summaryrefslogtreecommitdiffstats
path: root/src/shared/rm-rf.c
blob: 4e5c53e3ca4f69586cf4b0983458b826f1762e26 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
/* SPDX-License-Identifier: LGPL-2.1-or-later */

#include <errno.h>
#include <fcntl.h>
#include <stdbool.h>
#include <stddef.h>
#include <unistd.h>

#include "alloc-util.h"
#include "btrfs-util.h"
#include "cgroup-util.h"
#include "dirent-util.h"
#include "fd-util.h"
#include "fs-util.h"
#include "log.h"
#include "macro.h"
#include "mountpoint-util.h"
#include "path-util.h"
#include "rm-rf.h"
#include "stat-util.h"
#include "string-util.h"

/* We treat tmpfs/ramfs + cgroupfs as non-physical file systems. cgroupfs is similar to tmpfs in a way
 * after all: we can create arbitrary directory hierarchies in it, and hence can also use rm_rf() on it
 * to remove those again. */
static bool is_physical_fs(const struct statfs *sfs) {
        return !is_temporary_fs(sfs) && !is_cgroup_fs(sfs);
}

static int patch_dirfd_mode(
                int dfd,
                bool refuse_already_set,
                mode_t *ret_old_mode) {

        struct stat st;
        int r;

        assert(dfd >= 0);
        assert(ret_old_mode);

        if (fstat(dfd, &st) < 0)
                return -errno;
        if (!S_ISDIR(st.st_mode))
                return -ENOTDIR;

        if (FLAGS_SET(st.st_mode, 0700)) { /* Already set? */
                if (refuse_already_set)
                        return -EACCES; /* original error */

                *ret_old_mode = st.st_mode;
                return 0;
        }

        if (st.st_uid != geteuid())  /* this only works if the UID matches ours */
                return -EACCES;

        r = fchmod_opath(dfd, (st.st_mode | 0700) & 07777);
        if (r < 0)
                return r;

        *ret_old_mode = st.st_mode;
        return 1;
}

int unlinkat_harder(int dfd, const char *filename, int unlink_flags, RemoveFlags remove_flags) {
        mode_t old_mode;
        int r;

        /* Like unlinkat(), but tries harder: if we get EACCESS we'll try to set the r/w/x bits on the
         * directory. This is useful if we run unprivileged and have some files where the w bit is
         * missing. */

        if (unlinkat(dfd, filename, unlink_flags) >= 0)
                return 0;
        if (errno != EACCES || !FLAGS_SET(remove_flags, REMOVE_CHMOD))
                return -errno;

        r = patch_dirfd_mode(dfd, /* refuse_already_set = */ true, &old_mode);
        if (r < 0)
                return r;

        if (unlinkat(dfd, filename, unlink_flags) < 0) {
                r = -errno;
                /* Try to restore the original access mode if this didn't work */
                (void) fchmod(dfd, old_mode & 07777);
                return r;
        }

        if (FLAGS_SET(remove_flags, REMOVE_CHMOD_RESTORE) && fchmod(dfd, old_mode & 07777) < 0)
                return -errno;

        /* If this worked, we won't reset the old mode by default, since we'll need it for other entries too,
         * and we should destroy the whole thing */
        return 0;
}

int fstatat_harder(int dfd,
                const char *filename,
                struct stat *ret,
                int fstatat_flags,
                RemoveFlags remove_flags) {

        mode_t old_mode;
        int r;

        /* Like unlink_harder() but does the same for fstatat() */

        if (fstatat(dfd, filename, ret, fstatat_flags) >= 0)
                return 0;
        if (errno != EACCES || !FLAGS_SET(remove_flags, REMOVE_CHMOD))
                return -errno;

        r = patch_dirfd_mode(dfd, /* refuse_already_set = */ true, &old_mode);
        if (r < 0)
                return r;

        if (fstatat(dfd, filename, ret, fstatat_flags) < 0) {
                r = -errno;
                (void) fchmod(dfd, old_mode & 07777);
                return r;
        }

        if (FLAGS_SET(remove_flags, REMOVE_CHMOD_RESTORE) && fchmod(dfd, old_mode & 07777) < 0)
                return -errno;

        return 0;
}

static int openat_harder(int dfd, const char *path, int open_flags, RemoveFlags remove_flags, mode_t *ret_old_mode) {
        _cleanup_close_ int pfd = -EBADF, fd = -EBADF;
        bool chmod_done = false;
        mode_t old_mode;
        int r;

        assert(dfd >= 0 || dfd == AT_FDCWD);
        assert(path);

        /* Unlike unlink_harder() and fstatat_harder(), this chmod the specified path. */

        if (FLAGS_SET(open_flags, O_PATH) ||
            !FLAGS_SET(open_flags, O_DIRECTORY) ||
            !FLAGS_SET(remove_flags, REMOVE_CHMOD)) {

                fd = RET_NERRNO(openat(dfd, path, open_flags));
                if (fd < 0)
                        return fd;

                if (ret_old_mode) {
                        struct stat st;

                        if (fstat(fd, &st) < 0)
                                return -errno;

                        *ret_old_mode = st.st_mode;
                }

                return TAKE_FD(fd);
        }

        pfd = RET_NERRNO(openat(dfd, path, (open_flags & (O_CLOEXEC|O_DIRECTORY|O_NOFOLLOW)) | O_PATH));
        if (pfd < 0)
                return pfd;

        if (FLAGS_SET(remove_flags, REMOVE_CHMOD)) {
                r = patch_dirfd_mode(pfd, /* refuse_already_set = */ false, &old_mode);
                if (r < 0)
                        return r;

                chmod_done = r;
        }

        fd = fd_reopen(pfd, open_flags);
        if (fd < 0) {
                if (chmod_done)
                        (void) fchmod_opath(pfd, old_mode & 07777);
                return fd;
        }

        if (ret_old_mode)
                *ret_old_mode = old_mode;

        return TAKE_FD(fd);
}

static int rm_rf_children_impl(
                int fd,
                RemoveFlags flags,
                const struct stat *root_dev,
                mode_t old_mode);

static int rm_rf_inner_child(
                int fd,
                const char *fname,
                int is_dir,
                RemoveFlags flags,
                const struct stat *root_dev,
                bool allow_recursion) {

        struct stat st;
        int r, q = 0;

        assert(fd >= 0);
        assert(fname);

        if (is_dir < 0 ||
            root_dev ||
            (is_dir > 0 && (root_dev || (flags & REMOVE_SUBVOLUME)))) {

                r = fstatat_harder(fd, fname, &st, AT_SYMLINK_NOFOLLOW, flags);
                if (r < 0)
                        return r;

                is_dir = S_ISDIR(st.st_mode);
        }

        if (is_dir) {
                /* If root_dev is set, remove subdirectories only if device is same */
                if (root_dev && st.st_dev != root_dev->st_dev)
                        return 0;

                /* Stop at mount points */
                r = fd_is_mount_point(fd, fname, 0);
                if (r < 0)
                        return r;
                if (r > 0)
                        return 0;

                if ((flags & REMOVE_SUBVOLUME) && btrfs_might_be_subvol(&st)) {
                        /* This could be a subvolume, try to remove it */

                        r = btrfs_subvol_remove_fd(fd, fname, BTRFS_REMOVE_RECURSIVE|BTRFS_REMOVE_QUOTA);
                        if (r < 0) {
                                if (!IN_SET(r, -ENOTTY, -EINVAL))
                                        return r;

                                /* ENOTTY, then it wasn't a btrfs subvolume, continue below. */
                        } else
                                /* It was a subvolume, done. */
                                return 1;
                }

                if (!allow_recursion)
                        return -EISDIR;

                mode_t old_mode;
                int subdir_fd = openat_harder(fd, fname,
                                              O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC|O_NOFOLLOW|O_NOATIME,
                                              flags, &old_mode);
                if (subdir_fd < 0)
                        return subdir_fd;

                /* We pass REMOVE_PHYSICAL here, to avoid doing the fstatfs() to check the file system type
                 * again for each directory */
                q = rm_rf_children_impl(subdir_fd, flags | REMOVE_PHYSICAL, root_dev, old_mode);

        } else if (flags & REMOVE_ONLY_DIRECTORIES)
                return 0;

        r = unlinkat_harder(fd, fname, is_dir ? AT_REMOVEDIR : 0, flags);
        if (r < 0)
                return r;
        if (q < 0)
                return q;
        return 1;
}

typedef struct TodoEntry {
        DIR *dir;         /* A directory that we were operating on. */
        char *dirname;    /* The filename of that directory itself. */
        mode_t old_mode;  /* The original file mode. */
} TodoEntry;

static void free_todo_entries(TodoEntry **todos) {
        for (TodoEntry *x = *todos; x && x->dir; x++) {
                closedir(x->dir);
                free(x->dirname);
        }

        freep(todos);
}

int rm_rf_children(
                int fd,
                RemoveFlags flags,
                const struct stat *root_dev) {

        struct stat st;

        assert(fd >= 0);

        if (fstat(fd, &st) < 0)
                return -errno;

        return rm_rf_children_impl(fd, flags, root_dev, st.st_mode);
}

static int rm_rf_children_impl(
                int fd,
                RemoveFlags flags,
                const struct stat *root_dev,
                mode_t old_mode) {

        _cleanup_(free_todo_entries) TodoEntry *todos = NULL;
        size_t n_todo = 0;
        _cleanup_free_ char *dirname = NULL; /* Set when we are recursing and want to delete ourselves */
        int ret = 0, r;

        /* Return the first error we run into, but nevertheless try to go on.
         * The passed fd is closed in all cases, including on failure. */

        for (;;) {  /* This loop corresponds to the directory nesting level. */
                _cleanup_closedir_ DIR *d = NULL;

                if (n_todo > 0) {
                        /* We know that we are in recursion here, because n_todo is set.
                         * We need to remove the inner directory we were operating on. */
                        assert(dirname);
                        r = unlinkat_harder(dirfd(todos[n_todo-1].dir), dirname, AT_REMOVEDIR, flags);
                        if (r < 0 && r != -ENOENT) {
                                if (ret == 0)
                                        ret = r;

                                if (FLAGS_SET(flags, REMOVE_CHMOD_RESTORE))
                                        (void) fchmodat(dirfd(todos[n_todo-1].dir), dirname, old_mode & 07777, 0);
                        }
                        dirname = mfree(dirname);

                        /* And now let's back out one level up */
                        n_todo --;
                        d = TAKE_PTR(todos[n_todo].dir);
                        dirname = TAKE_PTR(todos[n_todo].dirname);
                        old_mode = todos[n_todo].old_mode;

                        assert(d);
                        fd = dirfd(d); /* Retrieve the file descriptor from the DIR object */
                        assert(fd >= 0);
                } else {
        next_fd:
                        assert(fd >= 0);
                        d = fdopendir(fd);
                        if (!d) {
                                safe_close(fd);
                                return -errno;
                        }
                        fd = dirfd(d); /* We donated the fd to fdopendir(). Let's make sure we sure we have
                                        * the right descriptor even if it were to internally invalidate the
                                        * one we passed. */

                        if (!(flags & REMOVE_PHYSICAL)) {
                                struct statfs sfs;

                                if (fstatfs(fd, &sfs) < 0)
                                        return -errno;

                                if (is_physical_fs(&sfs)) {
                                        /* We refuse to clean physical file systems with this call, unless
                                         * explicitly requested. This is extra paranoia just to be sure we
                                         * never ever remove non-state data. */

                                        _cleanup_free_ char *path = NULL;

                                        (void) fd_get_path(fd, &path);
                                        return log_error_errno(SYNTHETIC_ERRNO(EPERM),
                                                               "Attempted to remove disk file system under \"%s\", and we can't allow that.",
                                                               strna(path));
                                }
                        }
                }

                FOREACH_DIRENT_ALL(de, d, return -errno) {
                        int is_dir;

                        if (dot_or_dot_dot(de->d_name))
                                continue;

                        is_dir = de->d_type == DT_UNKNOWN ? -1 : de->d_type == DT_DIR;

                        r = rm_rf_inner_child(fd, de->d_name, is_dir, flags, root_dev, false);
                        if (r == -EISDIR) {
                                /* Push the current working state onto the todo list */

                                 if (!GREEDY_REALLOC0(todos, n_todo + 2))
                                         return log_oom();

                                 _cleanup_free_ char *newdirname = strdup(de->d_name);
                                 if (!newdirname)
                                         return log_oom();

                                 mode_t mode;
                                 int newfd = openat_harder(fd, de->d_name,
                                                           O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC|O_NOFOLLOW|O_NOATIME,
                                                           flags, &mode);
                                 if (newfd >= 0) {
                                         todos[n_todo++] = (TodoEntry) {
                                                 .dir = TAKE_PTR(d),
                                                 .dirname = TAKE_PTR(dirname),
                                                 .old_mode = old_mode
                                         };

                                         fd = newfd;
                                         dirname = TAKE_PTR(newdirname);
                                         old_mode = mode;

                                         goto next_fd;

                                 } else if (newfd != -ENOENT && ret == 0)
                                         ret = newfd;

                        } else if (r < 0 && r != -ENOENT && ret == 0)
                                ret = r;
                }

                if (FLAGS_SET(flags, REMOVE_SYNCFS) && syncfs(fd) < 0 && ret >= 0)
                        ret = -errno;

                if (n_todo == 0) {
                        if (FLAGS_SET(flags, REMOVE_CHMOD_RESTORE) &&
                            fchmod(fd, old_mode & 07777) < 0 && ret >= 0)
                                ret = -errno;

                        break;
                }
        }

        return ret;
}

int rm_rf(const char *path, RemoveFlags flags) {
        mode_t old_mode;
        int fd, r, q = 0;

        assert(path);

        /* For now, don't support dropping subvols when also only dropping directories, since we can't do
         * this race-freely. */
        if (FLAGS_SET(flags, REMOVE_ONLY_DIRECTORIES|REMOVE_SUBVOLUME))
                return -EINVAL;

        /* We refuse to clean the root file system with this call. This is extra paranoia to never cause a
         * really seriously broken system. */
        if (path_equal_or_files_same(path, "/", AT_SYMLINK_NOFOLLOW))
                return log_error_errno(SYNTHETIC_ERRNO(EPERM),
                                       "Attempted to remove entire root file system (\"%s\"), and we can't allow that.",
                                       path);

        if (FLAGS_SET(flags, REMOVE_SUBVOLUME | REMOVE_ROOT | REMOVE_PHYSICAL)) {
                /* Try to remove as subvolume first */
                r = btrfs_subvol_remove(path, BTRFS_REMOVE_RECURSIVE|BTRFS_REMOVE_QUOTA);
                if (r >= 0)
                        return r;

                if (FLAGS_SET(flags, REMOVE_MISSING_OK) && r == -ENOENT)
                        return 0;

                if (!IN_SET(r, -ENOTTY, -EINVAL, -ENOTDIR))
                        return r;

                /* Not btrfs or not a subvolume */
        }

        fd = openat_harder(AT_FDCWD, path, O_RDONLY|O_NONBLOCK|O_DIRECTORY|O_CLOEXEC|O_NOFOLLOW|O_NOATIME, flags, &old_mode);
        if (fd >= 0) {
                /* We have a dir */
                r = rm_rf_children_impl(fd, flags, NULL, old_mode);

                if (FLAGS_SET(flags, REMOVE_ROOT))
                        q = RET_NERRNO(rmdir(path));
        } else {
                if (FLAGS_SET(flags, REMOVE_MISSING_OK) && errno == ENOENT)
                        return 0;

                if (!IN_SET(errno, ENOTDIR, ELOOP))
                        return -errno;

                if (FLAGS_SET(flags, REMOVE_ONLY_DIRECTORIES) || !FLAGS_SET(flags, REMOVE_ROOT))
                        return 0;

                if (!FLAGS_SET(flags, REMOVE_PHYSICAL)) {
                        struct statfs s;

                        if (statfs(path, &s) < 0)
                                return -errno;
                        if (is_physical_fs(&s))
                                return log_error_errno(SYNTHETIC_ERRNO(EPERM),
                                                       "Attempted to remove files from a disk file system under \"%s\", refusing.",
                                                       path);
                }

                r = 0;
                q = RET_NERRNO(unlink(path));
        }

        if (r < 0)
                return r;
        if (q < 0 && (q != -ENOENT || !FLAGS_SET(flags, REMOVE_MISSING_OK)))
                return q;
        return 0;
}

int rm_rf_child(int fd, const char *name, RemoveFlags flags) {

        /* Removes one specific child of the specified directory */

        if (fd < 0)
                return -EBADF;

        if (!filename_is_valid(name))
                return -EINVAL;

        if ((flags & (REMOVE_ROOT|REMOVE_MISSING_OK)) != 0) /* Doesn't really make sense here, we are not supposed to remove 'fd' anyway */
                return -EINVAL;

        if (FLAGS_SET(flags, REMOVE_ONLY_DIRECTORIES|REMOVE_SUBVOLUME))
                return -EINVAL;

        return rm_rf_inner_child(fd, name, -1, flags, NULL, true);
}