1 // SPDX-License-Identifier: GPL-2.0
12 #include <linux/compat.h>
13 #include <linux/fsnotify.h>
14 #include <linux/mount.h>
15 #include <linux/namei.h>
16 #include <linux/security.h>
17 #include <linux/writeback.h>
19 #define FS_IOC_GOINGDOWN _IOR('X', 125, __u32)
20 #define FSOP_GOING_FLAGS_DEFAULT 0x0 /* going down */
21 #define FSOP_GOING_FLAGS_LOGFLUSH 0x1 /* flush log but not data */
22 #define FSOP_GOING_FLAGS_NOLOGFLUSH 0x2 /* don't flush log nor data */
34 static int bch2_inode_flags_set(struct bch_inode_info *inode,
35 struct bch_inode_unpacked *bi,
38 struct bch_fs *c = inode->v.i_sb->s_fs_info;
40 * We're relying on btree locking here for exclusion with other ioctl
41 * calls - use the flags in the btree (@bi), not inode->i_flags:
43 struct flags_set *s = p;
44 unsigned newflags = s->flags;
45 unsigned oldflags = bi->bi_flags & s->mask;
47 if (((newflags ^ oldflags) & (BCH_INODE_APPEND|BCH_INODE_IMMUTABLE)) &&
48 !capable(CAP_LINUX_IMMUTABLE))
51 if (!S_ISREG(bi->bi_mode) &&
52 !S_ISDIR(bi->bi_mode) &&
53 (newflags & (BCH_INODE_NODUMP|BCH_INODE_NOATIME)) != newflags)
56 if (s->set_projinherit) {
57 bi->bi_fields_set &= ~(1 << Inode_opt_project);
58 bi->bi_fields_set |= ((int) s->projinherit << Inode_opt_project);
61 bi->bi_flags &= ~s->mask;
62 bi->bi_flags |= newflags;
64 bi->bi_ctime = timespec_to_bch2_time(c, current_time(&inode->v));
68 static int bch2_ioc_getflags(struct bch_inode_info *inode, int __user *arg)
70 unsigned flags = map_flags(bch_flags_to_uflags, inode->ei_inode.bi_flags);
72 return put_user(flags, arg);
75 static int bch2_ioc_setflags(struct bch_fs *c,
77 struct bch_inode_info *inode,
80 struct flags_set s = { .mask = map_defined(bch_flags_to_uflags) };
84 if (get_user(uflags, (int __user *) arg))
87 s.flags = map_flags_rev(bch_flags_to_uflags, uflags);
91 ret = mnt_want_write_file(file);
95 inode_lock(&inode->v);
96 if (!inode_owner_or_capable(file_mnt_idmap(file), &inode->v)) {
101 mutex_lock(&inode->ei_update_lock);
102 ret = bch2_write_inode(c, inode, bch2_inode_flags_set, &s,
104 mutex_unlock(&inode->ei_update_lock);
107 inode_unlock(&inode->v);
108 mnt_drop_write_file(file);
112 static int bch2_ioc_fsgetxattr(struct bch_inode_info *inode,
113 struct fsxattr __user *arg)
115 struct fsxattr fa = { 0 };
117 fa.fsx_xflags = map_flags(bch_flags_to_xflags, inode->ei_inode.bi_flags);
119 if (inode->ei_inode.bi_fields_set & (1 << Inode_opt_project))
120 fa.fsx_xflags |= FS_XFLAG_PROJINHERIT;
122 fa.fsx_projid = inode->ei_qid.q[QTYP_PRJ];
124 return copy_to_user(arg, &fa, sizeof(fa));
127 static int fssetxattr_inode_update_fn(struct bch_inode_info *inode,
128 struct bch_inode_unpacked *bi,
131 struct flags_set *s = p;
133 if (s->projid != bi->bi_project) {
134 bi->bi_fields_set |= 1U << Inode_opt_project;
135 bi->bi_project = s->projid;
138 return bch2_inode_flags_set(inode, bi, p);
141 static int bch2_ioc_fssetxattr(struct bch_fs *c,
143 struct bch_inode_info *inode,
144 struct fsxattr __user *arg)
146 struct flags_set s = { .mask = map_defined(bch_flags_to_xflags) };
150 if (copy_from_user(&fa, arg, sizeof(fa)))
153 s.set_projinherit = true;
154 s.projinherit = (fa.fsx_xflags & FS_XFLAG_PROJINHERIT) != 0;
155 fa.fsx_xflags &= ~FS_XFLAG_PROJINHERIT;
157 s.flags = map_flags_rev(bch_flags_to_xflags, fa.fsx_xflags);
161 if (fa.fsx_projid >= U32_MAX)
165 * inode fields accessible via the xattr interface are stored with a +1
166 * bias, so that 0 means unset:
168 s.projid = fa.fsx_projid + 1;
170 ret = mnt_want_write_file(file);
174 inode_lock(&inode->v);
175 if (!inode_owner_or_capable(file_mnt_idmap(file), &inode->v)) {
180 mutex_lock(&inode->ei_update_lock);
181 ret = bch2_set_projid(c, inode, fa.fsx_projid);
185 ret = bch2_write_inode(c, inode, fssetxattr_inode_update_fn, &s,
188 mutex_unlock(&inode->ei_update_lock);
190 inode_unlock(&inode->v);
191 mnt_drop_write_file(file);
195 static int bch2_reinherit_attrs_fn(struct bch_inode_info *inode,
196 struct bch_inode_unpacked *bi,
199 struct bch_inode_info *dir = p;
201 return !bch2_reinherit_attrs(bi, &dir->ei_inode);
204 static int bch2_ioc_reinherit_attrs(struct bch_fs *c,
206 struct bch_inode_info *src,
207 const char __user *name)
209 struct bch_hash_info hash = bch2_hash_info_init(c, &src->ei_inode);
210 struct bch_inode_info *dst;
211 struct inode *vinode = NULL;
217 kname = kmalloc(BCH_NAME_MAX + 1, GFP_KERNEL);
221 ret = strncpy_from_user(kname, name, BCH_NAME_MAX);
222 if (unlikely(ret < 0))
228 ret = bch2_dirent_lookup(c, inode_inum(src), &hash, &qstr, &inum);
232 vinode = bch2_vfs_inode_get(c, inum);
233 ret = PTR_ERR_OR_ZERO(vinode);
237 dst = to_bch_ei(vinode);
239 ret = mnt_want_write_file(file);
243 bch2_lock_inodes(INODE_UPDATE_LOCK, src, dst);
245 if (inode_attr_changing(src, dst, Inode_opt_project)) {
246 ret = bch2_fs_quota_transfer(c, dst,
249 KEY_TYPE_QUOTA_PREALLOC);
254 ret = bch2_write_inode(c, dst, bch2_reinherit_attrs_fn, src, 0);
256 bch2_unlock_inodes(INODE_UPDATE_LOCK, src, dst);
258 /* return true if we did work */
262 mnt_drop_write_file(file);
271 static int bch2_ioc_goingdown(struct bch_fs *c, u32 __user *arg)
276 if (!capable(CAP_SYS_ADMIN))
279 if (get_user(flags, arg))
282 bch_notice(c, "shutdown by ioctl type %u", flags);
284 down_write(&c->vfs_sb->s_umount);
287 case FSOP_GOING_FLAGS_DEFAULT:
288 ret = freeze_bdev(c->vfs_sb->s_bdev);
292 bch2_journal_flush(&c->journal);
293 c->vfs_sb->s_flags |= SB_RDONLY;
294 bch2_fs_emergency_read_only(c);
295 thaw_bdev(c->vfs_sb->s_bdev);
298 case FSOP_GOING_FLAGS_LOGFLUSH:
299 bch2_journal_flush(&c->journal);
302 case FSOP_GOING_FLAGS_NOLOGFLUSH:
303 c->vfs_sb->s_flags |= SB_RDONLY;
304 bch2_fs_emergency_read_only(c);
311 up_write(&c->vfs_sb->s_umount);
315 static long bch2_ioctl_subvolume_create(struct bch_fs *c, struct file *filp,
316 struct bch_ioctl_subvolume arg)
319 struct bch_inode_info *inode;
320 struct user_namespace *s_user_ns;
321 struct dentry *dst_dentry;
322 struct path src_path, dst_path;
323 int how = LOOKUP_FOLLOW;
325 subvol_inum snapshot_src = { 0 };
326 unsigned lookup_flags = 0;
327 unsigned create_flags = BCH_CREATE_SUBVOL;
329 if (arg.flags & ~(BCH_SUBVOL_SNAPSHOT_CREATE|
330 BCH_SUBVOL_SNAPSHOT_RO))
333 if (!(arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE) &&
335 (arg.flags & BCH_SUBVOL_SNAPSHOT_RO)))
338 if (arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE)
339 create_flags |= BCH_CREATE_SNAPSHOT;
341 if (arg.flags & BCH_SUBVOL_SNAPSHOT_RO)
342 create_flags |= BCH_CREATE_SNAPSHOT_RO;
344 /* why do we need this lock? */
345 down_read(&c->vfs_sb->s_umount);
347 if (arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE)
348 sync_inodes_sb(c->vfs_sb);
351 error = user_path_at(arg.dirfd,
352 (const char __user *)(unsigned long)arg.src_ptr,
357 if (src_path.dentry->d_sb->s_fs_info != c) {
363 snapshot_src = inode_inum(to_bch_ei(src_path.dentry->d_inode));
366 dst_dentry = user_path_create(arg.dirfd,
367 (const char __user *)(unsigned long)arg.dst_ptr,
368 &dst_path, lookup_flags);
369 error = PTR_ERR_OR_ZERO(dst_dentry);
373 if (dst_dentry->d_sb->s_fs_info != c) {
378 if (dst_dentry->d_inode) {
383 dir = dst_path.dentry->d_inode;
384 if (IS_DEADDIR(dir)) {
385 error = -BCH_ERR_ENOENT_directory_dead;
389 s_user_ns = dir->i_sb->s_user_ns;
390 if (!kuid_has_mapping(s_user_ns, current_fsuid()) ||
391 !kgid_has_mapping(s_user_ns, current_fsgid())) {
396 error = inode_permission(file_mnt_idmap(filp),
397 dir, MAY_WRITE | MAY_EXEC);
401 if (!IS_POSIXACL(dir))
402 arg.mode &= ~current_umask();
404 error = security_path_mkdir(&dst_path, dst_dentry, arg.mode);
408 if ((arg.flags & BCH_SUBVOL_SNAPSHOT_CREATE) &&
410 snapshot_src.subvol = to_bch_ei(dir)->ei_inode.bi_subvol;
412 inode = __bch2_create(file_mnt_idmap(filp), to_bch_ei(dir),
413 dst_dentry, arg.mode|S_IFDIR,
414 0, snapshot_src, create_flags);
415 error = PTR_ERR_OR_ZERO(inode);
419 d_instantiate(dst_dentry, &inode->v);
420 fsnotify_mkdir(dir, dst_dentry);
422 done_path_create(&dst_path, dst_dentry);
427 if (retry_estale(error, lookup_flags)) {
428 lookup_flags |= LOOKUP_REVAL;
432 up_read(&c->vfs_sb->s_umount);
437 static long bch2_ioctl_subvolume_destroy(struct bch_fs *c, struct file *filp,
438 struct bch_ioctl_subvolume arg)
447 ret = user_path_at(arg.dirfd,
448 (const char __user *)(unsigned long)arg.dst_ptr,
449 LOOKUP_FOLLOW, &path);
453 if (path.dentry->d_sb->s_fs_info != c) {
458 dir = path.dentry->d_parent->d_inode;
460 ret = __bch2_unlink(dir, path.dentry, true);
464 fsnotify_rmdir(dir, path.dentry);
465 d_delete(path.dentry);
471 long bch2_fs_file_ioctl(struct file *file, unsigned cmd, unsigned long arg)
473 struct bch_inode_info *inode = file_bch_inode(file);
474 struct bch_fs *c = inode->v.i_sb->s_fs_info;
478 case FS_IOC_GETFLAGS:
479 ret = bch2_ioc_getflags(inode, (int __user *) arg);
482 case FS_IOC_SETFLAGS:
483 ret = bch2_ioc_setflags(c, file, inode, (int __user *) arg);
486 case FS_IOC_FSGETXATTR:
487 ret = bch2_ioc_fsgetxattr(inode, (void __user *) arg);
490 case FS_IOC_FSSETXATTR:
491 ret = bch2_ioc_fssetxattr(c, file, inode,
492 (void __user *) arg);
495 case BCHFS_IOC_REINHERIT_ATTRS:
496 ret = bch2_ioc_reinherit_attrs(c, file, inode,
497 (void __user *) arg);
500 case FS_IOC_GETVERSION:
504 case FS_IOC_SETVERSION:
508 case FS_IOC_GOINGDOWN:
509 ret = bch2_ioc_goingdown(c, (u32 __user *) arg);
512 case BCH_IOCTL_SUBVOLUME_CREATE: {
513 struct bch_ioctl_subvolume i;
515 ret = copy_from_user(&i, (void __user *) arg, sizeof(i))
517 : bch2_ioctl_subvolume_create(c, file, i);
521 case BCH_IOCTL_SUBVOLUME_DESTROY: {
522 struct bch_ioctl_subvolume i;
524 ret = copy_from_user(&i, (void __user *) arg, sizeof(i))
526 : bch2_ioctl_subvolume_destroy(c, file, i);
531 ret = bch2_fs_ioctl(c, cmd, (void __user *) arg);
535 return bch2_err_class(ret);
539 long bch2_compat_fs_ioctl(struct file *file, unsigned cmd, unsigned long arg)
541 /* These are just misnamed, they actually get/put from/to user an int */
543 case FS_IOC_GETFLAGS:
544 cmd = FS_IOC_GETFLAGS;
546 case FS_IOC32_SETFLAGS:
547 cmd = FS_IOC_SETFLAGS;
552 return bch2_fs_file_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
556 #endif /* NO_BCACHEFS_FS */