fscrypt: don't evict dirty inodes after removing key

After FS_IOC_REMOVE_ENCRYPTION_KEY removes a key, it syncs the
filesystem and tries to get and put all inodes that were unlocked by the
key so that unused inodes get evicted via fscrypt_drop_inode().
Normally, the inodes are all clean due to the sync.

However, after the filesystem is sync'ed, userspace can modify and close
one of the files.  (Userspace is *supposed* to close the files before
removing the key.  But it doesn't always happen, and the kernel can't
assume it.)  This causes the inode to be dirtied and have i_count == 0.
Then, fscrypt_drop_inode() failed to consider this case and indicated
that the inode can be dropped, causing the write to be lost.

On f2fs, other problems such as a filesystem freeze could occur due to
the inode being freed while still on f2fs's dirty inode list.

Fix this bug by making fscrypt_drop_inode() only drop clean inodes.

I've written an xfstest which detects this bug on ext4, f2fs, and ubifs.

Fixes: b1c0ec3599 ("fscrypt: add FS_IOC_REMOVE_ENCRYPTION_KEY ioctl")
Cc: <stable@vger.kernel.org> # v5.4+
Link: https://lore.kernel.org/r/20200305084138.653498-1-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>

fs/crypto/keysetup.c

@@ -538,6 +538,15 @@ int fscrypt_drop_inode(struct inode *inode)
 		return 0;
 	mk = ci->ci_master_key->payload.data[0];
 
+	/*
+	 * With proper, non-racy use of FS_IOC_REMOVE_ENCRYPTION_KEY, all inodes
+	 * protected by the key were cleaned by sync_filesystem().  But if
+	 * userspace is still using the files, inodes can be dirtied between
+	 * then and now.  We mustn't lose any writes, so skip dirty inodes here.
+	 */
+	if (inode->i_state & I_DIRTY_ALL)
+		return 0;
+
 	/*
 	 * Note: since we aren't holding ->mk_secret_sem, the result here can
 	 * immediately become outdated.  But there's no correctness problem with