2 * Copyright (C) 2011 STRATO. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include "kerncompat.h"
20 #include "androidcompat.h"
22 #include <sys/ioctl.h>
25 #include <sys/types.h>
26 #include <sys/socket.h>
28 #include <sys/syscall.h>
31 #include <uuid/uuid.h>
48 static const char * const scrub_cmd_group_usage[] = {
49 "btrfs scrub <command> [options] <path>|<device>",
53 #define SCRUB_DATA_FILE "/var/lib/btrfs/scrub.status"
54 #define SCRUB_PROGRESS_SOCKET_PATH "/var/lib/btrfs/scrub.progress"
55 #define SCRUB_FILE_VERSION_PREFIX "scrub status"
56 #define SCRUB_FILE_VERSION "1"
67 /* TBD: replace with #include "linux/ioprio.h" in some years */
68 #if !defined (IOPRIO_H)
69 #define IOPRIO_WHO_PROCESS 1
70 #define IOPRIO_CLASS_SHIFT 13
71 #define IOPRIO_PRIO_VALUE(class, data) \
72 (((class) << IOPRIO_CLASS_SHIFT) | (data))
73 #define IOPRIO_CLASS_IDLE 3
76 struct scrub_progress {
77 struct btrfs_ioctl_scrub_args scrub_args;
81 struct scrub_stats stats;
82 struct scrub_file_record *resumed;
84 pthread_mutex_t progress_mutex;
89 struct scrub_file_record {
90 u8 fsid[BTRFS_FSID_SIZE];
92 struct scrub_stats stats;
93 struct btrfs_scrub_progress p;
96 struct scrub_progress_cycle {
100 struct btrfs_ioctl_fs_info_args *fi;
101 struct scrub_progress *progress;
102 struct scrub_progress *shared_progress;
103 pthread_mutex_t *write_mutex;
106 struct scrub_fs_stat {
107 struct btrfs_scrub_progress p;
108 struct scrub_stats s;
112 static void print_scrub_full(struct btrfs_scrub_progress *sp)
114 printf("\tdata_extents_scrubbed: %lld\n", sp->data_extents_scrubbed);
115 printf("\ttree_extents_scrubbed: %lld\n", sp->tree_extents_scrubbed);
116 printf("\tdata_bytes_scrubbed: %lld\n", sp->data_bytes_scrubbed);
117 printf("\ttree_bytes_scrubbed: %lld\n", sp->tree_bytes_scrubbed);
118 printf("\tread_errors: %lld\n", sp->read_errors);
119 printf("\tcsum_errors: %lld\n", sp->csum_errors);
120 printf("\tverify_errors: %lld\n", sp->verify_errors);
121 printf("\tno_csum: %lld\n", sp->no_csum);
122 printf("\tcsum_discards: %lld\n", sp->csum_discards);
123 printf("\tsuper_errors: %lld\n", sp->super_errors);
124 printf("\tmalloc_errors: %lld\n", sp->malloc_errors);
125 printf("\tuncorrectable_errors: %lld\n", sp->uncorrectable_errors);
126 printf("\tunverified_errors: %lld\n", sp->unverified_errors);
127 printf("\tcorrected_errors: %lld\n", sp->corrected_errors);
128 printf("\tlast_physical: %lld\n", sp->last_physical);
131 #define ERR(test, ...) do { \
133 fprintf(stderr, __VA_ARGS__); \
136 #define PRINT_SCRUB_ERROR(test, desc) do { \
138 printf(" %s=%llu", desc, test); \
141 static void print_scrub_summary(struct btrfs_scrub_progress *p)
146 err_cnt = p->read_errors +
151 err_cnt2 = p->corrected_errors + p->uncorrectable_errors;
153 if (p->malloc_errors)
154 printf("*** WARNING: memory allocation failed while scrubbing. "
155 "results may be inaccurate\n");
157 printf("\ttotal bytes scrubbed: %s with %llu errors\n",
158 pretty_size(p->data_bytes_scrubbed + p->tree_bytes_scrubbed),
159 max(err_cnt, err_cnt2));
161 if (err_cnt || err_cnt2) {
162 printf("\terror details:");
163 PRINT_SCRUB_ERROR(p->read_errors, "read");
164 PRINT_SCRUB_ERROR(p->super_errors, "super");
165 PRINT_SCRUB_ERROR(p->verify_errors, "verify");
166 PRINT_SCRUB_ERROR(p->csum_errors, "csum");
168 printf("\tcorrected errors: %llu, uncorrectable errors: %llu, "
169 "unverified errors: %llu\n", p->corrected_errors,
170 p->uncorrectable_errors, p->unverified_errors);
174 #define _SCRUB_FS_STAT(p, name, fs_stat) do { \
175 fs_stat->p.name += p->name; \
178 #define _SCRUB_FS_STAT_MIN(ss, name, fs_stat) \
180 if (fs_stat->s.name > ss->name) { \
181 fs_stat->s.name = ss->name; \
185 #define _SCRUB_FS_STAT_ZMIN(ss, name, fs_stat) \
187 if (!fs_stat->s.name || fs_stat->s.name > ss->name) { \
188 fs_stat->s.name = ss->name; \
192 #define _SCRUB_FS_STAT_ZMAX(ss, name, fs_stat) \
194 if (!(fs_stat)->s.name || (fs_stat)->s.name < (ss)->name) { \
195 (fs_stat)->s.name = (ss)->name; \
199 static void add_to_fs_stat(struct btrfs_scrub_progress *p,
200 struct scrub_stats *ss,
201 struct scrub_fs_stat *fs_stat)
203 _SCRUB_FS_STAT(p, data_extents_scrubbed, fs_stat);
204 _SCRUB_FS_STAT(p, tree_extents_scrubbed, fs_stat);
205 _SCRUB_FS_STAT(p, data_bytes_scrubbed, fs_stat);
206 _SCRUB_FS_STAT(p, tree_bytes_scrubbed, fs_stat);
207 _SCRUB_FS_STAT(p, read_errors, fs_stat);
208 _SCRUB_FS_STAT(p, csum_errors, fs_stat);
209 _SCRUB_FS_STAT(p, verify_errors, fs_stat);
210 _SCRUB_FS_STAT(p, no_csum, fs_stat);
211 _SCRUB_FS_STAT(p, csum_discards, fs_stat);
212 _SCRUB_FS_STAT(p, super_errors, fs_stat);
213 _SCRUB_FS_STAT(p, malloc_errors, fs_stat);
214 _SCRUB_FS_STAT(p, uncorrectable_errors, fs_stat);
215 _SCRUB_FS_STAT(p, corrected_errors, fs_stat);
216 _SCRUB_FS_STAT(p, last_physical, fs_stat);
217 _SCRUB_FS_STAT_ZMIN(ss, t_start, fs_stat);
218 _SCRUB_FS_STAT_ZMIN(ss, t_resumed, fs_stat);
219 _SCRUB_FS_STAT_ZMAX(ss, duration, fs_stat);
220 _SCRUB_FS_STAT_ZMAX(ss, canceled, fs_stat);
221 _SCRUB_FS_STAT_MIN(ss, finished, fs_stat);
224 static void init_fs_stat(struct scrub_fs_stat *fs_stat)
226 memset(fs_stat, 0, sizeof(*fs_stat));
227 fs_stat->s.finished = 1;
230 static void _print_scrub_ss(struct scrub_stats *ss)
237 if (!ss || !ss->t_start) {
238 printf("\tno stats available\n");
242 localtime_r(&ss->t_resumed, &tm);
243 strftime(t, sizeof(t), "%c", &tm);
244 t[sizeof(t) - 1] = '\0';
245 printf("\tscrub resumed at %s", t);
247 localtime_r(&ss->t_start, &tm);
248 strftime(t, sizeof(t), "%c", &tm);
249 t[sizeof(t) - 1] = '\0';
250 printf("\tscrub started at %s", t);
253 seconds = ss->duration;
254 hours = ss->duration / (60 * 60);
255 gmtime_r(&seconds, &tm);
256 strftime(t, sizeof(t), "%M:%S", &tm);
257 if (ss->finished && !ss->canceled) {
258 printf(" and finished after %02u:%s\n", hours, t);
259 } else if (ss->canceled) {
260 printf(" and was aborted after %02u:%s\n", hours, t);
263 printf(", running for %02u:%s\n", hours, t);
265 printf(", interrupted after %02u:%s, not running\n",
270 static void print_scrub_dev(struct btrfs_ioctl_dev_info_args *di,
271 struct btrfs_scrub_progress *p, int raw,
272 const char *append, struct scrub_stats *ss)
274 printf("scrub device %s (id %llu) %s\n", di->path, di->devid,
275 append ? append : "");
283 print_scrub_summary(p);
287 static void print_fs_stat(struct scrub_fs_stat *fs_stat, int raw)
289 _print_scrub_ss(&fs_stat->s);
292 print_scrub_full(&fs_stat->p);
294 print_scrub_summary(&fs_stat->p);
297 static void free_history(struct scrub_file_record **last_scrubs)
299 struct scrub_file_record **l = last_scrubs;
308 * cancels a running scrub and makes the master process record the current
309 * progress status before exiting.
311 static int cancel_fd = -1;
312 static void scrub_sigint_record_progress(int signal)
316 ret = ioctl(cancel_fd, BTRFS_IOC_SCRUB_CANCEL, NULL);
318 perror("Scrub cancel failed");
321 static int scrub_handle_sigint_parent(void)
323 struct sigaction sa = {
324 .sa_handler = SIG_IGN,
325 .sa_flags = SA_RESTART,
328 return sigaction(SIGINT, &sa, NULL);
331 static int scrub_handle_sigint_child(int fd)
333 struct sigaction sa = {
334 .sa_handler = fd == -1 ? SIG_DFL : scrub_sigint_record_progress,
338 return sigaction(SIGINT, &sa, NULL);
341 static int scrub_datafile(const char *fn_base, const char *fn_local,
342 const char *fn_tmp, char *datafile, int size)
347 datafile[end + 1] = '\0';
348 strncpy(datafile, fn_base, end);
349 ret = strlen(datafile);
355 strncpy(datafile + ret + 1, fn_local, end - ret - 1);
356 ret = strlen(datafile);
363 strncpy(datafile + ret + 1, fn_tmp, end - ret - 1);
364 ret = strlen(datafile);
373 static int scrub_open_file(const char *datafile, int m)
378 fd = open(datafile, m, 0600);
382 ret = flock(fd, LOCK_EX|LOCK_NB);
392 static int scrub_open_file_r(const char *fn_base, const char *fn_local)
395 char datafile[PATH_MAX];
396 ret = scrub_datafile(fn_base, fn_local, NULL,
397 datafile, sizeof(datafile));
400 return scrub_open_file(datafile, O_RDONLY);
403 static int scrub_open_file_w(const char *fn_base, const char *fn_local,
407 char datafile[PATH_MAX];
408 ret = scrub_datafile(fn_base, fn_local, tmp,
409 datafile, sizeof(datafile));
412 return scrub_open_file(datafile, O_WRONLY|O_CREAT);
415 static int scrub_rename_file(const char *fn_base, const char *fn_local,
419 char datafile_old[PATH_MAX];
420 char datafile_new[PATH_MAX];
421 ret = scrub_datafile(fn_base, fn_local, tmp,
422 datafile_old, sizeof(datafile_old));
425 ret = scrub_datafile(fn_base, fn_local, NULL,
426 datafile_new, sizeof(datafile_new));
429 ret = rename(datafile_old, datafile_new);
430 return ret ? -errno : 0;
433 #define _SCRUB_KVREAD(ret, i, name, avail, l, dest) if (ret == 0) { \
434 ret = scrub_kvread(i, sizeof(#name), avail, l, #name, dest.name); \
438 * returns 0 if the key did not match (nothing was read)
439 * 1 if the key did match (success)
440 * -1 if the key did match and an error occured
442 static int scrub_kvread(int *i, int len, int avail, const char *buf,
443 const char *key, u64 *dest)
447 if (*i + len + 1 < avail && strncmp(&buf[*i], key, len - 1) == 0) {
452 for (j = 0; isdigit(buf[*i + j]) && *i + j < avail; ++j)
456 *dest = atoll(&buf[*i]);
464 #define _SCRUB_INVALID do { \
466 fprintf(stderr, "WARNING: invalid data in line %d pos " \
467 "%d state %d (near \"%.*s\") at %s:%d\n", \
468 lineno, i, state, 20 > avail ? avail : 20, \
469 l + i, __FILE__, __LINE__); \
473 static struct scrub_file_record **scrub_read_file(int fd, int report_errors)
486 char empty_uuid[BTRFS_FSID_SIZE] = {0};
487 struct scrub_file_record **p = NULL;
490 old_avail = avail - i;
491 BUG_ON(old_avail < 0);
493 memmove(l, l + i, old_avail);
494 avail = read(fd, l + old_avail, sizeof(l) - old_avail);
497 if (avail == 0 && old_avail == 0) {
499 memcmp(p[curr]->fsid, empty_uuid, BTRFS_FSID_SIZE) == 0) {
501 } else if (curr == -1) {
502 p = ERR_PTR(-ENODATA);
508 return ERR_PTR(-errno);
517 case 0: /* start of file */
518 ret = scrub_kvread(&i,
519 sizeof(SCRUB_FILE_VERSION_PREFIX), avail, l,
520 SCRUB_FILE_VERSION_PREFIX, &version);
523 if (version != atoll(SCRUB_FILE_VERSION))
524 return ERR_PTR(-ENOTSUP);
527 case 1: /* start of line, alloc */
529 * this state makes sure we have a complete line in
530 * further processing, so we don't need wrap-tracking
533 if (!eof && !memchr(l + i, '\n', avail - i))
536 if (curr > -1 && memcmp(p[curr]->fsid, empty_uuid,
537 BTRFS_FSID_SIZE) == 0) {
543 p = realloc(p, (curr + 2) * sizeof(*p));
546 return ERR_PTR(-errno);
548 p[curr] = malloc(sizeof(**p));
551 return ERR_PTR(-errno);
553 memset(p[curr], 0, sizeof(**p));
557 case 2: /* start of line, skip space */
558 while (isspace(l[i]) && i < avail) {
564 (!eof && !memchr(l + i, '\n', avail - i)))
568 case 3: /* read fsid */
571 for (j = 0; l[i + j] != ':' && i + j < avail; ++j)
573 if (i + j + 1 >= avail)
575 if (j != BTRFS_UUID_UNPARSED_SIZE - 1)
578 ret = uuid_parse(l + i, p[curr]->fsid);
584 case 4: /* read dev id */
585 for (j = 0; isdigit(l[i + j]) && i+j < avail; ++j)
587 if (j == 0 || i + j + 1 >= avail)
589 p[curr]->devid = atoll(&l[i]);
593 case 5: /* read key/value pair */
595 _SCRUB_KVREAD(ret, &i, data_extents_scrubbed, avail, l,
597 _SCRUB_KVREAD(ret, &i, data_extents_scrubbed, avail, l,
599 _SCRUB_KVREAD(ret, &i, tree_extents_scrubbed, avail, l,
601 _SCRUB_KVREAD(ret, &i, data_bytes_scrubbed, avail, l,
603 _SCRUB_KVREAD(ret, &i, tree_bytes_scrubbed, avail, l,
605 _SCRUB_KVREAD(ret, &i, read_errors, avail, l,
607 _SCRUB_KVREAD(ret, &i, csum_errors, avail, l,
609 _SCRUB_KVREAD(ret, &i, verify_errors, avail, l,
611 _SCRUB_KVREAD(ret, &i, no_csum, avail, l,
613 _SCRUB_KVREAD(ret, &i, csum_discards, avail, l,
615 _SCRUB_KVREAD(ret, &i, super_errors, avail, l,
617 _SCRUB_KVREAD(ret, &i, malloc_errors, avail, l,
619 _SCRUB_KVREAD(ret, &i, uncorrectable_errors, avail, l,
621 _SCRUB_KVREAD(ret, &i, corrected_errors, avail, l,
623 _SCRUB_KVREAD(ret, &i, last_physical, avail, l,
625 _SCRUB_KVREAD(ret, &i, finished, avail, l,
627 _SCRUB_KVREAD(ret, &i, t_start, avail, l,
628 (u64 *)&p[curr]->stats);
629 _SCRUB_KVREAD(ret, &i, t_resumed, avail, l,
630 (u64 *)&p[curr]->stats);
631 _SCRUB_KVREAD(ret, &i, duration, avail, l,
632 (u64 *)&p[curr]->stats);
633 _SCRUB_KVREAD(ret, &i, canceled, avail, l,
639 case 6: /* after number */
642 else if (l[i] == '\n')
648 case 99: /* skip rest of line */
653 if (l[i - 1] == '\n') {
665 static int scrub_write_buf(int fd, const void *data, int len)
668 ret = write(fd, data, len);
672 static int scrub_writev(int fd, char *buf, int max, const char *fmt, ...)
673 __attribute__ ((format (printf, 4, 5)));
674 static int scrub_writev(int fd, char *buf, int max, const char *fmt, ...)
680 ret = vsnprintf(buf, max, fmt, args);
684 return scrub_write_buf(fd, buf, ret);
687 #define _SCRUB_SUM(dest, data, name) dest->scrub_args.progress.name = \
688 data->resumed->p.name + data->scrub_args.progress.name
690 static struct scrub_progress *scrub_resumed_stats(struct scrub_progress *data,
691 struct scrub_progress *dest)
693 if (!data->resumed || data->skip)
696 _SCRUB_SUM(dest, data, data_extents_scrubbed);
697 _SCRUB_SUM(dest, data, tree_extents_scrubbed);
698 _SCRUB_SUM(dest, data, data_bytes_scrubbed);
699 _SCRUB_SUM(dest, data, tree_bytes_scrubbed);
700 _SCRUB_SUM(dest, data, read_errors);
701 _SCRUB_SUM(dest, data, csum_errors);
702 _SCRUB_SUM(dest, data, verify_errors);
703 _SCRUB_SUM(dest, data, no_csum);
704 _SCRUB_SUM(dest, data, csum_discards);
705 _SCRUB_SUM(dest, data, super_errors);
706 _SCRUB_SUM(dest, data, malloc_errors);
707 _SCRUB_SUM(dest, data, uncorrectable_errors);
708 _SCRUB_SUM(dest, data, corrected_errors);
709 _SCRUB_SUM(dest, data, last_physical);
710 dest->stats.canceled = data->stats.canceled;
711 dest->stats.finished = data->stats.finished;
712 dest->stats.t_resumed = data->stats.t_start;
713 dest->stats.t_start = data->resumed->stats.t_start;
714 dest->stats.duration = data->resumed->stats.duration +
715 data->stats.duration;
716 dest->scrub_args.devid = data->scrub_args.devid;
720 #define _SCRUB_KVWRITE(fd, buf, name, use) \
721 scrub_kvwrite(fd, buf, sizeof(buf), #name, \
722 use->scrub_args.progress.name)
724 #define _SCRUB_KVWRITE_STATS(fd, buf, name, use) \
725 scrub_kvwrite(fd, buf, sizeof(buf), #name, \
728 static int scrub_kvwrite(int fd, char *buf, int max, const char *key, u64 val)
730 return scrub_writev(fd, buf, max, "|%s:%lld", key, val);
733 static int scrub_write_file(int fd, const char *fsid,
734 struct scrub_progress *data, int n)
739 struct scrub_progress local;
740 struct scrub_progress *use;
745 /* each -1 is to subtract one \0 byte, the + 2 is for ':' and '\n' */
746 ret = scrub_write_buf(fd, SCRUB_FILE_VERSION_PREFIX ":"
747 SCRUB_FILE_VERSION "\n",
748 (sizeof(SCRUB_FILE_VERSION_PREFIX) - 1) +
749 (sizeof(SCRUB_FILE_VERSION) - 1) + 2);
753 for (i = 0; i < n; ++i) {
754 use = scrub_resumed_stats(&data[i], &local);
755 if (scrub_write_buf(fd, fsid, strlen(fsid)) ||
756 scrub_write_buf(fd, ":", 1) ||
757 scrub_writev(fd, buf, sizeof(buf), "%lld",
758 use->scrub_args.devid) ||
759 scrub_write_buf(fd, buf, ret) ||
760 _SCRUB_KVWRITE(fd, buf, data_extents_scrubbed, use) ||
761 _SCRUB_KVWRITE(fd, buf, tree_extents_scrubbed, use) ||
762 _SCRUB_KVWRITE(fd, buf, data_bytes_scrubbed, use) ||
763 _SCRUB_KVWRITE(fd, buf, tree_bytes_scrubbed, use) ||
764 _SCRUB_KVWRITE(fd, buf, read_errors, use) ||
765 _SCRUB_KVWRITE(fd, buf, csum_errors, use) ||
766 _SCRUB_KVWRITE(fd, buf, verify_errors, use) ||
767 _SCRUB_KVWRITE(fd, buf, no_csum, use) ||
768 _SCRUB_KVWRITE(fd, buf, csum_discards, use) ||
769 _SCRUB_KVWRITE(fd, buf, super_errors, use) ||
770 _SCRUB_KVWRITE(fd, buf, malloc_errors, use) ||
771 _SCRUB_KVWRITE(fd, buf, uncorrectable_errors, use) ||
772 _SCRUB_KVWRITE(fd, buf, corrected_errors, use) ||
773 _SCRUB_KVWRITE(fd, buf, last_physical, use) ||
774 _SCRUB_KVWRITE_STATS(fd, buf, t_start, use) ||
775 _SCRUB_KVWRITE_STATS(fd, buf, t_resumed, use) ||
776 _SCRUB_KVWRITE_STATS(fd, buf, duration, use) ||
777 _SCRUB_KVWRITE_STATS(fd, buf, canceled, use) ||
778 _SCRUB_KVWRITE_STATS(fd, buf, finished, use) ||
779 scrub_write_buf(fd, "\n", 1)) {
787 static int scrub_write_progress(pthread_mutex_t *m, const char *fsid,
788 struct scrub_progress *data, int n)
795 ret = pthread_setcancelstate(PTHREAD_CANCEL_DISABLE, &old);
801 ret = pthread_mutex_lock(m);
807 fd = scrub_open_file_w(SCRUB_DATA_FILE, fsid, "tmp");
812 err = scrub_write_file(fd, fsid, data, n);
815 err = scrub_rename_file(SCRUB_DATA_FILE, fsid, "tmp");
826 ret = pthread_mutex_unlock(m);
831 ret = pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, &old);
839 static void *scrub_one_dev(void *ctx)
841 struct scrub_progress *sp = ctx;
845 sp->stats.canceled = 0;
846 sp->stats.duration = 0;
847 sp->stats.finished = 0;
849 ret = syscall(SYS_ioprio_set, IOPRIO_WHO_PROCESS, 0,
850 IOPRIO_PRIO_VALUE(sp->ioprio_class,
851 sp->ioprio_classdata));
854 "WARNING: setting ioprio failed: %s (ignored).\n",
857 ret = ioctl(sp->fd, BTRFS_IOC_SCRUB, &sp->scrub_args);
858 gettimeofday(&tv, NULL);
860 sp->stats.duration = tv.tv_sec - sp->stats.t_start;
861 sp->stats.canceled = !!ret;
862 sp->ioctl_errno = errno;
863 ret = pthread_mutex_lock(&sp->progress_mutex);
865 return ERR_PTR(-ret);
866 sp->stats.finished = 1;
867 ret = pthread_mutex_unlock(&sp->progress_mutex);
869 return ERR_PTR(-ret);
874 static void *progress_one_dev(void *ctx)
876 struct scrub_progress *sp = ctx;
878 sp->ret = ioctl(sp->fd, BTRFS_IOC_SCRUB_PROGRESS, &sp->scrub_args);
879 sp->ioctl_errno = errno;
884 /* nb: returns a negative errno via ERR_PTR */
885 static void *scrub_progress_cycle(void *ctx)
888 int perr = 0; /* positive / pthread error returns */
891 char fsid[BTRFS_UUID_UNPARSED_SIZE];
892 struct scrub_progress *sp;
893 struct scrub_progress *sp_last;
894 struct scrub_progress *sp_shared;
896 struct scrub_progress_cycle *spc = ctx;
897 int ndev = spc->fi->num_devices;
901 struct pollfd accept_poll_fd = {
906 struct pollfd write_poll_fd = {
910 struct sockaddr_un peer;
911 socklen_t peer_size = sizeof(peer);
913 perr = pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, &old);
917 uuid_unparse(spc->fi->fsid, fsid);
919 for (i = 0; i < ndev; ++i) {
920 sp = &spc->progress[i];
921 sp_last = &spc->progress[i + ndev];
922 sp_shared = &spc->shared_progress[i];
923 sp->scrub_args.devid = sp_last->scrub_args.devid =
924 sp_shared->scrub_args.devid;
925 sp->fd = sp_last->fd = spc->fdmnt;
926 sp->stats.t_start = sp_last->stats.t_start =
927 sp_shared->stats.t_start;
928 sp->resumed = sp_last->resumed = sp_shared->resumed;
929 sp->skip = sp_last->skip = sp_shared->skip;
930 sp->stats.finished = sp_last->stats.finished =
931 sp_shared->stats.finished;
935 ret = poll(&accept_poll_fd, 1, 5 * 1000);
941 peer_fd = accept(spc->prg_fd, (struct sockaddr *)&peer,
943 gettimeofday(&tv, NULL);
946 for (i = 0; i < ndev; ++i) {
947 sp = &spc->progress[this * ndev + i];
948 sp_last = &spc->progress[last * ndev + i];
949 sp_shared = &spc->shared_progress[i];
950 if (sp->stats.finished)
952 progress_one_dev(sp);
953 sp->stats.duration = tv.tv_sec - sp->stats.t_start;
956 if (sp->ioctl_errno != ENOTCONN &&
957 sp->ioctl_errno != ENODEV) {
958 ret = -sp->ioctl_errno;
962 * scrub finished or device removed, check the
963 * finished flag. if unset, just use the last
964 * result we got for the current write and go
965 * on. flag should be set on next cycle, then.
967 perr = pthread_setcancelstate(
968 PTHREAD_CANCEL_DISABLE, &old);
971 perr = pthread_mutex_lock(&sp_shared->progress_mutex);
974 if (!sp_shared->stats.finished) {
975 perr = pthread_mutex_unlock(
976 &sp_shared->progress_mutex);
979 perr = pthread_setcancelstate(
980 PTHREAD_CANCEL_ENABLE, &old);
983 memcpy(sp, sp_last, sizeof(*sp));
986 perr = pthread_mutex_unlock(&sp_shared->progress_mutex);
989 perr = pthread_setcancelstate(
990 PTHREAD_CANCEL_ENABLE, &old);
993 memcpy(sp, sp_shared, sizeof(*sp));
994 memcpy(sp_last, sp_shared, sizeof(*sp));
997 write_poll_fd.fd = peer_fd;
998 ret = poll(&write_poll_fd, 1, 0);
1004 ret = scrub_write_file(
1006 &spc->progress[this * ndev], ndev);
1013 if (!spc->do_record)
1015 ret = scrub_write_progress(spc->write_mutex, fsid,
1016 &spc->progress[this * ndev], ndev);
1025 return ERR_PTR(ret);
1028 static struct scrub_file_record *last_dev_scrub(
1029 struct scrub_file_record *const *const past_scrubs, u64 devid)
1033 if (!past_scrubs || IS_ERR(past_scrubs))
1036 for (i = 0; past_scrubs[i]; ++i)
1037 if (past_scrubs[i]->devid == devid)
1038 return past_scrubs[i];
1043 static int mkdir_p(char *path)
1048 for (i = 1; i < strlen(path); ++i) {
1052 ret = mkdir(path, 0777);
1053 if (ret && errno != EEXIST)
1061 static int is_scrub_running_on_fs(struct btrfs_ioctl_fs_info_args *fi_args,
1062 struct btrfs_ioctl_dev_info_args *di_args,
1063 struct scrub_file_record **past_scrubs)
1067 if (!fi_args || !di_args || !past_scrubs)
1070 for (i = 0; i < fi_args->num_devices; i++) {
1071 struct scrub_file_record *sfr =
1072 last_dev_scrub(past_scrubs, di_args[i].devid);
1076 if (!(sfr->stats.finished || sfr->stats.canceled))
1082 static int is_scrub_running_in_kernel(int fd,
1083 struct btrfs_ioctl_dev_info_args *di_args, u64 max_devices)
1085 struct scrub_progress sp;
1089 for (i = 0; i < max_devices; i++) {
1090 memset(&sp, 0, sizeof(sp));
1091 sp.scrub_args.devid = di_args[i].devid;
1092 ret = ioctl(fd, BTRFS_IOC_SCRUB_PROGRESS, &sp.scrub_args);
1100 static const char * const cmd_scrub_start_usage[];
1101 static const char * const cmd_scrub_resume_usage[];
1103 static int scrub_start(int argc, char **argv, int resume)
1113 int e_uncorrectable = 0;
1114 int e_correctable = 0;
1117 int do_background = 1;
1123 int do_stats_per_dev = 0;
1124 int ioprio_class = IOPRIO_CLASS_IDLE;
1125 int ioprio_classdata = 0;
1129 struct btrfs_ioctl_fs_info_args fi_args;
1130 struct btrfs_ioctl_dev_info_args *di_args = NULL;
1131 struct scrub_progress *sp = NULL;
1132 struct scrub_fs_stat fs_stat;
1134 struct sockaddr_un addr = {
1135 .sun_family = AF_UNIX,
1137 pthread_t *t_devs = NULL;
1139 struct scrub_file_record **past_scrubs = NULL;
1140 struct scrub_file_record *last_scrub = NULL;
1141 char *datafile = strdup(SCRUB_DATA_FILE);
1142 char fsid[BTRFS_UUID_UNPARSED_SIZE];
1143 char sock_path[PATH_MAX] = "";
1144 struct scrub_progress_cycle spc;
1145 pthread_mutex_t spc_write_mutex = PTHREAD_MUTEX_INITIALIZER;
1148 DIR *dirstream = NULL;
1150 int nothing_to_resume = 0;
1153 while ((c = getopt(argc, argv, "BdqrRc:n:f")) != -1) {
1161 do_stats_per_dev = 1;
1173 ioprio_class = (int)strtol(optarg, NULL, 10);
1176 ioprio_classdata = (int)strtol(optarg, NULL, 10);
1183 usage(resume ? cmd_scrub_resume_usage :
1184 cmd_scrub_start_usage);
1188 /* try to catch most error cases before forking */
1190 if (check_argc_exact(argc - optind, 1)) {
1191 usage(resume ? cmd_scrub_resume_usage :
1192 cmd_scrub_start_usage);
1195 spc.progress = NULL;
1196 if (do_quiet && do_print)
1199 if (mkdir_p(datafile)) {
1200 ERR(!do_quiet, "WARNING: cannot create scrub data "
1201 "file, mkdir %s failed: %s. Status recording "
1202 "disabled\n", datafile, strerror(errno));
1207 path = argv[optind];
1209 fdmnt = open_path_or_dev_mnt(path, &dirstream);
1212 if (errno == EINVAL)
1214 "ERROR: '%s' is not a mounted btrfs device\n",
1217 ERR(!do_quiet, "ERROR: can't access '%s': %s\n",
1218 path, strerror(errno));
1222 ret = get_fs_info(path, &fi_args, &di_args);
1224 ERR(!do_quiet, "ERROR: getting dev info for scrub failed: "
1225 "%s\n", strerror(-ret));
1229 if (!fi_args.num_devices) {
1230 ERR(!do_quiet, "ERROR: no devices found\n");
1235 uuid_unparse(fi_args.fsid, fsid);
1236 fdres = scrub_open_file_r(SCRUB_DATA_FILE, fsid);
1237 if (fdres < 0 && fdres != -ENOENT) {
1238 ERR(!do_quiet, "WARNING: failed to open status file: "
1239 "%s\n", strerror(-fdres));
1240 } else if (fdres >= 0) {
1241 past_scrubs = scrub_read_file(fdres, !do_quiet);
1242 if (IS_ERR(past_scrubs))
1243 ERR(!do_quiet, "WARNING: failed to read status file: "
1244 "%s\n", strerror(-PTR_ERR(past_scrubs)));
1249 * Check for stale information in the status file, ie. if it's
1250 * canceled=0, finished=0 but no scrub is running.
1252 if (!is_scrub_running_in_kernel(fdmnt, di_args, fi_args.num_devices))
1256 * check whether any involved device is already busy running a
1257 * scrub. This would cause damaged status messages and the state
1258 * "aborted" without the explanation that a scrub was already
1259 * running. Therefore check it first, prevent it and give some
1260 * feedback to the user if scrub is already running.
1261 * Note that if scrub is started with a block device as the
1262 * parameter, only that particular block device is checked. It
1263 * is a normal mode of operation to start scrub on multiple
1264 * single devices, there is no reason to prevent this.
1266 if (!force && is_scrub_running_on_fs(&fi_args, di_args, past_scrubs)) {
1268 "ERROR: scrub is already running.\n"
1269 "To cancel use 'btrfs scrub cancel %s'.\n"
1270 "To see the status use 'btrfs scrub status [-d] %s'.\n",
1276 t_devs = malloc(fi_args.num_devices * sizeof(*t_devs));
1277 sp = calloc(fi_args.num_devices, sizeof(*sp));
1278 spc.progress = calloc(fi_args.num_devices * 2, sizeof(*spc.progress));
1280 if (!t_devs || !sp || !spc.progress) {
1281 ERR(!do_quiet, "ERROR: scrub failed: %s", strerror(errno));
1286 for (i = 0; i < fi_args.num_devices; ++i) {
1287 devid = di_args[i].devid;
1288 ret = pthread_mutex_init(&sp[i].progress_mutex, NULL);
1290 ERR(!do_quiet, "ERROR: pthread_mutex_init failed: "
1291 "%s\n", strerror(ret));
1295 last_scrub = last_dev_scrub(past_scrubs, devid);
1296 sp[i].scrub_args.devid = devid;
1298 if (resume && last_scrub && (last_scrub->stats.canceled ||
1299 !last_scrub->stats.finished)) {
1301 sp[i].scrub_args.start = last_scrub->p.last_physical;
1302 sp[i].resumed = last_scrub;
1303 } else if (resume) {
1306 sp[i].resumed = last_scrub;
1310 sp[i].scrub_args.start = 0ll;
1311 sp[i].resumed = NULL;
1314 sp[i].scrub_args.end = (u64)-1ll;
1315 sp[i].scrub_args.flags = readonly ? BTRFS_SCRUB_READONLY : 0;
1316 sp[i].ioprio_class = ioprio_class;
1317 sp[i].ioprio_classdata = ioprio_classdata;
1320 if (!n_start && !n_resume) {
1322 printf("scrub: nothing to resume for %s, fsid %s\n",
1324 nothing_to_resume = 1;
1328 ret = prg_fd = socket(AF_UNIX, SOCK_STREAM, 0);
1330 ret = scrub_datafile(SCRUB_PROGRESS_SOCKET_PATH, fsid, NULL,
1331 sock_path, sizeof(sock_path));
1332 /* ignore EOVERFLOW, try using a shorter path for the socket */
1333 addr.sun_path[sizeof(addr.sun_path) - 1] = '\0';
1334 strncpy(addr.sun_path, sock_path, sizeof(addr.sun_path) - 1);
1335 ret = bind(prg_fd, (struct sockaddr *)&addr, sizeof(addr));
1336 if (ret != -1 || errno != EADDRINUSE)
1339 * bind failed with EADDRINUSE. so let's see if anyone answers
1340 * when we make a call to the socket ...
1342 ret = connect(prg_fd, (struct sockaddr *)&addr, sizeof(addr));
1343 if (!ret || errno != ECONNREFUSED) {
1344 /* ... yes, so scrub must be running. error out */
1345 fprintf(stderr, "ERROR: scrub already running\n");
1351 * ... no, this means someone left us alone with an unused
1352 * socket in the file system. remove it and try again.
1354 ret = unlink(sock_path);
1357 ret = listen(prg_fd, 100);
1359 ERR(!do_quiet, "WARNING: failed to open the progress status "
1360 "socket at %s: %s. Progress cannot be queried\n",
1361 sock_path[0] ? sock_path : SCRUB_PROGRESS_SOCKET_PATH,
1372 /* write all-zero progress file for a start */
1373 ret = scrub_write_progress(&spc_write_mutex, fsid, sp,
1374 fi_args.num_devices);
1376 ERR(!do_quiet, "WARNING: failed to write the progress "
1377 "status file: %s. Status recording disabled\n",
1383 if (do_background) {
1386 ERR(!do_quiet, "ERROR: cannot scrub, fork failed: "
1387 "%s\n", strerror(errno));
1394 scrub_handle_sigint_parent();
1396 printf("scrub %s on %s, fsid %s (pid=%d)\n",
1397 n_start ? "started" : "resumed",
1405 ERR(!do_quiet, "ERROR: wait failed: (ret=%d) "
1406 "%s\n", ret, strerror(errno));
1410 if (!WIFEXITED(stat) || WEXITSTATUS(stat)) {
1411 ERR(!do_quiet, "ERROR: scrub process failed\n");
1412 err = WIFEXITED(stat) ? WEXITSTATUS(stat) : -1;
1420 scrub_handle_sigint_child(fdmnt);
1422 for (i = 0; i < fi_args.num_devices; ++i) {
1424 sp[i].scrub_args.progress = sp[i].resumed->p;
1425 sp[i].stats = sp[i].resumed->stats;
1427 sp[i].stats.finished = 1;
1430 devid = di_args[i].devid;
1431 gettimeofday(&tv, NULL);
1432 sp[i].stats.t_start = tv.tv_sec;
1433 ret = pthread_create(&t_devs[i], NULL,
1434 scrub_one_dev, &sp[i]);
1437 fprintf(stderr, "ERROR: creating "
1438 "scrub_one_dev[%llu] thread failed: "
1439 "%s\n", devid, strerror(ret));
1446 spc.prg_fd = prg_fd;
1447 spc.do_record = do_record;
1448 spc.write_mutex = &spc_write_mutex;
1449 spc.shared_progress = sp;
1451 ret = pthread_create(&t_prog, NULL, scrub_progress_cycle, &spc);
1454 fprintf(stderr, "ERROR: creating progress thread "
1455 "failed: %s\n", strerror(ret));
1461 for (i = 0; i < fi_args.num_devices; ++i) {
1464 devid = di_args[i].devid;
1465 ret = pthread_join(t_devs[i], NULL);
1468 fprintf(stderr, "ERROR: pthread_join failed "
1469 "for scrub_one_dev[%llu]: %s\n", devid,
1474 if (sp[i].ret && sp[i].ioctl_errno == ENODEV) {
1476 fprintf(stderr, "WARNING: device %lld not "
1477 "present\n", devid);
1480 if (sp[i].ret && sp[i].ioctl_errno == ECANCELED) {
1482 } else if (sp[i].ret) {
1484 fprintf(stderr, "ERROR: scrubbing %s failed "
1485 "for device id %lld (%s)\n", path,
1486 devid, strerror(sp[i].ioctl_errno));
1490 if (sp[i].scrub_args.progress.uncorrectable_errors > 0)
1492 if (sp[i].scrub_args.progress.corrected_errors > 0
1493 || sp[i].scrub_args.progress.unverified_errors > 0)
1498 const char *append = "done";
1499 if (!do_stats_per_dev)
1500 init_fs_stat(&fs_stat);
1501 for (i = 0; i < fi_args.num_devices; ++i) {
1502 if (do_stats_per_dev) {
1503 print_scrub_dev(&di_args[i],
1504 &sp[i].scrub_args.progress,
1506 sp[i].ret ? "canceled" : "done",
1510 append = "canceled";
1511 add_to_fs_stat(&sp[i].scrub_args.progress,
1512 &sp[i].stats, &fs_stat);
1515 if (!do_stats_per_dev) {
1516 printf("scrub %s for %s\n", append, fsid);
1517 print_fs_stat(&fs_stat, print_raw);
1521 ret = pthread_cancel(t_prog);
1523 ret = pthread_join(t_prog, &terr);
1525 /* check for errors from the handling of the progress thread */
1526 if (do_print && ret) {
1527 fprintf(stderr, "ERROR: progress thread handling failed: %s\n",
1531 /* check for errors returned from the progress thread itself */
1532 if (do_print && terr && terr != PTHREAD_CANCELED) {
1533 fprintf(stderr, "ERROR: recording progress "
1534 "failed: %s\n", strerror(-PTR_ERR(terr)));
1538 ret = scrub_write_progress(&spc_write_mutex, fsid, sp,
1539 fi_args.num_devices);
1540 if (ret && do_print) {
1541 fprintf(stderr, "ERROR: failed to record the result: "
1542 "%s\n", strerror(-ret));
1546 scrub_handle_sigint_child(-1);
1549 free_history(past_scrubs);
1559 close_file_or_dir(fdmnt, dirstream);
1563 if (nothing_to_resume)
1565 if (e_uncorrectable) {
1566 ERR(!do_quiet, "ERROR: There are uncorrectable errors.\n");
1570 ERR(!do_quiet, "WARNING: errors detected during scrubbing, corrected.\n");
1575 static const char * const cmd_scrub_start_usage[] = {
1576 "btrfs scrub start [-BdqrRf] [-c ioprio_class -n ioprio_classdata] <path>|<device>",
1577 "Start a new scrub. If a scrub is already running, the new one fails.",
1579 "-B do not background",
1580 "-d stats per device (-B only)",
1582 "-r read only mode",
1583 "-R raw print mode, print full data instead of summary",
1584 "-c set ioprio class (see ionice(1) manpage)",
1585 "-n set ioprio classdata (see ionice(1) manpage)",
1586 "-f force starting new scrub even if a scrub is already running",
1587 " this is useful when scrub stats record file is damaged",
1591 static int cmd_scrub_start(int argc, char **argv)
1593 return scrub_start(argc, argv, 0);
1596 static const char * const cmd_scrub_cancel_usage[] = {
1597 "btrfs scrub cancel <path>|<device>",
1598 "Cancel a running scrub",
1602 static int cmd_scrub_cancel(int argc, char **argv)
1607 DIR *dirstream = NULL;
1609 if (check_argc_exact(argc, 2))
1610 usage(cmd_scrub_cancel_usage);
1614 fdmnt = open_path_or_dev_mnt(path, &dirstream);
1616 if (errno == EINVAL)
1618 "ERROR: '%s' is not a mounted btrfs device\n",
1621 fprintf(stderr, "ERROR: can't access '%s': %s\n",
1622 path, strerror(errno));
1627 ret = ioctl(fdmnt, BTRFS_IOC_SCRUB_CANCEL, NULL);
1630 fprintf(stderr, "ERROR: scrub cancel failed on %s: %s\n", path,
1631 errno == ENOTCONN ? "not running" : strerror(errno));
1632 if (errno == ENOTCONN)
1640 printf("scrub cancelled\n");
1643 close_file_or_dir(fdmnt, dirstream);
1647 static const char * const cmd_scrub_resume_usage[] = {
1648 "btrfs scrub resume [-BdqrR] [-c ioprio_class -n ioprio_classdata] <path>|<device>",
1649 "Resume previously canceled or interrupted scrub",
1651 "-B do not background",
1652 "-d stats per device (-B only)",
1654 "-r read only mode",
1655 "-R raw print mode, print full data instead of summary",
1656 "-c set ioprio class (see ionice(1) manpage)",
1657 "-n set ioprio classdata (see ionice(1) manpage)",
1661 static int cmd_scrub_resume(int argc, char **argv)
1663 return scrub_start(argc, argv, 1);
1666 static const char * const cmd_scrub_status_usage[] = {
1667 "btrfs scrub status [-dR] <path>|<device>",
1668 "Show status of running or finished scrub",
1670 "-d stats per device",
1671 "-R print raw stats",
1675 static int cmd_scrub_status(int argc, char **argv)
1678 struct btrfs_ioctl_fs_info_args fi_args;
1679 struct btrfs_ioctl_dev_info_args *di_args = NULL;
1680 struct scrub_file_record **past_scrubs = NULL;
1681 struct scrub_file_record *last_scrub;
1682 struct scrub_fs_stat fs_stat;
1683 struct sockaddr_un addr = {
1684 .sun_family = AF_UNIX,
1691 int do_stats_per_dev = 0;
1693 char fsid[BTRFS_UUID_UNPARSED_SIZE];
1696 DIR *dirstream = NULL;
1699 while ((c = getopt(argc, argv, "dR")) != -1) {
1702 do_stats_per_dev = 1;
1709 usage(cmd_scrub_status_usage);
1713 if (check_argc_exact(argc - optind, 1))
1714 usage(cmd_scrub_status_usage);
1716 path = argv[optind];
1718 fdmnt = open_path_or_dev_mnt(path, &dirstream);
1721 if (errno == EINVAL)
1723 "ERROR: '%s' is not a mounted btrfs device\n",
1726 fprintf(stderr, "ERROR: can't access '%s': %s\n",
1727 path, strerror(errno));
1731 ret = get_fs_info(path, &fi_args, &di_args);
1733 fprintf(stderr, "ERROR: getting dev info for scrub failed: "
1734 "%s\n", strerror(-ret));
1738 if (!fi_args.num_devices) {
1739 fprintf(stderr, "ERROR: no devices found\n");
1744 uuid_unparse(fi_args.fsid, fsid);
1746 fdres = socket(AF_UNIX, SOCK_STREAM, 0);
1748 fprintf(stderr, "ERROR: failed to create socket to "
1749 "receive progress information: %s\n",
1754 scrub_datafile(SCRUB_PROGRESS_SOCKET_PATH, fsid,
1755 NULL, addr.sun_path, sizeof(addr.sun_path));
1756 /* ignore EOVERFLOW, just use shorter name and hope for the best */
1757 addr.sun_path[sizeof(addr.sun_path) - 1] = '\0';
1758 ret = connect(fdres, (struct sockaddr *)&addr, sizeof(addr));
1761 fdres = scrub_open_file_r(SCRUB_DATA_FILE, fsid);
1762 if (fdres < 0 && fdres != -ENOENT) {
1763 fprintf(stderr, "WARNING: failed to open status file: "
1764 "%s\n", strerror(-fdres));
1771 past_scrubs = scrub_read_file(fdres, 1);
1772 if (IS_ERR(past_scrubs))
1773 fprintf(stderr, "WARNING: failed to read status: %s\n",
1774 strerror(-PTR_ERR(past_scrubs)));
1776 in_progress = is_scrub_running_in_kernel(fdmnt, di_args, fi_args.num_devices);
1778 printf("scrub status for %s\n", fsid);
1780 if (do_stats_per_dev) {
1781 for (i = 0; i < fi_args.num_devices; ++i) {
1782 last_scrub = last_dev_scrub(past_scrubs,
1785 print_scrub_dev(&di_args[i], NULL, print_raw,
1789 last_scrub->stats.in_progress = in_progress;
1790 print_scrub_dev(&di_args[i], &last_scrub->p, print_raw,
1791 last_scrub->stats.finished ?
1792 "history" : "status",
1793 &last_scrub->stats);
1796 init_fs_stat(&fs_stat);
1797 fs_stat.s.in_progress = in_progress;
1798 for (i = 0; i < fi_args.num_devices; ++i) {
1799 last_scrub = last_dev_scrub(past_scrubs,
1803 add_to_fs_stat(&last_scrub->p, &last_scrub->stats,
1806 print_fs_stat(&fs_stat, print_raw);
1810 free_history(past_scrubs);
1814 close_file_or_dir(fdmnt, dirstream);
1819 static const char scrub_cmd_group_info[] =
1820 "verify checksums of data and metadata";
1822 const struct cmd_group scrub_cmd_group = {
1823 scrub_cmd_group_usage, scrub_cmd_group_info, {
1824 { "start", cmd_scrub_start, cmd_scrub_start_usage, NULL, 0 },
1825 { "cancel", cmd_scrub_cancel, cmd_scrub_cancel_usage, NULL, 0 },
1826 { "resume", cmd_scrub_resume, cmd_scrub_resume_usage, NULL, 0 },
1827 { "status", cmd_scrub_status, cmd_scrub_status_usage, NULL, 0 },
1832 int cmd_scrub(int argc, char **argv)
1834 return handle_command_group(&scrub_cmd_group, argc, argv);