27 files changed, 21138 insertions, 0 deletions

diff --git a/fs/nfs/Makefile b/fs/nfs/Makefile
new file mode 100644
index 00000000000..b4baa031edf
--- /dev/null
+++ b/fs/nfs/Makefile
@@ -0,0 +1,15 @@
+#
+# Makefile for the Linux nfs filesystem routines.
+#
+
+obj-$(CONFIG_NFS_FS) += nfs.o
+
+nfs-y 			:= dir.o file.o inode.o nfs2xdr.o pagelist.o \
+			   proc.o read.o symlink.o unlink.o write.o
+nfs-$(CONFIG_ROOT_NFS)	+= nfsroot.o mount_clnt.o      
+nfs-$(CONFIG_NFS_V3)	+= nfs3proc.o nfs3xdr.o
+nfs-$(CONFIG_NFS_V4)	+= nfs4proc.o nfs4xdr.o nfs4state.o nfs4renewd.o \
+			   delegation.o idmap.o \
+			   callback.o callback_xdr.o callback_proc.o
+nfs-$(CONFIG_NFS_DIRECTIO) += direct.o
+nfs-objs		:= $(nfs-y)
diff --git a/fs/nfs/callback.c b/fs/nfs/callback.c
new file mode 100644
index 00000000000..560d6175dd5
--- /dev/null
+++ b/fs/nfs/callback.c
@@ -0,0 +1,187 @@
+/*
+ * linux/fs/nfs/callback.c
+ *
+ * Copyright (C) 2004 Trond Myklebust
+ *
+ * NFSv4 callback handling
+ */
+
+#include <linux/config.h>
+#include <linux/completion.h>
+#include <linux/ip.h>
+#include <linux/module.h>
+#include <linux/smp_lock.h>
+#include <linux/sunrpc/svc.h>
+#include <linux/sunrpc/svcsock.h>
+#include <linux/nfs_fs.h>
+#include "callback.h"
+
+#define NFSDBG_FACILITY NFSDBG_CALLBACK
+
+struct nfs_callback_data {
+	unsigned int users;
+	struct svc_serv *serv;
+	pid_t pid;
+	struct completion started;
+	struct completion stopped;
+};
+
+static struct nfs_callback_data nfs_callback_info;
+static DECLARE_MUTEX(nfs_callback_sema);
+static struct svc_program nfs4_callback_program;
+
+unsigned short nfs_callback_tcpport;
+
+/*
+ * This is the callback kernel thread.
+ */
+static void nfs_callback_svc(struct svc_rqst *rqstp)
+{
+	struct svc_serv *serv = rqstp->rq_server;
+	int err;
+
+	__module_get(THIS_MODULE);
+	lock_kernel();
+
+	nfs_callback_info.pid = current->pid;
+	daemonize("nfsv4-svc");
+	/* Process request with signals blocked, but allow SIGKILL.  */
+	allow_signal(SIGKILL);
+
+	complete(&nfs_callback_info.started);
+
+	while (nfs_callback_info.users != 0 || !signalled()) {
+		/*
+		 * Listen for a request on the socket
+		 */
+		err = svc_recv(serv, rqstp, MAX_SCHEDULE_TIMEOUT);
+		if (err == -EAGAIN || err == -EINTR)
+			continue;
+		if (err < 0) {
+			printk(KERN_WARNING
+					"%s: terminating on error %d\n",
+					__FUNCTION__, -err);
+			break;
+		}
+		dprintk("%s: request from %u.%u.%u.%u\n", __FUNCTION__,
+				NIPQUAD(rqstp->rq_addr.sin_addr.s_addr));
+		svc_process(serv, rqstp);
+	}
+
+	nfs_callback_info.pid = 0;
+	complete(&nfs_callback_info.stopped);
+	unlock_kernel();
+	module_put_and_exit(0);
+}
+
+/*
+ * Bring up the server process if it is not already up.
+ */
+int nfs_callback_up(void)
+{
+	struct svc_serv *serv;
+	struct svc_sock *svsk;
+	int ret = 0;
+
+	lock_kernel();
+	down(&nfs_callback_sema);
+	if (nfs_callback_info.users++ || nfs_callback_info.pid != 0)
+		goto out;
+	init_completion(&nfs_callback_info.started);
+	init_completion(&nfs_callback_info.stopped);
+	serv = svc_create(&nfs4_callback_program, NFS4_CALLBACK_BUFSIZE);
+	ret = -ENOMEM;
+	if (!serv)
+		goto out_err;
+	/* FIXME: We don't want to register this socket with the portmapper */
+	ret = svc_makesock(serv, IPPROTO_TCP, 0);
+	if (ret < 0)
+		goto out_destroy;
+	if (!list_empty(&serv->sv_permsocks)) {
+		svsk = list_entry(serv->sv_permsocks.next,
+				struct svc_sock, sk_list);
+		nfs_callback_tcpport = ntohs(inet_sk(svsk->sk_sk)->sport);
+		dprintk ("Callback port = 0x%x\n", nfs_callback_tcpport);
+	} else
+		BUG();
+	ret = svc_create_thread(nfs_callback_svc, serv);
+	if (ret < 0)
+		goto out_destroy;
+	nfs_callback_info.serv = serv;
+	wait_for_completion(&nfs_callback_info.started);
+out:
+	up(&nfs_callback_sema);
+	unlock_kernel();
+	return ret;
+out_destroy:
+	svc_destroy(serv);
+out_err:
+	nfs_callback_info.users--;
+	goto out;
+}
+
+/*
+ * Kill the server process if it is not already up.
+ */
+int nfs_callback_down(void)
+{
+	int ret = 0;
+
+	lock_kernel();
+	down(&nfs_callback_sema);
+	if (--nfs_callback_info.users || nfs_callback_info.pid == 0)
+		goto out;
+	kill_proc(nfs_callback_info.pid, SIGKILL, 1);
+	wait_for_completion(&nfs_callback_info.stopped);
+out:
+	up(&nfs_callback_sema);
+	unlock_kernel();
+	return ret;
+}
+
+static int nfs_callback_authenticate(struct svc_rqst *rqstp)
+{
+	struct in_addr *addr = &rqstp->rq_addr.sin_addr;
+	struct nfs4_client *clp;
+
+	/* Don't talk to strangers */
+	clp = nfs4_find_client(addr);
+	if (clp == NULL)
+		return SVC_DROP;
+	dprintk("%s: %u.%u.%u.%u NFSv4 callback!\n", __FUNCTION__, NIPQUAD(addr));
+	nfs4_put_client(clp);
+	switch (rqstp->rq_authop->flavour) {
+		case RPC_AUTH_NULL:
+			if (rqstp->rq_proc != CB_NULL)
+				return SVC_DENIED;
+			break;
+		case RPC_AUTH_UNIX:
+			break;
+		case RPC_AUTH_GSS:
+			/* FIXME: RPCSEC_GSS handling? */
+		default:
+			return SVC_DENIED;
+	}
+	return SVC_OK;
+}
+
+/*
+ * Define NFS4 callback program
+ */
+extern struct svc_version nfs4_callback_version1;
+
+static struct svc_version *nfs4_callback_version[] = {
+	[1] = &nfs4_callback_version1,
+};
+
+static struct svc_stat nfs4_callback_stats;
+
+static struct svc_program nfs4_callback_program = {
+	.pg_prog = NFS4_CALLBACK,			/* RPC service number */
+	.pg_nvers = ARRAY_SIZE(nfs4_callback_version),	/* Number of entries */
+	.pg_vers = nfs4_callback_version,		/* version table */
+	.pg_name = "NFSv4 callback",			/* service name */
+	.pg_class = "nfs",				/* authentication class */
+	.pg_stats = &nfs4_callback_stats,
+	.pg_authenticate = nfs_callback_authenticate,
+};
diff --git a/fs/nfs/callback.h b/fs/nfs/callback.h
new file mode 100644
index 00000000000..a0db2d4f941
--- /dev/null
+++ b/fs/nfs/callback.h
@@ -0,0 +1,70 @@
+/*
+ * linux/fs/nfs/callback.h
+ *
+ * Copyright (C) 2004 Trond Myklebust
+ *
+ * NFSv4 callback definitions
+ */
+#ifndef __LINUX_FS_NFS_CALLBACK_H
+#define __LINUX_FS_NFS_CALLBACK_H
+
+#define NFS4_CALLBACK 0x40000000
+#define NFS4_CALLBACK_XDRSIZE 2048
+#define NFS4_CALLBACK_BUFSIZE (1024 + NFS4_CALLBACK_XDRSIZE)
+
+enum nfs4_callback_procnum {
+	CB_NULL = 0,
+	CB_COMPOUND = 1,
+};
+
+enum nfs4_callback_opnum {
+	OP_CB_GETATTR = 3,
+	OP_CB_RECALL  = 4,
+	OP_CB_ILLEGAL = 10044,
+};
+
+struct cb_compound_hdr_arg {
+	int taglen;
+	const char *tag;
+	unsigned int callback_ident;
+	unsigned nops;
+};
+
+struct cb_compound_hdr_res {
+	uint32_t *status;
+	int taglen;
+	const char *tag;
+	uint32_t *nops;
+};
+
+struct cb_getattrargs {
+	struct sockaddr_in *addr;
+	struct nfs_fh fh;
+	uint32_t bitmap[2];
+};
+
+struct cb_getattrres {
+	uint32_t status;
+	uint32_t bitmap[2];
+	uint64_t size;
+	uint64_t change_attr;
+	struct timespec ctime;
+	struct timespec mtime;
+};
+
+struct cb_recallargs {
+	struct sockaddr_in *addr;
+	struct nfs_fh fh;
+	nfs4_stateid stateid;
+	uint32_t truncate;
+};
+
+extern unsigned nfs4_callback_getattr(struct cb_getattrargs *args, struct cb_getattrres *res);
+extern unsigned nfs4_callback_recall(struct cb_recallargs *args, void *dummy);
+
+extern int nfs_callback_up(void);
+extern int nfs_callback_down(void);
+
+extern unsigned short nfs_callback_tcpport;
+
+#endif /* __LINUX_FS_NFS_CALLBACK_H */
diff --git a/fs/nfs/callback_proc.c b/fs/nfs/callback_proc.c
new file mode 100644
index 00000000000..ece27e42b93
--- /dev/null
+++ b/fs/nfs/callback_proc.c
@@ -0,0 +1,85 @@
+/*
+ * linux/fs/nfs/callback_proc.c
+ *
+ * Copyright (C) 2004 Trond Myklebust
+ *
+ * NFSv4 callback procedures
+ */
+#include <linux/config.h>
+#include <linux/nfs4.h>
+#include <linux/nfs_fs.h>
+#include "callback.h"
+#include "delegation.h"
+
+#define NFSDBG_FACILITY NFSDBG_CALLBACK
+ 
+unsigned nfs4_callback_getattr(struct cb_getattrargs *args, struct cb_getattrres *res)
+{
+	struct nfs4_client *clp;
+	struct nfs_delegation *delegation;
+	struct nfs_inode *nfsi;
+	struct inode *inode;
+	
+	res->bitmap[0] = res->bitmap[1] = 0;
+	res->status = htonl(NFS4ERR_BADHANDLE);
+	clp = nfs4_find_client(&args->addr->sin_addr);
+	if (clp == NULL)
+		goto out;
+	inode = nfs_delegation_find_inode(clp, &args->fh);
+	if (inode == NULL)
+		goto out_putclient;
+	nfsi = NFS_I(inode);
+	down_read(&nfsi->rwsem);
+	delegation = nfsi->delegation;
+	if (delegation == NULL || (delegation->type & FMODE_WRITE) == 0)
+		goto out_iput;
+	res->size = i_size_read(inode);
+	res->change_attr = NFS_CHANGE_ATTR(inode);
+	res->ctime = inode->i_ctime;
+	res->mtime = inode->i_mtime;
+	res->bitmap[0] = (FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE) &
+		args->bitmap[0];
+	res->bitmap[1] = (FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY) &
+		args->bitmap[1];
+	res->status = 0;
+out_iput:
+	up_read(&nfsi->rwsem);
+	iput(inode);
+out_putclient:
+	nfs4_put_client(clp);
+out:
+	dprintk("%s: exit with status = %d\n", __FUNCTION__, ntohl(res->status));
+	return res->status;
+}
+
+unsigned nfs4_callback_recall(struct cb_recallargs *args, void *dummy)
+{
+	struct nfs4_client *clp;
+	struct inode *inode;
+	unsigned res;
+	
+	res = htonl(NFS4ERR_BADHANDLE);
+	clp = nfs4_find_client(&args->addr->sin_addr);
+	if (clp == NULL)
+		goto out;
+	inode = nfs_delegation_find_inode(clp, &args->fh);
+	if (inode == NULL)
+		goto out_putclient;
+	/* Set up a helper thread to actually return the delegation */
+	switch(nfs_async_inode_return_delegation(inode, &args->stateid)) {
+		case 0:
+			res = 0;
+			break;
+		case -ENOENT:
+			res = htonl(NFS4ERR_BAD_STATEID);
+			break;
+		default:
+			res = htonl(NFS4ERR_RESOURCE);
+	}
+	iput(inode);
+out_putclient:
+	nfs4_put_client(clp);
+out:
+	dprintk("%s: exit with status = %d\n", __FUNCTION__, ntohl(res));
+	return res;
+}
diff --git a/fs/nfs/callback_xdr.c b/fs/nfs/callback_xdr.c
new file mode 100644
index 00000000000..d271df9df2b
--- /dev/null
+++ b/fs/nfs/callback_xdr.c
@@ -0,0 +1,481 @@
+/*
+ * linux/fs/nfs/callback_xdr.c
+ *
+ * Copyright (C) 2004 Trond Myklebust
+ *
+ * NFSv4 callback encode/decode procedures
+ */
+#include <linux/config.h>
+#include <linux/kernel.h>
+#include <linux/sunrpc/svc.h>
+#include <linux/nfs4.h>
+#include <linux/nfs_fs.h>
+#include "callback.h"
+
+#define CB_OP_TAGLEN_MAXSZ	(512)
+#define CB_OP_HDR_RES_MAXSZ	(2 + CB_OP_TAGLEN_MAXSZ)
+#define CB_OP_GETATTR_BITMAP_MAXSZ	(4)
+#define CB_OP_GETATTR_RES_MAXSZ	(CB_OP_HDR_RES_MAXSZ + \
+				CB_OP_GETATTR_BITMAP_MAXSZ + \
+				2 + 2 + 3 + 3)
+#define CB_OP_RECALL_RES_MAXSZ	(CB_OP_HDR_RES_MAXSZ)
+
+#define NFSDBG_FACILITY NFSDBG_CALLBACK
+
+typedef unsigned (*callback_process_op_t)(void *, void *);
+typedef unsigned (*callback_decode_arg_t)(struct svc_rqst *, struct xdr_stream *, void *);
+typedef unsigned (*callback_encode_res_t)(struct svc_rqst *, struct xdr_stream *, void *);
+
+
+struct callback_op {
+	callback_process_op_t process_op;
+	callback_decode_arg_t decode_args;
+	callback_encode_res_t encode_res;
+	long res_maxsize;
+};
+
+static struct callback_op callback_ops[];
+
+static int nfs4_callback_null(struct svc_rqst *rqstp, void *argp, void *resp)
+{
+	return htonl(NFS4_OK);
+}
+
+static int nfs4_decode_void(struct svc_rqst *rqstp, uint32_t *p, void *dummy)
+{
+	return xdr_argsize_check(rqstp, p);
+}
+
+static int nfs4_encode_void(struct svc_rqst *rqstp, uint32_t *p, void *dummy)
+{
+	return xdr_ressize_check(rqstp, p);
+}
+
+static uint32_t *read_buf(struct xdr_stream *xdr, int nbytes)
+{
+	uint32_t *p;
+
+	p = xdr_inline_decode(xdr, nbytes);
+	if (unlikely(p == NULL))
+		printk(KERN_WARNING "NFSv4 callback reply buffer overflowed!\n");
+	return p;
+}
+
+static unsigned decode_string(struct xdr_stream *xdr, unsigned int *len, const char **str)
+{
+	uint32_t *p;
+
+	p = read_buf(xdr, 4);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	*len = ntohl(*p);
+
+	if (*len != 0) {
+		p = read_buf(xdr, *len);
+		if (unlikely(p == NULL))
+			return htonl(NFS4ERR_RESOURCE);
+		*str = (const char *)p;
+	} else
+		*str = NULL;
+
+	return 0;
+}
+
+static unsigned decode_fh(struct xdr_stream *xdr, struct nfs_fh *fh)
+{
+	uint32_t *p;
+
+	p = read_buf(xdr, 4);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	fh->size = ntohl(*p);
+	if (fh->size > NFS4_FHSIZE)
+		return htonl(NFS4ERR_BADHANDLE);
+	p = read_buf(xdr, fh->size);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	memcpy(&fh->data[0], p, fh->size);
+	memset(&fh->data[fh->size], 0, sizeof(fh->data) - fh->size);
+	return 0;
+}
+
+static unsigned decode_bitmap(struct xdr_stream *xdr, uint32_t *bitmap)
+{
+	uint32_t *p;
+	unsigned int attrlen;
+
+	p = read_buf(xdr, 4);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	attrlen = ntohl(*p);
+	p = read_buf(xdr, attrlen << 2);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	if (likely(attrlen > 0))
+		bitmap[0] = ntohl(*p++);
+	if (attrlen > 1)
+		bitmap[1] = ntohl(*p);
+	return 0;
+}
+
+static unsigned decode_stateid(struct xdr_stream *xdr, nfs4_stateid *stateid)
+{
+	uint32_t *p;
+
+	p = read_buf(xdr, 16);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	memcpy(stateid->data, p, 16);
+	return 0;
+}
+
+static unsigned decode_compound_hdr_arg(struct xdr_stream *xdr, struct cb_compound_hdr_arg *hdr)
+{
+	uint32_t *p;
+	unsigned int minor_version;
+	unsigned status;
+
+	status = decode_string(xdr, &hdr->taglen, &hdr->tag);
+	if (unlikely(status != 0))
+		return status;
+	/* We do not like overly long tags! */
+	if (hdr->taglen > CB_OP_TAGLEN_MAXSZ-12 || hdr->taglen < 0) {
+		printk("NFSv4 CALLBACK %s: client sent tag of length %u\n",
+				__FUNCTION__, hdr->taglen);
+		return htonl(NFS4ERR_RESOURCE);
+	}
+	p = read_buf(xdr, 12);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	minor_version = ntohl(*p++);
+	/* Check minor version is zero. */
+	if (minor_version != 0) {
+		printk(KERN_WARNING "%s: NFSv4 server callback with illegal minor version %u!\n",
+				__FUNCTION__, minor_version);
+		return htonl(NFS4ERR_MINOR_VERS_MISMATCH);
+	}
+	hdr->callback_ident = ntohl(*p++);
+	hdr->nops = ntohl(*p);
+	return 0;
+}
+
+static unsigned decode_op_hdr(struct xdr_stream *xdr, unsigned int *op)
+{
+	uint32_t *p;
+	p = read_buf(xdr, 4);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	*op = ntohl(*p);
+	return 0;
+}
+
+static unsigned decode_getattr_args(struct svc_rqst *rqstp, struct xdr_stream *xdr, struct cb_getattrargs *args)
+{
+	unsigned status;
+
+	status = decode_fh(xdr, &args->fh);
+	if (unlikely(status != 0))
+		goto out;
+	args->addr = &rqstp->rq_addr;
+	status = decode_bitmap(xdr, args->bitmap);
+out:
+	dprintk("%s: exit with status = %d\n", __FUNCTION__, status);
+	return status;
+}
+
+static unsigned decode_recall_args(struct svc_rqst *rqstp, struct xdr_stream *xdr, struct cb_recallargs *args)
+{
+	uint32_t *p;
+	unsigned status;
+
+	args->addr = &rqstp->rq_addr;
+	status = decode_stateid(xdr, &args->stateid);
+	if (unlikely(status != 0))
+		goto out;
+	p = read_buf(xdr, 4);
+	if (unlikely(p == NULL)) {
+		status = htonl(NFS4ERR_RESOURCE);
+		goto out;
+	}
+	args->truncate = ntohl(*p);
+	status = decode_fh(xdr, &args->fh);
+out:
+	dprintk("%s: exit with status = %d\n", __FUNCTION__, status);
+	return 0;
+}
+
+static unsigned encode_string(struct xdr_stream *xdr, unsigned int len, const char *str)
+{
+	uint32_t *p;
+
+	p = xdr_reserve_space(xdr, 4 + len);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	xdr_encode_opaque(p, str, len);
+	return 0;
+}
+
+#define CB_SUPPORTED_ATTR0 (FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE)
+#define CB_SUPPORTED_ATTR1 (FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY)
+static unsigned encode_attr_bitmap(struct xdr_stream *xdr, const uint32_t *bitmap, uint32_t **savep)
+{
+	uint32_t bm[2];
+	uint32_t *p;
+
+	bm[0] = htonl(bitmap[0] & CB_SUPPORTED_ATTR0);
+	bm[1] = htonl(bitmap[1] & CB_SUPPORTED_ATTR1);
+	if (bm[1] != 0) {
+		p = xdr_reserve_space(xdr, 16);
+		if (unlikely(p == NULL))
+			return htonl(NFS4ERR_RESOURCE);
+		*p++ = htonl(2);
+		*p++ = bm[0];
+		*p++ = bm[1];
+	} else if (bm[0] != 0) {
+		p = xdr_reserve_space(xdr, 12);
+		if (unlikely(p == NULL))
+			return htonl(NFS4ERR_RESOURCE);
+		*p++ = htonl(1);
+		*p++ = bm[0];
+	} else {
+		p = xdr_reserve_space(xdr, 8);
+		if (unlikely(p == NULL))
+			return htonl(NFS4ERR_RESOURCE);
+		*p++ = htonl(0);
+	}
+	*savep = p;
+	return 0;
+}
+
+static unsigned encode_attr_change(struct xdr_stream *xdr, const uint32_t *bitmap, uint64_t change)
+{
+	uint32_t *p;
+
+	if (!(bitmap[0] & FATTR4_WORD0_CHANGE))
+		return 0;
+	p = xdr_reserve_space(xdr, 8);
+	if (unlikely(p == 0))
+		return htonl(NFS4ERR_RESOURCE);
+	p = xdr_encode_hyper(p, change);
+	return 0;
+}
+
+static unsigned encode_attr_size(struct xdr_stream *xdr, const uint32_t *bitmap, uint64_t size)
+{
+	uint32_t *p;
+
+	if (!(bitmap[0] & FATTR4_WORD0_SIZE))
+		return 0;
+	p = xdr_reserve_space(xdr, 8);
+	if (unlikely(p == 0))
+		return htonl(NFS4ERR_RESOURCE);
+	p = xdr_encode_hyper(p, size);
+	return 0;
+}
+
+static unsigned encode_attr_time(struct xdr_stream *xdr, const struct timespec *time)
+{
+	uint32_t *p;
+
+	p = xdr_reserve_space(xdr, 12);
+	if (unlikely(p == 0))
+		return htonl(NFS4ERR_RESOURCE);
+	p = xdr_encode_hyper(p, time->tv_sec);
+	*p = htonl(time->tv_nsec);
+	return 0;
+}
+
+static unsigned encode_attr_ctime(struct xdr_stream *xdr, const uint32_t *bitmap, const struct timespec *time)
+{
+	if (!(bitmap[1] & FATTR4_WORD1_TIME_METADATA))
+		return 0;
+	return encode_attr_time(xdr,time);
+}
+
+static unsigned encode_attr_mtime(struct xdr_stream *xdr, const uint32_t *bitmap, const struct timespec *time)
+{
+	if (!(bitmap[1] & FATTR4_WORD1_TIME_MODIFY))
+		return 0;
+	return encode_attr_time(xdr,time);
+}
+
+static unsigned encode_compound_hdr_res(struct xdr_stream *xdr, struct cb_compound_hdr_res *hdr)
+{
+	unsigned status;
+
+	hdr->status = xdr_reserve_space(xdr, 4);
+	if (unlikely(hdr->status == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	status = encode_string(xdr, hdr->taglen, hdr->tag);
+	if (unlikely(status != 0))
+		return status;
+	hdr->nops = xdr_reserve_space(xdr, 4);
+	if (unlikely(hdr->nops == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	return 0;
+}
+
+static unsigned encode_op_hdr(struct xdr_stream *xdr, uint32_t op, uint32_t res)
+{
+	uint32_t *p;
+	
+	p = xdr_reserve_space(xdr, 8);
+	if (unlikely(p == NULL))
+		return htonl(NFS4ERR_RESOURCE);
+	*p++ = htonl(op);
+	*p = res;
+	return 0;
+}
+
+static unsigned encode_getattr_res(struct svc_rqst *rqstp, struct xdr_stream *xdr, const struct cb_getattrres *res)
+{
+	uint32_t *savep;
+	unsigned status = res->status;
+	
+	if (unlikely(status != 0))
+		goto out;
+	status = encode_attr_bitmap(xdr, res->bitmap, &savep);
+	if (unlikely(status != 0))
+		goto out;
+	status = encode_attr_change(xdr, res->bitmap, res->change_attr);
+	if (unlikely(status != 0))
+		goto out;
+	status = encode_attr_size(xdr, res->bitmap, res->size);
+	if (unlikely(status != 0))
+		goto out;
+	status = encode_attr_ctime(xdr, res->bitmap, &res->ctime);
+	if (unlikely(status != 0))
+		goto out;
+	status = encode_attr_mtime(xdr, res->bitmap, &res->mtime);
+	*savep = htonl((unsigned int)((char *)xdr->p - (char *)(savep+1)));
+out:
+	dprintk("%s: exit with status = %d\n", __FUNCTION__, status);
+	return status;
+}
+
+static unsigned process_op(struct svc_rqst *rqstp,
+		struct xdr_stream *xdr_in, void *argp,
+		struct xdr_stream *xdr_out, void *resp)
+{
+	struct callback_op *op;
+	unsigned int op_nr;
+	unsigned int status = 0;
+	long maxlen;
+	unsigned res;
+
+	dprintk("%s: start\n", __FUNCTION__);
+	status = decode_op_hdr(xdr_in, &op_nr);
+	if (unlikely(status != 0)) {
+		op_nr = OP_CB_ILLEGAL;
+		op = &callback_ops[0];
+	} else if (unlikely(op_nr != OP_CB_GETATTR && op_nr != OP_CB_RECALL)) {
+		op_nr = OP_CB_ILLEGAL;
+		op = &callback_ops[0];
+		status = htonl(NFS4ERR_OP_ILLEGAL);
+	} else
+		op = &callback_ops[op_nr];
+
+	maxlen = xdr_out->end - xdr_out->p;
+	if (maxlen > 0 && maxlen < PAGE_SIZE) {
+		if (likely(status == 0 && op->decode_args != NULL))
+			status = op->decode_args(rqstp, xdr_in, argp);
+		if (likely(status == 0 && op->process_op != NULL))
+			status = op->process_op(argp, resp);
+	} else
+		status = htonl(NFS4ERR_RESOURCE);
+
+	res = encode_op_hdr(xdr_out, op_nr, status);
+	if (status == 0)
+		status = res;
+	if (op->encode_res != NULL && status == 0)
+		status = op->encode_res(rqstp, xdr_out, resp);
+	dprintk("%s: done, status = %d\n", __FUNCTION__, status);
+	return status;
+}
+
+/*
+ * Decode, process and encode a COMPOUND
+ */
+static int nfs4_callback_compound(struct svc_rqst *rqstp, void *argp, void *resp)
+{
+	struct cb_compound_hdr_arg hdr_arg;
+	struct cb_compound_hdr_res hdr_res;
+	struct xdr_stream xdr_in, xdr_out;
+	uint32_t *p;
+	unsigned int status;
+	unsigned int nops = 1;
+
+	dprintk("%s: start\n", __FUNCTION__);
+
+	xdr_init_decode(&xdr_in, &rqstp->rq_arg, rqstp->rq_arg.head[0].iov_base);
+
+	p = (uint32_t*)((char *)rqstp->rq_res.head[0].iov_base + rqstp->rq_res.head[0].iov_len);
+	rqstp->rq_res.head[0].iov_len = PAGE_SIZE;
+	xdr_init_encode(&xdr_out, &rqstp->rq_res, p);
+
+	decode_compound_hdr_arg(&xdr_in, &hdr_arg);
+	hdr_res.taglen = hdr_arg.taglen;
+	hdr_res.tag = hdr_arg.tag;
+	encode_compound_hdr_res(&xdr_out, &hdr_res);
+
+	for (;;) {
+		status = process_op(rqstp, &xdr_in, argp, &xdr_out, resp);
+		if (status != 0)
+			break;
+		if (nops == hdr_arg.nops)
+			break;
+		nops++;
+	}
+	*hdr_res.status = status;
+	*hdr_res.nops = htonl(nops);
+	dprintk("%s: done, status = %u\n", __FUNCTION__, status);
+	return rpc_success;
+}
+
+/*
+ * Define NFS4 callback COMPOUND ops.
+ */
+static struct callback_op callback_ops[] = {
+	[0] = {
+		.res_maxsize = CB_OP_HDR_RES_MAXSZ,
+	},
+	[OP_CB_GETATTR] = {
+		.process_op = (callback_process_op_t)nfs4_callback_getattr,
+		.decode_args = (callback_decode_arg_t)decode_getattr_args,
+		.encode_res = (callback_encode_res_t)encode_getattr_res,
+		.res_maxsize = CB_OP_GETATTR_RES_MAXSZ,
+	},
+	[OP_CB_RECALL] = {
+		.process_op = (callback_process_op_t)nfs4_callback_recall,
+		.decode_args = (callback_decode_arg_t)decode_recall_args,
+		.res_maxsize = CB_OP_RECALL_RES_MAXSZ,
+	}
+};
+
+/*
+ * Define NFS4 callback procedures
+ */
+static struct svc_procedure nfs4_callback_procedures1[] = {
+	[CB_NULL] = {
+		.pc_func = nfs4_callback_null,
+		.pc_decode = (kxdrproc_t)nfs4_decode_void,
+		.pc_encode = (kxdrproc_t)nfs4_encode_void,
+		.pc_xdrressize = 1,
+	},
+	[CB_COMPOUND] = {
+		.pc_func = nfs4_callback_compound,
+		.pc_encode = (kxdrproc_t)nfs4_encode_void,
+		.pc_argsize = 256,
+		.pc_ressize = 256,
+		.pc_xdrressize = NFS4_CALLBACK_BUFSIZE,
+	}
+};
+
+struct svc_version nfs4_callback_version1 = {
+	.vs_vers = 1,
+	.vs_nproc = ARRAY_SIZE(nfs4_callback_procedures1),
+	.vs_proc = nfs4_callback_procedures1,
+	.vs_xdrsize = NFS4_CALLBACK_XDRSIZE,
+	.vs_dispatch = NULL,
+};
+
diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
new file mode 100644
index 00000000000..5b9c60f9779
--- /dev/null
+++ b/fs/nfs/delegation.c
@@ -0,0 +1,342 @@
+/*
+ * linux/fs/nfs/delegation.c
+ *
+ * Copyright (C) 2004 Trond Myklebust
+ *
+ * NFS file delegation management
+ *
+ */
+#include <linux/config.h>
+#include <linux/completion.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/spinlock.h>
+
+#include <linux/nfs4.h>
+#include <linux/nfs_fs.h>
+#include <linux/nfs_xdr.h>
+
+#include "delegation.h"
+
+static struct nfs_delegation *nfs_alloc_delegation(void)
+{
+	return (struct nfs_delegation *)kmalloc(sizeof(struct nfs_delegation), GFP_KERNEL);
+}
+
+static void nfs_free_delegation(struct nfs_delegation *delegation)
+{
+	if (delegation->cred)
+		put_rpccred(delegation->cred);
+	kfree(delegation);
+}
+
+static void nfs_delegation_claim_opens(struct inode *inode)
+{
+	struct nfs_inode *nfsi = NFS_I(inode);
+	struct nfs_open_context *ctx;
+	struct nfs4_state *state;
+
+again:
+	spin_lock(&inode->i_lock);
+	list_for_each_entry(ctx, &nfsi->open_files, list) {
+		state = ctx->state;
+		if (state == NULL)
+			continue;
+		if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
+			continue;
+		get_nfs_open_context(ctx);
+		spin_unlock(&inode->i_lock);
+		if (nfs4_open_delegation_recall(ctx->dentry, state) < 0)
+			return;
+		put_nfs_open_context(ctx);
+		goto again;
+	}
+	spin_unlock(&inode->i_lock);
+}
+
+/*
+ * Set up a delegation on an inode
+ */
+void nfs_inode_reclaim_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res)
+{
+	struct nfs_delegation *delegation = NFS_I(inode)->delegation;
+
+	if (delegation == NULL)
+		return;
+	memcpy(delegation->stateid.data, res->delegation.data,
+			sizeof(delegation->stateid.data));
+	delegation->type = res->delegation_type;
+	delegation->maxsize = res->maxsize;
+	put_rpccred(cred);
+	delegation->cred = get_rpccred(cred);
+	delegation->flags &= ~NFS_DELEGATION_NEED_RECLAIM;
+	NFS_I(inode)->delegation_state = delegation->type;
+	smp_wmb();
+}
+
+/*
+ * Set up a delegation on an inode
+ */
+int nfs_inode_set_delegation(struct inode *inode, struct rpc_cred *cred, struct nfs_openres *res)
+{
+	struct nfs4_client *clp = NFS_SERVER(inode)->nfs4_state;
+	struct nfs_inode *nfsi = NFS_I(inode);
+	struct nfs_delegation *delegation;
+	int status = 0;
+
+	delegation = nfs_alloc_delegation();
+	if (delegation == NULL)
+		return -ENOMEM;
+	memcpy(delegation->stateid.data, res->delegation.data,
+			sizeof(delegation->stateid.data));
+	delegation->type = res->delegation_type;
+	delegation->maxsize = res->maxsize;
+	delegation->cred = get_rpccred(cred);
+	delegation->inode = inode;
+
+	spin_lock(&clp->cl_lock);
+	if (nfsi->delegation == NULL) {
+		list_add(&delegation->super_list, &clp->cl_delegations);
+		nfsi->delegation = delegation;
+		nfsi->delegation_state = delegation->type;
+		delegation = NULL;
+	} else {
+		if (memcmp(&delegation->stateid, &nfsi->delegation->stateid,
+					sizeof(delegation->stateid)) != 0 ||
+				delegation->type != nfsi->delegation->type) {
+			printk("%s: server %u.%u.%u.%u, handed out a duplicate delegation!\n",
+					__FUNCTION__, NIPQUAD(clp->cl_addr));
+			status = -EIO;
+		}
+	}
+	spin_unlock(&clp->cl_lock);
+	if (delegation != NULL)
+		kfree(delegation);
+	return status;
+}
+
+static int nfs_do_return_delegation(struct inode *inode, struct nfs_delegation *delegation)
+{
+	int res = 0;
+
+	__nfs_revalidate_inode(NFS_SERVER(inode), inode);
+
+	res = nfs4_proc_delegreturn(inode, delegation->cred, &delegation->stateid);
+	nfs_free_delegation(delegation);
+	return res;
+}
+
+/* Sync all data to disk upon delegation return */
+static void nfs_msync_inode(struct inode *inode)
+{
+	filemap_fdatawrite(inode->i_mapping);
+	nfs_wb_all(inode);
+	filemap_fdatawait(inode->i_mapping);
+}
+
+/*
+ * Basic procedure for returning a delegation to the server
+ */<