Merge git://git.linux-nfs.org/pub/linux/nfs-2.6 (a942b57b) · Commits · e / devices / android_kernel_fairphone_FP5

fs/lockd/clntproc.c

+17 −9

Original line number	Original line	Diff line number	Diff line
	@@ -454,7 +454,7 @@ static void nlmclnt_locks_init_private(struct file_lock fl, struct nlm_host ho
	fl->fl_ops = &nlmclnt_lock_ops;		fl->fl_ops = &nlmclnt_lock_ops;
	}		}

	static void do_vfs_lock(struct file_lock *fl)		static int do_vfs_lock(struct file_lock *fl)
	{		{
	int res = 0;		int res = 0;
	switch (fl->fl_flags & (FL_POSIX\|FL_FLOCK)) {		switch (fl->fl_flags & (FL_POSIX\|FL_FLOCK)) {
	@@ -467,9 +467,7 @@ static void do_vfs_lock(struct file_lock *fl)
	default:		default:
	BUG();		BUG();
	}		}
	if (res < 0)		return res;
	printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n",
	__FUNCTION__);
	}		}

	/*		/*
	@@ -498,6 +496,7 @@ nlmclnt_lock(struct nlm_rqst req, struct file_lock fl)
	struct nlm_host *host = req->a_host;		struct nlm_host *host = req->a_host;
	struct nlm_res *resp = &req->a_res;		struct nlm_res *resp = &req->a_res;
	struct nlm_wait *block = NULL;		struct nlm_wait *block = NULL;
			unsigned char fl_flags = fl->fl_flags;
	int status = -ENOLCK;		int status = -ENOLCK;

	if (!host->h_monitored && nsm_monitor(host) < 0) {		if (!host->h_monitored && nsm_monitor(host) < 0) {
	@@ -505,6 +504,10 @@ nlmclnt_lock(struct nlm_rqst req, struct file_lock fl)
	host->h_name);		host->h_name);
	goto out;		goto out;
	}		}
			fl->fl_flags \|= FL_ACCESS;
			status = do_vfs_lock(fl);
			if (status < 0)
			goto out;

	block = nlmclnt_prepare_block(host, fl);		block = nlmclnt_prepare_block(host, fl);
	again:		again:
	@@ -539,9 +542,10 @@ nlmclnt_lock(struct nlm_rqst req, struct file_lock fl)
	up_read(&host->h_rwsem);		up_read(&host->h_rwsem);
	goto again;		goto again;
	}		}
	fl->fl_flags \|= FL_SLEEP;
	/* Ensure the resulting lock will get added to granted list */		/* Ensure the resulting lock will get added to granted list */
	do_vfs_lock(fl);		fl->fl_flags = fl_flags \| FL_SLEEP;
			if (do_vfs_lock(fl) < 0)
			printk(KERN_WARNING "%s: VFS is out of sync with lock manager!\n", __FUNCTION__);
	up_read(&host->h_rwsem);		up_read(&host->h_rwsem);
	}		}
	status = nlm_stat_to_errno(resp->status);		status = nlm_stat_to_errno(resp->status);
	@@ -552,6 +556,7 @@ nlmclnt_lock(struct nlm_rqst req, struct file_lock fl)
	nlmclnt_cancel(host, req->a_args.block, fl);		nlmclnt_cancel(host, req->a_args.block, fl);
	out:		out:
	nlm_release_call(req);		nlm_release_call(req);
			fl->fl_flags = fl_flags;
	return status;		return status;
	}		}

	@@ -606,15 +611,19 @@ nlmclnt_unlock(struct nlm_rqst req, struct file_lock fl)
	{		{
	struct nlm_host *host = req->a_host;		struct nlm_host *host = req->a_host;
	struct nlm_res *resp = &req->a_res;		struct nlm_res *resp = &req->a_res;
	int status;		int status = 0;

	/*		/*
	* Note: the server is supposed to either grant us the unlock		* Note: the server is supposed to either grant us the unlock
	* request, or to deny it with NLM_LCK_DENIED_GRACE_PERIOD. In either		* request, or to deny it with NLM_LCK_DENIED_GRACE_PERIOD. In either
	* case, we want to unlock.		* case, we want to unlock.
	*/		*/
			fl->fl_flags \|= FL_EXISTS;
	down_read(&host->h_rwsem);		down_read(&host->h_rwsem);
	do_vfs_lock(fl);		if (do_vfs_lock(fl) == -ENOENT) {
			up_read(&host->h_rwsem);
			goto out;
			}
	up_read(&host->h_rwsem);		up_read(&host->h_rwsem);

	if (req->a_flags & RPC_TASK_ASYNC)		if (req->a_flags & RPC_TASK_ASYNC)
	@@ -624,7 +633,6 @@ nlmclnt_unlock(struct nlm_rqst req, struct file_lock fl)
	if (status < 0)		if (status < 0)
	goto out;		goto out;

	status = 0;
	if (resp->status == NLM_LCK_GRANTED)		if (resp->status == NLM_LCK_GRANTED)
	goto out;		goto out;

fs/locks.c

+21 −2

Original line number	Original line	Diff line number	Diff line
	@@ -725,6 +725,10 @@ static int posix_locks_deadlock(struct file_lock *caller_fl,
	/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks		/* Try to create a FLOCK lock on filp. We always insert new FLOCK locks
	* at the head of the list, but that's secret knowledge known only to		* at the head of the list, but that's secret knowledge known only to
	* flock_lock_file and posix_lock_file.		* flock_lock_file and posix_lock_file.
			*
			* Note that if called with an FL_EXISTS argument, the caller may determine
			* whether or not a lock was successfully freed by testing the return
			* value for -ENOENT.
	*/		*/
	static int flock_lock_file(struct file filp, struct file_lock request)		static int flock_lock_file(struct file filp, struct file_lock request)
	{		{
	@@ -735,6 +739,8 @@ static int flock_lock_file(struct file filp, struct file_lock request)
	int found = 0;		int found = 0;

	lock_kernel();		lock_kernel();
			if (request->fl_flags & FL_ACCESS)
			goto find_conflict;
	for_each_lock(inode, before) {		for_each_lock(inode, before) {
	struct file_lock fl = before;		struct file_lock fl = before;
	if (IS_POSIX(fl))		if (IS_POSIX(fl))
	@@ -750,8 +756,11 @@ static int flock_lock_file(struct file filp, struct file_lock request)
	break;		break;
	}		}

	if (request->fl_type == F_UNLCK)		if (request->fl_type == F_UNLCK) {
			if ((request->fl_flags & FL_EXISTS) && !found)
			error = -ENOENT;
	goto out;		goto out;
			}

	error = -ENOMEM;		error = -ENOMEM;
	new_fl = locks_alloc_lock();		new_fl = locks_alloc_lock();
	@@ -764,6 +773,7 @@ static int flock_lock_file(struct file filp, struct file_lock request)
	if (found)		if (found)
	cond_resched();		cond_resched();

			find_conflict:
	for_each_lock(inode, before) {		for_each_lock(inode, before) {
	struct file_lock fl = before;		struct file_lock fl = before;
	if (IS_POSIX(fl))		if (IS_POSIX(fl))
	@@ -777,6 +787,8 @@ static int flock_lock_file(struct file filp, struct file_lock request)
	locks_insert_block(fl, request);		locks_insert_block(fl, request);
	goto out;		goto out;
	}		}
			if (request->fl_flags & FL_ACCESS)
			goto out;
	locks_copy_lock(new_fl, request);		locks_copy_lock(new_fl, request);
	locks_insert_lock(&inode->i_flock, new_fl);		locks_insert_lock(&inode->i_flock, new_fl);
	new_fl = NULL;		new_fl = NULL;
	@@ -948,8 +960,11 @@ static int __posix_lock_file_conf(struct inode inode, struct file_lock request

	error = 0;		error = 0;
	if (!added) {		if (!added) {
	if (request->fl_type == F_UNLCK)		if (request->fl_type == F_UNLCK) {
			if (request->fl_flags & FL_EXISTS)
			error = -ENOENT;
	goto out;		goto out;
			}

	if (!new_fl) {		if (!new_fl) {
	error = -ENOLCK;		error = -ENOLCK;
	@@ -996,6 +1011,10 @@ static int __posix_lock_file_conf(struct inode inode, struct file_lock request
	* Add a POSIX style lock to a file.		* Add a POSIX style lock to a file.
	* We merge adjacent & overlapping locks whenever possible.		* We merge adjacent & overlapping locks whenever possible.
	* POSIX locks are sorted by owner task, then by starting address		* POSIX locks are sorted by owner task, then by starting address
			*
			* Note that if called with an FL_EXISTS argument, the caller may determine
			* whether or not a lock was successfully freed by testing the return
			* value for -ENOENT.
	*/		*/
	int posix_lock_file(struct file filp, struct file_lock fl)		int posix_lock_file(struct file filp, struct file_lock fl)
	{		{

fs/nfs/dir.c

+3 −1

Original line number	Original line	Diff line number	Diff line
	@@ -690,7 +690,9 @@ int nfs_lookup_verify_inode(struct inode inode, struct nameidata nd)
	goto out_force;		goto out_force;
	/* This is an open(2) */		/* This is an open(2) */
	if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&		if (nfs_lookup_check_intent(nd, LOOKUP_OPEN) != 0 &&
	!(server->flags & NFS_MOUNT_NOCTO))		!(server->flags & NFS_MOUNT_NOCTO) &&
			(S_ISREG(inode->i_mode) \|\|
			S_ISDIR(inode->i_mode)))
	goto out_force;		goto out_force;
	}		}
	return nfs_revalidate_inode(server, inode);		return nfs_revalidate_inode(server, inode);

fs/nfs/direct.c

+201 −234

Original line number	Original line	Diff line number	Diff line
	@@ -67,25 +67,19 @@ struct nfs_direct_req {
	struct kref kref; /* release manager */		struct kref kref; /* release manager */

	/* I/O parameters */		/* I/O parameters */
	struct list_head list, /* nfs_read/write_data structs */
	rewrite_list; /* saved nfs_write_data structs */
	struct nfs_open_context ctx; / file open context info */		struct nfs_open_context ctx; / file open context info */
	struct kiocb * iocb; /* controlling i/o request */		struct kiocb * iocb; /* controlling i/o request */
	struct inode * inode; /* target file of i/o */		struct inode * inode; /* target file of i/o */
	unsigned long user_addr; /* location of user's buffer */
	size_t user_count; /* total bytes to move */
	loff_t pos; /* starting offset in file */
	struct page ** pages; /* pages in our buffer */
	unsigned int npages; /* count of pages */

	/* completion state */		/* completion state */
			atomic_t io_count; /* i/os we're waiting for */
	spinlock_t lock; /* protect completion state */		spinlock_t lock; /* protect completion state */
	int outstanding; /* i/os we're waiting for */
	ssize_t count, /* bytes actually processed */		ssize_t count, /* bytes actually processed */
	error; /* any reported error */		error; /* any reported error */
	struct completion completion; /* wait for i/o completion */		struct completion completion; /* wait for i/o completion */

	/* commit state */		/* commit state */
			struct list_head rewrite_list; /* saved nfs_write_data structs */
	struct nfs_write_data * commit_data; /* special write_data for commits */		struct nfs_write_data * commit_data; /* special write_data for commits */
	int flags;		int flags;
	#define NFS_ODIRECT_DO_COMMIT (1) /* an unstable reply was received */		#define NFS_ODIRECT_DO_COMMIT (1) /* an unstable reply was received */
	@@ -93,8 +87,37 @@ struct nfs_direct_req {
	struct nfs_writeverf verf; /* unstable write verifier */		struct nfs_writeverf verf; /* unstable write verifier */
	};		};

	static void nfs_direct_write_schedule(struct nfs_direct_req *dreq, int sync);
	static void nfs_direct_write_complete(struct nfs_direct_req dreq, struct inode inode);		static void nfs_direct_write_complete(struct nfs_direct_req dreq, struct inode inode);
			static const struct rpc_call_ops nfs_write_direct_ops;

			static inline void get_dreq(struct nfs_direct_req *dreq)
			{
			atomic_inc(&dreq->io_count);
			}

			static inline int put_dreq(struct nfs_direct_req *dreq)
			{
			return atomic_dec_and_test(&dreq->io_count);
			}

			/*
			* "size" is never larger than rsize or wsize.
			*/
			static inline int nfs_direct_count_pages(unsigned long user_addr, size_t size)
			{
			int page_count;

			page_count = (user_addr + size + PAGE_SIZE - 1) >> PAGE_SHIFT;
			page_count -= user_addr >> PAGE_SHIFT;
			BUG_ON(page_count < 0);

			return page_count;
			}

			static inline unsigned int nfs_max_pages(unsigned int size)
			{
			return (size + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
			}

	/**		/**
	* nfs_direct_IO - NFS address space operation for direct I/O		* nfs_direct_IO - NFS address space operation for direct I/O
	@@ -118,50 +141,21 @@ ssize_t nfs_direct_IO(int rw, struct kiocb iocb, const struct iovec iov, loff_
	return -EINVAL;		return -EINVAL;
	}		}

	static void nfs_free_user_pages(struct page **pages, int npages, int do_dirty)		static void nfs_direct_dirty_pages(struct page **pages, int npages)
	{		{
	int i;		int i;
	for (i = 0; i < npages; i++) {		for (i = 0; i < npages; i++) {
	struct page *page = pages[i];		struct page *page = pages[i];
	if (do_dirty && !PageCompound(page))		if (!PageCompound(page))
	set_page_dirty_lock(page);		set_page_dirty_lock(page);
	page_cache_release(page);
	}		}
	kfree(pages);
	}		}

	static inline int nfs_get_user_pages(int rw, unsigned long user_addr, size_t size, struct page ***pages)		static void nfs_direct_release_pages(struct page **pages, int npages)
	{		{
	int result = -ENOMEM;		int i;
	unsigned long page_count;		for (i = 0; i < npages; i++)
	size_t array_size;		page_cache_release(pages[i]);

	page_count = (user_addr + size + PAGE_SIZE - 1) >> PAGE_SHIFT;
	page_count -= user_addr >> PAGE_SHIFT;

	array_size = (page_count * sizeof(struct page *));
	*pages = kmalloc(array_size, GFP_KERNEL);
	if (*pages) {
	down_read(&current->mm->mmap_sem);
	result = get_user_pages(current, current->mm, user_addr,
	page_count, (rw == READ), 0,
	*pages, NULL);
	up_read(&current->mm->mmap_sem);
	if (result != page_count) {
	/*
	* If we got fewer pages than expected from
	* get_user_pages(), the user buffer runs off the
	* end of a mapping; return EFAULT.
	*/
	if (result >= 0) {
	nfs_free_user_pages(*pages, result, 0);
	result = -EFAULT;
	} else
	kfree(*pages);
	*pages = NULL;
	}
	}
	return result;
	}		}

	static inline struct nfs_direct_req *nfs_direct_req_alloc(void)		static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
	@@ -173,13 +167,13 @@ static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
	return NULL;		return NULL;

	kref_init(&dreq->kref);		kref_init(&dreq->kref);
			kref_get(&dreq->kref);
	init_completion(&dreq->completion);		init_completion(&dreq->completion);
	INIT_LIST_HEAD(&dreq->list);
	INIT_LIST_HEAD(&dreq->rewrite_list);		INIT_LIST_HEAD(&dreq->rewrite_list);
	dreq->iocb = NULL;		dreq->iocb = NULL;
	dreq->ctx = NULL;		dreq->ctx = NULL;
	spin_lock_init(&dreq->lock);		spin_lock_init(&dreq->lock);
	dreq->outstanding = 0;		atomic_set(&dreq->io_count, 0);
	dreq->count = 0;		dreq->count = 0;
	dreq->error = 0;		dreq->error = 0;
	dreq->flags = 0;		dreq->flags = 0;
	@@ -220,18 +214,11 @@ static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
	}		}

	/*		/*
	* We must hold a reference to all the pages in this direct read request		* Synchronous I/O uses a stack-allocated iocb. Thus we can't trust
	* until the RPCs complete. This could be long after we are woken up in		* the iocb is still valid here if this is a synchronous request.
	* nfs_direct_wait (for instance, if someone hits ^C on a slow server).
	*
	* In addition, synchronous I/O uses a stack-allocated iocb. Thus we
	* can't trust the iocb is still valid here if this is a synchronous
	* request. If the waiter is woken prematurely, the iocb is long gone.
	*/		*/
	static void nfs_direct_complete(struct nfs_direct_req *dreq)		static void nfs_direct_complete(struct nfs_direct_req *dreq)
	{		{
	nfs_free_user_pages(dreq->pages, dreq->npages, 1);

	if (dreq->iocb) {		if (dreq->iocb) {
	long res = (long) dreq->error;		long res = (long) dreq->error;
	if (!res)		if (!res)
	@@ -244,48 +231,10 @@ static void nfs_direct_complete(struct nfs_direct_req *dreq)
	}		}

	/*		/*
	* Note we also set the number of requests we have in the dreq when we are		* We must hold a reference to all the pages in this direct read request
	* done. This prevents races with I/O completion so we will always wait		* until the RPCs complete. This could be long after we are woken up in
	* until all requests have been dispatched and completed.		* nfs_direct_wait (for instance, if someone hits ^C on a slow server).
	*/		*/
	static struct nfs_direct_req *nfs_direct_read_alloc(size_t nbytes, size_t rsize)
	{
	struct list_head *list;
	struct nfs_direct_req *dreq;
	unsigned int rpages = (rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;

	dreq = nfs_direct_req_alloc();
	if (!dreq)
	return NULL;

	list = &dreq->list;
	for(;;) {
	struct nfs_read_data *data = nfs_readdata_alloc(rpages);

	if (unlikely(!data)) {
	while (!list_empty(list)) {
	data = list_entry(list->next,
	struct nfs_read_data, pages);
	list_del(&data->pages);
	nfs_readdata_free(data);
	}
	kref_put(&dreq->kref, nfs_direct_req_release);
	return NULL;
	}

	INIT_LIST_HEAD(&data->pages);
	list_add(&data->pages, list);

	data->req = (struct nfs_page *) dreq;
	dreq->outstanding++;
	if (nbytes <= rsize)
	break;
	nbytes -= rsize;
	}
	kref_get(&dreq->kref);
	return dreq;
	}

	static void nfs_direct_read_result(struct rpc_task task, void calldata)		static void nfs_direct_read_result(struct rpc_task task, void calldata)
	{		{
	struct nfs_read_data *data = calldata;		struct nfs_read_data *data = calldata;
	@@ -294,6 +243,9 @@ static void nfs_direct_read_result(struct rpc_task task, void calldata)
	if (nfs_readpage_result(task, data) != 0)		if (nfs_readpage_result(task, data) != 0)
	return;		return;

			nfs_direct_dirty_pages(data->pagevec, data->npages);
			nfs_direct_release_pages(data->pagevec, data->npages);

	spin_lock(&dreq->lock);		spin_lock(&dreq->lock);

	if (likely(task->tk_status >= 0))		if (likely(task->tk_status >= 0))
	@@ -301,12 +253,9 @@ static void nfs_direct_read_result(struct rpc_task task, void calldata)
	else		else
	dreq->error = task->tk_status;		dreq->error = task->tk_status;

	if (--dreq->outstanding) {
	spin_unlock(&dreq->lock);		spin_unlock(&dreq->lock);
	return;
	}

	spin_unlock(&dreq->lock);		if (put_dreq(dreq))
	nfs_direct_complete(dreq);		nfs_direct_complete(dreq);
	}		}

	@@ -316,41 +265,60 @@ static const struct rpc_call_ops nfs_read_direct_ops = {
	};		};

	/*		/*
	* For each nfs_read_data struct that was allocated on the list, dispatch		* For each rsize'd chunk of the user's buffer, dispatch an NFS READ
	* an NFS READ operation		* operation. If nfs_readdata_alloc() or get_user_pages() fails,
			* bail and stop sending more reads. Read length accounting is
			* handled automatically by nfs_direct_read_result(). Otherwise, if
			* no requests have been sent, just return an error.
	*/		*/
	static void nfs_direct_read_schedule(struct nfs_direct_req *dreq)		static ssize_t nfs_direct_read_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos)
	{		{
	struct nfs_open_context *ctx = dreq->ctx;		struct nfs_open_context *ctx = dreq->ctx;
	struct inode *inode = ctx->dentry->d_inode;		struct inode *inode = ctx->dentry->d_inode;
	struct list_head *list = &dreq->list;
	struct page **pages = dreq->pages;
	size_t count = dreq->user_count;
	loff_t pos = dreq->pos;
	size_t rsize = NFS_SERVER(inode)->rsize;		size_t rsize = NFS_SERVER(inode)->rsize;
	unsigned int curpage, pgbase;		unsigned int rpages = nfs_max_pages(rsize);
			unsigned int pgbase;
			int result;
			ssize_t started = 0;

	curpage = 0;		get_dreq(dreq);
	pgbase = dreq->user_addr & ~PAGE_MASK;
			pgbase = user_addr & ~PAGE_MASK;
	do {		do {
	struct nfs_read_data *data;		struct nfs_read_data *data;
	size_t bytes;		size_t bytes;

			result = -ENOMEM;
			data = nfs_readdata_alloc(rpages);
			if (unlikely(!data))
			break;

	bytes = rsize;		bytes = rsize;
	if (count < rsize)		if (count < rsize)
	bytes = count;		bytes = count;

	BUG_ON(list_empty(list));		data->npages = nfs_direct_count_pages(user_addr, bytes);
	data = list_entry(list->next, struct nfs_read_data, pages);		down_read(&current->mm->mmap_sem);
	list_del_init(&data->pages);		result = get_user_pages(current, current->mm, user_addr,
			data->npages, 1, 0, data->pagevec, NULL);
			up_read(&current->mm->mmap_sem);
			if (unlikely(result < data->npages)) {
			if (result > 0)
			nfs_direct_release_pages(data->pagevec, result);
			nfs_readdata_release(data);
			break;
			}

			get_dreq(dreq);

			data->req = (struct nfs_page *) dreq;
	data->inode = inode;		data->inode = inode;
	data->cred = ctx->cred;		data->cred = ctx->cred;
	data->args.fh = NFS_FH(inode);		data->args.fh = NFS_FH(inode);
	data->args.context = ctx;		data->args.context = ctx;
	data->args.offset = pos;		data->args.offset = pos;
	data->args.pgbase = pgbase;		data->args.pgbase = pgbase;
	data->args.pages = &pages[curpage];		data->args.pages = data->pagevec;
	data->args.count = bytes;		data->args.count = bytes;
	data->res.fattr = &data->fattr;		data->res.fattr = &data->fattr;
	data->res.eof = 0;		data->res.eof = 0;
	@@ -373,33 +341,35 @@ static void nfs_direct_read_schedule(struct nfs_direct_req *dreq)
	bytes,		bytes,
	(unsigned long long)data->args.offset);		(unsigned long long)data->args.offset);

			started += bytes;
			user_addr += bytes;
	pos += bytes;		pos += bytes;
	pgbase += bytes;		pgbase += bytes;
	curpage += pgbase >> PAGE_SHIFT;
	pgbase &= ~PAGE_MASK;		pgbase &= ~PAGE_MASK;

	count -= bytes;		count -= bytes;
	} while (count != 0);		} while (count != 0);
	BUG_ON(!list_empty(list));
			if (put_dreq(dreq))
			nfs_direct_complete(dreq);

			if (started)
			return 0;
			return result < 0 ? (ssize_t) result : -EFAULT;
	}		}

	static ssize_t nfs_direct_read(struct kiocb iocb, unsigned long user_addr, size_t count, loff_t pos, struct page *pages, unsigned int nr_pages)		static ssize_t nfs_direct_read(struct kiocb *iocb, unsigned long user_addr, size_t count, loff_t pos)
	{		{
	ssize_t result;		ssize_t result = 0;
	sigset_t oldset;		sigset_t oldset;
	struct inode *inode = iocb->ki_filp->f_mapping->host;		struct inode *inode = iocb->ki_filp->f_mapping->host;
	struct rpc_clnt *clnt = NFS_CLIENT(inode);		struct rpc_clnt *clnt = NFS_CLIENT(inode);
	struct nfs_direct_req *dreq;		struct nfs_direct_req *dreq;

	dreq = nfs_direct_read_alloc(count, NFS_SERVER(inode)->rsize);		dreq = nfs_direct_req_alloc();
	if (!dreq)		if (!dreq)
	return -ENOMEM;		return -ENOMEM;

	dreq->user_addr = user_addr;
	dreq->user_count = count;
	dreq->pos = pos;
	dreq->pages = pages;
	dreq->npages = nr_pages;
	dreq->inode = inode;		dreq->inode = inode;
	dreq->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);		dreq->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);
	if (!is_sync_kiocb(iocb))		if (!is_sync_kiocb(iocb))
	@@ -407,7 +377,8 @@ static ssize_t nfs_direct_read(struct kiocb *iocb, unsigned long user_addr, size

	nfs_add_stats(inode, NFSIOS_DIRECTREADBYTES, count);		nfs_add_stats(inode, NFSIOS_DIRECTREADBYTES, count);
	rpc_clnt_sigmask(clnt, &oldset);		rpc_clnt_sigmask(clnt, &oldset);
	nfs_direct_read_schedule(dreq);		result = nfs_direct_read_schedule(dreq, user_addr, count, pos);
			if (!result)
	result = nfs_direct_wait(dreq);		result = nfs_direct_wait(dreq);
	rpc_clnt_sigunmask(clnt, &oldset);		rpc_clnt_sigunmask(clnt, &oldset);

	@@ -416,10 +387,10 @@ static ssize_t nfs_direct_read(struct kiocb *iocb, unsigned long user_addr, size

	static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)		static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)
	{		{
	list_splice_init(&dreq->rewrite_list, &dreq->list);		while (!list_empty(&dreq->rewrite_list)) {
	while (!list_empty(&dreq->list)) {		struct nfs_write_data *data = list_entry(dreq->rewrite_list.next, struct nfs_write_data, pages);
	struct nfs_write_data *data = list_entry(dreq->list.next, struct nfs_write_data, pages);
	list_del(&data->pages);		list_del(&data->pages);
			nfs_direct_release_pages(data->pagevec, data->npages);
	nfs_writedata_release(data);		nfs_writedata_release(data);
	}		}
	}		}
	@@ -427,14 +398,51 @@ static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)
	#if defined(CONFIG_NFS_V3) \|\| defined(CONFIG_NFS_V4)		#if defined(CONFIG_NFS_V3) \|\| defined(CONFIG_NFS_V4)
	static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)		static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
	{		{
	struct list_head *pos;		struct inode *inode = dreq->inode;
			struct list_head *p;
			struct nfs_write_data *data;

	list_splice_init(&dreq->rewrite_list, &dreq->list);
	list_for_each(pos, &dreq->list)
	dreq->outstanding++;
	dreq->count = 0;		dreq->count = 0;
			get_dreq(dreq);

			list_for_each(p, &dreq->rewrite_list) {
			data = list_entry(p, struct nfs_write_data, pages);

	nfs_direct_write_schedule(dreq, FLUSH_STABLE);		get_dreq(dreq);

			/*
			* Reset data->res.
			*/
			nfs_fattr_init(&data->fattr);
			data->res.count = data->args.count;
			memset(&data->verf, 0, sizeof(data->verf));

			/*
			* Reuse data->task; data->args should not have changed
			* since the original request was sent.
			*/
			rpc_init_task(&data->task, NFS_CLIENT(inode), RPC_TASK_ASYNC,
			&nfs_write_direct_ops, data);
			NFS_PROTO(inode)->write_setup(data, FLUSH_STABLE);

			data->task.tk_priority = RPC_PRIORITY_NORMAL;
			data->task.tk_cookie = (unsigned long) inode;

			/*
			* We're called via an RPC callback, so BKL is already held.
			*/
			rpc_execute(&data->task);

			dprintk("NFS: %5u rescheduled direct write call (req %s/%Ld, %u bytes @ offset %Lu)\n",
			data->task.tk_pid,
			inode->i_sb->s_id,
			(long long)NFS_FILEID(inode),
			data->args.count,
			(unsigned long long)data->args.offset);
			}

			if (put_dreq(dreq))
			nfs_direct_write_complete(dreq, inode);
	}		}

	static void nfs_direct_commit_result(struct rpc_task task, void calldata)		static void nfs_direct_commit_result(struct rpc_task task, void calldata)
	@@ -471,8 +479,8 @@ static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
	data->cred = dreq->ctx->cred;		data->cred = dreq->ctx->cred;

	data->args.fh = NFS_FH(data->inode);		data->args.fh = NFS_FH(data->inode);
	data->args.offset = dreq->pos;		data->args.offset = 0;
	data->args.count = dreq->user_count;		data->args.count = 0;
	data->res.count = 0;		data->res.count = 0;
	data->res.fattr = &data->fattr;		data->res.fattr = &data->fattr;
	data->res.verf = &data->verf;		data->res.verf = &data->verf;
	@@ -534,47 +542,6 @@ static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode
	}		}
	#endif		#endif

	static struct nfs_direct_req *nfs_direct_write_alloc(size_t nbytes, size_t wsize)
	{
	struct list_head *list;
	struct nfs_direct_req *dreq;
	unsigned int wpages = (wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;

	dreq = nfs_direct_req_alloc();
	if (!dreq)
	return NULL;

	list = &dreq->list;
	for(;;) {
	struct nfs_write_data *data = nfs_writedata_alloc(wpages);

	if (unlikely(!data)) {
	while (!list_empty(list)) {
	data = list_entry(list->next,
	struct nfs_write_data, pages);
	list_del(&data->pages);
	nfs_writedata_free(data);
	}
	kref_put(&dreq->kref, nfs_direct_req_release);
	return NULL;
	}

	INIT_LIST_HEAD(&data->pages);
	list_add(&data->pages, list);

	data->req = (struct nfs_page *) dreq;
	dreq->outstanding++;
	if (nbytes <= wsize)
	break;
	nbytes -= wsize;
	}

	nfs_alloc_commit_data(dreq);

	kref_get(&dreq->kref);
	return dreq;
	}

	static void nfs_direct_write_result(struct rpc_task task, void calldata)		static void nfs_direct_write_result(struct rpc_task task, void calldata)
	{		{
	struct nfs_write_data *data = calldata;		struct nfs_write_data *data = calldata;
	@@ -604,8 +571,6 @@ static void nfs_direct_write_result(struct rpc_task task, void calldata)
	}		}
	}		}
	}		}
	/* In case we have to resend */
	data->args.stable = NFS_FILE_SYNC;

	spin_unlock(&dreq->lock);		spin_unlock(&dreq->lock);
	}		}
	@@ -619,13 +584,7 @@ static void nfs_direct_write_release(void *calldata)
	struct nfs_write_data *data = calldata;		struct nfs_write_data *data = calldata;
	struct nfs_direct_req dreq = (struct nfs_direct_req ) data->req;		struct nfs_direct_req dreq = (struct nfs_direct_req ) data->req;

	spin_lock(&dreq->lock);		if (put_dreq(dreq))
	if (--dreq->outstanding) {
	spin_unlock(&dreq->lock);
	return;
	}
	spin_unlock(&dreq->lock);

	nfs_direct_write_complete(dreq, data->inode);		nfs_direct_write_complete(dreq, data->inode);
	}		}

	@@ -635,41 +594,62 @@ static const struct rpc_call_ops nfs_write_direct_ops = {
	};		};

	/*		/*
	* For each nfs_write_data struct that was allocated on the list, dispatch		* For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
	* an NFS WRITE operation		* operation. If nfs_writedata_alloc() or get_user_pages() fails,
			* bail and stop sending more writes. Write length accounting is
			* handled automatically by nfs_direct_write_result(). Otherwise, if
			* no requests have been sent, just return an error.
	*/		*/
	static void nfs_direct_write_schedule(struct nfs_direct_req *dreq, int sync)		static ssize_t nfs_direct_write_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos, int sync)
	{		{
	struct nfs_open_context *ctx = dreq->ctx;		struct nfs_open_context *ctx = dreq->ctx;
	struct inode *inode = ctx->dentry->d_inode;		struct inode *inode = ctx->dentry->d_inode;
	struct list_head *list = &dreq->list;
	struct page **pages = dreq->pages;
	size_t count = dreq->user_count;
	loff_t pos = dreq->pos;
	size_t wsize = NFS_SERVER(inode)->wsize;		size_t wsize = NFS_SERVER(inode)->wsize;
	unsigned int curpage, pgbase;		unsigned int wpages = nfs_max_pages(wsize);
			unsigned int pgbase;
			int result;
			ssize_t started = 0;

	curpage = 0;		get_dreq(dreq);
	pgbase = dreq->user_addr & ~PAGE_MASK;
			pgbase = user_addr & ~PAGE_MASK;
	do {		do {
	struct nfs_write_data *data;		struct nfs_write_data *data;
	size_t bytes;		size_t bytes;

			result = -ENOMEM;
			data = nfs_writedata_alloc(wpages);
			if (unlikely(!data))
			break;

	bytes = wsize;		bytes = wsize;
	if (count < wsize)		if (count < wsize)
	bytes = count;		bytes = count;

	BUG_ON(list_empty(list));		data->npages = nfs_direct_count_pages(user_addr, bytes);
	data = list_entry(list->next, struct nfs_write_data, pages);		down_read(&current->mm->mmap_sem);
			result = get_user_pages(current, current->mm, user_addr,
			data->npages, 0, 0, data->pagevec, NULL);
			up_read(&current->mm->mmap_sem);
			if (unlikely(result < data->npages)) {
			if (result > 0)
			nfs_direct_release_pages(data->pagevec, result);
			nfs_writedata_release(data);
			break;
			}

			get_dreq(dreq);

	list_move_tail(&data->pages, &dreq->rewrite_list);		list_move_tail(&data->pages, &dreq->rewrite_list);

			data->req = (struct nfs_page *) dreq;
	data->inode = inode;		data->inode = inode;
	data->cred = ctx->cred;		data->cred = ctx->cred;
	data->args.fh = NFS_FH(inode);		data->args.fh = NFS_FH(inode);
	data->args.context = ctx;		data->args.context = ctx;
	data->args.offset = pos;		data->args.offset = pos;
	data->args.pgbase = pgbase;		data->args.pgbase = pgbase;
	data->args.pages = &pages[curpage];		data->args.pages = data->pagevec;
	data->args.count = bytes;		data->args.count = bytes;
	data->res.fattr = &data->fattr;		data->res.fattr = &data->fattr;
	data->res.count = bytes;		data->res.count = bytes;
	@@ -693,19 +673,26 @@ static void nfs_direct_write_schedule(struct nfs_direct_req *dreq, int sync)
	bytes,		bytes,
	(unsigned long long)data->args.offset);		(unsigned long long)data->args.offset);

			started += bytes;
			user_addr += bytes;
	pos += bytes;		pos += bytes;
	pgbase += bytes;		pgbase += bytes;
	curpage += pgbase >> PAGE_SHIFT;
	pgbase &= ~PAGE_MASK;		pgbase &= ~PAGE_MASK;

	count -= bytes;		count -= bytes;
	} while (count != 0);		} while (count != 0);
	BUG_ON(!list_empty(list));
			if (put_dreq(dreq))
			nfs_direct_write_complete(dreq, inode);

			if (started)
			return 0;
			return result < 0 ? (ssize_t) result : -EFAULT;
	}		}

	static ssize_t nfs_direct_write(struct kiocb iocb, unsigned long user_addr, size_t count, loff_t pos, struct page *pages, int nr_pages)		static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, size_t count, loff_t pos)
	{		{
	ssize_t result;		ssize_t result = 0;
	sigset_t oldset;		sigset_t oldset;
	struct inode *inode = iocb->ki_filp->f_mapping->host;		struct inode *inode = iocb->ki_filp->f_mapping->host;
	struct rpc_clnt *clnt = NFS_CLIENT(inode);		struct rpc_clnt *clnt = NFS_CLIENT(inode);
	@@ -713,17 +700,14 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, siz
	size_t wsize = NFS_SERVER(inode)->wsize;		size_t wsize = NFS_SERVER(inode)->wsize;
	int sync = 0;		int sync = 0;

	dreq = nfs_direct_write_alloc(count, wsize);		dreq = nfs_direct_req_alloc();
	if (!dreq)		if (!dreq)
	return -ENOMEM;		return -ENOMEM;
			nfs_alloc_commit_data(dreq);

	if (dreq->commit_data == NULL \|\| count < wsize)		if (dreq->commit_data == NULL \|\| count < wsize)
	sync = FLUSH_STABLE;		sync = FLUSH_STABLE;

	dreq->user_addr = user_addr;
	dreq->user_count = count;
	dreq->pos = pos;
	dreq->pages = pages;
	dreq->npages = nr_pages;
	dreq->inode = inode;		dreq->inode = inode;
	dreq->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);		dreq->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);
	if (!is_sync_kiocb(iocb))		if (!is_sync_kiocb(iocb))
	@@ -734,7 +718,8 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, siz
	nfs_begin_data_update(inode);		nfs_begin_data_update(inode);

	rpc_clnt_sigmask(clnt, &oldset);		rpc_clnt_sigmask(clnt, &oldset);
	nfs_direct_write_schedule(dreq, sync);		result = nfs_direct_write_schedule(dreq, user_addr, count, pos, sync);
			if (!result)
	result = nfs_direct_wait(dreq);		result = nfs_direct_wait(dreq);
	rpc_clnt_sigunmask(clnt, &oldset);		rpc_clnt_sigunmask(clnt, &oldset);

	@@ -765,8 +750,6 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, siz
	ssize_t nfs_file_direct_read(struct kiocb iocb, char __user buf, size_t count, loff_t pos)		ssize_t nfs_file_direct_read(struct kiocb iocb, char __user buf, size_t count, loff_t pos)
	{		{
	ssize_t retval = -EINVAL;		ssize_t retval = -EINVAL;
	int page_count;
	struct page **pages;
	struct file *file = iocb->ki_filp;		struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;		struct address_space *mapping = file->f_mapping;

	@@ -788,14 +771,7 @@ ssize_t nfs_file_direct_read(struct kiocb iocb, char __user buf, size_t count,
	if (retval)		if (retval)
	goto out;		goto out;

	retval = nfs_get_user_pages(READ, (unsigned long) buf,		retval = nfs_direct_read(iocb, (unsigned long) buf, count, pos);
	count, &pages);
	if (retval < 0)
	goto out;
	page_count = retval;

	retval = nfs_direct_read(iocb, (unsigned long) buf, count, pos,
	pages, page_count);
	if (retval > 0)		if (retval > 0)
	iocb->ki_pos = pos + retval;		iocb->ki_pos = pos + retval;

	@@ -831,8 +807,6 @@ ssize_t nfs_file_direct_read(struct kiocb iocb, char __user buf, size_t count,
	ssize_t nfs_file_direct_write(struct kiocb iocb, const char __user buf, size_t count, loff_t pos)		ssize_t nfs_file_direct_write(struct kiocb iocb, const char __user buf, size_t count, loff_t pos)
	{		{
	ssize_t retval;		ssize_t retval;
	int page_count;
	struct page **pages;
	struct file *file = iocb->ki_filp;		struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;		struct address_space *mapping = file->f_mapping;

	@@ -860,14 +834,7 @@ ssize_t nfs_file_direct_write(struct kiocb iocb, const char __user buf, size_t
	if (retval)		if (retval)
	goto out;		goto out;

	retval = nfs_get_user_pages(WRITE, (unsigned long) buf,		retval = nfs_direct_write(iocb, (unsigned long) buf, count, pos);
	count, &pages);
	if (retval < 0)
	goto out;
	page_count = retval;

	retval = nfs_direct_write(iocb, (unsigned long) buf, count,
	pos, pages, page_count);

	/*		/*
	* XXX: nfs_end_data_update() already ensures this file's		* XXX: nfs_end_data_update() already ensures this file's