Merge tag 'fuse-update-6.0' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse

Pull fuse updates from Miklos Szeredi:

- Fix an issue with reusing the bdi in case of block based filesystems

- Allow root (in init namespace) to access fuse filesystems in user
namespaces if expicitly enabled with a module param

- Misc fixes

* tag 'fuse-update-6.0' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
fuse: retire block-device-based superblock on force unmount
vfs: function to prevent re-use of block-device-based superblocks
virtio_fs: Modify format for virtio_fs_direct_access
virtiofs: delete unused parameter for virtio_fs_cleanup_vqs
fuse: Add module param for CAP_SYS_ADMIN access bypassing allow_other
fuse: Remove the control interface for virtio-fs
fuse: ioctl: translate ENOSYS
fuse: limit nsec
fuse: avoid unnecessary spinlock bump
fuse: fix deadlock between atomic O_TRUNC and page invalidation
fuse: write inode in fuse_release()

Documentation/filesystems/fuse.rst

···
 	the filesystem or not.
 
 	Note that the *ptrace* check is not strictly necessary to
-	prevent B/2/i, it is enough to check if mount owner has enough
+	prevent C/2/i, it is enough to check if mount owner has enough
 	privilege to send signal to the process accessing the
 	filesystem, since *SIGSTOP* can be used to get a similar effect.
 
···
 
 If a sysadmin trusts the users enough, or can ensure through other
 measures, that system processes will never enter non-privileged
-mounts, it can relax the last limitation with a 'user_allow_other'
-config option.  If this config option is set, the mounting user can
-add the 'allow_other' mount option which disables the check for other
-users' processes.
+mounts, it can relax the last limitation in several ways:
+
+  - With the 'user_allow_other' config option. If this config option is
+    set, the mounting user can add the 'allow_other' mount option which
+    disables the check for other users' processes.
+
+    User namespaces have an unintuitive interaction with 'allow_other':
+    an unprivileged user - normally restricted from mounting with
+    'allow_other' - could do so in a user namespace where they're
+    privileged. If any process could access such an 'allow_other' mount
+    this would give the mounting user the ability to manipulate
+    processes in user namespaces where they're unprivileged. For this
+    reason 'allow_other' restricts access to users in the same userns
+    or a descendant.
+
+  - With the 'allow_sys_admin_access' module option. If this option is
+    set, super user's processes have unrestricted access to mounts
+    irrespective of allow_other setting or user namespace of the
+    mounting user.
+
+Note that both of these relaxations expose the system to potential
+information leak or *DoS* as described in points B and C/2/i-ii in the
+preceding section.
 
 Kernel - userspace interface
 ============================

fs/fuse/control.c

···
 	struct dentry *parent;
 	char name[32];
 
-	if (!fuse_control_sb)
+	if (!fuse_control_sb || fc->no_control)
 		return 0;
 
 	parent = fuse_control_sb->s_root;
···
 {
 	int i;
 
-	if (!fuse_control_sb)
+	if (!fuse_control_sb || fc->no_control)
 		return;
 
 	for (i = fc->ctl_ndents - 1; i >= 0; i--) {

fs/fuse/dax.c

···
 		WARN_ON(fcd->nr_free_ranges <= 0);
 		fcd->nr_free_ranges--;
 	}
+	__kick_dmap_free_worker(fcd, 0);
 	spin_unlock(&fcd->lock);
 
-	kick_dmap_free_worker(fcd, 0);
 	return dmap;
 }
 

fs/fuse/dir.c

···
 #include <linux/pagemap.h>
 #include <linux/file.h>
 #include <linux/fs_context.h>
+#include <linux/moduleparam.h>
 #include <linux/sched.h>
 #include <linux/namei.h>
 #include <linux/slab.h>
···
 #include <linux/security.h>
 #include <linux/types.h>
 #include <linux/kernel.h>
+
+static bool __read_mostly allow_sys_admin_access;
+module_param(allow_sys_admin_access, bool, 0644);
+MODULE_PARM_DESC(allow_sys_admin_access,
+		 "Allow users with CAP_SYS_ADMIN in initial userns to bypass allow_other access check");
 
 static void fuse_advise_use_readdirplus(struct inode *dir)
 {
···
 	struct fuse_file *ff;
 	void *security_ctx = NULL;
 	u32 security_ctxlen;
+	bool trunc = flags & O_TRUNC;
 
 	/* Userspace expects S_IFREG in create mode */
 	BUG_ON((mode & S_IFMT) != S_IFREG);
···
 	inarg.mode = mode;
 	inarg.umask = current_umask();
 
-	if (fm->fc->handle_killpriv_v2 && (flags & O_TRUNC) &&
+	if (fm->fc->handle_killpriv_v2 && trunc &&
 	    !(flags & O_EXCL) && !capable(CAP_FSETID)) {
 		inarg.open_flags |= FUSE_OPEN_KILL_SUIDGID;
 	}
···
 	} else {
 		file->private_data = ff;
 		fuse_finish_open(inode, file);
+		if (fm->fc->atomic_o_trunc && trunc)
+			truncate_pagecache(inode, 0);
+		else if (!(ff->open_flags & FOPEN_KEEP_CACHE))
+			invalidate_inode_pages2(inode->i_mapping);
 	}
 	return err;
 
···
 int fuse_allow_current_process(struct fuse_conn *fc)
 {
 	const struct cred *cred;
+
+	if (allow_sys_admin_access && capable(CAP_SYS_ADMIN))
+		return 1;
 
 	if (fc->allow_other)
 		return current_in_userns(fc->user_ns);

fs/fuse/file.c

···
 		fi->attr_version = atomic64_inc_return(&fc->attr_version);
 		i_size_write(inode, 0);
 		spin_unlock(&fi->lock);
-		truncate_pagecache(inode, 0);
 		file_update_time(file);
 		fuse_invalidate_attr_mask(inode, FUSE_STATX_MODSIZE);
-	} else if (!(ff->open_flags & FOPEN_KEEP_CACHE)) {
-		invalidate_inode_pages2(inode->i_mapping);
 	}
-
 	if ((file->f_mode & FMODE_WRITE) && fc->writeback_cache)
 		fuse_link_write_file(file);
 }
···
 	if (err)
 		return err;
 
-	if (is_wb_truncate || dax_truncate) {
+	if (is_wb_truncate || dax_truncate)
 		inode_lock(inode);
-		fuse_set_nowrite(inode);
-	}
 
 	if (dax_truncate) {
 		filemap_invalidate_lock(inode->i_mapping);
 		err = fuse_dax_break_layouts(inode, 0, 0);
 		if (err)
-			goto out;
+			goto out_inode_unlock;
 	}
+
+	if (is_wb_truncate || dax_truncate)
+		fuse_set_nowrite(inode);
 
 	err = fuse_do_open(fm, get_node_id(inode), file, isdir);
 	if (!err)
 		fuse_finish_open(inode, file);
 
-out:
+	if (is_wb_truncate || dax_truncate)
+		fuse_release_nowrite(inode);
+	if (!err) {
+		struct fuse_file *ff = file->private_data;
+
+		if (fc->atomic_o_trunc && (file->f_flags & O_TRUNC))
+			truncate_pagecache(inode, 0);
+		else if (!(ff->open_flags & FOPEN_KEEP_CACHE))
+			invalidate_inode_pages2(inode->i_mapping);
+	}
 	if (dax_truncate)
 		filemap_invalidate_unlock(inode->i_mapping);
-
-	if (is_wb_truncate | dax_truncate) {
-		fuse_release_nowrite(inode);
+out_inode_unlock:
+	if (is_wb_truncate || dax_truncate)
 		inode_unlock(inode);
-	}
 
 	return err;
 }
···
 
 static int fuse_release(struct inode *inode, struct file *file)
 {
+	struct fuse_conn *fc = get_fuse_conn(inode);
+
+	/*
+	 * Dirty pages might remain despite write_inode_now() call from
+	 * fuse_flush() due to writes racing with the close.
+	 */
+	if (fc->writeback_cache)
+		write_inode_now(inode, 1);
+
 	fuse_release_common(file, false);
 
 	/* return value is ignored by VFS */

fs/fuse/inode.c

···
 	inode->i_uid     = make_kuid(fc->user_ns, attr->uid);
 	inode->i_gid     = make_kgid(fc->user_ns, attr->gid);
 	inode->i_blocks  = attr->blocks;
+
+	/* Sanitize nsecs */
+	attr->atimensec = min_t(u32, attr->atimensec, NSEC_PER_SEC - 1);
+	attr->mtimensec = min_t(u32, attr->mtimensec, NSEC_PER_SEC - 1);
+	attr->ctimensec = min_t(u32, attr->ctimensec, NSEC_PER_SEC - 1);
+
 	inode->i_atime.tv_sec   = attr->atime;
 	inode->i_atime.tv_nsec  = attr->atimensec;
 	/* mtime from server may be stale due to local buffered write */
···
 {
 	struct fuse_conn *fc = get_fuse_conn_super(sb);
 
-	if (!fc->no_force_umount)
-		fuse_abort_conn(fc);
+	if (fc->no_force_umount)
+		return;
+
+	fuse_abort_conn(fc);
+
+	// Only retire block-device-based superblocks.
+	if (sb->s_bdev != NULL)
+		retire_super(sb);
 }
 
 static void fuse_send_destroy(struct fuse_mount *fm)

fs/fuse/ioctl.c

···
 #include <linux/compat.h>
 #include <linux/fileattr.h>
 
+static ssize_t fuse_send_ioctl(struct fuse_mount *fm, struct fuse_args *args)
+{
+	ssize_t ret = fuse_simple_request(fm, args);
+
+	/* Translate ENOSYS, which shouldn't be returned from fs */
+	if (ret == -ENOSYS)
+		ret = -ENOTTY;
+
+	return ret;
+}
+
 /*
  * CUSE servers compiled on 32bit broke on 64bit kernels because the
  * ABI was defined to be 'struct iovec' which is different on 32bit
···
 	ap.args.out_pages = true;
 	ap.args.out_argvar = true;
 
-	transferred = fuse_simple_request(fm, &ap.args);
+	transferred = fuse_send_ioctl(fm, &ap.args);
 	err = transferred;
 	if (transferred < 0)
 		goto out;
···
 	args.out_args[1].size = inarg.out_size;
 	args.out_args[1].value = ptr;
 
-	err = fuse_simple_request(fm, &args);
+	err = fuse_send_ioctl(fm, &args);
 	if (!err) {
 		if (outarg.result < 0)
 			err = outarg.result;

fs/fuse/virtio_fs.c

···
 }
 
 /* Free virtqueues (device must already be reset) */
-static void virtio_fs_cleanup_vqs(struct virtio_device *vdev,
-				  struct virtio_fs *fs)
+static void virtio_fs_cleanup_vqs(struct virtio_device *vdev)
 {
 	vdev->config->del_vqs(vdev);
 }
···
 {
 	struct virtio_fs *fs = dax_get_private(dax_dev);
 	phys_addr_t offset = PFN_PHYS(pgoff);
-	size_t max_nr_pages = fs->window_len/PAGE_SIZE - pgoff;
+	size_t max_nr_pages = fs->window_len / PAGE_SIZE - pgoff;
 
 	if (kaddr)
 		*kaddr = fs->window_kaddr + offset;
···
 
 out_vqs:
 	virtio_reset_device(vdev);
-	virtio_fs_cleanup_vqs(vdev, fs);
+	virtio_fs_cleanup_vqs(vdev);
 	kfree(fs->vqs);
 
 out:
···
 	virtio_fs_stop_all_queues(fs);
 	virtio_fs_drain_all_queues_locked(fs);
 	virtio_reset_device(vdev);
-	virtio_fs_cleanup_vqs(vdev, fs);
+	virtio_fs_cleanup_vqs(vdev);
 
 	vdev->priv = NULL;
 	/* Put device reference on virtio_fs object */

fs/super.c

···
 }
 
 /**
+ *	retire_super	-	prevents superblock from being reused
+ *	@sb: superblock to retire
+ *
+ *	The function marks superblock to be ignored in superblock test, which
+ *	prevents it from being reused for any new mounts.  If the superblock has
+ *	a private bdi, it also unregisters it, but doesn't reduce the refcount
+ *	of the superblock to prevent potential races.  The refcount is reduced
+ *	by generic_shutdown_super().  The function can not be called
+ *	concurrently with generic_shutdown_super().  It is safe to call the
+ *	function multiple times, subsequent calls have no effect.
+ *
+ *	The marker will affect the re-use only for block-device-based
+ *	superblocks.  Other superblocks will still get marked if this function
+ *	is used, but that will not affect their reusability.
+ */
+void retire_super(struct super_block *sb)
+{
+	WARN_ON(!sb->s_bdev);
+	down_write(&sb->s_umount);
+	if (sb->s_iflags & SB_I_PERSB_BDI) {
+		bdi_unregister(sb->s_bdi);
+		sb->s_iflags &= ~SB_I_PERSB_BDI;
+	}
+	sb->s_iflags |= SB_I_RETIRED;
+	up_write(&sb->s_umount);
+}
+EXPORT_SYMBOL(retire_super);
+
+/**
  *	generic_shutdown_super	-	common helper for ->kill_sb()
  *	@sb: superblock to kill
  *
···
 
 static int test_bdev_super_fc(struct super_block *s, struct fs_context *fc)
 {
-	return s->s_bdev == fc->sget_key;
+	return !(s->s_iflags & SB_I_RETIRED) && s->s_bdev == fc->sget_key;
 }
 
 /**
···
 
 static int test_bdev_super(struct super_block *s, void *data)
 {
-	return (void *)s->s_bdev == data;
+	return !(s->s_iflags & SB_I_RETIRED) && (void *)s->s_bdev == data;
 }
 
 struct dentry *mount_bdev(struct file_system_type *fs_type,

include/linux/fs.h

···
 #define SB_I_SKIP_SYNC	0x00000100	/* Skip superblock at global sync */
 #define SB_I_PERSB_BDI	0x00000200	/* has a per-sb bdi */
 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
+#define SB_I_RETIRED	0x00000800	/* superblock shouldn't be reused */
 
 /* Possible states of 'frozen' field */
 enum {
···
 	int flags, void *data,
 	int (*fill_super)(struct super_block *, void *, int));
 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
+void retire_super(struct super_block *sb);
 void generic_shutdown_super(struct super_block *sb);
 void kill_block_super(struct super_block *sb);
 void kill_anon_super(struct super_block *sb);