/*
 * nfs.c - barebox NFS driver
 *
 * Copyright (c) 2014 Uwe Kleine-König, Pengutronix
 * Copyright (c) 2012 Sascha Hauer <s.hauer@pengutronix.de>, Pengutronix
 * Copyright (c) Masami Komiya <mkomiya@sonare.it> 2004
 *
 * Based on U-Boot NFS code which is based on NetBSD code with
 * major changes to support nfs3.
 *
 * See file CREDITS for list of people who contributed to this
 * project.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 */
#define pr_fmt(fmt) "NFS: " fmt

#include <common.h>
#include <net.h>
#include <driver.h>
#include <fs.h>
#include <errno.h>
#include <libgen.h>
#include <fcntl.h>
#include <fs.h>
#include <init.h>
#include <linux/stat.h>
#include <linux/err.h>
#include <kfifo.h>
#include <linux/sizes.h>
#include <byteorder.h>
#include <globalvar.h>
#include <parseopt.h>

#define SUNRPC_PORT     111

#define PROG_PORTMAP    100000
#define PROG_NFS        100003
#define PROG_MOUNT      100005

#define MSG_CALL        0
#define MSG_REPLY       1

#define PORTMAP_GETPORT 3

#define MOUNT_ADDENTRY	1
#define MOUNT_UMOUNT	3

#define NFSPROC3_GETATTR	1
#define NFSPROC3_LOOKUP		3
#define NFSPROC3_READLINK	5
#define NFSPROC3_READ		6
#define NFSPROC3_READDIR	16

#define NFS3_FHSIZE      64
#define NFS3_COOKIEVERFSIZE	8

/* values of enum ftype3 */
#define NF3REG	1
#define NF3DIR	2
#define NF3BLK	3
#define NF3CHR	4
#define NF3LNK	5
#define NF3SOCK	6
#define NF3FIFO	7

/* values for enum nfsstat3 */
#define NFS3_OK			    0
#define NFS3ERR_PERM		    1
#define NFS3ERR_NOENT		    2
#define NFS3ERR_IO   		    5
#define NFS3ERR_NXIO 		    6
#define NFS3ERR_ACCES		   13
#define NFS3ERR_EXIST		   17
#define NFS3ERR_XDEV		   18
#define NFS3ERR_NODEV		   19
#define NFS3ERR_NOTDIR		   20
#define NFS3ERR_ISDIR		   21
#define NFS3ERR_INVAL		   22
#define NFS3ERR_FBIG		   27
#define NFS3ERR_NOSPC		   28
#define NFS3ERR_ROFS		   30
#define NFS3ERR_MLINK		   31
#define NFS3ERR_NAMETOOLONG	   63
#define NFS3ERR_NOTEMPTY	   66
#define NFS3ERR_DQUOT		   69
#define NFS3ERR_STALE		   70
#define NFS3ERR_REMOTE		   71
#define NFS3ERR_BADHANDLE 	10001
#define NFS3ERR_NOT_SYNC	10002
#define NFS3ERR_BAD_COOKIE	10003
#define NFS3ERR_NOTSUPP		10004
#define NFS3ERR_TOOSMALL	10005
#define NFS3ERR_SERVERFAULT	10006
#define NFS3ERR_BADTYPE		10007
#define NFS3ERR_JUKEBOX		10008

struct rpc_call {
	uint32_t id;
	uint32_t type;
	uint32_t rpcvers;
	uint32_t prog;
	uint32_t vers;
	uint32_t proc;
	uint32_t data[0];
};

struct rpc_reply {
	uint32_t id;
	uint32_t type;
	uint32_t rstatus;
	uint32_t verifier;
	uint32_t v2;
	uint32_t astatus;
	uint32_t data[0];
};

#define NFS_TIMEOUT	(2 * SECOND)
#define NFS_MAX_RESEND	5

struct nfs_fh {
	unsigned short size;
	unsigned char data[NFS3_FHSIZE];
};

struct packet {
	int len;
	char data[];
};

struct nfs_priv {
	struct net_connection *con;
	IPaddr_t server;
	char *path;
	uint16_t mount_port;
	unsigned manual_mount_port:1;
	uint16_t nfs_port;
	unsigned manual_nfs_port:1;
	uint32_t rpc_id;
	struct nfs_fh rootfh;
	struct packet *nfs_packet;
};

struct file_priv {
	struct kfifo *fifo;
	void *buf;
	struct nfs_priv *npriv;
	struct nfs_fh fh;
};

struct nfs_inode {
	struct inode inode;
	struct nfs_fh fh;
	struct nfs_priv *npriv;
};

static inline struct nfs_inode *nfsi(struct inode *inode)
{
	return container_of(inode, struct nfs_inode, inode);
}

static void nfs_set_fh(struct inode *inode, struct nfs_fh *fh)
{
	struct nfs_inode *ninode = nfsi(inode);

	ninode->fh = *fh;
}

static uint64_t nfs_timer_start;

static int nfs_state;
#define STATE_DONE			1
#define STATE_START			2

/*
 * common types used in more than one request:
 *
 * typedef uint32 count3;
 * typedef uint32 gid3;
 * typedef uint32 mode3;
 * typedef uint32 uid3;
 *
 * typedef uint64 cookie3;
 * typedef uint64 fileid3;
 * typedef uint64 size3;
 *
 * typedef string filename3<>;
 * typedef string nfspath3<>;
 *
 * typedef opaque cookieverf3[NFS3_COOKIEVERFSIZE];
 *
 * enum ftype3 {
 * 	NF3REG = 1,
 * 	NF3DIR = 2,
 * 	NF3BLK = 3,
 * 	NF3CHR = 4,
 * 	NF3LNK = 5,
 * 	NF3SOCK = 6,
 * 	NF3FIFO = 7
 * };
 *
 * struct specdata3 {
 * 	uint32 specdata1;
 * 	uint32 specdata2;
 * };
 *
 * struct nfs_fh3 {
 * 	opaque data<NFS3_FHSIZE>;
 * }
 *
 * struct nfstime3 {
 * 	uint32 seconds;
 * 	uint32 nseconds;
 * };
 *
 * struct fattr3 {
 * 	ftype3 type;
 * 	mode3 mode;
 * 	uint32_t nlink;
 * 	uid3 uid;
 * 	gid3 gid;
 * 	size3 size;
 * 	size3 used;
 * 	specdata3 rdev;
 * 	uint64_t fsid;
 * 	fileid3 fileid;
 * 	nfstime3 atime;
 * 	nfstime3 mtime;
 * 	nfstime3 ctime;
 * };
 *
 * struct diropargs3 {
 * 	nfs_fh3 dir;
 * 	filename3 name;
 * }
 *
 * union post_op_attr switch (bool attributes_follow) {
 * case TRUE:
 * 	fattr3 attributes;
 * case FALSE:
 * 	void;
 * };
 */

struct xdr_stream {
	__be32 *p;
	void *buf;
	__be32 *end;
};

#define xdr_zero 0
#define XDR_QUADLEN(l)		(((l) + 3) >> 2)

struct nfs_dir {
	/*
	 * stream points to the next entry3 in the reply member of READDIR3res
	 * (if any, to the end indicator otherwise).
	 */
	struct xdr_stream stream;
	uint64_t cookie;
	char cookieverf[NFS3_COOKIEVERFSIZE];
	struct nfs_fh fh;
};

static void xdr_init(struct xdr_stream *stream, void *buf, int len)
{
	stream->p = stream->buf = buf;
	stream->end = stream->buf + len;
}

static __be32 *__xdr_inline_decode(struct xdr_stream *xdr, size_t nbytes)
{
	__be32 *p = xdr->p;
	__be32 *q = p + XDR_QUADLEN(nbytes);

        if (q > xdr->end || q < p)
		return NULL;
	xdr->p = q;
	return p;
}

static __be32 *xdr_inline_decode(struct xdr_stream *xdr, size_t nbytes)
{
	__be32 *p;

	if (nbytes == 0)
		return xdr->p;
	if (xdr->p == xdr->end)
		return NULL;
	p = __xdr_inline_decode(xdr, nbytes);

	return p;
}

/*
 * name is expected to point to a buffer with a size of at least 256 bytes.
 */
static int decode_filename(struct xdr_stream *xdr, char *name, u32 *length)
{
	__be32 *p;
	u32 count;

	p = xdr_inline_decode(xdr, 4);
	if (!p)
		goto out_overflow;
	count = ntoh32(net_read_uint32(p));
	if (count > 255)
		goto out_nametoolong;
	p = xdr_inline_decode(xdr, count);
	if (!p)
		goto out_overflow;
	memcpy(name, p, count);
	name[count] = 0;
	*length = count;
	return 0;

out_nametoolong:
	pr_err("%s: returned a too long filename: %u\n", __func__, count);
	return -ENAMETOOLONG;

out_overflow:
	pr_err("%s: premature end of packet\n", __func__);
	return -EIO;
}

/*
 * rpc_add_credentials - Add RPC authentication/verifier entries
 */
static uint32_t *rpc_add_credentials(uint32_t *p)
{
	/*
	 * *BSD refuses AUTH_NONE, so use AUTH_UNIX. An empty hostname is OK for
	 * both Linux and *BSD.
	 */

	/* Provide an AUTH_UNIX credential.  */
	*p++ = hton32(1);		/* AUTH_UNIX */
	*p++ = hton32(20);		/* auth length: 20 + strlen(hostname) */
	*p++ = hton32(0);		/* stamp */
	*p++ = hton32(0);		/* hostname string length */
	/* memcpy(p, "", 0); p += 0; <- empty host name */

	*p++ = 0;			/* uid */
	*p++ = 0;			/* gid */
	*p++ = 0;			/* auxiliary gid list */

	/* Provide an AUTH_NONE verifier.  */
	*p++ = 0;			/* AUTH_NONE */
	*p++ = 0;			/* auth length */

	return p;
}

static int rpc_check_reply(struct packet *pkt, int rpc_prog,
			   uint32_t rpc_id, int *nfserr)
{
	uint32_t *data;
	struct rpc_reply rpc;

	*nfserr = 0;

	if (!pkt)
		return -EAGAIN;

	memcpy(&rpc, pkt->data, sizeof(rpc));

	if (ntoh32(rpc.id) != rpc_id) {
		if (rpc_id - ntoh32(rpc.id) == 1)
			/* stale packet, wait a bit longer */
			return 0;

		return -EINVAL;
	}

	if (rpc.rstatus  ||
	    rpc.verifier ||
	    rpc.astatus ) {
		return -EINVAL;
	}

	if (rpc_prog != PROG_NFS)
		return 0;

	data = (uint32_t *)(pkt->data + sizeof(struct rpc_reply));
	*nfserr = ntoh32(net_read_uint32(data));
	*nfserr = -*nfserr;

	debug("%s: state: %d, err %d\n", __func__, nfs_state, *nfserr);

	return 0;
}

/*
 * rpc_req - synchronous RPC request
 */
static struct packet *rpc_req(struct nfs_priv *npriv, int rpc_prog,
			      int rpc_proc, uint32_t *data, int datalen)
{
	struct rpc_call pkt;
	unsigned short dport;
	int ret;
	unsigned char *payload = net_udp_get_payload(npriv->con);
	int nfserr;
	int tries = 0;

	npriv->rpc_id++;

	pkt.id = hton32(npriv->rpc_id);
	pkt.type = hton32(MSG_CALL);
	pkt.rpcvers = hton32(2);	/* use RPC version 2 */
	pkt.prog = hton32(rpc_prog);
	pkt.proc = hton32(rpc_proc);

	debug("%s: prog: %d, proc: %d\n", __func__, rpc_prog, rpc_proc);

	if (rpc_prog == PROG_PORTMAP) {
		dport = SUNRPC_PORT;
		pkt.vers = hton32(2);
	} else if (rpc_prog == PROG_MOUNT) {
		dport = npriv->mount_port;
		pkt.vers = hton32(3);
	} else {
		dport = npriv->nfs_port;
		pkt.vers = hton32(3);
	}

	memcpy(payload, &pkt, sizeof(pkt));
	memcpy(payload + sizeof(pkt), data, datalen * sizeof(uint32_t));

	npriv->con->udp->uh_dport = hton16(dport);

again:
	ret = net_udp_send(npriv->con,
			sizeof(pkt) + datalen * sizeof(uint32_t));

	nfs_timer_start = get_time_ns();

	nfs_state = STATE_START;

	while (nfs_state != STATE_DONE) {
		if (ctrlc())
			return ERR_PTR(-EINTR);

		net_poll();

		if (is_timeout(nfs_timer_start, NFS_TIMEOUT)) {
			tries++;
			if (tries == NFS_MAX_RESEND)
				return ERR_PTR(-ETIMEDOUT);
			goto again;
		}

		ret = rpc_check_reply(npriv->nfs_packet, rpc_prog,
				      npriv->rpc_id, &nfserr);
		if (!ret) {
			if (rpc_prog == PROG_NFS && nfserr) {
				free(npriv->nfs_packet);
				return ERR_PTR(nfserr);
			} else {
				return npriv->nfs_packet;
			}
		}
	}

	return npriv->nfs_packet;
}

/*
 * rpc_lookup_req - Lookup RPC Port numbers
 */
static int rpc_lookup_req(struct nfs_priv *npriv, uint32_t prog, uint32_t ver)
{
	uint32_t data[16];
	struct packet *nfs_packet;
	uint32_t port;

	data[0] = 0; data[1] = 0;	/* auth credential */
	data[2] = 0; data[3] = 0;	/* auth verifier */
	data[4] = hton32(prog);
	data[5] = hton32(ver);
	data[6] = hton32(17);	/* IP_UDP */
	data[7] = 0;

	nfs_packet = rpc_req(npriv, PROG_PORTMAP, PORTMAP_GETPORT, data, 8);
	if (IS_ERR(nfs_packet))
		return PTR_ERR(nfs_packet);

	port = ntoh32(net_read_uint32(nfs_packet->data + sizeof(struct rpc_reply)));

	free(nfs_packet);

	return port;
}

static uint32_t *nfs_add_uint32(uint32_t *p, uint32_t val)
{
	*p++ = hton32(val);
	return p;
}

static uint32_t *nfs_add_uint64(uint32_t *p, uint64_t val)
{
	uint64_t nval = hton64(val);

	memcpy(p, &nval, 8);
	return p + 2;
}

static uint32_t *nfs_add_fh3(uint32_t *p, struct nfs_fh *fh)
{
	*p++ = hton32(fh->size);

	/* zero padding */
	if (fh->size & 3)
		p[fh->size / 4] = 0;

	memcpy(p, fh->data, fh->size);
	p += DIV_ROUND_UP(fh->size, 4);
	return p;
}

static uint32_t *nfs_add_filename(uint32_t *p,
		uint32_t filename_len, const char *filename)
{
	*p++ = hton32(filename_len);

	/* zero padding */
	if (filename_len & 3)
		p[filename_len / 4] = 0;

	memcpy(p, filename, filename_len);
	p += DIV_ROUND_UP(filename_len, 4);
	return p;
}

/* This is a 1:1 mapping for Linux, the compiler optimizes it out */
static const struct {
	uint32_t nfsmode;
	unsigned short statmode;
} nfs3_mode_bits[] = {
	{ 0x00001, S_IXOTH },
	{ 0x00002, S_IWOTH },
	{ 0x00004, S_IROTH },
	{ 0x00008, S_IXGRP },
	{ 0x00010, S_IWGRP },
	{ 0x00020, S_IRGRP },
	{ 0x00040, S_IXUSR },
	{ 0x00080, S_IWUSR },
	{ 0x00100, S_IRUSR },
	{ 0x00200, S_ISVTX },
	{ 0x00400, S_ISGID },
	{ 0x00800, S_ISUID },
};

static int nfs_fattr3_to_stat(uint32_t *p, struct inode *inode)
{
	uint32_t mode;
	size_t i;

	if (!inode)
		return 0;

	/* offsetof(struct fattr3, type) = 0 */
	switch (ntoh32(net_read_uint32(p + 0))) {
	case NF3REG:
		inode->i_mode = S_IFREG;
		break;
	case NF3DIR:
		inode->i_mode = S_IFDIR;
		break;
	case NF3BLK:
		inode->i_mode = S_IFBLK;
		break;
	case NF3CHR:
		inode->i_mode = S_IFCHR;
		break;
	case NF3LNK:
		inode->i_mode = S_IFLNK;
		break;
	case NF3SOCK:
		inode->i_mode = S_IFSOCK;
		break;
	case NF3FIFO:
		inode->i_mode = S_IFIFO;
		break;
	default:
		printf("%s: invalid mode %x\n",
				__func__, ntoh32(net_read_uint32(p + 0)));
		return -EIO;
	}

	/* offsetof(struct fattr3, mode) = 4 */
	mode = ntoh32(net_read_uint32(p + 1));
	for (i = 0; i < ARRAY_SIZE(nfs3_mode_bits); ++i) {
		if (mode & nfs3_mode_bits[i].nfsmode)
			inode->i_mode |= nfs3_mode_bits[i].statmode;
	}

	/* offsetof(struct fattr3, size) = 20 */
	inode->i_size = ntoh64(net_read_uint64(p + 5));

	return 0;
}

static uint32_t *nfs_read_post_op_attr(uint32_t *p, struct inode *inode)
{
	/*
	 * union post_op_attr switch (bool attributes_follow) {
	 * case TRUE:
	 * 	fattr3 attributes;
	 * case FALSE:
	 * 	void;
	 * };
	 */

	if (ntoh32(net_read_uint32(p++))) {
		nfs_fattr3_to_stat(p, inode);
		p += 21;
	}

	return p;
}

/*
 * nfs_mount_req - Mount an NFS Filesystem
 */
static int nfs_mount_req(struct nfs_priv *npriv)
{
	uint32_t data[1024];
	uint32_t *p;
	int len;
	int pathlen;
	struct packet *nfs_packet;

	pathlen = strlen(npriv->path);

	debug("%s: %s\n", __func__, npriv->path);

	p = &(data[0]);
	p = rpc_add_credentials(p);

	*p++ = hton32(pathlen);
	if (pathlen & 3)
		*(p + pathlen / 4) = 0;

	memcpy (p, npriv->path, pathlen);
	p += (pathlen + 3) / 4;

	len = p - &(data[0]);

	nfs_packet = rpc_req(npriv, PROG_MOUNT, MOUNT_ADDENTRY, data, len);
	if (IS_ERR(nfs_packet))
		return PTR_ERR(nfs_packet);

	p = (void *)nfs_packet->data + sizeof(struct rpc_reply) + 4;

	npriv->rootfh.size = ntoh32(net_read_uint32(p++));
	if (npriv->rootfh.size > NFS3_FHSIZE) {
		printf("%s: file handle too big: %lu\n",
		       __func__, (unsigned long)npriv->rootfh.size);
		free(nfs_packet);
		return -EIO;
	}
	memcpy(npriv->rootfh.data, p, npriv->rootfh.size);

	free(nfs_packet);

	return 0;
}

/*
 * nfs_umountall_req - Unmount all our NFS Filesystems on the Server
 */
static void nfs_umount_req(struct nfs_priv *npriv)
{
	uint32_t data[1024];
	uint32_t *p;
	int len;
	int pathlen;
	struct packet *nfs_packet;

	pathlen = strlen(npriv->path);

	p = &(data[0]);
	p = rpc_add_credentials(p);

	p = nfs_add_filename(p, pathlen, npriv->path);

	len = p - &(data[0]);

	nfs_packet = rpc_req(npriv, PROG_MOUNT, MOUNT_UMOUNT, data, len);

	if (!IS_ERR(nfs_packet))
		free(nfs_packet);
}

/*
 * nfs_lookup_req - Lookup Pathname
 *
 * *s is set to NULL if LOOKUP3resok doesn't contain obj_attributes.
 */
static int nfs_lookup_req(struct nfs_priv *npriv, struct nfs_fh *fh,
			  const char *filename, struct inode *inode)
{
	struct nfs_inode *ninode = nfsi(inode);
	uint32_t data[1024];
	uint32_t *p;
	int len;
	struct packet *nfs_packet;

	/*
	 * struct LOOKUP3args {
	 * 	diropargs3 what;
	 * };
	 *
	 * struct LOOKUP3resok {
	 * 	nfs_fh3 object;
	 * 	post_op_attr obj_attributes;
	 * 	post_op_attr dir_attributes;
	 * };
	 *
	 * struct LOOKUP3resfail {
	 * 	post_op_attr dir_attributes;
	 * };
	 *
	 * union LOOKUP3res switch (nfsstat3 status) {
	 * case NFS3_OK:
	 * 	LOOKUP3resok resok;
	 * default:
	 * 	LOOKUP3resfail resfail;
	 * };
	 */

	p = &(data[0]);
	p = rpc_add_credentials(p);

	/* what.dir */
	p = nfs_add_fh3(p, fh);

	/* what.name */
	p = nfs_add_filename(p, strlen(filename), filename);

	len = p - &(data[0]);

	nfs_packet = rpc_req(npriv, PROG_NFS, NFSPROC3_LOOKUP, data, len);
	if (IS_ERR(nfs_packet))
		return PTR_ERR(nfs_packet);

	p = (void *)nfs_packet->data + sizeof(struct rpc_reply) + 4;

	ninode->fh.size = ntoh32(net_read_uint32(p++));
	if (ninode->fh.size > NFS3_FHSIZE) {
		debug("%s: file handle too big: %u\n", __func__,
		      ninode->fh.size);
		return -EIO;
	}
	memcpy(ninode->fh.data, p, ninode->fh.size);
	p += DIV_ROUND_UP(ninode->fh.size, 4);

	nfs_read_post_op_attr(p, inode);

	free(nfs_packet);

	return 0;
}

/*
 * returns with dir->stream pointing to the first entry
 * of dirlist3 res.resok.reply
 */
static void *nfs_readdirattr_req(struct nfs_priv *npriv, struct nfs_dir *dir)
{
	uint32_t data[1024];
	uint32_t *p;
	int len;
	struct packet *nfs_packet;
	void *buf;

	/*
	 * struct READDIR3args {
	 * 	nfs_fh3 dir;
	 * 	cookie3 cookie;
	 * 	cookieverf3 cookieverf;
	 * 	count3 count;
	 * };
	 *
	 * struct entry3 {
	 * 	fileid3 fileid;
	 * 	filename3 name;
	 * 	cookie3 cookie;
	 * 	entry3 *nextentry;
	 * };
	 *
	 * struct dirlist3 {
	 * 	entry3 *entries;
	 * 	bool eof;
	 * };
	 *
	 * struct READDIR3resok {
	 * 	post_op_attr dir_attributes;
	 * 	cookieverf3 cookieverf;
	 * 	dirlist3 reply;
	 * };
	 *
	 * struct READDIR3resfail {
	 * 	post_op_attr dir_attributes;
	 * };
	 *
	 * union READDIR3res switch (nfsstat3 status) {
	 * case NFS3_OK:
	 * 	READDIR3resok resok;
	 * default:
	 * 	READDIR3resfail resfail;
	 * };
	 */

	p = &(data[0]);
	p = rpc_add_credentials(p);

	p = nfs_add_fh3(p, &dir->fh);
	p = nfs_add_uint64(p, dir->cookie);

	memcpy(p, dir->cookieverf, NFS3_COOKIEVERFSIZE);
	p += NFS3_COOKIEVERFSIZE / 4;

	p = nfs_add_uint32(p, 1024); /* count */

	nfs_packet = rpc_req(npriv, PROG_NFS, NFSPROC3_READDIR, data, p - data);
	if (IS_ERR(nfs_packet))
		return NULL;

	p = (void *)nfs_packet->data + sizeof(struct rpc_reply) + 4;
	p = nfs_read_post_op_attr(p, NULL);

	/* update cookieverf */
	memcpy(dir->cookieverf, p, NFS3_COOKIEVERFSIZE);
	p += NFS3_COOKIEVERFSIZE / 4;

	len = (void *)nfs_packet->data + nfs_packet->len - (void *)p;
	if (!len) {
		printf("%s: huh, no payload left\n", __func__);
		free(nfs_packet);
		return NULL;
	}

	buf = xzalloc(len);

	memcpy(buf, p, len);

	free(nfs_packet);

	xdr_init(&dir->stream, buf, len);

	/* now xdr points to dirlist3 res.resok.reply */

	return buf;
}

/*
 * nfs_read_req - Read File on NFS Server
 */
static int nfs_read_req(struct file_priv *priv, uint64_t offset,
		uint32_t readlen)
{
	uint32_t data[1024];
	uint32_t *p;
	int len;
	struct packet *nfs_packet;
	uint32_t rlen, eof;

	/*
	 * struct READ3args {
	 * 	nfs_fh3 file;
	 * 	offset3 offset;
	 * 	count3 count;
	 * };
	 *
	 * struct READ3resok {
	 * 	post_op_attr file_attributes;
	 * 	count3 count;
	 * 	bool eof;
	 * 	opaque data<>;
	 * };
	 *
	 * struct READ3resfail {
	 * 	post_op_attr file_attributes;
	 * };
	 *
	 * union READ3res switch (nfsstat3 status) {
	 * case NFS3_OK:
	 * 	READ3resok resok;
	 * default:
	 * 	READ3resfail resfail;
	 * };
	 */
	p = &(data[0]);
	p = rpc_add_credentials(p);

	p = nfs_add_fh3(p, &priv->fh);
	p = nfs_add_uint64(p, offset);
	p = nfs_add_uint32(p, readlen);

	len = p - &(data[0]);

	nfs_packet = rpc_req(priv->npriv, PROG_NFS, NFSPROC3_READ, data, len);
	if (IS_ERR(nfs_packet))
		return PTR_ERR(nfs_packet);

	p = (void *)nfs_packet->data + sizeof(struct rpc_reply) + 4;

	p = nfs_read_post_op_attr(p, NULL);

	rlen = ntoh32(net_read_uint32(p));

	/* skip over count */
	p += 1;

	eof = ntoh32(net_read_uint32(p));

	/*
	 * skip over eof and count embedded in the representation of data
	 * assuming it equals rlen above.
	 */
	p += 2;

	if (readlen && !rlen && !eof) {
		free(nfs_packet);
		return -EIO;
	}

	kfifo_put(priv->fifo, (char *)p, rlen);

	free(nfs_packet);

	return 0;
}

static void nfs_handler(void *ctx, char *packet, unsigned len)
{
	char *pkt = net_eth_to_udp_payload(packet);
	struct nfs_priv *npriv = ctx;

	nfs_state = STATE_DONE;

	npriv->nfs_packet = xmalloc(sizeof(*npriv->nfs_packet) + len);
	memcpy(npriv->nfs_packet->data, pkt, len);
	npriv->nfs_packet->len = len;
}

static int nfs_truncate(struct device_d *dev, FILE *f, loff_t size)
{
	return -ENOSYS;
}

static void nfs_do_close(struct file_priv *priv)
{
	if (priv->fifo)
		kfifo_free(priv->fifo);

	free(priv);
}

static int nfs_readlink_req(struct nfs_priv *npriv, struct nfs_fh *fh,
			    char **target)
{
	uint32_t data[1024];
	uint32_t *p;
	uint32_t len;
	struct packet *nfs_packet;

	/*
	 * struct READLINK3args {
	 * 	nfs_fh3 symlink;
	 * };
	 *
	 * struct READLINK3resok {
	 * 	post_op_attr symlink_attributes;
	 * 	nfspath3 data;
	 * };
	 *
	 * struct READLINK3resfail {
	 * 	post_op_attr symlink_attributes;
	 * }
	 *
	 * union READLINK3res switch (nfsstat3 status) {
	 * case NFS3_OK:
	 * 	READLINK3resok resok;
	 * default:
	 * 	READLINK3resfail resfail;
	 * };
	 */
	p = &(data[0]);
	p = rpc_add_credentials(p);

	p = nfs_add_fh3(p, fh);

	len = p - &(data[0]);

	nfs_packet = rpc_req(npriv, PROG_NFS, NFSPROC3_READLINK, data, len);
	if (IS_ERR(nfs_packet))
		return PTR_ERR(nfs_packet);

	p = (void *)nfs_packet->data + sizeof(struct rpc_reply) + 4;

	p = nfs_read_post_op_attr(p, NULL);

	len = ntoh32(net_read_uint32(p)); /* new path length */
	p++;

	*target = xzalloc(len + 1);
	memcpy(*target, p, len);

	free(nfs_packet);

	return 0;
}

static const char *nfs_get_link(struct dentry *dentry, struct inode *inode)
{
	struct nfs_inode *ninode = nfsi(inode);
	struct nfs_priv *npriv = ninode->npriv;
	int ret;

	ret = nfs_readlink_req(npriv, &ninode->fh, &inode->i_link);
	if (ret)
		return ERR_PTR(ret);

	return inode->i_link;
}

static int nfs_open(struct device_d *dev, FILE *file, const char *filename)
{
	struct inode *inode = file->f_inode;
	struct nfs_inode *ninode = nfsi(inode);
	struct nfs_priv *npriv = ninode->npriv;
	struct file_priv *priv;

	priv = xzalloc(sizeof(*priv));
	priv->fh = ninode->fh;
	priv->npriv = npriv;
	file->priv = priv;
	file->size = inode->i_size;

	priv->fifo = kfifo_alloc(1024);
	if (!priv->fifo) {
		free(priv);
		return -ENOMEM;
	}

	return 0;
}

static int nfs_close(struct device_d *dev, FILE *file)
{
	struct file_priv *priv = file->priv;

	nfs_do_close(priv);

	return 0;
}

static int nfs_write(struct device_d *_dev, FILE *file, const void *inbuf,
		size_t insize)
{
	return -ENOSYS;
}

static int nfs_read(struct device_d *dev, FILE *file, void *buf, size_t insize)
{
	struct file_priv *priv = file->priv;

	if (insize > 1024)
		insize = 1024;

	if (insize && !kfifo_len(priv->fifo)) {
		int ret = nfs_read_req(priv, file->pos, insize);
		if (ret)
			return ret;
	}

	return kfifo_get(priv->fifo, buf, insize);
}

static int nfs_lseek(struct device_d *dev, FILE *file, loff_t pos)
{
	struct file_priv *priv = file->priv;

	kfifo_reset(priv->fifo);

	return 0;
}

static int nfs_iterate(struct file *file, struct dir_context *ctx)
{
	struct dentry *dentry = file->f_path.dentry;
	struct inode *dir = d_inode(dentry);
	struct nfs_priv *npriv = nfsi(dir)->npriv;
	void *buf = NULL;
	struct nfs_dir *ndir;
	struct xdr_stream *xdr;
	int ret;
	uint32_t *p, len;

	ndir = xzalloc(sizeof(*ndir));
	ndir->fh = nfsi(dir)->fh;

	while (1) {
		/* cookie == 0 and cookieverf == 0 means start of dir */
		buf = nfs_readdirattr_req(npriv, ndir);
		if (!buf) {
			pr_err("%s: nfs_readdirattr_req failed\n", __func__);
			ret = -EINVAL;
			goto out;
		}

		xdr = &ndir->stream;

		while (1) {
			char name[256];

			p = xdr_inline_decode(xdr, 4);
			if (!p)
				goto err_eop;

			if (!net_read_uint32(p)) {
				/* eof? */
				p = xdr_inline_decode(xdr, 4);
				if (!p)
					goto err_eop;

				if (net_read_uint32(p)) {
					ret = 0;
					goto out;
				}

				break;
			}

			/* skip over fileid */
			p = xdr_inline_decode(xdr, 8);
			if (!p)
				goto err_eop;

			ret = decode_filename(xdr, name, &len);
			if (ret)
				goto out;

			dir_emit(ctx, name, len, 0, DT_UNKNOWN);

			p = xdr_inline_decode(xdr, 8);
			if (!p)
				goto err_eop;

			ndir->cookie = ntoh64(net_read_uint64(p));
		}
		free(buf);
	}

	ret = 0;

out:
	free(ndir->stream.buf);
	free(ndir);

	return ret;

err_eop:
	pr_err("Unexpected end of packet\n");

	return -EIO;
}

static struct inode *nfs_alloc_inode(struct super_block *sb)
{
	struct nfs_inode *node;

	node = xzalloc(sizeof(*node));
	if (!node)
		return NULL;

	return &node->inode;
}

static const struct inode_operations nfs_file_inode_operations;
static const struct file_operations nfs_dir_operations;
static const struct inode_operations nfs_dir_inode_operations;
static const struct file_operations nfs_file_operations;
static const struct inode_operations nfs_symlink_inode_operations = {
	.get_link = nfs_get_link,
};

static int nfs_init_inode(struct nfs_priv *npriv, struct inode *inode,
			  unsigned int mode)
{
	struct nfs_inode *ninode = nfsi(inode);

	ninode->npriv = npriv;

	inode->i_ino = get_next_ino();
	inode->i_mode = mode;

	switch (inode->i_mode & S_IFMT) {
	default:
		return -EINVAL;
	case S_IFREG:
		inode->i_op = &nfs_file_inode_operations;
		inode->i_fop = &nfs_file_operations;
		break;
	case S_IFDIR:
		inode->i_op = &nfs_dir_inode_operations;
		inode->i_fop = &nfs_dir_operations;
		inc_nlink(inode);
		break;
	case S_IFLNK:
		inode->i_op = &nfs_symlink_inode_operations;
		break;
	}

	return 0;
}

static struct dentry *nfs_lookup(struct inode *dir, struct dentry *dentry,
				   unsigned int flags)
{
	struct nfs_inode *ndir = nfsi(dir);
	struct inode *inode = new_inode(dir->i_sb);
	struct nfs_priv *npriv = ndir->npriv;
	int ret;

	if (!inode)
		return NULL;

	ret = nfs_lookup_req(npriv, &ndir->fh, dentry->name, inode);
	if (ret)
		return NULL;

	nfs_init_inode(npriv, inode, inode->i_mode);

	d_add(dentry, inode);

	return NULL;
}

static const struct file_operations nfs_dir_operations = {
	.iterate = nfs_iterate,
};

static const struct inode_operations nfs_dir_inode_operations =
{
	.lookup = nfs_lookup,
};

static const struct super_operations nfs_ops = {
	.alloc_inode = nfs_alloc_inode,
};

static char *rootnfsopts;

static void nfs_set_rootarg(struct nfs_priv *npriv, struct fs_device_d *fsdev)
{
	char *str, *tmp;
	const char *bootargs;

	str = basprintf("root=/dev/nfs nfsroot=%pI4:%s%s%s", &npriv->server, npriv->path,
			  rootnfsopts[0] ? "," : "", rootnfsopts);

	/* forward specific mount options on demand */
	if (npriv->manual_nfs_port == 1) {
		tmp = basprintf("%s,port=%hu", str, npriv->nfs_port);
		free(str);
		str = tmp;
	}

	if (npriv->manual_mount_port == 1) {
		tmp = basprintf("%s,mountport=%hu", str, npriv->mount_port);
		free(str);
		str = tmp;
	}

	bootargs = dev_get_param(&npriv->con->edev->dev, "linux.bootargs");
	if (bootargs) {
		tmp = basprintf("%s %s", str, bootargs);
		free(str);
		str = tmp;
	}

	fsdev_set_linux_rootarg(fsdev, str);

	free(str);
}

static int nfs_probe(struct device_d *dev)
{
	struct fs_device_d *fsdev = dev_to_fs_device(dev);
	struct nfs_priv *npriv = xzalloc(sizeof(struct nfs_priv));
	struct super_block *sb = &fsdev->sb;
	char *tmp = xstrdup(fsdev->backingstore);
	char *path;
	struct inode *inode;
	int ret;

	dev->priv = npriv;

	debug("nfs: mount: %s\n", fsdev->backingstore);

	path = strchr(tmp, ':');
	if (!path) {
		ret = -EINVAL;
		goto err;
	}

	*path = 0;

	npriv->path = xstrdup(path + 1);

	ret = resolv(tmp, &npriv->server);
	if (ret) {
		printf("cannot resolve \"%s\": %s\n", tmp, strerror(-ret));
		goto err1;
	}

	debug("nfs: server: %s path: %s\n", tmp, npriv->path);

	npriv->con = net_udp_new(npriv->server, SUNRPC_PORT, nfs_handler, npriv);
	if (IS_ERR(npriv->con)) {
		ret = PTR_ERR(npriv->con);
		goto err1;
	}

	/* Need a priviliged source port */
	net_udp_bind(npriv->con, 1000);

	parseopt_hu(fsdev->options, "mountport", &npriv->mount_port);
	if (!npriv->mount_port) {
		ret = rpc_lookup_req(npriv, PROG_MOUNT, 3);
		if (ret < 0) {
			printf("lookup mount port failed with %d\n", ret);
			goto err2;
		}
		npriv->mount_port = ret;
	} else {
		npriv->manual_mount_port = 1;
	}
	debug("mount port: %hu\n", npriv->mount_port);

	parseopt_hu(fsdev->options, "port", &npriv->nfs_port);
	if (!npriv->nfs_port) {
		ret = rpc_lookup_req(npriv, PROG_NFS, 3);
		if (ret < 0) {
			printf("lookup nfs port failed with %d\n", ret);
			goto err2;
		}
		npriv->nfs_port = ret;
	} else {
		npriv->manual_nfs_port = 1;
	}
	debug("nfs port: %d\n", npriv->nfs_port);

	ret = nfs_mount_req(npriv);
	if (ret) {
		printf("mounting failed with %d\n", ret);
		goto err2;
	}

	nfs_set_rootarg(npriv, fsdev);

	free(tmp);

	sb->s_op = &nfs_ops;

	inode = new_inode(sb);
	nfs_set_fh(inode, &npriv->rootfh);
	nfs_init_inode(npriv, inode, S_IFDIR);
	sb->s_root = d_make_root(inode);

	return 0;

err2:
	net_unregister(npriv->con);
err1:
	free(npriv->path);
err:
	free(tmp);
	free(npriv);

	return ret;
}

static void nfs_remove(struct device_d *dev)
{
	struct nfs_priv *npriv = dev->priv;

	nfs_umount_req(npriv);

	net_unregister(npriv->con);
	free(npriv->path);
	free(npriv);
}

static struct fs_driver_d nfs_driver = {
	.open      = nfs_open,
	.close     = nfs_close,
	.read      = nfs_read,
	.lseek     = nfs_lseek,
	.write     = nfs_write,
	.truncate  = nfs_truncate,
	.flags     = 0,
	.drv = {
		.probe  = nfs_probe,
		.remove = nfs_remove,
		.name = "nfs",
	}
};

static int nfs_init(void)
{
	rootnfsopts = xstrdup("v3,tcp");

	globalvar_add_simple_string("linux.rootnfsopts", &rootnfsopts);

	return register_fs_driver(&nfs_driver);
}
coredevice_initcall(nfs_init);