nfs4proc.c

/*
 *  fs/nfs/nfs4proc.c
 *
 *  Client-side procedure declarations for NFSv4.
 *
 *  Copyright (c) 2002 The Regents of the University of Michigan.
 *  All rights reserved.
 *
 *  Kendrick Smith <kmsmith@umich.edu>
 *  Andy Adamson   <andros@umich.edu>
 *
 *  Redistribution and use in source and binary forms, with or without
 *  modification, are permitted provided that the following conditions
 *  are met:
 *
 *  1. Redistributions of source code must retain the above copyright
 *     notice, this list of conditions and the following disclaimer.
 *  2. Redistributions in binary form must reproduce the above copyright
 *     notice, this list of conditions and the following disclaimer in the
 *     documentation and/or other materials provided with the distribution.
 *  3. Neither the name of the University nor the names of its
 *     contributors may be used to endorse or promote products derived
 *     from this software without specific prior written permission.
 *
 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
 *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
 *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
 *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include <linux/mm.h>
#include <linux/utsname.h>
#include <linux/delay.h>
#include <linux/errno.h>
#include <linux/string.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs.h>
#include <linux/nfs4.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_page.h>
#include <linux/smp_lock.h>
#include <linux/namei.h>
#include <linux/mount.h>

#include "nfs4_fs.h"
#include "delegation.h"

#define NFSDBG_FACILITY		NFSDBG_PROC

#define NFS4_POLL_RETRY_MIN	(1*HZ)
#define NFS4_POLL_RETRY_MAX	(15*HZ)

static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid);
static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
static int nfs4_async_handle_error(struct rpc_task *, const struct nfs_server *);
static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry);
static int nfs4_handle_exception(const struct nfs_server *server, int errorcode, struct nfs4_exception *exception);
extern u32 *nfs4_decode_dirent(u32 *p, struct nfs_entry *entry, int plus);
extern struct rpc_procinfo nfs4_procedures[];

/* Prevent leaks of NFSv4 errors into userland */
int nfs4_map_errors(int err)
{
	if (err < -1000) {
		dprintk("%s could not handle NFSv4 error %d\n",
				__FUNCTION__, -err);
		return -EIO;
	}
	return err;
}

/*
 * This is our standard bitmap for GETATTR requests.
 */
const u32 nfs4_fattr_bitmap[2] = {
	FATTR4_WORD0_TYPE
	| FATTR4_WORD0_CHANGE
	| FATTR4_WORD0_SIZE
	| FATTR4_WORD0_FSID
	| FATTR4_WORD0_FILEID,
	FATTR4_WORD1_MODE
	| FATTR4_WORD1_NUMLINKS
	| FATTR4_WORD1_OWNER
	| FATTR4_WORD1_OWNER_GROUP
	| FATTR4_WORD1_RAWDEV
	| FATTR4_WORD1_SPACE_USED
	| FATTR4_WORD1_TIME_ACCESS
	| FATTR4_WORD1_TIME_METADATA
	| FATTR4_WORD1_TIME_MODIFY
};

const u32 nfs4_statfs_bitmap[2] = {
	FATTR4_WORD0_FILES_AVAIL
	| FATTR4_WORD0_FILES_FREE
	| FATTR4_WORD0_FILES_TOTAL,
	FATTR4_WORD1_SPACE_AVAIL
	| FATTR4_WORD1_SPACE_FREE
	| FATTR4_WORD1_SPACE_TOTAL
};

const u32 nfs4_pathconf_bitmap[2] = {
	FATTR4_WORD0_MAXLINK
	| FATTR4_WORD0_MAXNAME,
	0
};

const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
			| FATTR4_WORD0_MAXREAD
			| FATTR4_WORD0_MAXWRITE
			| FATTR4_WORD0_LEASE_TIME,
			0
};

static void nfs4_setup_readdir(u64 cookie, u32 *verifier, struct dentry *dentry,
		struct nfs4_readdir_arg *readdir)
{
	u32 *start, *p;

	BUG_ON(readdir->count < 80);
	if (cookie > 2) {
		readdir->cookie = cookie;
		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
		return;
	}

	readdir->cookie = 0;
	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
	if (cookie == 2)
		return;
	
	/*
	 * NFSv4 servers do not return entries for '.' and '..'
	 * Therefore, we fake these entries here.  We let '.'
	 * have cookie 0 and '..' have cookie 1.  Note that
	 * when talking to the server, we always send cookie 0
	 * instead of 1 or 2.
	 */
	start = p = (u32 *)kmap_atomic(*readdir->pages, KM_USER0);
	
	if (cookie == 0) {
		*p++ = xdr_one;                                  /* next */
		*p++ = xdr_zero;                   /* cookie, first word */
		*p++ = xdr_one;                   /* cookie, second word */
		*p++ = xdr_one;                             /* entry len */
		memcpy(p, ".\0\0\0", 4);                        /* entry */
		p++;
		*p++ = xdr_one;                         /* bitmap length */
		*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
		*p++ = htonl(8);              /* attribute buffer length */
		p = xdr_encode_hyper(p, dentry->d_inode->i_ino);
	}
	
	*p++ = xdr_one;                                  /* next */
	*p++ = xdr_zero;                   /* cookie, first word */
	*p++ = xdr_two;                   /* cookie, second word */
	*p++ = xdr_two;                             /* entry len */
	memcpy(p, "..\0\0", 4);                         /* entry */
	p++;
	*p++ = xdr_one;                         /* bitmap length */
	*p++ = htonl(FATTR4_WORD0_FILEID);             /* bitmap */
	*p++ = htonl(8);              /* attribute buffer length */
	p = xdr_encode_hyper(p, dentry->d_parent->d_inode->i_ino);

	readdir->pgbase = (char *)p - (char *)start;
	readdir->count -= readdir->pgbase;
	kunmap_atomic(start, KM_USER0);
}

static void
renew_lease(struct nfs_server *server, unsigned long timestamp)
{
	struct nfs4_client *clp = server->nfs4_state;
	spin_lock(&clp->cl_lock);
	if (time_before(clp->cl_last_renewal,timestamp))
		clp->cl_last_renewal = timestamp;
	spin_unlock(&clp->cl_lock);
}

static void update_changeattr(struct inode *inode, struct nfs4_change_info *cinfo)
{
	struct nfs_inode *nfsi = NFS_I(inode);

	spin_lock(&inode->i_lock);
	nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
	if (cinfo->before == nfsi->change_attr && cinfo->atomic)
		nfsi->change_attr = cinfo->after;
	spin_unlock(&inode->i_lock);
}

/* Helper for asynchronous RPC calls */
static int nfs4_call_async(struct rpc_clnt *clnt,
		const struct rpc_call_ops *tk_ops, void *calldata)
{
	struct rpc_task *task;

	if (!(task = rpc_new_task(clnt, RPC_TASK_ASYNC, tk_ops, calldata)))
		return -ENOMEM;
	rpc_execute(task);
	return 0;
}

static int nfs4_wait_for_completion_rpc_task(struct rpc_task *task)
{
	sigset_t oldset;
	int ret;

	rpc_clnt_sigmask(task->tk_client, &oldset);
	ret = rpc_wait_for_completion_task(task);
	rpc_clnt_sigunmask(task->tk_client, &oldset);
	return ret;
}

static void update_open_stateid(struct nfs4_state *state, nfs4_stateid *stateid, int open_flags)
{
	struct inode *inode = state->inode;

	open_flags &= (FMODE_READ|FMODE_WRITE);
	/* Protect against nfs4_find_state_byowner() */
	spin_lock(&state->owner->so_lock);
	spin_lock(&inode->i_lock);
	memcpy(&state->stateid, stateid, sizeof(state->stateid));
	if ((open_flags & FMODE_WRITE))
		state->nwriters++;
	if (open_flags & FMODE_READ)
		state->nreaders++;
	nfs4_state_set_mode_locked(state, state->state | open_flags);
	spin_unlock(&inode->i_lock);
	spin_unlock(&state->owner->so_lock);
}

/*
 * OPEN_RECLAIM:
 * 	reclaim state on the server after a reboot.
 */
static int _nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
{
	struct inode *inode = state->inode;
	struct nfs_server *server = NFS_SERVER(inode);
	struct nfs_delegation *delegation = NFS_I(inode)->delegation;
	struct nfs_openargs o_arg = {
		.fh = NFS_FH(inode),
		.id = sp->so_id,
		.open_flags = state->state,
		.clientid = server->nfs4_state->cl_clientid,
		.claim = NFS4_OPEN_CLAIM_PREVIOUS,
		.bitmask = server->attr_bitmask,
	};
	struct nfs_openres o_res = {
		.server = server,	/* Grrr */
	};
	struct rpc_message msg = {
		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
		.rpc_argp       = &o_arg,
		.rpc_resp	= &o_res,
		.rpc_cred	= sp->so_cred,
	};
	int status;

	if (delegation != NULL) {
		if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
			memcpy(&state->stateid, &delegation->stateid,
					sizeof(state->stateid));
			set_bit(NFS_DELEGATED_STATE, &state->flags);
			return 0;
		}
		o_arg.u.delegation_type = delegation->type;
	}
	o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
	if (o_arg.seqid == NULL)
		return -ENOMEM;
	status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
	/* Confirm the sequence as being established */
	nfs_confirm_seqid(&sp->so_seqid, status);
	nfs_increment_open_seqid(status, o_arg.seqid);
	if (status == 0) {
		memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
		if (o_res.delegation_type != 0) {
			nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
			/* Did the server issue an immediate delegation recall? */
			if (o_res.do_recall)
				nfs_async_inode_return_delegation(inode, &o_res.stateid);
		}
	}
	nfs_free_seqid(o_arg.seqid);
	clear_bit(NFS_DELEGATED_STATE, &state->flags);
	/* Ensure we update the inode attributes */
	NFS_CACHEINV(inode);
	return status;
}

static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
{
	struct nfs_server *server = NFS_SERVER(state->inode);
	struct nfs4_exception exception = { };
	int err;
	do {
		err = _nfs4_open_reclaim(sp, state);
		if (err != -NFS4ERR_DELAY)
			break;
		nfs4_handle_exception(server, err, &exception);
	} while (exception.retry);
	return err;
}

static int _nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
{
	struct nfs4_state_owner  *sp  = state->owner;
	struct inode *inode = dentry->d_inode;
	struct nfs_server *server = NFS_SERVER(inode);
	struct dentry *parent = dget_parent(dentry);
	struct nfs_openargs arg = {
		.fh = NFS_FH(parent->d_inode),
		.clientid = server->nfs4_state->cl_clientid,
		.name = &dentry->d_name,
		.id = sp->so_id,
		.server = server,
		.bitmask = server->attr_bitmask,
		.claim = NFS4_OPEN_CLAIM_DELEGATE_CUR,
	};
	struct nfs_openres res = {
		.server = server,
	};
	struct 	rpc_message msg = {
		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR],
		.rpc_argp       = &arg,
		.rpc_resp       = &res,
		.rpc_cred	= sp->so_cred,
	};
	int status = 0;

	if (!test_bit(NFS_DELEGATED_STATE, &state->flags))
		goto out;
	if (state->state == 0)
		goto out;
	arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
	status = -ENOMEM;
	if (arg.seqid == NULL)
		goto out;
	arg.open_flags = state->state;
	memcpy(arg.u.delegation.data, state->stateid.data, sizeof(arg.u.delegation.data));
	status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
	nfs_increment_open_seqid(status, arg.seqid);
	if (status != 0)
		goto out_free;
	if(res.rflags & NFS4_OPEN_RESULT_CONFIRM) {
		status = _nfs4_proc_open_confirm(server->client, NFS_FH(inode),
				sp, &res.stateid, arg.seqid);
		if (status != 0)
			goto out_free;
	}
	nfs_confirm_seqid(&sp->so_seqid, 0);
	if (status >= 0) {
		memcpy(state->stateid.data, res.stateid.data,
				sizeof(state->stateid.data));
		clear_bit(NFS_DELEGATED_STATE, &state->flags);
	}
out_free:
	nfs_free_seqid(arg.seqid);
out:
	dput(parent);
	return status;
}

int nfs4_open_delegation_recall(struct dentry *dentry, struct nfs4_state *state)
{
	struct nfs4_exception exception = { };
	struct nfs_server *server = NFS_SERVER(dentry->d_inode);
	int err;
	do {
		err = _nfs4_open_delegation_recall(dentry, state);
		switch (err) {
			case 0:
				return err;
			case -NFS4ERR_STALE_CLIENTID:
			case -NFS4ERR_STALE_STATEID:
			case -NFS4ERR_EXPIRED:
				/* Don't recall a delegation if it was lost */
				nfs4_schedule_state_recovery(server->nfs4_state);
				return err;
		}
		err = nfs4_handle_exception(server, err, &exception);
	} while (exception.retry);
	return err;
}

static int _nfs4_proc_open_confirm(struct rpc_clnt *clnt, const struct nfs_fh *fh, struct nfs4_state_owner *sp, nfs4_stateid *stateid, struct nfs_seqid *seqid)
{
	struct nfs_open_confirmargs arg = {
		.fh             = fh,
		.seqid          = seqid,
		.stateid	= *stateid,
	};
	struct nfs_open_confirmres res;
	struct 	rpc_message msg = {
		.rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
		.rpc_argp       = &arg,
		.rpc_resp       = &res,
		.rpc_cred	= sp->so_cred,
	};
	int status;

	status = rpc_call_sync(clnt, &msg, RPC_TASK_NOINTR);
	/* Confirm the sequence as being established */
	nfs_confirm_seqid(&sp->so_seqid, status);
	nfs_increment_open_seqid(status, seqid);
	if (status >= 0)
		memcpy(stateid, &res.stateid, sizeof(*stateid));
	return status;
}

static int _nfs4_proc_open(struct inode *dir, struct nfs4_state_owner  *sp, struct nfs_openargs *o_arg, struct nfs_openres *o_res)
{
	struct nfs_server *server = NFS_SERVER(dir);
	struct rpc_message msg = {
		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
		.rpc_argp = o_arg,
		.rpc_resp = o_res,
		.rpc_cred = sp->so_cred,
	};
	int status;

	/* Update sequence id. The caller must serialize! */
	o_arg->id = sp->so_id;
	o_arg->clientid = sp->so_client->cl_clientid;

	status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
	if (status == 0) {
		/* OPEN on anything except a regular file is disallowed in NFSv4 */
		switch (o_res->f_attr->mode & S_IFMT) {
			case S_IFREG:
				break;
			case S_IFLNK:
				status = -ELOOP;
				break;
			case S_IFDIR:
				status = -EISDIR;
				break;
			default:
				status = -ENOTDIR;
		}
	}

	nfs_increment_open_seqid(status, o_arg->seqid);
	if (status != 0)
		goto out;
	if (o_arg->open_flags & O_CREAT) {
		update_changeattr(dir, &o_res->cinfo);
		nfs_post_op_update_inode(dir, o_res->dir_attr);
	} else
		nfs_refresh_inode(dir, o_res->dir_attr);
	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
		status = _nfs4_proc_open_confirm(server->client, &o_res->fh,
				sp, &o_res->stateid, o_arg->seqid);
		if (status != 0)
			goto out;
	}
	nfs_confirm_seqid(&sp->so_seqid, 0);
	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR))
		status = server->rpc_ops->getattr(server, &o_res->fh, o_res->f_attr);
out:
	return status;
}

static int _nfs4_do_access(struct inode *inode, struct rpc_cred *cred, int openflags)
{
	struct nfs_access_entry cache;
	int mask = 0;
	int status;

	if (openflags & FMODE_READ)
		mask |= MAY_READ;
	if (openflags & FMODE_WRITE)
		mask |= MAY_WRITE;
	status = nfs_access_get_cached(inode, cred, &cache);
	if (status == 0)
		goto out;

	/* Be clever: ask server to check for all possible rights */
	cache.mask = MAY_EXEC | MAY_WRITE | MAY_READ;
	cache.cred = cred;
	cache.jiffies = jiffies;
	status = _nfs4_proc_access(inode, &cache);
	if (status != 0)
		return status;
	nfs_access_add_cache(inode, &cache);
out:
	if ((cache.mask & mask) == mask)
		return 0;
	return -EACCES;
}

/*
 * OPEN_EXPIRED:
 * 	reclaim state on the server after a network partition.
 * 	Assumes caller holds the appropriate lock
 */
static int _nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry)
{
	struct dentry *parent = dget_parent(dentry);
	struct inode *dir = parent->d_inode;
	struct inode *inode = state->inode;
	struct nfs_server *server = NFS_SERVER(dir);
	struct nfs_delegation *delegation = NFS_I(inode)->delegation;
	struct nfs_fattr f_attr, dir_attr;
	struct nfs_openargs o_arg = {
		.fh = NFS_FH(dir),
		.open_flags = state->state,
		.name = &dentry->d_name,
		.bitmask = server->attr_bitmask,
		.claim = NFS4_OPEN_CLAIM_NULL,
	};
	struct nfs_openres o_res = {
		.f_attr = &f_attr,
		.dir_attr = &dir_attr,
		.server = server,
	};
	int status = 0;

	if (delegation != NULL && !(delegation->flags & NFS_DELEGATION_NEED_RECLAIM)) {
		status = _nfs4_do_access(inode, sp->so_cred, state->state);
		if (status < 0)
			goto out;
		memcpy(&state->stateid, &delegation->stateid, sizeof(state->stateid));
		set_bit(NFS_DELEGATED_STATE, &state->flags);
		goto out;
	}
	o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
	status = -ENOMEM;
	if (o_arg.seqid == NULL)
		goto out;
	nfs_fattr_init(&f_attr);
	nfs_fattr_init(&dir_attr);
	status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
	if (status != 0)
		goto out_nodeleg;
	/* Check if files differ */
	if ((f_attr.mode & S_IFMT) != (inode->i_mode & S_IFMT))
		goto out_stale;
	/* Has the file handle changed? */
	if (nfs_compare_fh(&o_res.fh, NFS_FH(inode)) != 0) {
		/* Verify if the change attributes are the same */
		if (f_attr.change_attr != NFS_I(inode)->change_attr)
			goto out_stale;
		if (nfs_size_to_loff_t(f_attr.size) != inode->i_size)
			goto out_stale;
		/* Lets just pretend that this is the same file */
		nfs_copy_fh(NFS_FH(inode), &o_res.fh);
		NFS_I(inode)->fileid = f_attr.fileid;
	}
	memcpy(&state->stateid, &o_res.stateid, sizeof(state->stateid));
	if (o_res.delegation_type != 0) {
		if (!(delegation->flags & NFS_DELEGATION_NEED_RECLAIM))
			nfs_inode_set_delegation(inode, sp->so_cred, &o_res);
		else
			nfs_inode_reclaim_delegation(inode, sp->so_cred, &o_res);
	}
out_nodeleg:
	nfs_free_seqid(o_arg.seqid);
	clear_bit(NFS_DELEGATED_STATE, &state->flags);
out:
	dput(parent);
	return status;
out_stale:
	status = -ESTALE;
	/* Invalidate the state owner so we don't ever use it again */
	nfs4_drop_state_owner(sp);
	d_drop(dentry);
	/* Should we be trying to close that stateid? */
	goto out_nodeleg;
}

static inline int nfs4_do_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state, struct dentry *dentry)
{
	struct nfs_server *server = NFS_SERVER(dentry->d_inode);
	struct nfs4_exception exception = { };
	int err;

	do {
		err = _nfs4_open_expired(sp, state, dentry);
		if (err == -NFS4ERR_DELAY)
			nfs4_handle_exception(server, err, &exception);
	} while (exception.retry);
	return err;
}

static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
{
	struct nfs_inode *nfsi = NFS_I(state->inode);
	struct nfs_open_context *ctx;
	int status;

	spin_lock(&state->inode->i_lock);
	list_for_each_entry(ctx, &nfsi->open_files, list) {
		if (ctx->state != state)
			continue;
		get_nfs_open_context(ctx);
		spin_unlock(&state->inode->i_lock);
		status = nfs4_do_open_expired(sp, state, ctx->dentry);
		put_nfs_open_context(ctx);
		return status;
	}
	spin_unlock(&state->inode->i_lock);
	return -ENOENT;
}

/*
 * Returns an nfs4_state + an extra reference to the inode
 */
static int _nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred, struct nfs4_state **res)
{
	struct nfs_delegation *delegation;
	struct nfs_server *server = NFS_SERVER(inode);
	struct nfs4_client *clp = server->nfs4_state;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct nfs4_state_owner *sp = NULL;
	struct nfs4_state *state = NULL;
	int open_flags = flags & (FMODE_READ|FMODE_WRITE);
	int err;

	/* Protect against reboot recovery - NOTE ORDER! */
	down_read(&clp->cl_sem);
	/* Protect against delegation recall */
	down_read(&nfsi->rwsem);
	delegation = NFS_I(inode)->delegation;
	err = -ENOENT;
	if (delegation == NULL || (delegation->type & open_flags) != open_flags)
		goto out_err;
	err = -ENOMEM;
	if (!(sp = nfs4_get_state_owner(server, cred))) {
		dprintk("%s: nfs4_get_state_owner failed!\n", __FUNCTION__);
		goto out_err;
	}
	state = nfs4_get_open_state(inode, sp);
	if (state == NULL)
		goto out_err;

	err = -ENOENT;
	if ((state->state & open_flags) == open_flags) {
		spin_lock(&inode->i_lock);
		if (open_flags & FMODE_READ)
			state->nreaders++;
		if (open_flags & FMODE_WRITE)
			state->nwriters++;
		spin_unlock(&inode->i_lock);
		goto out_ok;
	} else if (state->state != 0)
		goto out_err;

	lock_kernel();
	err = _nfs4_do_access(inode, cred, open_flags);
	unlock_kernel();
	if (err != 0)
		goto out_err;
	set_bit(NFS_DELEGATED_STATE, &state->flags);
	update_open_stateid(state, &delegation->stateid, open_flags);
out_ok:
	nfs4_put_state_owner(sp);
	up_read(&nfsi->rwsem);
	up_read(&clp->cl_sem);
	igrab(inode);
	*res = state;
	return 0; 
out_err:
	if (sp != NULL) {
		if (state != NULL)
			nfs4_put_open_state(state);
		nfs4_put_state_owner(sp);
	}
	up_read(&nfsi->rwsem);
	up_read(&clp->cl_sem);
	if (err != -EACCES)
		nfs_inode_return_delegation(inode);
	return err;
}

static struct nfs4_state *nfs4_open_delegated(struct inode *inode, int flags, struct rpc_cred *cred)
{
	struct nfs4_exception exception = { };
	struct nfs4_state *res;
	int err;

	do {
		err = _nfs4_open_delegated(inode, flags, cred, &res);
		if (err == 0)
			break;
		res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(inode),
					err, &exception));
	} while (exception.retry);
	return res;
}

/*
 * Returns an nfs4_state + an referenced inode
 */
static int _nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred, struct nfs4_state **res)
{
	struct nfs4_state_owner  *sp;
	struct nfs4_state     *state = NULL;
	struct nfs_server       *server = NFS_SERVER(dir);
	struct nfs4_client *clp = server->nfs4_state;
	struct inode *inode = NULL;
	int                     status;
	struct nfs_fattr f_attr, dir_attr;
	struct nfs_openargs o_arg = {
		.fh             = NFS_FH(dir),
		.open_flags	= flags,
		.name           = &dentry->d_name,
		.server         = server,
		.bitmask = server->attr_bitmask,
		.claim = NFS4_OPEN_CLAIM_NULL,
	};
	struct nfs_openres o_res = {
		.f_attr         = &f_attr,
		.dir_attr	= &dir_attr,
		.server         = server,
	};

	/* Protect against reboot recovery conflicts */
	down_read(&clp->cl_sem);
	status = -ENOMEM;
	if (!(sp = nfs4_get_state_owner(server, cred))) {
		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
		goto out_err;
	}
	if (flags & O_EXCL) {
		u32 *p = (u32 *) o_arg.u.verifier.data;
		p[0] = jiffies;
		p[1] = current->pid;
	} else
		o_arg.u.attrs = sattr;
	/* Serialization for the sequence id */

	o_arg.seqid = nfs_alloc_seqid(&sp->so_seqid);
	if (o_arg.seqid == NULL)
		return -ENOMEM;
	nfs_fattr_init(&f_attr);
	nfs_fattr_init(&dir_attr);
	status = _nfs4_proc_open(dir, sp, &o_arg, &o_res);
	if (status != 0)
		goto out_err;

	status = -ENOMEM;
	inode = nfs_fhget(dir->i_sb, &o_res.fh, &f_attr);
	if (!inode)
		goto out_err;
	state = nfs4_get_open_state(inode, sp);
	if (!state)
		goto out_err;
	update_open_stateid(state, &o_res.stateid, flags);
	if (o_res.delegation_type != 0)
		nfs_inode_set_delegation(inode, cred, &o_res);
	nfs_free_seqid(o_arg.seqid);
	nfs4_put_state_owner(sp);
	up_read(&clp->cl_sem);
	*res = state;
	return 0;
out_err:
	if (sp != NULL) {
		if (state != NULL)
			nfs4_put_open_state(state);
		nfs_free_seqid(o_arg.seqid);
		nfs4_put_state_owner(sp);
	}
	/* Note: clp->cl_sem must be released before nfs4_put_open_state()! */
	up_read(&clp->cl_sem);
	if (inode != NULL)
		iput(inode);
	*res = NULL;
	return status;
}


static struct nfs4_state *nfs4_do_open(struct inode *dir, struct dentry *dentry, int flags, struct iattr *sattr, struct rpc_cred *cred)
{
	struct nfs4_exception exception = { };
	struct nfs4_state *res;
	int status;

	do {
		status = _nfs4_do_open(dir, dentry, flags, sattr, cred, &res);
		if (status == 0)
			break;
		/* NOTE: BAD_SEQID means the server and client disagree about the
		 * book-keeping w.r.t. state-changing operations
		 * (OPEN/CLOSE/LOCK/LOCKU...)
		 * It is actually a sign of a bug on the client or on the server.
		 *
		 * If we receive a BAD_SEQID error in the particular case of
		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
		 * have unhashed the old state_owner for us, and that we can
		 * therefore safely retry using a new one. We should still warn
		 * the user though...
		 */
		if (status == -NFS4ERR_BAD_SEQID) {
			printk(KERN_WARNING "NFS: v4 server returned a bad sequence-id error!\n");
			exception.retry = 1;
			continue;
		}
		/*
		 * BAD_STATEID on OPEN means that the server cancelled our
		 * state before it received the OPEN_CONFIRM.
		 * Recover by retrying the request as per the discussion
		 * on Page 181 of RFC3530.
		 */
		if (status == -NFS4ERR_BAD_STATEID) {
			exception.retry = 1;
			continue;
		}
		res = ERR_PTR(nfs4_handle_exception(NFS_SERVER(dir),
					status, &exception));
	} while (exception.retry);
	return res;
}

static int _nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
                struct nfs_fh *fhandle, struct iattr *sattr,
                struct nfs4_state *state)
{
        struct nfs_setattrargs  arg = {
                .fh             = fhandle,
                .iap            = sattr,
		.server		= server,
		.bitmask = server->attr_bitmask,
        };
        struct nfs_setattrres  res = {
		.fattr		= fattr,
		.server		= server,
        };
        struct rpc_message msg = {
                .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
                .rpc_argp       = &arg,
                .rpc_resp       = &res,
        };
	int status;

	nfs_fattr_init(fattr);

	if (state != NULL) {
		msg.rpc_cred = state->owner->so_cred;
		nfs4_copy_stateid(&arg.stateid, state, current->files);
	} else
		memcpy(&arg.stateid, &zero_stateid, sizeof(arg.stateid));

	status = rpc_call_sync(server->client, &msg, 0);
	return status;
}

static int nfs4_do_setattr(struct nfs_server *server, struct nfs_fattr *fattr,
                struct nfs_fh *fhandle, struct iattr *sattr,
                struct nfs4_state *state)
{
	struct nfs4_exception exception = { };
	int err;
	do {
		err = nfs4_handle_exception(server,
				_nfs4_do_setattr(server, fattr, fhandle, sattr,
					state),
				&exception);
	} while (exception.retry);
	return err;
}

struct nfs4_closedata {
	struct inode *inode;
	struct nfs4_state *state;
	struct nfs_closeargs arg;
	struct nfs_closeres res;
	struct nfs_fattr fattr;
};

static void nfs4_free_closedata(void *data)
{
	struct nfs4_closedata *calldata = data;
	struct nfs4_state_owner *sp = calldata->state->owner;

	nfs4_put_open_state(calldata->state);
	nfs_free_seqid(calldata->arg.seqid);
	nfs4_put_state_owner(sp);
	kfree(calldata);
}

static void nfs4_close_done(struct rpc_task *task, void *data)
{
	struct nfs4_closedata *calldata = data;
	struct nfs4_state *state = calldata->state;
	struct nfs_server *server = NFS_SERVER(calldata->inode);

	if (RPC_ASSASSINATED(task))
		return;
        /* hmm. we are done with the inode, and in the process of freeing
	 * the state_owner. we keep this around to process errors
	 */
	nfs_increment_open_seqid(task->tk_status, calldata->arg.seqid);
	switch (task->tk_status) {
		case 0:
			memcpy(&state->stateid, &calldata->res.stateid,
					sizeof(state->stateid));
			break;
		case -NFS4ERR_STALE_STATEID:
		case -NFS4ERR_EXPIRED:
			nfs4_schedule_state_recovery(server->nfs4_state);
			break;
		default:
			if (nfs4_async_handle_error(task, server) == -EAGAIN) {
				rpc_restart_call(task);
				return;
			}
	}
	nfs_refresh_inode(calldata->inode, calldata->res.fattr);
}

static void nfs4_close_prepare(struct rpc_task *task, void *data)
{
	struct nfs4_closedata *calldata = data;
	struct nfs4_state *state = calldata->state;
	struct rpc_message msg = {
		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
		.rpc_argp = &calldata->arg,
		.rpc_resp = &calldata->res,
		.rpc_cred = state->owner->so_cred,
	};
	int mode = 0, old_mode;

	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
		return;
	/* Recalculate the new open mode in case someone reopened the file
	 * while we were waiting in line to be scheduled.
	 */
	spin_lock(&state->owner->so_lock);
	spin_lock(&calldata->inode->i_lock);
	mode = old_mode = state->state;
	if (state->nreaders == 0)
		mode &= ~FMODE_READ;
	if (state->nwriters == 0)
		mode &= ~FMODE_WRITE;
	nfs4_state_set_mode_locked(state, mode);
	spin_unlock(&calldata->inode->i_lock);
	spin_unlock(&state->owner->so_lock);
	if (mode == old_mode || test_bit(NFS_DELEGATED_STATE, &state->flags)) {
		/* Note: exit _without_ calling nfs4_close_done */
		task->tk_action = NULL;
		return;
	}
	nfs_fattr_init(calldata->res.fattr);
	if (mode != 0)
		msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
	calldata->arg.open_flags = mode;
	rpc_call_setup(task, &msg, 0);
}

static const struct rpc_call_ops nfs4_close_ops = {
	.rpc_call_prepare = nfs4_close_prepare,
	.rpc_call_done = nfs4_close_done,
	.rpc_release = nfs4_free_closedata,
};

/* 
 * It is possible for data to be read/written from a mem-mapped file 
 * after the sys_close call (which hits the vfs layer as a flush).
 * This means that we can't safely call nfsv4 close on a file until 
 * the inode is cleared. This in turn means that we are not good
 * NFSv4 citizens - we do not indicate to the server to update the file's 
 * share state even when we are done with one of the three share 
 * stateid's in the inode.
 *
 * NOTE: Caller must be holding the sp->so_owner semaphore!
 */
int nfs4_do_close(struct inode *inode, struct nfs4_state *state) 
{
	struct nfs_server *server = NFS_SERVER(inode);
	struct nfs4_closedata *calldata;
	int status = -ENOMEM;

	calldata = kmalloc(sizeof(*calldata), GFP_KERNEL);
	if (calldata == NULL)
		goto out;
	calldata->inode = inode;
	calldata->state = state;
	calldata->arg.fh = NFS_FH(inode);
	calldata->arg.stateid = &state->stateid;
	/* Serialization for the sequence id */
	calldata->arg.seqid = nfs_alloc_seqid(&state->owner->so_seqid);
	if (calldata->arg.seqid == NULL)
		goto out_free_calldata;
	calldata->arg.bitmask = server->attr_bitmask;
	calldata->res.fattr = &calldata->fattr;
	calldata->res.server = server;

	status = nfs4_call_async(server->client, &nfs4_close_ops, calldata);
	if (status == 0)
		goto out;

	nfs_free_seqid(calldata->arg.seqid);