount_list(&mounts);
747     /* We may not issue any FIFREEZE here.
748      * Just unset ga_state here and ready for the next call.
749      */
750     if (ret == 0) {
751         ga_unset_frozen(ga_state);
752     } else if (ret < 0) {
753         qmp_guest_fsfreeze_thaw(NULL);
754     }
755     return ret;
756 }
757 
758 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
759 {
760     int ret;
761 
762     ret = qmp_guest_fsfreeze_do_thaw(errp);
763     if (ret >= 0) {
764         ga_unset_frozen(ga_state);
765         execute_fsfreeze_hook(FSFREEZE_HOOK_THAW, errp);
766     } else {
767         ret = 0;
768     }
769 
770     return ret;
771 }
772 
773 static void guest_fsfreeze_cleanup(void)
774 {
775     Error *err = NULL;
776 
777     if (ga_is_frozen(ga_state) == GUEST_FSFREEZE_STATUS_FROZEN) {
778         qmp_guest_fsfreeze_thaw(&err);
779         if (err) {
780             slog("failed to clean up frozen filesystems: %s",
781                  error_get_pretty(err));
782             error_free(err);
783         }
784     }
785 }
786 #endif
787 
788 /* linux-specific implementations. avoid this if at all possible. */
789 #if defined(__linux__)
790 #if defined(CONFIG_FSFREEZE)
791 
792 static char *get_pci_driver(char const *syspath, int pathlen, Error **errp)
793 {
794     char *path;
795     char *dpath;
796     char *driver = NULL;
797     char buf[PATH_MAX];
798     ssize_t len;
799 
800     path = g_strndup(syspath, pathlen);
801     dpath = g_strdup_printf("%s/driver", path);
802     len = readlink(dpath, buf, sizeof(buf) - 1);
803     if (len != -1) {
804         buf[len] = 0;
805         driver = g_path_get_basename(buf);
806     }
807     g_free(dpath);
808     g_free(path);
809     return driver;
810 }
811 
812 static int compare_uint(const void *_a, const void *_b)
813 {
814     unsigned int a = *(unsigned int *)_a;
815     unsigned int b = *(unsigned int *)_b;
816 
817     return a < b ? -1 : a > b ? 1 : 0;
818 }
819 
820 /* Walk the specified sysfs and build a sorted list of host or ata numbers */
821 static int build_hosts(char const *syspath, char const *host, bool ata,
822                        unsigned int *hosts, int hosts_max, Error **errp)
823 {
824     char *path;
825     DIR *dir;
826     struct dirent *entry;
827     int i = 0;
828 
829     path = g_strndup(syspath, host - syspath);
830     dir = opendir(path);
831     if (!dir) {
832         error_setg_errno(errp, errno, "opendir(\"%s\")", path);
833         g_free(path);
834         return -1;
835     }
836 
837     while (i < hosts_max) {
838         entry = readdir(dir);
839         if (!entry) {
840             break;
841         }
842         if (ata && sscanf(entry->d_name, "ata%d", hosts + i) == 1) {
843             ++i;
844         } else if (!ata && sscanf(entry->d_name, "host%d", hosts + i) == 1) {
845             ++i;
846         }
847     }
848 
849     qsort(hosts, i, sizeof(hosts[0]), compare_uint);
850 
851     g_free(path);
852     closedir(dir);
853     return i;
854 }
855 
856 /*
857  * Store disk device info for devices on the PCI bus.
858  * Returns true if information has been stored, or false for failure.
859  */
860 static bool build_guest_fsinfo_for_pci_dev(char const *syspath,
861                                            GuestDiskAddress *disk,
862                                            Error **errp)
863 {
864     unsigned int pci[4], host, hosts[8], tgt[3];
865     int i, nhosts = 0, pcilen;
866     GuestPCIAddress *pciaddr = disk->pci_controller;
867     bool has_ata = false, has_host = false, has_tgt = false;
868     char *p, *q, *driver = NULL;
869     bool ret = false;
870 
871     p = strstr(syspath, "/devices/pci");
872     if (!p || sscanf(p + 12, "%*x:%*x/%x:%x:%x.%x%n",
873                      pci, pci + 1, pci + 2, pci + 3, &pcilen) < 4) {
874         g_debug("only pci device is supported: sysfs path '%s'", syspath);
875         return false;
876     }
877 
878     p += 12 + pcilen;
879     while (true) {
880         driver = get_pci_driver(syspath, p - syspath, errp);
881         if (driver && (g_str_equal(driver, "ata_piix") ||
882                        g_str_equal(driver, "sym53c8xx") ||
883                        g_str_equal(driver, "virtio-pci") ||
884                        g_str_equal(driver, "ahci") ||
885                        g_str_equal(driver, "nvme") ||
886                        g_str_equal(driver, "xhci_hcd") ||
887                        g_str_equal(driver, "ehci-pci"))) {
888             break;
889         }
890 
891         g_free(driver);
892         if (sscanf(p, "/%x:%x:%x.%x%n",
893                           pci, pci + 1, pci + 2, pci + 3, &pcilen) == 4) {
894             p += pcilen;
895             continue;
896         }
897 
898         g_debug("unsupported driver or sysfs path '%s'", syspath);
899         return false;
900     }
901 
902     p = strstr(syspath, "/target");
903     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
904                     tgt, tgt + 1, tgt + 2) == 3) {
905         has_tgt = true;
906     }
907 
908     p = strstr(syspath, "/ata");
909     if (p) {
910         q = p + 4;
911         has_ata = true;
912     } else {
913         p = strstr(syspath, "/host");
914         q = p + 5;
915     }
916     if (p && sscanf(q, "%u", &host) == 1) {
917         has_host = true;
918         nhosts = build_hosts(syspath, p, has_ata, hosts,
919                              ARRAY_SIZE(hosts), errp);
920         if (nhosts < 0) {
921             goto cleanup;
922         }
923     }
924 
925     pciaddr->domain = pci[0];
926     pciaddr->bus = pci[1];
927     pciaddr->slot = pci[2];
928     pciaddr->function = pci[3];
929 
930     if (strcmp(driver, "ata_piix") == 0) {
931         /* a host per ide bus, target*:0:<unit>:0 */
932         if (!has_host || !has_tgt) {
933             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
934             goto cleanup;
935         }
936         for (i = 0; i < nhosts; i++) {
937             if (host == hosts[i]) {
938                 disk->bus_type = GUEST_DISK_BUS_TYPE_IDE;
939                 disk->bus = i;
940                 disk->unit = tgt[1];
941                 break;
942             }
943         }
944         if (i >= nhosts) {
945             g_debug("no host for '%s' (driver '%s')", syspath, driver);
946             goto cleanup;
947         }
948     } else if (strcmp(driver, "sym53c8xx") == 0) {
949         /* scsi(LSI Logic): target*:0:<unit>:0 */
950         if (!has_tgt) {
951             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
952             goto cleanup;
953         }
954         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
955         disk->unit = tgt[1];
956     } else if (strcmp(driver, "virtio-pci") == 0) {
957         if (has_tgt) {
958             /* virtio-scsi: target*:0:0:<unit> */
959             disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
960             disk->unit = tgt[2];
961         } else {
962             /* virtio-blk: 1 disk per 1 device */
963             disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
964         }
965     } else if (strcmp(driver, "ahci") == 0) {
966         /* ahci: 1 host per 1 unit */
967         if (!has_host || !has_tgt) {
968             g_debug("invalid sysfs path '%s' (driver '%s')", syspath, driver);
969             goto cleanup;
970         }
971         for (i = 0; i < nhosts; i++) {
972             if (host == hosts[i]) {
973                 disk->unit = i;
974                 disk->bus_type = GUEST_DISK_BUS_TYPE_SATA;
975                 break;
976             }
977         }
978         if (i >= nhosts) {
979             g_debug("no host for '%s' (driver '%s')", syspath, driver);
980             goto cleanup;
981         }
982     } else if (strcmp(driver, "nvme") == 0) {
983         disk->bus_type = GUEST_DISK_BUS_TYPE_NVME;
984     } else if (strcmp(driver, "ehci-pci") == 0 || strcmp(driver, "xhci_hcd") == 0) {
985         disk->bus_type = GUEST_DISK_BUS_TYPE_USB;
986     } else {
987         g_debug("unknown driver '%s' (sysfs path '%s')", driver, syspath);
988         goto cleanup;
989     }
990 
991     ret = true;
992 
993 cleanup:
994     g_free(driver);
995     return ret;
996 }
997 
998 /*
999  * Store disk device info for non-PCI virtio devices (for example s390x
1000  * channel I/O devices). Returns true if information has been stored, or
1001  * false for failure.
1002  */
1003 static bool build_guest_fsinfo_for_nonpci_virtio(char const *syspath,
1004                                                  GuestDiskAddress *disk,
1005                                                  Error **errp)
1006 {
1007     unsigned int tgt[3];
1008     char *p;
1009 
1010     if (!strstr(syspath, "/virtio") || !strstr(syspath, "/block")) {
1011         g_debug("Unsupported virtio device '%s'", syspath);
1012         return false;
1013     }
1014 
1015     p = strstr(syspath, "/target");
1016     if (p && sscanf(p + 7, "%*u:%*u:%*u/%*u:%u:%u:%u",
1017                     &tgt[0], &tgt[1], &tgt[2]) == 3) {
1018         /* virtio-scsi: target*:0:<target>:<unit> */
1019         disk->bus_type = GUEST_DISK_BUS_TYPE_SCSI;
1020         disk->bus = tgt[0];
1021         disk->target = tgt[1];
1022         disk->unit = tgt[2];
1023     } else {
1024         /* virtio-blk: 1 disk per 1 device */
1025         disk->bus_type = GUEST_DISK_BUS_TYPE_VIRTIO;
1026     }
1027 
1028     return true;
1029 }
1030 
1031 /*
1032  * Store disk device info for CCW devices (s390x channel I/O devices).
1033  * Returns true if information has been stored, or false for failure.
1034  */
1035 static bool build_guest_fsinfo_for_ccw_dev(char const *syspath,
1036                                            GuestDiskAddress *disk,
1037                                            Error **errp)
1038 {
1039     unsigned int cssid, ssid, subchno, devno;
1040     char *p;
1041 
1042     p = strstr(syspath, "/devices/css");
1043     if (!p || sscanf(p + 12, "%*x/%x.%x.%x/%*x.%*x.%x/",
1044                      &cssid, &ssid, &subchno, &devno) < 4) {
1045         g_debug("could not parse ccw device sysfs path: %s", syspath);
1046         return false;
1047     }
1048 
1049     disk->ccw_address = g_new0(GuestCCWAddress, 1);
1050     disk->ccw_address->cssid = cssid;
1051     disk->ccw_address->ssid = ssid;
1052     disk->ccw_address->subchno = subchno;
1053     disk->ccw_address->devno = devno;
1054 
1055     if (strstr(p, "/virtio")) {
1056         build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1057     }
1058 
1059     return true;
1060 }
1061 
1062 /* Store disk device info specified by @sysfs into @fs */
1063 static void build_guest_fsinfo_for_real_device(char const *syspath,
1064                                                GuestFilesystemInfo *fs,
1065                                                Error **errp)
1066 {
1067     GuestDiskAddress *disk;
1068     GuestPCIAddress *pciaddr;
1069     bool has_hwinf;
1070 #ifdef CONFIG_LIBUDEV
1071     struct udev *udev = NULL;
1072     struct udev_device *udevice = NULL;
1073 #endif
1074 
1075     pciaddr = g_new0(GuestPCIAddress, 1);
1076     pciaddr->domain = -1;                       /* -1 means field is invalid */
1077     pciaddr->bus = -1;
1078     pciaddr->slot = -1;
1079     pciaddr->function = -1;
1080 
1081     disk = g_new0(GuestDiskAddress, 1);
1082     disk->pci_controller = pciaddr;
1083     disk->bus_type = GUEST_DISK_BUS_TYPE_UNKNOWN;
1084 
1085 #ifdef CONFIG_LIBUDEV
1086     udev = udev_new();
1087     udevice = udev_device_new_from_syspath(udev, syspath);
1088     if (udev == NULL || udevice == NULL) {
1089         g_debug("failed to query udev");
1090     } else {
1091         const char *devnode, *serial;
1092         devnode = udev_device_get_devnode(udevice);
1093         if (devnode != NULL) {
1094             disk->dev = g_strdup(devnode);
1095         }
1096         serial = udev_device_get_property_value(udevice, "ID_SERIAL");
1097         if (serial != NULL && *serial != 0) {
1098             disk->serial = g_strdup(serial);
1099         }
1100     }
1101 
1102     udev_unref(udev);
1103     udev_device_unref(udevice);
1104 #endif
1105 
1106     if (strstr(syspath, "/devices/pci")) {
1107         has_hwinf = build_guest_fsinfo_for_pci_dev(syspath, disk, errp);
1108     } else if (strstr(syspath, "/devices/css")) {
1109         has_hwinf = build_guest_fsinfo_for_ccw_dev(syspath, disk, errp);
1110     } else if (strstr(syspath, "/virtio")) {
1111         has_hwinf = build_guest_fsinfo_for_nonpci_virtio(syspath, disk, errp);
1112     } else {
1113         g_debug("Unsupported device type for '%s'", syspath);
1114         has_hwinf = false;
1115     }
1116 
1117     if (has_hwinf || disk->dev || disk->serial) {
1118         QAPI_LIST_PREPEND(fs->disk, disk);
1119     } else {
1120         qapi_free_GuestDiskAddress(disk);
1121     }
1122 }
1123 
1124 static void build_guest_fsinfo_for_device(char const *devpath,
1125                                           GuestFilesystemInfo *fs,
1126                                           Error **errp);
1127 
1128 /* Store a list of slave devices of virtual volume specified by @syspath into
1129  * @fs */
1130 static void build_guest_fsinfo_for_virtual_device(char const *syspath,
1131                                                   GuestFilesystemInfo *fs,
1132                                                   Error **errp)
1133 {
1134     Error *err = NULL;
1135     DIR *dir;
1136     char *dirpath;
1137     struct dirent *entry;
1138 
1139     dirpath = g_strdup_printf("%s/slaves", syspath);
1140     dir = opendir(dirpath);
1141     if (!dir) {
1142         if (errno != ENOENT) {
1143             error_setg_errno(errp, errno, "opendir(\"%s\")", dirpath);
1144         }
1145         g_free(dirpath);
1146         return;
1147     }
1148 
1149     for (;;) {
1150         errno = 0;
1151         entry = readdir(dir);
1152         if (entry == NULL) {
1153             if (errno) {
1154                 error_setg_errno(errp, errno, "readdir(\"%s\")", dirpath);
1155             }
1156             break;
1157         }
1158 
1159         if (entry->d_type == DT_LNK) {
1160             char *path;
1161 
1162             g_debug(" slave device '%s'", entry->d_name);
1163             path = g_strdup_printf("%s/slaves/%s", syspath, entry->d_name);
1164             build_guest_fsinfo_for_device(path, fs, &err);
1165             g_free(path);
1166 
1167             if (err) {
1168                 error_propagate(errp, err);
1169                 break;
1170             }
1171         }
1172     }
1173 
1174     g_free(dirpath);
1175     closedir(dir);
1176 }
1177 
1178 static bool is_disk_virtual(const char *devpath, Error **errp)
1179 {
1180     g_autofree char *syspath = realpath(devpath, NULL);
1181 
1182     if (!syspath) {
1183         error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1184         return false;
1185     }
1186     return strstr(syspath, "/devices/virtual/block/") != NULL;
1187 }
1188 
1189 /* Dispatch to functions for virtual/real device */
1190 static void build_guest_fsinfo_for_device(char const *devpath,
1191                                           GuestFilesystemInfo *fs,
1192                                           Error **errp)
1193 {
1194     ERRP_GUARD();
1195     g_autofree char *syspath = NULL;
1196     bool is_virtual = false;
1197 
1198     syspath = realpath(devpath, NULL);
1199     if (!syspath) {
1200         if (errno != ENOENT) {
1201             error_setg_errno(errp, errno, "realpath(\"%s\")", devpath);
1202             return;
1203         }
1204 
1205         /* ENOENT: This devpath may not exist because of container config */
1206         if (!fs->name) {
1207             fs->name = g_path_get_basename(devpath);
1208         }
1209         return;
1210     }
1211 
1212     if (!fs->name) {
1213         fs->name = g_path_get_basename(syspath);
1214     }
1215 
1216     g_debug("  parse sysfs path '%s'", syspath);
1217     is_virtual = is_disk_virtual(syspath, errp);
1218     if (*errp != NULL) {
1219         return;
1220     }
1221     if (is_virtual) {
1222         build_guest_fsinfo_for_virtual_device(syspath, fs, errp);
1223     } else {
1224         build_guest_fsinfo_for_real_device(syspath, fs, errp);
1225     }
1226 }
1227 
1228 #ifdef CONFIG_LIBUDEV
1229 
1230 /*
1231  * Wrapper around build_guest_fsinfo_for_device() for getting just
1232  * the disk address.
1233  */
1234 static GuestDiskAddress *get_disk_address(const char *syspath, Error **errp)
1235 {
1236     g_autoptr(GuestFilesystemInfo) fs = NULL;
1237 
1238     fs = g_new0(GuestFilesystemInfo, 1);
1239     build_guest_fsinfo_for_device(syspath, fs, errp);
1240     if (fs->disk != NULL) {
1241         return g_steal_pointer(&fs->disk->value);
1242     }
1243     return NULL;
1244 }
1245 
1246 static char *get_alias_for_syspath(const char *syspath)
1247 {
1248     struct udev *udev = NULL;
1249     struct udev_device *udevice = NULL;
1250     char *ret = NULL;
1251 
1252     udev = udev_new();
1253     if (udev == NULL) {
1254         g_debug("failed to query udev");
1255         goto out;
1256     }
1257     udevice = udev_device_new_from_syspath(udev, syspath);
1258     if (udevice == NULL) {
1259         g_debug("failed to query udev for path: %s", syspath);
1260         goto out;
1261     } else {
1262         const char *alias = udev_device_get_property_value(
1263             udevice, "DM_NAME");
1264         /*
1265          * NULL means there was an error and empty string means there is no
1266          * alias. In case of no alias we return NULL instead of empty string.
1267          */
1268         if (alias == NULL) {
1269             g_debug("failed to query udev for device alias for: %s",
1270                 syspath);
1271         } else if (*alias != 0) {
1272             ret = g_strdup(alias);
1273         }
1274     }
1275 
1276 out:
1277     udev_unref(udev);
1278     udev_device_unref(udevice);
1279     return ret;
1280 }
1281 
1282 static char *get_device_for_syspath(const char *syspath)
1283 {
1284     struct udev *udev = NULL;
1285     struct udev_device *udevice = NULL;
1286     char *ret = NULL;
1287 
1288     udev = udev_new();
1289     if (udev == NULL) {
1290         g_debug("failed to query udev");
1291         goto out;
1292     }
1293     udevice = udev_device_new_from_syspath(udev, syspath);
1294     if (udevice == NULL) {
1295         g_debug("failed to query udev for path: %s", syspath);
1296         goto out;
1297     } else {
1298         ret = g_strdup(udev_device_get_devnode(udevice));
1299     }
1300 
1301 out:
1302     udev_unref(udev);
1303     udev_device_unref(udevice);
1304     return ret;
1305 }
1306 
1307 static void get_disk_deps(const char *disk_dir, GuestDiskInfo *disk)
1308 {
1309     g_autofree char *deps_dir = NULL;
1310     const gchar *dep;
1311     GDir *dp_deps = NULL;
1312 
1313     /* List dependent disks */
1314     deps_dir = g_strdup_printf("%s/slaves", disk_dir);
1315     g_debug("  listing entries in: %s", deps_dir);
1316     dp_deps = g_dir_open(deps_dir, 0, NULL);
1317     if (dp_deps == NULL) {
1318         g_debug("failed to list entries in %s", deps_dir);
1319         return;
1320     }
1321     disk->has_dependencies = true;
1322     while ((dep = g_dir_read_name(dp_deps)) != NULL) {
1323         g_autofree char *dep_dir = NULL;
1324         char *dev_name;
1325 
1326         /* Add dependent disks */
1327         dep_dir = g_strdup_printf("%s/%s", deps_dir, dep);
1328         dev_name = get_device_for_syspath(dep_dir);
1329         if (dev_name != NULL) {
1330             g_debug("  adding dependent device: %s", dev_name);
1331             QAPI_LIST_PREPEND(disk->dependencies, dev_name);
1332         }
1333     }
1334     g_dir_close(dp_deps);
1335 }
1336 
1337 /*
1338  * Detect partitions subdirectory, name is "<disk_name><number>" or
1339  * "<disk_name>p<number>"
1340  *
1341  * @disk_name -- last component of /sys path (e.g. sda)
1342  * @disk_dir -- sys path of the disk (e.g. /sys/block/sda)
1343  * @disk_dev -- device node of the disk (e.g. /dev/sda)
1344  */
1345 static GuestDiskInfoList *get_disk_partitions(
1346     GuestDiskInfoList *list,
1347     const char *disk_name, const char *disk_dir,
1348     const char *disk_dev)
1349 {
1350     GuestDiskInfoList *ret = list;
1351     struct dirent *de_disk;
1352     DIR *dp_disk = NULL;
1353     size_t len = strlen(disk_name);
1354 
1355     dp_disk = opendir(disk_dir);
1356     while ((de_disk = readdir(dp_disk)) != NULL) {
1357         g_autofree char *partition_dir = NULL;
1358         char *dev_name;
1359         GuestDiskInfo *partition;
1360 
1361         if (!(de_disk->d_type & DT_DIR)) {
1362             continue;
1363         }
1364 
1365         if (!(strncmp(disk_name, de_disk->d_name, len) == 0 &&
1366             ((*(de_disk->d_name + len) == 'p' &&
1367             isdigit(*(de_disk->d_name + len + 1))) ||
1368                 isdigit(*(de_disk->d_name + len))))) {
1369             continue;
1370         }
1371 
1372         partition_dir = g_strdup_printf("%s/%s",
1373             disk_dir, de_disk->d_name);
1374         dev_name = get_device_for_syspath(partition_dir);
1375         if (dev_name == NULL) {
1376             g_debug("Failed to get device name for syspath: %s",
1377                 disk_dir);
1378             continue;
1379         }
1380         partition = g_new0(GuestDiskInfo, 1);
1381         partition->name = dev_name;
1382         partition->partition = true;
1383         partition->has_dependencies = true;
1384         /* Add parent disk as dependent for easier tracking of hierarchy */
1385         QAPI_LIST_PREPEND(partition->dependencies, g_strdup(disk_dev));
1386 
1387         QAPI_LIST_PREPEND(ret, partition);
1388     }
1389     closedir(dp_disk);
1390 
1391     return ret;
1392 }
1393 
1394 static void get_nvme_smart(GuestDiskInfo *disk)
1395 {
1396     int fd;
1397     GuestNVMeSmart *smart;
1398     NvmeSmartLog log = {0};
1399     struct nvme_admin_cmd cmd = {
1400         .opcode = NVME_ADM_CMD_GET_LOG_PAGE,
1401         .nsid = NVME_NSID_BROADCAST,
1402         .addr = (uintptr_t)&log,
1403         .data_len = sizeof(log),
1404         .cdw10 = NVME_LOG_SMART_INFO | (1 << 15) /* RAE bit */
1405                  | (((sizeof(log) >> 2) - 1) << 16)
1406     };
1407 
1408     fd = qga_open_cloexec(disk->name, O_RDONLY, 0);
1409     if (fd == -1) {
1410         g_debug("Failed to open device: %s: %s", disk->name, g_strerror(errno));
1411         return;
1412     }
1413 
1414     if (ioctl(fd, NVME_IOCTL_ADMIN_CMD, &cmd)) {
1415         g_debug("Failed to get smart: %s: %s", disk->name, g_strerror(errno));
1416         close(fd);
1417         return;
1418     }
1419 
1420     disk->smart = g_new0(GuestDiskSmart, 1);
1421     disk->smart->type = GUEST_DISK_BUS_TYPE_NVME;
1422 
1423     smart = &disk->smart->u.nvme;
1424     smart->critical_warning = log.critical_warning;
1425     smart->temperature = lduw_le_p(&log.temperature); /* unaligned field */
1426     smart->available_spare = log.available_spare;
1427     smart->available_spare_threshold = log.available_spare_threshold;
1428     smart->percentage_used = log.percentage_used;
1429     smart->data_units_read_lo = le64_to_cpu(log.data_units_read[0]);
1430     smart->data_units_read_hi = le64_to_cpu(log.data_units_read[1]);
1431     smart->data_units_written_lo = le64_to_cpu(log.data_units_written[0]);
1432     smart->data_units_written_hi = le64_to_cpu(log.data_units_written[1]);
1433     smart->host_read_commands_lo = le64_to_cpu(log.host_read_commands[0]);
1434     smart->host_read_commands_hi = le64_to_cpu(log.host_read_commands[1]);
1435     smart->host_write_commands_lo = le64_to_cpu(log.host_write_commands[0]);
1436     smart->host_write_commands_hi = le64_to_cpu(log.host_write_commands[1]);
1437     smart->controller_busy_time_lo = le64_to_cpu(log.controller_busy_time[0]);
1438     smart->controller_busy_time_hi = le64_to_cpu(log.controller_busy_time[1]);
1439     smart->power_cycles_lo = le64_to_cpu(log.power_cycles[0]);
1440     smart->power_cycles_hi = le64_to_cpu(log.power_cycles[1]);
1441     smart->power_on_hours_lo = le64_to_cpu(log.power_on_hours[0]);
1442     smart->power_on_hours_hi = le64_to_cpu(log.power_on_hours[1]);
1443     smart->unsafe_shutdowns_lo = le64_to_cpu(log.unsafe_shutdowns[0]);
1444     smart->unsafe_shutdowns_hi = le64_to_cpu(log.unsafe_shutdowns[1]);
1445     smart->media_errors_lo = le64_to_cpu(log.media_errors[0]);
1446     smart->media_errors_hi = le64_to_cpu(log.media_errors[1]);
1447     smart->number_of_error_log_entries_lo =
1448         le64_to_cpu(log.number_of_error_log_entries[0]);
1449     smart->number_of_error_log_entries_hi =
1450         le64_to_cpu(log.number_of_error_log_entries[1]);
1451 
1452     close(fd);
1453 }
1454 
1455 static void get_disk_smart(GuestDiskInfo *disk)
1456 {
1457     if (disk->address
1458         && (disk->address->bus_type == GUEST_DISK_BUS_TYPE_NVME)) {
1459         get_nvme_smart(disk);
1460     }
1461 }
1462 
1463 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1464 {
1465     GuestDiskInfoList *ret = NULL;
1466     GuestDiskInfo *disk;
1467     DIR *dp = NULL;
1468     struct dirent *de = NULL;
1469 
1470     g_debug("listing /sys/block directory");
1471     dp = opendir("/sys/block");
1472     if (dp == NULL) {
1473         error_setg_errno(errp, errno, "Can't open directory \"/sys/block\"");
1474         return NULL;
1475     }
1476     while ((de = readdir(dp)) != NULL) {
1477         g_autofree char *disk_dir = NULL, *line = NULL,
1478             *size_path = NULL;
1479         char *dev_name;
1480         Error *local_err = NULL;
1481         if (de->d_type != DT_LNK) {
1482             g_debug("  skipping entry: %s", de->d_name);
1483             continue;
1484         }
1485 
1486         /* Check size and skip zero-sized disks */
1487         g_debug("  checking disk size");
1488         size_path = g_strdup_printf("/sys/block/%s/size", de->d_name);
1489         if (!g_file_get_contents(size_path, &line, NULL, NULL)) {
1490             g_debug("  failed to read disk size");
1491             continue;
1492         }
1493         if (g_strcmp0(line, "0\n") == 0) {
1494             g_debug("  skipping zero-sized disk");
1495             continue;
1496         }
1497 
1498         g_debug("  adding %s", de->d_name);
1499         disk_dir = g_strdup_printf("/sys/block/%s", de->d_name);
1500         dev_name = get_device_for_syspath(disk_dir);
1501         if (dev_name == NULL) {
1502             g_debug("Failed to get device name for syspath: %s",
1503                 disk_dir);
1504             continue;
1505         }
1506         disk = g_new0(GuestDiskInfo, 1);
1507         disk->name = dev_name;
1508         disk->partition = false;
1509         disk->alias = get_alias_for_syspath(disk_dir);
1510         QAPI_LIST_PREPEND(ret, disk);
1511 
1512         /* Get address for non-virtual devices */
1513         bool is_virtual = is_disk_virtual(disk_dir, &local_err);
1514         if (local_err != NULL) {
1515             g_debug("  failed to check disk path, ignoring error: %s",
1516                 error_get_pretty(local_err));
1517             error_free(local_err);
1518             local_err = NULL;
1519             /* Don't try to get the address */
1520             is_virtual = true;
1521         }
1522         if (!is_virtual) {
1523             disk->address = get_disk_address(disk_dir, &local_err);
1524             if (local_err != NULL) {
1525                 g_debug("  failed to get device info, ignoring error: %s",
1526                     error_get_pretty(local_err));
1527                 error_free(local_err);
1528                 local_err = NULL;
1529             }
1530         }
1531 
1532         get_disk_deps(disk_dir, disk);
1533         get_disk_smart(disk);
1534         ret = get_disk_partitions(ret, de->d_name, disk_dir, dev_name);
1535     }
1536 
1537     closedir(dp);
1538 
1539     return ret;
1540 }
1541 
1542 #else
1543 
1544 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
1545 {
1546     error_setg(errp, QERR_UNSUPPORTED);
1547     return NULL;
1548 }
1549 
1550 #endif
1551 
1552 /* Return a list of the disk device(s)' info which @mount lies on */
1553 static GuestFilesystemInfo *build_guest_fsinfo(struct FsMount *mount,
1554                                                Error **errp)
1555 {
1556     GuestFilesystemInfo *fs = g_malloc0(sizeof(*fs));
1557     struct statvfs buf;
1558     unsigned long used, nonroot_total, fr_size;
1559     char *devpath = g_strdup_printf("/sys/dev/block/%u:%u",
1560                                     mount->devmajor, mount->devminor);
1561 
1562     fs->mountpoint = g_strdup(mount->dirname);
1563     fs->type = g_strdup(mount->devtype);
1564     build_guest_fsinfo_for_device(devpath, fs, errp);
1565 
1566     if (statvfs(fs->mountpoint, &buf) == 0) {
1567         fr_size = buf.f_frsize;
1568         used = buf.f_blocks - buf.f_bfree;
1569         nonroot_total = used + buf.f_bavail;
1570         fs->used_bytes = used * fr_size;
1571         fs->total_bytes = nonroot_total * fr_size;
1572 
1573         fs->has_total_bytes = true;
1574         fs->has_used_bytes = true;
1575     }
1576 
1577     g_free(devpath);
1578 
1579     return fs;
1580 }
1581 
1582 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
1583 {
1584     FsMountList mounts;
1585     struct FsMount *mount;
1586     GuestFilesystemInfoList *ret = NULL;
1587     Error *local_err = NULL;
1588 
1589     QTAILQ_INIT(&mounts);
1590     if (!build_fs_mount_list(&mounts, &local_err)) {
1591         error_propagate(errp, local_err);
1592         return NULL;
1593     }
1594 
1595     QTAILQ_FOREACH(mount, &mounts, next) {
1596         g_debug("Building guest fsinfo for '%s'", mount->dirname);
1597 
1598         QAPI_LIST_PREPEND(ret, build_guest_fsinfo(mount, &local_err));
1599         if (local_err) {
1600             error_propagate(errp, local_err);
1601             qapi_free_GuestFilesystemInfoList(ret);
1602             ret = NULL;
1603             break;
1604         }
1605     }
1606 
1607     free_fs_mount_list(&mounts);
1608     return ret;
1609 }
1610 #endif /* CONFIG_FSFREEZE */
1611 
1612 #if defined(CONFIG_FSTRIM)
1613 /*
1614  * Walk list of mounted file systems in the guest, and trim them.
1615  */
1616 GuestFilesystemTrimResponse *
1617 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
1618 {
1619     GuestFilesystemTrimResponse *response;
1620     GuestFilesystemTrimResult *result;
1621     int ret = 0;
1622     FsMountList mounts;
1623     struct FsMount *mount;
1624     int fd;
1625     struct fstrim_range r;
1626 
1627     slog("guest-fstrim called");
1628 
1629     QTAILQ_INIT(&mounts);
1630     if (!build_fs_mount_list(&mounts, errp)) {
1631         return NULL;
1632     }
1633 
1634     response = g_malloc0(sizeof(*response));
1635 
1636     QTAILQ_FOREACH(mount, &mounts, next) {
1637         result = g_malloc0(sizeof(*result));
1638         result->path = g_strdup(mount->dirname);
1639 
1640         QAPI_LIST_PREPEND(response->paths, result);
1641 
1642         fd = qga_open_cloexec(mount->dirname, O_RDONLY, 0);
1643         if (fd == -1) {
1644             result->error = g_strdup_printf("failed to open: %s",
1645                                             strerror(errno));
1646             continue;
1647         }
1648 
1649         /* We try to cull filesystems we know won't work in advance, but other
1650          * filesystems may not implement fstrim for less obvious reasons.
1651          * These will report EOPNOTSUPP; while in some other cases ENOTTY
1652          * will be reported (e.g. CD-ROMs).
1653          * Any other error means an unexpected error.
1654          */
1655         r.start = 0;
1656         r.len = -1;
1657         r.minlen = has_minimum ? minimum : 0;
1658         ret = ioctl(fd, FITRIM, &r);
1659         if (ret == -1) {
1660             if (errno == ENOTTY || errno == EOPNOTSUPP) {
1661                 result->error = g_strdup("trim not supported");
1662             } else {
1663                 result->error = g_strdup_printf("failed to trim: %s",
1664                                                 strerror(errno));
1665             }
1666             close(fd);
1667             continue;
1668         }
1669 
1670         result->has_minimum = true;
1671         result->minimum = r.minlen;
1672         result->has_trimmed = true;
1673         result->trimmed = r.len;
1674         close(fd);
1675     }
1676 
1677     free_fs_mount_list(&mounts);
1678     return response;
1679 }
1680 #endif /* CONFIG_FSTRIM */
1681 
1682 
1683 #define LINUX_SYS_STATE_FILE "/sys/power/state"
1684 #define SUSPEND_SUPPORTED 0
1685 #define SUSPEND_NOT_SUPPORTED 1
1686 
1687 typedef enum {
1688     SUSPEND_MODE_DISK = 0,
1689     SUSPEND_MODE_RAM = 1,
1690     SUSPEND_MODE_HYBRID = 2,
1691 } SuspendMode;
1692 
1693 /*
1694  * Executes a command in a child process using g_spawn_sync,
1695  * returning an int >= 0 representing the exit status of the
1696  * process.
1697  *
1698  * If the program wasn't found in path, returns -1.
1699  *
1700  * If a problem happened when creating the child process,
1701  * returns -1 and errp is set.
1702  */
1703 static int run_process_child(const char *command[], Error **errp)
1704 {
1705     int exit_status, spawn_flag;
1706     GError *g_err = NULL;
1707     bool success;
1708 
1709     spawn_flag = G_SPAWN_SEARCH_PATH | G_SPAWN_STDOUT_TO_DEV_NULL |
1710                  G_SPAWN_STDERR_TO_DEV_NULL;
1711 
1712     success =  g_spawn_sync(NULL, (char **)command, NULL, spawn_flag,
1713                             NULL, NULL, NULL, NULL,
1714                             &exit_status, &g_err);
1715 
1716     if (success) {
1717         return WEXITSTATUS(exit_status);
1718     }
1719 
1720     if (g_err && (g_err->code != G_SPAWN_ERROR_NOENT)) {
1721         error_setg(errp, "failed to create child process, error '%s'",
1722                    g_err->message);
1723     }
1724 
1725     g_error_free(g_err);
1726     return -1;
1727 }
1728 
1729 static bool systemd_supports_mode(SuspendMode mode, Error **errp)
1730 {
1731     const char *systemctl_args[3] = {"systemd-hibernate", "systemd-suspend",
1732                                      "systemd-hybrid-sleep"};
1733     const char *cmd[4] = {"systemctl", "status", systemctl_args[mode], NULL};
1734     int status;
1735 
1736     status = run_process_child(cmd, errp);
1737 
1738     /*
1739      * systemctl status uses LSB return codes so we can expect
1740      * status > 0 and be ok. To assert if the guest has support
1741      * for the selected suspend mode, status should be < 4. 4 is
1742      * the code for unknown service status, the return value when
1743      * the service does not exist. A common value is status = 3
1744      * (program is not running).
1745      */
1746     if (status > 0 && status < 4) {
1747         return true;
1748     }
1749 
1750     return false;
1751 }
1752 
1753 static void systemd_suspend(SuspendMode mode, Error **errp)
1754 {
1755     Error *local_err = NULL;
1756     const char *systemctl_args[3] = {"hibernate", "suspend", "hybrid-sleep"};
1757     const char *cmd[3] = {"systemctl", systemctl_args[mode], NULL};
1758     int status;
1759 
1760     status = run_process_child(cmd, &local_err);
1761 
1762     if (status == 0) {
1763         return;
1764     }
1765 
1766     if ((status == -1) && !local_err) {
1767         error_setg(errp, "the helper program 'systemctl %s' was not found",
1768                    systemctl_args[mode]);
1769         return;
1770     }
1771 
1772     if (local_err) {
1773         error_propagate(errp, local_err);
1774     } else {
1775         error_setg(errp, "the helper program 'systemctl %s' returned an "
1776                    "unexpected exit status code (%d)",
1777                    systemctl_args[mode], status);
1778     }
1779 }
1780 
1781 static bool pmutils_supports_mode(SuspendMode mode, Error **errp)
1782 {
1783     Error *local_err = NULL;
1784     const char *pmutils_args[3] = {"--hibernate", "--suspend",
1785                                    "--suspend-hybrid"};
1786     const char *cmd[3] = {"pm-is-supported", pmutils_args[mode], NULL};
1787     int status;
1788 
1789     status = run_process_child(cmd, &local_err);
1790 
1791     if (status == SUSPEND_SUPPORTED) {
1792         return true;
1793     }
1794 
1795     if ((status == -1) && !local_err) {
1796         return false;
1797     }
1798 
1799     if (local_err) {
1800         error_propagate(errp, local_err);
1801     } else {
1802         error_setg(errp,
1803                    "the helper program '%s' returned an unexpected exit"
1804                    " status code (%d)", "pm-is-supported", status);
1805     }
1806 
1807     return false;
1808 }
1809 
1810 static void pmutils_suspend(SuspendMode mode, Error **errp)
1811 {
1812     Error *local_err = NULL;
1813     const char *pmutils_binaries[3] = {"pm-hibernate", "pm-suspend",
1814                                        "pm-suspend-hybrid"};
1815     const char *cmd[2] = {pmutils_binaries[mode], NULL};
1816     int status;
1817 
1818     status = run_process_child(cmd, &local_err);
1819 
1820     if (status == 0) {
1821         return;
1822     }
1823 
1824     if ((status == -1) && !local_err) {
1825         error_setg(errp, "the helper program '%s' was not found",
1826                    pmutils_binaries[mode]);
1827         return;
1828     }
1829 
1830     if (local_err) {
1831         error_propagate(errp, local_err);
1832     } else {
1833         error_setg(errp,
1834                    "the helper program '%s' returned an unexpected exit"
1835                    " status code (%d)", pmutils_binaries[mode], status);
1836     }
1837 }
1838 
1839 static bool linux_sys_state_supports_mode(SuspendMode mode, Error **errp)
1840 {
1841     const char *sysfile_strs[3] = {"disk", "mem", NULL};
1842     const char *sysfile_str = sysfile_strs[mode];
1843     char buf[32]; /* hopefully big enough */
1844     int fd;
1845     ssize_t ret;
1846 
1847     if (!sysfile_str) {
1848         error_setg(errp, "unknown guest suspend mode");
1849         return false;
1850     }
1851 
1852     fd = open(LINUX_SYS_STATE_FILE, O_RDONLY);
1853     if (fd < 0) {
1854         return false;
1855     }
1856 
1857     ret = read(fd, buf, sizeof(buf) - 1);
1858     close(fd);
1859     if (ret <= 0) {
1860         return false;
1861     }
1862     buf[ret] = '\0';
1863 
1864     if (strstr(buf, sysfile_str)) {
1865         return true;
1866     }
1867     return false;
1868 }
1869 
1870 static void linux_sys_state_suspend(SuspendMode mode, Error **errp)
1871 {
1872     Error *local_err = NULL;
1873     const char *sysfile_strs[3] = {"disk", "mem", NULL};
1874     const char *sysfile_str = sysfile_strs[mode];
1875     pid_t pid;
1876     int status;
1877 
1878     if (!sysfile_str) {
1879         error_setg(errp, "unknown guest suspend mode");
1880         return;
1881     }
1882 
1883     pid = fork();
1884     if (!pid) {
1885         /* child */
1886         int fd;
1887 
1888         setsid();
1889         reopen_fd_to_null(0);
1890         reopen_fd_to_null(1);
1891         reopen_fd_to_null(2);
1892 
1893         fd = open(LINUX_SYS_STATE_FILE, O_WRONLY);
1894         if (fd < 0) {
1895             _exit(EXIT_FAILURE);
1896         }
1897 
1898         if (write(fd, sysfile_str, strlen(sysfile_str)) < 0) {
1899             _exit(EXIT_FAILURE);
1900         }
1901 
1902         _exit(EXIT_SUCCESS);
1903     } else if (pid < 0) {
1904         error_setg_errno(errp, errno, "failed to create child process");
1905         return;
1906     }
1907 
1908     ga_wait_child(pid, &status, &local_err);
1909     if (local_err) {
1910         error_propagate(errp, local_err);
1911         return;
1912     }
1913 
1914     if (WEXITSTATUS(status)) {
1915         error_setg(errp, "child process has failed to suspend");
1916     }
1917 
1918 }
1919 
1920 static void guest_suspend(SuspendMode mode, Error **errp)
1921 {
1922     Error *local_err = NULL;
1923     bool mode_supported = false;
1924 
1925     if (systemd_supports_mode(mode, &local_err)) {
1926         mode_supported = true;
1927         systemd_suspend(mode, &local_err);
1928 
1929         if (!local_err) {
1930             return;
1931         }
1932     }
1933 
1934     error_free(local_err);
1935     local_err = NULL;
1936 
1937     if (pmutils_supports_mode(mode, &local_err)) {
1938         mode_supported = true;
1939         pmutils_suspend(mode, &local_err);
1940 
1941         if (!local_err) {
1942             return;
1943         }
1944     }
1945 
1946     error_free(local_err);
1947     local_err = NULL;
1948 
1949     if (linux_sys_state_supports_mode(mode, &local_err)) {
1950         mode_supported = true;
1951         linux_sys_state_suspend(mode, &local_err);
1952     }
1953 
1954     if (!mode_supported) {
1955         error_free(local_err);
1956         error_setg(errp,
1957                    "the requested suspend mode is not supported by the guest");
1958     } else {
1959         error_propagate(errp, local_err);
1960     }
1961 }
1962 
1963 void qmp_guest_suspend_disk(Error **errp)
1964 {
1965     guest_suspend(SUSPEND_MODE_DISK, errp);
1966 }
1967 
1968 void qmp_guest_suspend_ram(Error **errp)
1969 {
1970     guest_suspend(SUSPEND_MODE_RAM, errp);
1971 }
1972 
1973 void qmp_guest_suspend_hybrid(Error **errp)
1974 {
1975     guest_suspend(SUSPEND_MODE_HYBRID, errp);
1976 }
1977 
1978 /* Transfer online/offline status between @vcpu and the guest system.
1979  *
1980  * On input either @errp or *@errp must be NULL.
1981  *
1982  * In system-to-@vcpu direction, the following @vcpu fields are accessed:
1983  * - R: vcpu->logical_id
1984  * - W: vcpu->online
1985  * - W: vcpu->can_offline
1986  *
1987  * In @vcpu-to-system direction, the following @vcpu fields are accessed:
1988  * - R: vcpu->logical_id
1989  * - R: vcpu->online
1990  *
1991  * Written members remain unmodified on error.
1992  */
1993 static void transfer_vcpu(GuestLogicalProcessor *vcpu, bool sys2vcpu,
1994                           char *dirpath, Error **errp)
1995 {
1996     int fd;
1997     int res;
1998     int dirfd;
1999     static const char fn[] = "online";
2000 
2001     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2002     if (dirfd == -1) {
2003         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2004         return;
2005     }
2006 
2007     fd = openat(dirfd, fn, sys2vcpu ? O_RDONLY : O_RDWR);
2008     if (fd == -1) {
2009         if (errno != ENOENT) {
2010             error_setg_errno(errp, errno, "open(\"%s/%s\")", dirpath, fn);
2011         } else if (sys2vcpu) {
2012             vcpu->online = true;
2013             vcpu->can_offline = false;
2014         } else if (!vcpu->online) {
2015             error_setg(errp, "logical processor #%" PRId64 " can't be "
2016                        "offlined", vcpu->logical_id);
2017         } /* otherwise pretend successful re-onlining */
2018     } else {
2019         unsigned char status;
2020 
2021         res = pread(fd, &status, 1, 0);
2022         if (res == -1) {
2023             error_setg_errno(errp, errno, "pread(\"%s/%s\")", dirpath, fn);
2024         } else if (res == 0) {
2025             error_setg(errp, "pread(\"%s/%s\"): unexpected EOF", dirpath,
2026                        fn);
2027         } else if (sys2vcpu) {
2028             vcpu->online = (status != '0');
2029             vcpu->can_offline = true;
2030         } else if (vcpu->online != (status != '0')) {
2031             status = '0' + vcpu->online;
2032             if (pwrite(fd, &status, 1, 0) == -1) {
2033                 error_setg_errno(errp, errno, "pwrite(\"%s/%s\")", dirpath,
2034                                  fn);
2035             }
2036         } /* otherwise pretend successful re-(on|off)-lining */
2037 
2038         res = close(fd);
2039         g_assert(res == 0);
2040     }
2041 
2042     res = close(dirfd);
2043     g_assert(res == 0);
2044 }
2045 
2046 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2047 {
2048     GuestLogicalProcessorList *head, **tail;
2049     const char *cpu_dir = "/sys/devices/system/cpu";
2050     const gchar *line;
2051     g_autoptr(GDir) cpu_gdir = NULL;
2052     Error *local_err = NULL;
2053 
2054     head = NULL;
2055     tail = &head;
2056     cpu_gdir = g_dir_open(cpu_dir, 0, NULL);
2057 
2058     if (cpu_gdir == NULL) {
2059         error_setg_errno(errp, errno, "failed to list entries: %s", cpu_dir);
2060         return NULL;
2061     }
2062 
2063     while (local_err == NULL && (line = g_dir_read_name(cpu_gdir)) != NULL) {
2064         GuestLogicalProcessor *vcpu;
2065         int64_t id;
2066         if (sscanf(line, "cpu%" PRId64, &id)) {
2067             g_autofree char *path = g_strdup_printf("/sys/devices/system/cpu/"
2068                                                     "cpu%" PRId64 "/", id);
2069             vcpu = g_malloc0(sizeof *vcpu);
2070             vcpu->logical_id = id;
2071             vcpu->has_can_offline = true; /* lolspeak ftw */
2072             transfer_vcpu(vcpu, true, path, &local_err);
2073             QAPI_LIST_APPEND(tail, vcpu);
2074         }
2075     }
2076 
2077     if (local_err == NULL) {
2078         /* there's no guest with zero VCPUs */
2079         g_assert(head != NULL);
2080         return head;
2081     }
2082 
2083     qapi_free_GuestLogicalProcessorList(head);
2084     error_propagate(errp, local_err);
2085     return NULL;
2086 }
2087 
2088 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2089 {
2090     int64_t processed;
2091     Error *local_err = NULL;
2092 
2093     processed = 0;
2094     while (vcpus != NULL) {
2095         char *path = g_strdup_printf("/sys/devices/system/cpu/cpu%" PRId64 "/",
2096                                      vcpus->value->logical_id);
2097 
2098         transfer_vcpu(vcpus->value, false, path, &local_err);
2099         g_free(path);
2100         if (local_err != NULL) {
2101             break;
2102         }
2103         ++processed;
2104         vcpus = vcpus->next;
2105     }
2106 
2107     if (local_err != NULL) {
2108         if (processed == 0) {
2109             error_propagate(errp, local_err);
2110         } else {
2111             error_free(local_err);
2112         }
2113     }
2114 
2115     return processed;
2116 }
2117 #endif /* __linux__ */
2118 
2119 #if defined(__linux__) || defined(__FreeBSD__)
2120 void qmp_guest_set_user_password(const char *username,
2121                                  const char *password,
2122                                  bool crypted,
2123                                  Error **errp)
2124 {
2125     Error *local_err = NULL;
2126     char *passwd_path = NULL;
2127     pid_t pid;
2128     int status;
2129     int datafd[2] = { -1, -1 };
2130     char *rawpasswddata = NULL;
2131     size_t rawpasswdlen;
2132     char *chpasswddata = NULL;
2133     size_t chpasswdlen;
2134 
2135     rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
2136     if (!rawpasswddata) {
2137         return;
2138     }
2139     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
2140     rawpasswddata[rawpasswdlen] = '\0';
2141 
2142     if (strchr(rawpasswddata, '\n')) {
2143         error_setg(errp, "forbidden characters in raw password");
2144         goto out;
2145     }
2146 
2147     if (strchr(username, '\n') ||
2148         strchr(username, ':')) {
2149         error_setg(errp, "forbidden characters in username");
2150         goto out;
2151     }
2152 
2153 #ifdef __FreeBSD__
2154     chpasswddata = g_strdup(rawpasswddata);
2155     passwd_path = g_find_program_in_path("pw");
2156 #else
2157     chpasswddata = g_strdup_printf("%s:%s\n", username, rawpasswddata);
2158     passwd_path = g_find_program_in_path("chpasswd");
2159 #endif
2160 
2161     chpasswdlen = strlen(chpasswddata);
2162 
2163     if (!passwd_path) {
2164         error_setg(errp, "cannot find 'passwd' program in PATH");
2165         goto out;
2166     }
2167 
2168     if (!g_unix_open_pipe(datafd, FD_CLOEXEC, NULL)) {
2169         error_setg(errp, "cannot create pipe FDs");
2170         goto out;
2171     }
2172 
2173     pid = fork();
2174     if (pid == 0) {
2175         close(datafd[1]);
2176         /* child */
2177         setsid();
2178         dup2(datafd[0], 0);
2179         reopen_fd_to_null(1);
2180         reopen_fd_to_null(2);
2181 
2182 #ifdef __FreeBSD__
2183         const char *h_arg;
2184         h_arg = (crypted) ? "-H" : "-h";
2185         execl(passwd_path, "pw", "usermod", "-n", username, h_arg, "0", NULL);
2186 #else
2187         if (crypted) {
2188             execl(passwd_path, "chpasswd", "-e", NULL);
2189         } else {
2190             execl(passwd_path, "chpasswd", NULL);
2191         }
2192 #endif
2193         _exit(EXIT_FAILURE);
2194     } else if (pid < 0) {
2195         error_setg_errno(errp, errno, "failed to create child process");
2196         goto out;
2197     }
2198     close(datafd[0]);
2199     datafd[0] = -1;
2200 
2201     if (qemu_write_full(datafd[1], chpasswddata, chpasswdlen) != chpasswdlen) {
2202         error_setg_errno(errp, errno, "cannot write new account password");
2203         goto out;
2204     }
2205     close(datafd[1]);
2206     datafd[1] = -1;
2207 
2208     ga_wait_child(pid, &status, &local_err);
2209     if (local_err) {
2210         error_propagate(errp, local_err);
2211         goto out;
2212     }
2213 
2214     if (!WIFEXITED(status)) {
2215         error_setg(errp, "child process has terminated abnormally");
2216         goto out;
2217     }
2218 
2219     if (WEXITSTATUS(status)) {
2220         error_setg(errp, "child process has failed to set user password");
2221         goto out;
2222     }
2223 
2224 out:
2225     g_free(chpasswddata);
2226     g_free(rawpasswddata);
2227     g_free(passwd_path);
2228     if (datafd[0] != -1) {
2229         close(datafd[0]);
2230     }
2231     if (datafd[1] != -1) {
2232         close(datafd[1]);
2233     }
2234 }
2235 #else /* __linux__ || __FreeBSD__ */
2236 void qmp_guest_set_user_password(const char *username,
2237                                  const char *password,
2238                                  bool crypted,
2239                                  Error **errp)
2240 {
2241     error_setg(errp, QERR_UNSUPPORTED);
2242 }
2243 #endif /* __linux__ || __FreeBSD__ */
2244 
2245 #ifdef __linux__
2246 static void ga_read_sysfs_file(int dirfd, const char *pathname, char *buf,
2247                                int size, Error **errp)
2248 {
2249     int fd;
2250     int res;
2251 
2252     errno = 0;
2253     fd = openat(dirfd, pathname, O_RDONLY);
2254     if (fd == -1) {
2255         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2256         return;
2257     }
2258 
2259     res = pread(fd, buf, size, 0);
2260     if (res == -1) {
2261         error_setg_errno(errp, errno, "pread sysfs file \"%s\"", pathname);
2262     } else if (res == 0) {
2263         error_setg(errp, "pread sysfs file \"%s\": unexpected EOF", pathname);
2264     }
2265     close(fd);
2266 }
2267 
2268 static void ga_write_sysfs_file(int dirfd, const char *pathname,
2269                                 const char *buf, int size, Error **errp)
2270 {
2271     int fd;
2272 
2273     errno = 0;
2274     fd = openat(dirfd, pathname, O_WRONLY);
2275     if (fd == -1) {
2276         error_setg_errno(errp, errno, "open sysfs file \"%s\"", pathname);
2277         return;
2278     }
2279 
2280     if (pwrite(fd, buf, size, 0) == -1) {
2281         error_setg_errno(errp, errno, "pwrite sysfs file \"%s\"", pathname);
2282     }
2283 
2284     close(fd);
2285 }
2286 
2287 /* Transfer online/offline status between @mem_blk and the guest system.
2288  *
2289  * On input either @errp or *@errp must be NULL.
2290  *
2291  * In system-to-@mem_blk direction, the following @mem_blk fields are accessed:
2292  * - R: mem_blk->phys_index
2293  * - W: mem_blk->online
2294  * - W: mem_blk->can_offline
2295  *
2296  * In @mem_blk-to-system direction, the following @mem_blk fields are accessed:
2297  * - R: mem_blk->phys_index
2298  * - R: mem_blk->online
2299  *-  R: mem_blk->can_offline
2300  * Written members remain unmodified on error.
2301  */
2302 static void transfer_memory_block(GuestMemoryBlock *mem_blk, bool sys2memblk,
2303                                   GuestMemoryBlockResponse *result,
2304                                   Error **errp)
2305 {
2306     char *dirpath;
2307     int dirfd;
2308     char *status;
2309     Error *local_err = NULL;
2310 
2311     if (!sys2memblk) {
2312         DIR *dp;
2313 
2314         if (!result) {
2315             error_setg(errp, "Internal error, 'result' should not be NULL");
2316             return;
2317         }
2318         errno = 0;
2319         dp = opendir("/sys/devices/system/memory/");
2320          /* if there is no 'memory' directory in sysfs,
2321          * we think this VM does not support online/offline memory block,
2322          * any other solution?
2323          */
2324         if (!dp) {
2325             if (errno == ENOENT) {
2326                 result->response =
2327                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2328             }
2329             goto out1;
2330         }
2331         closedir(dp);
2332     }
2333 
2334     dirpath = g_strdup_printf("/sys/devices/system/memory/memory%" PRId64 "/",
2335                               mem_blk->phys_index);
2336     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2337     if (dirfd == -1) {
2338         if (sys2memblk) {
2339             error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2340         } else {
2341             if (errno == ENOENT) {
2342                 result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_NOT_FOUND;
2343             } else {
2344                 result->response =
2345                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2346             }
2347         }
2348         g_free(dirpath);
2349         goto out1;
2350     }
2351     g_free(dirpath);
2352 
2353     status = g_malloc0(10);
2354     ga_read_sysfs_file(dirfd, "state", status, 10, &local_err);
2355     if (local_err) {
2356         /* treat with sysfs file that not exist in old kernel */
2357         if (errno == ENOENT) {
2358             error_free(local_err);
2359             if (sys2memblk) {
2360                 mem_blk->online = true;
2361                 mem_blk->can_offline = false;
2362             } else if (!mem_blk->online) {
2363                 result->response =
2364                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_NOT_SUPPORTED;
2365             }
2366         } else {
2367             if (sys2memblk) {
2368                 error_propagate(errp, local_err);
2369             } else {
2370                 error_free(local_err);
2371                 result->response =
2372                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2373             }
2374         }
2375         goto out2;
2376     }
2377 
2378     if (sys2memblk) {
2379         char removable = '0';
2380 
2381         mem_blk->online = (strncmp(status, "online", 6) == 0);
2382 
2383         ga_read_sysfs_file(dirfd, "removable", &removable, 1, &local_err);
2384         if (local_err) {
2385             /* if no 'removable' file, it doesn't support offline mem blk */
2386             if (errno == ENOENT) {
2387                 error_free(local_err);
2388                 mem_blk->can_offline = false;
2389             } else {
2390                 error_propagate(errp, local_err);
2391             }
2392         } else {
2393             mem_blk->can_offline = (removable != '0');
2394         }
2395     } else {
2396         if (mem_blk->online != (strncmp(status, "online", 6) == 0)) {
2397             const char *new_state = mem_blk->online ? "online" : "offline";
2398 
2399             ga_write_sysfs_file(dirfd, "state", new_state, strlen(new_state),
2400                                 &local_err);
2401             if (local_err) {
2402                 error_free(local_err);
2403                 result->response =
2404                     GUEST_MEMORY_BLOCK_RESPONSE_TYPE_OPERATION_FAILED;
2405                 goto out2;
2406             }
2407 
2408             result->response = GUEST_MEMORY_BLOCK_RESPONSE_TYPE_SUCCESS;
2409             result->has_error_code = false;
2410         } /* otherwise pretend successful re-(on|off)-lining */
2411     }
2412     g_free(status);
2413     close(dirfd);
2414     return;
2415 
2416 out2:
2417     g_free(status);
2418     close(dirfd);
2419 out1:
2420     if (!sys2memblk) {
2421         result->has_error_code = true;
2422         result->error_code = errno;
2423     }
2424 }
2425 
2426 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2427 {
2428     GuestMemoryBlockList *head, **tail;
2429     Error *local_err = NULL;
2430     struct dirent *de;
2431     DIR *dp;
2432 
2433     head = NULL;
2434     tail = &head;
2435 
2436     dp = opendir("/sys/devices/system/memory/");
2437     if (!dp) {
2438         /* it's ok if this happens to be a system that doesn't expose
2439          * memory blocks via sysfs, but otherwise we should report
2440          * an error
2441          */
2442         if (errno != ENOENT) {
2443             error_setg_errno(errp, errno, "Can't open directory"
2444                              "\"/sys/devices/system/memory/\"");
2445         }
2446         return NULL;
2447     }
2448 
2449     /* Note: the phys_index of memory block may be discontinuous,
2450      * this is because a memblk is the unit of the Sparse Memory design, which
2451      * allows discontinuous memory ranges (ex. NUMA), so here we should
2452      * traverse the memory block directory.
2453      */
2454     while ((de = readdir(dp)) != NULL) {
2455         GuestMemoryBlock *mem_blk;
2456 
2457         if ((strncmp(de->d_name, "memory", 6) != 0) ||
2458             !(de->d_type & DT_DIR)) {
2459             continue;
2460         }
2461 
2462         mem_blk = g_malloc0(sizeof *mem_blk);
2463         /* The d_name is "memoryXXX",  phys_index is block id, same as XXX */
2464         mem_blk->phys_index = strtoul(&de->d_name[6], NULL, 10);
2465         mem_blk->has_can_offline = true; /* lolspeak ftw */
2466         transfer_memory_block(mem_blk, true, NULL, &local_err);
2467         if (local_err) {
2468             break;
2469         }
2470 
2471         QAPI_LIST_APPEND(tail, mem_blk);
2472     }
2473 
2474     closedir(dp);
2475     if (local_err == NULL) {
2476         /* there's no guest with zero memory blocks */
2477         if (head == NULL) {
2478             error_setg(errp, "guest reported zero memory blocks!");
2479         }
2480         return head;
2481     }
2482 
2483     qapi_free_GuestMemoryBlockList(head);
2484     error_propagate(errp, local_err);
2485     return NULL;
2486 }
2487 
2488 GuestMemoryBlockResponseList *
2489 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2490 {
2491     GuestMemoryBlockResponseList *head, **tail;
2492     Error *local_err = NULL;
2493 
2494     head = NULL;
2495     tail = &head;
2496 
2497     while (mem_blks != NULL) {
2498         GuestMemoryBlockResponse *result;
2499         GuestMemoryBlock *current_mem_blk = mem_blks->value;
2500 
2501         result = g_malloc0(sizeof(*result));
2502         result->phys_index = current_mem_blk->phys_index;
2503         transfer_memory_block(current_mem_blk, false, result, &local_err);
2504         if (local_err) { /* should never happen */
2505             goto err;
2506         }
2507 
2508         QAPI_LIST_APPEND(tail, result);
2509         mem_blks = mem_blks->next;
2510     }
2511 
2512     return head;
2513 err:
2514     qapi_free_GuestMemoryBlockResponseList(head);
2515     error_propagate(errp, local_err);
2516     return NULL;
2517 }
2518 
2519 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2520 {
2521     Error *local_err = NULL;
2522     char *dirpath;
2523     int dirfd;
2524     char *buf;
2525     GuestMemoryBlockInfo *info;
2526 
2527     dirpath = g_strdup_printf("/sys/devices/system/memory/");
2528     dirfd = open(dirpath, O_RDONLY | O_DIRECTORY);
2529     if (dirfd == -1) {
2530         error_setg_errno(errp, errno, "open(\"%s\")", dirpath);
2531         g_free(dirpath);
2532         return NULL;
2533     }
2534     g_free(dirpath);
2535 
2536     buf = g_malloc0(20);
2537     ga_read_sysfs_file(dirfd, "block_size_bytes", buf, 20, &local_err);
2538     close(dirfd);
2539     if (local_err) {
2540         g_free(buf);
2541         error_propagate(errp, local_err);
2542         return NULL;
2543     }
2544 
2545     info = g_new0(GuestMemoryBlockInfo, 1);
2546     info->size = strtol(buf, NULL, 16); /* the unit is bytes */
2547 
2548     g_free(buf);
2549 
2550     return info;
2551 }
2552 
2553 #define MAX_NAME_LEN 128
2554 static GuestDiskStatsInfoList *guest_get_diskstats(Error **errp)
2555 {
2556 #ifdef CONFIG_LINUX
2557     GuestDiskStatsInfoList *head = NULL, **tail = &head;
2558     const char *diskstats = "/proc/diskstats";
2559     FILE *fp;
2560     size_t n;
2561     char *line = NULL;
2562 
2563     fp = fopen(diskstats, "r");
2564     if (fp  == NULL) {
2565         error_setg_errno(errp, errno, "open(\"%s\")", diskstats);
2566         return NULL;
2567     }
2568 
2569     while (getline(&line, &n, fp) != -1) {
2570         g_autofree GuestDiskStatsInfo *diskstatinfo = NULL;
2571         g_autofree GuestDiskStats *diskstat = NULL;
2572         char dev_name[MAX_NAME_LEN];
2573         unsigned int ios_pgr, tot_ticks, rq_ticks, wr_ticks, dc_ticks, fl_ticks;
2574         unsigned long rd_ios, rd_merges_or_rd_sec, rd_ticks_or_wr_sec, wr_ios;
2575         unsigned long wr_merges, rd_sec_or_wr_ios, wr_sec;
2576         unsigned long dc_ios, dc_merges, dc_sec, fl_ios;
2577         unsigned int major, minor;
2578         int i;
2579 
2580         i = sscanf(line, "%u %u %s %lu %lu %lu"
2581                    "%lu %lu %lu %lu %u %u %u %u"
2582                    "%lu %lu %lu %u %lu %u",
2583                    &major, &minor, dev_name,
2584                    &rd_ios, &rd_merges_or_rd_sec, &rd_sec_or_wr_ios,
2585                    &rd_ticks_or_wr_sec, &wr_ios, &wr_merges, &wr_sec,
2586                    &wr_ticks, &ios_pgr, &tot_ticks, &rq_ticks,
2587                    &dc_ios, &dc_merges, &dc_sec, &dc_ticks,
2588                    &fl_ios, &fl_ticks);
2589 
2590         if (i < 7) {
2591             continue;
2592         }
2593 
2594         diskstatinfo = g_new0(GuestDiskStatsInfo, 1);
2595         diskstatinfo->name = g_strdup(dev_name);
2596         diskstatinfo->major = major;
2597         diskstatinfo->minor = minor;
2598 
2599         diskstat = g_new0(GuestDiskStats, 1);
2600         if (i == 7) {
2601             diskstat->has_read_ios = true;
2602             diskstat->read_ios = rd_ios;
2603             diskstat->has_read_sectors = true;
2604             diskstat->read_sectors = rd_merges_or_rd_sec;
2605             diskstat->has_write_ios = true;
2606             diskstat->write_ios = rd_sec_or_wr_ios;
2607             diskstat->has_write_sectors = true;
2608             diskstat->write_sectors = rd_ticks_or_wr_sec;
2609         }
2610         if (i >= 14) {
2611             diskstat->has_read_ios = true;
2612             diskstat->read_ios = rd_ios;
2613             diskstat->has_read_sectors = true;
2614             diskstat->read_sectors = rd_sec_or_wr_ios;
2615             diskstat->has_read_merges = true;
2616             diskstat->read_merges = rd_merges_or_rd_sec;
2617             diskstat->has_read_ticks = true;
2618             diskstat->read_ticks = rd_ticks_or_wr_sec;
2619             diskstat->has_write_ios = true;
2620             diskstat->write_ios = wr_ios;
2621             diskstat->has_write_sectors = true;
2622             diskstat->write_sectors = wr_sec;
2623             diskstat->has_write_merges = true;
2624             diskstat->write_merges = wr_merges;
2625             diskstat->has_write_ticks = true;
2626             diskstat->write_ticks = wr_ticks;
2627             diskstat->has_ios_pgr = true;
2628             diskstat->ios_pgr = ios_pgr;
2629             diskstat->has_total_ticks = true;
2630             diskstat->total_ticks = tot_ticks;
2631             diskstat->has_weight_ticks = true;
2632             diskstat->weight_ticks = rq_ticks;
2633         }
2634         if (i >= 18) {
2635             diskstat->has_discard_ios = true;
2636             diskstat->discard_ios = dc_ios;
2637             diskstat->has_discard_merges = true;
2638             diskstat->discard_merges = dc_merges;
2639             diskstat->has_discard_sectors = true;
2640             diskstat->discard_sectors = dc_sec;
2641             diskstat->has_discard_ticks = true;
2642             diskstat->discard_ticks = dc_ticks;
2643         }
2644         if (i >= 20) {
2645             diskstat->has_flush_ios = true;
2646             diskstat->flush_ios = fl_ios;
2647             diskstat->has_flush_ticks = true;
2648             diskstat->flush_ticks = fl_ticks;
2649         }
2650 
2651         diskstatinfo->stats = g_steal_pointer(&diskstat);
2652         QAPI_LIST_APPEND(tail, diskstatinfo);
2653         diskstatinfo = NULL;
2654     }
2655     free(line);
2656     fclose(fp);
2657     return head;
2658 #else
2659     g_debug("disk stats reporting available only for Linux");
2660     return NULL;
2661 #endif
2662 }
2663 
2664 GuestDiskStatsInfoList *qmp_guest_get_diskstats(Error **errp)
2665 {
2666     return guest_get_diskstats(errp);
2667 }
2668 
2669 GuestCpuStatsList *qmp_guest_get_cpustats(Error **errp)
2670 {
2671     GuestCpuStatsList *head = NULL, **tail = &head;
2672     const char *cpustats = "/proc/stat";
2673     int clk_tck = sysconf(_SC_CLK_TCK);
2674     FILE *fp;
2675     size_t n;
2676     char *line = NULL;
2677 
2678     fp = fopen(cpustats, "r");
2679     if (fp  == NULL) {
2680         error_setg_errno(errp, errno, "open(\"%s\")", cpustats);
2681         return NULL;
2682     }
2683 
2684     while (getline(&line, &n, fp) != -1) {
2685         GuestCpuStats *cpustat = NULL;
2686         GuestLinuxCpuStats *linuxcpustat;
2687         int i;
2688         unsigned long user, system, idle, iowait, irq, softirq, steal, guest;
2689         unsigned long nice, guest_nice;
2690         char name[64];
2691 
2692         i = sscanf(line, "%s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu",
2693                    name, &user, &nice, &system, &idle, &iowait, &irq, &softirq,
2694                    &steal, &guest, &guest_nice);
2695 
2696         /* drop "cpu 1 2 3 ...", get "cpuX 1 2 3 ..." only */
2697         if ((i == EOF) || strncmp(name, "cpu", 3) || (name[3] == '\0')) {
2698             continue;
2699         }
2700 
2701         if (i < 5) {
2702             slog("Parsing cpu stat from %s failed, see \"man proc\"", cpustats);
2703             break;
2704         }
2705 
2706         cpustat = g_new0(GuestCpuStats, 1);
2707         cpustat->type = GUEST_CPU_STATS_TYPE_LINUX;
2708 
2709         linuxcpustat = &cpustat->u.q_linux;
2710         linuxcpustat->cpu = atoi(&name[3]);
2711         linuxcpustat->user = user * 1000 / clk_tck;
2712         linuxcpustat->nice = nice * 1000 / clk_tck;
2713         linuxcpustat->system = system * 1000 / clk_tck;
2714         linuxcpustat->idle = idle * 1000 / clk_tck;
2715 
2716         if (i > 5) {
2717             linuxcpustat->has_iowait = true;
2718             linuxcpustat->iowait = iowait * 1000 / clk_tck;
2719         }
2720 
2721         if (i > 6) {
2722             linuxcpustat->has_irq = true;
2723             linuxcpustat->irq = irq * 1000 / clk_tck;
2724             linuxcpustat->has_softirq = true;
2725             linuxcpustat->softirq = softirq * 1000 / clk_tck;
2726         }
2727 
2728         if (i > 8) {
2729             linuxcpustat->has_steal = true;
2730             linuxcpustat->steal = steal * 1000 / clk_tck;
2731         }
2732 
2733         if (i > 9) {
2734             linuxcpustat->has_guest = true;
2735             linuxcpustat->guest = guest * 1000 / clk_tck;
2736         }
2737 
2738         if (i > 10) {
2739             linuxcpustat->has_guest = true;
2740             linuxcpustat->guest = guest * 1000 / clk_tck;
2741             linuxcpustat->has_guestnice = true;
2742             linuxcpustat->guestnice = guest_nice * 1000 / clk_tck;
2743         }
2744 
2745         QAPI_LIST_APPEND(tail, cpustat);
2746     }
2747 
2748     free(line);
2749     fclose(fp);
2750     return head;
2751 }
2752 
2753 #else /* defined(__linux__) */
2754 
2755 void qmp_guest_suspend_disk(Error **errp)
2756 {
2757     error_setg(errp, QERR_UNSUPPORTED);
2758 }
2759 
2760 void qmp_guest_suspend_ram(Error **errp)
2761 {
2762     error_setg(errp, QERR_UNSUPPORTED);
2763 }
2764 
2765 void qmp_guest_suspend_hybrid(Error **errp)
2766 {
2767     error_setg(errp, QERR_UNSUPPORTED);
2768 }
2769 
2770 GuestLogicalProcessorList *qmp_guest_get_vcpus(Error **errp)
2771 {
2772     error_setg(errp, QERR_UNSUPPORTED);
2773     return NULL;
2774 }
2775 
2776 int64_t qmp_guest_set_vcpus(GuestLogicalProcessorList *vcpus, Error **errp)
2777 {
2778     error_setg(errp, QERR_UNSUPPORTED);
2779     return -1;
2780 }
2781 
2782 GuestMemoryBlockList *qmp_guest_get_memory_blocks(Error **errp)
2783 {
2784     error_setg(errp, QERR_UNSUPPORTED);
2785     return NULL;
2786 }
2787 
2788 GuestMemoryBlockResponseList *
2789 qmp_guest_set_memory_blocks(GuestMemoryBlockList *mem_blks, Error **errp)
2790 {
2791     error_setg(errp, QERR_UNSUPPORTED);
2792     return NULL;
2793 }
2794 
2795 GuestMemoryBlockInfo *qmp_guest_get_memory_block_info(Error **errp)
2796 {
2797     error_setg(errp, QERR_UNSUPPORTED);
2798     return NULL;
2799 }
2800 
2801 #endif
2802 
2803 #ifdef HAVE_GETIFADDRS
2804 static GuestNetworkInterface *
2805 guest_find_interface(GuestNetworkInterfaceList *head,
2806                      const char *name)
2807 {
2808     for (; head; head = head->next) {
2809         if (strcmp(head->value->name, name) == 0) {
2810             return head->value;
2811         }
2812     }
2813 
2814     return NULL;
2815 }
2816 
2817 static int guest_get_network_stats(const char *name,
2818                        GuestNetworkInterfaceStat *stats)
2819 {
2820 #ifdef CONFIG_LINUX
2821     int name_len;
2822     char const *devinfo = "/proc/net/dev";
2823     FILE *fp;
2824     char *line = NULL, *colon;
2825     size_t n = 0;
2826     fp = fopen(devinfo, "r");
2827     if (!fp) {
2828         g_debug("failed to open network stats %s: %s", devinfo,
2829                 g_strerror(errno));
2830         return -1;
2831     }
2832     name_len = strlen(name);
2833     while (getline(&line, &n, fp) != -1) {
2834         long long dummy;
2835         long long rx_bytes;
2836         long long rx_packets;
2837         long long rx_errs;
2838         long long rx_dropped;
2839         long long tx_bytes;
2840         long long tx_packets;
2841         long long tx_errs;
2842         long long tx_dropped;
2843         char *trim_line;
2844         trim_line = g_strchug(line);
2845         if (trim_line[0] == '\0') {
2846             continue;
2847         }
2848         colon = strchr(trim_line, ':');
2849         if (!colon) {
2850             continue;
2851         }
2852         if (colon - name_len  == trim_line &&
2853            strncmp(trim_line, name, name_len) == 0) {
2854             if (sscanf(colon + 1,
2855                 "%lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld %lld",
2856                   &rx_bytes, &rx_packets, &rx_errs, &rx_dropped,
2857                   &dummy, &dummy, &dummy, &dummy,
2858                   &tx_bytes, &tx_packets, &tx_errs, &tx_dropped,
2859                   &dummy, &dummy, &dummy, &dummy) != 16) {
2860                 continue;
2861             }
2862             stats->rx_bytes = rx_bytes;
2863             stats->rx_packets = rx_packets;
2864             stats->rx_errs = rx_errs;
2865             stats->rx_dropped = rx_dropped;
2866             stats->tx_bytes = tx_bytes;
2867             stats->tx_packets = tx_packets;
2868             stats->tx_errs = tx_errs;
2869             stats->tx_dropped = tx_dropped;
2870             fclose(fp);
2871             g_free(line);
2872             return 0;
2873         }
2874     }
2875     fclose(fp);
2876     g_free(line);
2877     g_debug("/proc/net/dev: Interface '%s' not found", name);
2878 #else /* !CONFIG_LINUX */
2879     g_debug("Network stats reporting available only for Linux");
2880 #endif /* !CONFIG_LINUX */
2881     return -1;
2882 }
2883 
2884 #ifndef CONFIG_BSD
2885 /*
2886  * Fill "buf" with MAC address by ifaddrs. Pointer buf must point to a
2887  * buffer with ETHER_ADDR_LEN length at least.
2888  *
2889  * Returns false in case of an error, otherwise true. "obtained" argument
2890  * is true if a MAC address was obtained successful, otherwise false.
2891  */
2892 bool guest_get_hw_addr(struct ifaddrs *ifa, unsigned char *buf,
2893                        bool *obtained, Error **errp)
2894 {
2895     struct ifreq ifr;
2896     int sock;
2897 
2898     *obtained = false;
2899 
2900     /* we haven't obtained HW address yet */
2901     sock = socket(PF_INET, SOCK_STREAM, 0);
2902     if (sock == -1) {
2903         error_setg_errno(errp, errno, "failed to create socket");
2904         return false;
2905     }
2906 
2907     memset(&ifr, 0, sizeof(ifr));
2908     pstrcpy(ifr.ifr_name, IF_NAMESIZE, ifa->ifa_name);
2909     if (ioctl(sock, SIOCGIFHWADDR, &ifr) == -1) {
2910         /*
2911          * We can't get the hw addr of this interface, but that's not a
2912          * fatal error.
2913          */
2914         if (errno == EADDRNOTAVAIL) {
2915             /* The interface doesn't have a hw addr (e.g. loopback). */
2916             g_debug("failed to get MAC address of %s: %s",
2917                     ifa->ifa_name, strerror(errno));
2918         } else{
2919             g_warning("failed to get MAC address of %s: %s",
2920                       ifa->ifa_name, strerror(errno));
2921         }
2922     } else {
2923 #ifdef CONFIG_SOLARIS
2924         memcpy(buf, &ifr.ifr_addr.sa_data, ETHER_ADDR_LEN);
2925 #else
2926         memcpy(buf, &ifr.ifr_hwaddr.sa_data, ETHER_ADDR_LEN);
2927 #endif
2928         *obtained = true;
2929     }
2930     close(sock);
2931     return true;
2932 }
2933 #endif /* CONFIG_BSD */
2934 
2935 /*
2936  * Build information about guest interfaces
2937  */
2938 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
2939 {
2940     GuestNetworkInterfaceList *head = NULL, **tail = &head;
2941     struct ifaddrs *ifap, *ifa;
2942 
2943     if (getifaddrs(&ifap) < 0) {
2944         error_setg_errno(errp, errno, "getifaddrs failed");
2945         goto error;
2946     }
2947 
2948     for (ifa = ifap; ifa; ifa = ifa->ifa_next) {
2949         GuestNetworkInterface *info;
2950         GuestIpAddressList **address_tail;
2951         GuestIpAddress *address_item = NULL;
2952         GuestNetworkInterfaceStat *interface_stat = NULL;
2953         char addr4[INET_ADDRSTRLEN];
2954         char addr6[INET6_ADDRSTRLEN];
2955         unsigned char mac_addr[ETHER_ADDR_LEN];
2956         bool obtained;
2957         void *p;
2958 
2959         g_debug("Processing %s interface", ifa->ifa_name);
2960 
2961         info = guest_find_interface(head, ifa->ifa_name);
2962 
2963         if (!info) {
2964             info = g_malloc0(sizeof(*info));
2965             info->name = g_strdup(ifa->ifa_name);
2966 
2967             QAPI_LIST_APPEND(tail, info);
2968         }
2969 
2970         if (!info->hardware_address) {
2971             if (!guest_get_hw_addr(ifa, mac_addr, &obtained, errp)) {
2972                 goto error;
2973             }
2974             if (obtained) {
2975                 info->hardware_address =
2976                     g_strdup_printf("%02x:%02x:%02x:%02x:%02x:%02x",
2977                                     (int) mac_addr[0], (int) mac_addr[1],
2978                                     (int) mac_addr[2], (int) mac_addr[3],
2979                                     (int) mac_addr[4], (int) mac_addr[5]);
2980             }
2981         }
2982 
2983         if (ifa->ifa_addr &&
2984             ifa->ifa_addr->sa_family == AF_INET) {
2985             /* interface with IPv4 address */
2986             p = &((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
2987             if (!inet_ntop(AF_INET, p, addr4, sizeof(addr4))) {
2988                 error_setg_errno(errp, errno, "inet_ntop failed");
2989                 goto error;
2990             }
2991 
2992             address_item = g_malloc0(sizeof(*address_item));
2993             address_item->ip_address = g_strdup(addr4);
2994             address_item->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV4;
2995 
2996             if (ifa->ifa_netmask) {
2997                 /* Count the number of set bits in netmask.
2998                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
2999                 p = &((struct sockaddr_in *)ifa->ifa_netmask)->sin_addr;
3000                 address_item->prefix = ctpop32(((uint32_t *) p)[0]);
3001             }
3002         } else if (ifa->ifa_addr &&
3003                    ifa->ifa_addr->sa_family == AF_INET6) {
3004             /* interface with IPv6 address */
3005             p = &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr;
3006             if (!inet_ntop(AF_INET6, p, addr6, sizeof(addr6))) {
3007                 error_setg_errno(errp, errno, "inet_ntop failed");
3008                 goto error;
3009             }
3010 
3011             address_item = g_malloc0(sizeof(*address_item));
3012             address_item->ip_address = g_strdup(addr6);
3013             address_item->ip_address_type = GUEST_IP_ADDRESS_TYPE_IPV6;
3014 
3015             if (ifa->ifa_netmask) {
3016                 /* Count the number of set bits in netmask.
3017                  * This is safe as '1' and '0' cannot be shuffled in netmask. */
3018                 p = &((struct sockaddr_in6 *)ifa->ifa_netmask)->sin6_addr;
3019                 address_item->prefix =
3020                     ctpop32(((uint32_t *) p)[0]) +
3021                     ctpop32(((uint32_t *) p)[1]) +
3022                     ctpop32(((uint32_t *) p)[2]) +
3023                     ctpop32(((uint32_t *) p)[3]);
3024             }
3025         }
3026 
3027         if (!address_item) {
3028             continue;
3029         }
3030 
3031         address_tail = &info->ip_addresses;
3032         while (*address_tail) {
3033             address_tail = &(*address_tail)->next;
3034         }
3035         QAPI_LIST_APPEND(address_tail, address_item);
3036 
3037         info->has_ip_addresses = true;
3038 
3039         if (!info->statistics) {
3040             interface_stat = g_malloc0(sizeof(*interface_stat));
3041             if (guest_get_network_stats(info->name, interface_stat) == -1) {
3042                 g_free(interface_stat);
3043             } else {
3044                 info->statistics = interface_stat;
3045             }
3046         }
3047     }
3048 
3049     freeifaddrs(ifap);
3050     return head;
3051 
3052 error:
3053     freeifaddrs(ifap);
3054     qapi_free_GuestNetworkInterfaceList(head);
3055     return NULL;
3056 }
3057 
3058 #else
3059 
3060 GuestNetworkInterfaceList *qmp_guest_network_get_interfaces(Error **errp)
3061 {
3062     error_setg(errp, QERR_UNSUPPORTED);
3063     return NULL;
3064 }
3065 
3066 #endif /* HAVE_GETIFADDRS */
3067 
3068 #if !defined(CONFIG_FSFREEZE)
3069 
3070 GuestFilesystemInfoList *qmp_guest_get_fsinfo(Error **errp)
3071 {
3072     error_setg(errp, QERR_UNSUPPORTED);
3073     return NULL;
3074 }
3075 
3076 GuestFsfreezeStatus qmp_guest_fsfreeze_status(Error **errp)
3077 {
3078     error_setg(errp, QERR_UNSUPPORTED);
3079 
3080     return 0;
3081 }
3082 
3083 int64_t qmp_guest_fsfreeze_freeze(Error **errp)
3084 {
3085     error_setg(errp, QERR_UNSUPPORTED);
3086 
3087     return 0;
3088 }
3089 
3090 int64_t qmp_guest_fsfreeze_freeze_list(bool has_mountpoints,
3091                                        strList *mountpoints,
3092                                        Error **errp)
3093 {
3094     error_setg(errp, QERR_UNSUPPORTED);
3095 
3096     return 0;
3097 }
3098 
3099 int64_t qmp_guest_fsfreeze_thaw(Error **errp)
3100 {
3101     error_setg(errp, QERR_UNSUPPORTED);
3102 
3103     return 0;
3104 }
3105 
3106 GuestDiskInfoList *qmp_guest_get_disks(Error **errp)
3107 {
3108     error_setg(errp, QERR_UNSUPPORTED);
3109     return NULL;
3110 }
3111 
3112 GuestDiskStatsInfoList *qmp_guest_get_diskstats(Error **errp)
3113 {
3114     error_setg(errp, QERR_UNSUPPORTED);
3115     return NULL;
3116 }
3117 
3118 GuestCpuStatsList *qmp_guest_get_cpustats(Error **errp)
3119 {
3120     error_setg(errp, QERR_UNSUPPORTED);
3121     return NULL;
3122 }
3123 
3124 #endif /* CONFIG_FSFREEZE */
3125 
3126 #if !defined(CONFIG_FSTRIM)
3127 GuestFilesystemTrimResponse *
3128 qmp_guest_fstrim(bool has_minimum, int64_t minimum, Error **errp)
3129 {
3130     error_setg(errp, QERR_UNSUPPORTED);
3131     return NULL;
3132 }
3133 #endif
3134 
3135 /* add unsupported commands to the list of blocked RPCs */
3136 GList *ga_command_init_blockedrpcs(GList *blockedrpcs)
3137 {
3138 #if !defined(__linux__)
3139     {
3140         const char *list[] = {
3141             "guest-suspend-disk", "guest-suspend-ram",
3142             "guest-suspend-hybrid", "guest-get-vcpus", "guest-set-vcpus",
3143             "guest-get-memory-blocks", "guest-set-memory-blocks",
3144             "guest-get-memory-block-size", "guest-get-memory-block-info",
3145             NULL};
3146         char **p = (char **)list;
3147 
3148         while (*p) {
3149             blockedrpcs = g_list_append(blockedrpcs, g_strdup(*p++));
3150         }
3151     }
3152 #endif
3153 
3154 #if !defined(HAVE_GETIFADDRS)
3155     blockedrpcs = g_list_append(blockedrpcs,
3156                               g_strdup("guest-network-get-interfaces"));
3157 #endif
3158 
3159 #if !defined(CONFIG_FSFREEZE)
3160     {
3161         const char *list[] = {
3162             "guest-get-fsinfo", "guest-fsfreeze-status",
3163             "guest-fsfreeze-freeze", "guest-fsfreeze-freeze-list",
3164             "guest-fsfreeze-thaw", "guest-get-fsinfo",
3165             "guest-get-disks", NULL};
3166         char **p = (char **)list;
3167 
3168         while (*p) {
3169             blockedrpcs = g_list_append(blockedrpcs, g_strdup(*p++));
3170         }
3171     }
3172 #endif
3173 
3174 #if !defined(CONFIG_FSTRIM)
3175     blockedrpcs = g_list_append(blockedrpcs, g_strdup("guest-fstrim"));
3176 #endif
3177 
3178     blockedrpcs = g_list_append(blockedrpcs, g_strdup("guest-get-devices"));
3179 
3180     return blockedrpcs;
3181 }
3182 
3183 /* register init/cleanup routines for stateful command groups */
3184 void ga_command_state_init(GAState *s, GACommandState *cs)
3185 {
3186 #if defined(CONFIG_FSFREEZE)
3187     ga_command_state_add(cs, NULL, guest_fsfreeze_cleanup);
3188 #endif
3189 }
3190 
3191 #ifdef HAVE_UTMPX
3192 
3193 #define QGA_MICRO_SECOND_TO_SECOND 1000000
3194 
3195 static double ga_get_login_time(struct utmpx *user_info)
3196 {
3197     double seconds = (double)user_info->ut_tv.tv_sec;
3198     double useconds = (double)user_info->ut_tv.tv_usec;
3199     useconds /= QGA_MICRO_SECOND_TO_SECOND;
3200     return seconds + useconds;
3201 }
3202 
3203 GuestUserList *qmp_guest_get_users(Error **errp)
3204 {
3205     GHashTable *cache = NULL;
3206     GuestUserList *head = NULL, **tail = &head;
3207     struct utmpx *user_info = NULL;
3208     gpointer value = NULL;
3209     GuestUser *user = NULL;
3210     double login_time = 0;
3211 
3212     cache = g_hash_table_new(g_str_hash, g_str_equal);
3213     setutxent();
3214 
3215     for (;;) {
3216         user_info = getutxent();
3217         if (user_info == NULL) {
3218             break;
3219         } else if (user_info->ut_type != USER_PROCESS) {
3220             continue;
3221         } else if (g_hash_table_contains(cache, user_info->ut_user)) {
3222             value = g_hash_table_lookup(cache, user_info->ut_user);
3223             user = (GuestUser *)value;
3224             login_time = ga_get_login_time(user_info);
3225             /* We're ensuring the earliest login time to be sent */
3226             if (login_time < user->login_time) {
3227                 user->login_time = login_time;
3228             }
3229             continue;
3230         }
3231 
3232         user = g_new0(GuestUser, 1);
3233         user->user = g_strdup(user_info->ut_user);
3234         user->login_time = ga_get_login_time(user_info);
3235 
3236         g_hash_table_insert(cache, user->user, user);
3237 
3238         QAPI_LIST_APPEND(tail, user);
3239     }
3240     endutxent();
3241     g_hash_table_destroy(cache);
3242     return head;
3243 }
3244 
3245 #else
3246 
3247 GuestUserList *qmp_guest_get_users(Error **errp)
3248 {
3249     error_setg(errp, QERR_UNSUPPORTED);
3250     return NULL;
3251 }
3252 
3253 #endif
3254 
3255 /* Replace escaped special characters with their real values. The replacement
3256  * is done in place -- returned value is in the original string.
3257  */
3258 static void ga_osrelease_replace_special(gchar *value)
3259 {
3260     gchar *p, *p2, quote;
3261 
3262     /* Trim the string at first space or semicolon if it is not enclosed in
3263      * single or double quotes. */
3264     if ((value[0] != '"') || (value[0] == '\'')) {
3265         p = strchr(value, ' ');
3266         if (p != NULL) {
3267             *p = 0;
3268         }
3269         p = strchr(value, ';');
3270         if (p != NULL) {
3271             *p = 0;
3272         }
3273         return;
3274     }
3275 
3276     quote = value[0];
3277     p2 = value;
3278     p = value + 1;
3279     while (*p != 0) {
3280         if (*p == '\\') {
3281             p++;
3282             switch (*p) {
3283             case '$':
3284             case '\'':
3285             case '"':
3286             case '\\':
3287             case '`':
3288                 break;
3289             default:
3290                 /* Keep literal backslash followed by whatever is there */
3291                 p--;
3292                 break;
3293             }
3294         } else if (*p == quote) {
3295             *p2 = 0;
3296             break;
3297         }
3298         *(p2++) = *(p++);
3299     }
3300 }
3301 
3302 static GKeyFile *ga_parse_osrelease(const char *fname)
3303 {
3304     gchar *content = NULL;
3305     gchar *content2 = NULL;
3306     GError *err = NULL;
3307     GKeyFile *keys = g_key_file_new();
3308     const char *group = "[os-release]\n";
3309 
3310     if (!g_file_get_contents(fname, &content, NULL, &err)) {
3311         slog("failed to read '%s', error: %s", fname, err->message);
3312         goto fail;
3313     }
3314 
3315     if (!g_utf8_validate(content, -1, NULL)) {
3316         slog("file is not utf-8 encoded: %s", fname);
3317         goto fail;
3318     }
3319     content2 = g_strdup_printf("%s%s", group, content);
3320 
3321     if (!g_key_file_load_from_data(keys, content2, -1, G_KEY_FILE_NONE,
3322                                    &err)) {
3323         slog("failed to parse file '%s', error: %s", fname, err->message);
3324         goto fail;
3325     }
3326 
3327     g_free(content);
3328     g_free(content2);
3329     return keys;
3330 
3331 fail:
3332     g_error_free(err);
3333     g_free(content);
3334     g_free(content2);
3335     g_key_file_free(keys);
3336     return NULL;
3337 }
3338 
3339 GuestOSInfo *qmp_guest_get_osinfo(Error **errp)
3340 {
3341     GuestOSInfo *info = NULL;
3342     struct utsname kinfo;
3343     GKeyFile *osrelease = NULL;
3344     const char *qga_os_release = g_getenv("QGA_OS_RELEASE");
3345 
3346     info = g_new0(GuestOSInfo, 1);
3347 
3348     if (uname(&kinfo) != 0) {
3349         error_setg_errno(errp, errno, "uname failed");
3350     } else {
3351         info->kernel_version = g_strdup(kinfo.version);
3352         info->kernel_release = g_strdup(kinfo.release);
3353         info->machine = g_strdup(kinfo.machine);
3354     }
3355 
3356     if (qga_os_release != NULL) {
3357         osrelease = ga_parse_osrelease(qga_os_release);
3358     } else {
3359         osrelease = ga_parse_osrelease("/etc/os-release");
3360         if (osrelease == NULL) {
3361             osrelease = ga_parse_osrelease("/usr/lib/os-release");
3362         }
3363     }
3364 
3365     if (osrelease != NULL) {
3366         char *value;
3367 
3368 #define GET_FIELD(field, osfield) do { \
3369     value = g_key_file_get_value(osrelease, "os-release", osfield, NULL); \
3370     if (value != NULL) { \
3371         ga_osrelease_replace_special(value); \
3372         info->field = value; \
3373     } \
3374 } while (0)
3375         GET_FIELD(id, "ID");
3376         GET_FIELD(name, "NAME");
3377         GET_FIELD(pretty_name, "PRETTY_NAME");
3378         GET_FIELD(version, "VERSION");
3379         GET_FIELD(version_id, "VERSION_ID");
3380         GET_FIELD(variant, "VARIANT");
3381         GET_FIELD(variant_id, "VARIANT_ID");
3382 #undef GET_FIELD
3383 
3384         g_key_file_free(osrelease);
3385     }
3386 
3387     return info;
3388 }
3389 
3390 GuestDeviceInfoList *qmp_guest_get_devices(Error **errp)
3391 {
3392     error_setg(errp, QERR_UNSUPPORTED);
3393 
3394     return NULL;
3395 }
3396 
3397 #ifndef HOST_NAME_MAX
3398 # ifdef _POSIX_HOST_NAME_MAX
3399 #  define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
3400 # else
3401 #  define HOST_NAME_MAX 255
3402 # endif
3403 #endif
3404 
3405 char *qga_get_host_name(Error **errp)
3406 {
3407     long len = -1;
3408     g_autofree char *hostname = NULL;
3409 
3410 #ifdef _SC_HOST_NAME_MAX
3411     len = sysconf(_SC_HOST_NAME_MAX);
3412 #endif /* _SC_HOST_NAME_MAX */
3413 
3414     if (len < 0) {
3415         len = HOST_NAME_MAX;
3416     }
3417 
3418     /* Unfortunately, gethostname() below does not guarantee a
3419      * NULL terminated string. Therefore, allocate one byte more
3420      * to be sure. */
3421     hostname = g_new0(char, len + 1);
3422 
3423     if (gethostname(hostname, len) < 0) {
3424         error_setg_errno(errp, errno,
3425                          "cannot get hostname");
3426         return NULL;
3427     }
3428 
3429     return g_steal_pointer(&hostname);
3430 }
3431