Creating Linux virtual filesystems (LWN, 2003)

myfs.c

/*
 * Author: Harrison Turton <[email protected]>
 *
 * This follows the LWN article "Creating Linux virtual filesystems":
 *
 *   https://lwn.net/Articles/57369/
 *
 * However, this was published in 2003. The VFS has changed quite a
 * bit since then, and so the code had to be tweaked to work. So
 * isn't a byte-for-byte copy, I had to go digging through the kernel
 * to figure out the specifics.
 *
 * Maybe this will help someone else :)
 *
 * To use this, build it as a kernel module, load it, and mount a directory
 * as "myfs". There's nothing particularly special about this process. Just
 * know that you can't ls the mounted directory, because that's not
 * implemented. Instead, run cat /dir/counter and cat/subdir/subcounter a
 * few times, and see what happens.
 */
#include <linux/module.h>
#include <linux/printk.h>
#include <linux/fs.h> /* libfs */
#include <linux/kernel.h>
#include <linux/time.h>
#include <asm/atomic.h>

MODULE_LICENSE("MIT");
MODULE_AUTHOR("Harrison Turton");
MODULE_DESCRIPTION("A small demo filesystem for learning");

#define MYFS_MAGIC 0xdeadbeef
#define TMPSIZE 20
#define PERM_RX_RWX 0755
#define PERM_R_RWX 0644

static struct super_operations myfs_s_ops = {
	.statfs = simple_statfs,
	.drop_inode = generic_delete_inode,
};

static ssize_t myfs_read(struct file *filp, char *buf,
				size_t count, loff_t *offset)
{
	atomic_t *counter = (atomic_t *) filp->private_data;
	int v, len;
	char tmp[TMPSIZE];

	v = atomic_read(counter);
	if (*offset > 0) {
		v -= 1;
	} else {
		atomic_inc(counter);
	}

	len = snprintf(tmp, TMPSIZE, "%d\n", v);
	if (*offset > len)
		return 0;
	if (count > len - *offset)
		count = len - *offset;
	
	if (copy_to_user(buf, tmp + *offset, count))
		return -EFAULT;
	*offset += count;

	return count;
}

static ssize_t myfs_write(struct file *filp, const char *buf,
				size_t count, loff_t *offset)
{
	atomic_t *counter = (atomic_t *) filp->private_data;
	char tmp[TMPSIZE];

	if (*offset != 0)
		return -EINVAL;

	if (count >= TMPSIZE)
		return -EINVAL;

	memset(tmp, 0, TMPSIZE);
	if (copy_from_user(tmp, buf, count))
		return -EFAULT;

	atomic_set(counter, simple_strtol(tmp, NULL, 10));
	return count;
}

static int myfs_open(struct inode *inode, struct file *filp)
{
	filp->private_data = inode->i_private;
	return 0;
} 

static struct file_operations myfs_f_ops = {
	.open = myfs_open,
	.read = myfs_read,
	.write = myfs_write,
};

static struct inode *myfs_make_inode(struct super_block *sb, int mode)
{
	struct inode *inode = new_inode(sb);
	if (inode == NULL)
		return NULL;
	
	inode->i_mode = mode;
	i_uid_write(inode, 0);
	i_gid_write(inode, 0);
	inode->i_blocks = 0;
	inode->i_atime = current_time(inode);
	inode->i_mtime = current_time(inode);
	inode->i_ctime = current_time(inode);
	return inode;
}

static struct dentry *myfs_create_file(struct super_block *sb,
				struct dentry *dir, const char *name, atomic_t *counter)
{
	struct dentry *dentry;
	struct inode *inode;
	struct qstr qname;

	qname.name = name;
	qname.len = strlen(name);
	qname.hash = full_name_hash(dir, name, qname.len);

	dentry = d_alloc(dir, &qname);
	if (!dentry)
		goto out;
	
	inode = myfs_make_inode(sb, S_IFREG | PERM_RX_RWX);
	if (!inode)
		goto out_dput;

	inode->i_fop = &myfs_f_ops;
	inode->i_private = counter;

	d_add(dentry, inode);
	return dentry;

	out_dput:
		dput(dentry);
	out:
		return NULL;
}

static struct dentry *myfs_create_dir(struct super_block *sb,
				struct dentry *parent, const char *name)
{
	struct dentry *dentry;
	struct inode *inode;
	struct qstr qname;

	dentry = d_alloc(parent, &qname);
	if (!dentry)
		goto out;

	qname.name = name;
	qname.len = strlen(name);
	qname.hash = full_name_hash(parent, name, qname.len);

	inode = myfs_make_inode(sb, S_IFDIR | PERM_RX_RWX);
	if (!inode)
		goto out_dput;

	inode->i_op = &simple_dir_inode_operations;
	inode->i_fop = &simple_dir_operations;

	d_add(dentry, inode);
	return dentry;

	out_dput:
		dput(dentry);
	out:
		return NULL;
}

static atomic_t counter, subcounter;

static void myfs_create_files(struct super_block *sb, struct dentry *root)
{
	struct dentry *subdir;

	// One counter file in the top-level directory
	atomic_set(&counter, 0);
	myfs_create_file(sb, root, "counter", &counter);

	// One in the subdirectory
	atomic_set(&subcounter, 0);
	subdir = myfs_create_dir(sb, root, "subdir");
	if (subdir) {
		myfs_create_file(sb, subdir, "subcounter", &subcounter);
	} else {
		pr_info("Could not create subdir");
	}
}

static int myfs_fill_super(struct super_block *sb, void *data, int flags)
{
	struct inode *root;
	struct dentry *root_dentry;

	sb->s_blocksize = PAGE_SIZE;
	sb->s_blocksize_bits = PAGE_SHIFT;
	sb->s_magic = MYFS_MAGIC;
	sb->s_op = &myfs_s_ops;

	root = myfs_make_inode(sb, S_IFDIR | PERM_R_RWX);
	if (!root)
		goto out;
	root->i_op = &simple_dir_inode_operations;
	root->i_fop = &simple_dir_operations;

	root_dentry = d_make_root(root);
	if (!root_dentry)
		goto out_iput;
	sb->s_root = root_dentry;

	myfs_create_files(sb, root_dentry);
	return 0;

	out_iput:
		iput(root);
	out:
		return -ENOMEM;
}

static struct dentry *myfs_mount(struct file_system_type *fs_type, int flags,
				const char *dev_name, void *data)
{
	return mount_nodev(fs_type, flags, data, &myfs_fill_super);
}

static struct file_system_type myfs_type = {
	.name = "myfs",
	.owner = THIS_MODULE,
	.mount = myfs_mount,
	.kill_sb = kill_litter_super,
};

static int __init myfs_init(void)
{
	pr_info("myfs_init: Hello world!\n");
	pr_info("myfs_init: Moduled loaded\n");
	return register_filesystem(&myfs_type);
}

static void __exit myfs_exit(void)
{
	pr_info("myfs_init: Goodbye world :(\n");
	pr_info("myfs_exit: Module unloaded\n");
	unregister_filesystem(&myfs_type);
}

module_init(myfs_init);
module_exit(myfs_exit);