1 files changed, 611 insertions, 619 deletions

diff --git a/fs/nfs/direct.c b/fs/nfs/direct.c
index 84d3c8b9020..f11b9eed0de 100644
--- a/fs/nfs/direct.c
+++ b/fs/nfs/direct.c
@@ -45,17 +45,19 @@
 #include <linux/pagemap.h>
 #include <linux/kref.h>
 #include <linux/slab.h>
+#include <linux/task_io_accounting_ops.h>
+#include <linux/module.h>
 
 #include <linux/nfs_fs.h>
 #include <linux/nfs_page.h>
 #include <linux/sunrpc/clnt.h>
 
-#include <asm/system.h>
 #include <asm/uaccess.h>
-#include <asm/atomic.h>
+#include <linux/atomic.h>
 
 #include "internal.h"
 #include "iostat.h"
+#include "pnfs.h"
 
 #define NFSDBG_FACILITY		NFSDBG_VFS
 
@@ -77,20 +79,24 @@ struct nfs_direct_req {
 	atomic_t		io_count;	/* i/os we're waiting for */
 	spinlock_t		lock;		/* protect completion state */
 	ssize_t			count,		/* bytes actually processed */
+				bytes_left,	/* bytes left to be sent */
 				error;		/* any reported error */
 	struct completion	completion;	/* wait for i/o completion */
 
 	/* 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_mds_commit_info mds_cinfo;	/* Storage for cinfo */
+	struct pnfs_ds_commit_info ds_cinfo;	/* Storage for cinfo */
+	struct work_struct	work;
 	int			flags;
 #define NFS_ODIRECT_DO_COMMIT		(1)	/* an unstable reply was received */
 #define NFS_ODIRECT_RESCHED_WRITES	(2)	/* write verification failed */
 	struct nfs_writeverf	verf;		/* unstable write verifier */
 };
 
+static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops;
+static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops;
 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 void nfs_direct_write_schedule_work(struct work_struct *work);
 
 static inline void get_dreq(struct nfs_direct_req *dreq)
 {
@@ -102,6 +108,97 @@ static inline int put_dreq(struct nfs_direct_req *dreq)
 	return atomic_dec_and_test(&dreq->io_count);
 }
 
+/*
+ * nfs_direct_select_verf - select the right verifier
+ * @dreq - direct request possibly spanning multiple servers
+ * @ds_clp - nfs_client of data server or NULL if MDS / non-pnfs
+ * @ds_idx - index of data server in data server list, only valid if ds_clp set
+ *
+ * returns the correct verifier to use given the role of the server
+ */
+static struct nfs_writeverf *
+nfs_direct_select_verf(struct nfs_direct_req *dreq,
+		       struct nfs_client *ds_clp,
+		       int ds_idx)
+{
+	struct nfs_writeverf *verfp = &dreq->verf;
+
+#ifdef CONFIG_NFS_V4_1
+	if (ds_clp) {
+		/* pNFS is in use, use the DS verf */
+		if (ds_idx >= 0 && ds_idx < dreq->ds_cinfo.nbuckets)
+			verfp = &dreq->ds_cinfo.buckets[ds_idx].direct_verf;
+		else
+			WARN_ON_ONCE(1);
+	}
+#endif
+	return verfp;
+}
+
+
+/*
+ * nfs_direct_set_hdr_verf - set the write/commit verifier
+ * @dreq - direct request possibly spanning multiple servers
+ * @hdr - pageio header to validate against previously seen verfs
+ *
+ * Set the server's (MDS or DS) "seen" verifier
+ */
+static void nfs_direct_set_hdr_verf(struct nfs_direct_req *dreq,
+				    struct nfs_pgio_header *hdr)
+{
+	struct nfs_writeverf *verfp;
+
+	verfp = nfs_direct_select_verf(dreq, hdr->data->ds_clp,
+				      hdr->data->ds_idx);
+	WARN_ON_ONCE(verfp->committed >= 0);
+	memcpy(verfp, &hdr->verf, sizeof(struct nfs_writeverf));
+	WARN_ON_ONCE(verfp->committed < 0);
+}
+
+/*
+ * nfs_direct_cmp_hdr_verf - compare verifier for pgio header
+ * @dreq - direct request possibly spanning multiple servers
+ * @hdr - pageio header to validate against previously seen verf
+ *
+ * set the server's "seen" verf if not initialized.
+ * returns result of comparison between @hdr->verf and the "seen"
+ * verf of the server used by @hdr (DS or MDS)
+ */
+static int nfs_direct_set_or_cmp_hdr_verf(struct nfs_direct_req *dreq,
+					  struct nfs_pgio_header *hdr)
+{
+	struct nfs_writeverf *verfp;
+
+	verfp = nfs_direct_select_verf(dreq, hdr->data->ds_clp,
+					 hdr->data->ds_idx);
+	if (verfp->committed < 0) {
+		nfs_direct_set_hdr_verf(dreq, hdr);
+		return 0;
+	}
+	return memcmp(verfp, &hdr->verf, sizeof(struct nfs_writeverf));
+}
+
+#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
+/*
+ * nfs_direct_cmp_commit_data_verf - compare verifier for commit data
+ * @dreq - direct request possibly spanning multiple servers
+ * @data - commit data to validate against previously seen verf
+ *
+ * returns result of comparison between @data->verf and the verf of
+ * the server used by @data (DS or MDS)
+ */
+static int nfs_direct_cmp_commit_data_verf(struct nfs_direct_req *dreq,
+					   struct nfs_commit_data *data)
+{
+	struct nfs_writeverf *verfp;
+
+	verfp = nfs_direct_select_verf(dreq, data->ds_clp,
+					 data->ds_commit_index);
+	WARN_ON_ONCE(verfp->committed < 0);
+	return memcmp(verfp, &data->verf, sizeof(struct nfs_writeverf));
+}
+#endif
+
 /**
  * nfs_direct_IO - NFS address space operation for direct I/O
  * @rw: direction (read or write)
@@ -111,33 +208,26 @@ static inline int put_dreq(struct nfs_direct_req *dreq)
  * @nr_segs: size of iovec array
  *
  * The presence of this routine in the address space ops vector means
- * the NFS client supports direct I/O.  However, we shunt off direct
- * read and write requests before the VFS gets them, so this method
- * should never be called.
+ * the NFS client supports direct I/O. However, for most direct IO, we
+ * shunt off direct read and write requests before the VFS gets them,
+ * so this method is only ever called for swap.
  */
-ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t pos, unsigned long nr_segs)
+ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, struct iov_iter *iter, loff_t pos)
 {
-	dprintk("NFS: nfs_direct_IO (%s) off/no(%Ld/%lu) EINVAL\n",
-			iocb->ki_filp->f_path.dentry->d_name.name,
-			(long long) pos, nr_segs);
+#ifndef CONFIG_NFS_SWAP
+	dprintk("NFS: nfs_direct_IO (%pD) off/no(%Ld/%lu) EINVAL\n",
+			iocb->ki_filp, (long long) pos, iter->nr_segs);
 
 	return -EINVAL;
-}
-
-static void nfs_direct_dirty_pages(struct page **pages, unsigned int pgbase, size_t count)
-{
-	unsigned int npages;
-	unsigned int i;
+#else
+	VM_BUG_ON(iocb->ki_nbytes != PAGE_SIZE);
 
-	if (count == 0)
-		return;
-	pages += (pgbase >> PAGE_SHIFT);
-	npages = (count + (pgbase & ~PAGE_MASK) + PAGE_SIZE - 1) >> PAGE_SHIFT;
-	for (i = 0; i < npages; i++) {
-		struct page *page = pages[i];
-		if (!PageCompound(page))
-			set_page_dirty(page);
-	}
+	if (rw == READ || rw == KERNEL_READ)
+		return nfs_file_direct_read(iocb, iter, pos,
+				rw == READ ? true : false);
+	return nfs_file_direct_write(iocb, iter, pos,
+				rw == WRITE ? true : false);
+#endif /* CONFIG_NFS_SWAP */
 }
 
 static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
@@ -147,26 +237,31 @@ static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
 		page_cache_release(pages[i]);
 }
 
+void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo,
+			      struct nfs_direct_req *dreq)
+{
+	cinfo->lock = &dreq->lock;
+	cinfo->mds = &dreq->mds_cinfo;
+	cinfo->ds = &dreq->ds_cinfo;
+	cinfo->dreq = dreq;
+	cinfo->completion_ops = &nfs_direct_commit_completion_ops;
+}
+
 static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
 {
 	struct nfs_direct_req *dreq;
 
-	dreq = kmem_cache_alloc(nfs_direct_cachep, GFP_KERNEL);
+	dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL);
 	if (!dreq)
 		return NULL;
 
 	kref_init(&dreq->kref);
 	kref_get(&dreq->kref);
 	init_completion(&dreq->completion);
-	INIT_LIST_HEAD(&dreq->rewrite_list);
-	dreq->iocb = NULL;
-	dreq->ctx = NULL;
-	dreq->l_ctx = NULL;
+	INIT_LIST_HEAD(&dreq->mds_cinfo.list);
+	dreq->verf.committed = NFS_INVALID_STABLE_HOW;	/* not set yet */
+	INIT_WORK(&dreq->work, nfs_direct_write_schedule_work);
 	spin_lock_init(&dreq->lock);
-	atomic_set(&dreq->io_count, 0);
-	dreq->count = 0;
-	dreq->error = 0;
-	dreq->flags = 0;
 
 	return dreq;
 }
@@ -187,6 +282,12 @@ static void nfs_direct_req_release(struct nfs_direct_req *dreq)
 	kref_put(&dreq->kref, nfs_direct_req_free);
 }
 
+ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq)
+{
+	return dreq->bytes_left;
+}
+EXPORT_SYMBOL_GPL(nfs_dreq_bytes_left);
+
 /*
  * Collects and returns the final error value/byte-count.
  */
@@ -213,62 +314,97 @@ out:
  * 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.
  */
-static void nfs_direct_complete(struct nfs_direct_req *dreq)
+static void nfs_direct_complete(struct nfs_direct_req *dreq, bool write)
 {
+	struct inode *inode = dreq->inode;
+
+	if (dreq->iocb && write) {
+		loff_t pos = dreq->iocb->ki_pos + dreq->count;
+
+		spin_lock(&inode->i_lock);
+		if (i_size_read(inode) < pos)
+			i_size_write(inode, pos);
+		spin_unlock(&inode->i_lock);
+	}
+
+	if (write)
+		nfs_zap_mapping(inode, inode->i_mapping);
+
+	inode_dio_done(inode);
+
 	if (dreq->iocb) {
 		long res = (long) dreq->error;
 		if (!res)
 			res = (long) dreq->count;
 		aio_complete(dreq->iocb, res, 0);
 	}
+
 	complete_all(&dreq->completion);
 
 	nfs_direct_req_release(dreq);
 }
 
-/*
- * We must hold a reference to all the pages in this direct read request
- * until the RPCs complete.  This could be long *after* we are woken up in
- * nfs_direct_wait (for instance, if someone hits ^C on a slow server).
- */
-static void nfs_direct_read_result(struct rpc_task *task, void *calldata)
+static void nfs_direct_readpage_release(struct nfs_page *req)
 {
-	struct nfs_read_data *data = calldata;
-
-	nfs_readpage_result(task, data);
+	dprintk("NFS: direct read done (%s/%llu %d@%lld)\n",
+		req->wb_context->dentry->d_inode->i_sb->s_id,
+		(unsigned long long)NFS_FILEID(req->wb_context->dentry->d_inode),
+		req->wb_bytes,
+		(long long)req_offset(req));
+	nfs_release_request(req);
 }
 
-static void nfs_direct_read_release(void *calldata)
+static void nfs_direct_read_completion(struct nfs_pgio_header *hdr)
 {
+	unsigned long bytes = 0;
+	struct nfs_direct_req *dreq = hdr->dreq;
 
-	struct nfs_read_data *data = calldata;
-	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
-	int status = data->task.tk_status;
+	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
+		goto out_put;
 
 	spin_lock(&dreq->lock);
-	if (unlikely(status < 0)) {
-		dreq->error = status;
-		spin_unlock(&dreq->lock);
-	} else {
-		dreq->count += data->res.count;
-		spin_unlock(&dreq->lock);
-		nfs_direct_dirty_pages(data->pagevec,
-				data->args.pgbase,
-				data->res.count);
-	}
-	nfs_direct_release_pages(data->pagevec, data->npages);
+	if (test_bit(NFS_IOHDR_ERROR, &hdr->flags) && (hdr->good_bytes == 0))
+		dreq->error = hdr->error;
+	else
+		dreq->count += hdr->good_bytes;
+	spin_unlock(&dreq->lock);
+
+	while (!list_empty(&hdr->pages)) {
+		struct nfs_page *req = nfs_list_entry(hdr->pages.next);
+		struct page *page = req->wb_page;
 
+		if (!PageCompound(page) && bytes < hdr->good_bytes)
+			set_page_dirty(page);
+		bytes += req->wb_bytes;
+		nfs_list_remove_request(req);
+		nfs_direct_readpage_release(req);
+	}
+out_put:
 	if (put_dreq(dreq))
-		nfs_direct_complete(dreq);
-	nfs_readdata_free(data);
+		nfs_direct_complete(dreq, false);
+	hdr->release(hdr);
+}
+
+static void nfs_read_sync_pgio_error(struct list_head *head)
+{
+	struct nfs_page *req;
+
+	while (!list_empty(head)) {
+		req = nfs_list_entry(head->next);
+		nfs_list_remove_request(req);
+		nfs_release_request(req);
+	}
 }
 
-static const struct rpc_call_ops nfs_read_direct_ops = {
-#if defined(CONFIG_NFS_V4_1)
-	.rpc_call_prepare = nfs_read_prepare,
-#endif /* CONFIG_NFS_V4_1 */
-	.rpc_call_done = nfs_direct_read_result,
-	.rpc_release = nfs_direct_read_release,
+static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr)
+{
+	get_dreq(hdr->dreq);
+}
+
+static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = {
+	.error_cleanup = nfs_read_sync_pgio_error,
+	.init_hdr = nfs_direct_pgio_init,
+	.completion = nfs_direct_read_completion,
 };
 
 /*
@@ -278,331 +414,272 @@ static const struct rpc_call_ops nfs_read_direct_ops = {
  * handled automatically by nfs_direct_read_result().  Otherwise, if
  * no requests have been sent, just return an error.
  */
-static ssize_t nfs_direct_read_schedule_segment(struct nfs_direct_req *dreq,
-						const struct iovec *iov,
-						loff_t pos)
-{
-	struct nfs_open_context *ctx = dreq->ctx;
-	struct inode *inode = ctx->path.dentry->d_inode;
-	unsigned long user_addr = (unsigned long)iov->iov_base;
-	size_t count = iov->iov_len;
-	size_t rsize = NFS_SERVER(inode)->rsize;
-	struct rpc_task *task;
-	struct rpc_message msg = {
-		.rpc_cred = ctx->cred,
-	};
-	struct rpc_task_setup task_setup_data = {
-		.rpc_client = NFS_CLIENT(inode),
-		.rpc_message = &msg,
-		.callback_ops = &nfs_read_direct_ops,
-		.workqueue = nfsiod_workqueue,
-		.flags = RPC_TASK_ASYNC,
-	};
-	unsigned int pgbase;
-	int result;
-	ssize_t started = 0;
-
-	do {
-		struct nfs_read_data *data;
-		size_t bytes;
-
-		pgbase = user_addr & ~PAGE_MASK;
-		bytes = min(rsize,count);
-
-		result = -ENOMEM;
-		data = nfs_readdata_alloc(nfs_page_array_len(pgbase, bytes));
-		if (unlikely(!data))
-			break;
-
-		down_read(&current->mm->mmap_sem);
-		result = get_user_pages(current, current->mm, user_addr,
-					data->npages, 1, 0, data->pagevec, NULL);
-		up_read(&current->mm->mmap_sem);
-		if (result < 0) {
-			nfs_readdata_free(data);
-			break;
-		}
-		if ((unsigned)result < data->npages) {
-			bytes = result * PAGE_SIZE;
-			if (bytes <= pgbase) {
-				nfs_direct_release_pages(data->pagevec, result);
-				nfs_readdata_free(data);
-				break;
-			}
-			bytes -= pgbase;
-			data->npages = result;
-		}
-
-		get_dreq(dreq);
-
-		data->req = (struct nfs_page *) dreq;
-		data->inode = inode;
-		data->cred = msg.rpc_cred;
-		data->args.fh = NFS_FH(inode);
-		data->args.context = ctx;
-		data->args.lock_context = dreq->l_ctx;
-		data->args.offset = pos;
-		data->args.pgbase = pgbase;
-		data->args.pages = data->pagevec;
-		data->args.count = bytes;
-		data->res.fattr = &data->fattr;
-		data->res.eof = 0;
-		data->res.count = bytes;
-		nfs_fattr_init(&data->fattr);
-		msg.rpc_argp = &data->args;
-		msg.rpc_resp = &data->res;
-
-		task_setup_data.task = &data->task;
-		task_setup_data.callback_data = data;
-		NFS_PROTO(inode)->read_setup(data, &msg);
-
-		task = rpc_run_task(&task_setup_data);
-		if (IS_ERR(task))
-			break;
-		rpc_put_task(task);
-
-		dprintk("NFS: %5u initiated direct read call "
-			"(req %s/%Ld, %zu bytes @ offset %Lu)\n",
-				data->task.tk_pid,
-				inode->i_sb->s_id,
-				(long long)NFS_FILEID(inode),
-				bytes,
-				(unsigned long long)data->args.offset);
-
-		started += bytes;
-		user_addr += bytes;
-		pos += bytes;
-		/* FIXME: Remove this unnecessary math from final patch */
-		pgbase += bytes;
-		pgbase &= ~PAGE_MASK;
-		BUG_ON(pgbase != (user_addr & ~PAGE_MASK));
-
-		count -= bytes;
-	} while (count != 0);
-
-	if (started)
-		return started;
-	return result < 0 ? (ssize_t) result : -EFAULT;
-}
 
 static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq,
-					      const struct iovec *iov,
-					      unsigned long nr_segs,
+					      struct iov_iter *iter,
 					      loff_t pos)
 {
+	struct nfs_pageio_descriptor desc;
+	struct inode *inode = dreq->inode;
 	ssize_t result = -EINVAL;
 	size_t requested_bytes = 0;
-	unsigned long seg;
+	size_t rsize = max_t(size_t, NFS_SERVER(inode)->rsize, PAGE_SIZE);
 
+	nfs_pageio_init_read(&desc, dreq->inode, false,
+			     &nfs_direct_read_completion_ops);
 	get_dreq(dreq);
+	desc.pg_dreq = dreq;
+	atomic_inc(&inode->i_dio_count);
+
+	while (iov_iter_count(iter)) {
+		struct page **pagevec;
+		size_t bytes;
+		size_t pgbase;
+		unsigned npages, i;
 
-	for (seg = 0; seg < nr_segs; seg++) {
-		const struct iovec *vec = &iov[seg];
-		result = nfs_direct_read_schedule_segment(dreq, vec, pos);
+		result = iov_iter_get_pages_alloc(iter, &pagevec, 
+						  rsize, &pgbase);
 		if (result < 0)
 			break;
-		requested_bytes += result;
-		if ((size_t)result < vec->iov_len)
+	
+		bytes = result;
+		iov_iter_advance(iter, bytes);
+		npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE;
+		for (i = 0; i < npages; i++) {
+			struct nfs_page *req;
+			unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
+			/* XXX do we need to do the eof zeroing found in async_filler? */
+			req = nfs_create_request(dreq->ctx, pagevec[i], NULL,
+						 pgbase, req_len);
+			if (IS_ERR(req)) {
+				result = PTR_ERR(req);
+				break;
+			}
+			req->wb_index = pos >> PAGE_SHIFT;
+			req->wb_offset = pos & ~PAGE_MASK;
+			if (!nfs_pageio_add_request(&desc, req)) {
+				result = desc.pg_error;
+				nfs_release_request(req);
+				break;
+			}
+			pgbase = 0;
+			bytes -= req_len;
+			requested_bytes += req_len;
+			pos += req_len;
+			dreq->bytes_left -= req_len;
+		}
+		nfs_direct_release_pages(pagevec, npages);
+		kvfree(pagevec);
+		if (result < 0)
 			break;
-		pos += vec->iov_len;
 	}
 
-	if (put_dreq(dreq))
-		nfs_direct_complete(dreq);
+	nfs_pageio_complete(&desc);
 
-	if (requested_bytes != 0)
-		return 0;
+	/*
+	 * If no bytes were started, return the error, and let the
+	 * generic layer handle the completion.
+	 */
+	if (requested_bytes == 0) {
+		inode_dio_done(inode);
+		nfs_direct_req_release(dreq);
+		return result < 0 ? result : -EIO;
+	}
 
-	if (result < 0)
-		return result;
-	return -EIO;
+	if (put_dreq(dreq))
+		nfs_direct_complete(dreq, false);
+	return 0;
 }
 
-static ssize_t nfs_direct_read(struct kiocb *iocb, const struct iovec *iov,
-			       unsigned long nr_segs, loff_t pos)
+/**
+ * nfs_file_direct_read - file direct read operation for NFS files
+ * @iocb: target I/O control block
+ * @iter: vector of user buffers into which to read data
+ * @pos: byte offset in file where reading starts
+ *
+ * We use this function for direct reads instead of calling
+ * generic_file_aio_read() in order to avoid gfar's check to see if
+ * the request starts before the end of the file.  For that check
+ * to work, we must generate a GETATTR before each direct read, and
+ * even then there is a window between the GETATTR and the subsequent
+ * READ where the file size could change.  Our preference is simply
+ * to do all reads the application wants, and the server will take
+ * care of managing the end of file boundary.
+ *
+ * This function also eliminates unnecessarily updating the file's
+ * atime locally, as the NFS server sets the file's atime, and this
+ * client must read the updated atime from the server back into its
+ * cache.
+ */
+ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter,
+				loff_t pos, bool uio)
 {
-	ssize_t result = -ENOMEM;
-	struct inode *inode = iocb->ki_filp->f_mapping->host;
+	struct file *file = iocb->ki_filp;
+	struct address_space *mapping = file->f_mapping;
+	struct inode *inode = mapping->host;
 	struct nfs_direct_req *dreq;
+	struct nfs_lock_context *l_ctx;
+	ssize_t result = -EINVAL;
+	size_t count = iov_iter_count(iter);
+	nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
+
+	dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n",
+		file, count, (long long) pos);
+
+	result = 0;
+	if (!count)
+		goto out;
 
+	mutex_lock(&inode->i_mutex);
+	result = nfs_sync_mapping(mapping);
+	if (result)
+		goto out_unlock;
+
+	task_io_account_read(count);
+
+	result = -ENOMEM;
 	dreq = nfs_direct_req_alloc();
 	if (dreq == NULL)
-		goto out;
+		goto out_unlock;
 
 	dreq->inode = inode;
+	dreq->bytes_left = count;
 	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
-	dreq->l_ctx = nfs_get_lock_context(dreq->ctx);
-	if (dreq->l_ctx == NULL)
+	l_ctx = nfs_get_lock_context(dreq->ctx);
+	if (IS_ERR(l_ctx)) {
+		result = PTR_ERR(l_ctx);
 		goto out_release;
+	}
+	dreq->l_ctx = l_ctx;
 	if (!is_sync_kiocb(iocb))
 		dreq->iocb = iocb;
 
-	result = nfs_direct_read_schedule_iovec(dreq, iov, nr_segs, pos);
-	if (!result)
+	NFS_I(inode)->read_io += count;
+	result = nfs_direct_read_schedule_iovec(dreq, iter, pos);
+
+	mutex_unlock(&inode->i_mutex);
+
+	if (!result) {
 		result = nfs_direct_wait(dreq);
+		if (result > 0)
+			iocb->ki_pos = pos + result;
+	}
+
+	nfs_direct_req_release(dreq);
+	return result;
+
 out_release:
 	nfs_direct_req_release(dreq);
+out_unlock:
+	mutex_unlock(&inode->i_mutex);
 out:
 	return result;
 }
 
-static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)
-{
-	while (!list_empty(&dreq->rewrite_list)) {
-		struct nfs_write_data *data = list_entry(dreq->rewrite_list.next, struct nfs_write_data, pages);
-		list_del(&data->pages);
-		nfs_direct_release_pages(data->pagevec, data->npages);
-		nfs_writedata_free(data);
-	}
-}
-
-#if defined(CONFIG_NFS_V3) || defined(CONFIG_NFS_V4)
+#if IS_ENABLED(CONFIG_NFS_V3) || IS_ENABLED(CONFIG_NFS_V4)
 static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
 {
-	struct inode *inode = dreq->inode;
-	struct list_head *p;
-	struct nfs_write_data *data;
-	struct rpc_task *task;
-	struct rpc_message msg = {
-		.rpc_cred = dreq->ctx->cred,
-	};
-	struct rpc_task_setup task_setup_data = {
-		.rpc_client = NFS_CLIENT(inode),
-		.rpc_message = &msg,
-		.callback_ops = &nfs_write_direct_ops,
-		.workqueue = nfsiod_workqueue,
-		.flags = RPC_TASK_ASYNC,
-	};
+	struct nfs_pageio_descriptor desc;
+	struct nfs_page *req, *tmp;
+	LIST_HEAD(reqs);
+	struct nfs_commit_info cinfo;
+	LIST_HEAD(failed);
+
+	nfs_init_cinfo_from_dreq(&cinfo, dreq);
+	pnfs_recover_commit_reqs(dreq->inode, &reqs, &cinfo);
+	spin_lock(cinfo.lock);
+	nfs_scan_commit_list(&cinfo.mds->list, &reqs, &cinfo, 0);
+	spin_unlock(cinfo.lock);
 
 	dreq->count = 0;
 	get_dreq(dreq);
 
-	list_for_each(p, &dreq->rewrite_list) {
-		data = list_entry(p, struct nfs_write_data, pages);
-
-		get_dreq(dreq);
-
-		/* Use stable writes */
-		data->args.stable = NFS_FILE_SYNC;
-
-		/*
-		 * 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.
-		 */
-		task_setup_data.task = &data->task;
-		task_setup_data.callback_data = data;
-		msg.rpc_argp = &data->args;
-		msg.rpc_resp = &data->res;
-		NFS_PROTO(inode)->write_setup(data, &msg);
-
-		/*
-		 * We're called via an RPC callback, so BKL is already held.
-		 */
-		task = rpc_run_task(&task_setup_data);
-		if (!IS_ERR(task))
-			rpc_put_task(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);
+	nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, false,
+			      &nfs_direct_write_completion_ops);
+	desc.pg_dreq = dreq;
+
+	list_for_each_entry_safe(req, tmp, &reqs, wb_list) {
+		if (!nfs_pageio_add_request(&desc, req)) {
+			nfs_list_remove_request(req);
+			nfs_list_add_request(req, &failed);
+			spin_lock(cinfo.lock);
+			dreq->flags = 0;
+			dreq->error = -EIO;
+			spin_unlock(cinfo.lock);
+		}
+		nfs_release_request(req);
 	}
+	nfs_pageio_complete(&desc);
 
-	if (put_dreq(dreq))
-		nfs_direct_write_complete(dreq, inode);
-}
-
-static void nfs_direct_commit_result(struct rpc_task *task, void *calldata)
-{
-	struct nfs_write_data *data = calldata;
+	while (!list_empty(&failed)) {
+		req = nfs_list_entry(failed.next);
+		nfs_list_remove_request(req);
+		nfs_unlock_and_release_request(req);
+	}
 
-	/* Call the NFS version-specific code */
-	NFS_PROTO(data->inode)->commit_done(task, data);
+	if (put_dreq(dreq))
+		nfs_direct_write_complete(dreq, dreq->inode);
 }
 
-static void nfs_direct_commit_release(void *calldata)
+static void nfs_direct_commit_complete(struct nfs_commit_data *data)
 {
-	struct nfs_write_data *data = calldata;
-	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
+	struct nfs_direct_req *dreq = data->dreq;
+	struct nfs_commit_info cinfo;
+	struct nfs_page *req;
 	int status = data->task.tk_status;
 
+	nfs_init_cinfo_from_dreq(&cinfo, dreq);
 	if (status < 0) {
 		dprintk("NFS: %5u commit failed with error %d.\n",
-				data->task.tk_pid, status);
+			data->task.tk_pid, status);
 		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
-	} else if (memcmp(&dreq->verf, &data->verf, sizeof(data->verf))) {
+	} else if (nfs_direct_cmp_commit_data_verf(dreq, data)) {
 		dprintk("NFS: %5u commit verify failed\n", data->task.tk_pid);
 		dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
 	}
 
 	dprintk("NFS: %5u commit returned %d\n", data->task.tk_pid, status);
-	nfs_direct_write_complete(dreq, data->inode);
-	nfs_commit_free(data);
+	while (!list_empty(&data->pages)) {
+		req = nfs_list_entry(data->pages.next);
+		nfs_list_remove_request(req);
+		if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES) {
+			/* Note the rewrite will go through mds */
+			nfs_mark_request_commit(req, NULL, &cinfo);
+		} else
+			nfs_release_request(req);
+		nfs_unlock_and_release_request(req);
+	}
+
+	if (atomic_dec_and_test(&cinfo.mds->rpcs_out))
+		nfs_direct_write_complete(dreq, data->inode);
+}
+
+static void nfs_direct_error_cleanup(struct nfs_inode *nfsi)
+{
+	/* There is no lock to clear */
 }
 
-static const struct rpc_call_ops nfs_commit_direct_ops = {
-#if defined(CONFIG_NFS_V4_1)
-	.rpc_call_prepare = nfs_write_prepare,
-#endif /* CONFIG_NFS_V4_1 */
-	.rpc_call_done = nfs_direct_commit_result,
-	.rpc_release = nfs_direct_commit_release,
+static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = {
+	.completion = nfs_direct_commit_complete,
+	.error_cleanup = nfs_direct_error_cleanup,
 };
 
 static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
 {
-	struct nfs_write_data *data = dreq->commit_data;
-	struct rpc_task *task;
-	struct rpc_message msg = {
-		.rpc_argp = &data->args,
-		.rpc_resp = &data->res,
-		.rpc_cred = dreq->ctx->cred,
-	};
-	struct rpc_task_setup task_setup_data = {
-		.task = &data->task,
-		.rpc_client = NFS_CLIENT(dreq->inode),
-		.rpc_message = &msg,
-		.callback_ops = &nfs_commit_direct_ops,
-		.callback_data = data,
-		.workqueue = nfsiod_workqueue,
-		.flags = RPC_TASK_ASYNC,
-	};
-
-	data->inode = dreq->inode;
-	data->cred = msg.rpc_cred;
-
-	data->args.fh = NFS_FH(data->inode);
-	data->args.offset = 0;
-	data->args.count = 0;
-	data->args.context = dreq->ctx;
-	data->args.lock_context = dreq->l_ctx;
-	data->res.count = 0;
-	data->res.fattr = &data->fattr;
-	data->res.verf = &data->verf;
-	nfs_fattr_init(&data->fattr);
-
-	NFS_PROTO(data->inode)->commit_setup(data, &msg);
-
-	/* Note: task.tk_ops->rpc_release will free dreq->commit_data */
-	dreq->commit_data = NULL;
-
-	dprintk("NFS: %5u initiated commit call\n", data->task.tk_pid);
-
-	task = rpc_run_task(&task_setup_data);
-	if (!IS_ERR(task))
-		rpc_put_task(task);
+	int res;
+	struct nfs_commit_info cinfo;
+	LIST_HEAD(mds_list);
+
+	nfs_init_cinfo_from_dreq(&cinfo, dreq);
+	nfs_scan_commit(dreq->inode, &mds_list, &cinfo);
+	res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo);
+	if (res < 0) /* res == -ENOMEM */
+		nfs_direct_write_reschedule(dreq);
 }
 
-static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
+static void nfs_direct_write_schedule_work(struct work_struct *work)
 {
+	struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work);
 	int flags = dreq->flags;
 
 	dreq->flags = 0;
@@ -614,92 +691,111 @@ static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode
 			nfs_direct_write_reschedule(dreq);
 			break;
 		default:
-			if (dreq->commit_data != NULL)
-				nfs_commit_free(dreq->commit_data);
-			nfs_direct_free_writedata(dreq);
-			nfs_zap_mapping(inode, inode->i_mapping);
-			nfs_direct_complete(dreq);
+			nfs_direct_complete(dreq, true);
 	}
 }
 
-static void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
+static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
 {
-	dreq->commit_data = nfs_commitdata_alloc();
-	if (dreq->commit_data != NULL)
-		dreq->commit_data->req = (struct nfs_page *) dreq;
+	schedule_work(&dreq->work); /* Calls nfs_direct_write_schedule_work */
 }
+
 #else
-static inline void nfs_alloc_commit_data(struct nfs_direct_req *dreq)
+static void nfs_direct_write_schedule_work(struct work_struct *work)
 {
-	dreq->commit_data = NULL;
 }
 
 static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode *inode)
 {
-	nfs_direct_free_writedata(dreq);
-	nfs_zap_mapping(inode, inode->i_mapping);
-	nfs_direct_complete(dreq);
+	nfs_direct_complete(dreq, true);
 }
 #endif
 
-static void nfs_direct_write_result(struct rpc_task *task, void *calldata)
+static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
 {
-	struct nfs_write_data *data = calldata;
+	struct nfs_direct_req *dreq = hdr->dreq;
+	struct nfs_commit_info cinfo;
+	int bit = -1;
+	struct nfs_page *req = nfs_list_entry(hdr->pages.next);
 
-	if (nfs_writeback_done(task, data) != 0)
-		return;
-}
+	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
+		goto out_put;
 
-/*
- * NB: Return the value of the first error return code.  Subsequent
- *     errors after the first one are ignored.
- */
-static void nfs_direct_write_release(void *calldata)
-{
-	struct nfs_write_data *data = calldata;
-	struct nfs_direct_req *dreq = (struct nfs_direct_req *) data->req;
-	int status = data->task.tk_status;
+	nfs_init_cinfo_from_dreq(&cinfo, dreq);
 
 	spin_lock(&dreq->lock);
 
-	if (unlikely(status < 0)) {
-		/* An error has occurred, so we should not commit */
+	if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
 		dreq->flags = 0;
-		dreq->error = status;
+		dreq->error = hdr->error;
 	}
-	if (unlikely(dreq->error != 0))
-		goto out_unlock;
-
-	dreq->count += data->res.count;
-
-	if (data->res.verf->committed != NFS_FILE_SYNC) {
-		switch (dreq->flags) {
-			case 0:
-				memcpy(&dreq->verf, &data->verf, sizeof(dreq->verf));
+	if (dreq->error != 0)
+		bit = NFS_IOHDR_ERROR;
+	else {
+		dreq->count += hdr->good_bytes;
+		if (test_bit(NFS_IOHDR_NEED_RESCHED, &hdr->flags)) {
+			dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
+			bit = NFS_IOHDR_NEED_RESCHED;
+		} else if (test_bit(NFS_IOHDR_NEED_COMMIT, &hdr->flags)) {
+			if (dreq->flags == NFS_ODIRECT_RESCHED_WRITES)
+				bit = NFS_IOHDR_NEED_RESCHED;
+			else if (dreq->flags == 0) {
+				nfs_direct_set_hdr_verf(dreq, hdr);
+				bit = NFS_IOHDR_NEED_COMMIT;
 				dreq->flags = NFS_ODIRECT_DO_COMMIT;
-				break;
-			case NFS_ODIRECT_DO_COMMIT:
-				if (memcmp(&dreq->verf, &data->verf, sizeof(dreq->verf))) {
-					dprintk("NFS: %5u write verify failed\n", data->task.tk_pid);
-					dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
-				}
+			} else if (dreq->flags == NFS_ODIRECT_DO_COMMIT) {
+				if (nfs_direct_set_or_cmp_hdr_verf(dreq, hdr)) {
+					dreq->flags =
+						NFS_ODIRECT_RESCHED_WRITES;
+					bit = NFS_IOHDR_NEED_RESCHED;
+				} else
+					bit = NFS_IOHDR_NEED_COMMIT;
+			}
 		}
 	}
-out_unlock:
 	spin_unlock(&dreq->lock);
 
+	while (!list_empty(&hdr->pages)) {
+
+		req = nfs_list_entry(hdr->pages.next);
+		nfs_list_remove_request(req);
+		switch (bit) {
+		case NFS_IOHDR_NEED_RESCHED:
+		case NFS_IOHDR_NEED_COMMIT:
+			kref_get(&req->wb_kref);
+			nfs_mark_request_commit(req, hdr->lseg, &cinfo);
+		}
+		nfs_unlock_and_release_request(req);
+	}
+
+out_put:
 	if (put_dreq(dreq))
-		nfs_direct_write_complete(dreq, data->inode);
+		nfs_direct_write_complete(dreq, hdr->inode);
+	hdr->release(hdr);
 }
 
-static const struct rpc_call_ops nfs_write_direct_ops = {
-#if defined(CONFIG_NFS_V4_1)
-	.rpc_call_prepare = nfs_write_prepare,
-#endif /* CONFIG_NFS_V4_1 */
-	.rpc_call_done = nfs_direct_write_result,
-	.rpc_release = nfs_direct_write_release,
+static void nfs_write_sync_pgio_error(struct list_head *head)
+{
+	struct nfs_page *req;
+
+	while (!list_empty(head)) {
+		req = nfs_list_entry(head->next);
+		nfs_list_remove_request(req);
+		nfs_unlock_and_release_request(req);
+	}
+}
+
+static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = {
+	.error_cleanup = nfs_write_sync_pgio_error,
+	.init_hdr = nfs_direct_pgio_init,
+	.completion = nfs_direct_write_completion,
 };
 
+
+/*
+ * NB: Return the value of the first error return code.  Subsequent
+ *     errors after the first one are ignored.
+ */
 /*
  * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
  * operation.  If nfs_writedata_alloc() or get_user_pages() fails,
@@ -707,244 +803,87 @@ static const struct rpc_call_ops nfs_write_direct_ops = {
  * handled automatically by nfs_direct_write_result().  Otherwise, if
  * no requests have been sent, just return an error.
  */
-static ssize_t nfs_direct_write_schedule_segment(struct nfs_direct_req *dreq,
-						 const struct iovec *iov,
-						 loff_t pos, int sync)
-{
-	struct nfs_open_context *ctx = dreq->ctx;
-	struct inode *inode = ctx->path.dentry->d_inode;
-	unsigned long user_addr = (unsigned long)iov->iov_base;
-	size_t count = iov->iov_len;
-	struct rpc_task *task;
-	struct rpc_message msg = {
-		.rpc_cred = ctx->cred,
-	};
-	struct rpc_task_setup task_setup_data = {
-		.rpc_client = NFS_CLIENT(inode),
-		.rpc_message = &msg,
-		.callback_ops = &nfs_write_direct_ops,
-		.workqueue = nfsiod_workqueue,
-		.flags = RPC_TASK_ASYNC,
-	};
-	size_t wsize = NFS_SERVER(inode)->wsize;
-	unsigned int pgbase;
-	int result;
-	ssize_t started = 0;
-
-	do {
-		struct nfs_write_data *data;
-		size_t bytes;
-
-		pgbase = user_addr & ~PAGE_MASK;
-		bytes = min(wsize,count);
-
-		result = -ENOMEM;
-		data = nfs_writedata_alloc(nfs_page_array_len(pgbase, bytes));
-		if (unlikely(!data))
-			break;
-
-		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 (result < 0) {
-			nfs_writedata_free(data);
-			break;
-		}
-		if ((unsigned)result < data->npages) {
-			bytes = result * PAGE_SIZE;
-			if (bytes <= pgbase) {
-				nfs_direct_release_pages(data->pagevec, result);
-				nfs_writedata_free(data);
-				break;
-			}
-			bytes -= pgbase;
-			data->npages = result;
-		}
-
-		get_dreq(dreq);
-
-		list_move_tail(&data->pages, &dreq->rewrite_list);
-
-		data->req = (struct nfs_page *) dreq;
-		data->inode = inode;
-		data->cred = msg.rpc_cred;
-		data->args.fh = NFS_FH(inode);
-		data->args.context = ctx;
-		data->args.lock_context = dreq->l_ctx;
-		data->args.offset = pos;
-		data->args.pgbase = pgbase;
-		data->args.pages = data->pagevec;
-		data->args.count = bytes;
-		data->args.stable = sync;
-		data->res.fattr = &data->fattr;
-		data->res.count = bytes;
-		data->res.verf = &data->verf;
-		nfs_fattr_init(&data->fattr);
-
-		task_setup_data.task = &data->task;
-		task_setup_data.callback_data = data;
-		msg.rpc_argp = &data->args;
-		msg.rpc_resp = &data->res;
-		NFS_PROTO(inode)->write_setup(data, &msg);
-
-		task = rpc_run_task(&task_setup_data);
-		if (IS_ERR(task))
-			break;
-		rpc_put_task(task);
-
-		dprintk("NFS: %5u initiated direct write call "
-			"(req %s/%Ld, %zu bytes @ offset %Lu)\n",
-				data->task.tk_pid,
-				inode->i_sb->s_id,
-				(long long)NFS_FILEID(inode),
-				bytes,
-				(unsigned long long)data->args.offset);
-
-		started += bytes;
-		user_addr += bytes;
-		pos += bytes;
-
-		/* FIXME: Remove this useless math from the final patch */
-		pgbase += bytes;
-		pgbase &= ~PAGE_MASK;
-		BUG_ON(pgbase != (user_addr & ~PAGE_MASK));
-
-		count -= bytes;
-	} while (count != 0);
-
-	if (started)
-		return started;
-	return result < 0 ? (ssize_t) result : -EFAULT;
-}
-
 static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
-					       const struct iovec *iov,
-					       unsigned long nr_segs,
-					       loff_t pos, int sync)
+					       struct iov_iter *iter,
+					       loff_t pos)
 {
+	struct nfs_pageio_descriptor desc;
+	struct inode *inode = dreq->inode;
 	ssize_t result = 0;
 	size_t requested_bytes = 0;
-	unsigned long seg;
+	size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE);
 
+	nfs_pageio_init_write(&desc, inode, FLUSH_COND_STABLE, false,
+			      &nfs_direct_write_completion_ops);
+	desc.pg_dreq = dreq;
 	get_dreq(dreq);
+	atomic_inc(&inode->i_dio_count);
 
-	for (seg = 0; seg < nr_segs; seg++) {
-		const struct iovec *vec = &iov[seg];
-		result = nfs_direct_write_schedule_segment(dreq, vec,
-							   pos, sync);
+	NFS_I(inode)->write_io += iov_iter_count(iter);
+	while (iov_iter_count(iter)) {
+		struct page **pagevec;
+		size_t bytes;
+		size_t pgbase;
+		unsigned npages, i;
+
+		result = iov_iter_get_pages_alloc(iter, &pagevec, 
+						  wsize, &pgbase);
 		if (result < 0)
 			break;
-		requested_bytes += result;
-		if ((size_t)result < vec->iov_len)
+
+		bytes = result;
+		iov_iter_advance(iter, bytes);
+		npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE;
+		for (i = 0; i < npages; i++) {
+			struct nfs_page *req;
+			unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
+
+			req = nfs_create_request(dreq->ctx, pagevec[i], NULL,
+						 pgbase, req_len);
+			if (IS_ERR(req)) {
+				result = PTR_ERR(req);
+				break;
+			}
+			nfs_lock_request(req);
+			req->wb_index = pos >> PAGE_SHIFT;
+			req->wb_offset = pos & ~PAGE_MASK;
+			if (!nfs_pageio_add_request(&desc, req)) {
+				result = desc.pg_error;
+				nfs_unlock_and_release_request(req);
+				break;
+			}
+			pgbase = 0;
+			bytes -= req_len;
+			requested_bytes += req_len;
+			pos += req_len;
+			dreq->bytes_left -= req_len;
+		}
+		nfs_direct_release_pages(pagevec, npages);
+		kvfree(pagevec);
+		if (result < 0)
 			break;
-		pos += vec->iov_len;
+	}
+	nfs_pageio_complete(&desc);
+
+	/*
+	 * If no bytes were started, return the error, and let the
+	 * generic layer handle the completion.
+	 */
+	if (requested_bytes == 0) {
+		inode_dio_done(inode);
+		nfs_direct_req_release(dreq);
+		return result < 0 ? result : -EIO;
 	}
 
 	if (put_dreq(dreq))
 		nfs_direct_write_complete(dreq, dreq->inode);
-
-	if (requested_bytes != 0)
-		return 0;
-
-	if (result < 0)
-		return result;
-	return -EIO;
-}
-
-static ssize_t nfs_direct_write(struct kiocb *iocb, const struct iovec *iov,
-				unsigned long nr_segs, loff_t pos,
-				size_t count)
-{
-	ssize_t result = -ENOMEM;
-	struct inode *inode = iocb->ki_filp->f_mapping->host;
-	struct nfs_direct_req *dreq;
-	size_t wsize = NFS_SERVER(inode)->wsize;
-	int sync = NFS_UNSTABLE;
-
-	dreq = nfs_direct_req_alloc();
-	if (!dreq)
-		goto out;
-	nfs_alloc_commit_data(dreq);
-
-	if (dreq->commit_data == NULL || count < wsize)
-		sync = NFS_FILE_SYNC;
-
-	dreq->inode = inode;
-	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
-	dreq->l_ctx = nfs_get_lock_context(dreq->ctx);
-	if (dreq->l_ctx == NULL)
-		goto out_release;
-	if (!is_sync_kiocb(iocb))
-		dreq->iocb = iocb;
-
-	result = nfs_direct_write_schedule_iovec(dreq, iov, nr_segs, pos, sync);
-	if (!result)
-		result = nfs_direct_wait(dreq);
-out_release:
-	nfs_direct_req_release(dreq);
-out:
-	return result;
-}
-
-/**
- * nfs_file_direct_read - file direct read operation for NFS files
- * @iocb: target I/O control block
- * @iov: vector of user buffers into which to read data
- * @nr_segs: size of iov vector
- * @pos: byte offset in file where reading starts
- *
- * We use this function for direct reads instead of calling
- * generic_file_aio_read() in order to avoid gfar's check to see if
- * the request starts before the end of the file.  For that check
- * to work, we must generate a GETATTR before each direct read, and
- * even then there is a window between the GETATTR and the subsequent
- * READ where the file size could change.  Our preference is simply
- * to do all reads the application wants, and the server will take
- * care of managing the end of file boundary.
- *
- * This function also eliminates unnecessarily updating the file's
- * atime locally, as the NFS server sets the file's atime, and this
- * client must read the updated atime from the server back into its
- * cache.
- */
-ssize_t nfs_file_direct_read(struct kiocb *iocb, const struct iovec *iov,
-				unsigned long nr_segs, loff_t pos)
-{
-	ssize_t retval = -EINVAL;
-	struct file *file = iocb->ki_filp;
-	struct address_space *mapping = file->f_mapping;
-	size_t count;
-
-	count = iov_length(iov, nr_segs);
-	nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
-
-	dfprintk(FILE, "NFS: direct read(%s/%s, %zd@%Ld)\n",
-		file->f_path.dentry->d_parent->d_name.name,
-		file->f_path.dentry->d_name.name,
-		count, (long long) pos);
-
-	retval = 0;
-	if (!count)
-		goto out;
-
-	retval = nfs_sync_mapping(mapping);
-	if (retval)
-		goto out;
-
-	retval = nfs_direct_read(iocb, iov, nr_segs, pos);
-	if (retval > 0)
-		iocb->ki_pos = pos + retval;
-
-out:
-	return retval;
+	return 0;
 }
 
 /**
  * nfs_file_direct_write - file direct write operation for NFS files
  * @iocb: target I/O control block
- * @iov: vector of user buffers from which to write data
- * @nr_segs: size of iov vector
+ * @iter: vector of user buffers from which to write data
  * @pos: byte offset in file where writing starts
  *
  * We use this function for direct writes instead of calling
@@ -962,44 +901,97 @@ out:
  * Note that O_APPEND is not supported for NFS direct writes, as there
  * is no atomic O_APPEND write facility in the NFS protocol.
  */
-ssize_t nfs_file_direct_write(struct kiocb *iocb, const struct iovec *iov,
-				unsigned long nr_segs, loff_t pos)
+ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter,
+				loff_t pos, bool uio)
 {
-	ssize_t retval = -EINVAL;
+	ssize_t result = -EINVAL;
 	struct file *file = iocb->ki_filp;
 	struct address_space *mapping = file->f_mapping;
-	size_t count;
+	struct inode *inode = mapping->host;
+	struct nfs_direct_req *dreq;
+	struct nfs_lock_context *l_ctx;
+	loff_t end;
+	size_t count = iov_iter_count(iter);
+	end = (pos + count - 1) >> PAGE_CACHE_SHIFT;
 
-	count = iov_length(iov, nr_segs);
 	nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);
 
-	dfprintk(FILE, "NFS: direct write(%s/%s, %zd@%Ld)\n",
-		file->f_path.dentry->d_parent->d_name.name,
-		file->f_path.dentry->d_name.name,
-		count, (long long) pos);
+	dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n",
+		file, count, (long long) pos);
 
-	retval = generic_write_checks(file, &pos, &count, 0);
-	if (retval)
+	result = generic_write_checks(file, &pos, &count, 0);
+	if (result)
 		goto out;
 
-	retval = -EINVAL;
+	result = -EINVAL;
 	if ((ssize_t) count < 0)
 		goto out;
-	retval = 0;
+	result = 0;
 	if (!count)
 		goto out;
 
-	retval = nfs_sync_mapping(mapping);
-	if (retval)
-		goto out;
+	mutex_lock(&inode->i_mutex);
+
+	result = nfs_sync_mapping(mapping);
+	if (result)
+		goto out_unlock;
+
+	if (mapping->nrpages) {
+		result = invalidate_inode_pages2_range(mapping,
+					pos >> PAGE_CACHE_SHIFT, end);
+		if (result)
+			goto out_unlock;
+	}
 
-	retval = nfs_direct_write(iocb, iov, nr_segs, pos, count);
+	task_io_account_write(count);
 
-	if (retval > 0)
-		iocb->ki_pos = pos + retval;
+	result = -ENOMEM;
+	dreq = nfs_direct_req_alloc();
+	if (!dreq)
+		goto out_unlock;
 
+	dreq->inode = inode;
+	dreq->bytes_left = count;
+	dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
+	l_ctx = nfs_get_lock_context(dreq->ctx);
+	if (IS_ERR(l_ctx)) {
+		result = PTR_ERR(l_ctx);
+		goto out_release;
+	}
+	dreq->l_ctx = l_ctx;
+	if (!is_sync_kiocb(iocb))
+		dreq->iocb = iocb;
+
+	result = nfs_direct_write_schedule_iovec(dreq, iter, pos);
+
+	if (mapping->nrpages) {
+		invalidate_inode_pages2_range(mapping,
+					      pos >> PAGE_CACHE_SHIFT, end);
+	}
+
+	mutex_unlock(&inode->i_mutex);
+
+	if (!result) {
+		result = nfs_direct_wait(dreq);
+		if (result > 0) {
+			struct inode *inode = mapping->host;
+
+			iocb->ki_pos = pos + result;
+			spin_lock(&inode->i_lock);
+			if (i_size_read(inode) < iocb->ki_pos)
+				i_size_write(inode, iocb->ki_pos);
+			spin_unlock(&inode->i_lock);
+		}
+	}
+	nfs_direct_req_release(dreq);
+	return result;
+
+out_release:
+	nfs_direct_req_release(dreq);
+out_unlock:
+	mutex_unlock(&inode->i_mutex);
 out:
-	return retval;
+	return result;
 }
 
 /**