Merge branch 'odirect'
Trond Myklebust [Sun, 25 Jun 2006 10:27:31 +0000 (06:27 -0400)]
fs/nfs/direct.c
include/linux/nfs_xdr.h

index e25b759..402005c 100644 (file)
@@ -68,19 +68,25 @@ struct nfs_direct_req {
        struct kref             kref;           /* release manager */
 
        /* 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 kiocb *          iocb;           /* controlling i/o request */
        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 */
-       atomic_t                io_count;       /* i/os we're waiting for */
        spinlock_t              lock;           /* protect completion state */
+       int                     outstanding;    /* i/os we're waiting for */
        ssize_t                 count,          /* bytes actually processed */
                                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 */
        int                     flags;
 #define NFS_ODIRECT_DO_COMMIT          (1)     /* an unstable reply was received */
@@ -88,37 +94,8 @@ struct nfs_direct_req {
        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 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
@@ -142,21 +119,50 @@ ssize_t nfs_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_
        return -EINVAL;
 }
 
-static void nfs_direct_dirty_pages(struct page **pages, int npages)
+static void nfs_free_user_pages(struct page **pages, int npages, int do_dirty)
 {
        int i;
        for (i = 0; i < npages; i++) {
                struct page *page = pages[i];
-               if (!PageCompound(page))
+               if (do_dirty && !PageCompound(page))
                        set_page_dirty_lock(page);
+               page_cache_release(page);
        }
+       kfree(pages);
 }
 
-static void nfs_direct_release_pages(struct page **pages, int npages)
+static inline int nfs_get_user_pages(int rw, unsigned long user_addr, size_t size, struct page ***pages)
 {
-       int i;
-       for (i = 0; i < npages; i++)
-               page_cache_release(pages[i]);
+       int result = -ENOMEM;
+       unsigned long page_count;
+       size_t array_size;
+
+       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)
@@ -168,13 +174,13 @@ static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
                return NULL;
 
        kref_init(&dreq->kref);
-       kref_get(&dreq->kref);
        init_completion(&dreq->completion);
+       INIT_LIST_HEAD(&dreq->list);
        INIT_LIST_HEAD(&dreq->rewrite_list);
        dreq->iocb = NULL;
        dreq->ctx = NULL;
        spin_lock_init(&dreq->lock);
-       atomic_set(&dreq->io_count, 0);
+       dreq->outstanding = 0;
        dreq->count = 0;
        dreq->error = 0;
        dreq->flags = 0;
@@ -215,11 +221,18 @@ 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.
+ * 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).
+ *
+ * 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)
 {
+       nfs_free_user_pages(dreq->pages, dreq->npages, 1);
+
        if (dreq->iocb) {
                long res = (long) dreq->error;
                if (!res)
@@ -232,10 +245,48 @@ static void nfs_direct_complete(struct nfs_direct_req *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).
+ * Note we also set the number of requests we have in the dreq when we are
+ * done.  This prevents races with I/O completion so we will always wait
+ * until all requests have been dispatched and completed.
  */
+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)
 {
        struct nfs_read_data *data = calldata;
@@ -244,9 +295,6 @@ static void nfs_direct_read_result(struct rpc_task *task, void *calldata)
        if (nfs_readpage_result(task, data) != 0)
                return;
 
-       nfs_direct_dirty_pages(data->pagevec, data->npages);
-       nfs_direct_release_pages(data->pagevec, data->npages);
-
        spin_lock(&dreq->lock);
 
        if (likely(task->tk_status >= 0))
@@ -254,10 +302,13 @@ static void nfs_direct_read_result(struct rpc_task *task, void *calldata)
        else
                dreq->error = task->tk_status;
 
-       spin_unlock(&dreq->lock);
+       if (--dreq->outstanding) {
+               spin_unlock(&dreq->lock);
+               return;
+       }
 
-       if (put_dreq(dreq))
-               nfs_direct_complete(dreq);
+       spin_unlock(&dreq->lock);
+       nfs_direct_complete(dreq);
 }
 
 static const struct rpc_call_ops nfs_read_direct_ops = {
@@ -266,60 +317,41 @@ static const struct rpc_call_ops nfs_read_direct_ops = {
 };
 
 /*
- * For each rsize'd chunk of the user's buffer, dispatch an NFS READ
- * 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.
+ * For each nfs_read_data struct that was allocated on the list, dispatch
+ * an NFS READ operation
  */
-static ssize_t nfs_direct_read_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos)
+static void nfs_direct_read_schedule(struct nfs_direct_req *dreq)
 {
        struct nfs_open_context *ctx = dreq->ctx;
        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;
-       unsigned int rpages = nfs_max_pages(rsize);
-       unsigned int pgbase;
-       int result;
-       ssize_t started = 0;
-
-       get_dreq(dreq);
+       unsigned int curpage, pgbase;
 
-       pgbase = user_addr & ~PAGE_MASK;
+       curpage = 0;
+       pgbase = dreq->user_addr & ~PAGE_MASK;
        do {
                struct nfs_read_data *data;
                size_t bytes;
 
-               result = -ENOMEM;
-               data = nfs_readdata_alloc(rpages);
-               if (unlikely(!data))
-                       break;
-
                bytes = rsize;
                if (count < rsize)
                        bytes = count;
 
-               data->npages = nfs_direct_count_pages(user_addr, bytes);
-               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 (unlikely(result < data->npages)) {
-                       if (result > 0)
-                               nfs_direct_release_pages(data->pagevec, result);
-                       nfs_readdata_release(data);
-                       break;
-               }
-
-               get_dreq(dreq);
+               BUG_ON(list_empty(list));
+               data = list_entry(list->next, struct nfs_read_data, pages);
+               list_del_init(&data->pages);
 
-               data->req = (struct nfs_page *) dreq;
                data->inode = inode;
                data->cred = ctx->cred;
                data->args.fh = NFS_FH(inode);
                data->args.context = ctx;
                data->args.offset = pos;
                data->args.pgbase = pgbase;
-               data->args.pages = data->pagevec;
+               data->args.pages = &pages[curpage];
                data->args.count = bytes;
                data->res.fattr = &data->fattr;
                data->res.eof = 0;
@@ -342,35 +374,33 @@ static ssize_t nfs_direct_read_schedule(struct nfs_direct_req *dreq, unsigned lo
                                bytes,
                                (unsigned long long)data->args.offset);
 
-               started += bytes;
-               user_addr += bytes;
                pos += bytes;
                pgbase += bytes;
+               curpage += pgbase >> PAGE_SHIFT;
                pgbase &= ~PAGE_MASK;
 
                count -= bytes;
        } while (count != 0);
-
-       if (put_dreq(dreq))
-               nfs_direct_complete(dreq);
-
-       if (started)
-               return 0;
-       return result < 0 ? (ssize_t) result : -EFAULT;
+       BUG_ON(!list_empty(list));
 }
 
-static ssize_t nfs_direct_read(struct kiocb *iocb, unsigned long user_addr, size_t count, loff_t pos)
+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)
 {
-       ssize_t result = 0;
+       ssize_t result;
        sigset_t oldset;
        struct inode *inode = iocb->ki_filp->f_mapping->host;
        struct rpc_clnt *clnt = NFS_CLIENT(inode);
        struct nfs_direct_req *dreq;
 
-       dreq = nfs_direct_req_alloc();
+       dreq = nfs_direct_read_alloc(count, NFS_SERVER(inode)->rsize);
        if (!dreq)
                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->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);
        if (!is_sync_kiocb(iocb))
@@ -378,9 +408,8 @@ static ssize_t nfs_direct_read(struct kiocb *iocb, unsigned long user_addr, size
 
        nfs_add_stats(inode, NFSIOS_DIRECTREADBYTES, count);
        rpc_clnt_sigmask(clnt, &oldset);
-       result = nfs_direct_read_schedule(dreq, user_addr, count, pos);
-       if (!result)
-               result = nfs_direct_wait(dreq);
+       nfs_direct_read_schedule(dreq);
+       result = nfs_direct_wait(dreq);
        rpc_clnt_sigunmask(clnt, &oldset);
 
        return result;
@@ -388,10 +417,10 @@ 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)
 {
-       while (!list_empty(&dreq->rewrite_list)) {
-               struct nfs_write_data *data = list_entry(dreq->rewrite_list.next, struct nfs_write_data, pages);
+       list_splice_init(&dreq->rewrite_list, &dreq->list);
+       while (!list_empty(&dreq->list)) {
+               struct nfs_write_data *data = list_entry(dreq->list.next, struct nfs_write_data, pages);
                list_del(&data->pages);
-               nfs_direct_release_pages(data->pagevec, data->npages);
                nfs_writedata_release(data);
        }
 }
@@ -399,51 +428,14 @@ static void nfs_direct_free_writedata(struct nfs_direct_req *dreq)
 #if defined(CONFIG_NFS_V3) || defined(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 list_head *pos;
 
+       list_splice_init(&dreq->rewrite_list, &dreq->list);
+       list_for_each(pos, &dreq->list)
+               dreq->outstanding++;
        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);
-
-               /*
-                * 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);
+       nfs_direct_write_schedule(dreq, FLUSH_STABLE);
 }
 
 static void nfs_direct_commit_result(struct rpc_task *task, void *calldata)
@@ -480,8 +472,8 @@ static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
        data->cred = dreq->ctx->cred;
 
        data->args.fh = NFS_FH(data->inode);
-       data->args.offset = 0;
-       data->args.count = 0;
+       data->args.offset = dreq->pos;
+       data->args.count = dreq->user_count;
        data->res.count = 0;
        data->res.fattr = &data->fattr;
        data->res.verf = &data->verf;
@@ -543,6 +535,47 @@ static void nfs_direct_write_complete(struct nfs_direct_req *dreq, struct inode
 }
 #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)
 {
        struct nfs_write_data *data = calldata;
@@ -572,6 +605,8 @@ 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);
 }
@@ -585,8 +620,14 @@ 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;
 
-       if (put_dreq(dreq))
-               nfs_direct_write_complete(dreq, data->inode);
+       spin_lock(&dreq->lock);
+       if (--dreq->outstanding) {
+               spin_unlock(&dreq->lock);
+               return;
+       }
+       spin_unlock(&dreq->lock);
+
+       nfs_direct_write_complete(dreq, data->inode);
 }
 
 static const struct rpc_call_ops nfs_write_direct_ops = {
@@ -595,62 +636,41 @@ static const struct rpc_call_ops nfs_write_direct_ops = {
 };
 
 /*
- * For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
- * 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.
+ * For each nfs_write_data struct that was allocated on the list, dispatch
+ * an NFS WRITE operation
  */
-static ssize_t nfs_direct_write_schedule(struct nfs_direct_req *dreq, unsigned long user_addr, size_t count, loff_t pos, int sync)
+static void nfs_direct_write_schedule(struct nfs_direct_req *dreq, int sync)
 {
        struct nfs_open_context *ctx = dreq->ctx;
        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;
-       unsigned int wpages = nfs_max_pages(wsize);
-       unsigned int pgbase;
-       int result;
-       ssize_t started = 0;
+       unsigned int curpage, pgbase;
 
-       get_dreq(dreq);
-
-       pgbase = user_addr & ~PAGE_MASK;
+       curpage = 0;
+       pgbase = dreq->user_addr & ~PAGE_MASK;
        do {
                struct nfs_write_data *data;
                size_t bytes;
 
-               result = -ENOMEM;
-               data = nfs_writedata_alloc(wpages);
-               if (unlikely(!data))
-                       break;
-
                bytes = wsize;
                if (count < wsize)
                        bytes = count;
 
-               data->npages = nfs_direct_count_pages(user_addr, bytes);
-               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);
-
+               BUG_ON(list_empty(list));
+               data = list_entry(list->next, struct nfs_write_data, pages);
                list_move_tail(&data->pages, &dreq->rewrite_list);
 
-               data->req = (struct nfs_page *) dreq;
                data->inode = inode;
                data->cred = ctx->cred;
                data->args.fh = NFS_FH(inode);
                data->args.context = ctx;
                data->args.offset = pos;
                data->args.pgbase = pgbase;
-               data->args.pages = data->pagevec;
+               data->args.pages = &pages[curpage];
                data->args.count = bytes;
                data->res.fattr = &data->fattr;
                data->res.count = bytes;
@@ -674,26 +694,19 @@ static ssize_t nfs_direct_write_schedule(struct nfs_direct_req *dreq, unsigned l
                                bytes,
                                (unsigned long long)data->args.offset);
 
-               started += bytes;
-               user_addr += bytes;
                pos += bytes;
                pgbase += bytes;
+               curpage += pgbase >> PAGE_SHIFT;
                pgbase &= ~PAGE_MASK;
 
                count -= bytes;
        } while (count != 0);
-
-       if (put_dreq(dreq))
-               nfs_direct_write_complete(dreq, inode);
-
-       if (started)
-               return 0;
-       return result < 0 ? (ssize_t) result : -EFAULT;
+       BUG_ON(!list_empty(list));
 }
 
-static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, size_t count, loff_t pos)
+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)
 {
-       ssize_t result = 0;
+       ssize_t result;
        sigset_t oldset;
        struct inode *inode = iocb->ki_filp->f_mapping->host;
        struct rpc_clnt *clnt = NFS_CLIENT(inode);
@@ -701,14 +714,17 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, siz
        size_t wsize = NFS_SERVER(inode)->wsize;
        int sync = 0;
 
-       dreq = nfs_direct_req_alloc();
+       dreq = nfs_direct_write_alloc(count, wsize);
        if (!dreq)
                return -ENOMEM;
-       nfs_alloc_commit_data(dreq);
-
        if (dreq->commit_data == NULL || count < wsize)
                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->ctx = get_nfs_open_context((struct nfs_open_context *)iocb->ki_filp->private_data);
        if (!is_sync_kiocb(iocb))
@@ -719,9 +735,8 @@ static ssize_t nfs_direct_write(struct kiocb *iocb, unsigned long user_addr, siz
        nfs_begin_data_update(inode);
 
        rpc_clnt_sigmask(clnt, &oldset);
-       result = nfs_direct_write_schedule(dreq, user_addr, count, pos, sync);
-       if (!result)
-               result = nfs_direct_wait(dreq);
+       nfs_direct_write_schedule(dreq, sync);
+       result = nfs_direct_wait(dreq);
        rpc_clnt_sigunmask(clnt, &oldset);
 
        return result;
@@ -751,6 +766,8 @@ 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 retval = -EINVAL;
+       int page_count;
+       struct page **pages;
        struct file *file = iocb->ki_filp;
        struct address_space *mapping = file->f_mapping;
 
@@ -772,7 +789,14 @@ ssize_t nfs_file_direct_read(struct kiocb *iocb, char __user *buf, size_t count,
        if (retval)
                goto out;
 
-       retval = nfs_direct_read(iocb, (unsigned long) buf, count, pos);
+       retval = nfs_get_user_pages(READ, (unsigned long) buf,
+                                               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)
                iocb->ki_pos = pos + retval;
 
@@ -808,6 +832,8 @@ out:
 ssize_t nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t count, loff_t pos)
 {
        ssize_t retval;
+       int page_count;
+       struct page **pages;
        struct file *file = iocb->ki_filp;
        struct address_space *mapping = file->f_mapping;
 
@@ -835,7 +861,14 @@ ssize_t nfs_file_direct_write(struct kiocb *iocb, const char __user *buf, size_t
        if (retval)
                goto out;
 
-       retval = nfs_direct_write(iocb, (unsigned long) buf, count, pos);
+       retval = nfs_get_user_pages(WRITE, (unsigned long) buf,
+                                               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
index 2d3fb64..7c7320f 100644 (file)
@@ -729,7 +729,6 @@ struct nfs_read_data {
        struct list_head        pages;  /* Coalesced read requests */
        struct nfs_page         *req;   /* multi ops per nfs_page */
        struct page             **pagevec;
-       unsigned int            npages; /* active pages in pagevec */
        struct nfs_readargs args;
        struct nfs_readres  res;
 #ifdef CONFIG_NFS_V4
@@ -748,7 +747,6 @@ struct nfs_write_data {
        struct list_head        pages;          /* Coalesced requests we wish to flush */
        struct nfs_page         *req;           /* multi ops per nfs_page */
        struct page             **pagevec;
-       unsigned int            npages;         /* active pages in pagevec */
        struct nfs_writeargs    args;           /* argument struct */
        struct nfs_writeres     res;            /* result struct */
 #ifdef CONFIG_NFS_V4