From d43dbf7322a356733b3e3a997cad51dce174d83c Mon Sep 17 00:00:00 2001 From: Christian Brauner Date: Wed, 9 Apr 2025 10:00:23 +0200 Subject: mount: ensure we don't pointlessly walk the mount tree This logic got broken recently. Add it back. Fixes: 474f7825d533 ("fs: add copy_mount_setattr() helper") Link: https://lore.kernel.org/20250409-sektflaschen-gecko-27c021fbd222@brauner Tested-by: Mikhail Gavrilov Signed-off-by: Christian Brauner --- fs/namespace.c | 30 ++++++++++++++++++------------ 1 file changed, 18 insertions(+), 12 deletions(-) (limited to 'fs/namespace.c') diff --git a/fs/namespace.c b/fs/namespace.c index 14935a0500a2..1eb76d6312de 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -5189,8 +5189,8 @@ static void finish_mount_kattr(struct mount_kattr *kattr) mnt_idmap_put(kattr->mnt_idmap); } -static int copy_mount_setattr(struct mount_attr __user *uattr, size_t usize, - struct mount_kattr *kattr) +static int wants_mount_setattr(struct mount_attr __user *uattr, size_t usize, + struct mount_kattr *kattr) { int ret; struct mount_attr attr; @@ -5213,9 +5213,13 @@ static int copy_mount_setattr(struct mount_attr __user *uattr, size_t usize, if (attr.attr_set == 0 && attr.attr_clr == 0 && attr.propagation == 0) - return 0; + return 0; /* Tell caller to not bother. */ + + ret = build_mount_kattr(&attr, usize, kattr); + if (ret < 0) + return ret; - return build_mount_kattr(&attr, usize, kattr); + return 1; } SYSCALL_DEFINE5(mount_setattr, int, dfd, const char __user *, path, @@ -5247,8 +5251,8 @@ SYSCALL_DEFINE5(mount_setattr, int, dfd, const char __user *, path, if (flags & AT_RECURSIVE) kattr.kflags |= MOUNT_KATTR_RECURSE; - err = copy_mount_setattr(uattr, usize, &kattr); - if (err) + err = wants_mount_setattr(uattr, usize, &kattr); + if (err <= 0) return err; err = user_path_at(dfd, path, kattr.lookup_flags, &target); @@ -5282,15 +5286,17 @@ SYSCALL_DEFINE5(open_tree_attr, int, dfd, const char __user *, filename, if (flags & AT_RECURSIVE) kattr.kflags |= MOUNT_KATTR_RECURSE; - ret = copy_mount_setattr(uattr, usize, &kattr); - if (ret) + ret = wants_mount_setattr(uattr, usize, &kattr); + if (ret < 0) return ret; - ret = do_mount_setattr(&file->f_path, &kattr); - if (ret) - return ret; + if (ret) { + ret = do_mount_setattr(&file->f_path, &kattr); + if (ret) + return ret; - finish_mount_kattr(&kattr); + finish_mount_kattr(&kattr); + } } fd = get_unused_fd_flags(flags & O_CLOEXEC); -- cgit v1.2.3 From 47a742fd977a7a8c39fea890712e9bfdf76f98f1 Mon Sep 17 00:00:00 2001 From: Jan Stancek Date: Thu, 10 Apr 2025 17:05:42 +0200 Subject: fs: use namespace_{lock,unlock} in dissolve_on_fput() In commit b73ec10a4587 ("fs: add fastpath for dissolve_on_fput()"), the namespace_{lock,unlock} has been replaced with scoped_guard using the namespace_sem. This however now also skips processing of 'unmounted' list in namespace_unlock(), and mount is not (immediately) cleaned up. For example, this causes LTP move_mount02 fail: ... move_mount02.c:80: TPASS: invalid-from-fd: move_mount() failed as expected: EBADF (9) move_mount02.c:80: TPASS: invalid-from-path: move_mount() failed as expected: ENOENT (2) move_mount02.c:80: TPASS: invalid-to-fd: move_mount() failed as expected: EBADF (9) move_mount02.c:80: TPASS: invalid-to-path: move_mount() failed as expected: ENOENT (2) move_mount02.c:80: TPASS: invalid-flags: move_mount() failed as expected: EINVAL (22) tst_test.c:1833: TINFO: === Testing on ext3 === tst_test.c:1170: TINFO: Formatting /dev/loop0 with ext3 opts='' extra opts='' mke2fs 1.47.2 (1-Jan-2025) /dev/loop0 is apparently in use by the system; will not make a filesystem here! tst_test.c:1170: TBROK: mkfs.ext3 failed with exit code 1 The test makes number of move_mount() calls but these are all designed to fail with specific errno. Even after test, 'losetup -d' can't detach loop device. Define a new guard for dissolve_on_fput, that will use namespace_{lock,unlock}. Fixes: b73ec10a4587 ("fs: add fastpath for dissolve_on_fput()") Signed-off-by: Jan Stancek Link: https://lore.kernel.org/cad2f042b886bf0ced3d8e3aff120ec5e0125d61.1744297468.git.jstancek@redhat.com Signed-off-by: Christian Brauner --- fs/namespace.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs/namespace.c') diff --git a/fs/namespace.c b/fs/namespace.c index 1eb76d6312de..d9ca80dcc544 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -1830,6 +1830,8 @@ static inline void namespace_lock(void) down_write(&namespace_sem); } +DEFINE_GUARD(namespace_lock, struct rw_semaphore *, namespace_lock(), namespace_unlock()) + enum umount_tree_flags { UMOUNT_SYNC = 1, UMOUNT_PROPAGATE = 2, @@ -2383,7 +2385,7 @@ void dissolve_on_fput(struct vfsmount *mnt) return; } - scoped_guard(rwsem_write, &namespace_sem) { + scoped_guard(namespace_lock, &namespace_sem) { ns = m->mnt_ns; if (!must_dissolve(ns)) return; -- cgit v1.2.3 From 0d039eac6e5950f9d1ecc9e410c2fd1feaeab3b6 Mon Sep 17 00:00:00 2001 From: Al Viro Date: Wed, 23 Apr 2025 02:30:34 +0100 Subject: fix a couple of races in MNT_TREE_BENEATH handling by do_move_mount() Normally do_lock_mount(path, _) is locking a mountpoint pinned by *path and at the time when matching unlock_mount() unlocks that location it is still pinned by the same thing. Unfortunately, for 'beneath' case it's no longer that simple - the object being locked is not the one *path points to. It's the mountpoint of path->mnt. The thing is, without sufficient locking ->mnt_parent may change under us and none of the locks are held at that point. The rules are * mount_lock stabilizes m->mnt_parent for any mount m. * namespace_sem stabilizes m->mnt_parent, provided that m is mounted. * if either of the above holds and refcount of m is positive, we are guaranteed the same for refcount of m->mnt_parent. namespace_sem nests inside inode_lock(), so do_lock_mount() has to take inode_lock() before grabbing namespace_sem. It does recheck that path->mnt is still mounted in the same place after getting namespace_sem, and it does take care to pin the dentry. It is needed, since otherwise we might end up with racing mount --move (or umount) happening while we were getting locks; in that case dentry would no longer be a mountpoint and could've been evicted on memory pressure along with its inode - not something you want when grabbing lock on that inode. However, pinning a dentry is not enough - the matching mount is also pinned only by the fact that path->mnt is mounted on top it and at that point we are not holding any locks whatsoever, so the same kind of races could end up with all references to that mount gone just as we are about to enter inode_lock(). If that happens, we are left with filesystem being shut down while we are holding a dentry reference on it; results are not pretty. What we need to do is grab both dentry and mount at the same time; that makes inode_lock() safe *and* avoids the problem with fs getting shut down under us. After taking namespace_sem we verify that path->mnt is still mounted (which stabilizes its ->mnt_parent) and check that it's still mounted at the same place. From that point on to the matching namespace_unlock() we are guaranteed that mount/dentry pair we'd grabbed are also pinned by being the mountpoint of path->mnt, so we can quietly drop both the dentry reference (as the current code does) and mnt one - it's OK to do under namespace_sem, since we are not dropping the final refs. That solves the problem on do_lock_mount() side; unlock_mount() also has one, since dentry is guaranteed to stay pinned only until the namespace_unlock(). That's easy to fix - just have inode_unlock() done earlier, while it's still pinned by mp->m_dentry. Fixes: 6ac392815628 "fs: allow to mount beneath top mount" # v6.5+ Signed-off-by: Al Viro Signed-off-by: Christian Brauner --- fs/namespace.c | 69 ++++++++++++++++++++++++++++++---------------------------- 1 file changed, 36 insertions(+), 33 deletions(-) (limited to 'fs/namespace.c') diff --git a/fs/namespace.c b/fs/namespace.c index d9ca80dcc544..98a5cd756e9a 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -2826,56 +2826,62 @@ static struct mountpoint *do_lock_mount(struct path *path, bool beneath) struct vfsmount *mnt = path->mnt; struct dentry *dentry; struct mountpoint *mp = ERR_PTR(-ENOENT); + struct path under = {}; for (;;) { - struct mount *m; + struct mount *m = real_mount(mnt); if (beneath) { - m = real_mount(mnt); + path_put(&under); read_seqlock_excl(&mount_lock); - dentry = dget(m->mnt_mountpoint); + under.mnt = mntget(&m->mnt_parent->mnt); + under.dentry = dget(m->mnt_mountpoint); read_sequnlock_excl(&mount_lock); + dentry = under.dentry; } else { dentry = path->dentry; } inode_lock(dentry->d_inode); - if (unlikely(cant_mount(dentry))) { - inode_unlock(dentry->d_inode); - goto out; - } - namespace_lock(); - if (beneath && (!is_mounted(mnt) || m->mnt_mountpoint != dentry)) { + if (unlikely(cant_mount(dentry) || !is_mounted(mnt))) + break; // not to be mounted on + + if (beneath && unlikely(m->mnt_mountpoint != dentry || + &m->mnt_parent->mnt != under.mnt)) { namespace_unlock(); inode_unlock(dentry->d_inode); - goto out; + continue; // got moved } mnt = lookup_mnt(path); - if (likely(!mnt)) + if (unlikely(mnt)) { + namespace_unlock(); + inode_unlock(dentry->d_inode); + path_put(path); + path->mnt = mnt; + path->dentry = dget(mnt->mnt_root); + continue; // got overmounted + } + mp = get_mountpoint(dentry); + if (IS_ERR(mp)) break; - - namespace_unlock(); - inode_unlock(dentry->d_inode); - if (beneath) - dput(dentry); - path_put(path); - path->mnt = mnt; - path->dentry = dget(mnt->mnt_root); - } - - mp = get_mountpoint(dentry); - if (IS_ERR(mp)) { - namespace_unlock(); - inode_unlock(dentry->d_inode); + if (beneath) { + /* + * @under duplicates the references that will stay + * at least until namespace_unlock(), so the path_put() + * below is safe (and OK to do under namespace_lock - + * we are not dropping the final references here). + */ + path_put(&under); + } + return mp; } - -out: + namespace_unlock(); + inode_unlock(dentry->d_inode); if (beneath) - dput(dentry); - + path_put(&under); return mp; } @@ -2886,14 +2892,11 @@ static inline struct mountpoint *lock_mount(struct path *path) static void unlock_mount(struct mountpoint *where) { - struct dentry *dentry = where->m_dentry; - + inode_unlock(where->m_dentry->d_inode); read_seqlock_excl(&mount_lock); put_mountpoint(where); read_sequnlock_excl(&mount_lock); - namespace_unlock(); - inode_unlock(dentry->d_inode); } static int graft_tree(struct mount *mnt, struct mount *p, struct mountpoint *mp) -- cgit v1.2.3